User login
Does the Family APGAR Effectively Measure Family Functioning?
METHODS: Data were collected from 401 pediatricians and family physicians from the Pediatric Research in Office Settings network and the Ambulatory Sentinel Practice Network. The physicians enrolled 22,059 consecutive office visits by children aged 4 to 15 years. Parents completed a survey that included the Family APGAR and the Pediatric Symptom Checklist. Clinicians completed a survey that described child psychosocial problems, treatments initiated or continued, and specialty care referrals.
RESULTS: Family dysfunction on the index visit often differed from dysfunction at follow-up (k=0.24). Only 31% of the families with positive Family APGAR scores at baseline were positive at follow-up, and only 43% of those with positive scores at follow-up had a positive score at the initial visit. There were many disagreements between the Family APGAR and the clinician. The Family APGAR was negative for 73% of clinician-identified dysfunctional families, and clinicians did not identify dysfunction for 83% of Family APGAR–identified dysfunctions (k=0.06).
CONCLUSIONS: Our data do not support the use of the Family APGAR as a measure of family dysfunction in the primary care setting. Future research should clarify what it does measure.
A strong family orientation has been a cornerstone of family practice since its emergence in the late 1960s1-4 and is alo important in pediatrics.5 The development of family medicine as a dominant primary care specialty has occurred in parallel with the development of clinical applications of family systems theory.6-9 More recently the Institute of Medicine report on primary care in America10 has reaffirmed provision of care in the context of family and the community as a central component of primary care.
Integrating an effective family orientation into everyday practice has proved feasible and extant in family practice.11-13 Several approaches to examining and characterizing family function for research purposes have been proposed.14 These include a combination of analysis of communication, observation of interaction, and individual patient report. Many of these approaches are time consuming and not practical for use in large sample studies requiring a brief instrument. The ability to assess the family context, however, is critical to many primary care studies and particularly those that deal with behavior, mental health, and psychosocial problems.
The Family APGAR was introduced by Gabriel Smilkstein in 1978 to assess adult satisfaction with social support from the family.15 It draws its name from a 5-item measure of perceived family support in the domains of adaptation, partnership, growth, affection, and resolve. The statements focus on the emotional, communicative, and social interactive relationships between the respondent and his or her family, for example: “I find that my family accepts my wishes to take on new activities or make changes in my lifestyle.”
Several studies have examined the psychometric properties of the instrument. We focus on evidence about the validity of the instrument, as it has regularly been found to be internally consistent.16,17 Good and colleagues16 found that Family APGAR scores correlated highly (r=0.80) with scores on the Pless-Satterwhite Family Function Index18 in a small nonclinical sample (N=38). In a small sample of mental health outpatients (N=20), the same authors found that Family APGAR scores correlated (r=0.64) with therapists’ ratings of the degree of family distress. Foulke and coworkers19 administered the Family APGAR and the Family Adaptation and Cohesion Evaluation Scales20 (FACES II) to 140 families and found that the Family APGAR correlated with the FACES Cohesion scale (r=0.70) and with the Adaptability scale (r=0.59; Stephen Zyzanski, personal communication, June 2000). However, when Clover and colleagues21 administered the Family APGAR and the FACES II to 66 families they reported that there was no association between the 2 scales.
Smucker and coworkers22 found no association (k=-0.05) between the Family APGAR and physicians’ judgments about the presence of family dysfunction among 152 families. This lack of association, however, could have resulted from the physicians’ difficulties in recognizing family dysfunction, problems in the Family APGAR, or both. North and colleagues23 obtained ratings of the usefulness of family assessment tools from 299 family physicians. The Family APGAR was rated less useful than any other tool.
Smilkstein and coworkers17 found that adults in counseling perceived their families as more dysfunctional than adults in other samples. There was, however, no assessment of the family; thus the finding does not directly support the validity of the Family APGAR as a measure of family dysfunction. Smilkstein and colleagues also found that adopted children were more likely to perceive their families as dysfunctional than were biological children. This would not validate the Family APGAR as a measure of family dysfunction, because it seems unlikely that families who adopt are more dysfunctional than other families.
A few studies have examined whether low Family APGAR ratings (which mean higher perceived family dysfunction) predict other clinical phenomena,12,22,24-27 with mixed results. However, an association between a patient’s report on the Family APGAR and later mental health service use does not directly bear on whether the instrument is a valid measure of family dysfunction. Therefore, the evidence of whether the Family APGAR is a valid measure of family dysfunction is mixed.
We used the Family APGAR as a measure of family dysfunction in a large study of psychosocial problems in children. Our study accomplished 3 goals that had not been achieved in previous research. First, we examined the internal consistency of the Family APGAR in a very large sample of office-based visits (N=21,285). In an internally consistent survey the items essentially measure one thing. Researchers who use the Family APGAR to compare families on the basis of functionality are assuming that there is a single dimension of family characteristics that is tapped by the survey.
Second, we used a large sample of repeat office visits (N=1146) to examine whether positive (dysfunctional) Family APGAR scores are stable over 6 months. When health service workers speak of dysfunctional families they often mean those that are persistently dysfunctional. Similarly, clinicians often adopt a watch and wait strategy for dealing with an initial report of psychosocial problems.28 If a positive score on the Family APGAR signals persistent dysfunction, then a positive score at the index visit should usually be matched by a positive score at follow-up.
Third, in a large sample of office visits (N=4050) we examined whether an adult family member’s report of problems on the family APGAR matched the clinician’s independent judgments of whether there were family problems. Although there are many reasons to expect disagreements between a valid survey measure of family dysfunction and clinicians’ judgments, a very weak level of agreement would raise questions about whether the Family APGAR measured dysfunctionality.
Methods
Study Sites
The Child Behavior Study (CBS) was conducted in several large primary care research networks in North America. The Ambulatory Sentinel Practice Network (ASPN), a family practice research network, included 141 practices in 41 states and 6 Canadian provinces and was composed of approximately 680 clinicians. Eightyfive percent of the ASPN clinicians were family physicians, 7% were nurse practitioners, and 8% were physician assistants. Additional family physician participants came from the Wisconsin Research Network and the Minnesota Academy of Family Physicians Research Network, which had characteristics similar to ASPN. The primary care practice–based Pediatric Research in Office Settings (PROS) network included more than 1500 clinicians from more than 480 pediatric practices in all 50 states and the Commonwealth of Puerto Rico. Eightynine percent of the PROS clinicians were pediatricians, 10% were nurse practitioners, and 1% were physician assistants. Of the 206 practices participating in the CBS, 30% were urban, 38% were suburban, and 32% were rural.
All clinicians participating in the CBS were included for our analysis (401 clinicians in 44 states, the Commonwealth of Puerto Rico, and 4 Canadian provinces). The clinicians included 267 pediatricians, 134 family practitioners, and 29 nurse clinicians. Previous research from both ASPN and PROS confirmed the comparability of patients, clinicians, and practices in primary care network studies with those identified in national samples.29,30 In addition, we compared participating pediatric clinicians with a random sample of pediatricians from the American Academy of Pediatrics31 on demographic factors and practice characteristics. We found few differences between participating clinicians and other clinicians.
Patient Sample
Each participating clinician enrolled a consecutive sample of approximately 55 children aged 4 to 15 years presenting for nonemergent care with a parent or primary caretaker. We enrolled each child only once and excluded children seen for procedures only. Some eligible children were not recruited, primarily because of parental refusal (63% of eligible but not participating children) and occasionally because the opportunity was either overlooked by the office staff (25%) or because the family dropped out of the study (12%). We compared participating children with those who were eligible but not participating on the basis of age and sex, and found no differences. In addition, we examined whether clinician or practice characteristics might affect patient participation, including clinician discipline, geographic region, practice population size, percentage of managed care patients, and clinician attitudes toward mental health treatment. Only those clinicians located in the South and West seemed to include a higher percentage of their eligible participants (94% to 89% for each); none of the other measured sources of selection bias were statistically significant.
This procedure produced a sample of 22,059 children seen in office visits. Among those visits 774 (3.5%) with missing data on 1 or more of the 5 APGAR items were excluded, resulting in a final study sample of 21,285 visits.
Procedures
Procedures and consent forms for the CBS were approved by institutional review boards affiliated with PROS, ASPN, and the University of Pittsburgh. Study procedures have been described in detail elsewhere32 Consenting parents (or the accompanying primary caregiver) filled out a parent questionnaire while waiting to see the clinician. The questionnaire included demographic data, the Family APGAR, and the Pediatric Symptom Checklist (a psychosocial screening instrument). The clinician did not see the completed Family APGAR, Pediatric Symptom Checklist, or other parent questionnaire data.
After seeing a patient the clinician completed a survey about the encounter, documenting whether a new, ongoing, or recurrent psychosocial problem was present, including an explicit statement of family dysfunction. Finally, the survey also included a checklist of a series of psychosocial problems that the clinician might have recognized in the child (clinicians could and often did check more than one problem).
Procedures for Follow-up
A random sample of children with clinician-identified psychosocial problems was identified for follow-up. African American children were oversampled for follow-up to obtain a sufficient sample. A total of 1970 children were included in the follow-up, and 1354 (69%) were successfully followed up. For this analysis, we used the 1146 patients with complete APGAR data for whom the adult respondent was the same at enrollment and follow-up.
Results
Table 1 shows the associations between the Family APGAR scores and several demographic variables. The strongest predictor of a low Family APGAR score was when the child’s parents were either not married or were separated. Table 2 presents the results for individual items of the Family APGAR.
Internal Consistency
We examined the intercorrelation of the Family APGAR items to determine whether the scale measured a single dimension of family functioning. The correlations of items with the total score ranged from r=0.63 to 0.71. Coefficient a, a summary measure of the intercorrelation of items, equaled 0.85, and deletion of any item from the scale reduced the a. This is a respectable level of internal consistency, suggesting that the Family APGAR items can all be viewed as measures of a single underlying dimension.
Stability of Family APGAR Over Time
Table 3 compares response on the Family APGAR on the initial and follow-up visits. There was a slight but statistically significant difference between the frequency of positive scores (Ž5, indicating family dysfunction) with families appearing more dysfunctional at the follow-up (McNemar’s test: (c2[1]=29.02; P <.001).
If the Family APGAR measures a stable characteristic of family functioning, a family’s dysfunctional status at the index visit should usually agree with its status at the 6-month follow-up. However, only 31% of families appearing dysfunctional during the initial visit still seemed so during the follow-up, and only 43% of those appearing dysfunctional during the follow-up appeared so during the initial visit. The k statistic, a chance-corrected measure of the agreement between the time 1 and time 2 scales, was only 0.24.
Clinician Assessment of Psychosocial Problem and the Family APGAR
Table 4 presents the concordance between a positive score on the Family APGAR and clinicians’ identification of family dysfunction. This Table includes the subset of children for which clinicians recognized a psychosocial problem, because this is the group for which a clinician would be likely to use the Family APGAR. There were high rates of disagreement between clinicians and the scale concerning positive cases. The Family APGAR was negative for 73% of clinician-identified dysfunctional families, and clinicians did not identify dysfunction for 83% of APGAR-identified dysfunctions. Although there was a significant positive association between the Family APGAR and clinician identifications (c2[1]) =19.12; P <.002), the k agreement statistic was only 0.06.
Discussion
Our study adds important new information on the performance of the Family APGAR as a measure of family support and dysfunction. Our results confirm some of the previous work that found that the Family APGAR is an internally consistent measure. Nevertheless, it is unclear exactly what it measures. The Family APGAR did not remain stable across assessments that averaged 6 months apart. On the one hand, it is correlated with both parental reports of symptoms and physician treatment decisions. Previously we reported27 an association of Family APGAR with behavioral problems in children as assessed by both physicians’ reports and scores on the Pediatric Symptom Checklist.22 Also, positive Family APGARs were more frequent among single or separated parents. This could reflect a higher level of dysfunction among such families, but it is equally consistent with the premise that the Family APGAR is a measure of family support.
Less than a third of the families who screened positive at time 1 on the Family APGAR also screened positive at follow-up. By the design of our study, only families in which the clinician had recognized a psychosocial problem at time 1 were followed up. It is plausible that the prevalence of family dysfunction among such families is higher than the general population. If so, it implies that in the general population, the rate of positive time-2 Family APGAR scores given positive time-1 scores would be even lower. The pattern of strong internal consistency and weak temporal stability suggests that the Family APGAR tracks a labile characteristic of families. By itself the transience of positive Family APGAR scores does not imply that it is an invalid measure of dysfunction. It is possible that family dysfunction can be serious but transient. Given what the Family APGAR actually measures, however, this interpretation is hard to support. A positive Family APGAR score is a report by a single individual that the family does not adequately communicate with, emotionally support, adapt to, or problem-solve with him or her. Is one such report evidence of serious family dysfunction, or does normal family functioning include occasional transient disturbances of the relations between a family and one of its members? We incline to the latter view.
Also, the Family APGAR was not associated with clinician reports of family dysfunction. Disagreement between clinicians and the Family APGAR does not necessarily imply that the Family APGAR is wrong. It is likely that clinicians have difficulty recognizing family dysfunction, as would be suggested by the literature showing that clinicians often fail to recognize psychiatric disorder.32,33 In the latter case, however, it has been found that primary care clinicians’ judgments about the presence of psychiatric disorders fail to agree with well-validated psychiatric instruments. Given the scarcity of previous evidence about the validity of the Family APGAR, we do not believe that the lack of agreement between the clinicians and the Family APGAR implies that the clinicians were in error. All we can say is that there is little agreement between the instrument and clinician-identified family dysfunction.
Limitations
Our data have important limitations for examining the ability of the Family APGAR to provide a measure of family function. In our study, the Family APGAR was reported by a single adult in the family. We do not have data from other adults in the family or the index child in the study. Also, we do not have a gold standard criterion assessment of family dysfunction. Our study focused on psychosocial problems in the index child, and thus we did not document a complete picture of the psychosocial problems of the family. Finally, entry into the study was through a child’s visit; this was not a systematic sample of families visiting primary care offices. Therefore, our study oversampled families of children who were presenting for a medical or psychosocial problem.
Future Research
The Family APGAR appears to have utility in family practice research, but researchers should carefully consider how they are using it. Further research could provide a more complete explanation of the association between distress as measured by the Family APGAR and psychosocial problems in children. Our speculation is that because the Family APGAR assesses an adult’s perceptions of family support, low scores may measure parental distress, which will sometimes reflect parental depression. The detrimental effects of parental depression on children are well established. This suggests that the Family APGAR may be an important variable to investigate as a determinant of care-seeking behavior, parent and physician treatment decisions, and as a marker for problems in one or more children. It might be more efficient, however, to screen for parental depression.
Conclusions
Although originally introduced as an assessment of adult satisfaction with family support, the Family APGAR has developed a research following as a measure of family functioning. We present data from a large community-based study of behavioral problems in children that raise questions about the Family APGAR as a measure of family dysfunction. Viewed in the light of the scarcity of previous evidence on the validity of the Family APGAR, we do not believe it should be used as a measure of family functioning. However, because the Family APGAR is associated with child psychosocial problems, it remains of interest for clinical research. One of the goals of future research should be to clarify what the Family APGAR does measure.
We note, however, that the fundamental problems we have discussed may not be in the Family APGAR but rather in the lack of clarity about the meaning of family dysfunction.13 What is the justification for using a measure of social support as a measure of family dysfunction? Many other issues arise in discussion of dysfunctional families, such as parental drug use, the lack of a stable family residence, neglectful child rearing, and the occurrence of domestic violence. The Family APGAR was never intended to measure these issues. A prerequisite for future research must be a clarification of what it means to label a family as dysfunctional.
Acknowledgments
This work was supported by a grant from the National Institute of Mental Health (MH 50629; PI: Kelleher), the Bureau of Maternal and Child Health, and the Staunton Farm Foundation. The authors gratefully acknowledge the contributions of the Pediatric Research in Office Settings (PROS) network of the American Academy of Pediatrics, Elk Grove Village, Ill; the Ambulatory Sentinel Practice Network (ASPN), Denver, Colo; the Wisconsin Research Network (WReN), Madison, Wis; and the Minnesota Academy of Family Physicians Research Network (MAFPRN), St. Paul, Minn. We are particularly grateful for the effort of Diane Comer in both the research and the preparation of this manuscript.
PARTICIPATING CBS PRACTICES
PROS Practices: The pediatric practices or individual practitioners who completed this study are listed by American Academy of Pediatrics chapter: Alabama: Drs Heilpern and Reynolds, PC (Birmingham); Alaska: Anchorage Neighborhood Health Center (Anchorage); Arizona: Mesa Pediatrics Professional Association (Mesa), Pediatric Ambulatory Care Clinic (Phoenix), Orange Grove Pediatrics (Tucson); California 1: Anita Tolentino-Macaraeg, MD (Hollister), Palo Alto Medical Foundation (Los Altos); Colorado: Arvada Pediatric Associates (Arvada), Family Health Center (Denver), Gino Figlio, MD (Lamar); Connecticut: Gerald Jensen, MD (Bristol), Barry Keller, MD (Danbury), Community Health Services (Hartford), St. Francis Pediatric Primary Care Center (Hartford); Florida: Atlantic Coast Pediatrics (Merritt Island), Children’s Clinic (Tallahassee); Georgia: The Pediatric Center (Stone Mountain); Hawaii: Melinda Ashton, MD (Honolulu), Straub Clinic—Pediatrics (Aiea); Iowa: Newborn & Pediatric Specialist, PC (Des Moines), David Kelly, MD (Marshalltown); Illinois: SIU Physicians & Surgeons (Auburn), Emalee Flaherty, MD (Chicago), Southwest Pediatrics (Palos Park); Indiana: Bloomington Pediatric Association (Bloomington), Community Health Access Program (Bloomington), Drs. Mary Jo Stine and Richard Weiner (Indianapolis), Jeffersonville Pediatrics (Jeffersonville), Pediatric Advocates (Peru); Kansas: Bethel Pediatrics (Newton); Kentucky: Tri-State Pediatrics, PSC (Ashland); Louisiana: Children’s Clinic of Southwest LA (Lake Charles); Maine: John Salvato, MD (Waterville), Intermed Pediatrics (Yarmouth); Maryland: O’Donovan & Ahluwalia, MD, PA (Baltimore), Children’s Medical Group (Cumberland), Shore Pediatrics (Easton), Clinical Associates Pediatrics (Towson/Woodlawn); Massachusetts: Holyoke Pediatric Associates (Holyoke), Medical Associates (Leominster), The Fallon Clinic (Worcester); Michigan: University Pediatricians, P.C. (Detroit), Pediatric Associates of Farmington (Farmington), Mott Children’s Health Center (Flint), H.. Hildebrandt, MD (Ypsilanti); Montana: Stevensville Pediatrics (Stevensville); Nebraska: Southwest Pediatrics (Omaha); Nevada: Capital Medical Associates (Carson City), Physician’s Center West (Fallon); New Hampshire: Exeter Pediatric Associates (Exeter); New Jersey: Delaware Valley Pediatric Association (Lawrenceville); New Mexico: Albuquerque Pediatric Associates (Albuquerque); New York 1: Pediatric Associates (Camillus), Elmwood Pediatric Group (Rochester), Park Medical Group (Rochester), Edward D. Lewis, MD (Rochester), Panorama Pediatric Group (Rochester), Amherst Pediatric Associates (Williamsville); New York 2: Centro Medico (Jackson Heights); New York 3: Pediatric Office at Roosevelt Island (New York); North Carolina, Triangle Pediatric Center (Cary), Goldsboro Pediatrics (Goldsboro), Medical Association of Surry (Mount Airy), Peace Haven Family Health Center (Winston-Salem); North Dakota: MeritCare MedicalGroup-Pediatrics (Fargo), Altru Clinic (Grand Forks), Dakota Clinic (Jamestown), Medical Arts Clinic (Minot); Ohio: Oxford Pediatrics & Adolescents (Oxford), Pediatrics (Portsmouth), St. Elizabeth Health Center (Youngstown); Oklahoma: Eastern Oklahoma Medical Plaza (Poteau), Shawnee Medical Center Clinic (Shawnee), Pediatric & Adolescent Care (Tulsa); Pennsylvania: Pediatric Practice of Northeastern (Honesdale), Schuylkill Pediatrics (Pottsville), Cevallos and Moise Pediatric Associates, PC (Quakertown), Pennridge Pediatric Associates (Sellersville); Puerto Rico: Ethel Lamela, MD (Isabela), Primary Care Pediatric Clinic Catano (Rio Piedras); Rhode Island: Marvin Wasser, MD (Cranston); South Carolina: Carolina Primary Care (Columbia); Tennessee: Johnson City Pediatrics (Johnson City); Texas: The Pediatric Clinic (Greenville), Department of Pediatrics (Lackland Air Force Base), MD Pediatric Associates (Lewisville), Winnsboro Pediatrics (Winnsboro); Utah: Gordon Glade, MD (American Fork), Mountain View Pediatrics (Sandy), Salt Lake Clinic (Sandy), Granger Medical Center (West Valley City); Vermont: CHP Brattleboro Pediatrics (Brattleboro), University Pediatrics (Burlington), Rebecca Collman, MD (Colchester), Essex Pediatrics (Essex Junction), Mousetrap Pediatrics (Milton), CHP Timber Lane Pediatrics (South Burlington), Joseph Hagan, Jr, MD (South Burlington), Practitioners of Pediatric Medicine (South Burlington), University Pediatrics (Williston); Virginia: Drs. Casey, Goldman, Lischwe, Garrett & Kim (Arlington), James River Pediatrics (Midlothian), Pediatric Faculty Practice Office (Richmond); Washington: Jemima Tso, MD (Auburn), Redmond Pediatrics (Redmond), Rockwood Clinic (Spokane); West Virginia: Tess Alejo (Martinsburg), Medical & Pediatric Associates (Parkersburg), Grant Memorial Pediatrics (Petersburg); Wisconsin: Beloit Clinic SC (Beloit), Middleton Pediatric Clinic (Middleton), Waukesha Pediatric Associates (Waukesha), Gundersen Clinic-Whitehall (Whitehall); Wyoming: Cheyenne Children’s Clinic (Cheyenne), Jackson Pediatrics (Jackson).
ASPN Practices: Arkansas: Batesville Family Practice Center (Batesville); California: Foothills Family Medical Group (Auburn), Loma Linda Family Medical Group (Loma Linda); Colorado: Renate Justin, MD (Fort Collins), Harrington, Knaus, & Spence, PC (Carbondale), La Mariposa Clinic (Denver), Colorado Springs Health Partners (Monument), Penrose Family Health Center (Penrose); Florida: The Family Doctors of Belleview (Belleview); Georgia: Titus Taube, MD (Warner Robbins); Louisiana: Family Medicine Center of Baton Rouge (Baton Rouge); Minnesota: Eagle Medical (Excelsior), Ramsey Clinic—Maplewood (Maplewood); New Hampshire: Mascoma Valley Community Care (Enfield) Hillsboro Medical Services (Hillsboro); New Jersey: A John Orzano, MD (Flemington), Community Care Center (Lebanon); New Mexico: Santa Fe Family Practice (Santa Fe); New York: Raj B. Kachoria, MD (Macedon), Canal Park Family Practice (Palmyra), Montefiore Comprehensive Family Care (Bronx), Mary Kay Ness, MD (Honeoye Falls); North Carolina: Bakersville Community Medical Clinic (Bakersville), Nalli Clinic (Matthews); North Dakota: University of North Dakota Family Practice Center—Minot (Minot) Minot Center for Family Medicine (Minot); Ohio: Center for Family Medicine (Cleveland); Oregon: Dunes Family Health Care, Inc. (Reedsport); Pennsylvania: John Farmer, DO (Waynesboro), Good Samaritan Family Practice (Lebanon); Tennessee: Michael H. Hartsell, MD (Greeneville), Mountain City Extended Hours Clinic (Mountain City); Texas: Van Horn Rural Health Clinic (Van Horn); Virginia: June Tunstall, MD (Surry); Tappahannock Family Practice (Tappahannock); West Virginia: North Fayette Family Health Center (Hico); Wisconsin: Kronenwetter Clinic (Mosinee); Alberta, Foothills Family Medicine Centre (Black Diamond); New Brunswick, David Ross, MD (Moncton); Newfoundland: Newhook Community Health Center (Whitbourne), Ross Thomas, MD (Sackville); Ontario: Steve Nantes, MD (Kitchener), Metcalfe & Dowdell (Kitchener), Bryan Alton, MD (Hamilton).
MAFPRN Practices: Family Medical Practice, PA (Willman), Family Medicine of Winona (Winona), River Valley Clinic (Hastings), Family Medicine Clinic of Lake Crystal (Lake Crystal), Gateway Family Health Clinic (Moose Lake), Eagan Medical Associates (Eagan), Fairview Uptown Clinic (Minneapolis), Bay Area Health Center (Silver Bay), West Side Health Center (St. Paul), Hopkins Family Physicians (Hopkins), Family Practice Center (St. Cloud), Mt. Royal Medical Center (Duluth), North Memorial Family Practice (Minneapolis).
WREN Practices: Poynette Family Practice Center (Poynette), Medical Associates (Baraboo), Plymouth Family Physicians (Plymouth), Monroe Clinic (Monroe), UCC/Mona Grove (Madison), Family Doctors-Black Creek (Black Creek), Southwestern Family Practice (South Milwaukee), Family Health Plan (Elm Grove), LaSalle Clinic (Appleton), Marshfield Clinic—Merril Center (Merrill), Tigerton Clinic (Tigerton), Dean Medical, (Oregon), Physicians Plus/Fitchburg (Fitchburg), Family Health Plan (Glendale), Franciscan Skemp Clinic (Tomah), Galesville Medical Center (Galesville), Medical Associates (Beaver Dam), LaSalle Clinic (Waupaca).
1. Carmichael LP. The family in medicine. J Fam Pract 1976;3:562-63.
2. Geyman JP. The family as object of care in family practice. J Fam Pract 1977;5:571-75.
3. Curry HB. The family as our patient. J Fam Pract 1977;4:757-58.
4. Medalie JH. A family-oriented approach in primary care. Boston, Mass: Little, Brown, and Company; 1976.
5. Kemper KJ, Kelleher KJ. Rationale for family psychosocial screening. Ambulatory Child Health 1996;1:311-24.
6. Ransom DC. The evolution from an individual to a family approach. New York, NY: Brunner-Mazel; 1985.
7. McDaniel S, Campbell TL, Seaburn DB. Family-oriented primary care. New York, NY: Springer-Verlag; 1990.
8. Doherty WB. Family therapy and family medicine. New York, NY: Guilford Press; 1983.
9. Crouch M, Roberts L, eds. The family in medical practice: a family systems primer. New York, NY: Springer-Verlag; 1987.
10. Institute of Medicine. Primary care: America’s health in a new era. Washington, DC: National Academy Press; 1996.
11. Medalie JH, Zyzanski S, Langa D, Stange KC. The family in family practice: is it a reality? J Fam Pract 1998;46:390-96.
12. Mengel M. The use of the family apgar in screening for family dysfunction in a family practice center. J Fam Pract 1987;24:394-98.
13. Mengel MB. The Family APGAR in a research setting. Fam Med 1988;20:143-44.
14. Jacob T, Tennenbaum DL. Family assessment methods. In: Assessment and diagnosis in Child Psychopathol New York, NY: Guilford Press; 1988;196-231.
15. Smilkstein G. The Family APGAR: a proposal for a family function test and its use by physicians. J Fam Pract 1978;6:1231-39.
16. Good MJ, Smilkstein G, Good BJ, Shaffer T, Arrons T. The family APGAR index: a study of construct validity. J Fam Pract 1979;8:577-82.
17. Smilkstein G, Ashworth C, Montano D. Validity and reliability of the Family APGAR as a test of family function. J Fam Pract 1982;15:303-11.
18. Pless JB, Satterwhite B. A measure of family functioning and its application. Soc Sci Med 1973;7:613-21.
19. Foulke FG, Reeb KG, Graham AV, Zyzanski SJ. Family function. respiratory illness, and otitis media in urban black infants. Fam Med 1988;20:128-32.
20. Olson DH, Bell R, Porter J. Family Adaptation and Cohesion Evaluation Scale II. Minneapolis, Minn: Family Inventories Project; 1982.
21. Clover RD, Abell T, Becker LA, Crawford S, Ramsey CN. Family functioning and stress as predictors of influenza B infection. J Fam Pract 1989;28:535-39.
22. Smucker WD, Wildman BG, Lynch TR, et al. Relationship between the Family APGAR and behavioral problems in children. Arch Fam Med 1995;4:535-39.
23. North S, Marvel MK, Hendricks B, Morphew P, North D. Physicians’ usefulness ratings of family-oriented clinical tools. J Fam Pract 1993;37:30-34.
24. Hilliard R, Gjerde C, Parker L. Validity of two psychological screening measures in family practice: personal inventroy and Family APGAR. J Fam Pract 1986;23:345-49.
25. Gwyther RE, Bentz EJ, Drossman DA, Berolzheimer N. Validity of the Family APGAR in patients with irritable bowel syndrome. Fam Med 1993;25:21-25.
26. Ramsey CN, Abell TD, Baker LC. The relationship between family functioning, life events, family structure, and the outcome of pregnancy. J Fam Pract 1986;22:521-27.
27. Murphy M, Kelleher K, Pagano M, et al. The Family APGAR and psychosocial problems in children: a report from ASPN and PROS. J Fam Pract 1998;46:54-64.
28. Gardner W, Kelleher K, Wasserman R, et al. Primary care treatment of pediatric psychosocial problems: a study from PROS and ASPN. Pediatrics 2000;106:E44.-
29. Green L, Miller R, Reed F, Iverson D, Barley G. How representative of typical practice are practice based research networks? A report from the Ambulatory Sentinel Practice Network, Inc (ASPN). Arch Fam Med 1993;2:939-49.
30. Nutting PA, Baier M, Werner JJ, Cutter G, Reed FM, Orzano AJ. Practice patterns of family physicians in practice-based research networks: A report from ASPN. J Am Board Fam Pract 1999;12:278-84.
31. American Academy of Pediatrics. Periodic survey of fellows. Elk Grove, Ill: AAP; 1995. Report No: 32.
32. Kelleher KJ, Childs GE, Wasserman RC, McInerney TK, Nutting PA, Gardner WP. Insurance status and recognition of psychosocial problems: a report from PROS and ASPN. Arch Pediatr Adolesc Med 1997;151:1109-15.
33. Costello EJ. Primary care pediatrics and child psychopathology: a review of diagnosis, treatment, and referral practices. Pediatrics 1986;78:1044-51.
34. Downey G, Coyne J. Children of depressed parents: an integrative review. Psychological Bull 1990;108:50-76.
METHODS: Data were collected from 401 pediatricians and family physicians from the Pediatric Research in Office Settings network and the Ambulatory Sentinel Practice Network. The physicians enrolled 22,059 consecutive office visits by children aged 4 to 15 years. Parents completed a survey that included the Family APGAR and the Pediatric Symptom Checklist. Clinicians completed a survey that described child psychosocial problems, treatments initiated or continued, and specialty care referrals.
RESULTS: Family dysfunction on the index visit often differed from dysfunction at follow-up (k=0.24). Only 31% of the families with positive Family APGAR scores at baseline were positive at follow-up, and only 43% of those with positive scores at follow-up had a positive score at the initial visit. There were many disagreements between the Family APGAR and the clinician. The Family APGAR was negative for 73% of clinician-identified dysfunctional families, and clinicians did not identify dysfunction for 83% of Family APGAR–identified dysfunctions (k=0.06).
CONCLUSIONS: Our data do not support the use of the Family APGAR as a measure of family dysfunction in the primary care setting. Future research should clarify what it does measure.
A strong family orientation has been a cornerstone of family practice since its emergence in the late 1960s1-4 and is alo important in pediatrics.5 The development of family medicine as a dominant primary care specialty has occurred in parallel with the development of clinical applications of family systems theory.6-9 More recently the Institute of Medicine report on primary care in America10 has reaffirmed provision of care in the context of family and the community as a central component of primary care.
Integrating an effective family orientation into everyday practice has proved feasible and extant in family practice.11-13 Several approaches to examining and characterizing family function for research purposes have been proposed.14 These include a combination of analysis of communication, observation of interaction, and individual patient report. Many of these approaches are time consuming and not practical for use in large sample studies requiring a brief instrument. The ability to assess the family context, however, is critical to many primary care studies and particularly those that deal with behavior, mental health, and psychosocial problems.
The Family APGAR was introduced by Gabriel Smilkstein in 1978 to assess adult satisfaction with social support from the family.15 It draws its name from a 5-item measure of perceived family support in the domains of adaptation, partnership, growth, affection, and resolve. The statements focus on the emotional, communicative, and social interactive relationships between the respondent and his or her family, for example: “I find that my family accepts my wishes to take on new activities or make changes in my lifestyle.”
Several studies have examined the psychometric properties of the instrument. We focus on evidence about the validity of the instrument, as it has regularly been found to be internally consistent.16,17 Good and colleagues16 found that Family APGAR scores correlated highly (r=0.80) with scores on the Pless-Satterwhite Family Function Index18 in a small nonclinical sample (N=38). In a small sample of mental health outpatients (N=20), the same authors found that Family APGAR scores correlated (r=0.64) with therapists’ ratings of the degree of family distress. Foulke and coworkers19 administered the Family APGAR and the Family Adaptation and Cohesion Evaluation Scales20 (FACES II) to 140 families and found that the Family APGAR correlated with the FACES Cohesion scale (r=0.70) and with the Adaptability scale (r=0.59; Stephen Zyzanski, personal communication, June 2000). However, when Clover and colleagues21 administered the Family APGAR and the FACES II to 66 families they reported that there was no association between the 2 scales.
Smucker and coworkers22 found no association (k=-0.05) between the Family APGAR and physicians’ judgments about the presence of family dysfunction among 152 families. This lack of association, however, could have resulted from the physicians’ difficulties in recognizing family dysfunction, problems in the Family APGAR, or both. North and colleagues23 obtained ratings of the usefulness of family assessment tools from 299 family physicians. The Family APGAR was rated less useful than any other tool.
Smilkstein and coworkers17 found that adults in counseling perceived their families as more dysfunctional than adults in other samples. There was, however, no assessment of the family; thus the finding does not directly support the validity of the Family APGAR as a measure of family dysfunction. Smilkstein and colleagues also found that adopted children were more likely to perceive their families as dysfunctional than were biological children. This would not validate the Family APGAR as a measure of family dysfunction, because it seems unlikely that families who adopt are more dysfunctional than other families.
A few studies have examined whether low Family APGAR ratings (which mean higher perceived family dysfunction) predict other clinical phenomena,12,22,24-27 with mixed results. However, an association between a patient’s report on the Family APGAR and later mental health service use does not directly bear on whether the instrument is a valid measure of family dysfunction. Therefore, the evidence of whether the Family APGAR is a valid measure of family dysfunction is mixed.
We used the Family APGAR as a measure of family dysfunction in a large study of psychosocial problems in children. Our study accomplished 3 goals that had not been achieved in previous research. First, we examined the internal consistency of the Family APGAR in a very large sample of office-based visits (N=21,285). In an internally consistent survey the items essentially measure one thing. Researchers who use the Family APGAR to compare families on the basis of functionality are assuming that there is a single dimension of family characteristics that is tapped by the survey.
Second, we used a large sample of repeat office visits (N=1146) to examine whether positive (dysfunctional) Family APGAR scores are stable over 6 months. When health service workers speak of dysfunctional families they often mean those that are persistently dysfunctional. Similarly, clinicians often adopt a watch and wait strategy for dealing with an initial report of psychosocial problems.28 If a positive score on the Family APGAR signals persistent dysfunction, then a positive score at the index visit should usually be matched by a positive score at follow-up.
Third, in a large sample of office visits (N=4050) we examined whether an adult family member’s report of problems on the family APGAR matched the clinician’s independent judgments of whether there were family problems. Although there are many reasons to expect disagreements between a valid survey measure of family dysfunction and clinicians’ judgments, a very weak level of agreement would raise questions about whether the Family APGAR measured dysfunctionality.
Methods
Study Sites
The Child Behavior Study (CBS) was conducted in several large primary care research networks in North America. The Ambulatory Sentinel Practice Network (ASPN), a family practice research network, included 141 practices in 41 states and 6 Canadian provinces and was composed of approximately 680 clinicians. Eightyfive percent of the ASPN clinicians were family physicians, 7% were nurse practitioners, and 8% were physician assistants. Additional family physician participants came from the Wisconsin Research Network and the Minnesota Academy of Family Physicians Research Network, which had characteristics similar to ASPN. The primary care practice–based Pediatric Research in Office Settings (PROS) network included more than 1500 clinicians from more than 480 pediatric practices in all 50 states and the Commonwealth of Puerto Rico. Eightynine percent of the PROS clinicians were pediatricians, 10% were nurse practitioners, and 1% were physician assistants. Of the 206 practices participating in the CBS, 30% were urban, 38% were suburban, and 32% were rural.
All clinicians participating in the CBS were included for our analysis (401 clinicians in 44 states, the Commonwealth of Puerto Rico, and 4 Canadian provinces). The clinicians included 267 pediatricians, 134 family practitioners, and 29 nurse clinicians. Previous research from both ASPN and PROS confirmed the comparability of patients, clinicians, and practices in primary care network studies with those identified in national samples.29,30 In addition, we compared participating pediatric clinicians with a random sample of pediatricians from the American Academy of Pediatrics31 on demographic factors and practice characteristics. We found few differences between participating clinicians and other clinicians.
Patient Sample
Each participating clinician enrolled a consecutive sample of approximately 55 children aged 4 to 15 years presenting for nonemergent care with a parent or primary caretaker. We enrolled each child only once and excluded children seen for procedures only. Some eligible children were not recruited, primarily because of parental refusal (63% of eligible but not participating children) and occasionally because the opportunity was either overlooked by the office staff (25%) or because the family dropped out of the study (12%). We compared participating children with those who were eligible but not participating on the basis of age and sex, and found no differences. In addition, we examined whether clinician or practice characteristics might affect patient participation, including clinician discipline, geographic region, practice population size, percentage of managed care patients, and clinician attitudes toward mental health treatment. Only those clinicians located in the South and West seemed to include a higher percentage of their eligible participants (94% to 89% for each); none of the other measured sources of selection bias were statistically significant.
This procedure produced a sample of 22,059 children seen in office visits. Among those visits 774 (3.5%) with missing data on 1 or more of the 5 APGAR items were excluded, resulting in a final study sample of 21,285 visits.
Procedures
Procedures and consent forms for the CBS were approved by institutional review boards affiliated with PROS, ASPN, and the University of Pittsburgh. Study procedures have been described in detail elsewhere32 Consenting parents (or the accompanying primary caregiver) filled out a parent questionnaire while waiting to see the clinician. The questionnaire included demographic data, the Family APGAR, and the Pediatric Symptom Checklist (a psychosocial screening instrument). The clinician did not see the completed Family APGAR, Pediatric Symptom Checklist, or other parent questionnaire data.
After seeing a patient the clinician completed a survey about the encounter, documenting whether a new, ongoing, or recurrent psychosocial problem was present, including an explicit statement of family dysfunction. Finally, the survey also included a checklist of a series of psychosocial problems that the clinician might have recognized in the child (clinicians could and often did check more than one problem).
Procedures for Follow-up
A random sample of children with clinician-identified psychosocial problems was identified for follow-up. African American children were oversampled for follow-up to obtain a sufficient sample. A total of 1970 children were included in the follow-up, and 1354 (69%) were successfully followed up. For this analysis, we used the 1146 patients with complete APGAR data for whom the adult respondent was the same at enrollment and follow-up.
Results
Table 1 shows the associations between the Family APGAR scores and several demographic variables. The strongest predictor of a low Family APGAR score was when the child’s parents were either not married or were separated. Table 2 presents the results for individual items of the Family APGAR.
Internal Consistency
We examined the intercorrelation of the Family APGAR items to determine whether the scale measured a single dimension of family functioning. The correlations of items with the total score ranged from r=0.63 to 0.71. Coefficient a, a summary measure of the intercorrelation of items, equaled 0.85, and deletion of any item from the scale reduced the a. This is a respectable level of internal consistency, suggesting that the Family APGAR items can all be viewed as measures of a single underlying dimension.
Stability of Family APGAR Over Time
Table 3 compares response on the Family APGAR on the initial and follow-up visits. There was a slight but statistically significant difference between the frequency of positive scores (Ž5, indicating family dysfunction) with families appearing more dysfunctional at the follow-up (McNemar’s test: (c2[1]=29.02; P <.001).
If the Family APGAR measures a stable characteristic of family functioning, a family’s dysfunctional status at the index visit should usually agree with its status at the 6-month follow-up. However, only 31% of families appearing dysfunctional during the initial visit still seemed so during the follow-up, and only 43% of those appearing dysfunctional during the follow-up appeared so during the initial visit. The k statistic, a chance-corrected measure of the agreement between the time 1 and time 2 scales, was only 0.24.
Clinician Assessment of Psychosocial Problem and the Family APGAR
Table 4 presents the concordance between a positive score on the Family APGAR and clinicians’ identification of family dysfunction. This Table includes the subset of children for which clinicians recognized a psychosocial problem, because this is the group for which a clinician would be likely to use the Family APGAR. There were high rates of disagreement between clinicians and the scale concerning positive cases. The Family APGAR was negative for 73% of clinician-identified dysfunctional families, and clinicians did not identify dysfunction for 83% of APGAR-identified dysfunctions. Although there was a significant positive association between the Family APGAR and clinician identifications (c2[1]) =19.12; P <.002), the k agreement statistic was only 0.06.
Discussion
Our study adds important new information on the performance of the Family APGAR as a measure of family support and dysfunction. Our results confirm some of the previous work that found that the Family APGAR is an internally consistent measure. Nevertheless, it is unclear exactly what it measures. The Family APGAR did not remain stable across assessments that averaged 6 months apart. On the one hand, it is correlated with both parental reports of symptoms and physician treatment decisions. Previously we reported27 an association of Family APGAR with behavioral problems in children as assessed by both physicians’ reports and scores on the Pediatric Symptom Checklist.22 Also, positive Family APGARs were more frequent among single or separated parents. This could reflect a higher level of dysfunction among such families, but it is equally consistent with the premise that the Family APGAR is a measure of family support.
Less than a third of the families who screened positive at time 1 on the Family APGAR also screened positive at follow-up. By the design of our study, only families in which the clinician had recognized a psychosocial problem at time 1 were followed up. It is plausible that the prevalence of family dysfunction among such families is higher than the general population. If so, it implies that in the general population, the rate of positive time-2 Family APGAR scores given positive time-1 scores would be even lower. The pattern of strong internal consistency and weak temporal stability suggests that the Family APGAR tracks a labile characteristic of families. By itself the transience of positive Family APGAR scores does not imply that it is an invalid measure of dysfunction. It is possible that family dysfunction can be serious but transient. Given what the Family APGAR actually measures, however, this interpretation is hard to support. A positive Family APGAR score is a report by a single individual that the family does not adequately communicate with, emotionally support, adapt to, or problem-solve with him or her. Is one such report evidence of serious family dysfunction, or does normal family functioning include occasional transient disturbances of the relations between a family and one of its members? We incline to the latter view.
Also, the Family APGAR was not associated with clinician reports of family dysfunction. Disagreement between clinicians and the Family APGAR does not necessarily imply that the Family APGAR is wrong. It is likely that clinicians have difficulty recognizing family dysfunction, as would be suggested by the literature showing that clinicians often fail to recognize psychiatric disorder.32,33 In the latter case, however, it has been found that primary care clinicians’ judgments about the presence of psychiatric disorders fail to agree with well-validated psychiatric instruments. Given the scarcity of previous evidence about the validity of the Family APGAR, we do not believe that the lack of agreement between the clinicians and the Family APGAR implies that the clinicians were in error. All we can say is that there is little agreement between the instrument and clinician-identified family dysfunction.
Limitations
Our data have important limitations for examining the ability of the Family APGAR to provide a measure of family function. In our study, the Family APGAR was reported by a single adult in the family. We do not have data from other adults in the family or the index child in the study. Also, we do not have a gold standard criterion assessment of family dysfunction. Our study focused on psychosocial problems in the index child, and thus we did not document a complete picture of the psychosocial problems of the family. Finally, entry into the study was through a child’s visit; this was not a systematic sample of families visiting primary care offices. Therefore, our study oversampled families of children who were presenting for a medical or psychosocial problem.
Future Research
The Family APGAR appears to have utility in family practice research, but researchers should carefully consider how they are using it. Further research could provide a more complete explanation of the association between distress as measured by the Family APGAR and psychosocial problems in children. Our speculation is that because the Family APGAR assesses an adult’s perceptions of family support, low scores may measure parental distress, which will sometimes reflect parental depression. The detrimental effects of parental depression on children are well established. This suggests that the Family APGAR may be an important variable to investigate as a determinant of care-seeking behavior, parent and physician treatment decisions, and as a marker for problems in one or more children. It might be more efficient, however, to screen for parental depression.
Conclusions
Although originally introduced as an assessment of adult satisfaction with family support, the Family APGAR has developed a research following as a measure of family functioning. We present data from a large community-based study of behavioral problems in children that raise questions about the Family APGAR as a measure of family dysfunction. Viewed in the light of the scarcity of previous evidence on the validity of the Family APGAR, we do not believe it should be used as a measure of family functioning. However, because the Family APGAR is associated with child psychosocial problems, it remains of interest for clinical research. One of the goals of future research should be to clarify what the Family APGAR does measure.
We note, however, that the fundamental problems we have discussed may not be in the Family APGAR but rather in the lack of clarity about the meaning of family dysfunction.13 What is the justification for using a measure of social support as a measure of family dysfunction? Many other issues arise in discussion of dysfunctional families, such as parental drug use, the lack of a stable family residence, neglectful child rearing, and the occurrence of domestic violence. The Family APGAR was never intended to measure these issues. A prerequisite for future research must be a clarification of what it means to label a family as dysfunctional.
Acknowledgments
This work was supported by a grant from the National Institute of Mental Health (MH 50629; PI: Kelleher), the Bureau of Maternal and Child Health, and the Staunton Farm Foundation. The authors gratefully acknowledge the contributions of the Pediatric Research in Office Settings (PROS) network of the American Academy of Pediatrics, Elk Grove Village, Ill; the Ambulatory Sentinel Practice Network (ASPN), Denver, Colo; the Wisconsin Research Network (WReN), Madison, Wis; and the Minnesota Academy of Family Physicians Research Network (MAFPRN), St. Paul, Minn. We are particularly grateful for the effort of Diane Comer in both the research and the preparation of this manuscript.
PARTICIPATING CBS PRACTICES
PROS Practices: The pediatric practices or individual practitioners who completed this study are listed by American Academy of Pediatrics chapter: Alabama: Drs Heilpern and Reynolds, PC (Birmingham); Alaska: Anchorage Neighborhood Health Center (Anchorage); Arizona: Mesa Pediatrics Professional Association (Mesa), Pediatric Ambulatory Care Clinic (Phoenix), Orange Grove Pediatrics (Tucson); California 1: Anita Tolentino-Macaraeg, MD (Hollister), Palo Alto Medical Foundation (Los Altos); Colorado: Arvada Pediatric Associates (Arvada), Family Health Center (Denver), Gino Figlio, MD (Lamar); Connecticut: Gerald Jensen, MD (Bristol), Barry Keller, MD (Danbury), Community Health Services (Hartford), St. Francis Pediatric Primary Care Center (Hartford); Florida: Atlantic Coast Pediatrics (Merritt Island), Children’s Clinic (Tallahassee); Georgia: The Pediatric Center (Stone Mountain); Hawaii: Melinda Ashton, MD (Honolulu), Straub Clinic—Pediatrics (Aiea); Iowa: Newborn & Pediatric Specialist, PC (Des Moines), David Kelly, MD (Marshalltown); Illinois: SIU Physicians & Surgeons (Auburn), Emalee Flaherty, MD (Chicago), Southwest Pediatrics (Palos Park); Indiana: Bloomington Pediatric Association (Bloomington), Community Health Access Program (Bloomington), Drs. Mary Jo Stine and Richard Weiner (Indianapolis), Jeffersonville Pediatrics (Jeffersonville), Pediatric Advocates (Peru); Kansas: Bethel Pediatrics (Newton); Kentucky: Tri-State Pediatrics, PSC (Ashland); Louisiana: Children’s Clinic of Southwest LA (Lake Charles); Maine: John Salvato, MD (Waterville), Intermed Pediatrics (Yarmouth); Maryland: O’Donovan & Ahluwalia, MD, PA (Baltimore), Children’s Medical Group (Cumberland), Shore Pediatrics (Easton), Clinical Associates Pediatrics (Towson/Woodlawn); Massachusetts: Holyoke Pediatric Associates (Holyoke), Medical Associates (Leominster), The Fallon Clinic (Worcester); Michigan: University Pediatricians, P.C. (Detroit), Pediatric Associates of Farmington (Farmington), Mott Children’s Health Center (Flint), H.. Hildebrandt, MD (Ypsilanti); Montana: Stevensville Pediatrics (Stevensville); Nebraska: Southwest Pediatrics (Omaha); Nevada: Capital Medical Associates (Carson City), Physician’s Center West (Fallon); New Hampshire: Exeter Pediatric Associates (Exeter); New Jersey: Delaware Valley Pediatric Association (Lawrenceville); New Mexico: Albuquerque Pediatric Associates (Albuquerque); New York 1: Pediatric Associates (Camillus), Elmwood Pediatric Group (Rochester), Park Medical Group (Rochester), Edward D. Lewis, MD (Rochester), Panorama Pediatric Group (Rochester), Amherst Pediatric Associates (Williamsville); New York 2: Centro Medico (Jackson Heights); New York 3: Pediatric Office at Roosevelt Island (New York); North Carolina, Triangle Pediatric Center (Cary), Goldsboro Pediatrics (Goldsboro), Medical Association of Surry (Mount Airy), Peace Haven Family Health Center (Winston-Salem); North Dakota: MeritCare MedicalGroup-Pediatrics (Fargo), Altru Clinic (Grand Forks), Dakota Clinic (Jamestown), Medical Arts Clinic (Minot); Ohio: Oxford Pediatrics & Adolescents (Oxford), Pediatrics (Portsmouth), St. Elizabeth Health Center (Youngstown); Oklahoma: Eastern Oklahoma Medical Plaza (Poteau), Shawnee Medical Center Clinic (Shawnee), Pediatric & Adolescent Care (Tulsa); Pennsylvania: Pediatric Practice of Northeastern (Honesdale), Schuylkill Pediatrics (Pottsville), Cevallos and Moise Pediatric Associates, PC (Quakertown), Pennridge Pediatric Associates (Sellersville); Puerto Rico: Ethel Lamela, MD (Isabela), Primary Care Pediatric Clinic Catano (Rio Piedras); Rhode Island: Marvin Wasser, MD (Cranston); South Carolina: Carolina Primary Care (Columbia); Tennessee: Johnson City Pediatrics (Johnson City); Texas: The Pediatric Clinic (Greenville), Department of Pediatrics (Lackland Air Force Base), MD Pediatric Associates (Lewisville), Winnsboro Pediatrics (Winnsboro); Utah: Gordon Glade, MD (American Fork), Mountain View Pediatrics (Sandy), Salt Lake Clinic (Sandy), Granger Medical Center (West Valley City); Vermont: CHP Brattleboro Pediatrics (Brattleboro), University Pediatrics (Burlington), Rebecca Collman, MD (Colchester), Essex Pediatrics (Essex Junction), Mousetrap Pediatrics (Milton), CHP Timber Lane Pediatrics (South Burlington), Joseph Hagan, Jr, MD (South Burlington), Practitioners of Pediatric Medicine (South Burlington), University Pediatrics (Williston); Virginia: Drs. Casey, Goldman, Lischwe, Garrett & Kim (Arlington), James River Pediatrics (Midlothian), Pediatric Faculty Practice Office (Richmond); Washington: Jemima Tso, MD (Auburn), Redmond Pediatrics (Redmond), Rockwood Clinic (Spokane); West Virginia: Tess Alejo (Martinsburg), Medical & Pediatric Associates (Parkersburg), Grant Memorial Pediatrics (Petersburg); Wisconsin: Beloit Clinic SC (Beloit), Middleton Pediatric Clinic (Middleton), Waukesha Pediatric Associates (Waukesha), Gundersen Clinic-Whitehall (Whitehall); Wyoming: Cheyenne Children’s Clinic (Cheyenne), Jackson Pediatrics (Jackson).
ASPN Practices: Arkansas: Batesville Family Practice Center (Batesville); California: Foothills Family Medical Group (Auburn), Loma Linda Family Medical Group (Loma Linda); Colorado: Renate Justin, MD (Fort Collins), Harrington, Knaus, & Spence, PC (Carbondale), La Mariposa Clinic (Denver), Colorado Springs Health Partners (Monument), Penrose Family Health Center (Penrose); Florida: The Family Doctors of Belleview (Belleview); Georgia: Titus Taube, MD (Warner Robbins); Louisiana: Family Medicine Center of Baton Rouge (Baton Rouge); Minnesota: Eagle Medical (Excelsior), Ramsey Clinic—Maplewood (Maplewood); New Hampshire: Mascoma Valley Community Care (Enfield) Hillsboro Medical Services (Hillsboro); New Jersey: A John Orzano, MD (Flemington), Community Care Center (Lebanon); New Mexico: Santa Fe Family Practice (Santa Fe); New York: Raj B. Kachoria, MD (Macedon), Canal Park Family Practice (Palmyra), Montefiore Comprehensive Family Care (Bronx), Mary Kay Ness, MD (Honeoye Falls); North Carolina: Bakersville Community Medical Clinic (Bakersville), Nalli Clinic (Matthews); North Dakota: University of North Dakota Family Practice Center—Minot (Minot) Minot Center for Family Medicine (Minot); Ohio: Center for Family Medicine (Cleveland); Oregon: Dunes Family Health Care, Inc. (Reedsport); Pennsylvania: John Farmer, DO (Waynesboro), Good Samaritan Family Practice (Lebanon); Tennessee: Michael H. Hartsell, MD (Greeneville), Mountain City Extended Hours Clinic (Mountain City); Texas: Van Horn Rural Health Clinic (Van Horn); Virginia: June Tunstall, MD (Surry); Tappahannock Family Practice (Tappahannock); West Virginia: North Fayette Family Health Center (Hico); Wisconsin: Kronenwetter Clinic (Mosinee); Alberta, Foothills Family Medicine Centre (Black Diamond); New Brunswick, David Ross, MD (Moncton); Newfoundland: Newhook Community Health Center (Whitbourne), Ross Thomas, MD (Sackville); Ontario: Steve Nantes, MD (Kitchener), Metcalfe & Dowdell (Kitchener), Bryan Alton, MD (Hamilton).
MAFPRN Practices: Family Medical Practice, PA (Willman), Family Medicine of Winona (Winona), River Valley Clinic (Hastings), Family Medicine Clinic of Lake Crystal (Lake Crystal), Gateway Family Health Clinic (Moose Lake), Eagan Medical Associates (Eagan), Fairview Uptown Clinic (Minneapolis), Bay Area Health Center (Silver Bay), West Side Health Center (St. Paul), Hopkins Family Physicians (Hopkins), Family Practice Center (St. Cloud), Mt. Royal Medical Center (Duluth), North Memorial Family Practice (Minneapolis).
WREN Practices: Poynette Family Practice Center (Poynette), Medical Associates (Baraboo), Plymouth Family Physicians (Plymouth), Monroe Clinic (Monroe), UCC/Mona Grove (Madison), Family Doctors-Black Creek (Black Creek), Southwestern Family Practice (South Milwaukee), Family Health Plan (Elm Grove), LaSalle Clinic (Appleton), Marshfield Clinic—Merril Center (Merrill), Tigerton Clinic (Tigerton), Dean Medical, (Oregon), Physicians Plus/Fitchburg (Fitchburg), Family Health Plan (Glendale), Franciscan Skemp Clinic (Tomah), Galesville Medical Center (Galesville), Medical Associates (Beaver Dam), LaSalle Clinic (Waupaca).
METHODS: Data were collected from 401 pediatricians and family physicians from the Pediatric Research in Office Settings network and the Ambulatory Sentinel Practice Network. The physicians enrolled 22,059 consecutive office visits by children aged 4 to 15 years. Parents completed a survey that included the Family APGAR and the Pediatric Symptom Checklist. Clinicians completed a survey that described child psychosocial problems, treatments initiated or continued, and specialty care referrals.
RESULTS: Family dysfunction on the index visit often differed from dysfunction at follow-up (k=0.24). Only 31% of the families with positive Family APGAR scores at baseline were positive at follow-up, and only 43% of those with positive scores at follow-up had a positive score at the initial visit. There were many disagreements between the Family APGAR and the clinician. The Family APGAR was negative for 73% of clinician-identified dysfunctional families, and clinicians did not identify dysfunction for 83% of Family APGAR–identified dysfunctions (k=0.06).
CONCLUSIONS: Our data do not support the use of the Family APGAR as a measure of family dysfunction in the primary care setting. Future research should clarify what it does measure.
A strong family orientation has been a cornerstone of family practice since its emergence in the late 1960s1-4 and is alo important in pediatrics.5 The development of family medicine as a dominant primary care specialty has occurred in parallel with the development of clinical applications of family systems theory.6-9 More recently the Institute of Medicine report on primary care in America10 has reaffirmed provision of care in the context of family and the community as a central component of primary care.
Integrating an effective family orientation into everyday practice has proved feasible and extant in family practice.11-13 Several approaches to examining and characterizing family function for research purposes have been proposed.14 These include a combination of analysis of communication, observation of interaction, and individual patient report. Many of these approaches are time consuming and not practical for use in large sample studies requiring a brief instrument. The ability to assess the family context, however, is critical to many primary care studies and particularly those that deal with behavior, mental health, and psychosocial problems.
The Family APGAR was introduced by Gabriel Smilkstein in 1978 to assess adult satisfaction with social support from the family.15 It draws its name from a 5-item measure of perceived family support in the domains of adaptation, partnership, growth, affection, and resolve. The statements focus on the emotional, communicative, and social interactive relationships between the respondent and his or her family, for example: “I find that my family accepts my wishes to take on new activities or make changes in my lifestyle.”
Several studies have examined the psychometric properties of the instrument. We focus on evidence about the validity of the instrument, as it has regularly been found to be internally consistent.16,17 Good and colleagues16 found that Family APGAR scores correlated highly (r=0.80) with scores on the Pless-Satterwhite Family Function Index18 in a small nonclinical sample (N=38). In a small sample of mental health outpatients (N=20), the same authors found that Family APGAR scores correlated (r=0.64) with therapists’ ratings of the degree of family distress. Foulke and coworkers19 administered the Family APGAR and the Family Adaptation and Cohesion Evaluation Scales20 (FACES II) to 140 families and found that the Family APGAR correlated with the FACES Cohesion scale (r=0.70) and with the Adaptability scale (r=0.59; Stephen Zyzanski, personal communication, June 2000). However, when Clover and colleagues21 administered the Family APGAR and the FACES II to 66 families they reported that there was no association between the 2 scales.
Smucker and coworkers22 found no association (k=-0.05) between the Family APGAR and physicians’ judgments about the presence of family dysfunction among 152 families. This lack of association, however, could have resulted from the physicians’ difficulties in recognizing family dysfunction, problems in the Family APGAR, or both. North and colleagues23 obtained ratings of the usefulness of family assessment tools from 299 family physicians. The Family APGAR was rated less useful than any other tool.
Smilkstein and coworkers17 found that adults in counseling perceived their families as more dysfunctional than adults in other samples. There was, however, no assessment of the family; thus the finding does not directly support the validity of the Family APGAR as a measure of family dysfunction. Smilkstein and colleagues also found that adopted children were more likely to perceive their families as dysfunctional than were biological children. This would not validate the Family APGAR as a measure of family dysfunction, because it seems unlikely that families who adopt are more dysfunctional than other families.
A few studies have examined whether low Family APGAR ratings (which mean higher perceived family dysfunction) predict other clinical phenomena,12,22,24-27 with mixed results. However, an association between a patient’s report on the Family APGAR and later mental health service use does not directly bear on whether the instrument is a valid measure of family dysfunction. Therefore, the evidence of whether the Family APGAR is a valid measure of family dysfunction is mixed.
We used the Family APGAR as a measure of family dysfunction in a large study of psychosocial problems in children. Our study accomplished 3 goals that had not been achieved in previous research. First, we examined the internal consistency of the Family APGAR in a very large sample of office-based visits (N=21,285). In an internally consistent survey the items essentially measure one thing. Researchers who use the Family APGAR to compare families on the basis of functionality are assuming that there is a single dimension of family characteristics that is tapped by the survey.
Second, we used a large sample of repeat office visits (N=1146) to examine whether positive (dysfunctional) Family APGAR scores are stable over 6 months. When health service workers speak of dysfunctional families they often mean those that are persistently dysfunctional. Similarly, clinicians often adopt a watch and wait strategy for dealing with an initial report of psychosocial problems.28 If a positive score on the Family APGAR signals persistent dysfunction, then a positive score at the index visit should usually be matched by a positive score at follow-up.
Third, in a large sample of office visits (N=4050) we examined whether an adult family member’s report of problems on the family APGAR matched the clinician’s independent judgments of whether there were family problems. Although there are many reasons to expect disagreements between a valid survey measure of family dysfunction and clinicians’ judgments, a very weak level of agreement would raise questions about whether the Family APGAR measured dysfunctionality.
Methods
Study Sites
The Child Behavior Study (CBS) was conducted in several large primary care research networks in North America. The Ambulatory Sentinel Practice Network (ASPN), a family practice research network, included 141 practices in 41 states and 6 Canadian provinces and was composed of approximately 680 clinicians. Eightyfive percent of the ASPN clinicians were family physicians, 7% were nurse practitioners, and 8% were physician assistants. Additional family physician participants came from the Wisconsin Research Network and the Minnesota Academy of Family Physicians Research Network, which had characteristics similar to ASPN. The primary care practice–based Pediatric Research in Office Settings (PROS) network included more than 1500 clinicians from more than 480 pediatric practices in all 50 states and the Commonwealth of Puerto Rico. Eightynine percent of the PROS clinicians were pediatricians, 10% were nurse practitioners, and 1% were physician assistants. Of the 206 practices participating in the CBS, 30% were urban, 38% were suburban, and 32% were rural.
All clinicians participating in the CBS were included for our analysis (401 clinicians in 44 states, the Commonwealth of Puerto Rico, and 4 Canadian provinces). The clinicians included 267 pediatricians, 134 family practitioners, and 29 nurse clinicians. Previous research from both ASPN and PROS confirmed the comparability of patients, clinicians, and practices in primary care network studies with those identified in national samples.29,30 In addition, we compared participating pediatric clinicians with a random sample of pediatricians from the American Academy of Pediatrics31 on demographic factors and practice characteristics. We found few differences between participating clinicians and other clinicians.
Patient Sample
Each participating clinician enrolled a consecutive sample of approximately 55 children aged 4 to 15 years presenting for nonemergent care with a parent or primary caretaker. We enrolled each child only once and excluded children seen for procedures only. Some eligible children were not recruited, primarily because of parental refusal (63% of eligible but not participating children) and occasionally because the opportunity was either overlooked by the office staff (25%) or because the family dropped out of the study (12%). We compared participating children with those who were eligible but not participating on the basis of age and sex, and found no differences. In addition, we examined whether clinician or practice characteristics might affect patient participation, including clinician discipline, geographic region, practice population size, percentage of managed care patients, and clinician attitudes toward mental health treatment. Only those clinicians located in the South and West seemed to include a higher percentage of their eligible participants (94% to 89% for each); none of the other measured sources of selection bias were statistically significant.
This procedure produced a sample of 22,059 children seen in office visits. Among those visits 774 (3.5%) with missing data on 1 or more of the 5 APGAR items were excluded, resulting in a final study sample of 21,285 visits.
Procedures
Procedures and consent forms for the CBS were approved by institutional review boards affiliated with PROS, ASPN, and the University of Pittsburgh. Study procedures have been described in detail elsewhere32 Consenting parents (or the accompanying primary caregiver) filled out a parent questionnaire while waiting to see the clinician. The questionnaire included demographic data, the Family APGAR, and the Pediatric Symptom Checklist (a psychosocial screening instrument). The clinician did not see the completed Family APGAR, Pediatric Symptom Checklist, or other parent questionnaire data.
After seeing a patient the clinician completed a survey about the encounter, documenting whether a new, ongoing, or recurrent psychosocial problem was present, including an explicit statement of family dysfunction. Finally, the survey also included a checklist of a series of psychosocial problems that the clinician might have recognized in the child (clinicians could and often did check more than one problem).
Procedures for Follow-up
A random sample of children with clinician-identified psychosocial problems was identified for follow-up. African American children were oversampled for follow-up to obtain a sufficient sample. A total of 1970 children were included in the follow-up, and 1354 (69%) were successfully followed up. For this analysis, we used the 1146 patients with complete APGAR data for whom the adult respondent was the same at enrollment and follow-up.
Results
Table 1 shows the associations between the Family APGAR scores and several demographic variables. The strongest predictor of a low Family APGAR score was when the child’s parents were either not married or were separated. Table 2 presents the results for individual items of the Family APGAR.
Internal Consistency
We examined the intercorrelation of the Family APGAR items to determine whether the scale measured a single dimension of family functioning. The correlations of items with the total score ranged from r=0.63 to 0.71. Coefficient a, a summary measure of the intercorrelation of items, equaled 0.85, and deletion of any item from the scale reduced the a. This is a respectable level of internal consistency, suggesting that the Family APGAR items can all be viewed as measures of a single underlying dimension.
Stability of Family APGAR Over Time
Table 3 compares response on the Family APGAR on the initial and follow-up visits. There was a slight but statistically significant difference between the frequency of positive scores (Ž5, indicating family dysfunction) with families appearing more dysfunctional at the follow-up (McNemar’s test: (c2[1]=29.02; P <.001).
If the Family APGAR measures a stable characteristic of family functioning, a family’s dysfunctional status at the index visit should usually agree with its status at the 6-month follow-up. However, only 31% of families appearing dysfunctional during the initial visit still seemed so during the follow-up, and only 43% of those appearing dysfunctional during the follow-up appeared so during the initial visit. The k statistic, a chance-corrected measure of the agreement between the time 1 and time 2 scales, was only 0.24.
Clinician Assessment of Psychosocial Problem and the Family APGAR
Table 4 presents the concordance between a positive score on the Family APGAR and clinicians’ identification of family dysfunction. This Table includes the subset of children for which clinicians recognized a psychosocial problem, because this is the group for which a clinician would be likely to use the Family APGAR. There were high rates of disagreement between clinicians and the scale concerning positive cases. The Family APGAR was negative for 73% of clinician-identified dysfunctional families, and clinicians did not identify dysfunction for 83% of APGAR-identified dysfunctions. Although there was a significant positive association between the Family APGAR and clinician identifications (c2[1]) =19.12; P <.002), the k agreement statistic was only 0.06.
Discussion
Our study adds important new information on the performance of the Family APGAR as a measure of family support and dysfunction. Our results confirm some of the previous work that found that the Family APGAR is an internally consistent measure. Nevertheless, it is unclear exactly what it measures. The Family APGAR did not remain stable across assessments that averaged 6 months apart. On the one hand, it is correlated with both parental reports of symptoms and physician treatment decisions. Previously we reported27 an association of Family APGAR with behavioral problems in children as assessed by both physicians’ reports and scores on the Pediatric Symptom Checklist.22 Also, positive Family APGARs were more frequent among single or separated parents. This could reflect a higher level of dysfunction among such families, but it is equally consistent with the premise that the Family APGAR is a measure of family support.
Less than a third of the families who screened positive at time 1 on the Family APGAR also screened positive at follow-up. By the design of our study, only families in which the clinician had recognized a psychosocial problem at time 1 were followed up. It is plausible that the prevalence of family dysfunction among such families is higher than the general population. If so, it implies that in the general population, the rate of positive time-2 Family APGAR scores given positive time-1 scores would be even lower. The pattern of strong internal consistency and weak temporal stability suggests that the Family APGAR tracks a labile characteristic of families. By itself the transience of positive Family APGAR scores does not imply that it is an invalid measure of dysfunction. It is possible that family dysfunction can be serious but transient. Given what the Family APGAR actually measures, however, this interpretation is hard to support. A positive Family APGAR score is a report by a single individual that the family does not adequately communicate with, emotionally support, adapt to, or problem-solve with him or her. Is one such report evidence of serious family dysfunction, or does normal family functioning include occasional transient disturbances of the relations between a family and one of its members? We incline to the latter view.
Also, the Family APGAR was not associated with clinician reports of family dysfunction. Disagreement between clinicians and the Family APGAR does not necessarily imply that the Family APGAR is wrong. It is likely that clinicians have difficulty recognizing family dysfunction, as would be suggested by the literature showing that clinicians often fail to recognize psychiatric disorder.32,33 In the latter case, however, it has been found that primary care clinicians’ judgments about the presence of psychiatric disorders fail to agree with well-validated psychiatric instruments. Given the scarcity of previous evidence about the validity of the Family APGAR, we do not believe that the lack of agreement between the clinicians and the Family APGAR implies that the clinicians were in error. All we can say is that there is little agreement between the instrument and clinician-identified family dysfunction.
Limitations
Our data have important limitations for examining the ability of the Family APGAR to provide a measure of family function. In our study, the Family APGAR was reported by a single adult in the family. We do not have data from other adults in the family or the index child in the study. Also, we do not have a gold standard criterion assessment of family dysfunction. Our study focused on psychosocial problems in the index child, and thus we did not document a complete picture of the psychosocial problems of the family. Finally, entry into the study was through a child’s visit; this was not a systematic sample of families visiting primary care offices. Therefore, our study oversampled families of children who were presenting for a medical or psychosocial problem.
Future Research
The Family APGAR appears to have utility in family practice research, but researchers should carefully consider how they are using it. Further research could provide a more complete explanation of the association between distress as measured by the Family APGAR and psychosocial problems in children. Our speculation is that because the Family APGAR assesses an adult’s perceptions of family support, low scores may measure parental distress, which will sometimes reflect parental depression. The detrimental effects of parental depression on children are well established. This suggests that the Family APGAR may be an important variable to investigate as a determinant of care-seeking behavior, parent and physician treatment decisions, and as a marker for problems in one or more children. It might be more efficient, however, to screen for parental depression.
Conclusions
Although originally introduced as an assessment of adult satisfaction with family support, the Family APGAR has developed a research following as a measure of family functioning. We present data from a large community-based study of behavioral problems in children that raise questions about the Family APGAR as a measure of family dysfunction. Viewed in the light of the scarcity of previous evidence on the validity of the Family APGAR, we do not believe it should be used as a measure of family functioning. However, because the Family APGAR is associated with child psychosocial problems, it remains of interest for clinical research. One of the goals of future research should be to clarify what the Family APGAR does measure.
We note, however, that the fundamental problems we have discussed may not be in the Family APGAR but rather in the lack of clarity about the meaning of family dysfunction.13 What is the justification for using a measure of social support as a measure of family dysfunction? Many other issues arise in discussion of dysfunctional families, such as parental drug use, the lack of a stable family residence, neglectful child rearing, and the occurrence of domestic violence. The Family APGAR was never intended to measure these issues. A prerequisite for future research must be a clarification of what it means to label a family as dysfunctional.
Acknowledgments
This work was supported by a grant from the National Institute of Mental Health (MH 50629; PI: Kelleher), the Bureau of Maternal and Child Health, and the Staunton Farm Foundation. The authors gratefully acknowledge the contributions of the Pediatric Research in Office Settings (PROS) network of the American Academy of Pediatrics, Elk Grove Village, Ill; the Ambulatory Sentinel Practice Network (ASPN), Denver, Colo; the Wisconsin Research Network (WReN), Madison, Wis; and the Minnesota Academy of Family Physicians Research Network (MAFPRN), St. Paul, Minn. We are particularly grateful for the effort of Diane Comer in both the research and the preparation of this manuscript.
PARTICIPATING CBS PRACTICES
PROS Practices: The pediatric practices or individual practitioners who completed this study are listed by American Academy of Pediatrics chapter: Alabama: Drs Heilpern and Reynolds, PC (Birmingham); Alaska: Anchorage Neighborhood Health Center (Anchorage); Arizona: Mesa Pediatrics Professional Association (Mesa), Pediatric Ambulatory Care Clinic (Phoenix), Orange Grove Pediatrics (Tucson); California 1: Anita Tolentino-Macaraeg, MD (Hollister), Palo Alto Medical Foundation (Los Altos); Colorado: Arvada Pediatric Associates (Arvada), Family Health Center (Denver), Gino Figlio, MD (Lamar); Connecticut: Gerald Jensen, MD (Bristol), Barry Keller, MD (Danbury), Community Health Services (Hartford), St. Francis Pediatric Primary Care Center (Hartford); Florida: Atlantic Coast Pediatrics (Merritt Island), Children’s Clinic (Tallahassee); Georgia: The Pediatric Center (Stone Mountain); Hawaii: Melinda Ashton, MD (Honolulu), Straub Clinic—Pediatrics (Aiea); Iowa: Newborn & Pediatric Specialist, PC (Des Moines), David Kelly, MD (Marshalltown); Illinois: SIU Physicians & Surgeons (Auburn), Emalee Flaherty, MD (Chicago), Southwest Pediatrics (Palos Park); Indiana: Bloomington Pediatric Association (Bloomington), Community Health Access Program (Bloomington), Drs. Mary Jo Stine and Richard Weiner (Indianapolis), Jeffersonville Pediatrics (Jeffersonville), Pediatric Advocates (Peru); Kansas: Bethel Pediatrics (Newton); Kentucky: Tri-State Pediatrics, PSC (Ashland); Louisiana: Children’s Clinic of Southwest LA (Lake Charles); Maine: John Salvato, MD (Waterville), Intermed Pediatrics (Yarmouth); Maryland: O’Donovan & Ahluwalia, MD, PA (Baltimore), Children’s Medical Group (Cumberland), Shore Pediatrics (Easton), Clinical Associates Pediatrics (Towson/Woodlawn); Massachusetts: Holyoke Pediatric Associates (Holyoke), Medical Associates (Leominster), The Fallon Clinic (Worcester); Michigan: University Pediatricians, P.C. (Detroit), Pediatric Associates of Farmington (Farmington), Mott Children’s Health Center (Flint), H.. Hildebrandt, MD (Ypsilanti); Montana: Stevensville Pediatrics (Stevensville); Nebraska: Southwest Pediatrics (Omaha); Nevada: Capital Medical Associates (Carson City), Physician’s Center West (Fallon); New Hampshire: Exeter Pediatric Associates (Exeter); New Jersey: Delaware Valley Pediatric Association (Lawrenceville); New Mexico: Albuquerque Pediatric Associates (Albuquerque); New York 1: Pediatric Associates (Camillus), Elmwood Pediatric Group (Rochester), Park Medical Group (Rochester), Edward D. Lewis, MD (Rochester), Panorama Pediatric Group (Rochester), Amherst Pediatric Associates (Williamsville); New York 2: Centro Medico (Jackson Heights); New York 3: Pediatric Office at Roosevelt Island (New York); North Carolina, Triangle Pediatric Center (Cary), Goldsboro Pediatrics (Goldsboro), Medical Association of Surry (Mount Airy), Peace Haven Family Health Center (Winston-Salem); North Dakota: MeritCare MedicalGroup-Pediatrics (Fargo), Altru Clinic (Grand Forks), Dakota Clinic (Jamestown), Medical Arts Clinic (Minot); Ohio: Oxford Pediatrics & Adolescents (Oxford), Pediatrics (Portsmouth), St. Elizabeth Health Center (Youngstown); Oklahoma: Eastern Oklahoma Medical Plaza (Poteau), Shawnee Medical Center Clinic (Shawnee), Pediatric & Adolescent Care (Tulsa); Pennsylvania: Pediatric Practice of Northeastern (Honesdale), Schuylkill Pediatrics (Pottsville), Cevallos and Moise Pediatric Associates, PC (Quakertown), Pennridge Pediatric Associates (Sellersville); Puerto Rico: Ethel Lamela, MD (Isabela), Primary Care Pediatric Clinic Catano (Rio Piedras); Rhode Island: Marvin Wasser, MD (Cranston); South Carolina: Carolina Primary Care (Columbia); Tennessee: Johnson City Pediatrics (Johnson City); Texas: The Pediatric Clinic (Greenville), Department of Pediatrics (Lackland Air Force Base), MD Pediatric Associates (Lewisville), Winnsboro Pediatrics (Winnsboro); Utah: Gordon Glade, MD (American Fork), Mountain View Pediatrics (Sandy), Salt Lake Clinic (Sandy), Granger Medical Center (West Valley City); Vermont: CHP Brattleboro Pediatrics (Brattleboro), University Pediatrics (Burlington), Rebecca Collman, MD (Colchester), Essex Pediatrics (Essex Junction), Mousetrap Pediatrics (Milton), CHP Timber Lane Pediatrics (South Burlington), Joseph Hagan, Jr, MD (South Burlington), Practitioners of Pediatric Medicine (South Burlington), University Pediatrics (Williston); Virginia: Drs. Casey, Goldman, Lischwe, Garrett & Kim (Arlington), James River Pediatrics (Midlothian), Pediatric Faculty Practice Office (Richmond); Washington: Jemima Tso, MD (Auburn), Redmond Pediatrics (Redmond), Rockwood Clinic (Spokane); West Virginia: Tess Alejo (Martinsburg), Medical & Pediatric Associates (Parkersburg), Grant Memorial Pediatrics (Petersburg); Wisconsin: Beloit Clinic SC (Beloit), Middleton Pediatric Clinic (Middleton), Waukesha Pediatric Associates (Waukesha), Gundersen Clinic-Whitehall (Whitehall); Wyoming: Cheyenne Children’s Clinic (Cheyenne), Jackson Pediatrics (Jackson).
ASPN Practices: Arkansas: Batesville Family Practice Center (Batesville); California: Foothills Family Medical Group (Auburn), Loma Linda Family Medical Group (Loma Linda); Colorado: Renate Justin, MD (Fort Collins), Harrington, Knaus, & Spence, PC (Carbondale), La Mariposa Clinic (Denver), Colorado Springs Health Partners (Monument), Penrose Family Health Center (Penrose); Florida: The Family Doctors of Belleview (Belleview); Georgia: Titus Taube, MD (Warner Robbins); Louisiana: Family Medicine Center of Baton Rouge (Baton Rouge); Minnesota: Eagle Medical (Excelsior), Ramsey Clinic—Maplewood (Maplewood); New Hampshire: Mascoma Valley Community Care (Enfield) Hillsboro Medical Services (Hillsboro); New Jersey: A John Orzano, MD (Flemington), Community Care Center (Lebanon); New Mexico: Santa Fe Family Practice (Santa Fe); New York: Raj B. Kachoria, MD (Macedon), Canal Park Family Practice (Palmyra), Montefiore Comprehensive Family Care (Bronx), Mary Kay Ness, MD (Honeoye Falls); North Carolina: Bakersville Community Medical Clinic (Bakersville), Nalli Clinic (Matthews); North Dakota: University of North Dakota Family Practice Center—Minot (Minot) Minot Center for Family Medicine (Minot); Ohio: Center for Family Medicine (Cleveland); Oregon: Dunes Family Health Care, Inc. (Reedsport); Pennsylvania: John Farmer, DO (Waynesboro), Good Samaritan Family Practice (Lebanon); Tennessee: Michael H. Hartsell, MD (Greeneville), Mountain City Extended Hours Clinic (Mountain City); Texas: Van Horn Rural Health Clinic (Van Horn); Virginia: June Tunstall, MD (Surry); Tappahannock Family Practice (Tappahannock); West Virginia: North Fayette Family Health Center (Hico); Wisconsin: Kronenwetter Clinic (Mosinee); Alberta, Foothills Family Medicine Centre (Black Diamond); New Brunswick, David Ross, MD (Moncton); Newfoundland: Newhook Community Health Center (Whitbourne), Ross Thomas, MD (Sackville); Ontario: Steve Nantes, MD (Kitchener), Metcalfe & Dowdell (Kitchener), Bryan Alton, MD (Hamilton).
MAFPRN Practices: Family Medical Practice, PA (Willman), Family Medicine of Winona (Winona), River Valley Clinic (Hastings), Family Medicine Clinic of Lake Crystal (Lake Crystal), Gateway Family Health Clinic (Moose Lake), Eagan Medical Associates (Eagan), Fairview Uptown Clinic (Minneapolis), Bay Area Health Center (Silver Bay), West Side Health Center (St. Paul), Hopkins Family Physicians (Hopkins), Family Practice Center (St. Cloud), Mt. Royal Medical Center (Duluth), North Memorial Family Practice (Minneapolis).
WREN Practices: Poynette Family Practice Center (Poynette), Medical Associates (Baraboo), Plymouth Family Physicians (Plymouth), Monroe Clinic (Monroe), UCC/Mona Grove (Madison), Family Doctors-Black Creek (Black Creek), Southwestern Family Practice (South Milwaukee), Family Health Plan (Elm Grove), LaSalle Clinic (Appleton), Marshfield Clinic—Merril Center (Merrill), Tigerton Clinic (Tigerton), Dean Medical, (Oregon), Physicians Plus/Fitchburg (Fitchburg), Family Health Plan (Glendale), Franciscan Skemp Clinic (Tomah), Galesville Medical Center (Galesville), Medical Associates (Beaver Dam), LaSalle Clinic (Waupaca).
1. Carmichael LP. The family in medicine. J Fam Pract 1976;3:562-63.
2. Geyman JP. The family as object of care in family practice. J Fam Pract 1977;5:571-75.
3. Curry HB. The family as our patient. J Fam Pract 1977;4:757-58.
4. Medalie JH. A family-oriented approach in primary care. Boston, Mass: Little, Brown, and Company; 1976.
5. Kemper KJ, Kelleher KJ. Rationale for family psychosocial screening. Ambulatory Child Health 1996;1:311-24.
6. Ransom DC. The evolution from an individual to a family approach. New York, NY: Brunner-Mazel; 1985.
7. McDaniel S, Campbell TL, Seaburn DB. Family-oriented primary care. New York, NY: Springer-Verlag; 1990.
8. Doherty WB. Family therapy and family medicine. New York, NY: Guilford Press; 1983.
9. Crouch M, Roberts L, eds. The family in medical practice: a family systems primer. New York, NY: Springer-Verlag; 1987.
10. Institute of Medicine. Primary care: America’s health in a new era. Washington, DC: National Academy Press; 1996.
11. Medalie JH, Zyzanski S, Langa D, Stange KC. The family in family practice: is it a reality? J Fam Pract 1998;46:390-96.
12. Mengel M. The use of the family apgar in screening for family dysfunction in a family practice center. J Fam Pract 1987;24:394-98.
13. Mengel MB. The Family APGAR in a research setting. Fam Med 1988;20:143-44.
14. Jacob T, Tennenbaum DL. Family assessment methods. In: Assessment and diagnosis in Child Psychopathol New York, NY: Guilford Press; 1988;196-231.
15. Smilkstein G. The Family APGAR: a proposal for a family function test and its use by physicians. J Fam Pract 1978;6:1231-39.
16. Good MJ, Smilkstein G, Good BJ, Shaffer T, Arrons T. The family APGAR index: a study of construct validity. J Fam Pract 1979;8:577-82.
17. Smilkstein G, Ashworth C, Montano D. Validity and reliability of the Family APGAR as a test of family function. J Fam Pract 1982;15:303-11.
18. Pless JB, Satterwhite B. A measure of family functioning and its application. Soc Sci Med 1973;7:613-21.
19. Foulke FG, Reeb KG, Graham AV, Zyzanski SJ. Family function. respiratory illness, and otitis media in urban black infants. Fam Med 1988;20:128-32.
20. Olson DH, Bell R, Porter J. Family Adaptation and Cohesion Evaluation Scale II. Minneapolis, Minn: Family Inventories Project; 1982.
21. Clover RD, Abell T, Becker LA, Crawford S, Ramsey CN. Family functioning and stress as predictors of influenza B infection. J Fam Pract 1989;28:535-39.
22. Smucker WD, Wildman BG, Lynch TR, et al. Relationship between the Family APGAR and behavioral problems in children. Arch Fam Med 1995;4:535-39.
23. North S, Marvel MK, Hendricks B, Morphew P, North D. Physicians’ usefulness ratings of family-oriented clinical tools. J Fam Pract 1993;37:30-34.
24. Hilliard R, Gjerde C, Parker L. Validity of two psychological screening measures in family practice: personal inventroy and Family APGAR. J Fam Pract 1986;23:345-49.
25. Gwyther RE, Bentz EJ, Drossman DA, Berolzheimer N. Validity of the Family APGAR in patients with irritable bowel syndrome. Fam Med 1993;25:21-25.
26. Ramsey CN, Abell TD, Baker LC. The relationship between family functioning, life events, family structure, and the outcome of pregnancy. J Fam Pract 1986;22:521-27.
27. Murphy M, Kelleher K, Pagano M, et al. The Family APGAR and psychosocial problems in children: a report from ASPN and PROS. J Fam Pract 1998;46:54-64.
28. Gardner W, Kelleher K, Wasserman R, et al. Primary care treatment of pediatric psychosocial problems: a study from PROS and ASPN. Pediatrics 2000;106:E44.-
29. Green L, Miller R, Reed F, Iverson D, Barley G. How representative of typical practice are practice based research networks? A report from the Ambulatory Sentinel Practice Network, Inc (ASPN). Arch Fam Med 1993;2:939-49.
30. Nutting PA, Baier M, Werner JJ, Cutter G, Reed FM, Orzano AJ. Practice patterns of family physicians in practice-based research networks: A report from ASPN. J Am Board Fam Pract 1999;12:278-84.
31. American Academy of Pediatrics. Periodic survey of fellows. Elk Grove, Ill: AAP; 1995. Report No: 32.
32. Kelleher KJ, Childs GE, Wasserman RC, McInerney TK, Nutting PA, Gardner WP. Insurance status and recognition of psychosocial problems: a report from PROS and ASPN. Arch Pediatr Adolesc Med 1997;151:1109-15.
33. Costello EJ. Primary care pediatrics and child psychopathology: a review of diagnosis, treatment, and referral practices. Pediatrics 1986;78:1044-51.
34. Downey G, Coyne J. Children of depressed parents: an integrative review. Psychological Bull 1990;108:50-76.
1. Carmichael LP. The family in medicine. J Fam Pract 1976;3:562-63.
2. Geyman JP. The family as object of care in family practice. J Fam Pract 1977;5:571-75.
3. Curry HB. The family as our patient. J Fam Pract 1977;4:757-58.
4. Medalie JH. A family-oriented approach in primary care. Boston, Mass: Little, Brown, and Company; 1976.
5. Kemper KJ, Kelleher KJ. Rationale for family psychosocial screening. Ambulatory Child Health 1996;1:311-24.
6. Ransom DC. The evolution from an individual to a family approach. New York, NY: Brunner-Mazel; 1985.
7. McDaniel S, Campbell TL, Seaburn DB. Family-oriented primary care. New York, NY: Springer-Verlag; 1990.
8. Doherty WB. Family therapy and family medicine. New York, NY: Guilford Press; 1983.
9. Crouch M, Roberts L, eds. The family in medical practice: a family systems primer. New York, NY: Springer-Verlag; 1987.
10. Institute of Medicine. Primary care: America’s health in a new era. Washington, DC: National Academy Press; 1996.
11. Medalie JH, Zyzanski S, Langa D, Stange KC. The family in family practice: is it a reality? J Fam Pract 1998;46:390-96.
12. Mengel M. The use of the family apgar in screening for family dysfunction in a family practice center. J Fam Pract 1987;24:394-98.
13. Mengel MB. The Family APGAR in a research setting. Fam Med 1988;20:143-44.
14. Jacob T, Tennenbaum DL. Family assessment methods. In: Assessment and diagnosis in Child Psychopathol New York, NY: Guilford Press; 1988;196-231.
15. Smilkstein G. The Family APGAR: a proposal for a family function test and its use by physicians. J Fam Pract 1978;6:1231-39.
16. Good MJ, Smilkstein G, Good BJ, Shaffer T, Arrons T. The family APGAR index: a study of construct validity. J Fam Pract 1979;8:577-82.
17. Smilkstein G, Ashworth C, Montano D. Validity and reliability of the Family APGAR as a test of family function. J Fam Pract 1982;15:303-11.
18. Pless JB, Satterwhite B. A measure of family functioning and its application. Soc Sci Med 1973;7:613-21.
19. Foulke FG, Reeb KG, Graham AV, Zyzanski SJ. Family function. respiratory illness, and otitis media in urban black infants. Fam Med 1988;20:128-32.
20. Olson DH, Bell R, Porter J. Family Adaptation and Cohesion Evaluation Scale II. Minneapolis, Minn: Family Inventories Project; 1982.
21. Clover RD, Abell T, Becker LA, Crawford S, Ramsey CN. Family functioning and stress as predictors of influenza B infection. J Fam Pract 1989;28:535-39.
22. Smucker WD, Wildman BG, Lynch TR, et al. Relationship between the Family APGAR and behavioral problems in children. Arch Fam Med 1995;4:535-39.
23. North S, Marvel MK, Hendricks B, Morphew P, North D. Physicians’ usefulness ratings of family-oriented clinical tools. J Fam Pract 1993;37:30-34.
24. Hilliard R, Gjerde C, Parker L. Validity of two psychological screening measures in family practice: personal inventroy and Family APGAR. J Fam Pract 1986;23:345-49.
25. Gwyther RE, Bentz EJ, Drossman DA, Berolzheimer N. Validity of the Family APGAR in patients with irritable bowel syndrome. Fam Med 1993;25:21-25.
26. Ramsey CN, Abell TD, Baker LC. The relationship between family functioning, life events, family structure, and the outcome of pregnancy. J Fam Pract 1986;22:521-27.
27. Murphy M, Kelleher K, Pagano M, et al. The Family APGAR and psychosocial problems in children: a report from ASPN and PROS. J Fam Pract 1998;46:54-64.
28. Gardner W, Kelleher K, Wasserman R, et al. Primary care treatment of pediatric psychosocial problems: a study from PROS and ASPN. Pediatrics 2000;106:E44.-
29. Green L, Miller R, Reed F, Iverson D, Barley G. How representative of typical practice are practice based research networks? A report from the Ambulatory Sentinel Practice Network, Inc (ASPN). Arch Fam Med 1993;2:939-49.
30. Nutting PA, Baier M, Werner JJ, Cutter G, Reed FM, Orzano AJ. Practice patterns of family physicians in practice-based research networks: A report from ASPN. J Am Board Fam Pract 1999;12:278-84.
31. American Academy of Pediatrics. Periodic survey of fellows. Elk Grove, Ill: AAP; 1995. Report No: 32.
32. Kelleher KJ, Childs GE, Wasserman RC, McInerney TK, Nutting PA, Gardner WP. Insurance status and recognition of psychosocial problems: a report from PROS and ASPN. Arch Pediatr Adolesc Med 1997;151:1109-15.
33. Costello EJ. Primary care pediatrics and child psychopathology: a review of diagnosis, treatment, and referral practices. Pediatrics 1986;78:1044-51.
34. Downey G, Coyne J. Children of depressed parents: an integrative review. Psychological Bull 1990;108:50-76.
The Use of Complementary and Alternative Medicine by Primary Care Patients
METHODS: We conducted a survey of 542 patients attending 16 family practice clinics that belong to a community-based research network in San Diego, California, to determine patients’ reasons for using CAM therapies in conjunction with a visit to a family physician and the impact of these therapies on their health and well-being.
RESULTS: Approximately 21% of the patients reported using one or more forms of CAM therapy in conjunction with the most important health problem underlying their visit to the physician. The most common forms of therapy were visiting chiropractors (34.5% of CAM users), herbal remedies and supplements (26.7%), and massage therapy (17.2%). Recommendations from friends or coworkers, a desire to avoid the side effects of conventional treatments, or failure of conventional treatments to cure a problem were the most frequently cited reasons for using these therapies. Use of practitioner-based therapies was significantly and independently associated with poor perceived health status, poor emotional functioning, and a musculoskeletal disorder, usually low back pain. Use of self-care–based therapies was associated with high education and poor perceived general health compared with a year ago. Use of traditional folk remedies was associated with Hispanic ethnicity.
CONCLUSIONS: Sociodemographic characteristics and clinical conditions that predict use of CAM therapies by primary care patients in conjunction with a current health problem vary with the type of therapy used.
Within the past 5 years several studies have pointed to the widespread use of complementary and alternative medicine (CAM) in the United States. In a study conducted in 1991,1 1 in 3 respondents in a national sample of adults reported using at least one unconventional therapy in the past year. By 1997 that number had risen to more than 4 in 10.2 Similar studies conducted in Europe3 and Canada4 have reported utilization rates between 18% and 75%. Although several studies have found substantial use among patients attending specialty clinics,5-7 between 28% and 50% of family practice patients have been found to have used some form of CAM.8-10 In response to its widespread use, CAM has gained increasing acceptance among family physicians and other primary care physicians and in schools of medicine where more courses are being taught on the subject.11
Despite this increased use and acceptance by patients and health care providers, there is relatively little information available concerning the reasons for use of the various forms of therapies and treatments considered alternative or complementary. Recent surveys of the extent of use of these treatments provide little insight into why certain patients are more likely to use CAM therapies in general or specific therapies, such as chiropractic, message, herbal therapy, acupuncture, and homeopathy. Small studies of specific groups of patients suggest that use of these therapies is associated with the disease and with patient characteristics such as education and level of dissatisfaction with the primary care provider,9-10 but the extent to which these findings are generalizable to all primary care patients remains unclear.
A second limitation to our understanding of the use of CAM therapies by primary care patients is that the effectiveness of these therapies has not been subjected to rigorous examination. Their use by the general public appears to be based on anecdotal evidence, primarily personal experience or the experiences of others. Although there is a consensus within the medical establishment that most of these therapies are harmless,12 there is increasing evidence of the adverse consequences related to their use and misuse.13 However, much of this evidence is also anecdotal, based on case reports and not on large population-based studies.
Our objective was to address these 2 deficiencies in the understanding of the likelihood of use and the effectiveness of CAM therapies by conducting a large survey of a diverse population of patients attending family practice clinics in several different settings throughout the San Diego area, making use of a recently formed network of family physicians committed to community-based primary care research. Our goal for this survey was to examine the characteristics of primary care patients who use CAM therapies, determine whether these characteristics are significantly different for patients who have not used these therapies, and determine whether use of CAM therapies is associated with clinical condition, functional status, and quality of life.
Methods
Subjects
Our subjects included 541 patients aged 18 years and older visiting 16 family practice clinics in the San Diego area during a 3-month period (June 1999-August 1999). Each of these practices was a member the recently formed San Diego Unified Research in Family Medicine Network (SURF*NET). In this practice-based research initiative, community physicians, faculty and educators of academic family medicine programs combine research and clinical practice to develop a vital body of knowledge in the discipline of family medicine. The 16 clinics participating in the study represented more than 40 family physicians who were SURF*NET members and a patient population of more than 30,000, covering a broad and representative cross-section of the San Diego community.
To participate in the study patients had to identify a specific health complaint as the reason for a clinic visit. Individuals who made a visit for a general physical examination or to transport a pediatric patient were excluded. Our study participants represented 89% of all patients who met eligibility criteria and were invited to participate. Participants and nonparticipants exhibited no significant differences with respect to age, sex, ethnicity, or insurance status.
Data
Patients were asked to complete a questionnaire administered by a survey worker in the waiting room before the scheduled visit with a family physician. The survey instrument included questions about the social and demographic characteristics of the patient, including age, sex, marital status, ethnicity, place of birth, level of acculturation, education, and 1998 household income. Level of acculturation was assessed on the basis of a 5-item scale used in previous studies of patient populations.14 Patients were then grouped into acculturation categories: low, medium, and high. They were also grouped into categories based on their level of education (no college, some college, and college graduate), the method of payment for the clinic visit (cash, Medi-Cal/Medicare, and health maintenance organization or health insurance), and median 1998 household income (<$50,000 or Ž$50,000).
The Medical Outcomes Study Short Form (SF-36)15 was used to assess a patient’s current health status and quality of well-being. The patients were evaluated on the basis of physical and social functioning, physical and emotional role functioning, mental health, energy or fatigue, pain, general perceived health compared with others the same age, and general health compared with a year ago.
Finally, each patient was asked to describe the most important or significant health problems experienced during the past year and whether any of these problems had precipitated the current clinic visit. Health problems were then coded by investigators according to International Classification of Diseases—9th revision—Clinical Modification (ICD-9-CM) criteria.16 Symptoms and complaints that could not be attributed to a specific diagnosis were placed under the ICD-9-CM category of Symptoms and Ill-Defined Conditions. Using a list derived from previous studies,1,2 patients were then asked whether they had used one or more of 16 CAM therapies or therapists for their principal medical condition during the past 12 months. Information was also collected on the level of satisfaction with these CAM therapies, level of satisfaction with care provided by their family physician for the problem, reasons for using a CAM therapy or therapist, and whether the family physician had been notified by the patient that he or she was using such alternatives. Patients’ level of satisfaction with conventional and CAM treatments was rated on a scale from 1 (not at all satisfied) to 10 (completely satisfied). Reasons for using a CAM therapy were derived from a list compiled by Lazar and O’Connor.17
Statistics
Univariate statistics (percentages and means) were used to describe the characteristics of CAM use. Bivariate analyses (chi-square tests for categorical variables and paired-sample t tests and analysis of variance for continuous measures) were used to determine whether there were any significant differences between patients reporting use of any form of CAM therapy in the past year and those reporting no use of such therapy with respect to the following predictors: (1) social and demographic characteristics; (2) functional status and quality of wellbeing; (3) dissatisfaction with conventional treatments; and (4) ICD-9-CM diagnostic category of chief health complaint. Similar analyses were performed by 3 classes of therapy: (1) practitioner-based therapies (acupuncture, biofeedback, chiropractic, homeopathy, massage therapy, naturopathy); (2) self-care based therapies (energy healing, meditation and prayer, dietary interventions, herbal remedies, multivitamin supplements); and (3) traditional folk remedies. When appropriate (ie, based on the number of users), analyses were also conducted for individual types of therapy (chiropractic, acupuncture, herbal remedies, dietary interventions, massage therapy). Logistic regression models with stepwise entry of all potential independent variables were used to assess the odds of using a CAM therapy associated with each patient characteristic.
Results
Characteristics of CAM Users
Of the 541 adults participating in our study, 116 (21%) reported using 1 or more forms of CAM therapy or therapists within the past year for the primary health problem contributing to the present clinic visit. A visit to a chiropractor was the most frequent form of alternative therapy, followed by the use of herbal remedies or supplements, and message therapy Figure 1.
Approximately 60% of these patients had informed their physicians of the use of these CAM therapies. Of those who had not done so, 60% had indicated that there had been no previous opportunity to inform their physicians, since this had been the first visit for the health problem in question. When examined by class of CAM usage, approximately two thirds of those using practitioner-based and self-care– based therapies reported use to their physician, compared with 40% of those using traditional folk remedies.
The timing of initial use of CAM therapies in treating the current health problem is shown in Figure 2. In general, one third of all users of CAM therapies initiated treatment with one or more therapies before their initial visit to a primary care physician for the same clinical problem. Thirty-seven percent initiated use of CAM concurrent with (ie, within 2 weeks) of their initial visit to a primary care provider, and 1 in 5 (19%) initiated use of a CAM therapy after their initial primary care visit. One third (36%) of those using practitioner-based and self-care–based therapies and 46.2% of those using traditional folk remedies reported initiating therapy before a visit with a primary care provider. One third (36%) of the users of self-care–based therapies, 43% of users of practitioner-based therapies, and none of the users of traditional folk remedies reported using such a therapy concurrent with their initial clinic visit. One in 4 of the practitioner-based therapies (24.6%) and traditional folk remedies (23.1%) and 14% of users of self-care–based therapies reported initiating use after their initial visit to a primary care provider.
Approximately 1 in 4 patients reported using CAM to avoid side effects of regular treatment because a friend or coworker had recommended the treatment or because conventional treatment had failed to cure the problem Table 1. Between 10% and 15% of the patients reported using these therapies for philosophical reasons, because they preferred to deal with the problem by themselves, or because older family members had used these treatments for the same problem. Only 7 patients reported using therapies because they were unhappy with the attitude of family physicians. When examined by class of therapy, approximately 1 in 3 users of practitioner-based therapies reported using them to avoid the side effects of regular treatment, failure of regular treatment to cure their problem, and a recommendation from a friend or coworker. In addition to a preference for dealing with the problem by themselves, these 3 reasons (side effects, failure of regular treatment, and a recommendation from a friend) were also the primary reasons for use of self-care– based therapies. In contrast, use by parents and relatives for the same problem represented the primary reason for traditional folk remedies, accounting for slightly less than one third (30.8%) of the patients using them.
Predictors of CAM Use
A comparison of the social and demographic characteristics of users and nonusers of CAM is provided in Table 2. Use of CAM therapies was positively associated with level of education but inversely associated with level of acculturation. When examined by specific categories of CAM, women were significantly more likely than men to use herbal remedies (P <.05; data not shown) and other self-care– based forms of alternative medicine and traditional folk medicines. Level of education was positively associated with self-care–based forms of CAM in general and use of herbal (P <.001; data not shown) and dietary (P <.05; data not shown) remedies in particular. However, education was inversely associated with use of traditional folk remedies. Self-care-based therapies in general and herbal remedies in particular (P <.05; data not shown) were significantly associated with the level of household income. Use of traditional folk remedies was significantly associated with Hispanic ethnicity, place of birth, and low acculturation. Dietary remedies were positively associated with level of acculturation (P <.05; data not shown). Patients belonging to an health maintenance organization or possessing other forms of non-government-sponsored insurance were significantly more likely to use massage therapy (P <.05; data not shown) or herbal remedies (P <.05; data not shown).
The health status and quality of well-being of users and nonusers of CAM therapies and therapists is provided in Table 3. Users of CAM therapies reported significantly lower emotional role functioning and perceived general health compared with nonusers of the same age. Users of practitioner-based therapies reported significantly lower social functioning, physical and emotional role functioning, mental health, and perceived general health, and more pain than nonusers. Users of self-care–based therapies and traditional folk remedies reported significantly lower levels of general health than a year ago. Users of acupuncture (P=.03; data not shown) and chiropractors (P=.001; data not shown) reported significantly lower levels of general perceived health than nonusers (data not shown). Users of chiropractors also reported significantly higher levels of pain (P=.015; data not shown) than nonusers.
Musculoskeletal problems, usually back pain, were cited as the most common health problem associated with CAM use, followed by endocrine and metabolic diseases (primarily diabetes or obesity), diseases of the respiratory system (primarily asthma), and diseases of the genitourinary system Table 4. CAM users were approximately twice as likely as nonusers to have a musculoskeletal system disorder and 2.5 times as likely to have a genitourinary system disorder. Users of practitioner-based therapies were 2.7 times as likely to have a musculoskeletal system disorder as nonusers. Users of chiropractors were 3.7 times (P <.001; data not shown) and users of massage therapy were 2.2 times (P <.05; data not shown) as likely to have a musculoskeletal disorder as nonusers.
Users of CAM therapies in general (P <.01) and practitioner-based therapies (P <.01) and chiropractors (P <.001; data not shown) in particular reported significantly less satisfaction than nonusers with the conventional treatment they received from their family physician Figure 3. However, CAM users also reported no significant difference in the level of satisfaction with these therapies and the level of satisfaction with the care received from their family physician. Users of folk remedies reported significantly higher levels of satisfaction with conventional treatment than nonusers (P <.05), and higher levels of satisfaction with their CAM therapy than other CAM users (P <.05).
We examined the independent association between each of the sociodemographic and health status characteristics of patients and CAM use with a series of logistic regression models. Age, sex, marital status, place of birth, education, household income, method of payment, level of acculturation, satisfaction with primary care received, SF-36 measures of functional status, and presence or absence of a musculoskeletal disorder were entered into the model in a stepwise fashion. The best fitting models are presented in Table 5. Physical role functioning was the only significant independent predictor of use of one or more CAM therapies. The model correctly classified 70% of the patients. Musculoskeletal disorders, emotional functioning, and perceived general health were significant independent predictors of use of a practitioner-based therapy. The model correctly classified 86.4% of the patients. Patients with musculoskeletal disorders were 7.37 times (95% confidence interval [CI], 2.37-23.51) as likely to use some form of practitioner-based therapy (usually a chiropractor or massage therapist) as patients with other physical conditions. Level of education and general perceived health compared with a year ago were significant independent predictors of use of self-care–based therapies. The model correctly classified 86.4% of the patients. College graduates were 1.44 times (95% CI, 1.04-2.01) as likely to use some form of self-care–based therapy as patients with 12 or fewer years of education. Hispanic ethnicity was the only significant independent predictor of use of traditional folk remedies. The model correctly classified 97.8% of the patients. Hispanics were 10.27 times (95% CI, 3.10-33.95) as likely to use traditional folk remedies as members of other ethnic groups.
Logistic regression analyses predicting use of specific CAM therapies revealed that general perceived health was a significant independent predictor of use of acupuncture (b=-0.0665; standard error [SE]=0.0253; P=.0086) and chiropractors (b=-0.0314; SE=0.0139; P=.0243). A musculoskeletal disorder was a significant independent predictor of use of chiropractors (b=2.0123; SE=0.6720; P=.0005) and massage therapists (b=1.8467; SE=0.7792; P=.0178). Patients with musculoskeletal disorders were 10.23 times (95% CI, 2.00-27.92) as likely to use a chiropractor and 6.34 times (95% CI, 1.38-29.19) as likely to use a massage therapist as patients with some other clinical condition. Emotional functioning was also a significant independent predictor of use of massage therapy (b=-0.0902; SE=0.0347; P=.0094), and household income was a significant independent predictor of use of acupuncture (b=2.5045; SE=1.2428; P=.0439).
DISCUSSION
In our study approximately 1 in 5 patients reported having used some form of complementary or alternative therapy or therapist in the past year in conjunction with a current health problem. This percentage is smaller than the 42% reported by Eisenberg and colleagues.2 However, that study was a population-based survey and was not tied to use of primary care services. Although previous studies of primary care patients have reported higher percentages of patients using CAM therapies, differences in methods preclude a comparison with those percentages. Drivdahl and Miser,9 for example, reported that 28% of 177 family practice patients in a military clinic had ever used some form of alternative medicine. Elder and colleagues10 reported that 50% of 113 family practice clinic patients had used some form of CAM, but this includes use for reasons other than those precipitating the visit to the physician. To our knowledge this is the largest study conducted of a broad spectrum of family practice patients, focusing on use related to a specific health problem that precipitated a visit to a family physician.
Consistent with the findings of previous studies,1,2,9 use of CAM therapies in general and self-care–based therapies in particular by this group of primary care patients, was significantly associated with a high level of education. This may be attributed to the fact that better educated patients tend to be more informed or at least more likely to be exposed to information about the benefits of CAM.17,18 Consistent with previous research,1,2,17,18 use of self-care–based therapies was also associated with high household income in our study. Many of these therapies are not always covered by health insurance plans, which means they require cash payments that those with higher incomes are more likely to be able to afford.
In our study use of traditional folk remedies was inversely associated with levels of education and acculturation and positively associated with Hispanic ethnicity and being foreign-born. Low-acculturated, predominately Hispanic immigrants are more likely to be uninsured or underinsured,19 not able to afford cannot afford conventional treatment,20 and be more familiar with the efficacy of those traditional folk remedies.21
We also found that users of CAM therapies generally perceive their health to be worse compared with others their age and with their health a year ago than nonusers. Emotional functioning was also a significant independent predictor of use of practitioner-based therapies. Given the cross-sectional nature of the study design, it is difficult to determine whether health status is a cause or consequence of CAM use. However, a further examination of the experience of these patients with conventional treatment and their specific clinical condition provides some insight into this relationship.
Failure of regular treatment to cure the problem and a desire to avoid side effects were each cited by 1 in 4 users of CAM therapies in our study. Users were also significantly less satisfied than nonusers with the care they had received from their primary care provider. They were significantly more likely to report visiting a primary care provider for a musculoskeletal or genitourinary disorder than nonusers. Both of these disorders are noteworthy for the limited treatment effectiveness or inconsistency in treatment approach of conventional therapies.22,23 Taken together, these observations suggest that many users of CAM do so because they are dissatisfied with the care received from their primary care provider.
However, complementary and alternative medicine users were found to be no more satisfied with these alternatives than with the care received from their family physician. Furthermore, unhappiness with the attitude of their family physician and failure to correctly diagnose the problem was cited by a relatively small number of patients as a reason for using CAM therapies. Only 20% of users reported doing so after seeing a physician; the others did so either before or concurrent with a visit. This suggests that experience with the family physician played a relatively minor role in the decision of most patients to use CAM therapies.
Conclusions
The results of our study can be used to acquaint family physicians with the characteristics of users of complementary and alternative therapies and therapists in conjunction with specific health problems. Such information may have implications for diagnosis and treatment regimens, especially if contraindications to certain forms of treatment arise as a result of potential adverse reactions when used in combination with specific forms of CAM (eg, lead poisoning resulting from the use of certain traditional Mexican remedies used as laxatives24). In contrast to the results of previous studies,25 relatively few patients were reluctant to notify their primary care provider of their use of CAM therapies. However, patients were more likely to share this information if the physician made an effort to ask. An understanding of the characteristics associated with CAM use should enable the physician to obtain this information.
The results of our study suggest that patterns and predictors of CAM use by primary care patients vary with the type of therapy used. It is important to realize that all CAM therapies are not alike. Self-care–based therapies are more likely to be used by patients who are well educated, while traditional folk remedies are more likely to be used by Hispanic immigrants with low levels of education. Practitioner-based therapies are more likely to be used by patients with musculoskeletal disorders. Family physicians should keep these distinctions in mind when evaluating the likelihood of CAM use by a particular patient and potential implications for primary care delivery.
Acknowledgments
Our study was funded by a grant from the California Academy of Family Physicians.
Related Resources
National Center for Complementary and Alternative Medicine
A branch of the National Institutes of Health that conducts and supports basic and applied research and training and disseminates information on complementary and alternative medicine to physicians and the public. www.nccam.nih.gov
Whole Health MD
Information on combining alternative medicine, supplements, vitamins, herbs, and nutrition with conventional medicine. www.wholehealth.com
alternativeDr.com
Resources for books, reports, and practitioner searches serving various categories in alternative medicine therapies. www.alternativedr.com
HealthWorld Online
Alternative and Complementary Medicine Center-provides in-depth information about a range of therapies, as well as discussion forums, conference listings, and other resources. www.healthy.net/clinic/therapy/altmedcolumn
1. DM, Kessler RC, Foster C, Norlock FE, Calkins DR, Delbanco TL. Unconventional medicine in the United States: prevalence, costs, and patterns of use. N Engl J Med 1993;328:246-52.
2. DM, Davis RB, Ettner SL, et al. Trends in alternative medicine use in the United States, 1990-1997: results of a follow-up national survey. JAMA 1998;280:1569-75.
3. G. Alternative medicine in Europe: a quantitative comparison of the use and knowledge of alternative medicine and patient profiles in nine European countries. Brussels, Belgium: Belgium Consumers’ Association; 1987.
4. in five Canadians is using alternative therapies. Can Med Assoc J 1991;144:469.-
5. M, Fitzcharles MA. Alternative medicine use by rheumatology patients in a universal health care setting. J Rheumatol 1994;21:148-52.
6. W, O’Connor BB, MacGregor RR, Schwartz JS. Patient use and assessment of conventional and alternative therapies for HIV infection and AIDS. AIDS 1993;7:561-65.
7. MJ, Sutherland LR, Brkich L. Use of alternative medicine by patients attending a gastroenterology clinic. Can Med Assoc J 1990;142:121-25.
8. J, Shepherd S. Alternative or additional medicine? An exploratory study in general practice. Soc Sci Med 1993;37:983-88.
9. NC, Gillcrist A, Minz R. Use of alternative health care by family practice patients. Arch Fam Med 1997;6:181-84.
10. CE, Miser WF. The use of alternative health care by a family practice population. J Am Board Fam Pract 1998;11:193-99.
11. BM, Singh BB, Hartnoll SM, Singh BK, Reilly D. Primary care physicians and complementary-alternative medicine: training, attitudes, and practice patterns. J Am Board Fam Pract 1998;11:272-81.
12. M, Kassirer JP. Alternative medicine: the risk of untested and unregulated remedies. N Engl J Med 1998;339:839-41.
13. MJ, Anderson RA, Egeler RM, Wolff JEA. Alternative therapies for the treatment of childhood cancer. N Engl J Med 1998;339:846-47.
14. G, Sabogal F, Marin B, Otero-Sabogal R, Perez-Stable EJ. Development of a short acculturation scale for Hispanics. Hispanic J Behav Sci 1987;9:183-205.
15. JE,, Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36): I: conceptual framework and item selection. Med Care 1992;30:473-83.
16. Management Information Corporation. International classification of diseases-9th revision-clinical modification. Los Angeles, Calif: Practice Management Information Corporation; 1991.
17. JS, O’Connor BB. Talking with patients about their use of alternative therapies. Prim Care 1997;24:699-711.
18. JA. Why patients use alternative medicine: results of a national study. JAMA 1998;279:1548-53.
19. AM, ed. Acculturation: theory, models and some new findings. Boulder, Colo: Westview Press; 1980.
20. M, Richard C, Casebeer AW, Ray NF. Health insurance coverage among foreign-born US residents: the impact of race, ethnicity, and length of residence. Am J Public Health 1997;87:96-102.
21. AL, Mazur LJ. Use of folk remedies in a hispanic population. Arch Pediatr Adolesc Med 1995;149:978-81.
22. DC, MacCormack FA, Berg AO. Managing low back pain: a comparison of the beliefs and behaviors of family physicians and chiropractors. West J Med 1988;149:475-80.
23. AO. Variations among family physicians’ management strategies for lower urinary tract infection in women: a report from the Washington family physicians collaborative research network. J Am Board Fam Pract 1991;4:327-30.
24. for Disease Control and Prevention. Lead poisoning from Mexican folk remedies—California. MMWR Morb Mortal Wkly Rep 1993;42:521-24.
25. SR, Fosket JR. Disclosing complementary and alternative medicine use in the medical encounter: a qualitative study in women with breast cancer. J Fam Pract 1999;48:453-58.
METHODS: We conducted a survey of 542 patients attending 16 family practice clinics that belong to a community-based research network in San Diego, California, to determine patients’ reasons for using CAM therapies in conjunction with a visit to a family physician and the impact of these therapies on their health and well-being.
RESULTS: Approximately 21% of the patients reported using one or more forms of CAM therapy in conjunction with the most important health problem underlying their visit to the physician. The most common forms of therapy were visiting chiropractors (34.5% of CAM users), herbal remedies and supplements (26.7%), and massage therapy (17.2%). Recommendations from friends or coworkers, a desire to avoid the side effects of conventional treatments, or failure of conventional treatments to cure a problem were the most frequently cited reasons for using these therapies. Use of practitioner-based therapies was significantly and independently associated with poor perceived health status, poor emotional functioning, and a musculoskeletal disorder, usually low back pain. Use of self-care–based therapies was associated with high education and poor perceived general health compared with a year ago. Use of traditional folk remedies was associated with Hispanic ethnicity.
CONCLUSIONS: Sociodemographic characteristics and clinical conditions that predict use of CAM therapies by primary care patients in conjunction with a current health problem vary with the type of therapy used.
Within the past 5 years several studies have pointed to the widespread use of complementary and alternative medicine (CAM) in the United States. In a study conducted in 1991,1 1 in 3 respondents in a national sample of adults reported using at least one unconventional therapy in the past year. By 1997 that number had risen to more than 4 in 10.2 Similar studies conducted in Europe3 and Canada4 have reported utilization rates between 18% and 75%. Although several studies have found substantial use among patients attending specialty clinics,5-7 between 28% and 50% of family practice patients have been found to have used some form of CAM.8-10 In response to its widespread use, CAM has gained increasing acceptance among family physicians and other primary care physicians and in schools of medicine where more courses are being taught on the subject.11
Despite this increased use and acceptance by patients and health care providers, there is relatively little information available concerning the reasons for use of the various forms of therapies and treatments considered alternative or complementary. Recent surveys of the extent of use of these treatments provide little insight into why certain patients are more likely to use CAM therapies in general or specific therapies, such as chiropractic, message, herbal therapy, acupuncture, and homeopathy. Small studies of specific groups of patients suggest that use of these therapies is associated with the disease and with patient characteristics such as education and level of dissatisfaction with the primary care provider,9-10 but the extent to which these findings are generalizable to all primary care patients remains unclear.
A second limitation to our understanding of the use of CAM therapies by primary care patients is that the effectiveness of these therapies has not been subjected to rigorous examination. Their use by the general public appears to be based on anecdotal evidence, primarily personal experience or the experiences of others. Although there is a consensus within the medical establishment that most of these therapies are harmless,12 there is increasing evidence of the adverse consequences related to their use and misuse.13 However, much of this evidence is also anecdotal, based on case reports and not on large population-based studies.
Our objective was to address these 2 deficiencies in the understanding of the likelihood of use and the effectiveness of CAM therapies by conducting a large survey of a diverse population of patients attending family practice clinics in several different settings throughout the San Diego area, making use of a recently formed network of family physicians committed to community-based primary care research. Our goal for this survey was to examine the characteristics of primary care patients who use CAM therapies, determine whether these characteristics are significantly different for patients who have not used these therapies, and determine whether use of CAM therapies is associated with clinical condition, functional status, and quality of life.
Methods
Subjects
Our subjects included 541 patients aged 18 years and older visiting 16 family practice clinics in the San Diego area during a 3-month period (June 1999-August 1999). Each of these practices was a member the recently formed San Diego Unified Research in Family Medicine Network (SURF*NET). In this practice-based research initiative, community physicians, faculty and educators of academic family medicine programs combine research and clinical practice to develop a vital body of knowledge in the discipline of family medicine. The 16 clinics participating in the study represented more than 40 family physicians who were SURF*NET members and a patient population of more than 30,000, covering a broad and representative cross-section of the San Diego community.
To participate in the study patients had to identify a specific health complaint as the reason for a clinic visit. Individuals who made a visit for a general physical examination or to transport a pediatric patient were excluded. Our study participants represented 89% of all patients who met eligibility criteria and were invited to participate. Participants and nonparticipants exhibited no significant differences with respect to age, sex, ethnicity, or insurance status.
Data
Patients were asked to complete a questionnaire administered by a survey worker in the waiting room before the scheduled visit with a family physician. The survey instrument included questions about the social and demographic characteristics of the patient, including age, sex, marital status, ethnicity, place of birth, level of acculturation, education, and 1998 household income. Level of acculturation was assessed on the basis of a 5-item scale used in previous studies of patient populations.14 Patients were then grouped into acculturation categories: low, medium, and high. They were also grouped into categories based on their level of education (no college, some college, and college graduate), the method of payment for the clinic visit (cash, Medi-Cal/Medicare, and health maintenance organization or health insurance), and median 1998 household income (<$50,000 or Ž$50,000).
The Medical Outcomes Study Short Form (SF-36)15 was used to assess a patient’s current health status and quality of well-being. The patients were evaluated on the basis of physical and social functioning, physical and emotional role functioning, mental health, energy or fatigue, pain, general perceived health compared with others the same age, and general health compared with a year ago.
Finally, each patient was asked to describe the most important or significant health problems experienced during the past year and whether any of these problems had precipitated the current clinic visit. Health problems were then coded by investigators according to International Classification of Diseases—9th revision—Clinical Modification (ICD-9-CM) criteria.16 Symptoms and complaints that could not be attributed to a specific diagnosis were placed under the ICD-9-CM category of Symptoms and Ill-Defined Conditions. Using a list derived from previous studies,1,2 patients were then asked whether they had used one or more of 16 CAM therapies or therapists for their principal medical condition during the past 12 months. Information was also collected on the level of satisfaction with these CAM therapies, level of satisfaction with care provided by their family physician for the problem, reasons for using a CAM therapy or therapist, and whether the family physician had been notified by the patient that he or she was using such alternatives. Patients’ level of satisfaction with conventional and CAM treatments was rated on a scale from 1 (not at all satisfied) to 10 (completely satisfied). Reasons for using a CAM therapy were derived from a list compiled by Lazar and O’Connor.17
Statistics
Univariate statistics (percentages and means) were used to describe the characteristics of CAM use. Bivariate analyses (chi-square tests for categorical variables and paired-sample t tests and analysis of variance for continuous measures) were used to determine whether there were any significant differences between patients reporting use of any form of CAM therapy in the past year and those reporting no use of such therapy with respect to the following predictors: (1) social and demographic characteristics; (2) functional status and quality of wellbeing; (3) dissatisfaction with conventional treatments; and (4) ICD-9-CM diagnostic category of chief health complaint. Similar analyses were performed by 3 classes of therapy: (1) practitioner-based therapies (acupuncture, biofeedback, chiropractic, homeopathy, massage therapy, naturopathy); (2) self-care based therapies (energy healing, meditation and prayer, dietary interventions, herbal remedies, multivitamin supplements); and (3) traditional folk remedies. When appropriate (ie, based on the number of users), analyses were also conducted for individual types of therapy (chiropractic, acupuncture, herbal remedies, dietary interventions, massage therapy). Logistic regression models with stepwise entry of all potential independent variables were used to assess the odds of using a CAM therapy associated with each patient characteristic.
Results
Characteristics of CAM Users
Of the 541 adults participating in our study, 116 (21%) reported using 1 or more forms of CAM therapy or therapists within the past year for the primary health problem contributing to the present clinic visit. A visit to a chiropractor was the most frequent form of alternative therapy, followed by the use of herbal remedies or supplements, and message therapy Figure 1.
Approximately 60% of these patients had informed their physicians of the use of these CAM therapies. Of those who had not done so, 60% had indicated that there had been no previous opportunity to inform their physicians, since this had been the first visit for the health problem in question. When examined by class of CAM usage, approximately two thirds of those using practitioner-based and self-care– based therapies reported use to their physician, compared with 40% of those using traditional folk remedies.
The timing of initial use of CAM therapies in treating the current health problem is shown in Figure 2. In general, one third of all users of CAM therapies initiated treatment with one or more therapies before their initial visit to a primary care physician for the same clinical problem. Thirty-seven percent initiated use of CAM concurrent with (ie, within 2 weeks) of their initial visit to a primary care provider, and 1 in 5 (19%) initiated use of a CAM therapy after their initial primary care visit. One third (36%) of those using practitioner-based and self-care–based therapies and 46.2% of those using traditional folk remedies reported initiating therapy before a visit with a primary care provider. One third (36%) of the users of self-care–based therapies, 43% of users of practitioner-based therapies, and none of the users of traditional folk remedies reported using such a therapy concurrent with their initial clinic visit. One in 4 of the practitioner-based therapies (24.6%) and traditional folk remedies (23.1%) and 14% of users of self-care–based therapies reported initiating use after their initial visit to a primary care provider.
Approximately 1 in 4 patients reported using CAM to avoid side effects of regular treatment because a friend or coworker had recommended the treatment or because conventional treatment had failed to cure the problem Table 1. Between 10% and 15% of the patients reported using these therapies for philosophical reasons, because they preferred to deal with the problem by themselves, or because older family members had used these treatments for the same problem. Only 7 patients reported using therapies because they were unhappy with the attitude of family physicians. When examined by class of therapy, approximately 1 in 3 users of practitioner-based therapies reported using them to avoid the side effects of regular treatment, failure of regular treatment to cure their problem, and a recommendation from a friend or coworker. In addition to a preference for dealing with the problem by themselves, these 3 reasons (side effects, failure of regular treatment, and a recommendation from a friend) were also the primary reasons for use of self-care– based therapies. In contrast, use by parents and relatives for the same problem represented the primary reason for traditional folk remedies, accounting for slightly less than one third (30.8%) of the patients using them.
Predictors of CAM Use
A comparison of the social and demographic characteristics of users and nonusers of CAM is provided in Table 2. Use of CAM therapies was positively associated with level of education but inversely associated with level of acculturation. When examined by specific categories of CAM, women were significantly more likely than men to use herbal remedies (P <.05; data not shown) and other self-care– based forms of alternative medicine and traditional folk medicines. Level of education was positively associated with self-care–based forms of CAM in general and use of herbal (P <.001; data not shown) and dietary (P <.05; data not shown) remedies in particular. However, education was inversely associated with use of traditional folk remedies. Self-care-based therapies in general and herbal remedies in particular (P <.05; data not shown) were significantly associated with the level of household income. Use of traditional folk remedies was significantly associated with Hispanic ethnicity, place of birth, and low acculturation. Dietary remedies were positively associated with level of acculturation (P <.05; data not shown). Patients belonging to an health maintenance organization or possessing other forms of non-government-sponsored insurance were significantly more likely to use massage therapy (P <.05; data not shown) or herbal remedies (P <.05; data not shown).
The health status and quality of well-being of users and nonusers of CAM therapies and therapists is provided in Table 3. Users of CAM therapies reported significantly lower emotional role functioning and perceived general health compared with nonusers of the same age. Users of practitioner-based therapies reported significantly lower social functioning, physical and emotional role functioning, mental health, and perceived general health, and more pain than nonusers. Users of self-care–based therapies and traditional folk remedies reported significantly lower levels of general health than a year ago. Users of acupuncture (P=.03; data not shown) and chiropractors (P=.001; data not shown) reported significantly lower levels of general perceived health than nonusers (data not shown). Users of chiropractors also reported significantly higher levels of pain (P=.015; data not shown) than nonusers.
Musculoskeletal problems, usually back pain, were cited as the most common health problem associated with CAM use, followed by endocrine and metabolic diseases (primarily diabetes or obesity), diseases of the respiratory system (primarily asthma), and diseases of the genitourinary system Table 4. CAM users were approximately twice as likely as nonusers to have a musculoskeletal system disorder and 2.5 times as likely to have a genitourinary system disorder. Users of practitioner-based therapies were 2.7 times as likely to have a musculoskeletal system disorder as nonusers. Users of chiropractors were 3.7 times (P <.001; data not shown) and users of massage therapy were 2.2 times (P <.05; data not shown) as likely to have a musculoskeletal disorder as nonusers.
Users of CAM therapies in general (P <.01) and practitioner-based therapies (P <.01) and chiropractors (P <.001; data not shown) in particular reported significantly less satisfaction than nonusers with the conventional treatment they received from their family physician Figure 3. However, CAM users also reported no significant difference in the level of satisfaction with these therapies and the level of satisfaction with the care received from their family physician. Users of folk remedies reported significantly higher levels of satisfaction with conventional treatment than nonusers (P <.05), and higher levels of satisfaction with their CAM therapy than other CAM users (P <.05).
We examined the independent association between each of the sociodemographic and health status characteristics of patients and CAM use with a series of logistic regression models. Age, sex, marital status, place of birth, education, household income, method of payment, level of acculturation, satisfaction with primary care received, SF-36 measures of functional status, and presence or absence of a musculoskeletal disorder were entered into the model in a stepwise fashion. The best fitting models are presented in Table 5. Physical role functioning was the only significant independent predictor of use of one or more CAM therapies. The model correctly classified 70% of the patients. Musculoskeletal disorders, emotional functioning, and perceived general health were significant independent predictors of use of a practitioner-based therapy. The model correctly classified 86.4% of the patients. Patients with musculoskeletal disorders were 7.37 times (95% confidence interval [CI], 2.37-23.51) as likely to use some form of practitioner-based therapy (usually a chiropractor or massage therapist) as patients with other physical conditions. Level of education and general perceived health compared with a year ago were significant independent predictors of use of self-care–based therapies. The model correctly classified 86.4% of the patients. College graduates were 1.44 times (95% CI, 1.04-2.01) as likely to use some form of self-care–based therapy as patients with 12 or fewer years of education. Hispanic ethnicity was the only significant independent predictor of use of traditional folk remedies. The model correctly classified 97.8% of the patients. Hispanics were 10.27 times (95% CI, 3.10-33.95) as likely to use traditional folk remedies as members of other ethnic groups.
Logistic regression analyses predicting use of specific CAM therapies revealed that general perceived health was a significant independent predictor of use of acupuncture (b=-0.0665; standard error [SE]=0.0253; P=.0086) and chiropractors (b=-0.0314; SE=0.0139; P=.0243). A musculoskeletal disorder was a significant independent predictor of use of chiropractors (b=2.0123; SE=0.6720; P=.0005) and massage therapists (b=1.8467; SE=0.7792; P=.0178). Patients with musculoskeletal disorders were 10.23 times (95% CI, 2.00-27.92) as likely to use a chiropractor and 6.34 times (95% CI, 1.38-29.19) as likely to use a massage therapist as patients with some other clinical condition. Emotional functioning was also a significant independent predictor of use of massage therapy (b=-0.0902; SE=0.0347; P=.0094), and household income was a significant independent predictor of use of acupuncture (b=2.5045; SE=1.2428; P=.0439).
DISCUSSION
In our study approximately 1 in 5 patients reported having used some form of complementary or alternative therapy or therapist in the past year in conjunction with a current health problem. This percentage is smaller than the 42% reported by Eisenberg and colleagues.2 However, that study was a population-based survey and was not tied to use of primary care services. Although previous studies of primary care patients have reported higher percentages of patients using CAM therapies, differences in methods preclude a comparison with those percentages. Drivdahl and Miser,9 for example, reported that 28% of 177 family practice patients in a military clinic had ever used some form of alternative medicine. Elder and colleagues10 reported that 50% of 113 family practice clinic patients had used some form of CAM, but this includes use for reasons other than those precipitating the visit to the physician. To our knowledge this is the largest study conducted of a broad spectrum of family practice patients, focusing on use related to a specific health problem that precipitated a visit to a family physician.
Consistent with the findings of previous studies,1,2,9 use of CAM therapies in general and self-care–based therapies in particular by this group of primary care patients, was significantly associated with a high level of education. This may be attributed to the fact that better educated patients tend to be more informed or at least more likely to be exposed to information about the benefits of CAM.17,18 Consistent with previous research,1,2,17,18 use of self-care–based therapies was also associated with high household income in our study. Many of these therapies are not always covered by health insurance plans, which means they require cash payments that those with higher incomes are more likely to be able to afford.
In our study use of traditional folk remedies was inversely associated with levels of education and acculturation and positively associated with Hispanic ethnicity and being foreign-born. Low-acculturated, predominately Hispanic immigrants are more likely to be uninsured or underinsured,19 not able to afford cannot afford conventional treatment,20 and be more familiar with the efficacy of those traditional folk remedies.21
We also found that users of CAM therapies generally perceive their health to be worse compared with others their age and with their health a year ago than nonusers. Emotional functioning was also a significant independent predictor of use of practitioner-based therapies. Given the cross-sectional nature of the study design, it is difficult to determine whether health status is a cause or consequence of CAM use. However, a further examination of the experience of these patients with conventional treatment and their specific clinical condition provides some insight into this relationship.
Failure of regular treatment to cure the problem and a desire to avoid side effects were each cited by 1 in 4 users of CAM therapies in our study. Users were also significantly less satisfied than nonusers with the care they had received from their primary care provider. They were significantly more likely to report visiting a primary care provider for a musculoskeletal or genitourinary disorder than nonusers. Both of these disorders are noteworthy for the limited treatment effectiveness or inconsistency in treatment approach of conventional therapies.22,23 Taken together, these observations suggest that many users of CAM do so because they are dissatisfied with the care received from their primary care provider.
However, complementary and alternative medicine users were found to be no more satisfied with these alternatives than with the care received from their family physician. Furthermore, unhappiness with the attitude of their family physician and failure to correctly diagnose the problem was cited by a relatively small number of patients as a reason for using CAM therapies. Only 20% of users reported doing so after seeing a physician; the others did so either before or concurrent with a visit. This suggests that experience with the family physician played a relatively minor role in the decision of most patients to use CAM therapies.
Conclusions
The results of our study can be used to acquaint family physicians with the characteristics of users of complementary and alternative therapies and therapists in conjunction with specific health problems. Such information may have implications for diagnosis and treatment regimens, especially if contraindications to certain forms of treatment arise as a result of potential adverse reactions when used in combination with specific forms of CAM (eg, lead poisoning resulting from the use of certain traditional Mexican remedies used as laxatives24). In contrast to the results of previous studies,25 relatively few patients were reluctant to notify their primary care provider of their use of CAM therapies. However, patients were more likely to share this information if the physician made an effort to ask. An understanding of the characteristics associated with CAM use should enable the physician to obtain this information.
The results of our study suggest that patterns and predictors of CAM use by primary care patients vary with the type of therapy used. It is important to realize that all CAM therapies are not alike. Self-care–based therapies are more likely to be used by patients who are well educated, while traditional folk remedies are more likely to be used by Hispanic immigrants with low levels of education. Practitioner-based therapies are more likely to be used by patients with musculoskeletal disorders. Family physicians should keep these distinctions in mind when evaluating the likelihood of CAM use by a particular patient and potential implications for primary care delivery.
Acknowledgments
Our study was funded by a grant from the California Academy of Family Physicians.
Related Resources
National Center for Complementary and Alternative Medicine
A branch of the National Institutes of Health that conducts and supports basic and applied research and training and disseminates information on complementary and alternative medicine to physicians and the public. www.nccam.nih.gov
Whole Health MD
Information on combining alternative medicine, supplements, vitamins, herbs, and nutrition with conventional medicine. www.wholehealth.com
alternativeDr.com
Resources for books, reports, and practitioner searches serving various categories in alternative medicine therapies. www.alternativedr.com
HealthWorld Online
Alternative and Complementary Medicine Center-provides in-depth information about a range of therapies, as well as discussion forums, conference listings, and other resources. www.healthy.net/clinic/therapy/altmedcolumn
METHODS: We conducted a survey of 542 patients attending 16 family practice clinics that belong to a community-based research network in San Diego, California, to determine patients’ reasons for using CAM therapies in conjunction with a visit to a family physician and the impact of these therapies on their health and well-being.
RESULTS: Approximately 21% of the patients reported using one or more forms of CAM therapy in conjunction with the most important health problem underlying their visit to the physician. The most common forms of therapy were visiting chiropractors (34.5% of CAM users), herbal remedies and supplements (26.7%), and massage therapy (17.2%). Recommendations from friends or coworkers, a desire to avoid the side effects of conventional treatments, or failure of conventional treatments to cure a problem were the most frequently cited reasons for using these therapies. Use of practitioner-based therapies was significantly and independently associated with poor perceived health status, poor emotional functioning, and a musculoskeletal disorder, usually low back pain. Use of self-care–based therapies was associated with high education and poor perceived general health compared with a year ago. Use of traditional folk remedies was associated with Hispanic ethnicity.
CONCLUSIONS: Sociodemographic characteristics and clinical conditions that predict use of CAM therapies by primary care patients in conjunction with a current health problem vary with the type of therapy used.
Within the past 5 years several studies have pointed to the widespread use of complementary and alternative medicine (CAM) in the United States. In a study conducted in 1991,1 1 in 3 respondents in a national sample of adults reported using at least one unconventional therapy in the past year. By 1997 that number had risen to more than 4 in 10.2 Similar studies conducted in Europe3 and Canada4 have reported utilization rates between 18% and 75%. Although several studies have found substantial use among patients attending specialty clinics,5-7 between 28% and 50% of family practice patients have been found to have used some form of CAM.8-10 In response to its widespread use, CAM has gained increasing acceptance among family physicians and other primary care physicians and in schools of medicine where more courses are being taught on the subject.11
Despite this increased use and acceptance by patients and health care providers, there is relatively little information available concerning the reasons for use of the various forms of therapies and treatments considered alternative or complementary. Recent surveys of the extent of use of these treatments provide little insight into why certain patients are more likely to use CAM therapies in general or specific therapies, such as chiropractic, message, herbal therapy, acupuncture, and homeopathy. Small studies of specific groups of patients suggest that use of these therapies is associated with the disease and with patient characteristics such as education and level of dissatisfaction with the primary care provider,9-10 but the extent to which these findings are generalizable to all primary care patients remains unclear.
A second limitation to our understanding of the use of CAM therapies by primary care patients is that the effectiveness of these therapies has not been subjected to rigorous examination. Their use by the general public appears to be based on anecdotal evidence, primarily personal experience or the experiences of others. Although there is a consensus within the medical establishment that most of these therapies are harmless,12 there is increasing evidence of the adverse consequences related to their use and misuse.13 However, much of this evidence is also anecdotal, based on case reports and not on large population-based studies.
Our objective was to address these 2 deficiencies in the understanding of the likelihood of use and the effectiveness of CAM therapies by conducting a large survey of a diverse population of patients attending family practice clinics in several different settings throughout the San Diego area, making use of a recently formed network of family physicians committed to community-based primary care research. Our goal for this survey was to examine the characteristics of primary care patients who use CAM therapies, determine whether these characteristics are significantly different for patients who have not used these therapies, and determine whether use of CAM therapies is associated with clinical condition, functional status, and quality of life.
Methods
Subjects
Our subjects included 541 patients aged 18 years and older visiting 16 family practice clinics in the San Diego area during a 3-month period (June 1999-August 1999). Each of these practices was a member the recently formed San Diego Unified Research in Family Medicine Network (SURF*NET). In this practice-based research initiative, community physicians, faculty and educators of academic family medicine programs combine research and clinical practice to develop a vital body of knowledge in the discipline of family medicine. The 16 clinics participating in the study represented more than 40 family physicians who were SURF*NET members and a patient population of more than 30,000, covering a broad and representative cross-section of the San Diego community.
To participate in the study patients had to identify a specific health complaint as the reason for a clinic visit. Individuals who made a visit for a general physical examination or to transport a pediatric patient were excluded. Our study participants represented 89% of all patients who met eligibility criteria and were invited to participate. Participants and nonparticipants exhibited no significant differences with respect to age, sex, ethnicity, or insurance status.
Data
Patients were asked to complete a questionnaire administered by a survey worker in the waiting room before the scheduled visit with a family physician. The survey instrument included questions about the social and demographic characteristics of the patient, including age, sex, marital status, ethnicity, place of birth, level of acculturation, education, and 1998 household income. Level of acculturation was assessed on the basis of a 5-item scale used in previous studies of patient populations.14 Patients were then grouped into acculturation categories: low, medium, and high. They were also grouped into categories based on their level of education (no college, some college, and college graduate), the method of payment for the clinic visit (cash, Medi-Cal/Medicare, and health maintenance organization or health insurance), and median 1998 household income (<$50,000 or Ž$50,000).
The Medical Outcomes Study Short Form (SF-36)15 was used to assess a patient’s current health status and quality of well-being. The patients were evaluated on the basis of physical and social functioning, physical and emotional role functioning, mental health, energy or fatigue, pain, general perceived health compared with others the same age, and general health compared with a year ago.
Finally, each patient was asked to describe the most important or significant health problems experienced during the past year and whether any of these problems had precipitated the current clinic visit. Health problems were then coded by investigators according to International Classification of Diseases—9th revision—Clinical Modification (ICD-9-CM) criteria.16 Symptoms and complaints that could not be attributed to a specific diagnosis were placed under the ICD-9-CM category of Symptoms and Ill-Defined Conditions. Using a list derived from previous studies,1,2 patients were then asked whether they had used one or more of 16 CAM therapies or therapists for their principal medical condition during the past 12 months. Information was also collected on the level of satisfaction with these CAM therapies, level of satisfaction with care provided by their family physician for the problem, reasons for using a CAM therapy or therapist, and whether the family physician had been notified by the patient that he or she was using such alternatives. Patients’ level of satisfaction with conventional and CAM treatments was rated on a scale from 1 (not at all satisfied) to 10 (completely satisfied). Reasons for using a CAM therapy were derived from a list compiled by Lazar and O’Connor.17
Statistics
Univariate statistics (percentages and means) were used to describe the characteristics of CAM use. Bivariate analyses (chi-square tests for categorical variables and paired-sample t tests and analysis of variance for continuous measures) were used to determine whether there were any significant differences between patients reporting use of any form of CAM therapy in the past year and those reporting no use of such therapy with respect to the following predictors: (1) social and demographic characteristics; (2) functional status and quality of wellbeing; (3) dissatisfaction with conventional treatments; and (4) ICD-9-CM diagnostic category of chief health complaint. Similar analyses were performed by 3 classes of therapy: (1) practitioner-based therapies (acupuncture, biofeedback, chiropractic, homeopathy, massage therapy, naturopathy); (2) self-care based therapies (energy healing, meditation and prayer, dietary interventions, herbal remedies, multivitamin supplements); and (3) traditional folk remedies. When appropriate (ie, based on the number of users), analyses were also conducted for individual types of therapy (chiropractic, acupuncture, herbal remedies, dietary interventions, massage therapy). Logistic regression models with stepwise entry of all potential independent variables were used to assess the odds of using a CAM therapy associated with each patient characteristic.
Results
Characteristics of CAM Users
Of the 541 adults participating in our study, 116 (21%) reported using 1 or more forms of CAM therapy or therapists within the past year for the primary health problem contributing to the present clinic visit. A visit to a chiropractor was the most frequent form of alternative therapy, followed by the use of herbal remedies or supplements, and message therapy Figure 1.
Approximately 60% of these patients had informed their physicians of the use of these CAM therapies. Of those who had not done so, 60% had indicated that there had been no previous opportunity to inform their physicians, since this had been the first visit for the health problem in question. When examined by class of CAM usage, approximately two thirds of those using practitioner-based and self-care– based therapies reported use to their physician, compared with 40% of those using traditional folk remedies.
The timing of initial use of CAM therapies in treating the current health problem is shown in Figure 2. In general, one third of all users of CAM therapies initiated treatment with one or more therapies before their initial visit to a primary care physician for the same clinical problem. Thirty-seven percent initiated use of CAM concurrent with (ie, within 2 weeks) of their initial visit to a primary care provider, and 1 in 5 (19%) initiated use of a CAM therapy after their initial primary care visit. One third (36%) of those using practitioner-based and self-care–based therapies and 46.2% of those using traditional folk remedies reported initiating therapy before a visit with a primary care provider. One third (36%) of the users of self-care–based therapies, 43% of users of practitioner-based therapies, and none of the users of traditional folk remedies reported using such a therapy concurrent with their initial clinic visit. One in 4 of the practitioner-based therapies (24.6%) and traditional folk remedies (23.1%) and 14% of users of self-care–based therapies reported initiating use after their initial visit to a primary care provider.
Approximately 1 in 4 patients reported using CAM to avoid side effects of regular treatment because a friend or coworker had recommended the treatment or because conventional treatment had failed to cure the problem Table 1. Between 10% and 15% of the patients reported using these therapies for philosophical reasons, because they preferred to deal with the problem by themselves, or because older family members had used these treatments for the same problem. Only 7 patients reported using therapies because they were unhappy with the attitude of family physicians. When examined by class of therapy, approximately 1 in 3 users of practitioner-based therapies reported using them to avoid the side effects of regular treatment, failure of regular treatment to cure their problem, and a recommendation from a friend or coworker. In addition to a preference for dealing with the problem by themselves, these 3 reasons (side effects, failure of regular treatment, and a recommendation from a friend) were also the primary reasons for use of self-care– based therapies. In contrast, use by parents and relatives for the same problem represented the primary reason for traditional folk remedies, accounting for slightly less than one third (30.8%) of the patients using them.
Predictors of CAM Use
A comparison of the social and demographic characteristics of users and nonusers of CAM is provided in Table 2. Use of CAM therapies was positively associated with level of education but inversely associated with level of acculturation. When examined by specific categories of CAM, women were significantly more likely than men to use herbal remedies (P <.05; data not shown) and other self-care– based forms of alternative medicine and traditional folk medicines. Level of education was positively associated with self-care–based forms of CAM in general and use of herbal (P <.001; data not shown) and dietary (P <.05; data not shown) remedies in particular. However, education was inversely associated with use of traditional folk remedies. Self-care-based therapies in general and herbal remedies in particular (P <.05; data not shown) were significantly associated with the level of household income. Use of traditional folk remedies was significantly associated with Hispanic ethnicity, place of birth, and low acculturation. Dietary remedies were positively associated with level of acculturation (P <.05; data not shown). Patients belonging to an health maintenance organization or possessing other forms of non-government-sponsored insurance were significantly more likely to use massage therapy (P <.05; data not shown) or herbal remedies (P <.05; data not shown).
The health status and quality of well-being of users and nonusers of CAM therapies and therapists is provided in Table 3. Users of CAM therapies reported significantly lower emotional role functioning and perceived general health compared with nonusers of the same age. Users of practitioner-based therapies reported significantly lower social functioning, physical and emotional role functioning, mental health, and perceived general health, and more pain than nonusers. Users of self-care–based therapies and traditional folk remedies reported significantly lower levels of general health than a year ago. Users of acupuncture (P=.03; data not shown) and chiropractors (P=.001; data not shown) reported significantly lower levels of general perceived health than nonusers (data not shown). Users of chiropractors also reported significantly higher levels of pain (P=.015; data not shown) than nonusers.
Musculoskeletal problems, usually back pain, were cited as the most common health problem associated with CAM use, followed by endocrine and metabolic diseases (primarily diabetes or obesity), diseases of the respiratory system (primarily asthma), and diseases of the genitourinary system Table 4. CAM users were approximately twice as likely as nonusers to have a musculoskeletal system disorder and 2.5 times as likely to have a genitourinary system disorder. Users of practitioner-based therapies were 2.7 times as likely to have a musculoskeletal system disorder as nonusers. Users of chiropractors were 3.7 times (P <.001; data not shown) and users of massage therapy were 2.2 times (P <.05; data not shown) as likely to have a musculoskeletal disorder as nonusers.
Users of CAM therapies in general (P <.01) and practitioner-based therapies (P <.01) and chiropractors (P <.001; data not shown) in particular reported significantly less satisfaction than nonusers with the conventional treatment they received from their family physician Figure 3. However, CAM users also reported no significant difference in the level of satisfaction with these therapies and the level of satisfaction with the care received from their family physician. Users of folk remedies reported significantly higher levels of satisfaction with conventional treatment than nonusers (P <.05), and higher levels of satisfaction with their CAM therapy than other CAM users (P <.05).
We examined the independent association between each of the sociodemographic and health status characteristics of patients and CAM use with a series of logistic regression models. Age, sex, marital status, place of birth, education, household income, method of payment, level of acculturation, satisfaction with primary care received, SF-36 measures of functional status, and presence or absence of a musculoskeletal disorder were entered into the model in a stepwise fashion. The best fitting models are presented in Table 5. Physical role functioning was the only significant independent predictor of use of one or more CAM therapies. The model correctly classified 70% of the patients. Musculoskeletal disorders, emotional functioning, and perceived general health were significant independent predictors of use of a practitioner-based therapy. The model correctly classified 86.4% of the patients. Patients with musculoskeletal disorders were 7.37 times (95% confidence interval [CI], 2.37-23.51) as likely to use some form of practitioner-based therapy (usually a chiropractor or massage therapist) as patients with other physical conditions. Level of education and general perceived health compared with a year ago were significant independent predictors of use of self-care–based therapies. The model correctly classified 86.4% of the patients. College graduates were 1.44 times (95% CI, 1.04-2.01) as likely to use some form of self-care–based therapy as patients with 12 or fewer years of education. Hispanic ethnicity was the only significant independent predictor of use of traditional folk remedies. The model correctly classified 97.8% of the patients. Hispanics were 10.27 times (95% CI, 3.10-33.95) as likely to use traditional folk remedies as members of other ethnic groups.
Logistic regression analyses predicting use of specific CAM therapies revealed that general perceived health was a significant independent predictor of use of acupuncture (b=-0.0665; standard error [SE]=0.0253; P=.0086) and chiropractors (b=-0.0314; SE=0.0139; P=.0243). A musculoskeletal disorder was a significant independent predictor of use of chiropractors (b=2.0123; SE=0.6720; P=.0005) and massage therapists (b=1.8467; SE=0.7792; P=.0178). Patients with musculoskeletal disorders were 10.23 times (95% CI, 2.00-27.92) as likely to use a chiropractor and 6.34 times (95% CI, 1.38-29.19) as likely to use a massage therapist as patients with some other clinical condition. Emotional functioning was also a significant independent predictor of use of massage therapy (b=-0.0902; SE=0.0347; P=.0094), and household income was a significant independent predictor of use of acupuncture (b=2.5045; SE=1.2428; P=.0439).
DISCUSSION
In our study approximately 1 in 5 patients reported having used some form of complementary or alternative therapy or therapist in the past year in conjunction with a current health problem. This percentage is smaller than the 42% reported by Eisenberg and colleagues.2 However, that study was a population-based survey and was not tied to use of primary care services. Although previous studies of primary care patients have reported higher percentages of patients using CAM therapies, differences in methods preclude a comparison with those percentages. Drivdahl and Miser,9 for example, reported that 28% of 177 family practice patients in a military clinic had ever used some form of alternative medicine. Elder and colleagues10 reported that 50% of 113 family practice clinic patients had used some form of CAM, but this includes use for reasons other than those precipitating the visit to the physician. To our knowledge this is the largest study conducted of a broad spectrum of family practice patients, focusing on use related to a specific health problem that precipitated a visit to a family physician.
Consistent with the findings of previous studies,1,2,9 use of CAM therapies in general and self-care–based therapies in particular by this group of primary care patients, was significantly associated with a high level of education. This may be attributed to the fact that better educated patients tend to be more informed or at least more likely to be exposed to information about the benefits of CAM.17,18 Consistent with previous research,1,2,17,18 use of self-care–based therapies was also associated with high household income in our study. Many of these therapies are not always covered by health insurance plans, which means they require cash payments that those with higher incomes are more likely to be able to afford.
In our study use of traditional folk remedies was inversely associated with levels of education and acculturation and positively associated with Hispanic ethnicity and being foreign-born. Low-acculturated, predominately Hispanic immigrants are more likely to be uninsured or underinsured,19 not able to afford cannot afford conventional treatment,20 and be more familiar with the efficacy of those traditional folk remedies.21
We also found that users of CAM therapies generally perceive their health to be worse compared with others their age and with their health a year ago than nonusers. Emotional functioning was also a significant independent predictor of use of practitioner-based therapies. Given the cross-sectional nature of the study design, it is difficult to determine whether health status is a cause or consequence of CAM use. However, a further examination of the experience of these patients with conventional treatment and their specific clinical condition provides some insight into this relationship.
Failure of regular treatment to cure the problem and a desire to avoid side effects were each cited by 1 in 4 users of CAM therapies in our study. Users were also significantly less satisfied than nonusers with the care they had received from their primary care provider. They were significantly more likely to report visiting a primary care provider for a musculoskeletal or genitourinary disorder than nonusers. Both of these disorders are noteworthy for the limited treatment effectiveness or inconsistency in treatment approach of conventional therapies.22,23 Taken together, these observations suggest that many users of CAM do so because they are dissatisfied with the care received from their primary care provider.
However, complementary and alternative medicine users were found to be no more satisfied with these alternatives than with the care received from their family physician. Furthermore, unhappiness with the attitude of their family physician and failure to correctly diagnose the problem was cited by a relatively small number of patients as a reason for using CAM therapies. Only 20% of users reported doing so after seeing a physician; the others did so either before or concurrent with a visit. This suggests that experience with the family physician played a relatively minor role in the decision of most patients to use CAM therapies.
Conclusions
The results of our study can be used to acquaint family physicians with the characteristics of users of complementary and alternative therapies and therapists in conjunction with specific health problems. Such information may have implications for diagnosis and treatment regimens, especially if contraindications to certain forms of treatment arise as a result of potential adverse reactions when used in combination with specific forms of CAM (eg, lead poisoning resulting from the use of certain traditional Mexican remedies used as laxatives24). In contrast to the results of previous studies,25 relatively few patients were reluctant to notify their primary care provider of their use of CAM therapies. However, patients were more likely to share this information if the physician made an effort to ask. An understanding of the characteristics associated with CAM use should enable the physician to obtain this information.
The results of our study suggest that patterns and predictors of CAM use by primary care patients vary with the type of therapy used. It is important to realize that all CAM therapies are not alike. Self-care–based therapies are more likely to be used by patients who are well educated, while traditional folk remedies are more likely to be used by Hispanic immigrants with low levels of education. Practitioner-based therapies are more likely to be used by patients with musculoskeletal disorders. Family physicians should keep these distinctions in mind when evaluating the likelihood of CAM use by a particular patient and potential implications for primary care delivery.
Acknowledgments
Our study was funded by a grant from the California Academy of Family Physicians.
Related Resources
National Center for Complementary and Alternative Medicine
A branch of the National Institutes of Health that conducts and supports basic and applied research and training and disseminates information on complementary and alternative medicine to physicians and the public. www.nccam.nih.gov
Whole Health MD
Information on combining alternative medicine, supplements, vitamins, herbs, and nutrition with conventional medicine. www.wholehealth.com
alternativeDr.com
Resources for books, reports, and practitioner searches serving various categories in alternative medicine therapies. www.alternativedr.com
HealthWorld Online
Alternative and Complementary Medicine Center-provides in-depth information about a range of therapies, as well as discussion forums, conference listings, and other resources. www.healthy.net/clinic/therapy/altmedcolumn
1. DM, Kessler RC, Foster C, Norlock FE, Calkins DR, Delbanco TL. Unconventional medicine in the United States: prevalence, costs, and patterns of use. N Engl J Med 1993;328:246-52.
2. DM, Davis RB, Ettner SL, et al. Trends in alternative medicine use in the United States, 1990-1997: results of a follow-up national survey. JAMA 1998;280:1569-75.
3. G. Alternative medicine in Europe: a quantitative comparison of the use and knowledge of alternative medicine and patient profiles in nine European countries. Brussels, Belgium: Belgium Consumers’ Association; 1987.
4. in five Canadians is using alternative therapies. Can Med Assoc J 1991;144:469.-
5. M, Fitzcharles MA. Alternative medicine use by rheumatology patients in a universal health care setting. J Rheumatol 1994;21:148-52.
6. W, O’Connor BB, MacGregor RR, Schwartz JS. Patient use and assessment of conventional and alternative therapies for HIV infection and AIDS. AIDS 1993;7:561-65.
7. MJ, Sutherland LR, Brkich L. Use of alternative medicine by patients attending a gastroenterology clinic. Can Med Assoc J 1990;142:121-25.
8. J, Shepherd S. Alternative or additional medicine? An exploratory study in general practice. Soc Sci Med 1993;37:983-88.
9. NC, Gillcrist A, Minz R. Use of alternative health care by family practice patients. Arch Fam Med 1997;6:181-84.
10. CE, Miser WF. The use of alternative health care by a family practice population. J Am Board Fam Pract 1998;11:193-99.
11. BM, Singh BB, Hartnoll SM, Singh BK, Reilly D. Primary care physicians and complementary-alternative medicine: training, attitudes, and practice patterns. J Am Board Fam Pract 1998;11:272-81.
12. M, Kassirer JP. Alternative medicine: the risk of untested and unregulated remedies. N Engl J Med 1998;339:839-41.
13. MJ, Anderson RA, Egeler RM, Wolff JEA. Alternative therapies for the treatment of childhood cancer. N Engl J Med 1998;339:846-47.
14. G, Sabogal F, Marin B, Otero-Sabogal R, Perez-Stable EJ. Development of a short acculturation scale for Hispanics. Hispanic J Behav Sci 1987;9:183-205.
15. JE,, Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36): I: conceptual framework and item selection. Med Care 1992;30:473-83.
16. Management Information Corporation. International classification of diseases-9th revision-clinical modification. Los Angeles, Calif: Practice Management Information Corporation; 1991.
17. JS, O’Connor BB. Talking with patients about their use of alternative therapies. Prim Care 1997;24:699-711.
18. JA. Why patients use alternative medicine: results of a national study. JAMA 1998;279:1548-53.
19. AM, ed. Acculturation: theory, models and some new findings. Boulder, Colo: Westview Press; 1980.
20. M, Richard C, Casebeer AW, Ray NF. Health insurance coverage among foreign-born US residents: the impact of race, ethnicity, and length of residence. Am J Public Health 1997;87:96-102.
21. AL, Mazur LJ. Use of folk remedies in a hispanic population. Arch Pediatr Adolesc Med 1995;149:978-81.
22. DC, MacCormack FA, Berg AO. Managing low back pain: a comparison of the beliefs and behaviors of family physicians and chiropractors. West J Med 1988;149:475-80.
23. AO. Variations among family physicians’ management strategies for lower urinary tract infection in women: a report from the Washington family physicians collaborative research network. J Am Board Fam Pract 1991;4:327-30.
24. for Disease Control and Prevention. Lead poisoning from Mexican folk remedies—California. MMWR Morb Mortal Wkly Rep 1993;42:521-24.
25. SR, Fosket JR. Disclosing complementary and alternative medicine use in the medical encounter: a qualitative study in women with breast cancer. J Fam Pract 1999;48:453-58.
1. DM, Kessler RC, Foster C, Norlock FE, Calkins DR, Delbanco TL. Unconventional medicine in the United States: prevalence, costs, and patterns of use. N Engl J Med 1993;328:246-52.
2. DM, Davis RB, Ettner SL, et al. Trends in alternative medicine use in the United States, 1990-1997: results of a follow-up national survey. JAMA 1998;280:1569-75.
3. G. Alternative medicine in Europe: a quantitative comparison of the use and knowledge of alternative medicine and patient profiles in nine European countries. Brussels, Belgium: Belgium Consumers’ Association; 1987.
4. in five Canadians is using alternative therapies. Can Med Assoc J 1991;144:469.-
5. M, Fitzcharles MA. Alternative medicine use by rheumatology patients in a universal health care setting. J Rheumatol 1994;21:148-52.
6. W, O’Connor BB, MacGregor RR, Schwartz JS. Patient use and assessment of conventional and alternative therapies for HIV infection and AIDS. AIDS 1993;7:561-65.
7. MJ, Sutherland LR, Brkich L. Use of alternative medicine by patients attending a gastroenterology clinic. Can Med Assoc J 1990;142:121-25.
8. J, Shepherd S. Alternative or additional medicine? An exploratory study in general practice. Soc Sci Med 1993;37:983-88.
9. NC, Gillcrist A, Minz R. Use of alternative health care by family practice patients. Arch Fam Med 1997;6:181-84.
10. CE, Miser WF. The use of alternative health care by a family practice population. J Am Board Fam Pract 1998;11:193-99.
11. BM, Singh BB, Hartnoll SM, Singh BK, Reilly D. Primary care physicians and complementary-alternative medicine: training, attitudes, and practice patterns. J Am Board Fam Pract 1998;11:272-81.
12. M, Kassirer JP. Alternative medicine: the risk of untested and unregulated remedies. N Engl J Med 1998;339:839-41.
13. MJ, Anderson RA, Egeler RM, Wolff JEA. Alternative therapies for the treatment of childhood cancer. N Engl J Med 1998;339:846-47.
14. G, Sabogal F, Marin B, Otero-Sabogal R, Perez-Stable EJ. Development of a short acculturation scale for Hispanics. Hispanic J Behav Sci 1987;9:183-205.
15. JE,, Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36): I: conceptual framework and item selection. Med Care 1992;30:473-83.
16. Management Information Corporation. International classification of diseases-9th revision-clinical modification. Los Angeles, Calif: Practice Management Information Corporation; 1991.
17. JS, O’Connor BB. Talking with patients about their use of alternative therapies. Prim Care 1997;24:699-711.
18. JA. Why patients use alternative medicine: results of a national study. JAMA 1998;279:1548-53.
19. AM, ed. Acculturation: theory, models and some new findings. Boulder, Colo: Westview Press; 1980.
20. M, Richard C, Casebeer AW, Ray NF. Health insurance coverage among foreign-born US residents: the impact of race, ethnicity, and length of residence. Am J Public Health 1997;87:96-102.
21. AL, Mazur LJ. Use of folk remedies in a hispanic population. Arch Pediatr Adolesc Med 1995;149:978-81.
22. DC, MacCormack FA, Berg AO. Managing low back pain: a comparison of the beliefs and behaviors of family physicians and chiropractors. West J Med 1988;149:475-80.
23. AO. Variations among family physicians’ management strategies for lower urinary tract infection in women: a report from the Washington family physicians collaborative research network. J Am Board Fam Pract 1991;4:327-30.
24. for Disease Control and Prevention. Lead poisoning from Mexican folk remedies—California. MMWR Morb Mortal Wkly Rep 1993;42:521-24.
25. SR, Fosket JR. Disclosing complementary and alternative medicine use in the medical encounter: a qualitative study in women with breast cancer. J Fam Pract 1999;48:453-58.
Factors Associated with Repeat Mammography Screening
METHODS: Our data are from a baseline telephone interview conducted among 1287 female members of Blue Cross Blue Shield of North Carolina who were aged either 40 to 44 years or 50 to 54 years.
RESULTS: The 3 groups differed significantly on beliefs and health-related behaviors, with the off-schedule group almost consistently falling between the on-schedule and never screened groups. Off-schedule women were more likely than on-schedule women, but less likely than those never screened, to not have a clinical breast examination within 12 months, to be ambivalent about screening mammography, to be confused about screening guidelines, and to not be advised by a physician to get a mammogram in the past 2 years. Off-schedule women perceived their breast cancer risk as lower and were less likely to be up to date with other cancer screening tests.
CONCLUSIONS: Our findings suggest that women who are off schedule are in need of mammography-promoting interventions, including recommendations from and discussion with their health care providers. Because they are more positive and knowledgeable about mammography than women who have never been screened, they may benefit from brief interventions from health care providers that highlight the importance of repeat screening.
Because most research has shown that routine mammography screening saves lives,1,2 many medical organizations have developed guidelines and recommendations for mammography screening. For example, the National Cancer Institute (NCI) recommends mammograms every 1 to 2 years for women aged 40 years and older and annual screening for women aged 50 years and older. Other organizations, such as the United States Preventive Health Task Force and the American College of Preventive Medicine, differ from the NCI regarding optimal screening intervals, but all agree on the importance of regular screening to achieve a breast cancer mortality reduction benefit.
However, despite these recommendations and substantial increases in the percentage of women who have ever had mammograms, most women are not having regular screening at recommended intervals. There is also still a group of women who have never been screened.3-6 For the purpose of developing interventions to encourage routine screening, it may be important to understand differences among women who are and are not getting repeat screening and those who have never been screened. However, with a few exceptions,7-12 most mammography studies have focused on 1-time screening rather than repeat adherence. Few studies have identified factors that predict repeat mammography use according to recommended guidelines.
For our analysis we sought to better understand factors that differentiate 3 groups of insured women: those who have been screened and are adherent to the screening mammography guidelines of the NCI, those screened at least once but who are currently nonadherent, and those never screened. Study findings should guide the design of interventions to promote continued adherence for repeat screening mammography.
Methods
Study Population
We randomly selected 2165 women from a sampling frame of 4000 women aged 40 to 44 years and 50 to 54 years who were members of the Personal Care Plan of Blue Cross Blue Shield of North Carolina (BCBS of NC) in 1997. We stratified the sample by age and mammography compliance status. Women with previous breast cancer and those who were no longer BCBS members were excluded from the sample. The completion rate for the telephone interviews was 76%, and the nonresponse rate was 20%, leaving a sample of 1287 women.
We mailed introductory letters, and professional interviewers conducted telephone interviews between November 1997 and May 1998. The participants provided oral consent in accordance with regulations from the Department of Health and Human Services. The analyses reported in this paper were conducted on baseline data collected for a larger intervention trial designed to enhance informed decision making about mammography. Additional details regarding our study methodology have been published elsewhere.13
Measures
Screening History Measure. The main variable of interest was self-reported mammography history (including most recent and previous mammograms). We calculated 2 screening variables reflecting whether: (1) the most recent mammogram was within the recommended time frame according to NCI recommendations, and (2) the interval between the most recent and the previous mammography date was within the recommended time frame for the woman’s age. The second interval could be computed only for women with more than 1 previous mammogram.
We categorized the participants as “never had a mammogram,” “off schedule,” or “on schedule for their 2 most recent mammograms.” (We refer to the groups as never had, off schedule, and on schedule.)
We followed the recommendations of BCBS of NC which specified mammography consistent with the NCI recommendation: every 1 to 2 years for women in their 40s and every year for women in their 50s.14 However, because many women are not screened exactly 12 months following their previous mammogram, we added a 3-month window to the intervals; thus, the window was 15 months for women in their 50s and 27 months for women in their 40s. These interval windows are consistent with those adopted by other investigators.10
Our on- and off-schedule classification algorithms allow for women in their early 50s who passed from the “every 1 to 2 years” to the “every year” guidelines between their most recent and previous mammograms and those in their 40s who had not yet had time for 2 mammograms based on their age and the recommendations. The classification of women by screening mammography history is presented in Table 1
Because NCI recommendations indicate a single mammogram for average-risk women younger than 42 years, we could not consider women younger than this to be off schedule, so we excluded them (n=198) from our analysis. Women who had 2 mammograms within 11 months (n=32) were also excluded, because it is likely they were on diagnostic rather than screening schedules. The final analysis was based on 1057 women.
Information From the Telephone Interviews. The sociodemographics included age, ethnic background, educational level, marital status, employment, and financial status.
Medical and family history included whether the woman ever had an abnormal mammogram, a biopsy, or a first-degree relative with breast cancer.
Provider-related measures assessed whether the woman had a regular physician and a provider recommendation for mammography and whether she discussed decisions with her care providers.
Breast cancer screening measures assessed mammography and clinical breast examinations (CBE) using questions asked by the Breast Cancer Screening Consortium of the NCI.15
Other health-related behaviors included when women had their most recent cervical screening, if they exercised regularly (if so, how often), whether they smoked (if so, how frequently), and whether they had thought about, had ever used, and were currently using hormonal replacement therapy (HRT).
Mammography knowledge, beliefs, and perceptions included whether mammograms are effective for reducing breast cancer deaths, how often a woman should be screened, and at what age a woman is more likely to develop breast cancer. In addition, women were asked whether they agreed, disagreed, or were undecided about 20 statements (11 pro and 9 con) about mammography screening consistent with the Transtheoretical model.16,17 The 11 pro and 9 con statements were used to compute pro and con scores, respectively. A high pro score indicates positive beliefs about mammography, while a high con score indicates negative beliefs. Previous research indicates that women who have more pros are more likely to get regular screening mammograms.
Risk perception measures assessed perceived absolute and comparative (self vs other) breast cancer risks. The absolute risk questions included: “How likely are you to get breast cancer in (a) the next 10 years and (b) your lifetime?” Responses were on a 5-point scale from “very unlikely” to “very likely.” For comparative risk the women were asked, “Compared with other women your age, how likely are you to get breast cancer in (a) the next 10 years and (b) your lifetime?” Responses were on a 5-point scale from “much below” to “much above average.”
Worry about breast cancer was measured for the next 10 years and a woman’s lifetime. Five responses ranged from not at all to very worried.
The women were also asked whether they felt ambivalent about getting a mammogram within their age-specific recommended time frames. The responses were agree, disagree, and undecided.
Statistical Analysis
We used the Pearson chi-square test to compare differences in the never-had, off-schedule, and on-schedule groups on provider-related information about mammography screening, women’s mammography knowledge, risk perceptions, worry about breast cancer, ambivalence, and other health-related behaviors. In addition, we used the F test to compare differences in perceived pro and con scales in the 3 groups. Because we were testing several hypotheses, all tests were performed using a 2-sided a=0.01.18
Because we were interested in identifying factors that are associated with repeat mammography use and because the proportional odds assumption was violated, women who never had a mammogram were excluded from the logistic regression analysis. The same results were observed when the never-had group was included with the off-schedule group (data not shown). In a logistic regression analysis modeling the probability of being off schedule, candidate variables were those that had a P value less than or equal to .20 in bivariate analyses Table 2
Table 3. Variables retained in the final model were those significant at a P value less than or equal to .01 level. We calculated odds ratios and 95% confidence intervals for independent variables in the model.
Results
Demographic and Medical History Factors
Seventy-four percent of the participants had more than a high school education. Eighty-two percent were white Table 2. The majority were married (78%), worked for pay (89%), and reported adequate income (91%).
One fourth reported a previous abnormal mammogram; 13% reported at least 1 biopsy; and 9% reported a first-degree relative with breast cancer.
Off-schedule women were less likely than those who were on schedule to report first-degree relatives with breast cancer, previous abnormal mammograms, and breast biopsies.
Provider-Related Information and Knowledge About Screening
The majority of women reported having a regular physician who recommended a mammogram within the past 2 years Table 2. Although most reported that they shared in medical testing decisions with their physicians, only 12% reported having raised questions about breast cancer screening with their physicians during the past 2 years. Off-schedule women were less likely than those on schedule to report having a regular physician and receiving a mammography recommendation in the past 2 years. Off-schedule women were more likely to report these factors than those never screened.
Almost all women reported that they believed mammograms to be effective in reducing breast cancer deaths, but women who were on and off schedule were more likely to do so than the never-had group. Off-schedule women were less likely than the on-schedule and never-had groups to correctly report screening recommendations for women their age.
Risk Perceptions
Comparative risk perceptions were associated with mammography history Table 2. Women who never had a mammogram perceived their comparative risk as lower than those in the on- and off-schedule groups, with off-schedule women falling between the never-had and on-schedule groups.
Perceptions About Mammography
The 3 groups of women differed significantly in perceptions about mammography Table 2. Off-schedule women were more likely than on-schedule women, but less likely than the never-had group, to be ambivalent about mammography and confused about guidelines. Off-schedule women were less likely than on-schedule women to report insufficient information to decide to get a mammogram, but were also more likely than the never-had group to report enough information.
Pro and con mammography scores, which reflected women’s positive and negative beliefs about mammography screening and the likelihood of being screened, were associated with their mammography history (data not shown). The off-schedule group had a significantly lower mean pro score (mean=9.4, standard deviation [SD] =2.4) than on-schedule women (mean=10, SD=1.5), but had a significantly higher mean pro score than the never-had group (mean=7.7, SD=3.6, P <.001). The off-schedule group had a significantly higher mean con score (mean=-5.5, SD=3.1) than on-schedule women (mean=-6.3, SD=2.7), but had a significantly lower mean con score than the never-had group (mean=-3.7, SD=3.6, P <.001).
Screening and Other Health-Related Behaviors
Overall, the majority of the women reported recent CBEs and Papanicolaou (Pap) tests. Approximately half said they got regular exercise and had used HRT Table 3.
Recent CBE and Pap tests were associated with mammography history. Off-schedule women were less likely than those who were on schedule but more likely than the never-had group to report both having a CBE within the past 12 months and a Pap test within 24 months. This pattern persisted among younger (42 to 45 years) and older (50 to 55 years) women.
Previous and current HRT use were associated with mammography history for women aged 50 years and older. Off-schedule women were less likely than on-schedule women, but more likely than women who never had a mammogram, to have used HRT.
Cigarette smoking was associated with mammography history. Women who never had mammograms were more likely to be current smokers than those in the on- and off-schedule groups.
We tested whether women off schedule for mammography were also off schedule for CBEs and cervical screening Table 4. Women who had a CBE within the past 12 months and a Pap test within the past 24 months were considered on schedule for both tests. A chi-square test of trend (P <.001) revealed a strong relationship between being on schedule for mammography screening and being on schedule for CBE and cervical screening.
Multivariate Analysis
Important factors associated with being off schedule for screening mammography were: being aged 50 to 54 years, not having a CBE within the past 12 months, being ambivalent about mammography, low perception of breast cancer risk, not being advised to have a mammogram by a physician in the past 2 years, confusion about screening mammography, and never having an abnormal mammogram Table 5.
Discussion
Although there have been significant increases in use of screening mammography during the last decade,3-4,8,10,11 at least 40% of the women in the United States are not adherent to the recommended guidelines. This is an important problem, because regular screening is needed to yield maximal breast cancer mortality reductions.
All of the participants in our study were in age categories for which there are mammography recommendations. It is noteworthy that even though all of the women in our study had insurance covering mammography and were in a plan that actively promoted screening, approximately half were either off schedule or never had a mammogram. This is consistent with the findings of other studies that financial coverage is necessary but not sufficient for mammography use.19
Several provider-related factors were significantly associated with the screening group; off-schedule women were less likely than their on-schedule counterparts but more likely than the never-had group to report having a regular physician, a discussion of mammography with their physicians, or a mammography recommendation from a physician within the past 2 years. The relationship between physician discussion and recommendations could be bidirectional, in that on-schedule women may be more open to discussion or at least perceived by their physicians to be so. They may even be more likely to initiate such discussions. Previous research20 has shown that physician recommendations facilitate adherence. Our data further support the important role of physician discussion and recommendations in repeat adherence. Thus, physicians should continue to reinforce the importance of mammography even for women who have been on schedule.
Although the majority of women in our study knew that mammograms are effective in reducing breast cancer mortality, there were differences by group in knowledge. Women who never had a mammogram were less likely to report that mammograms are effective. Off-schedule women were less knowledgeable than either the on-schedule or never-had groups about how often women should be screened; perhaps this lack of knowledge about when to be rescreened contributes to their being off schedule. In any case, it is important for the physician to remind a woman about the appropriate schedule and to provide a referral.
Off-schedule women were more likely than on-schedule women to be ambivalent about mammography and confused about screening guidelines. Whether these findings can be attributed to the guideline debate of 1997 shortly before our data collection cannot be determined. However, these findings do indicate a need for mammography education about both the rationale for repeat screening and specific information about recommended guidelines.
There is increased interest in evaluating multiple risk behaviors. Our results confirm other findings21-24 that women who are off schedule for mammography are less likely to be adherent for other screening behaviors. Consistent with other studies,21,25-27 we found smokers were less likely to be on schedule for screening mammography. These findings suggest that it may be useful to address multiple screening behaviors rather than focusing on one test at a time.
There were associations between mammography history and variables related to HRT. Consistent with other research,28 off-schedule women were less likely to have ever used or to currently be using HRT. Because it is likely that physicians routinely order mammograms before prescribing HRT, this association may be due more to routine medical procedure than patient characteristics.29 However, whether decisions to use HRT and to have regular mammograms are associated should be explored.
Also consistent with previous findings,7,8 multivariate analyses revealed that younger age, having a CBE within the past 12 months, and physician recommendations were important factors associated with repeat mammography. As previously reported,13 “feeling torn” about mammography and being confused about screening guidelines were negatively associated with being on schedule for mammography.
Limitations
One limitation of our study is that our sample was drawn from women with health insurance rather than from the general population. Thus, we cannot generalize the results of our study to the entire population of North Carolina.
Also, because the sample was drawn for the purposes of a subsequent intervention, there are some other anomalies. We stratified the sample on the basis of age and adherence status, and thus the proportions per se cannot be generalized to the health plan.
Another limitation is that we collected self-report information only on the 2 most recent mammograms. Although long-term mammography history studies should be conducted in the future, ours is one of only a few studies to date that assessed more than 1-time mammography use. Thus, our findings set the stage for future assessments of repeat adherence. Previous research suggests that the correspondence between self-report and mammography use is very high in health maintenance organization settings,30,31 but there is a discrepancy in recall of timing of the mammogram.32 Although we cannot conclusively verify the date of last and previous mammograms, our findings show expected differences between those who reported being on versus off schedule. The 3-month window we allowed before categorizing women as off schedule may have limited misclassification of adherent women as nonadherent. Thus, we probably underestimated the number of women who were off schedule for repeat mammography.
Conclusions
Our study is one of a small number to analyze differences in beliefs and other health-related behaviors among groups of women who are on schedule or off schedule for a mamogram and those who never had mammograms. With a few exceptions, the results suggest a trend, as the off-schedule group almost consistently falls between the on-schedule and never-had groups. For instance, they were more likely than those never screened but less likely than on-schedule women to report the kind of provider support (discussions and recommendations) that facilitates screening and to understand the rationale and recommendations for regular screening. Off-schedule women also showed a need to change other health-related behaviors. Off-schedule women were also likely to perceive their breast cancer risk as lower, be less likely to be up to date with other cancer screening tests, and to have ever used HRT.
Because there are few studies comparing women who are on versus off schedule for their 2 most recent mammograms, we were not sure how, for instance, the off-schedule and the never-had group would compare. Our findings suggest that women who are off schedule are in need of mammography-promoting interventions, including recommendations from and discussion with their health care providers. Because they are more positive and knowledgeable about mammography than never screened women, they may benefit from brief interventions from health care providers that highlight the importance of regular screening.
Significant progress has been made in the proportion of women in the United States who have been screened. Further increases will be dependent not only on motivating women who have never been screened but also in enhancing levels of regular screening. Physicians have a central role to play in facilitating regular screening.
Acknowledgments
Our study was funded by the National Cancer Institute grant #5U19-CA-72099-03. We express our sincere appreciation to Don Bradley, MD, at Blue Cross Blue Shield of North Carolina for his leadership and the many women who are participating in this project. We thank Elizabeth Powell for the preparation of the manuscript. Our manuscript represents the perspective of the authors and not the National Cancer Institute.
Related resources
- National Cancer Institute Cancer information, news on research, funding and treatment recommendations. www.nci.nih.gov
- American Cancer Society News on cancer research. Search function identifies local resources. News on breast and other cancers. Information on ACS research and funding programs. Yearly statistics on incidence of cancer dating back to 1995. www.cancer.org
1. Baker LH. Breast cancer detection demonstration project: five-year summary report. Cancer 2. 1982;32:194-225
2. Shapiro S, Venet W, Strax P, Venet L, Roeser R. Ten-to-fourteen year effect of screening on breast cancer mortality. J Natl Cancer Inst 1982;69:349-55
3. Anonymous. Self-reported use of mammography and insurance status among women aged Ž 40 years—United States, 1991-1992 and 1996-1997. MMWR Morb Mortal Wkly Rep 1998;47:825-30
4. Anonymous. Self-reported use of mammography among women aged Ž40 years—United States, 1989 and 1995. MMWR Morb Mortal Wkly Rep 1997;46:937-41
5. Faulkner LA, Schauffler HH. The effect of health insurance coverage on the appropriate use of recommended clinical preventive services. Am J Prev Med 1997;13:453-58
6. Hahn RA, Teutsch SM, Franks AL, Chang MH, Lloyd EE. The prevalence of risk factors among women in the United States by race and age, 1992-1994: opportunities for primary and secondary prevention. J Am Med Womens Assoc 1998;53:96-104,107.
7. Lerman C, Rimer B, Trock B, Balshem A, Engstrom P. Factors associated with repeat adherence to breast cancer screening. Prev Med 1990;19:279-90
8. Bastani R, Kaplan CP, Maxwell AE, Nisenbaum R, Pearce J, Marcus AC. Initial and repeat mammography screening in a low income population in Los Angeles. Cancer Epidemiol Biomarkers Prev 1995;4:161-71.
9. Burack RC, Gimotty PA. Promoting screening mammography in inner-city settings: the sustained effectiveness of computerized reminders in a randomized controlled trial. Med Care 1997;35:921-31
10. Song L, Fletcher R. Breast cancer rescreening in low-income women. Am J Prev Med 1998;15:128-33
11. Yood MU, McCarthy BD, Lee NC, Jacobsen G, Johnson CC. Patterns and characteristics of repeat mammography among women 50 years and older. Cancer Epidemiol Biomarkers Prev 1999;8:595-99
12. Lipkus IM, Rimer BK, Halabi S, Strigo TS. Can tailored interventions increase mammography use among HMO women? Am J Prev Med 2000;18:1-10
13. Rimer BK, Halabi S, Strigo TS, Crawford Y, Lipkus IM. Confusion about mammography: prevalence and consequences. J Women’s Health Gender-Based Med 1999;8:509-20
14. National Cancer Institute and American Cancer Society. Joint statement on breast cancer screening for women in their 40s. The Cancer Information Service; 1997.
15. Stoddard AM, Rimer BK, Lane D, et al. for the NCI Breast Cancer Consortium. Underusers of mammogram screening: stage of adoption in five US subpopulations. Prev Med 1998;27:478-87
16. Rakowski W, Ehrich B, Golsetin M, et al. A stage-matched intervention for screening mammography. Ann Behav Med 1997;19:S063.-
17. Velicer W, DiClemente C, Prochaska J, et al. A decisional balance measure for assessing and predicting smoking status. J Personality Soc Psychol 1985;48:1279-89
18. Forthofer RN, Lehnen RF. Public program analysis: a new categorical data analysis approach. Belmont: Lifetime Learning Publications; 1981.
19. Rimer BK, Resch N, King E, et al. Multistrategy health education program to increase mammography use among women ages 65 and older. Public Health Rep 1992;107:369-80
20. Skinner, Strecher, Hospers. Physicians’ recommendations for mammography: do tailored messages make a difference? Am J Public Health 1994;84:43-49
21. Ronco G, Segnan N, Ponti A. Who has Pap tests? Variables associated with the use of Pap tests in absence of screening programmes. Int J Epidemiol 1991;20:349-53
22. Rakowski W, Rimer BK, Bryant SA. Integrating behavior and intention regarding mammography by respondents in the 1990 national health interview survey of health promotion and disease prevention. Pub Health Reports 1993;108:605-24
23. Hyman RB, Greewald ES, Hacker S. Smoking, dietary, and breast and cervical cancer screening knowledge and screening practices of employees in an urban medical center. J Cancer Educ 1995;10:82-87
24. Pearlman DN, Rakowski W, Ehrich B. Mammography, clinical breast exam and Pap testing: correlates of combined screening. Am J Prev Med 1996;12:52-64
25. Orleans CT, Rimer BK, Cristinzio S, Keintz MK, Fleisher L. A national survey of older smokers: treatment needs of a growing population. Health Psychol 1991;10:343-51.
26. McBride CM, Curry SJ, Taplin S, Anderman C, Grothaus L. Exploring environmental barriers to participation in mammography screening in an HMO. Can Epidemiol Biomarkers Prev 1993;2:559-605
27. Beaulieu MD, Beland F, Roy D, Falardeau M, Herbert G. Factors determining compliance with screening mammography. Can Med Assoc J 1996;154:1335-43
28. Bastian LA, Couchman GM, Rimer BK, McBride CM, Feaganes JR, Siegler IC. Perceptions of menopausal stage and patterns of hormone replacement therapy use. J Women’s Health 1997;6:467-75
29. Personal communication with Lori Bastian.
30. King ES, Rimer BK, Trock B, Balshem A, Engstrom P. How valid are mammography self-reports? Am J Public Health 1990;80:1386-88
31. Degnan D, Harris R, Ranney J, Quade D, Earp JA, Gonzalez J. Measuring the use of mammography: two methods compared. Am J Public Health 1992;82:1386-88
32. PM, Mickey RM, Worden JK. Reliability of self-reported breast screening information in a survey of lower income women. Prev Med 1997;26:287-91
METHODS: Our data are from a baseline telephone interview conducted among 1287 female members of Blue Cross Blue Shield of North Carolina who were aged either 40 to 44 years or 50 to 54 years.
RESULTS: The 3 groups differed significantly on beliefs and health-related behaviors, with the off-schedule group almost consistently falling between the on-schedule and never screened groups. Off-schedule women were more likely than on-schedule women, but less likely than those never screened, to not have a clinical breast examination within 12 months, to be ambivalent about screening mammography, to be confused about screening guidelines, and to not be advised by a physician to get a mammogram in the past 2 years. Off-schedule women perceived their breast cancer risk as lower and were less likely to be up to date with other cancer screening tests.
CONCLUSIONS: Our findings suggest that women who are off schedule are in need of mammography-promoting interventions, including recommendations from and discussion with their health care providers. Because they are more positive and knowledgeable about mammography than women who have never been screened, they may benefit from brief interventions from health care providers that highlight the importance of repeat screening.
Because most research has shown that routine mammography screening saves lives,1,2 many medical organizations have developed guidelines and recommendations for mammography screening. For example, the National Cancer Institute (NCI) recommends mammograms every 1 to 2 years for women aged 40 years and older and annual screening for women aged 50 years and older. Other organizations, such as the United States Preventive Health Task Force and the American College of Preventive Medicine, differ from the NCI regarding optimal screening intervals, but all agree on the importance of regular screening to achieve a breast cancer mortality reduction benefit.
However, despite these recommendations and substantial increases in the percentage of women who have ever had mammograms, most women are not having regular screening at recommended intervals. There is also still a group of women who have never been screened.3-6 For the purpose of developing interventions to encourage routine screening, it may be important to understand differences among women who are and are not getting repeat screening and those who have never been screened. However, with a few exceptions,7-12 most mammography studies have focused on 1-time screening rather than repeat adherence. Few studies have identified factors that predict repeat mammography use according to recommended guidelines.
For our analysis we sought to better understand factors that differentiate 3 groups of insured women: those who have been screened and are adherent to the screening mammography guidelines of the NCI, those screened at least once but who are currently nonadherent, and those never screened. Study findings should guide the design of interventions to promote continued adherence for repeat screening mammography.
Methods
Study Population
We randomly selected 2165 women from a sampling frame of 4000 women aged 40 to 44 years and 50 to 54 years who were members of the Personal Care Plan of Blue Cross Blue Shield of North Carolina (BCBS of NC) in 1997. We stratified the sample by age and mammography compliance status. Women with previous breast cancer and those who were no longer BCBS members were excluded from the sample. The completion rate for the telephone interviews was 76%, and the nonresponse rate was 20%, leaving a sample of 1287 women.
We mailed introductory letters, and professional interviewers conducted telephone interviews between November 1997 and May 1998. The participants provided oral consent in accordance with regulations from the Department of Health and Human Services. The analyses reported in this paper were conducted on baseline data collected for a larger intervention trial designed to enhance informed decision making about mammography. Additional details regarding our study methodology have been published elsewhere.13
Measures
Screening History Measure. The main variable of interest was self-reported mammography history (including most recent and previous mammograms). We calculated 2 screening variables reflecting whether: (1) the most recent mammogram was within the recommended time frame according to NCI recommendations, and (2) the interval between the most recent and the previous mammography date was within the recommended time frame for the woman’s age. The second interval could be computed only for women with more than 1 previous mammogram.
We categorized the participants as “never had a mammogram,” “off schedule,” or “on schedule for their 2 most recent mammograms.” (We refer to the groups as never had, off schedule, and on schedule.)
We followed the recommendations of BCBS of NC which specified mammography consistent with the NCI recommendation: every 1 to 2 years for women in their 40s and every year for women in their 50s.14 However, because many women are not screened exactly 12 months following their previous mammogram, we added a 3-month window to the intervals; thus, the window was 15 months for women in their 50s and 27 months for women in their 40s. These interval windows are consistent with those adopted by other investigators.10
Our on- and off-schedule classification algorithms allow for women in their early 50s who passed from the “every 1 to 2 years” to the “every year” guidelines between their most recent and previous mammograms and those in their 40s who had not yet had time for 2 mammograms based on their age and the recommendations. The classification of women by screening mammography history is presented in Table 1
Because NCI recommendations indicate a single mammogram for average-risk women younger than 42 years, we could not consider women younger than this to be off schedule, so we excluded them (n=198) from our analysis. Women who had 2 mammograms within 11 months (n=32) were also excluded, because it is likely they were on diagnostic rather than screening schedules. The final analysis was based on 1057 women.
Information From the Telephone Interviews. The sociodemographics included age, ethnic background, educational level, marital status, employment, and financial status.
Medical and family history included whether the woman ever had an abnormal mammogram, a biopsy, or a first-degree relative with breast cancer.
Provider-related measures assessed whether the woman had a regular physician and a provider recommendation for mammography and whether she discussed decisions with her care providers.
Breast cancer screening measures assessed mammography and clinical breast examinations (CBE) using questions asked by the Breast Cancer Screening Consortium of the NCI.15
Other health-related behaviors included when women had their most recent cervical screening, if they exercised regularly (if so, how often), whether they smoked (if so, how frequently), and whether they had thought about, had ever used, and were currently using hormonal replacement therapy (HRT).
Mammography knowledge, beliefs, and perceptions included whether mammograms are effective for reducing breast cancer deaths, how often a woman should be screened, and at what age a woman is more likely to develop breast cancer. In addition, women were asked whether they agreed, disagreed, or were undecided about 20 statements (11 pro and 9 con) about mammography screening consistent with the Transtheoretical model.16,17 The 11 pro and 9 con statements were used to compute pro and con scores, respectively. A high pro score indicates positive beliefs about mammography, while a high con score indicates negative beliefs. Previous research indicates that women who have more pros are more likely to get regular screening mammograms.
Risk perception measures assessed perceived absolute and comparative (self vs other) breast cancer risks. The absolute risk questions included: “How likely are you to get breast cancer in (a) the next 10 years and (b) your lifetime?” Responses were on a 5-point scale from “very unlikely” to “very likely.” For comparative risk the women were asked, “Compared with other women your age, how likely are you to get breast cancer in (a) the next 10 years and (b) your lifetime?” Responses were on a 5-point scale from “much below” to “much above average.”
Worry about breast cancer was measured for the next 10 years and a woman’s lifetime. Five responses ranged from not at all to very worried.
The women were also asked whether they felt ambivalent about getting a mammogram within their age-specific recommended time frames. The responses were agree, disagree, and undecided.
Statistical Analysis
We used the Pearson chi-square test to compare differences in the never-had, off-schedule, and on-schedule groups on provider-related information about mammography screening, women’s mammography knowledge, risk perceptions, worry about breast cancer, ambivalence, and other health-related behaviors. In addition, we used the F test to compare differences in perceived pro and con scales in the 3 groups. Because we were testing several hypotheses, all tests were performed using a 2-sided a=0.01.18
Because we were interested in identifying factors that are associated with repeat mammography use and because the proportional odds assumption was violated, women who never had a mammogram were excluded from the logistic regression analysis. The same results were observed when the never-had group was included with the off-schedule group (data not shown). In a logistic regression analysis modeling the probability of being off schedule, candidate variables were those that had a P value less than or equal to .20 in bivariate analyses Table 2
Table 3. Variables retained in the final model were those significant at a P value less than or equal to .01 level. We calculated odds ratios and 95% confidence intervals for independent variables in the model.
Results
Demographic and Medical History Factors
Seventy-four percent of the participants had more than a high school education. Eighty-two percent were white Table 2. The majority were married (78%), worked for pay (89%), and reported adequate income (91%).
One fourth reported a previous abnormal mammogram; 13% reported at least 1 biopsy; and 9% reported a first-degree relative with breast cancer.
Off-schedule women were less likely than those who were on schedule to report first-degree relatives with breast cancer, previous abnormal mammograms, and breast biopsies.
Provider-Related Information and Knowledge About Screening
The majority of women reported having a regular physician who recommended a mammogram within the past 2 years Table 2. Although most reported that they shared in medical testing decisions with their physicians, only 12% reported having raised questions about breast cancer screening with their physicians during the past 2 years. Off-schedule women were less likely than those on schedule to report having a regular physician and receiving a mammography recommendation in the past 2 years. Off-schedule women were more likely to report these factors than those never screened.
Almost all women reported that they believed mammograms to be effective in reducing breast cancer deaths, but women who were on and off schedule were more likely to do so than the never-had group. Off-schedule women were less likely than the on-schedule and never-had groups to correctly report screening recommendations for women their age.
Risk Perceptions
Comparative risk perceptions were associated with mammography history Table 2. Women who never had a mammogram perceived their comparative risk as lower than those in the on- and off-schedule groups, with off-schedule women falling between the never-had and on-schedule groups.
Perceptions About Mammography
The 3 groups of women differed significantly in perceptions about mammography Table 2. Off-schedule women were more likely than on-schedule women, but less likely than the never-had group, to be ambivalent about mammography and confused about guidelines. Off-schedule women were less likely than on-schedule women to report insufficient information to decide to get a mammogram, but were also more likely than the never-had group to report enough information.
Pro and con mammography scores, which reflected women’s positive and negative beliefs about mammography screening and the likelihood of being screened, were associated with their mammography history (data not shown). The off-schedule group had a significantly lower mean pro score (mean=9.4, standard deviation [SD] =2.4) than on-schedule women (mean=10, SD=1.5), but had a significantly higher mean pro score than the never-had group (mean=7.7, SD=3.6, P <.001). The off-schedule group had a significantly higher mean con score (mean=-5.5, SD=3.1) than on-schedule women (mean=-6.3, SD=2.7), but had a significantly lower mean con score than the never-had group (mean=-3.7, SD=3.6, P <.001).
Screening and Other Health-Related Behaviors
Overall, the majority of the women reported recent CBEs and Papanicolaou (Pap) tests. Approximately half said they got regular exercise and had used HRT Table 3.
Recent CBE and Pap tests were associated with mammography history. Off-schedule women were less likely than those who were on schedule but more likely than the never-had group to report both having a CBE within the past 12 months and a Pap test within 24 months. This pattern persisted among younger (42 to 45 years) and older (50 to 55 years) women.
Previous and current HRT use were associated with mammography history for women aged 50 years and older. Off-schedule women were less likely than on-schedule women, but more likely than women who never had a mammogram, to have used HRT.
Cigarette smoking was associated with mammography history. Women who never had mammograms were more likely to be current smokers than those in the on- and off-schedule groups.
We tested whether women off schedule for mammography were also off schedule for CBEs and cervical screening Table 4. Women who had a CBE within the past 12 months and a Pap test within the past 24 months were considered on schedule for both tests. A chi-square test of trend (P <.001) revealed a strong relationship between being on schedule for mammography screening and being on schedule for CBE and cervical screening.
Multivariate Analysis
Important factors associated with being off schedule for screening mammography were: being aged 50 to 54 years, not having a CBE within the past 12 months, being ambivalent about mammography, low perception of breast cancer risk, not being advised to have a mammogram by a physician in the past 2 years, confusion about screening mammography, and never having an abnormal mammogram Table 5.
Discussion
Although there have been significant increases in use of screening mammography during the last decade,3-4,8,10,11 at least 40% of the women in the United States are not adherent to the recommended guidelines. This is an important problem, because regular screening is needed to yield maximal breast cancer mortality reductions.
All of the participants in our study were in age categories for which there are mammography recommendations. It is noteworthy that even though all of the women in our study had insurance covering mammography and were in a plan that actively promoted screening, approximately half were either off schedule or never had a mammogram. This is consistent with the findings of other studies that financial coverage is necessary but not sufficient for mammography use.19
Several provider-related factors were significantly associated with the screening group; off-schedule women were less likely than their on-schedule counterparts but more likely than the never-had group to report having a regular physician, a discussion of mammography with their physicians, or a mammography recommendation from a physician within the past 2 years. The relationship between physician discussion and recommendations could be bidirectional, in that on-schedule women may be more open to discussion or at least perceived by their physicians to be so. They may even be more likely to initiate such discussions. Previous research20 has shown that physician recommendations facilitate adherence. Our data further support the important role of physician discussion and recommendations in repeat adherence. Thus, physicians should continue to reinforce the importance of mammography even for women who have been on schedule.
Although the majority of women in our study knew that mammograms are effective in reducing breast cancer mortality, there were differences by group in knowledge. Women who never had a mammogram were less likely to report that mammograms are effective. Off-schedule women were less knowledgeable than either the on-schedule or never-had groups about how often women should be screened; perhaps this lack of knowledge about when to be rescreened contributes to their being off schedule. In any case, it is important for the physician to remind a woman about the appropriate schedule and to provide a referral.
Off-schedule women were more likely than on-schedule women to be ambivalent about mammography and confused about screening guidelines. Whether these findings can be attributed to the guideline debate of 1997 shortly before our data collection cannot be determined. However, these findings do indicate a need for mammography education about both the rationale for repeat screening and specific information about recommended guidelines.
There is increased interest in evaluating multiple risk behaviors. Our results confirm other findings21-24 that women who are off schedule for mammography are less likely to be adherent for other screening behaviors. Consistent with other studies,21,25-27 we found smokers were less likely to be on schedule for screening mammography. These findings suggest that it may be useful to address multiple screening behaviors rather than focusing on one test at a time.
There were associations between mammography history and variables related to HRT. Consistent with other research,28 off-schedule women were less likely to have ever used or to currently be using HRT. Because it is likely that physicians routinely order mammograms before prescribing HRT, this association may be due more to routine medical procedure than patient characteristics.29 However, whether decisions to use HRT and to have regular mammograms are associated should be explored.
Also consistent with previous findings,7,8 multivariate analyses revealed that younger age, having a CBE within the past 12 months, and physician recommendations were important factors associated with repeat mammography. As previously reported,13 “feeling torn” about mammography and being confused about screening guidelines were negatively associated with being on schedule for mammography.
Limitations
One limitation of our study is that our sample was drawn from women with health insurance rather than from the general population. Thus, we cannot generalize the results of our study to the entire population of North Carolina.
Also, because the sample was drawn for the purposes of a subsequent intervention, there are some other anomalies. We stratified the sample on the basis of age and adherence status, and thus the proportions per se cannot be generalized to the health plan.
Another limitation is that we collected self-report information only on the 2 most recent mammograms. Although long-term mammography history studies should be conducted in the future, ours is one of only a few studies to date that assessed more than 1-time mammography use. Thus, our findings set the stage for future assessments of repeat adherence. Previous research suggests that the correspondence between self-report and mammography use is very high in health maintenance organization settings,30,31 but there is a discrepancy in recall of timing of the mammogram.32 Although we cannot conclusively verify the date of last and previous mammograms, our findings show expected differences between those who reported being on versus off schedule. The 3-month window we allowed before categorizing women as off schedule may have limited misclassification of adherent women as nonadherent. Thus, we probably underestimated the number of women who were off schedule for repeat mammography.
Conclusions
Our study is one of a small number to analyze differences in beliefs and other health-related behaviors among groups of women who are on schedule or off schedule for a mamogram and those who never had mammograms. With a few exceptions, the results suggest a trend, as the off-schedule group almost consistently falls between the on-schedule and never-had groups. For instance, they were more likely than those never screened but less likely than on-schedule women to report the kind of provider support (discussions and recommendations) that facilitates screening and to understand the rationale and recommendations for regular screening. Off-schedule women also showed a need to change other health-related behaviors. Off-schedule women were also likely to perceive their breast cancer risk as lower, be less likely to be up to date with other cancer screening tests, and to have ever used HRT.
Because there are few studies comparing women who are on versus off schedule for their 2 most recent mammograms, we were not sure how, for instance, the off-schedule and the never-had group would compare. Our findings suggest that women who are off schedule are in need of mammography-promoting interventions, including recommendations from and discussion with their health care providers. Because they are more positive and knowledgeable about mammography than never screened women, they may benefit from brief interventions from health care providers that highlight the importance of regular screening.
Significant progress has been made in the proportion of women in the United States who have been screened. Further increases will be dependent not only on motivating women who have never been screened but also in enhancing levels of regular screening. Physicians have a central role to play in facilitating regular screening.
Acknowledgments
Our study was funded by the National Cancer Institute grant #5U19-CA-72099-03. We express our sincere appreciation to Don Bradley, MD, at Blue Cross Blue Shield of North Carolina for his leadership and the many women who are participating in this project. We thank Elizabeth Powell for the preparation of the manuscript. Our manuscript represents the perspective of the authors and not the National Cancer Institute.
Related resources
- National Cancer Institute Cancer information, news on research, funding and treatment recommendations. www.nci.nih.gov
- American Cancer Society News on cancer research. Search function identifies local resources. News on breast and other cancers. Information on ACS research and funding programs. Yearly statistics on incidence of cancer dating back to 1995. www.cancer.org
METHODS: Our data are from a baseline telephone interview conducted among 1287 female members of Blue Cross Blue Shield of North Carolina who were aged either 40 to 44 years or 50 to 54 years.
RESULTS: The 3 groups differed significantly on beliefs and health-related behaviors, with the off-schedule group almost consistently falling between the on-schedule and never screened groups. Off-schedule women were more likely than on-schedule women, but less likely than those never screened, to not have a clinical breast examination within 12 months, to be ambivalent about screening mammography, to be confused about screening guidelines, and to not be advised by a physician to get a mammogram in the past 2 years. Off-schedule women perceived their breast cancer risk as lower and were less likely to be up to date with other cancer screening tests.
CONCLUSIONS: Our findings suggest that women who are off schedule are in need of mammography-promoting interventions, including recommendations from and discussion with their health care providers. Because they are more positive and knowledgeable about mammography than women who have never been screened, they may benefit from brief interventions from health care providers that highlight the importance of repeat screening.
Because most research has shown that routine mammography screening saves lives,1,2 many medical organizations have developed guidelines and recommendations for mammography screening. For example, the National Cancer Institute (NCI) recommends mammograms every 1 to 2 years for women aged 40 years and older and annual screening for women aged 50 years and older. Other organizations, such as the United States Preventive Health Task Force and the American College of Preventive Medicine, differ from the NCI regarding optimal screening intervals, but all agree on the importance of regular screening to achieve a breast cancer mortality reduction benefit.
However, despite these recommendations and substantial increases in the percentage of women who have ever had mammograms, most women are not having regular screening at recommended intervals. There is also still a group of women who have never been screened.3-6 For the purpose of developing interventions to encourage routine screening, it may be important to understand differences among women who are and are not getting repeat screening and those who have never been screened. However, with a few exceptions,7-12 most mammography studies have focused on 1-time screening rather than repeat adherence. Few studies have identified factors that predict repeat mammography use according to recommended guidelines.
For our analysis we sought to better understand factors that differentiate 3 groups of insured women: those who have been screened and are adherent to the screening mammography guidelines of the NCI, those screened at least once but who are currently nonadherent, and those never screened. Study findings should guide the design of interventions to promote continued adherence for repeat screening mammography.
Methods
Study Population
We randomly selected 2165 women from a sampling frame of 4000 women aged 40 to 44 years and 50 to 54 years who were members of the Personal Care Plan of Blue Cross Blue Shield of North Carolina (BCBS of NC) in 1997. We stratified the sample by age and mammography compliance status. Women with previous breast cancer and those who were no longer BCBS members were excluded from the sample. The completion rate for the telephone interviews was 76%, and the nonresponse rate was 20%, leaving a sample of 1287 women.
We mailed introductory letters, and professional interviewers conducted telephone interviews between November 1997 and May 1998. The participants provided oral consent in accordance with regulations from the Department of Health and Human Services. The analyses reported in this paper were conducted on baseline data collected for a larger intervention trial designed to enhance informed decision making about mammography. Additional details regarding our study methodology have been published elsewhere.13
Measures
Screening History Measure. The main variable of interest was self-reported mammography history (including most recent and previous mammograms). We calculated 2 screening variables reflecting whether: (1) the most recent mammogram was within the recommended time frame according to NCI recommendations, and (2) the interval between the most recent and the previous mammography date was within the recommended time frame for the woman’s age. The second interval could be computed only for women with more than 1 previous mammogram.
We categorized the participants as “never had a mammogram,” “off schedule,” or “on schedule for their 2 most recent mammograms.” (We refer to the groups as never had, off schedule, and on schedule.)
We followed the recommendations of BCBS of NC which specified mammography consistent with the NCI recommendation: every 1 to 2 years for women in their 40s and every year for women in their 50s.14 However, because many women are not screened exactly 12 months following their previous mammogram, we added a 3-month window to the intervals; thus, the window was 15 months for women in their 50s and 27 months for women in their 40s. These interval windows are consistent with those adopted by other investigators.10
Our on- and off-schedule classification algorithms allow for women in their early 50s who passed from the “every 1 to 2 years” to the “every year” guidelines between their most recent and previous mammograms and those in their 40s who had not yet had time for 2 mammograms based on their age and the recommendations. The classification of women by screening mammography history is presented in Table 1
Because NCI recommendations indicate a single mammogram for average-risk women younger than 42 years, we could not consider women younger than this to be off schedule, so we excluded them (n=198) from our analysis. Women who had 2 mammograms within 11 months (n=32) were also excluded, because it is likely they were on diagnostic rather than screening schedules. The final analysis was based on 1057 women.
Information From the Telephone Interviews. The sociodemographics included age, ethnic background, educational level, marital status, employment, and financial status.
Medical and family history included whether the woman ever had an abnormal mammogram, a biopsy, or a first-degree relative with breast cancer.
Provider-related measures assessed whether the woman had a regular physician and a provider recommendation for mammography and whether she discussed decisions with her care providers.
Breast cancer screening measures assessed mammography and clinical breast examinations (CBE) using questions asked by the Breast Cancer Screening Consortium of the NCI.15
Other health-related behaviors included when women had their most recent cervical screening, if they exercised regularly (if so, how often), whether they smoked (if so, how frequently), and whether they had thought about, had ever used, and were currently using hormonal replacement therapy (HRT).
Mammography knowledge, beliefs, and perceptions included whether mammograms are effective for reducing breast cancer deaths, how often a woman should be screened, and at what age a woman is more likely to develop breast cancer. In addition, women were asked whether they agreed, disagreed, or were undecided about 20 statements (11 pro and 9 con) about mammography screening consistent with the Transtheoretical model.16,17 The 11 pro and 9 con statements were used to compute pro and con scores, respectively. A high pro score indicates positive beliefs about mammography, while a high con score indicates negative beliefs. Previous research indicates that women who have more pros are more likely to get regular screening mammograms.
Risk perception measures assessed perceived absolute and comparative (self vs other) breast cancer risks. The absolute risk questions included: “How likely are you to get breast cancer in (a) the next 10 years and (b) your lifetime?” Responses were on a 5-point scale from “very unlikely” to “very likely.” For comparative risk the women were asked, “Compared with other women your age, how likely are you to get breast cancer in (a) the next 10 years and (b) your lifetime?” Responses were on a 5-point scale from “much below” to “much above average.”
Worry about breast cancer was measured for the next 10 years and a woman’s lifetime. Five responses ranged from not at all to very worried.
The women were also asked whether they felt ambivalent about getting a mammogram within their age-specific recommended time frames. The responses were agree, disagree, and undecided.
Statistical Analysis
We used the Pearson chi-square test to compare differences in the never-had, off-schedule, and on-schedule groups on provider-related information about mammography screening, women’s mammography knowledge, risk perceptions, worry about breast cancer, ambivalence, and other health-related behaviors. In addition, we used the F test to compare differences in perceived pro and con scales in the 3 groups. Because we were testing several hypotheses, all tests were performed using a 2-sided a=0.01.18
Because we were interested in identifying factors that are associated with repeat mammography use and because the proportional odds assumption was violated, women who never had a mammogram were excluded from the logistic regression analysis. The same results were observed when the never-had group was included with the off-schedule group (data not shown). In a logistic regression analysis modeling the probability of being off schedule, candidate variables were those that had a P value less than or equal to .20 in bivariate analyses Table 2
Table 3. Variables retained in the final model were those significant at a P value less than or equal to .01 level. We calculated odds ratios and 95% confidence intervals for independent variables in the model.
Results
Demographic and Medical History Factors
Seventy-four percent of the participants had more than a high school education. Eighty-two percent were white Table 2. The majority were married (78%), worked for pay (89%), and reported adequate income (91%).
One fourth reported a previous abnormal mammogram; 13% reported at least 1 biopsy; and 9% reported a first-degree relative with breast cancer.
Off-schedule women were less likely than those who were on schedule to report first-degree relatives with breast cancer, previous abnormal mammograms, and breast biopsies.
Provider-Related Information and Knowledge About Screening
The majority of women reported having a regular physician who recommended a mammogram within the past 2 years Table 2. Although most reported that they shared in medical testing decisions with their physicians, only 12% reported having raised questions about breast cancer screening with their physicians during the past 2 years. Off-schedule women were less likely than those on schedule to report having a regular physician and receiving a mammography recommendation in the past 2 years. Off-schedule women were more likely to report these factors than those never screened.
Almost all women reported that they believed mammograms to be effective in reducing breast cancer deaths, but women who were on and off schedule were more likely to do so than the never-had group. Off-schedule women were less likely than the on-schedule and never-had groups to correctly report screening recommendations for women their age.
Risk Perceptions
Comparative risk perceptions were associated with mammography history Table 2. Women who never had a mammogram perceived their comparative risk as lower than those in the on- and off-schedule groups, with off-schedule women falling between the never-had and on-schedule groups.
Perceptions About Mammography
The 3 groups of women differed significantly in perceptions about mammography Table 2. Off-schedule women were more likely than on-schedule women, but less likely than the never-had group, to be ambivalent about mammography and confused about guidelines. Off-schedule women were less likely than on-schedule women to report insufficient information to decide to get a mammogram, but were also more likely than the never-had group to report enough information.
Pro and con mammography scores, which reflected women’s positive and negative beliefs about mammography screening and the likelihood of being screened, were associated with their mammography history (data not shown). The off-schedule group had a significantly lower mean pro score (mean=9.4, standard deviation [SD] =2.4) than on-schedule women (mean=10, SD=1.5), but had a significantly higher mean pro score than the never-had group (mean=7.7, SD=3.6, P <.001). The off-schedule group had a significantly higher mean con score (mean=-5.5, SD=3.1) than on-schedule women (mean=-6.3, SD=2.7), but had a significantly lower mean con score than the never-had group (mean=-3.7, SD=3.6, P <.001).
Screening and Other Health-Related Behaviors
Overall, the majority of the women reported recent CBEs and Papanicolaou (Pap) tests. Approximately half said they got regular exercise and had used HRT Table 3.
Recent CBE and Pap tests were associated with mammography history. Off-schedule women were less likely than those who were on schedule but more likely than the never-had group to report both having a CBE within the past 12 months and a Pap test within 24 months. This pattern persisted among younger (42 to 45 years) and older (50 to 55 years) women.
Previous and current HRT use were associated with mammography history for women aged 50 years and older. Off-schedule women were less likely than on-schedule women, but more likely than women who never had a mammogram, to have used HRT.
Cigarette smoking was associated with mammography history. Women who never had mammograms were more likely to be current smokers than those in the on- and off-schedule groups.
We tested whether women off schedule for mammography were also off schedule for CBEs and cervical screening Table 4. Women who had a CBE within the past 12 months and a Pap test within the past 24 months were considered on schedule for both tests. A chi-square test of trend (P <.001) revealed a strong relationship between being on schedule for mammography screening and being on schedule for CBE and cervical screening.
Multivariate Analysis
Important factors associated with being off schedule for screening mammography were: being aged 50 to 54 years, not having a CBE within the past 12 months, being ambivalent about mammography, low perception of breast cancer risk, not being advised to have a mammogram by a physician in the past 2 years, confusion about screening mammography, and never having an abnormal mammogram Table 5.
Discussion
Although there have been significant increases in use of screening mammography during the last decade,3-4,8,10,11 at least 40% of the women in the United States are not adherent to the recommended guidelines. This is an important problem, because regular screening is needed to yield maximal breast cancer mortality reductions.
All of the participants in our study were in age categories for which there are mammography recommendations. It is noteworthy that even though all of the women in our study had insurance covering mammography and were in a plan that actively promoted screening, approximately half were either off schedule or never had a mammogram. This is consistent with the findings of other studies that financial coverage is necessary but not sufficient for mammography use.19
Several provider-related factors were significantly associated with the screening group; off-schedule women were less likely than their on-schedule counterparts but more likely than the never-had group to report having a regular physician, a discussion of mammography with their physicians, or a mammography recommendation from a physician within the past 2 years. The relationship between physician discussion and recommendations could be bidirectional, in that on-schedule women may be more open to discussion or at least perceived by their physicians to be so. They may even be more likely to initiate such discussions. Previous research20 has shown that physician recommendations facilitate adherence. Our data further support the important role of physician discussion and recommendations in repeat adherence. Thus, physicians should continue to reinforce the importance of mammography even for women who have been on schedule.
Although the majority of women in our study knew that mammograms are effective in reducing breast cancer mortality, there were differences by group in knowledge. Women who never had a mammogram were less likely to report that mammograms are effective. Off-schedule women were less knowledgeable than either the on-schedule or never-had groups about how often women should be screened; perhaps this lack of knowledge about when to be rescreened contributes to their being off schedule. In any case, it is important for the physician to remind a woman about the appropriate schedule and to provide a referral.
Off-schedule women were more likely than on-schedule women to be ambivalent about mammography and confused about screening guidelines. Whether these findings can be attributed to the guideline debate of 1997 shortly before our data collection cannot be determined. However, these findings do indicate a need for mammography education about both the rationale for repeat screening and specific information about recommended guidelines.
There is increased interest in evaluating multiple risk behaviors. Our results confirm other findings21-24 that women who are off schedule for mammography are less likely to be adherent for other screening behaviors. Consistent with other studies,21,25-27 we found smokers were less likely to be on schedule for screening mammography. These findings suggest that it may be useful to address multiple screening behaviors rather than focusing on one test at a time.
There were associations between mammography history and variables related to HRT. Consistent with other research,28 off-schedule women were less likely to have ever used or to currently be using HRT. Because it is likely that physicians routinely order mammograms before prescribing HRT, this association may be due more to routine medical procedure than patient characteristics.29 However, whether decisions to use HRT and to have regular mammograms are associated should be explored.
Also consistent with previous findings,7,8 multivariate analyses revealed that younger age, having a CBE within the past 12 months, and physician recommendations were important factors associated with repeat mammography. As previously reported,13 “feeling torn” about mammography and being confused about screening guidelines were negatively associated with being on schedule for mammography.
Limitations
One limitation of our study is that our sample was drawn from women with health insurance rather than from the general population. Thus, we cannot generalize the results of our study to the entire population of North Carolina.
Also, because the sample was drawn for the purposes of a subsequent intervention, there are some other anomalies. We stratified the sample on the basis of age and adherence status, and thus the proportions per se cannot be generalized to the health plan.
Another limitation is that we collected self-report information only on the 2 most recent mammograms. Although long-term mammography history studies should be conducted in the future, ours is one of only a few studies to date that assessed more than 1-time mammography use. Thus, our findings set the stage for future assessments of repeat adherence. Previous research suggests that the correspondence between self-report and mammography use is very high in health maintenance organization settings,30,31 but there is a discrepancy in recall of timing of the mammogram.32 Although we cannot conclusively verify the date of last and previous mammograms, our findings show expected differences between those who reported being on versus off schedule. The 3-month window we allowed before categorizing women as off schedule may have limited misclassification of adherent women as nonadherent. Thus, we probably underestimated the number of women who were off schedule for repeat mammography.
Conclusions
Our study is one of a small number to analyze differences in beliefs and other health-related behaviors among groups of women who are on schedule or off schedule for a mamogram and those who never had mammograms. With a few exceptions, the results suggest a trend, as the off-schedule group almost consistently falls between the on-schedule and never-had groups. For instance, they were more likely than those never screened but less likely than on-schedule women to report the kind of provider support (discussions and recommendations) that facilitates screening and to understand the rationale and recommendations for regular screening. Off-schedule women also showed a need to change other health-related behaviors. Off-schedule women were also likely to perceive their breast cancer risk as lower, be less likely to be up to date with other cancer screening tests, and to have ever used HRT.
Because there are few studies comparing women who are on versus off schedule for their 2 most recent mammograms, we were not sure how, for instance, the off-schedule and the never-had group would compare. Our findings suggest that women who are off schedule are in need of mammography-promoting interventions, including recommendations from and discussion with their health care providers. Because they are more positive and knowledgeable about mammography than never screened women, they may benefit from brief interventions from health care providers that highlight the importance of regular screening.
Significant progress has been made in the proportion of women in the United States who have been screened. Further increases will be dependent not only on motivating women who have never been screened but also in enhancing levels of regular screening. Physicians have a central role to play in facilitating regular screening.
Acknowledgments
Our study was funded by the National Cancer Institute grant #5U19-CA-72099-03. We express our sincere appreciation to Don Bradley, MD, at Blue Cross Blue Shield of North Carolina for his leadership and the many women who are participating in this project. We thank Elizabeth Powell for the preparation of the manuscript. Our manuscript represents the perspective of the authors and not the National Cancer Institute.
Related resources
- National Cancer Institute Cancer information, news on research, funding and treatment recommendations. www.nci.nih.gov
- American Cancer Society News on cancer research. Search function identifies local resources. News on breast and other cancers. Information on ACS research and funding programs. Yearly statistics on incidence of cancer dating back to 1995. www.cancer.org
1. Baker LH. Breast cancer detection demonstration project: five-year summary report. Cancer 2. 1982;32:194-225
2. Shapiro S, Venet W, Strax P, Venet L, Roeser R. Ten-to-fourteen year effect of screening on breast cancer mortality. J Natl Cancer Inst 1982;69:349-55
3. Anonymous. Self-reported use of mammography and insurance status among women aged Ž 40 years—United States, 1991-1992 and 1996-1997. MMWR Morb Mortal Wkly Rep 1998;47:825-30
4. Anonymous. Self-reported use of mammography among women aged Ž40 years—United States, 1989 and 1995. MMWR Morb Mortal Wkly Rep 1997;46:937-41
5. Faulkner LA, Schauffler HH. The effect of health insurance coverage on the appropriate use of recommended clinical preventive services. Am J Prev Med 1997;13:453-58
6. Hahn RA, Teutsch SM, Franks AL, Chang MH, Lloyd EE. The prevalence of risk factors among women in the United States by race and age, 1992-1994: opportunities for primary and secondary prevention. J Am Med Womens Assoc 1998;53:96-104,107.
7. Lerman C, Rimer B, Trock B, Balshem A, Engstrom P. Factors associated with repeat adherence to breast cancer screening. Prev Med 1990;19:279-90
8. Bastani R, Kaplan CP, Maxwell AE, Nisenbaum R, Pearce J, Marcus AC. Initial and repeat mammography screening in a low income population in Los Angeles. Cancer Epidemiol Biomarkers Prev 1995;4:161-71.
9. Burack RC, Gimotty PA. Promoting screening mammography in inner-city settings: the sustained effectiveness of computerized reminders in a randomized controlled trial. Med Care 1997;35:921-31
10. Song L, Fletcher R. Breast cancer rescreening in low-income women. Am J Prev Med 1998;15:128-33
11. Yood MU, McCarthy BD, Lee NC, Jacobsen G, Johnson CC. Patterns and characteristics of repeat mammography among women 50 years and older. Cancer Epidemiol Biomarkers Prev 1999;8:595-99
12. Lipkus IM, Rimer BK, Halabi S, Strigo TS. Can tailored interventions increase mammography use among HMO women? Am J Prev Med 2000;18:1-10
13. Rimer BK, Halabi S, Strigo TS, Crawford Y, Lipkus IM. Confusion about mammography: prevalence and consequences. J Women’s Health Gender-Based Med 1999;8:509-20
14. National Cancer Institute and American Cancer Society. Joint statement on breast cancer screening for women in their 40s. The Cancer Information Service; 1997.
15. Stoddard AM, Rimer BK, Lane D, et al. for the NCI Breast Cancer Consortium. Underusers of mammogram screening: stage of adoption in five US subpopulations. Prev Med 1998;27:478-87
16. Rakowski W, Ehrich B, Golsetin M, et al. A stage-matched intervention for screening mammography. Ann Behav Med 1997;19:S063.-
17. Velicer W, DiClemente C, Prochaska J, et al. A decisional balance measure for assessing and predicting smoking status. J Personality Soc Psychol 1985;48:1279-89
18. Forthofer RN, Lehnen RF. Public program analysis: a new categorical data analysis approach. Belmont: Lifetime Learning Publications; 1981.
19. Rimer BK, Resch N, King E, et al. Multistrategy health education program to increase mammography use among women ages 65 and older. Public Health Rep 1992;107:369-80
20. Skinner, Strecher, Hospers. Physicians’ recommendations for mammography: do tailored messages make a difference? Am J Public Health 1994;84:43-49
21. Ronco G, Segnan N, Ponti A. Who has Pap tests? Variables associated with the use of Pap tests in absence of screening programmes. Int J Epidemiol 1991;20:349-53
22. Rakowski W, Rimer BK, Bryant SA. Integrating behavior and intention regarding mammography by respondents in the 1990 national health interview survey of health promotion and disease prevention. Pub Health Reports 1993;108:605-24
23. Hyman RB, Greewald ES, Hacker S. Smoking, dietary, and breast and cervical cancer screening knowledge and screening practices of employees in an urban medical center. J Cancer Educ 1995;10:82-87
24. Pearlman DN, Rakowski W, Ehrich B. Mammography, clinical breast exam and Pap testing: correlates of combined screening. Am J Prev Med 1996;12:52-64
25. Orleans CT, Rimer BK, Cristinzio S, Keintz MK, Fleisher L. A national survey of older smokers: treatment needs of a growing population. Health Psychol 1991;10:343-51.
26. McBride CM, Curry SJ, Taplin S, Anderman C, Grothaus L. Exploring environmental barriers to participation in mammography screening in an HMO. Can Epidemiol Biomarkers Prev 1993;2:559-605
27. Beaulieu MD, Beland F, Roy D, Falardeau M, Herbert G. Factors determining compliance with screening mammography. Can Med Assoc J 1996;154:1335-43
28. Bastian LA, Couchman GM, Rimer BK, McBride CM, Feaganes JR, Siegler IC. Perceptions of menopausal stage and patterns of hormone replacement therapy use. J Women’s Health 1997;6:467-75
29. Personal communication with Lori Bastian.
30. King ES, Rimer BK, Trock B, Balshem A, Engstrom P. How valid are mammography self-reports? Am J Public Health 1990;80:1386-88
31. Degnan D, Harris R, Ranney J, Quade D, Earp JA, Gonzalez J. Measuring the use of mammography: two methods compared. Am J Public Health 1992;82:1386-88
32. PM, Mickey RM, Worden JK. Reliability of self-reported breast screening information in a survey of lower income women. Prev Med 1997;26:287-91
1. Baker LH. Breast cancer detection demonstration project: five-year summary report. Cancer 2. 1982;32:194-225
2. Shapiro S, Venet W, Strax P, Venet L, Roeser R. Ten-to-fourteen year effect of screening on breast cancer mortality. J Natl Cancer Inst 1982;69:349-55
3. Anonymous. Self-reported use of mammography and insurance status among women aged Ž 40 years—United States, 1991-1992 and 1996-1997. MMWR Morb Mortal Wkly Rep 1998;47:825-30
4. Anonymous. Self-reported use of mammography among women aged Ž40 years—United States, 1989 and 1995. MMWR Morb Mortal Wkly Rep 1997;46:937-41
5. Faulkner LA, Schauffler HH. The effect of health insurance coverage on the appropriate use of recommended clinical preventive services. Am J Prev Med 1997;13:453-58
6. Hahn RA, Teutsch SM, Franks AL, Chang MH, Lloyd EE. The prevalence of risk factors among women in the United States by race and age, 1992-1994: opportunities for primary and secondary prevention. J Am Med Womens Assoc 1998;53:96-104,107.
7. Lerman C, Rimer B, Trock B, Balshem A, Engstrom P. Factors associated with repeat adherence to breast cancer screening. Prev Med 1990;19:279-90
8. Bastani R, Kaplan CP, Maxwell AE, Nisenbaum R, Pearce J, Marcus AC. Initial and repeat mammography screening in a low income population in Los Angeles. Cancer Epidemiol Biomarkers Prev 1995;4:161-71.
9. Burack RC, Gimotty PA. Promoting screening mammography in inner-city settings: the sustained effectiveness of computerized reminders in a randomized controlled trial. Med Care 1997;35:921-31
10. Song L, Fletcher R. Breast cancer rescreening in low-income women. Am J Prev Med 1998;15:128-33
11. Yood MU, McCarthy BD, Lee NC, Jacobsen G, Johnson CC. Patterns and characteristics of repeat mammography among women 50 years and older. Cancer Epidemiol Biomarkers Prev 1999;8:595-99
12. Lipkus IM, Rimer BK, Halabi S, Strigo TS. Can tailored interventions increase mammography use among HMO women? Am J Prev Med 2000;18:1-10
13. Rimer BK, Halabi S, Strigo TS, Crawford Y, Lipkus IM. Confusion about mammography: prevalence and consequences. J Women’s Health Gender-Based Med 1999;8:509-20
14. National Cancer Institute and American Cancer Society. Joint statement on breast cancer screening for women in their 40s. The Cancer Information Service; 1997.
15. Stoddard AM, Rimer BK, Lane D, et al. for the NCI Breast Cancer Consortium. Underusers of mammogram screening: stage of adoption in five US subpopulations. Prev Med 1998;27:478-87
16. Rakowski W, Ehrich B, Golsetin M, et al. A stage-matched intervention for screening mammography. Ann Behav Med 1997;19:S063.-
17. Velicer W, DiClemente C, Prochaska J, et al. A decisional balance measure for assessing and predicting smoking status. J Personality Soc Psychol 1985;48:1279-89
18. Forthofer RN, Lehnen RF. Public program analysis: a new categorical data analysis approach. Belmont: Lifetime Learning Publications; 1981.
19. Rimer BK, Resch N, King E, et al. Multistrategy health education program to increase mammography use among women ages 65 and older. Public Health Rep 1992;107:369-80
20. Skinner, Strecher, Hospers. Physicians’ recommendations for mammography: do tailored messages make a difference? Am J Public Health 1994;84:43-49
21. Ronco G, Segnan N, Ponti A. Who has Pap tests? Variables associated with the use of Pap tests in absence of screening programmes. Int J Epidemiol 1991;20:349-53
22. Rakowski W, Rimer BK, Bryant SA. Integrating behavior and intention regarding mammography by respondents in the 1990 national health interview survey of health promotion and disease prevention. Pub Health Reports 1993;108:605-24
23. Hyman RB, Greewald ES, Hacker S. Smoking, dietary, and breast and cervical cancer screening knowledge and screening practices of employees in an urban medical center. J Cancer Educ 1995;10:82-87
24. Pearlman DN, Rakowski W, Ehrich B. Mammography, clinical breast exam and Pap testing: correlates of combined screening. Am J Prev Med 1996;12:52-64
25. Orleans CT, Rimer BK, Cristinzio S, Keintz MK, Fleisher L. A national survey of older smokers: treatment needs of a growing population. Health Psychol 1991;10:343-51.
26. McBride CM, Curry SJ, Taplin S, Anderman C, Grothaus L. Exploring environmental barriers to participation in mammography screening in an HMO. Can Epidemiol Biomarkers Prev 1993;2:559-605
27. Beaulieu MD, Beland F, Roy D, Falardeau M, Herbert G. Factors determining compliance with screening mammography. Can Med Assoc J 1996;154:1335-43
28. Bastian LA, Couchman GM, Rimer BK, McBride CM, Feaganes JR, Siegler IC. Perceptions of menopausal stage and patterns of hormone replacement therapy use. J Women’s Health 1997;6:467-75
29. Personal communication with Lori Bastian.
30. King ES, Rimer BK, Trock B, Balshem A, Engstrom P. How valid are mammography self-reports? Am J Public Health 1990;80:1386-88
31. Degnan D, Harris R, Ranney J, Quade D, Earp JA, Gonzalez J. Measuring the use of mammography: two methods compared. Am J Public Health 1992;82:1386-88
32. PM, Mickey RM, Worden JK. Reliability of self-reported breast screening information in a survey of lower income women. Prev Med 1997;26:287-91
Osteoporosis Prevention Counseling During Health Maintenance Examinations
METHODS: A total of 449 women aged 18 to 65 years participated in exit interviews immediately following a health maintenance examination at 1 of 8 Wisconsin family practice clinics.
RESULTS: Forty-six percent of these women reported discussing osteoporosis with their providers during their visit, and 51% reported discussing calcium intake. A total of 61% reported discussing either osteoporosis or calcium intake during the visit. Some providers were able to discuss these topics with more than 90% of their patients. A logistic regression model showed that providers were less likely to discuss either of these issues with women younger than 40 years (P=.019); they were more likely to discuss them with women older than 60 years (P=.002) than with women aged 40 to 60 years; and women providers were significantly more likely to discuss either issue (P=.004).
CONCLUSIONS: Primary care providers are in a good position to counsel women of all ages about their potential for avoiding osteoporosis and to recommend prevention strategies. The United States Preventive Services Task Force recommends that all women be counseled on adequate calcium intake yearly after the age of 18 years. Provider education and institutional changes may increase the frequency of this counseling for all primary care physicians.
Osteoporosis is an important cause of age-related mortality and morbidity. More than 1.5 million Americans have osteoporosis-related fractures, costing the United States health care system more than $10 billion annually.1 Although osteoporosis can occur in men, its incidence is higher in women. Those who are postmenopausal are at the highest risk because of the bone loss that occurs with decreasing estrogen levels. The National Osteoporosis Foundation estimates that 21% to 30% of all postmenopausal white women have osteoporosis, and an additional 54% have low bone density.2 Women aged older than 50 years have a 4 in 10 chance of incurring a fracture during their remaining lifetime.3
Bone mineral density measurements can identify women with low bone mass who are at risk for a fracture. Although measurements of bone mineral density may be clinically indicated in high-risk women, current evidence does not support using them as screening modalities.4 Medications can slightly increase bone mass and prevent further loss, but treatment options for osteoporosis are suboptimal. Most consensus recommendations focus on prevention as the best approach.5-7
Prevention of osteoporosis should begin in adolescence with education about risk factors, encouragement of adequate dietary calcium and Vitamin D, exercise, and other healthy behaviors; it must continue throughout a woman’s life.8-10 Several studies have shown that calcium supplementation can increase bone density in women from adolescence to postmenopause.9,11-15 The United States Preventive Services Task Force recommends that all women be counseled on adequate calcium intake yearly after the age of 18 years.11
It is unclear how many primary care providers discuss osteoporosis and calcium intake with women at their annual health maintenance examinations. A review by the lead author16 of 263 charts of women older than 50 years found an overall documented rate of osteoporosis risk assessment of 35%. A vitamin manufacturing company telephone survey of 505 women aged 18 to 65 years found that only 34% had discussed osteoporosis and 44% had discussed calcium intake with their physicians in the past year.17 A 1991 study reviewed 243 medical records of women aged between 40 and 65 years and found documentation that although 74% of the women had 2 or more risk factors for osteoporosis, only 19% had received an osteoporosis-specific intervention (ie, calcium supplementation or counseling about osteoporosis risk or hormone replacement therapy).18 Also, the medical records of only 10% of the women in our prevalence study had a documented assessment of osteoporosis risk.
We used patient exit interviews to assess the frequency of osteoporosis prevention counseling during women’s health maintenance examinations in a primary care setting. On the basis of our literature review, we hypothesized that woman providers would discuss calcium intake and osteoporosis prevention more often than men and that discussions would be more frequent as women aged. We also hypothesized that physician assistants and nurse practitioners would discuss prevention topics more frequently than physicians.
Methods
For our report of osteoporosis prevention in primary care we used data collected in 8 Wisconsin family practice clinics at the outset of an osteoporosis intervention pilot study. Seven of the clinics were residency training sites, and the eighth was a university-affiliated private practice. All faculty physicians, physician assistants, nurse practitioners, and second-year residents in each clinic were invited to participate. Of the 67 providers invited, 90% agreed to be part of the study. Each provider completed a brief questionnaire that included demographic information such as age, sex, and job title; adequacy of own calcium intake; and personal experience with osteoporosis. Each participating provider signed a consent form agreeing to allow a study invitation to be given to women patients registering for a well-woman visit. The providers knew the general content of the interviews but not the specific questions. There was a 1- to 2-month time lag from when the providers signed the consent form to the beginning of the interviews. Data collection was completed during a period of 2 to 3 months, and providers were not told on which days it would occur. The University of Wisconsin Human Subjects Committee approved the study protocol.
A researcher approached women patients aged 18 to 65 years as they were registering for an annual health maintenance examination with a participating provider and asked each if she would be willing to answer questions about her health care after her examination. No interview content was specified at the time of the request. After her appointment each woman who agreed to take part in the study completed a 5-minute interview about any discussion she had with her provider about calcium intake and osteoporosis risk factors and prevention. The following questions were part of the protocol: “Did your provider talk to you about osteoporosis today?” and “Did he/she talk to you about your calcium intake?” Each woman provided demographic information, her personal history of osteoporosis, smoking status, menopause status, exercise history, and any estrogen prescriptions Table 1. Each woman received $5 after completing the interview. If a woman was not able to complete the interview immediately after her appointment, she was called at home that evening or the following day. Our goal was to interview 5 to 10 women per provider.
Descriptive statistics were compiled to provide an overall summary of the data as well as a summary by provider and clinic. Pearson correlation coefficients were calculated to test the degree of hypothesized association between patient age and provider-patient discussion of either calcium intake or osteoporosis prevention. The chi-square test for independence was used to assess whether physician sex was associated with an increase in frequency of discussion. We used a logistic regression model to test the hypotheses that these prevention discussions about calcium intake or osteoporosis prevention occurred more often when a woman or midlevel provider was seen or if the patient was older, while controlling for other health factors (osteoporosis, race, and menopausal status).
Results
A total of 449 women were interviewed at the 8 clinics, an average of 7.5 per provider. The consent rate for patients approached for an interview was 90.4%. The average patient age was 40 years, with a range of 18 to 65 years. Three hundred and eighty-five (91%) of the women were white and only 24 (3%) said they had osteoporosis. Forty-six percent of the women interviewed reported discussing osteoporosis with their providers, and 54% reported discussing calcium intake during the visit. Overall, 61% reported a discussion of either calcium intake or osteoporosis. Two of the 8 clinics had significantly higher rates of either osteoporosis or calcium discussions (89% and 92%, respectively), while the discussion rate for the other 6 clinics was approximately 50%.
The providers included 37 faculty physicians, 15 second-year family practice resident physicians, and 8 nurse practitioners and physician assistants. The mean age of the providers was 42 years (mean=46 years for men and 38 years for women). The group was evenly split between men and women. Sixty-two percent of the providers were faculty, 25% residents, and 13% nurse practitioners or physician assistants.
Table 2 shows the significant increase in provider-patient discussion about osteoporosis and calcium intake as patient age increases. Regardless of provider sex, women of all ages reported discussions of calcium supplementation more often than discussion of osteoporosis risk only. Younger women reported conversations with their providers about osteoporosis approximately one third of the time during a health maintenance visit. This increased to 50% for women in their 40s and to more than 60% for women in their 50s and 60s. Almost half of the younger women reported discussing the importance of calcium intake, while women older than 60 years reported discussions of calcium in more than 60% of the interviews.
The logistic regression model presented in Table 3 shows the odds of the hypothesized variables having an influence on the outcome of a woman having a discussion of either calcium or osteoporosis with her provider. It shows that patient age is significant, with women younger than 40 years reporting these discussions half as often as the total patient group. Women in their 40s continued to be less likely to have prevention discussions, while women in their 60s were significantly more likely than those aged 40 to 60 years to have talked with their providers about topics important for osteoporosis prevention. The model also shows that ethnicity, smoking status, amount of exercise, menopause status, and patient history of osteoporosis were not related to occurrence of these discussions. Provider sex was significant, however, with women much more likely to discuss either calcium intake or osteoporosis prevention than men (P=.004). Men discussed these topics in only 24% of visits with women younger than 40 years, increasing to more than 40% when their patients reached their 40s, 53% for women in their 50s, and 62% of women in their 60s. A separate logistics regression model showed that provider dietary calcium intake and calcium supplements were not associated with differences in counseling rates. Provider personal experience with osteoporosis showed a borderline significant association with lower rates of counseling (P=.04).
Menopausal status was also associated with an increase in counseling rates, likely related to age. Two thirds of the menopausal women in this study were taking estrogen. Estrogen use was not associated with a change in counseling. Nurse practitioners and physician assistants were more likely to discuss osteoporosis prevention than their physician colleagues, although this difference was not statistically significant. A separate logistic regression, which added clinic site to the model, provided no additional explanation of the frequency of osteoporosis risk prevention discussion. Women at 2 of the clinics were significantly more likely to report discussions with their providers about either osteoporosis or calcium, but in those 2 clinics both patient age and provider sex remained significant.
Discussion
The 61% overall rate of osteoporosis and calcium discussions in our study is higher than rates documented in other studies.16-8 Since this is the only study to interview women immediately after a health maintenance visit, the accuracy of patient recall may be improved. Many providers discussed adequate calcium intake without specifically discussing osteoporosis, although the opposite was not true. It may be better to discuss a behavioral change than the risk of the disease without discussion of how to prevent it.
Woman providers discussed calcium and osteoporosis significantly more than men, which supports findings from a chart review study published by the lead author16 on the same topic. It has been well documented in the literature that women physicians provide more health screening, such as Papanicolaou tests and mammograms, than men.19,20 This is the first study to address osteoporosis prevention topics.
Provider personal experience with osteoporosis was associated with a lower rate of counseling about calcium intake and prevention strategies. This association was of borderline significance and deserves further study.
Women of older age groups more often reported discussion of osteoporosis during a health maintenance visit; however, age was not as well correlated with discussions of calcium intake. Although it is commendable that providers increased their attention to osteoporosis in older women, counseling young women is also essential to prevent osteoporosis. Adequate dietary calcium and risk reduction for osteoporosis through diet and exercise may provide young women with increased protection from osteoporosis before entering their menopausal years.
Two of the clinics in our study showed significantly higher rates of discussion of calcium and osteoporosis. These residency clinics were smaller than many of the more urban sites, and each had a faculty member who was very interested in women’s health. The percentage of woman providers is higher in one of these clinics, which may explain in part the increased rates of osteoporosis and calcium discussions in this particular clinic, but it cannot account for the other clinic, where only 33% of the providers were women. The high rates of osteoporosis prevention counseling in these 2 clinics imply that there are some institutional changes that can be implemented to improve rates of counseling in other clinics.
Limitations
There are several limitations to our study. We used patient reports to describe what occurred during a health maintenance examination but did not corroborate this data with chart reviews. A chart review done in some of the same clinics found a much lower recorded incidence of osteoporosis or calcium discussions,16 so patient reports may be a better measure. We did not collect information regarding length of visit. All of the providers routinely see women for annual examinations during a 30-minute time slot, although occasional variation may occur. We also did not collect information about each woman’s current calcium intake or use of hormone replacement therapy. Also, since with one exception this study was completed at academic practices, it may not reflect practices outside of an academic setting.
Conclusions
Major barriers to osteoporosis prevention include time constraints and competing issues brought to the visit by both the patient and the care provider. The limited time during health maintenance visits does not allow providers to address every prevention topic. Some providers feel osteoporosis is not as important an issue as tobacco smoking, cancer prevention, exercise, or diet. As a result, inclusion of osteoporosis in a universal primary care prevention agenda is currently controversial. However, according to the National Osteoporosis Foundation, a vast majority of postmenopausal white women have osteoporosis or low bone density, and others report 4 of every 10 women older than 50 years will fracture a bone over the course of their remaining lifetime.3 Because of this high prevalence, osteoporosis should be a priority prevention topic for women’s health providers.
Primary care providers are in a good position to counsel women about osteoporosis risk factors and prevention strategies. Provider education along with changes in the roles and responsibilities of staff to provide services may increase the frequency of this counseling. Further study should examine ways for primary care providers to consistently implement osteoporosis prevention.
Acknowledgments
The University of Wisconsin Department of Family Medicine Research Program provided funding for our study.
1. Riggs BL, Melton LJ. The prevention and treatment of osteoporosis. N Engl J Med 1992;327:620-27.
2. National Osteoporosis Foundation. Osteoporosis: review of the evidence for prevention, diagnosis and treatment and cost effectiveness analysis. Osteoporos Int 1998;8:S1-S88.
3. Lips P. Epidemiology and predictors of fractures associated with osteoporosis. Am J Med 1997;103:3S-8S;discussion 8S-11S.
4. Melton LJ, Eddy DM, Johnston CC. Screening for osteoporosis. Ann Intern Med 1990;112:516-28.
5. Consensus Development Statement. Who are candidates for prevention and treatment for osteoporosis? Osteoporos Int 1997;7:1-6.
6. Clinical practice guidelines for the diagnosis and management of osteoporosis. Can Med Assoc J 1996;155:1113-29.
7. Matkovic V, Ilich JZ, Skugor M, Saracoglu M. Primary prevention of osteoporosis. Phys Med Rehab Clin North Am 1995;6:595-627.
8. Nelson DA. An anthropological perspective on optimizing calcium consumption for the prevention of osteoporosis. Osteoporos Int 1996;6:325-28.
9. Recker RR, Davies KM, Hinders SM, Heaney RP, Stegman MR, Kimmel DB. Bone gain in young adult women. JAMA 1992;268:2403-08.
10. Nowson CA, Green RM, Hopper JL, et al. A co-twin study of the effect of calcium supplementation on bone density during adolescence. Osteoporos Int 1997;7:219-25.
11. US Preventive Services Task Force. Guide to clinical preventive services. Baltimore, Md: Williams and Wilkins; 1996.
12. Lloyd T, Martel JK, Rollings N, et al. The effect of calcium supplementation and tanner stage on bone density, content and area in teenage women. Osteoporos Int 1996;6:276-83.
13. Kanders B, Dempster DAW, Lindsay R. Interaction of calcium nutrition and physical activity on bone mass in young women. J Bone Miner Res 1988;3:145-49.
14. Dawson-Hughes B, Dallal GE, Krall EA, Sadowski L, Sahyoun N, Tannenbaum S. A controlled trial of the effect of calcium supplementation on bone density in postmenopausal women. N Engl J Med 1990;323:878-83.
15. Devine A, Dick IM, Heal SJ, Criddle RA, Prince RL. A 4-year follow-up study of the effects of calcium supplementation on bone density in elderly postmenopausal women. Osteoporos Int 1997;7:23-28.
16. Schrager S, Kausch T, Bobula JA. Osteoporosis risk assessment by family practice faculty and residents: a chart review. Wis Med J 1999;98:34-36.
17. Citracal osteoporosis prevention survey conducted by Opinion Research Corporation International for Mission Pharmacal; 1996.
18. Bourguet CC, Hamrick GA, Gilchrist VJ. The prevalence of osteoporosis risk factors and physician intervention. J Fam Pract 1991;32:265-71.
19. Lurie N, Slater J, McGovern P, Ekstrum J, Quam L, Margolis K. Preventive care for women: does the sex of the physician matter? N Engl J Med 1993;329:478-82.
20. Cassard SD, Weisman CS, Plichta SB, Johnson TL. Physician gender and women’s preventive services. J Women’s Health 1997;6:199-207.
METHODS: A total of 449 women aged 18 to 65 years participated in exit interviews immediately following a health maintenance examination at 1 of 8 Wisconsin family practice clinics.
RESULTS: Forty-six percent of these women reported discussing osteoporosis with their providers during their visit, and 51% reported discussing calcium intake. A total of 61% reported discussing either osteoporosis or calcium intake during the visit. Some providers were able to discuss these topics with more than 90% of their patients. A logistic regression model showed that providers were less likely to discuss either of these issues with women younger than 40 years (P=.019); they were more likely to discuss them with women older than 60 years (P=.002) than with women aged 40 to 60 years; and women providers were significantly more likely to discuss either issue (P=.004).
CONCLUSIONS: Primary care providers are in a good position to counsel women of all ages about their potential for avoiding osteoporosis and to recommend prevention strategies. The United States Preventive Services Task Force recommends that all women be counseled on adequate calcium intake yearly after the age of 18 years. Provider education and institutional changes may increase the frequency of this counseling for all primary care physicians.
Osteoporosis is an important cause of age-related mortality and morbidity. More than 1.5 million Americans have osteoporosis-related fractures, costing the United States health care system more than $10 billion annually.1 Although osteoporosis can occur in men, its incidence is higher in women. Those who are postmenopausal are at the highest risk because of the bone loss that occurs with decreasing estrogen levels. The National Osteoporosis Foundation estimates that 21% to 30% of all postmenopausal white women have osteoporosis, and an additional 54% have low bone density.2 Women aged older than 50 years have a 4 in 10 chance of incurring a fracture during their remaining lifetime.3
Bone mineral density measurements can identify women with low bone mass who are at risk for a fracture. Although measurements of bone mineral density may be clinically indicated in high-risk women, current evidence does not support using them as screening modalities.4 Medications can slightly increase bone mass and prevent further loss, but treatment options for osteoporosis are suboptimal. Most consensus recommendations focus on prevention as the best approach.5-7
Prevention of osteoporosis should begin in adolescence with education about risk factors, encouragement of adequate dietary calcium and Vitamin D, exercise, and other healthy behaviors; it must continue throughout a woman’s life.8-10 Several studies have shown that calcium supplementation can increase bone density in women from adolescence to postmenopause.9,11-15 The United States Preventive Services Task Force recommends that all women be counseled on adequate calcium intake yearly after the age of 18 years.11
It is unclear how many primary care providers discuss osteoporosis and calcium intake with women at their annual health maintenance examinations. A review by the lead author16 of 263 charts of women older than 50 years found an overall documented rate of osteoporosis risk assessment of 35%. A vitamin manufacturing company telephone survey of 505 women aged 18 to 65 years found that only 34% had discussed osteoporosis and 44% had discussed calcium intake with their physicians in the past year.17 A 1991 study reviewed 243 medical records of women aged between 40 and 65 years and found documentation that although 74% of the women had 2 or more risk factors for osteoporosis, only 19% had received an osteoporosis-specific intervention (ie, calcium supplementation or counseling about osteoporosis risk or hormone replacement therapy).18 Also, the medical records of only 10% of the women in our prevalence study had a documented assessment of osteoporosis risk.
We used patient exit interviews to assess the frequency of osteoporosis prevention counseling during women’s health maintenance examinations in a primary care setting. On the basis of our literature review, we hypothesized that woman providers would discuss calcium intake and osteoporosis prevention more often than men and that discussions would be more frequent as women aged. We also hypothesized that physician assistants and nurse practitioners would discuss prevention topics more frequently than physicians.
Methods
For our report of osteoporosis prevention in primary care we used data collected in 8 Wisconsin family practice clinics at the outset of an osteoporosis intervention pilot study. Seven of the clinics were residency training sites, and the eighth was a university-affiliated private practice. All faculty physicians, physician assistants, nurse practitioners, and second-year residents in each clinic were invited to participate. Of the 67 providers invited, 90% agreed to be part of the study. Each provider completed a brief questionnaire that included demographic information such as age, sex, and job title; adequacy of own calcium intake; and personal experience with osteoporosis. Each participating provider signed a consent form agreeing to allow a study invitation to be given to women patients registering for a well-woman visit. The providers knew the general content of the interviews but not the specific questions. There was a 1- to 2-month time lag from when the providers signed the consent form to the beginning of the interviews. Data collection was completed during a period of 2 to 3 months, and providers were not told on which days it would occur. The University of Wisconsin Human Subjects Committee approved the study protocol.
A researcher approached women patients aged 18 to 65 years as they were registering for an annual health maintenance examination with a participating provider and asked each if she would be willing to answer questions about her health care after her examination. No interview content was specified at the time of the request. After her appointment each woman who agreed to take part in the study completed a 5-minute interview about any discussion she had with her provider about calcium intake and osteoporosis risk factors and prevention. The following questions were part of the protocol: “Did your provider talk to you about osteoporosis today?” and “Did he/she talk to you about your calcium intake?” Each woman provided demographic information, her personal history of osteoporosis, smoking status, menopause status, exercise history, and any estrogen prescriptions Table 1. Each woman received $5 after completing the interview. If a woman was not able to complete the interview immediately after her appointment, she was called at home that evening or the following day. Our goal was to interview 5 to 10 women per provider.
Descriptive statistics were compiled to provide an overall summary of the data as well as a summary by provider and clinic. Pearson correlation coefficients were calculated to test the degree of hypothesized association between patient age and provider-patient discussion of either calcium intake or osteoporosis prevention. The chi-square test for independence was used to assess whether physician sex was associated with an increase in frequency of discussion. We used a logistic regression model to test the hypotheses that these prevention discussions about calcium intake or osteoporosis prevention occurred more often when a woman or midlevel provider was seen or if the patient was older, while controlling for other health factors (osteoporosis, race, and menopausal status).
Results
A total of 449 women were interviewed at the 8 clinics, an average of 7.5 per provider. The consent rate for patients approached for an interview was 90.4%. The average patient age was 40 years, with a range of 18 to 65 years. Three hundred and eighty-five (91%) of the women were white and only 24 (3%) said they had osteoporosis. Forty-six percent of the women interviewed reported discussing osteoporosis with their providers, and 54% reported discussing calcium intake during the visit. Overall, 61% reported a discussion of either calcium intake or osteoporosis. Two of the 8 clinics had significantly higher rates of either osteoporosis or calcium discussions (89% and 92%, respectively), while the discussion rate for the other 6 clinics was approximately 50%.
The providers included 37 faculty physicians, 15 second-year family practice resident physicians, and 8 nurse practitioners and physician assistants. The mean age of the providers was 42 years (mean=46 years for men and 38 years for women). The group was evenly split between men and women. Sixty-two percent of the providers were faculty, 25% residents, and 13% nurse practitioners or physician assistants.
Table 2 shows the significant increase in provider-patient discussion about osteoporosis and calcium intake as patient age increases. Regardless of provider sex, women of all ages reported discussions of calcium supplementation more often than discussion of osteoporosis risk only. Younger women reported conversations with their providers about osteoporosis approximately one third of the time during a health maintenance visit. This increased to 50% for women in their 40s and to more than 60% for women in their 50s and 60s. Almost half of the younger women reported discussing the importance of calcium intake, while women older than 60 years reported discussions of calcium in more than 60% of the interviews.
The logistic regression model presented in Table 3 shows the odds of the hypothesized variables having an influence on the outcome of a woman having a discussion of either calcium or osteoporosis with her provider. It shows that patient age is significant, with women younger than 40 years reporting these discussions half as often as the total patient group. Women in their 40s continued to be less likely to have prevention discussions, while women in their 60s were significantly more likely than those aged 40 to 60 years to have talked with their providers about topics important for osteoporosis prevention. The model also shows that ethnicity, smoking status, amount of exercise, menopause status, and patient history of osteoporosis were not related to occurrence of these discussions. Provider sex was significant, however, with women much more likely to discuss either calcium intake or osteoporosis prevention than men (P=.004). Men discussed these topics in only 24% of visits with women younger than 40 years, increasing to more than 40% when their patients reached their 40s, 53% for women in their 50s, and 62% of women in their 60s. A separate logistics regression model showed that provider dietary calcium intake and calcium supplements were not associated with differences in counseling rates. Provider personal experience with osteoporosis showed a borderline significant association with lower rates of counseling (P=.04).
Menopausal status was also associated with an increase in counseling rates, likely related to age. Two thirds of the menopausal women in this study were taking estrogen. Estrogen use was not associated with a change in counseling. Nurse practitioners and physician assistants were more likely to discuss osteoporosis prevention than their physician colleagues, although this difference was not statistically significant. A separate logistic regression, which added clinic site to the model, provided no additional explanation of the frequency of osteoporosis risk prevention discussion. Women at 2 of the clinics were significantly more likely to report discussions with their providers about either osteoporosis or calcium, but in those 2 clinics both patient age and provider sex remained significant.
Discussion
The 61% overall rate of osteoporosis and calcium discussions in our study is higher than rates documented in other studies.16-8 Since this is the only study to interview women immediately after a health maintenance visit, the accuracy of patient recall may be improved. Many providers discussed adequate calcium intake without specifically discussing osteoporosis, although the opposite was not true. It may be better to discuss a behavioral change than the risk of the disease without discussion of how to prevent it.
Woman providers discussed calcium and osteoporosis significantly more than men, which supports findings from a chart review study published by the lead author16 on the same topic. It has been well documented in the literature that women physicians provide more health screening, such as Papanicolaou tests and mammograms, than men.19,20 This is the first study to address osteoporosis prevention topics.
Provider personal experience with osteoporosis was associated with a lower rate of counseling about calcium intake and prevention strategies. This association was of borderline significance and deserves further study.
Women of older age groups more often reported discussion of osteoporosis during a health maintenance visit; however, age was not as well correlated with discussions of calcium intake. Although it is commendable that providers increased their attention to osteoporosis in older women, counseling young women is also essential to prevent osteoporosis. Adequate dietary calcium and risk reduction for osteoporosis through diet and exercise may provide young women with increased protection from osteoporosis before entering their menopausal years.
Two of the clinics in our study showed significantly higher rates of discussion of calcium and osteoporosis. These residency clinics were smaller than many of the more urban sites, and each had a faculty member who was very interested in women’s health. The percentage of woman providers is higher in one of these clinics, which may explain in part the increased rates of osteoporosis and calcium discussions in this particular clinic, but it cannot account for the other clinic, where only 33% of the providers were women. The high rates of osteoporosis prevention counseling in these 2 clinics imply that there are some institutional changes that can be implemented to improve rates of counseling in other clinics.
Limitations
There are several limitations to our study. We used patient reports to describe what occurred during a health maintenance examination but did not corroborate this data with chart reviews. A chart review done in some of the same clinics found a much lower recorded incidence of osteoporosis or calcium discussions,16 so patient reports may be a better measure. We did not collect information regarding length of visit. All of the providers routinely see women for annual examinations during a 30-minute time slot, although occasional variation may occur. We also did not collect information about each woman’s current calcium intake or use of hormone replacement therapy. Also, since with one exception this study was completed at academic practices, it may not reflect practices outside of an academic setting.
Conclusions
Major barriers to osteoporosis prevention include time constraints and competing issues brought to the visit by both the patient and the care provider. The limited time during health maintenance visits does not allow providers to address every prevention topic. Some providers feel osteoporosis is not as important an issue as tobacco smoking, cancer prevention, exercise, or diet. As a result, inclusion of osteoporosis in a universal primary care prevention agenda is currently controversial. However, according to the National Osteoporosis Foundation, a vast majority of postmenopausal white women have osteoporosis or low bone density, and others report 4 of every 10 women older than 50 years will fracture a bone over the course of their remaining lifetime.3 Because of this high prevalence, osteoporosis should be a priority prevention topic for women’s health providers.
Primary care providers are in a good position to counsel women about osteoporosis risk factors and prevention strategies. Provider education along with changes in the roles and responsibilities of staff to provide services may increase the frequency of this counseling. Further study should examine ways for primary care providers to consistently implement osteoporosis prevention.
Acknowledgments
The University of Wisconsin Department of Family Medicine Research Program provided funding for our study.
METHODS: A total of 449 women aged 18 to 65 years participated in exit interviews immediately following a health maintenance examination at 1 of 8 Wisconsin family practice clinics.
RESULTS: Forty-six percent of these women reported discussing osteoporosis with their providers during their visit, and 51% reported discussing calcium intake. A total of 61% reported discussing either osteoporosis or calcium intake during the visit. Some providers were able to discuss these topics with more than 90% of their patients. A logistic regression model showed that providers were less likely to discuss either of these issues with women younger than 40 years (P=.019); they were more likely to discuss them with women older than 60 years (P=.002) than with women aged 40 to 60 years; and women providers were significantly more likely to discuss either issue (P=.004).
CONCLUSIONS: Primary care providers are in a good position to counsel women of all ages about their potential for avoiding osteoporosis and to recommend prevention strategies. The United States Preventive Services Task Force recommends that all women be counseled on adequate calcium intake yearly after the age of 18 years. Provider education and institutional changes may increase the frequency of this counseling for all primary care physicians.
Osteoporosis is an important cause of age-related mortality and morbidity. More than 1.5 million Americans have osteoporosis-related fractures, costing the United States health care system more than $10 billion annually.1 Although osteoporosis can occur in men, its incidence is higher in women. Those who are postmenopausal are at the highest risk because of the bone loss that occurs with decreasing estrogen levels. The National Osteoporosis Foundation estimates that 21% to 30% of all postmenopausal white women have osteoporosis, and an additional 54% have low bone density.2 Women aged older than 50 years have a 4 in 10 chance of incurring a fracture during their remaining lifetime.3
Bone mineral density measurements can identify women with low bone mass who are at risk for a fracture. Although measurements of bone mineral density may be clinically indicated in high-risk women, current evidence does not support using them as screening modalities.4 Medications can slightly increase bone mass and prevent further loss, but treatment options for osteoporosis are suboptimal. Most consensus recommendations focus on prevention as the best approach.5-7
Prevention of osteoporosis should begin in adolescence with education about risk factors, encouragement of adequate dietary calcium and Vitamin D, exercise, and other healthy behaviors; it must continue throughout a woman’s life.8-10 Several studies have shown that calcium supplementation can increase bone density in women from adolescence to postmenopause.9,11-15 The United States Preventive Services Task Force recommends that all women be counseled on adequate calcium intake yearly after the age of 18 years.11
It is unclear how many primary care providers discuss osteoporosis and calcium intake with women at their annual health maintenance examinations. A review by the lead author16 of 263 charts of women older than 50 years found an overall documented rate of osteoporosis risk assessment of 35%. A vitamin manufacturing company telephone survey of 505 women aged 18 to 65 years found that only 34% had discussed osteoporosis and 44% had discussed calcium intake with their physicians in the past year.17 A 1991 study reviewed 243 medical records of women aged between 40 and 65 years and found documentation that although 74% of the women had 2 or more risk factors for osteoporosis, only 19% had received an osteoporosis-specific intervention (ie, calcium supplementation or counseling about osteoporosis risk or hormone replacement therapy).18 Also, the medical records of only 10% of the women in our prevalence study had a documented assessment of osteoporosis risk.
We used patient exit interviews to assess the frequency of osteoporosis prevention counseling during women’s health maintenance examinations in a primary care setting. On the basis of our literature review, we hypothesized that woman providers would discuss calcium intake and osteoporosis prevention more often than men and that discussions would be more frequent as women aged. We also hypothesized that physician assistants and nurse practitioners would discuss prevention topics more frequently than physicians.
Methods
For our report of osteoporosis prevention in primary care we used data collected in 8 Wisconsin family practice clinics at the outset of an osteoporosis intervention pilot study. Seven of the clinics were residency training sites, and the eighth was a university-affiliated private practice. All faculty physicians, physician assistants, nurse practitioners, and second-year residents in each clinic were invited to participate. Of the 67 providers invited, 90% agreed to be part of the study. Each provider completed a brief questionnaire that included demographic information such as age, sex, and job title; adequacy of own calcium intake; and personal experience with osteoporosis. Each participating provider signed a consent form agreeing to allow a study invitation to be given to women patients registering for a well-woman visit. The providers knew the general content of the interviews but not the specific questions. There was a 1- to 2-month time lag from when the providers signed the consent form to the beginning of the interviews. Data collection was completed during a period of 2 to 3 months, and providers were not told on which days it would occur. The University of Wisconsin Human Subjects Committee approved the study protocol.
A researcher approached women patients aged 18 to 65 years as they were registering for an annual health maintenance examination with a participating provider and asked each if she would be willing to answer questions about her health care after her examination. No interview content was specified at the time of the request. After her appointment each woman who agreed to take part in the study completed a 5-minute interview about any discussion she had with her provider about calcium intake and osteoporosis risk factors and prevention. The following questions were part of the protocol: “Did your provider talk to you about osteoporosis today?” and “Did he/she talk to you about your calcium intake?” Each woman provided demographic information, her personal history of osteoporosis, smoking status, menopause status, exercise history, and any estrogen prescriptions Table 1. Each woman received $5 after completing the interview. If a woman was not able to complete the interview immediately after her appointment, she was called at home that evening or the following day. Our goal was to interview 5 to 10 women per provider.
Descriptive statistics were compiled to provide an overall summary of the data as well as a summary by provider and clinic. Pearson correlation coefficients were calculated to test the degree of hypothesized association between patient age and provider-patient discussion of either calcium intake or osteoporosis prevention. The chi-square test for independence was used to assess whether physician sex was associated with an increase in frequency of discussion. We used a logistic regression model to test the hypotheses that these prevention discussions about calcium intake or osteoporosis prevention occurred more often when a woman or midlevel provider was seen or if the patient was older, while controlling for other health factors (osteoporosis, race, and menopausal status).
Results
A total of 449 women were interviewed at the 8 clinics, an average of 7.5 per provider. The consent rate for patients approached for an interview was 90.4%. The average patient age was 40 years, with a range of 18 to 65 years. Three hundred and eighty-five (91%) of the women were white and only 24 (3%) said they had osteoporosis. Forty-six percent of the women interviewed reported discussing osteoporosis with their providers, and 54% reported discussing calcium intake during the visit. Overall, 61% reported a discussion of either calcium intake or osteoporosis. Two of the 8 clinics had significantly higher rates of either osteoporosis or calcium discussions (89% and 92%, respectively), while the discussion rate for the other 6 clinics was approximately 50%.
The providers included 37 faculty physicians, 15 second-year family practice resident physicians, and 8 nurse practitioners and physician assistants. The mean age of the providers was 42 years (mean=46 years for men and 38 years for women). The group was evenly split between men and women. Sixty-two percent of the providers were faculty, 25% residents, and 13% nurse practitioners or physician assistants.
Table 2 shows the significant increase in provider-patient discussion about osteoporosis and calcium intake as patient age increases. Regardless of provider sex, women of all ages reported discussions of calcium supplementation more often than discussion of osteoporosis risk only. Younger women reported conversations with their providers about osteoporosis approximately one third of the time during a health maintenance visit. This increased to 50% for women in their 40s and to more than 60% for women in their 50s and 60s. Almost half of the younger women reported discussing the importance of calcium intake, while women older than 60 years reported discussions of calcium in more than 60% of the interviews.
The logistic regression model presented in Table 3 shows the odds of the hypothesized variables having an influence on the outcome of a woman having a discussion of either calcium or osteoporosis with her provider. It shows that patient age is significant, with women younger than 40 years reporting these discussions half as often as the total patient group. Women in their 40s continued to be less likely to have prevention discussions, while women in their 60s were significantly more likely than those aged 40 to 60 years to have talked with their providers about topics important for osteoporosis prevention. The model also shows that ethnicity, smoking status, amount of exercise, menopause status, and patient history of osteoporosis were not related to occurrence of these discussions. Provider sex was significant, however, with women much more likely to discuss either calcium intake or osteoporosis prevention than men (P=.004). Men discussed these topics in only 24% of visits with women younger than 40 years, increasing to more than 40% when their patients reached their 40s, 53% for women in their 50s, and 62% of women in their 60s. A separate logistics regression model showed that provider dietary calcium intake and calcium supplements were not associated with differences in counseling rates. Provider personal experience with osteoporosis showed a borderline significant association with lower rates of counseling (P=.04).
Menopausal status was also associated with an increase in counseling rates, likely related to age. Two thirds of the menopausal women in this study were taking estrogen. Estrogen use was not associated with a change in counseling. Nurse practitioners and physician assistants were more likely to discuss osteoporosis prevention than their physician colleagues, although this difference was not statistically significant. A separate logistic regression, which added clinic site to the model, provided no additional explanation of the frequency of osteoporosis risk prevention discussion. Women at 2 of the clinics were significantly more likely to report discussions with their providers about either osteoporosis or calcium, but in those 2 clinics both patient age and provider sex remained significant.
Discussion
The 61% overall rate of osteoporosis and calcium discussions in our study is higher than rates documented in other studies.16-8 Since this is the only study to interview women immediately after a health maintenance visit, the accuracy of patient recall may be improved. Many providers discussed adequate calcium intake without specifically discussing osteoporosis, although the opposite was not true. It may be better to discuss a behavioral change than the risk of the disease without discussion of how to prevent it.
Woman providers discussed calcium and osteoporosis significantly more than men, which supports findings from a chart review study published by the lead author16 on the same topic. It has been well documented in the literature that women physicians provide more health screening, such as Papanicolaou tests and mammograms, than men.19,20 This is the first study to address osteoporosis prevention topics.
Provider personal experience with osteoporosis was associated with a lower rate of counseling about calcium intake and prevention strategies. This association was of borderline significance and deserves further study.
Women of older age groups more often reported discussion of osteoporosis during a health maintenance visit; however, age was not as well correlated with discussions of calcium intake. Although it is commendable that providers increased their attention to osteoporosis in older women, counseling young women is also essential to prevent osteoporosis. Adequate dietary calcium and risk reduction for osteoporosis through diet and exercise may provide young women with increased protection from osteoporosis before entering their menopausal years.
Two of the clinics in our study showed significantly higher rates of discussion of calcium and osteoporosis. These residency clinics were smaller than many of the more urban sites, and each had a faculty member who was very interested in women’s health. The percentage of woman providers is higher in one of these clinics, which may explain in part the increased rates of osteoporosis and calcium discussions in this particular clinic, but it cannot account for the other clinic, where only 33% of the providers were women. The high rates of osteoporosis prevention counseling in these 2 clinics imply that there are some institutional changes that can be implemented to improve rates of counseling in other clinics.
Limitations
There are several limitations to our study. We used patient reports to describe what occurred during a health maintenance examination but did not corroborate this data with chart reviews. A chart review done in some of the same clinics found a much lower recorded incidence of osteoporosis or calcium discussions,16 so patient reports may be a better measure. We did not collect information regarding length of visit. All of the providers routinely see women for annual examinations during a 30-minute time slot, although occasional variation may occur. We also did not collect information about each woman’s current calcium intake or use of hormone replacement therapy. Also, since with one exception this study was completed at academic practices, it may not reflect practices outside of an academic setting.
Conclusions
Major barriers to osteoporosis prevention include time constraints and competing issues brought to the visit by both the patient and the care provider. The limited time during health maintenance visits does not allow providers to address every prevention topic. Some providers feel osteoporosis is not as important an issue as tobacco smoking, cancer prevention, exercise, or diet. As a result, inclusion of osteoporosis in a universal primary care prevention agenda is currently controversial. However, according to the National Osteoporosis Foundation, a vast majority of postmenopausal white women have osteoporosis or low bone density, and others report 4 of every 10 women older than 50 years will fracture a bone over the course of their remaining lifetime.3 Because of this high prevalence, osteoporosis should be a priority prevention topic for women’s health providers.
Primary care providers are in a good position to counsel women about osteoporosis risk factors and prevention strategies. Provider education along with changes in the roles and responsibilities of staff to provide services may increase the frequency of this counseling. Further study should examine ways for primary care providers to consistently implement osteoporosis prevention.
Acknowledgments
The University of Wisconsin Department of Family Medicine Research Program provided funding for our study.
1. Riggs BL, Melton LJ. The prevention and treatment of osteoporosis. N Engl J Med 1992;327:620-27.
2. National Osteoporosis Foundation. Osteoporosis: review of the evidence for prevention, diagnosis and treatment and cost effectiveness analysis. Osteoporos Int 1998;8:S1-S88.
3. Lips P. Epidemiology and predictors of fractures associated with osteoporosis. Am J Med 1997;103:3S-8S;discussion 8S-11S.
4. Melton LJ, Eddy DM, Johnston CC. Screening for osteoporosis. Ann Intern Med 1990;112:516-28.
5. Consensus Development Statement. Who are candidates for prevention and treatment for osteoporosis? Osteoporos Int 1997;7:1-6.
6. Clinical practice guidelines for the diagnosis and management of osteoporosis. Can Med Assoc J 1996;155:1113-29.
7. Matkovic V, Ilich JZ, Skugor M, Saracoglu M. Primary prevention of osteoporosis. Phys Med Rehab Clin North Am 1995;6:595-627.
8. Nelson DA. An anthropological perspective on optimizing calcium consumption for the prevention of osteoporosis. Osteoporos Int 1996;6:325-28.
9. Recker RR, Davies KM, Hinders SM, Heaney RP, Stegman MR, Kimmel DB. Bone gain in young adult women. JAMA 1992;268:2403-08.
10. Nowson CA, Green RM, Hopper JL, et al. A co-twin study of the effect of calcium supplementation on bone density during adolescence. Osteoporos Int 1997;7:219-25.
11. US Preventive Services Task Force. Guide to clinical preventive services. Baltimore, Md: Williams and Wilkins; 1996.
12. Lloyd T, Martel JK, Rollings N, et al. The effect of calcium supplementation and tanner stage on bone density, content and area in teenage women. Osteoporos Int 1996;6:276-83.
13. Kanders B, Dempster DAW, Lindsay R. Interaction of calcium nutrition and physical activity on bone mass in young women. J Bone Miner Res 1988;3:145-49.
14. Dawson-Hughes B, Dallal GE, Krall EA, Sadowski L, Sahyoun N, Tannenbaum S. A controlled trial of the effect of calcium supplementation on bone density in postmenopausal women. N Engl J Med 1990;323:878-83.
15. Devine A, Dick IM, Heal SJ, Criddle RA, Prince RL. A 4-year follow-up study of the effects of calcium supplementation on bone density in elderly postmenopausal women. Osteoporos Int 1997;7:23-28.
16. Schrager S, Kausch T, Bobula JA. Osteoporosis risk assessment by family practice faculty and residents: a chart review. Wis Med J 1999;98:34-36.
17. Citracal osteoporosis prevention survey conducted by Opinion Research Corporation International for Mission Pharmacal; 1996.
18. Bourguet CC, Hamrick GA, Gilchrist VJ. The prevalence of osteoporosis risk factors and physician intervention. J Fam Pract 1991;32:265-71.
19. Lurie N, Slater J, McGovern P, Ekstrum J, Quam L, Margolis K. Preventive care for women: does the sex of the physician matter? N Engl J Med 1993;329:478-82.
20. Cassard SD, Weisman CS, Plichta SB, Johnson TL. Physician gender and women’s preventive services. J Women’s Health 1997;6:199-207.
1. Riggs BL, Melton LJ. The prevention and treatment of osteoporosis. N Engl J Med 1992;327:620-27.
2. National Osteoporosis Foundation. Osteoporosis: review of the evidence for prevention, diagnosis and treatment and cost effectiveness analysis. Osteoporos Int 1998;8:S1-S88.
3. Lips P. Epidemiology and predictors of fractures associated with osteoporosis. Am J Med 1997;103:3S-8S;discussion 8S-11S.
4. Melton LJ, Eddy DM, Johnston CC. Screening for osteoporosis. Ann Intern Med 1990;112:516-28.
5. Consensus Development Statement. Who are candidates for prevention and treatment for osteoporosis? Osteoporos Int 1997;7:1-6.
6. Clinical practice guidelines for the diagnosis and management of osteoporosis. Can Med Assoc J 1996;155:1113-29.
7. Matkovic V, Ilich JZ, Skugor M, Saracoglu M. Primary prevention of osteoporosis. Phys Med Rehab Clin North Am 1995;6:595-627.
8. Nelson DA. An anthropological perspective on optimizing calcium consumption for the prevention of osteoporosis. Osteoporos Int 1996;6:325-28.
9. Recker RR, Davies KM, Hinders SM, Heaney RP, Stegman MR, Kimmel DB. Bone gain in young adult women. JAMA 1992;268:2403-08.
10. Nowson CA, Green RM, Hopper JL, et al. A co-twin study of the effect of calcium supplementation on bone density during adolescence. Osteoporos Int 1997;7:219-25.
11. US Preventive Services Task Force. Guide to clinical preventive services. Baltimore, Md: Williams and Wilkins; 1996.
12. Lloyd T, Martel JK, Rollings N, et al. The effect of calcium supplementation and tanner stage on bone density, content and area in teenage women. Osteoporos Int 1996;6:276-83.
13. Kanders B, Dempster DAW, Lindsay R. Interaction of calcium nutrition and physical activity on bone mass in young women. J Bone Miner Res 1988;3:145-49.
14. Dawson-Hughes B, Dallal GE, Krall EA, Sadowski L, Sahyoun N, Tannenbaum S. A controlled trial of the effect of calcium supplementation on bone density in postmenopausal women. N Engl J Med 1990;323:878-83.
15. Devine A, Dick IM, Heal SJ, Criddle RA, Prince RL. A 4-year follow-up study of the effects of calcium supplementation on bone density in elderly postmenopausal women. Osteoporos Int 1997;7:23-28.
16. Schrager S, Kausch T, Bobula JA. Osteoporosis risk assessment by family practice faculty and residents: a chart review. Wis Med J 1999;98:34-36.
17. Citracal osteoporosis prevention survey conducted by Opinion Research Corporation International for Mission Pharmacal; 1996.
18. Bourguet CC, Hamrick GA, Gilchrist VJ. The prevalence of osteoporosis risk factors and physician intervention. J Fam Pract 1991;32:265-71.
19. Lurie N, Slater J, McGovern P, Ekstrum J, Quam L, Margolis K. Preventive care for women: does the sex of the physician matter? N Engl J Med 1993;329:478-82.
20. Cassard SD, Weisman CS, Plichta SB, Johnson TL. Physician gender and women’s preventive services. J Women’s Health 1997;6:199-207.
Long-Term Follow-up of Depression Among Patients in the Community and in Family Practice Settings
METHODS: We performed a literature search to identify relevant papers published between 1970 and 1999 on original long-term follow-up studies of depression in community and primary care populations. The included studies were of adult populations with depression based on diagnostic criteria and a follow-up of at least 5 years. Data about recurrences, relapses, psychopathology, disability, or quality of life at follow-up were examined.
RESULTS: We found 8 studies that fulfilled our criteria. The reported rates of recurrence or depression at follow-up were between 30% and 40%. Higher rates were found in the younger and older age groups. Data about other predictors of outcome, health status, and the relation between treatment and outcome did not justify any hard conclusions.
CONCLUSIONS: The long-term outcome of depression in the community and in primary care is rarely studied. The results of available studies are difficult to compare because of the large differences in populations and methods. Nevertheless, these studies suggest that the long-term prognosis of depression in the community and in primary care is not as poor as in psychiatry.
Depression is regarded as a chronic illness with a high prevalence and a large impact on quality of life.1-6 Nevertheless, the long-term outcome of depression in primary care and in the community is not clear. Most long-term outcome studies of depression have been performed with populations of patients who have been referred to psychiatric specialists.7-11 However, not everyone with depression in the community consults a physician, and usually only the more severe and lasting cases—approximately 5% to 15% of all patients who seek medical attention and have received a diagnosis of depression in primary care—are referred for secondary care.12,13 It is unlikely that the outcome in the community and in primary care is identical to that of referred cases, because of the difference in the prevalence of the various severity levels of depression between these populations.14
The follow-up periods of most studies of depression performed in the community and in primary care have been relatively short.15-17 From these studies we know that patients experience disability during depressive episodes, but we do not have a clear picture of the long-term consequences from the patient perspective.2,4,5 Also, in short-term studies, rates of depression measured at follow-up are not conclusive in determining recurrence rates.
Concerning treatment, it has been established in many short-term studies that antidepressants are effective for the treatment of major depressive disorder11,18,19 and perhaps also for minor depression20-23 (with a high prevalence in community and primary care).24-27
However, papers can be found describing a totally different picture for the long-term outcome of chronic diseases.28,29 These studies demonstrate that short-term effectiveness and safety do not automatically predict long-term outcome. Therefore, we think that knowledge about the long-term course of depression in the community and in primary care, naturalistic as well as treated, is indispensable for determining what treatment strategy is justified. Studies about the negative effects of antidepressants have been published,30-33 and some suggest that these drugs might influence the course of depression in a negative way.34 Long-term outcome information should be available for all levels of depression, including cases for which no medical attention is sought, and differences between naturalistic outcome and outcome after treatment should be clear. We reviewed the literature for long-term outcome studies of depression in the community and in primary care, looking for answers to the following 3 research questions: What is the recurrence rate of depression? Can a relation be found between long-term outcome and treatment? What are the long-term consequences for the health status of the patients involved?
Methods
Retrieval of the Literature
We performed a computerized search of studies from 1970 to 1999 using MEDLINE, Psychlit, Current Contents, and The Cochrane Library. We chose 1970 as our starting point, because at that time modern classification systems were introduced and research diagnostic criteria became available to investigators.
Thesaurus and free text words were combined for “depression/depressive disorder” with “general practice/family practice/primary care” or “community” and “follow-up/course/outcome/prognosis.”
Selection of the Literature
Two reviewers (H.J.S., E.M.vW-B.) made a selection by screening titles and abstracts. If an abstract had been selected by only one of the reviewers it was discussed until consensus was reached.
Our inclusion criteria were: original longitudinal follow-up studies in English of adult populations in the community or primary care with at least 25 patients in the follow-up. In the included studies diagnosis of depression was according to: the International Classification of Primary Care or the International Classification of Health Problems in Primary Care (ICHPPC-2) in general practice studies; the Diagnostic and Statistical Manual of Mental Disorders (DSM), third edition, third edition revised, or fourth edition; the research diagnostic criteria (RDC) or immediate predecessor (St Louis); or the International Classification of Diseases (ICD)-9th Revision-Clinical Modification or the ICD-10th Revision.35,36 As inclusion criteria for outcome of depression we included studies reporting on recurrences, relapses, psychopathology, disability, or quality of life at follow-up. We defined long-term as a follow-up of at least 5 years, because recurrences usually occur within this time frame.9,13,37 A follow-up of at least 5 years should give an indication of percentages of single-episode depression and recurrence rates, and should provide more of an opportunity for distinguishing recurrence from relapse and no recovery (yet) than a shorter follow-up.
When abstracts met the inclusion criteria or remained unclear, full articles were retrieved for further evaluation. We also retrieved relevant reviews. All of these studies were screened with the ancestry approach. Additionally, a number of experts in the field from the Netherlands, the United Kingdom, and the United States were asked for additional references.
Data Abstraction and Presentation
Because of the wide variety of study designs, we limited ourselves to a qualitative evaluation. We abstracted data about design, setting, diagnostic criteria, number and specific diagnosis of depressive patients in the follow-up, age and sex, length of follow-up, treatment, and outcomes.
We calculated a total rate of recurrence or depression at follow up for all patients still alive and present at the end of the follow-up of each study. For that purpose we combined the rates for minor and major depression. This gave us the opportunity to compare the outcome results of depression diagnosed with family practice criteria with DSM cases and cases meeting RDC.
Results
Selection of Articles
The computer search supplied 421 potentially relevant articles. We selected 56 papers on the basis of that search, the reference lists of 4 review articles, and the suggestions of experts. Eight of those studies met all our inclusion criteria: 6 of these were community studies, and 2 were in primary care. Studies were excluded for 1 or more of the following reasons: no longitudinal follow-up (13), long-term follow-up shorter than 5 years (35), no diagnostic criteria mentioned in the article (4), population not from community or primary care (5) or too small (1), or outcome results of depression were mixed with other diagnoses (2).
Design, Aim, and Outcomes
Table 1 provides an overview of the included studies with outcomes as presented in the original articles. There was only one study from the 1970s38 meeting our criteria; all others were published in the last 10 years.
Initially there had either been a community survey with screening instruments, followed by diagnostic interviews or the whole population had been interviewed to identify cases of depression. Then there had been a follow up with the depressed subjects. In 4 of the 6 community studies the outcome was presented as depression at follow-up,39-42 and in the other 2 studies as recurrences.43,44 In 2 studies the results of depression at follow-up were based on 3 follow-up interviews: in the third, fourth, and fifth year after the initial assessment in the first study,41 and one every 5 years in the other.42 Three studies were performed on the elderly,39-41 one on young adults living in the community,42 and 2 in community samples in which all ages were represented.43,44 In one of these latter studies44 the population consisted of family members and relatives of affectively ill probands.
Both family practice studies had a historic cohort design13,38 and referred to patients recognized with depression in family practice. A cohort of depressed patients had been identified from a morbidity registry13 or practice,38 and was followed up longitudinally using the patients’ records (and registry13). The follow-up started on the date the diagnosis was made for the first time13 or the first time in the practice.38 Outcome in both studies was based on the reference to recurrences on the patients’ records over the entire study period.
The Diagnosis of Depression and Diagnostic Criteria
There was only one study using specific family practice criteria. For that study E-list criteria (the first classification for general practice, developed in the United Kingdom and derived from the ICD) was used initially and was later replaced with ICHPPC-2 criteria.45,13 In all the other studies DSM,40,42,43 RDC,38,42-44 or criteria derived from the RDC for use in elderly populations39,41 (obtained with GMS-AGECAT, a computerized diagnostic system for elderly subjects derived from the Geriatric Mental State46) were used. In the family practice study by Widmer and Cadoret38 the symptoms on the records were incorporated in the RDC retrospectively.
Length of Follow-up of Depression
In one study44 the length of the follow-up from the onset of depression was not clear because it did not mention when the depressive episodes occurred. In all other studies the follow-up began with a depressive episode. One study started the follow-up at first episodes only.13 That is the only study in which a time relation between the initial diagnosis and recurrences is presented longitudinally. One other study started at initial episodes in the practice38; all others start at index and recurrent episodes, but the proportions are not clear.
Recurrence or Depression at Follow-up
There were differences in levels of depression included in the outcome results Table 1. The rates of recurrence presented in the community studies ranged between 26% and 47%, and the rates of depression at follow-up were between 9% and 44%. Recurrence rates in the family practice studies ranged from 35% to 40%. The recurrence rate of 35% was calculated retrospectively, relying on the symptoms mentioned on the case records. The recurrence rate of 40% was found by extracting data from a morbidity registry and checking those data against symptoms on the patient records.13 Both studies also presented the number of recurrences. In one study 27% of the followed-up patients had 2 episodes; 6%, 3; and 3%, 4 or more38; in the other the percentages were 16% with 2, 12% with 3, and 12% with 4 or more.13
Treatment
The authors of 4 studies38-40,42 reported on treatment. None described the nature and length of treatment clearly, and treatment was not related to recurrence or depression at follow-up.
Mortality
Data for mortality were given in 4 studies13,39-41 (range of rates=14%-44%). The higher mortality rates refer to the elderly. In 2 studies mortality was similar to the expected rates (compared with a control group in one13 and with the National Mortality Statistics in the other39); in another study41 the rates were significantly higher than in a control group of nondepressed individuals; and in the last study13 the results of a comparison were not discussed. Data on suicide attempts and suicide can be found in only one study.13
Health Status
Two studies reported on disability or self-perceived physical health, but none of the studies used any of the well-known health status measurement instruments. One study39 reported on disability levels (with a modified version of the American Resources and Services) There was significantly more moderate to severe impairment among the depressed than among the recovered cases, as rated by a clinician; also, it is not clear whether this rating refers to the initial assessment or the follow-up. In the other study40 46% of the elderly patients reported poor health status, but it is not clear which instrument was used; no relation was found between outcome and perceived health status.
Total Rates of Recurrence or Depression at Follow-up
Table 2 shows the total rates of recurrence or depression at follow-up, adding up the results of minor and major depression when possible. The recurrence rates of the populations in which no specific age group was followed-up ranged from 30% to 40%. These studies had indications for higher recurrence rates in the younger age groups. The community studies of depression in the elderly and young adults reported outcome as depression at follow-up. The rates of depression at follow-up in 2 of the studies of the elderly39,41 were relatively high. One of these studies41 only reported on major depression at follow-up. This was also the case in the study with young adults.42
Discussion
Studies of the long-term outcome of depression in the community and in primary care are scarce and difficult to compare, and methodologic shortcomings hamper their generalizability.
Our data suggest that overall recurrence rates in the community and in family practice vary between 30% and 40%. The relationship between treatment and long-term outcome remains unclear, because none of the studies were controlled trials for treatment or looked into this matter adequately. This also applies to the patient’s qualitative experience. Almost all studies exclusively report physician-diagnosed recurrence or positive scores on diagnostic instruments of depression at follow-up.
Recurrence rates of 30% to 40% indicate that the prognosis for depression in community and family practice is not as poor as in psychiatry. In psychiatric settings much higher recurrence rates are found, with percentages of up to almost 90% depending on the length of follow-up and the setting.9,10,37 Prognosis seems to be related to age, with young adults and the elderly having poorer prognoses. In the study on young adults, between 30% and 40% of the patients had a major depressive disorder at follow-up, but those results did not include rates of minor depression. They also did not include recurrences between the follow-up interviews, and thus it is likely that recurrence rates were higher. Higher recurrence rates were also found in the younger age groups in 2 of the community studies.43,44 All the studies performed exclusively with the elderly reported depression at follow-up and gave a poor prognosis.39-41 This was not confirmed in the 2 community studies involving various ages and one general practice study.13,43,44 The higher rates in studies on the elderly might be explained by the use of diagnostic instruments more sensitive to detect depression specifically in that age group. Another explanation might be that differences were not found because of the relatively small number of elderly persons present in the other studies.
Limitations of the Studies
In the community studies subjects were identified with screening instruments. Therefore, a number of false-positive diagnoses may have been included in these studies that biased outcome results.47 This risk was minimized by using interviews in addition to the screening instruments. Another important limitation is the risk of missing part of the information about recurrences in the intervals between assessments. This risk is less in studies that include more than one follow-up assessment, as was done in 2 of the included studies,41,42 and also in studies where data about the interval are retrieved using information based on patients’ recall.40,43 Recall is known to introduce bias by not always giving sufficient details after longer periods of time.48
In the family practice studies where the specificity of the diagnosis is usually high,49,50 outcome results may have been biased because the results of undetected or misdiagnosed patients with depression are missing. In both studies the information was retrieved from the case records. Thus, accuracy depended on the completeness of the physicians’ notes. In one of the studies the case records were used in addition to data from a morbidity registry in which physicians were trained regularly to use criteria for diagnosis. Although suicide data can be found in this study, patients who left the practices or died within 10 years were excluded, and outcome results should be viewed taking that into consideration.
An important shortcoming is that most studies started at first or recurrent episodes, so it is not possible to give an exact percentage of single-episode depression versus recurrent illness. A description of the longitudinal course starting at first diagnosis is also not possible. Even in the one study starting at the first episode we can only draw conclusions about recurrence rates after diagnosis, because we have no certainty that the first diagnosis was in fact related to the first depressive episode. Other shortcomings are the small number of patients in the follow-up in 2 studies39,43 and uncertainty about the representativeness of the samples in one of the community studies.39,41,42,44 Since a family history of depression is a risk factor in an individual,51 the population of family members of affectively ill probands cannot be regarded as a representative community sample.44
The validity of the recurrence rates mentioned in the articles is difficult to assess because only one author gave confidence intervals of recurrence rates or depression at follow-up.13
Limitations of Our Review
Although we made our choice of inclusion criteria to ensure reasonable comparability, older studies52,53 and those in which data on depression could not be extracted from a broader variety of mental illness in family practice54 were excluded. We also made the choice to describe a limited number of outcomes, but the small number of studies included and the variety within the studies did not allow a review of more outcome results.
Although the calculation of a total recurrence rate may be criticized, we think that the fluctuating nature of depression justifies this procedure.
Conclusions
There are large gaps in the available knowledge about long-term outcome of depression in primary care, and future studies are required to fill in these gaps. We recommend the following:
- The outcome of all types of depression should be evaluated in prospective studies, with a follow-up of at least 5 years, of representative samples in both community and primary care.
- Continuous morbidity registration should be used. With this aim, data meeting fixed criteria that have been established beforehand should be collected longitudinally.
- Studies should include naturalistic follow-up and relate treatment to outcome.
- Quality-of-life assessment should be included.
Recommendations for clinical practice
Family physicians can reassure patients with depression by telling them that although the long-term outcome of the illness is not completely clear, there are indications that the majority of patients with depression do not have a poor prognosis. As the long term risk of recurrence seems to be approximately 40%, most patients in primary care settings only have 1 episode of depression. This information might aid a patient’s recovery.
1. Froom J, Aoyama H, Hermoni D, et al. Depressive disorders in three primary care populations: United States, Israel, Japan. Fam Pract 1995;12:274-78
2. Ormel J, VonKorff M, Ustun TB, et al. Common mental disorders and disability across cultures: results from the WHO Collaborative Study on Psychological Problems in General Health Care. JAMA 1994;272:1741-48
3. Robins LN. Lifetime prevalence of specific psychiatric disorders in three sites. Arch Gen Psychiatry 1984;41:949-58
4. Williams JW, Kerber CA, Mulrow CD, et al. Depressive disorders in primary care: prevalence, functional disability, and identification. J Gen Intern Med 1995;10:7-12
5. Wells KB, Stewart A, Hays RD, et al. The functioning and well-being of depressed patients: results from the Medical Outcomes Study. JAMA 1989;262:914-19
6. Katon W, Schulberg H. Epidemiology of depression in primary care. Gen Hosp Psychiatry 1992;14:237-47
7. Coryell W, Turvey C, Endicott J, et al. Bipolar I affective disorder: predictors of outcome after 15 years. J Affect Disord 1998;50:109-16
8. Paykel ES. Historical overview of outcome of depression. Br J Psychiatry Suppl 1994;26:6-8
9. Angst J, Preisig M. Course of a clinical cohort of unipolar, bipolar and schizoaffective patients: results of a prospective study from 1959 to 1985. Schweiz Archiv Neurol Psychiatr 1995;146:5-16
10. Piccinelli M, Wilkinson G. Outcome of depression in psychiatric settings. Br J Psychiatry 1994;164:297-304
11. Lavori PW, Keller MB, Mueller TI, et al. Recurrence after recovery in unipolar MDD: an observational follow-up study of clinical predictors and somatic treatment as a mediating factor. Int J Methods Psychiatric Res 1994;4:211-29
12. Blacker CV, Clare AW. Depressive disorder in primary care. Br J Psychiatry 1987;150:737-51
13. van Weel-Baumgarten E, van den Bosch W, van den Hoogen H, et al. Ten year follow-up of depression after diagnosis in general practice. Br J Gen Pract 1998;48:1643-46
14. Watts CA. Depressive disorders in the community: the scene in Great Britain, 1965. J Clin Psychiatry 1984;45:70-77
15. Klinkman MS, Schwenk TL, Coyne JC. Depression in primary care—more like asthma than appendicitis: the Michigan Depression Project. Can J Psychiatry 1997;42:966-73.
16. Lin EH, Simon GE, Katon WJ, et al. Can enhanced acute-phase treatment of depression improve long-term outcomes? A report of randomized trials in primary care. Am J Psychiatry 1999;156:643-45
17. Ormel J, Oldehinkel T, Brilman E, et al. Outcome of depression and anxiety in primary care: a three-wave 3 1/2-year study of psychopathology and disability. Arch Gen Psychiatry 1993;50:759-66
18. Kupfer DJ. Long-term treatment of depression. J Clin Psychiatry 1991;52 (suppl):28-34.
19. Prien RF. Efficacy of continuation drug therapy of depression and anxiety: issues and methodologies. J Clin Psychopharmacol 1990;10:86S-90S.
20. Tan RS, Barlow RJ, Abel C, et al. The effect of low dose lofepramine in depressed elderly patients in general medical wards. Br J Clin Pharmacol 1994;37:321-24
21. Thompson C, Thompson CM. The prescribing of antidepressants in general practice: II. A placebo-controlled trial of low-dose dothiepin. Human Psychopharma 1989;191-204.
22. Wernicke JF, Dunlop SR, Dornseif BE, et al. Low-dose fluoxetine therapy for depression. Psychopharmacol Bull 1988;24:183-88
23. Paykel ES, Freeling P, Hollyman JA. Are tricyclic antidepressants useful for mild depression? A placebo controlled trial. Pharmacopsychiatry 1988;21:15-18
24. Hollyman JA, Freeling P, Paykel ES, et al. Double-blind placebo-controlled trial of amitriptyline among depressed patients in general practice. J R Coll Gen Pract 1988;38:393-97
25. Mynors WL, Gath D. Predictors of treatment outcome for major depression in primary care. Psychol Med 1997;27:731-36
26. Scott AI, Freeman CP. Edinburgh primary care depression study: treatment outcome, patient satisfaction, and cost after 16 weeks. BMJ 1992;304:883-87
27. Schulberg HC, Block MR, Madonia MJ, et al. Treating major depression in primary care practice: eight-month clinical outcomes. Arch Gen Psychiatry 1996;53:913-19
28. Pincus T, Stein CM. Why randomized controlled clinical trials do not depict accurately long-term outcomes in rheumatoid arthritis: some explanations and suggestions for future studies. Clin Exp Rheumatol 1997;15 (suppl):S27-38.
29. Wolfe F, Hawley DJ, Cathey MA. Clinical and health status measures over time: prognosis and outcome assessment in rheumatoid arthritis. J Rheumatol 1991;18:1290-97.
30. de Abajo FJ, Rodriguez LA, Montero D. Association between selective serotonin reuptake inhibitors and upper gastrointestinal bleeding: population based case-control study. BMJ 1999;319:1106-09
31. Po AL. Antidepressants and upper gastrointestinal bleeding. BMJ 1999;319:1081-82
32. Mackay FR, Dunn NR, Martin RM, et al. Newer antidepressants: a comparison of tolerability in general practice. Br J Gen Pract 1999;49:892-96
33. Sampson E, Warner JP. Serotonin syndrome: potentially fatal but difficult to recognize. Br J Gen Pract 1999;49:867-68
34. Tondo L, Laddomada P, Serra G, et al. Rapid cyclers and antidepressants. Int Pharmacopsychiatry 1981;16:119-23
35. Classification committee of WONCA. ICHPPC-2-defined (International classification of Health Problems in Primary Care). 3rd ed. Oxford, England: Oxford University Press; 1983.
36. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th edition. Washington, DC: American Psychiatric Association; 1994.
37. Van Londen L, Molenaar RP, Goekoop JG, et al. Three- to 5-year prospective follow-up of outcome in major depression. Psychol Med 1998;28:731-35
38. Widmer RB, Cadoret RJ. Depression in primary care: changes in pattern of patient visits and complaints during a developing depression. J Fam Pract 1978;7:293-302
39. Kua EH. The depressed elderly Chinese living in the community: a five-year follow-up study. Int J Geriatr Psychiatry 1993;8:427-30
40. Kivela SL, Kongas SP, Kesti E, et al. Five-year prognosis for depression in old age. Int Psychogeriatr 1994;6:69-78
41. Sharma VK, Copeland JR, Dewey ME, et al. Outcome of the depressed elderly living in the community in Liverpool: a 5-year follow-up. Psychol Med 1998;28:1329-37
42. Angst J, Merikangas K. The depressive spectrum: diagnostic classification and course. J Affect Disord 1997;45:31-39
43. Eaton WW, Anthony JC, Gallo J, et al. Natural history of Diagnostic Interview Schedule/DSM-IV major depression: the Baltimore Epidemiologic Catchment Area follow-up. Arch Gen Psychiatry 1997;54:993-99
44. Coryell W, Endicott J, Keller MB. Predictors of relapse into major depressive disorder in a nonclinical population. Am J Psychiatry 1991;148:1353-58
45. Logan WPD, Cushion A. Morbidity statistics from general practice. Vol 1. London, England: Her Majesty’s Stationary Office; 1954.
46. Copeland JR, Dewey ME, Griffiths-Jones HM. A computerized psychiatric diagnostic system and case nomenclature for elderly subjects: GMS and AGECAT. Psychol Med 1986;16:89-99
47. Nagel R, Lynch D, Tamburrino M. Validity of the medical outcomes study depression screener in family practice training centers and community settings. Fam Med 1998;30:362-65
48. Andrews G, Anstey K, Brodaty H, et al. Recall of depressive episode 25 years previously. Psychol Med 1999;29:787-91
49. Wright AF. Should general practitioners be testing for depression? Br J Gen Pract 1994;44:132-35
50. Van Weel C. Validating long term morbidity recording. J Epidemiol Community Health 1995;49 (suppl):29-32.
51. Merikangas KR, Wicki W, Angst J. Heterogeneity of depression: classification of depressive subtypes by longitudinal course. Br J Psychiatry 1994;164:342-48
52. Murphy JM, Olivier DC, Sobol AM, et al. Diagnosis and outcome: depression and anxiety in a general population. Psychol Med 1986;16:117-26
53. Hagnell O, Lanke J, Rorsman B. Suicide and depression in the male part of the Lundby study: changes over time during a 25-year observation period. Neuropsychobiology 1982;8:182-87
54. Lloyd KR, Jenkins R, Mann A. Long-term outcome of patients with neurotic illness in general practice. BMJ 1996;313:26-28
METHODS: We performed a literature search to identify relevant papers published between 1970 and 1999 on original long-term follow-up studies of depression in community and primary care populations. The included studies were of adult populations with depression based on diagnostic criteria and a follow-up of at least 5 years. Data about recurrences, relapses, psychopathology, disability, or quality of life at follow-up were examined.
RESULTS: We found 8 studies that fulfilled our criteria. The reported rates of recurrence or depression at follow-up were between 30% and 40%. Higher rates were found in the younger and older age groups. Data about other predictors of outcome, health status, and the relation between treatment and outcome did not justify any hard conclusions.
CONCLUSIONS: The long-term outcome of depression in the community and in primary care is rarely studied. The results of available studies are difficult to compare because of the large differences in populations and methods. Nevertheless, these studies suggest that the long-term prognosis of depression in the community and in primary care is not as poor as in psychiatry.
Depression is regarded as a chronic illness with a high prevalence and a large impact on quality of life.1-6 Nevertheless, the long-term outcome of depression in primary care and in the community is not clear. Most long-term outcome studies of depression have been performed with populations of patients who have been referred to psychiatric specialists.7-11 However, not everyone with depression in the community consults a physician, and usually only the more severe and lasting cases—approximately 5% to 15% of all patients who seek medical attention and have received a diagnosis of depression in primary care—are referred for secondary care.12,13 It is unlikely that the outcome in the community and in primary care is identical to that of referred cases, because of the difference in the prevalence of the various severity levels of depression between these populations.14
The follow-up periods of most studies of depression performed in the community and in primary care have been relatively short.15-17 From these studies we know that patients experience disability during depressive episodes, but we do not have a clear picture of the long-term consequences from the patient perspective.2,4,5 Also, in short-term studies, rates of depression measured at follow-up are not conclusive in determining recurrence rates.
Concerning treatment, it has been established in many short-term studies that antidepressants are effective for the treatment of major depressive disorder11,18,19 and perhaps also for minor depression20-23 (with a high prevalence in community and primary care).24-27
However, papers can be found describing a totally different picture for the long-term outcome of chronic diseases.28,29 These studies demonstrate that short-term effectiveness and safety do not automatically predict long-term outcome. Therefore, we think that knowledge about the long-term course of depression in the community and in primary care, naturalistic as well as treated, is indispensable for determining what treatment strategy is justified. Studies about the negative effects of antidepressants have been published,30-33 and some suggest that these drugs might influence the course of depression in a negative way.34 Long-term outcome information should be available for all levels of depression, including cases for which no medical attention is sought, and differences between naturalistic outcome and outcome after treatment should be clear. We reviewed the literature for long-term outcome studies of depression in the community and in primary care, looking for answers to the following 3 research questions: What is the recurrence rate of depression? Can a relation be found between long-term outcome and treatment? What are the long-term consequences for the health status of the patients involved?
Methods
Retrieval of the Literature
We performed a computerized search of studies from 1970 to 1999 using MEDLINE, Psychlit, Current Contents, and The Cochrane Library. We chose 1970 as our starting point, because at that time modern classification systems were introduced and research diagnostic criteria became available to investigators.
Thesaurus and free text words were combined for “depression/depressive disorder” with “general practice/family practice/primary care” or “community” and “follow-up/course/outcome/prognosis.”
Selection of the Literature
Two reviewers (H.J.S., E.M.vW-B.) made a selection by screening titles and abstracts. If an abstract had been selected by only one of the reviewers it was discussed until consensus was reached.
Our inclusion criteria were: original longitudinal follow-up studies in English of adult populations in the community or primary care with at least 25 patients in the follow-up. In the included studies diagnosis of depression was according to: the International Classification of Primary Care or the International Classification of Health Problems in Primary Care (ICHPPC-2) in general practice studies; the Diagnostic and Statistical Manual of Mental Disorders (DSM), third edition, third edition revised, or fourth edition; the research diagnostic criteria (RDC) or immediate predecessor (St Louis); or the International Classification of Diseases (ICD)-9th Revision-Clinical Modification or the ICD-10th Revision.35,36 As inclusion criteria for outcome of depression we included studies reporting on recurrences, relapses, psychopathology, disability, or quality of life at follow-up. We defined long-term as a follow-up of at least 5 years, because recurrences usually occur within this time frame.9,13,37 A follow-up of at least 5 years should give an indication of percentages of single-episode depression and recurrence rates, and should provide more of an opportunity for distinguishing recurrence from relapse and no recovery (yet) than a shorter follow-up.
When abstracts met the inclusion criteria or remained unclear, full articles were retrieved for further evaluation. We also retrieved relevant reviews. All of these studies were screened with the ancestry approach. Additionally, a number of experts in the field from the Netherlands, the United Kingdom, and the United States were asked for additional references.
Data Abstraction and Presentation
Because of the wide variety of study designs, we limited ourselves to a qualitative evaluation. We abstracted data about design, setting, diagnostic criteria, number and specific diagnosis of depressive patients in the follow-up, age and sex, length of follow-up, treatment, and outcomes.
We calculated a total rate of recurrence or depression at follow up for all patients still alive and present at the end of the follow-up of each study. For that purpose we combined the rates for minor and major depression. This gave us the opportunity to compare the outcome results of depression diagnosed with family practice criteria with DSM cases and cases meeting RDC.
Results
Selection of Articles
The computer search supplied 421 potentially relevant articles. We selected 56 papers on the basis of that search, the reference lists of 4 review articles, and the suggestions of experts. Eight of those studies met all our inclusion criteria: 6 of these were community studies, and 2 were in primary care. Studies were excluded for 1 or more of the following reasons: no longitudinal follow-up (13), long-term follow-up shorter than 5 years (35), no diagnostic criteria mentioned in the article (4), population not from community or primary care (5) or too small (1), or outcome results of depression were mixed with other diagnoses (2).
Design, Aim, and Outcomes
Table 1 provides an overview of the included studies with outcomes as presented in the original articles. There was only one study from the 1970s38 meeting our criteria; all others were published in the last 10 years.
Initially there had either been a community survey with screening instruments, followed by diagnostic interviews or the whole population had been interviewed to identify cases of depression. Then there had been a follow up with the depressed subjects. In 4 of the 6 community studies the outcome was presented as depression at follow-up,39-42 and in the other 2 studies as recurrences.43,44 In 2 studies the results of depression at follow-up were based on 3 follow-up interviews: in the third, fourth, and fifth year after the initial assessment in the first study,41 and one every 5 years in the other.42 Three studies were performed on the elderly,39-41 one on young adults living in the community,42 and 2 in community samples in which all ages were represented.43,44 In one of these latter studies44 the population consisted of family members and relatives of affectively ill probands.
Both family practice studies had a historic cohort design13,38 and referred to patients recognized with depression in family practice. A cohort of depressed patients had been identified from a morbidity registry13 or practice,38 and was followed up longitudinally using the patients’ records (and registry13). The follow-up started on the date the diagnosis was made for the first time13 or the first time in the practice.38 Outcome in both studies was based on the reference to recurrences on the patients’ records over the entire study period.
The Diagnosis of Depression and Diagnostic Criteria
There was only one study using specific family practice criteria. For that study E-list criteria (the first classification for general practice, developed in the United Kingdom and derived from the ICD) was used initially and was later replaced with ICHPPC-2 criteria.45,13 In all the other studies DSM,40,42,43 RDC,38,42-44 or criteria derived from the RDC for use in elderly populations39,41 (obtained with GMS-AGECAT, a computerized diagnostic system for elderly subjects derived from the Geriatric Mental State46) were used. In the family practice study by Widmer and Cadoret38 the symptoms on the records were incorporated in the RDC retrospectively.
Length of Follow-up of Depression
In one study44 the length of the follow-up from the onset of depression was not clear because it did not mention when the depressive episodes occurred. In all other studies the follow-up began with a depressive episode. One study started the follow-up at first episodes only.13 That is the only study in which a time relation between the initial diagnosis and recurrences is presented longitudinally. One other study started at initial episodes in the practice38; all others start at index and recurrent episodes, but the proportions are not clear.
Recurrence or Depression at Follow-up
There were differences in levels of depression included in the outcome results Table 1. The rates of recurrence presented in the community studies ranged between 26% and 47%, and the rates of depression at follow-up were between 9% and 44%. Recurrence rates in the family practice studies ranged from 35% to 40%. The recurrence rate of 35% was calculated retrospectively, relying on the symptoms mentioned on the case records. The recurrence rate of 40% was found by extracting data from a morbidity registry and checking those data against symptoms on the patient records.13 Both studies also presented the number of recurrences. In one study 27% of the followed-up patients had 2 episodes; 6%, 3; and 3%, 4 or more38; in the other the percentages were 16% with 2, 12% with 3, and 12% with 4 or more.13
Treatment
The authors of 4 studies38-40,42 reported on treatment. None described the nature and length of treatment clearly, and treatment was not related to recurrence or depression at follow-up.
Mortality
Data for mortality were given in 4 studies13,39-41 (range of rates=14%-44%). The higher mortality rates refer to the elderly. In 2 studies mortality was similar to the expected rates (compared with a control group in one13 and with the National Mortality Statistics in the other39); in another study41 the rates were significantly higher than in a control group of nondepressed individuals; and in the last study13 the results of a comparison were not discussed. Data on suicide attempts and suicide can be found in only one study.13
Health Status
Two studies reported on disability or self-perceived physical health, but none of the studies used any of the well-known health status measurement instruments. One study39 reported on disability levels (with a modified version of the American Resources and Services) There was significantly more moderate to severe impairment among the depressed than among the recovered cases, as rated by a clinician; also, it is not clear whether this rating refers to the initial assessment or the follow-up. In the other study40 46% of the elderly patients reported poor health status, but it is not clear which instrument was used; no relation was found between outcome and perceived health status.
Total Rates of Recurrence or Depression at Follow-up
Table 2 shows the total rates of recurrence or depression at follow-up, adding up the results of minor and major depression when possible. The recurrence rates of the populations in which no specific age group was followed-up ranged from 30% to 40%. These studies had indications for higher recurrence rates in the younger age groups. The community studies of depression in the elderly and young adults reported outcome as depression at follow-up. The rates of depression at follow-up in 2 of the studies of the elderly39,41 were relatively high. One of these studies41 only reported on major depression at follow-up. This was also the case in the study with young adults.42
Discussion
Studies of the long-term outcome of depression in the community and in primary care are scarce and difficult to compare, and methodologic shortcomings hamper their generalizability.
Our data suggest that overall recurrence rates in the community and in family practice vary between 30% and 40%. The relationship between treatment and long-term outcome remains unclear, because none of the studies were controlled trials for treatment or looked into this matter adequately. This also applies to the patient’s qualitative experience. Almost all studies exclusively report physician-diagnosed recurrence or positive scores on diagnostic instruments of depression at follow-up.
Recurrence rates of 30% to 40% indicate that the prognosis for depression in community and family practice is not as poor as in psychiatry. In psychiatric settings much higher recurrence rates are found, with percentages of up to almost 90% depending on the length of follow-up and the setting.9,10,37 Prognosis seems to be related to age, with young adults and the elderly having poorer prognoses. In the study on young adults, between 30% and 40% of the patients had a major depressive disorder at follow-up, but those results did not include rates of minor depression. They also did not include recurrences between the follow-up interviews, and thus it is likely that recurrence rates were higher. Higher recurrence rates were also found in the younger age groups in 2 of the community studies.43,44 All the studies performed exclusively with the elderly reported depression at follow-up and gave a poor prognosis.39-41 This was not confirmed in the 2 community studies involving various ages and one general practice study.13,43,44 The higher rates in studies on the elderly might be explained by the use of diagnostic instruments more sensitive to detect depression specifically in that age group. Another explanation might be that differences were not found because of the relatively small number of elderly persons present in the other studies.
Limitations of the Studies
In the community studies subjects were identified with screening instruments. Therefore, a number of false-positive diagnoses may have been included in these studies that biased outcome results.47 This risk was minimized by using interviews in addition to the screening instruments. Another important limitation is the risk of missing part of the information about recurrences in the intervals between assessments. This risk is less in studies that include more than one follow-up assessment, as was done in 2 of the included studies,41,42 and also in studies where data about the interval are retrieved using information based on patients’ recall.40,43 Recall is known to introduce bias by not always giving sufficient details after longer periods of time.48
In the family practice studies where the specificity of the diagnosis is usually high,49,50 outcome results may have been biased because the results of undetected or misdiagnosed patients with depression are missing. In both studies the information was retrieved from the case records. Thus, accuracy depended on the completeness of the physicians’ notes. In one of the studies the case records were used in addition to data from a morbidity registry in which physicians were trained regularly to use criteria for diagnosis. Although suicide data can be found in this study, patients who left the practices or died within 10 years were excluded, and outcome results should be viewed taking that into consideration.
An important shortcoming is that most studies started at first or recurrent episodes, so it is not possible to give an exact percentage of single-episode depression versus recurrent illness. A description of the longitudinal course starting at first diagnosis is also not possible. Even in the one study starting at the first episode we can only draw conclusions about recurrence rates after diagnosis, because we have no certainty that the first diagnosis was in fact related to the first depressive episode. Other shortcomings are the small number of patients in the follow-up in 2 studies39,43 and uncertainty about the representativeness of the samples in one of the community studies.39,41,42,44 Since a family history of depression is a risk factor in an individual,51 the population of family members of affectively ill probands cannot be regarded as a representative community sample.44
The validity of the recurrence rates mentioned in the articles is difficult to assess because only one author gave confidence intervals of recurrence rates or depression at follow-up.13
Limitations of Our Review
Although we made our choice of inclusion criteria to ensure reasonable comparability, older studies52,53 and those in which data on depression could not be extracted from a broader variety of mental illness in family practice54 were excluded. We also made the choice to describe a limited number of outcomes, but the small number of studies included and the variety within the studies did not allow a review of more outcome results.
Although the calculation of a total recurrence rate may be criticized, we think that the fluctuating nature of depression justifies this procedure.
Conclusions
There are large gaps in the available knowledge about long-term outcome of depression in primary care, and future studies are required to fill in these gaps. We recommend the following:
- The outcome of all types of depression should be evaluated in prospective studies, with a follow-up of at least 5 years, of representative samples in both community and primary care.
- Continuous morbidity registration should be used. With this aim, data meeting fixed criteria that have been established beforehand should be collected longitudinally.
- Studies should include naturalistic follow-up and relate treatment to outcome.
- Quality-of-life assessment should be included.
Recommendations for clinical practice
Family physicians can reassure patients with depression by telling them that although the long-term outcome of the illness is not completely clear, there are indications that the majority of patients with depression do not have a poor prognosis. As the long term risk of recurrence seems to be approximately 40%, most patients in primary care settings only have 1 episode of depression. This information might aid a patient’s recovery.
METHODS: We performed a literature search to identify relevant papers published between 1970 and 1999 on original long-term follow-up studies of depression in community and primary care populations. The included studies were of adult populations with depression based on diagnostic criteria and a follow-up of at least 5 years. Data about recurrences, relapses, psychopathology, disability, or quality of life at follow-up were examined.
RESULTS: We found 8 studies that fulfilled our criteria. The reported rates of recurrence or depression at follow-up were between 30% and 40%. Higher rates were found in the younger and older age groups. Data about other predictors of outcome, health status, and the relation between treatment and outcome did not justify any hard conclusions.
CONCLUSIONS: The long-term outcome of depression in the community and in primary care is rarely studied. The results of available studies are difficult to compare because of the large differences in populations and methods. Nevertheless, these studies suggest that the long-term prognosis of depression in the community and in primary care is not as poor as in psychiatry.
Depression is regarded as a chronic illness with a high prevalence and a large impact on quality of life.1-6 Nevertheless, the long-term outcome of depression in primary care and in the community is not clear. Most long-term outcome studies of depression have been performed with populations of patients who have been referred to psychiatric specialists.7-11 However, not everyone with depression in the community consults a physician, and usually only the more severe and lasting cases—approximately 5% to 15% of all patients who seek medical attention and have received a diagnosis of depression in primary care—are referred for secondary care.12,13 It is unlikely that the outcome in the community and in primary care is identical to that of referred cases, because of the difference in the prevalence of the various severity levels of depression between these populations.14
The follow-up periods of most studies of depression performed in the community and in primary care have been relatively short.15-17 From these studies we know that patients experience disability during depressive episodes, but we do not have a clear picture of the long-term consequences from the patient perspective.2,4,5 Also, in short-term studies, rates of depression measured at follow-up are not conclusive in determining recurrence rates.
Concerning treatment, it has been established in many short-term studies that antidepressants are effective for the treatment of major depressive disorder11,18,19 and perhaps also for minor depression20-23 (with a high prevalence in community and primary care).24-27
However, papers can be found describing a totally different picture for the long-term outcome of chronic diseases.28,29 These studies demonstrate that short-term effectiveness and safety do not automatically predict long-term outcome. Therefore, we think that knowledge about the long-term course of depression in the community and in primary care, naturalistic as well as treated, is indispensable for determining what treatment strategy is justified. Studies about the negative effects of antidepressants have been published,30-33 and some suggest that these drugs might influence the course of depression in a negative way.34 Long-term outcome information should be available for all levels of depression, including cases for which no medical attention is sought, and differences between naturalistic outcome and outcome after treatment should be clear. We reviewed the literature for long-term outcome studies of depression in the community and in primary care, looking for answers to the following 3 research questions: What is the recurrence rate of depression? Can a relation be found between long-term outcome and treatment? What are the long-term consequences for the health status of the patients involved?
Methods
Retrieval of the Literature
We performed a computerized search of studies from 1970 to 1999 using MEDLINE, Psychlit, Current Contents, and The Cochrane Library. We chose 1970 as our starting point, because at that time modern classification systems were introduced and research diagnostic criteria became available to investigators.
Thesaurus and free text words were combined for “depression/depressive disorder” with “general practice/family practice/primary care” or “community” and “follow-up/course/outcome/prognosis.”
Selection of the Literature
Two reviewers (H.J.S., E.M.vW-B.) made a selection by screening titles and abstracts. If an abstract had been selected by only one of the reviewers it was discussed until consensus was reached.
Our inclusion criteria were: original longitudinal follow-up studies in English of adult populations in the community or primary care with at least 25 patients in the follow-up. In the included studies diagnosis of depression was according to: the International Classification of Primary Care or the International Classification of Health Problems in Primary Care (ICHPPC-2) in general practice studies; the Diagnostic and Statistical Manual of Mental Disorders (DSM), third edition, third edition revised, or fourth edition; the research diagnostic criteria (RDC) or immediate predecessor (St Louis); or the International Classification of Diseases (ICD)-9th Revision-Clinical Modification or the ICD-10th Revision.35,36 As inclusion criteria for outcome of depression we included studies reporting on recurrences, relapses, psychopathology, disability, or quality of life at follow-up. We defined long-term as a follow-up of at least 5 years, because recurrences usually occur within this time frame.9,13,37 A follow-up of at least 5 years should give an indication of percentages of single-episode depression and recurrence rates, and should provide more of an opportunity for distinguishing recurrence from relapse and no recovery (yet) than a shorter follow-up.
When abstracts met the inclusion criteria or remained unclear, full articles were retrieved for further evaluation. We also retrieved relevant reviews. All of these studies were screened with the ancestry approach. Additionally, a number of experts in the field from the Netherlands, the United Kingdom, and the United States were asked for additional references.
Data Abstraction and Presentation
Because of the wide variety of study designs, we limited ourselves to a qualitative evaluation. We abstracted data about design, setting, diagnostic criteria, number and specific diagnosis of depressive patients in the follow-up, age and sex, length of follow-up, treatment, and outcomes.
We calculated a total rate of recurrence or depression at follow up for all patients still alive and present at the end of the follow-up of each study. For that purpose we combined the rates for minor and major depression. This gave us the opportunity to compare the outcome results of depression diagnosed with family practice criteria with DSM cases and cases meeting RDC.
Results
Selection of Articles
The computer search supplied 421 potentially relevant articles. We selected 56 papers on the basis of that search, the reference lists of 4 review articles, and the suggestions of experts. Eight of those studies met all our inclusion criteria: 6 of these were community studies, and 2 were in primary care. Studies were excluded for 1 or more of the following reasons: no longitudinal follow-up (13), long-term follow-up shorter than 5 years (35), no diagnostic criteria mentioned in the article (4), population not from community or primary care (5) or too small (1), or outcome results of depression were mixed with other diagnoses (2).
Design, Aim, and Outcomes
Table 1 provides an overview of the included studies with outcomes as presented in the original articles. There was only one study from the 1970s38 meeting our criteria; all others were published in the last 10 years.
Initially there had either been a community survey with screening instruments, followed by diagnostic interviews or the whole population had been interviewed to identify cases of depression. Then there had been a follow up with the depressed subjects. In 4 of the 6 community studies the outcome was presented as depression at follow-up,39-42 and in the other 2 studies as recurrences.43,44 In 2 studies the results of depression at follow-up were based on 3 follow-up interviews: in the third, fourth, and fifth year after the initial assessment in the first study,41 and one every 5 years in the other.42 Three studies were performed on the elderly,39-41 one on young adults living in the community,42 and 2 in community samples in which all ages were represented.43,44 In one of these latter studies44 the population consisted of family members and relatives of affectively ill probands.
Both family practice studies had a historic cohort design13,38 and referred to patients recognized with depression in family practice. A cohort of depressed patients had been identified from a morbidity registry13 or practice,38 and was followed up longitudinally using the patients’ records (and registry13). The follow-up started on the date the diagnosis was made for the first time13 or the first time in the practice.38 Outcome in both studies was based on the reference to recurrences on the patients’ records over the entire study period.
The Diagnosis of Depression and Diagnostic Criteria
There was only one study using specific family practice criteria. For that study E-list criteria (the first classification for general practice, developed in the United Kingdom and derived from the ICD) was used initially and was later replaced with ICHPPC-2 criteria.45,13 In all the other studies DSM,40,42,43 RDC,38,42-44 or criteria derived from the RDC for use in elderly populations39,41 (obtained with GMS-AGECAT, a computerized diagnostic system for elderly subjects derived from the Geriatric Mental State46) were used. In the family practice study by Widmer and Cadoret38 the symptoms on the records were incorporated in the RDC retrospectively.
Length of Follow-up of Depression
In one study44 the length of the follow-up from the onset of depression was not clear because it did not mention when the depressive episodes occurred. In all other studies the follow-up began with a depressive episode. One study started the follow-up at first episodes only.13 That is the only study in which a time relation between the initial diagnosis and recurrences is presented longitudinally. One other study started at initial episodes in the practice38; all others start at index and recurrent episodes, but the proportions are not clear.
Recurrence or Depression at Follow-up
There were differences in levels of depression included in the outcome results Table 1. The rates of recurrence presented in the community studies ranged between 26% and 47%, and the rates of depression at follow-up were between 9% and 44%. Recurrence rates in the family practice studies ranged from 35% to 40%. The recurrence rate of 35% was calculated retrospectively, relying on the symptoms mentioned on the case records. The recurrence rate of 40% was found by extracting data from a morbidity registry and checking those data against symptoms on the patient records.13 Both studies also presented the number of recurrences. In one study 27% of the followed-up patients had 2 episodes; 6%, 3; and 3%, 4 or more38; in the other the percentages were 16% with 2, 12% with 3, and 12% with 4 or more.13
Treatment
The authors of 4 studies38-40,42 reported on treatment. None described the nature and length of treatment clearly, and treatment was not related to recurrence or depression at follow-up.
Mortality
Data for mortality were given in 4 studies13,39-41 (range of rates=14%-44%). The higher mortality rates refer to the elderly. In 2 studies mortality was similar to the expected rates (compared with a control group in one13 and with the National Mortality Statistics in the other39); in another study41 the rates were significantly higher than in a control group of nondepressed individuals; and in the last study13 the results of a comparison were not discussed. Data on suicide attempts and suicide can be found in only one study.13
Health Status
Two studies reported on disability or self-perceived physical health, but none of the studies used any of the well-known health status measurement instruments. One study39 reported on disability levels (with a modified version of the American Resources and Services) There was significantly more moderate to severe impairment among the depressed than among the recovered cases, as rated by a clinician; also, it is not clear whether this rating refers to the initial assessment or the follow-up. In the other study40 46% of the elderly patients reported poor health status, but it is not clear which instrument was used; no relation was found between outcome and perceived health status.
Total Rates of Recurrence or Depression at Follow-up
Table 2 shows the total rates of recurrence or depression at follow-up, adding up the results of minor and major depression when possible. The recurrence rates of the populations in which no specific age group was followed-up ranged from 30% to 40%. These studies had indications for higher recurrence rates in the younger age groups. The community studies of depression in the elderly and young adults reported outcome as depression at follow-up. The rates of depression at follow-up in 2 of the studies of the elderly39,41 were relatively high. One of these studies41 only reported on major depression at follow-up. This was also the case in the study with young adults.42
Discussion
Studies of the long-term outcome of depression in the community and in primary care are scarce and difficult to compare, and methodologic shortcomings hamper their generalizability.
Our data suggest that overall recurrence rates in the community and in family practice vary between 30% and 40%. The relationship between treatment and long-term outcome remains unclear, because none of the studies were controlled trials for treatment or looked into this matter adequately. This also applies to the patient’s qualitative experience. Almost all studies exclusively report physician-diagnosed recurrence or positive scores on diagnostic instruments of depression at follow-up.
Recurrence rates of 30% to 40% indicate that the prognosis for depression in community and family practice is not as poor as in psychiatry. In psychiatric settings much higher recurrence rates are found, with percentages of up to almost 90% depending on the length of follow-up and the setting.9,10,37 Prognosis seems to be related to age, with young adults and the elderly having poorer prognoses. In the study on young adults, between 30% and 40% of the patients had a major depressive disorder at follow-up, but those results did not include rates of minor depression. They also did not include recurrences between the follow-up interviews, and thus it is likely that recurrence rates were higher. Higher recurrence rates were also found in the younger age groups in 2 of the community studies.43,44 All the studies performed exclusively with the elderly reported depression at follow-up and gave a poor prognosis.39-41 This was not confirmed in the 2 community studies involving various ages and one general practice study.13,43,44 The higher rates in studies on the elderly might be explained by the use of diagnostic instruments more sensitive to detect depression specifically in that age group. Another explanation might be that differences were not found because of the relatively small number of elderly persons present in the other studies.
Limitations of the Studies
In the community studies subjects were identified with screening instruments. Therefore, a number of false-positive diagnoses may have been included in these studies that biased outcome results.47 This risk was minimized by using interviews in addition to the screening instruments. Another important limitation is the risk of missing part of the information about recurrences in the intervals between assessments. This risk is less in studies that include more than one follow-up assessment, as was done in 2 of the included studies,41,42 and also in studies where data about the interval are retrieved using information based on patients’ recall.40,43 Recall is known to introduce bias by not always giving sufficient details after longer periods of time.48
In the family practice studies where the specificity of the diagnosis is usually high,49,50 outcome results may have been biased because the results of undetected or misdiagnosed patients with depression are missing. In both studies the information was retrieved from the case records. Thus, accuracy depended on the completeness of the physicians’ notes. In one of the studies the case records were used in addition to data from a morbidity registry in which physicians were trained regularly to use criteria for diagnosis. Although suicide data can be found in this study, patients who left the practices or died within 10 years were excluded, and outcome results should be viewed taking that into consideration.
An important shortcoming is that most studies started at first or recurrent episodes, so it is not possible to give an exact percentage of single-episode depression versus recurrent illness. A description of the longitudinal course starting at first diagnosis is also not possible. Even in the one study starting at the first episode we can only draw conclusions about recurrence rates after diagnosis, because we have no certainty that the first diagnosis was in fact related to the first depressive episode. Other shortcomings are the small number of patients in the follow-up in 2 studies39,43 and uncertainty about the representativeness of the samples in one of the community studies.39,41,42,44 Since a family history of depression is a risk factor in an individual,51 the population of family members of affectively ill probands cannot be regarded as a representative community sample.44
The validity of the recurrence rates mentioned in the articles is difficult to assess because only one author gave confidence intervals of recurrence rates or depression at follow-up.13
Limitations of Our Review
Although we made our choice of inclusion criteria to ensure reasonable comparability, older studies52,53 and those in which data on depression could not be extracted from a broader variety of mental illness in family practice54 were excluded. We also made the choice to describe a limited number of outcomes, but the small number of studies included and the variety within the studies did not allow a review of more outcome results.
Although the calculation of a total recurrence rate may be criticized, we think that the fluctuating nature of depression justifies this procedure.
Conclusions
There are large gaps in the available knowledge about long-term outcome of depression in primary care, and future studies are required to fill in these gaps. We recommend the following:
- The outcome of all types of depression should be evaluated in prospective studies, with a follow-up of at least 5 years, of representative samples in both community and primary care.
- Continuous morbidity registration should be used. With this aim, data meeting fixed criteria that have been established beforehand should be collected longitudinally.
- Studies should include naturalistic follow-up and relate treatment to outcome.
- Quality-of-life assessment should be included.
Recommendations for clinical practice
Family physicians can reassure patients with depression by telling them that although the long-term outcome of the illness is not completely clear, there are indications that the majority of patients with depression do not have a poor prognosis. As the long term risk of recurrence seems to be approximately 40%, most patients in primary care settings only have 1 episode of depression. This information might aid a patient’s recovery.
1. Froom J, Aoyama H, Hermoni D, et al. Depressive disorders in three primary care populations: United States, Israel, Japan. Fam Pract 1995;12:274-78
2. Ormel J, VonKorff M, Ustun TB, et al. Common mental disorders and disability across cultures: results from the WHO Collaborative Study on Psychological Problems in General Health Care. JAMA 1994;272:1741-48
3. Robins LN. Lifetime prevalence of specific psychiatric disorders in three sites. Arch Gen Psychiatry 1984;41:949-58
4. Williams JW, Kerber CA, Mulrow CD, et al. Depressive disorders in primary care: prevalence, functional disability, and identification. J Gen Intern Med 1995;10:7-12
5. Wells KB, Stewart A, Hays RD, et al. The functioning and well-being of depressed patients: results from the Medical Outcomes Study. JAMA 1989;262:914-19
6. Katon W, Schulberg H. Epidemiology of depression in primary care. Gen Hosp Psychiatry 1992;14:237-47
7. Coryell W, Turvey C, Endicott J, et al. Bipolar I affective disorder: predictors of outcome after 15 years. J Affect Disord 1998;50:109-16
8. Paykel ES. Historical overview of outcome of depression. Br J Psychiatry Suppl 1994;26:6-8
9. Angst J, Preisig M. Course of a clinical cohort of unipolar, bipolar and schizoaffective patients: results of a prospective study from 1959 to 1985. Schweiz Archiv Neurol Psychiatr 1995;146:5-16
10. Piccinelli M, Wilkinson G. Outcome of depression in psychiatric settings. Br J Psychiatry 1994;164:297-304
11. Lavori PW, Keller MB, Mueller TI, et al. Recurrence after recovery in unipolar MDD: an observational follow-up study of clinical predictors and somatic treatment as a mediating factor. Int J Methods Psychiatric Res 1994;4:211-29
12. Blacker CV, Clare AW. Depressive disorder in primary care. Br J Psychiatry 1987;150:737-51
13. van Weel-Baumgarten E, van den Bosch W, van den Hoogen H, et al. Ten year follow-up of depression after diagnosis in general practice. Br J Gen Pract 1998;48:1643-46
14. Watts CA. Depressive disorders in the community: the scene in Great Britain, 1965. J Clin Psychiatry 1984;45:70-77
15. Klinkman MS, Schwenk TL, Coyne JC. Depression in primary care—more like asthma than appendicitis: the Michigan Depression Project. Can J Psychiatry 1997;42:966-73.
16. Lin EH, Simon GE, Katon WJ, et al. Can enhanced acute-phase treatment of depression improve long-term outcomes? A report of randomized trials in primary care. Am J Psychiatry 1999;156:643-45
17. Ormel J, Oldehinkel T, Brilman E, et al. Outcome of depression and anxiety in primary care: a three-wave 3 1/2-year study of psychopathology and disability. Arch Gen Psychiatry 1993;50:759-66
18. Kupfer DJ. Long-term treatment of depression. J Clin Psychiatry 1991;52 (suppl):28-34.
19. Prien RF. Efficacy of continuation drug therapy of depression and anxiety: issues and methodologies. J Clin Psychopharmacol 1990;10:86S-90S.
20. Tan RS, Barlow RJ, Abel C, et al. The effect of low dose lofepramine in depressed elderly patients in general medical wards. Br J Clin Pharmacol 1994;37:321-24
21. Thompson C, Thompson CM. The prescribing of antidepressants in general practice: II. A placebo-controlled trial of low-dose dothiepin. Human Psychopharma 1989;191-204.
22. Wernicke JF, Dunlop SR, Dornseif BE, et al. Low-dose fluoxetine therapy for depression. Psychopharmacol Bull 1988;24:183-88
23. Paykel ES, Freeling P, Hollyman JA. Are tricyclic antidepressants useful for mild depression? A placebo controlled trial. Pharmacopsychiatry 1988;21:15-18
24. Hollyman JA, Freeling P, Paykel ES, et al. Double-blind placebo-controlled trial of amitriptyline among depressed patients in general practice. J R Coll Gen Pract 1988;38:393-97
25. Mynors WL, Gath D. Predictors of treatment outcome for major depression in primary care. Psychol Med 1997;27:731-36
26. Scott AI, Freeman CP. Edinburgh primary care depression study: treatment outcome, patient satisfaction, and cost after 16 weeks. BMJ 1992;304:883-87
27. Schulberg HC, Block MR, Madonia MJ, et al. Treating major depression in primary care practice: eight-month clinical outcomes. Arch Gen Psychiatry 1996;53:913-19
28. Pincus T, Stein CM. Why randomized controlled clinical trials do not depict accurately long-term outcomes in rheumatoid arthritis: some explanations and suggestions for future studies. Clin Exp Rheumatol 1997;15 (suppl):S27-38.
29. Wolfe F, Hawley DJ, Cathey MA. Clinical and health status measures over time: prognosis and outcome assessment in rheumatoid arthritis. J Rheumatol 1991;18:1290-97.
30. de Abajo FJ, Rodriguez LA, Montero D. Association between selective serotonin reuptake inhibitors and upper gastrointestinal bleeding: population based case-control study. BMJ 1999;319:1106-09
31. Po AL. Antidepressants and upper gastrointestinal bleeding. BMJ 1999;319:1081-82
32. Mackay FR, Dunn NR, Martin RM, et al. Newer antidepressants: a comparison of tolerability in general practice. Br J Gen Pract 1999;49:892-96
33. Sampson E, Warner JP. Serotonin syndrome: potentially fatal but difficult to recognize. Br J Gen Pract 1999;49:867-68
34. Tondo L, Laddomada P, Serra G, et al. Rapid cyclers and antidepressants. Int Pharmacopsychiatry 1981;16:119-23
35. Classification committee of WONCA. ICHPPC-2-defined (International classification of Health Problems in Primary Care). 3rd ed. Oxford, England: Oxford University Press; 1983.
36. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th edition. Washington, DC: American Psychiatric Association; 1994.
37. Van Londen L, Molenaar RP, Goekoop JG, et al. Three- to 5-year prospective follow-up of outcome in major depression. Psychol Med 1998;28:731-35
38. Widmer RB, Cadoret RJ. Depression in primary care: changes in pattern of patient visits and complaints during a developing depression. J Fam Pract 1978;7:293-302
39. Kua EH. The depressed elderly Chinese living in the community: a five-year follow-up study. Int J Geriatr Psychiatry 1993;8:427-30
40. Kivela SL, Kongas SP, Kesti E, et al. Five-year prognosis for depression in old age. Int Psychogeriatr 1994;6:69-78
41. Sharma VK, Copeland JR, Dewey ME, et al. Outcome of the depressed elderly living in the community in Liverpool: a 5-year follow-up. Psychol Med 1998;28:1329-37
42. Angst J, Merikangas K. The depressive spectrum: diagnostic classification and course. J Affect Disord 1997;45:31-39
43. Eaton WW, Anthony JC, Gallo J, et al. Natural history of Diagnostic Interview Schedule/DSM-IV major depression: the Baltimore Epidemiologic Catchment Area follow-up. Arch Gen Psychiatry 1997;54:993-99
44. Coryell W, Endicott J, Keller MB. Predictors of relapse into major depressive disorder in a nonclinical population. Am J Psychiatry 1991;148:1353-58
45. Logan WPD, Cushion A. Morbidity statistics from general practice. Vol 1. London, England: Her Majesty’s Stationary Office; 1954.
46. Copeland JR, Dewey ME, Griffiths-Jones HM. A computerized psychiatric diagnostic system and case nomenclature for elderly subjects: GMS and AGECAT. Psychol Med 1986;16:89-99
47. Nagel R, Lynch D, Tamburrino M. Validity of the medical outcomes study depression screener in family practice training centers and community settings. Fam Med 1998;30:362-65
48. Andrews G, Anstey K, Brodaty H, et al. Recall of depressive episode 25 years previously. Psychol Med 1999;29:787-91
49. Wright AF. Should general practitioners be testing for depression? Br J Gen Pract 1994;44:132-35
50. Van Weel C. Validating long term morbidity recording. J Epidemiol Community Health 1995;49 (suppl):29-32.
51. Merikangas KR, Wicki W, Angst J. Heterogeneity of depression: classification of depressive subtypes by longitudinal course. Br J Psychiatry 1994;164:342-48
52. Murphy JM, Olivier DC, Sobol AM, et al. Diagnosis and outcome: depression and anxiety in a general population. Psychol Med 1986;16:117-26
53. Hagnell O, Lanke J, Rorsman B. Suicide and depression in the male part of the Lundby study: changes over time during a 25-year observation period. Neuropsychobiology 1982;8:182-87
54. Lloyd KR, Jenkins R, Mann A. Long-term outcome of patients with neurotic illness in general practice. BMJ 1996;313:26-28
1. Froom J, Aoyama H, Hermoni D, et al. Depressive disorders in three primary care populations: United States, Israel, Japan. Fam Pract 1995;12:274-78
2. Ormel J, VonKorff M, Ustun TB, et al. Common mental disorders and disability across cultures: results from the WHO Collaborative Study on Psychological Problems in General Health Care. JAMA 1994;272:1741-48
3. Robins LN. Lifetime prevalence of specific psychiatric disorders in three sites. Arch Gen Psychiatry 1984;41:949-58
4. Williams JW, Kerber CA, Mulrow CD, et al. Depressive disorders in primary care: prevalence, functional disability, and identification. J Gen Intern Med 1995;10:7-12
5. Wells KB, Stewart A, Hays RD, et al. The functioning and well-being of depressed patients: results from the Medical Outcomes Study. JAMA 1989;262:914-19
6. Katon W, Schulberg H. Epidemiology of depression in primary care. Gen Hosp Psychiatry 1992;14:237-47
7. Coryell W, Turvey C, Endicott J, et al. Bipolar I affective disorder: predictors of outcome after 15 years. J Affect Disord 1998;50:109-16
8. Paykel ES. Historical overview of outcome of depression. Br J Psychiatry Suppl 1994;26:6-8
9. Angst J, Preisig M. Course of a clinical cohort of unipolar, bipolar and schizoaffective patients: results of a prospective study from 1959 to 1985. Schweiz Archiv Neurol Psychiatr 1995;146:5-16
10. Piccinelli M, Wilkinson G. Outcome of depression in psychiatric settings. Br J Psychiatry 1994;164:297-304
11. Lavori PW, Keller MB, Mueller TI, et al. Recurrence after recovery in unipolar MDD: an observational follow-up study of clinical predictors and somatic treatment as a mediating factor. Int J Methods Psychiatric Res 1994;4:211-29
12. Blacker CV, Clare AW. Depressive disorder in primary care. Br J Psychiatry 1987;150:737-51
13. van Weel-Baumgarten E, van den Bosch W, van den Hoogen H, et al. Ten year follow-up of depression after diagnosis in general practice. Br J Gen Pract 1998;48:1643-46
14. Watts CA. Depressive disorders in the community: the scene in Great Britain, 1965. J Clin Psychiatry 1984;45:70-77
15. Klinkman MS, Schwenk TL, Coyne JC. Depression in primary care—more like asthma than appendicitis: the Michigan Depression Project. Can J Psychiatry 1997;42:966-73.
16. Lin EH, Simon GE, Katon WJ, et al. Can enhanced acute-phase treatment of depression improve long-term outcomes? A report of randomized trials in primary care. Am J Psychiatry 1999;156:643-45
17. Ormel J, Oldehinkel T, Brilman E, et al. Outcome of depression and anxiety in primary care: a three-wave 3 1/2-year study of psychopathology and disability. Arch Gen Psychiatry 1993;50:759-66
18. Kupfer DJ. Long-term treatment of depression. J Clin Psychiatry 1991;52 (suppl):28-34.
19. Prien RF. Efficacy of continuation drug therapy of depression and anxiety: issues and methodologies. J Clin Psychopharmacol 1990;10:86S-90S.
20. Tan RS, Barlow RJ, Abel C, et al. The effect of low dose lofepramine in depressed elderly patients in general medical wards. Br J Clin Pharmacol 1994;37:321-24
21. Thompson C, Thompson CM. The prescribing of antidepressants in general practice: II. A placebo-controlled trial of low-dose dothiepin. Human Psychopharma 1989;191-204.
22. Wernicke JF, Dunlop SR, Dornseif BE, et al. Low-dose fluoxetine therapy for depression. Psychopharmacol Bull 1988;24:183-88
23. Paykel ES, Freeling P, Hollyman JA. Are tricyclic antidepressants useful for mild depression? A placebo controlled trial. Pharmacopsychiatry 1988;21:15-18
24. Hollyman JA, Freeling P, Paykel ES, et al. Double-blind placebo-controlled trial of amitriptyline among depressed patients in general practice. J R Coll Gen Pract 1988;38:393-97
25. Mynors WL, Gath D. Predictors of treatment outcome for major depression in primary care. Psychol Med 1997;27:731-36
26. Scott AI, Freeman CP. Edinburgh primary care depression study: treatment outcome, patient satisfaction, and cost after 16 weeks. BMJ 1992;304:883-87
27. Schulberg HC, Block MR, Madonia MJ, et al. Treating major depression in primary care practice: eight-month clinical outcomes. Arch Gen Psychiatry 1996;53:913-19
28. Pincus T, Stein CM. Why randomized controlled clinical trials do not depict accurately long-term outcomes in rheumatoid arthritis: some explanations and suggestions for future studies. Clin Exp Rheumatol 1997;15 (suppl):S27-38.
29. Wolfe F, Hawley DJ, Cathey MA. Clinical and health status measures over time: prognosis and outcome assessment in rheumatoid arthritis. J Rheumatol 1991;18:1290-97.
30. de Abajo FJ, Rodriguez LA, Montero D. Association between selective serotonin reuptake inhibitors and upper gastrointestinal bleeding: population based case-control study. BMJ 1999;319:1106-09
31. Po AL. Antidepressants and upper gastrointestinal bleeding. BMJ 1999;319:1081-82
32. Mackay FR, Dunn NR, Martin RM, et al. Newer antidepressants: a comparison of tolerability in general practice. Br J Gen Pract 1999;49:892-96
33. Sampson E, Warner JP. Serotonin syndrome: potentially fatal but difficult to recognize. Br J Gen Pract 1999;49:867-68
34. Tondo L, Laddomada P, Serra G, et al. Rapid cyclers and antidepressants. Int Pharmacopsychiatry 1981;16:119-23
35. Classification committee of WONCA. ICHPPC-2-defined (International classification of Health Problems in Primary Care). 3rd ed. Oxford, England: Oxford University Press; 1983.
36. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th edition. Washington, DC: American Psychiatric Association; 1994.
37. Van Londen L, Molenaar RP, Goekoop JG, et al. Three- to 5-year prospective follow-up of outcome in major depression. Psychol Med 1998;28:731-35
38. Widmer RB, Cadoret RJ. Depression in primary care: changes in pattern of patient visits and complaints during a developing depression. J Fam Pract 1978;7:293-302
39. Kua EH. The depressed elderly Chinese living in the community: a five-year follow-up study. Int J Geriatr Psychiatry 1993;8:427-30
40. Kivela SL, Kongas SP, Kesti E, et al. Five-year prognosis for depression in old age. Int Psychogeriatr 1994;6:69-78
41. Sharma VK, Copeland JR, Dewey ME, et al. Outcome of the depressed elderly living in the community in Liverpool: a 5-year follow-up. Psychol Med 1998;28:1329-37
42. Angst J, Merikangas K. The depressive spectrum: diagnostic classification and course. J Affect Disord 1997;45:31-39
43. Eaton WW, Anthony JC, Gallo J, et al. Natural history of Diagnostic Interview Schedule/DSM-IV major depression: the Baltimore Epidemiologic Catchment Area follow-up. Arch Gen Psychiatry 1997;54:993-99
44. Coryell W, Endicott J, Keller MB. Predictors of relapse into major depressive disorder in a nonclinical population. Am J Psychiatry 1991;148:1353-58
45. Logan WPD, Cushion A. Morbidity statistics from general practice. Vol 1. London, England: Her Majesty’s Stationary Office; 1954.
46. Copeland JR, Dewey ME, Griffiths-Jones HM. A computerized psychiatric diagnostic system and case nomenclature for elderly subjects: GMS and AGECAT. Psychol Med 1986;16:89-99
47. Nagel R, Lynch D, Tamburrino M. Validity of the medical outcomes study depression screener in family practice training centers and community settings. Fam Med 1998;30:362-65
48. Andrews G, Anstey K, Brodaty H, et al. Recall of depressive episode 25 years previously. Psychol Med 1999;29:787-91
49. Wright AF. Should general practitioners be testing for depression? Br J Gen Pract 1994;44:132-35
50. Van Weel C. Validating long term morbidity recording. J Epidemiol Community Health 1995;49 (suppl):29-32.
51. Merikangas KR, Wicki W, Angst J. Heterogeneity of depression: classification of depressive subtypes by longitudinal course. Br J Psychiatry 1994;164:342-48
52. Murphy JM, Olivier DC, Sobol AM, et al. Diagnosis and outcome: depression and anxiety in a general population. Psychol Med 1986;16:117-26
53. Hagnell O, Lanke J, Rorsman B. Suicide and depression in the male part of the Lundby study: changes over time during a 25-year observation period. Neuropsychobiology 1982;8:182-87
54. Lloyd KR, Jenkins R, Mann A. Long-term outcome of patients with neurotic illness in general practice. BMJ 1996;313:26-28
Providing Primary Care for Long-Term Survivors of Childhood Acute Lymphoblastic Leukemia
Acute lymphoblastic leukemia (ALL), the most common childhood malignancy, accounts for almost one fourth of childhood cancers.1 The incidence of ALL has shown a moderate increase in the past 20 years. It is generally considered a cancer of younger children, with a peak incidence between the ages of 2 and 5 years. It is approximately 30% more common in boys than girls and approximately twice as common in white children as in black children. Improvements in ALL treatment during the past 20 years have increased the overall survival rate to approximately 80%. Thus, success in “curing” this childhood disease has resulted in a growing population of long-term survivors.
Since it is anticipated that the majority of long-term survivors of childhood ALL will seek health care from primary care physicians, it is important to understand the potential health problems that these patients may experience secondary to their cancer treatment.2-4 However, there are no articles in peer-reviewed family practice journals concerning the long-term follow-up of survivors of childhood ALL. Our clinical review briefly describes the evolution of the treatment for ALL, potential late effects of treatment, and recommendations for screening asymptomatic long-term survivors. Because this field of investigation is rapidly advancing and much of the available information is from cross-sectional and small cohort studies, these recommendations should not be viewed as a set of guidelines. Instead, our review is intended to contribute a foundation for primary care physicians providing longitudinal health care for ALL survivors while highlighting the areas needing further investigation. Also, because of the evolving changes in treatment protocols—and thus in potential late effects—it is essential to frequently communicate with our colleagues who specialize in the treatment of children with cancer.
Evolution of treatment for childhood all
During the 1940s childhood leukemias had a uniformly rapid fatal course over a short period of time, thus the designation of the term “acute.”5 In the late 1940s, Farber and colleagues6 found that aminopterin (a folic acid antagonist) could induce temporary remissions in leukemia. This discovery opened the era of clinical investigation into the uses of combined chemotherapy in treating childhood ALL Figure 1. The use of antimetabolite therapy for prolonged periods started in the late 1950s and early 1960s and suggested that it was possible for children to have an extended period of remission and possibly be cured. The addition of anthracyclines such as daunorubicin in the 1970s and the discovery that the enzyme L-asparaginase was useful in ALL therapy for depleting cells of the essential amino acid L-asparagine further boosted the ability to induce and sustain remission.7
A significant factor in morbidity and mortality from childhood ALL was the development of leukemia within the central nervous system (CNS). Left untreated, more than half the children with ALL developed leukemia in the CNS, even when bone marrow remission was sustained. In most patients, CNS relapse was followed by bone marrow relapse. Prophylactic radiation to the head and spine, introduced in the early 1970s, significantly decreased the incidence of CNS leukemia and resulted in significant advancement in long-term survival. However, in the early 1980s—as a consequence of the appreciation of neurodevelopmental delays and cognitive dysfunction secondary to relatively higher-dose (24 Gy) cranial irradiation (CRT), different methods of CNS treatment and prophylaxis evolved, either using lower-dose CRT (18 Gy), intensification of systemic methotrexate (MTX) dosaging, or intrathecal medications.8-11
Current treatment regimens divide therapy into remission induction, consolidation and CNS prophylaxis, and maintenance or continuous treatment. Induction chemotherapy (aimed at an initial reduction in blast cell percentage in the bone marrow to 5% or lower) consists of a 1-month schedule of vincristine, prednisone, and L-asparaginase alone or with other agents. Following induction, a consolidation phase consisting of an intensified period of treatment combines the use of antimetabolites and other agents with intrathecal chemotherapy for CNS prophylaxis. Maintenance therapy continues for a period of approximately 2 years and relies heavily on the use of methotrexate and 6-mercaptopurine. During the past 2 decades, recognized differences in the phenotype of the leukemic cells have resulted in protocol modifications to improve outcome and reduce toxicity. Increasingly, the T-cell phenotype of childhood ALL has been treated more effectively with intensified regimens that include cyclophosphamide, cytarabine, and anthracylines.12,13
Late effects of treatment for childhood all
A late effect is defined as any chronic or late occurring physical or psychosocial outcome persisting or developing more than 5 years after diagnosis of the cancer. In this section we describe potential late effects in order from more common or serious health problems to less common or serious ones Table 1. Many of these late effects may have long asymptomatic intervals before end-stage disease or serious health outcomes, such as survivors with hepatitis C who develop cirrhosis or those with a late-onset cardiomyopathy who present in congestive heart failure. Included in each section is a discussion about the screening tests commonly used in long-term follow-up programs that include asymptomatic survivors4Table 2. It should be stressed that the value of most of these tests has not been studied in this population in a prospective or a well-designed retrospective manner with adequate sample sizes, which limits the strength of the recommendations. Clinicians should be selective in ordering tests and providing preventive services and should actively incorporate the patient’s concerns and fears when arriving at an individualized decision on whether to perform a test. Figure 2 is a compilation of information pertinent to the follow-up of a survivor of childhood ALL, provided as a single-page template for clinical use.
Because bone marrow transplantation (BMT) is a relatively new therapy affecting a much smaller number of ALL survivors, our review does not include the late effects related to total body irradiation and BMT.
Cognitive dysfunction and performance at school and work
As described in the section on the evolution of treatment, 24 Gy CRT is associated with cognitive dysfunction. A meta-analysis of more than 30 retrospective and prospective studies established that 24 Gy CRT in combination with MTX resulted in a mean decrease of 10 points in full-scale intelligence quotient (IQ).9 Verbal scores were affected more than performance IQ, and changes were noted to be progressive. Although more than half the patients had mild to moderate learning problems, the outcomes were highly variable, and some patients experienced 20- to 30-point losses, while others had no discernable changes.9,14 Deficits have been noted in measures of visual-spatial abilities, attention-concentration, nonverbal memory, and somatosensory functioning.8-10,15-20 Studies have also shown that girls and patients treated with CRT before the age of 4 years are at significantly higher risk. Neuropathologic changes resulting from 24 Gy CRT include leukoencephalopathy, mineralizing microangiopathy, subacute necrotizing leukomyelopathy, and intracerebral calcifications, commonly with subsequent cerebral atrophy and microcephally.21,22
Treatment with 18 Gy CRT in combination with chemotherapy also affects cognition, though not as profoundly as with 24 Gy CRT. In a retrospective study of children with ALL, randomized by risk group to receive either 18 Gy CRT with chemotherapy or chemotherapy alone, 66 survivors were subsequently tested using several cognitive measures.23 Girls who were treated with CRT/chemotherapy had a mean IQ 9 points lower than those treated with chemotherapy alone. All patients had impairments in verbal coding and short-term memory regardless of CRT use or MTX dose, suggesting that another agent such as glucocorticoids may be responsible. Other small prospective and retrospective studies have found a mild decrease in full-scale IQ in patients treated with 18 Gy CRT/chemotherapy, although subanalysis generally showed that changes were only significant for girls and patients treated at a younger age.24-27
Recent studies suggest that neurodevelopmental outcomes for survivors treated with chemotherapy alone are generally positive.28 An analysis of 30 survivors whose condition was diagnosed before the age of 12 months showed no decrease in 6 cognitive and motor indices and no sex differences.29 Though full-scale IQ was normal, Brown and colleagues30 reported that girls had significantly decreased nonverbal scores in a study of 47 ALL survivors. Fine motor disturbances and manual dexterity difficulties, which may compound learning difficulties, have been seen in 25% to 33% of ALL survivors evaluated in 2 small cross-sectional studies.31,32 Changes in cerebellar-frontal subsystems that correlate with neuropsychological deficits have also been seen in ALL patients treated with chemotherapy alone.33
The Children’s Cancer Group investigated the impact of treatment on scholastic performance of 593 adult survivors, compared with 409 sibling controls.34 Patients treated with 24 Gy CRT were more likely to enter special education or learning-disabled programs, with relative risks of 4.1 and 5.3, respectively. Previous treatment with 18 Gy CRT had less impact, with a relative risk of 4.0 to enter a special education program but no increased risk of entering a learning-disabled program. Patients treated with CRT (18 or 24 Gy) were just as likely to enter gifted and talented programs as their sibling controls. In general, survivors were as likely to finish high school and enter college as controls, but those treated with 24 Gy or treated before the age of 6 years were less likely to enter college. There were no sex differences in educational achievements.
There are no studies that explore problems in job acquisition, promotion, and retention for ALL survivors with evidence of cognitive dysfunction. Abstract thinking abilities in higher-level decision making may be problematic for some ALL survivors, particularly those treated with 24 Gy CRT. Further study is warranted, particularly in evaluating methods to assist at-risk survivors in developing job skills and applying for a job.
Obesity, physical inactivity, and risk of premature cardiovascular disease
Several retrospective cohort and cross-sectional studies have shown an increased incidence and prevalence of obesity in ALL survivors. Early studies suggested that the resulting obesity was secondary to CRT, with 38% to 57% of the survivors having a body mass index (BMI) >2 standard deviations (SDs) above the norm at the time of attainment of final height.35-38 Two recent cross-sectional studies suggest that the increased prevalence of obesity may be due to other factors. Van Dongen-Melman and coworkers39 compared the weight gain and BMI of 113 ALL survivors who had received CRT/chemotherapy or chemotherapy alone and found that children treated with a combination of prednisone and dexamethasone had the highest prevalence of obesity (44%).39 Talvensaari and colleagues40 evaluated 50 childhood cancer survivors with a median age of 18 years (including 28 ALL patients) and found an increased prevalence of obesity in survivors that was not associated with CRT.
Obesity in ALL survivors may be due in part to reduced physical activity. In a small cross-sectional study with sibling controls, ALL survivors had decreased activity levels and total daily energy expenditures that correlated with their percentage of body fat.41 Maximal and submaximal exercise capacity were reduced in another cross-sectional study.42 Similarly, in a study of 53 ALL survivors with a longer interval from ALL diagnosis (mean=10.5 years), 25% and 31%, respectively, were unable to reach normal maximal oxygen uptake and normal oxygen uptake at the anaerobic threshold.43
Changes in gross motor skills may also affect the physical activity level of ALL survivors. Balance, strength, running speed and agility, and hand grip strength were decreased in a cohort of 36 ALL survivors with a median age of 9.3 years.44 In a follow-up of this cohort, Wright and coworkers45 reported that the ALL survivors had significantly less active and passive dorsiflexion range of motion of the ankle than did controls. Younger age at diagnosis and female sex were significant predictors, while treatment with CRT did not increase risk. These studies suggest that ALL survivors should be assessed for gross motor deficits that might alter exercise choices.
In the general population, obesity and physical inactivity are risk factors for cardiovascular disease. Obesity (an especially important risk factor during young adulthood) enhances the development of hypertension, dyslipidemia, and insulin resistance.46-48 Because the median age of ALL survivors is still relatively young, there are no cohort or case-control studies evaluating the treatment-related risk of premature onset of coronary artery disease. Talvensaari and coworkers40 reported that 50 childhood cancer survivors (including 28 ALL survivors) had an increased risk of fasting hyperinsulinemia and reduced high-density lipoprotein (HDL) cholesterol compared with 50 age- and sex-matched controls. Eight of the cancer survivors with reduced spontaneous growth hormone (GH) secretion (4/8 had received CRT) had obesity, hyperinsulinemia, and reduced HDL cholesterol, fitting the criteria for cardiac dysmetabolic syndrome, a clustering of metabolic problems associated with a markedly increased risk of cardiovascular disease.49
Studies of noncancer populations may shed light on the cardiovascular risk of ALL survivors with GH deficiency. Hypopituitarism with GH deficiency in adults is associated with increased vascular mortality.50-52 Adults with GH deficiency also have an increased prevalence of dyslipidemia53,54 and insulin resistance,55 that may improve with GH therapy.56,57
Counseling on the benefits of proper diet and exercise is an important component of long-term care for ALL survivors. Periodic analysis of lipoproteins has not been prospectively studied in ALL survivors, but the US Preventive Services Task Force states that adolescents and young adults who have major risk factors for cardiovascular disease should be screened.58
Psychosocial well-being of all survivors
The long-term psychosocial welfare of ALL survivors is complex. A population-based sibling-matched control study of 93 ALL survivors who were at least 15 years postdiagnosis showed no difference in quality of life or mental health.59 Similarly, no differences were found in symptoms of anxiety and posttraumatic stress in 130 leukemia survivors and 155 controls.60 In contrast, a large cooperative study of the Children’s Cancer Group and the National Institutes of Health evaluated 580 adult survivors and 396 sibling controls and reported that survivors had greater negative mood and reported more tension, depression, anger, and confusion.61 Female, minority, and unemployed survivors reported the highest total mood disturbance. Issues related to late effects, especially cognitive dysfunction, obesity, and physical inactivity, may have an impact on the mental health of survivors.
Few data are available on the risk behavior of ALL survivors. In a cohort study of 592 young adult ALL survivors and 409 sibling controls, Tao and colleagues62 reported that ALL survivors were less likely to start smoking, but once they started they were no more likely to quit than their siblings. Fourteen percent of the ALL survivors were smokers. Although no prospective studies have evaluated the effect of smoking on the incidence and severity of late effects of ALL treatment, it will have an impact on survivors with cardiovascular risk factors, restrictive pulmonary disease, and osteopenia. Counseling on smoking cessation is imperative in the long-term health care of ALL survivors.
Osteopenia and osteoporosis
Several well-designed small to medium-size cross-sectional studies of childhood cancer survivors63-65 and ALL survivors66-71 with median ages at evaluation ranging from 12 to 25 years consistently showed reduction in bone mineral density, bone mass content (BMC), and/or age-adjusted bone mass. Age at diagnosis, interval since treatment, sex, and cumulative dosages of MTX and corticosteroids have not been consistently associated with reduction in bone mass. In contrast, CRT has consistently been identified as a risk factor, although the 3 studies that evaluated GH status showed variation in the relationship of GH deficiency and reduced bone mass.69-71 Impairment of peak bone mass is likely multifactorial in etiology, with predisposing risk factors including altered bone metabolism at the time of onset of leukemia, interference in bone metabolism by corticosteroids and MTX, and impaired bone growth and skeletal maturation caused by pituitary dysfunction/GH deficiency. In an ongoing prospective cohort study, Atkinson and coworkers72 reported that by 6 months of therapy for ALL, 64% of the children had a reduction from baseline measures of BMC, and by the end of 2 years of therapy 83% were osteopenic. Hypomagnesemia due to renal wasting of magnesium after treatment with high-dose corticosteroids and/or aminoglycosides was associated with the progression in changes and may be a key factor in the alteration of bone metabolism.
Reduction in peak bone mass in young adults is a significant risk factor for developing osteoporosis and subsequent fracture, and measures to prevent or reverse bone loss are important. Exercise increases bone density in obese children73 and young adults74 and has recently been shown by meta-analysis75 to prevent or reverse almost 1% of bone loss per year in pre- and postmenopausal women. With ALL survivors likely to be less physically active,41-43 it is essential to counsel them on the benefits of exercise in preventing cardiovascular disease and osteoporosis and help them develop an exercise plan. Additionally, counseling on calcium intake and avoidance of smoking is important. Though bone densitometry has not been an effective screening test for the general population, it has value in high-risk groups.76,77 Prospective randomized trials are needed to evaluate the usefulness and frequency of screening.
GH deficiency
Cross-sectional and longitudinal studies have consistently shown that patients treated with 24 Gy CRT have a decrease in median height of approximately 1 to 1.5 SD score, or 5 to 10 cm.37,78-84 Treatment with 18 Gy CRT85 or chemotherapy alone86,87 affect the final height to a lesser degree. Sklar and coworkers88 reported a change in final height SD score of -0.65 for patients treated with 18 Gy CRT and -0.49 for those treated with chemotherapy alone. Girls and patients treated at a younger age (<5 years) have the greatest growth reduction.37,78,88,89 These changes are thought to be secondary to GH deficiency, resulting in a blunted pubertal growth spurt. The greater the deficiency, the more profound the impairment of growth.90 Brennan and colleagues71 reported a median decrement in final height of 2.1 SD in patients with severe GH deficiency. Treatment with GH in these patients usually results in near normalization of final height.
Though GH therapy is generally stopped when children reach their final height or by the age of 18 years, deficiency persists. In a small cross-sectional study of 30 ALL survivors, 9 of 15 patients who received 24 Gy CRT (median age=21.4 years) were GH deficient.91 In another cross-sectional analysis of the GH status of 32 ALL survivors (median age=23 years), 21 of 32 were GH deficient, including 9 who were severely deficient.71 The consequences of GH deficiency in adulthood are not well understood. Small studies suggest that GH replacement may improve bone mineral density,92 body composition,93 and quality of life.94
Late onset anthracycline-induced cardiomyopathy
Anthracyclines (notably daunorubicin and doxorubicin) are often used during the induction phase of treatment, with some protocols using moderate to high dosages (Ž350 mg/m2) for high-risk patients. In the past 10 years it has become apparent that childhood cancer patients treated with an anthracycline are at increased risk for developing late-onset cardiomyopathy.95-97 Classically, anthracycline-induced cardiomyopathy is characterized by elevated afterload followed by the development of a dilated thin-walled left ventricle. Over time this can lead to a stiff and poorly compliant left ventricle. Most patients are asymptomatic, but longitudinal studies suggest that a significant proportion will experience progressive changes and may develop congestive heart failure.96,97
Lipshultz and coworkers95 assessed the cardiac status of 115 ALL survivors treated with doxorubicin and found that 65% of those treated with 228 mg/m2 or more had increased left ventricular afterload.95 In a follow-up study, Lipshultz and colleagues96 reported that female sex, younger age at treatment, higher rate of administration of doxorubicin, and cumulative dose of doxorubicin were independent risk factors for the development of altered left ventricular function. Two recent cross-sectional studies suggest that the risk of left ventricular dysfunction is uncommon in children who received cumulative doses less than 300 mg per m2.98,99 In patients treated with cumulative doses less than 270 mg per m2, Sorensen and coworkers98 did not find that female sex and younger age at treatment were risk factors. However, because late cardiac abnormalities were seen in survivors who received only 90 mg per m2, there might be no absolute level below which cardiotoxicity can be prevented.
Because of the concerns about cardiotoxicity, most recent protocols limit anthracycline doses to less than 300 mg per m2, and the use of cardioprotectants such as dexrazoxane in children is under investigation.100 Primary care physicians who provide follow-up care for adult survivors should communicate with oncologists at the treating institution, obtain information about the cumulative dosage of anthracyclines, and discuss long-term screening. Because patients with anthracycline-induced cardiomyopathies generally have a prolonged asymptomatic interval before becoming symptomatic, interval screening is recommended. Optimal timing and testing modality for screening have not been prospectively studied. It is currently recommended that patients who received 300 mg/m2 or more of an anthracycline have a screening echocardiogram every 2 to 3 years to evaluate left ventricular function and shortening fraction.101 It is also important to question patients regarding symptoms of congestive heart failure and to aggressively evaluate them if present.
Hepatitis C
Because most ALL patients receive blood products during therapy, those treated before adequate blood donor screening for hepatitis C was initiated in the early 1990s are at risk for chronic liver disease.102 The prevalence of circulating hepatitis C virus (HCV) ribonucleic acid (RNA) in ALL patients treated in Italy before 1990 ranges from 23% to 49%.103-105 The natural history of ALL survivors with hepatitis C is not well understood. In an Italian study, only 4% of the 56 HCV-RNA seropositive patients had persistently elevated alanine aminotransferase (ALT) over the course of follow-up (mean=17 years).106 For a median of 14 years, 81 survivors of various childhood cancers who were HCV-RNA seropositive were followed, and none showed progression to liver failure.107 In contrast, Paul and coworkers108 reported that 12% of 75 leukemia survivors were anti-HCV positive, 6 of 9 had liver biopsies that showed at least moderate portal inflammation, and half had bridging fibrosis. The Centers for Disease Control and Prevention102 recommend universal screening with anti-HCV for all patients who received blood products before July 1992.
Second malignant neoplasms
Second malignant neoplasms (SMN) are rare in ALL survivors. Thirteen SMNs were diagnosed a median of 6.7 years from ALL diagnosis in a cohort study of 1597 ALL survivors and were associated with the use of radiation (8/13, CNS or head and neck) or chemotherapy (3/13, hematopoietic).109 The cumulative incidence of brain tumors at 20 years in a cohort of 1612 patients was only 1.39%, and more than half of these tumors were either low-grade or benign.110 CNS tumors did not occur in patients treated with chemotherapy alone. Thyroid tumors (predominantly papillary carcinoma) can rarely occur after treatment with cranial or craniospinal irradiation.111,112 Cases of basal cell carcinoma along the spinal axis have also been reported in patients treated with craniospinal irradiation.113,114
Therapy-related acute myelogenous leukemia (t-AML) has been seen following treatment of several childhood cancers, such as ALL and Hodgkin’s and non– Hodgkin’s lymphoma. Cohort studies have shown that agents with leukemogenic potential include alklyating agents and epidophyllotoxin chemotherapy.115-121 Most t-AMLs occur within 8 years of treatment, although cases occurring up to 13 years have been reported.115 Myelodysplasia (especially pancytopenia) generally precedes t-AML. The risk of t-AML following treatment for ALL has been small in 2 cohort studies.109,122 However, because precancerous states (myelodysplastic changes or myelodysplastic syndrome) are usually antecedent to t-AML and early diagnosis may improve outcomes, most institutions recommend obtaining a complete blood count (CBC) with a platelet count and a white blood cell differential in the routine follow-up of ALL survivors who have been treated with an alkylating agent, such as cyclophosphamide, or an epidophyllotoxin, such as etoposide. How long and how frequently a CBC should be obtained in follow-up of an ALL survivor have not been established.
Fertility and reproduction
Most antimetabolite-based treatment protocols for ALL do not affect long-term fertility for men or women.123,124 Craniospinal and abdominal irradiation have been associated with infertility in both sexes but are no longer used for ALL.125-127 Cyclophosphamide (an alkylating agent commonly used in earlier protocols but currently limited to high-risk patients) is also associated with infertility in a dosedependent fashion in both sexes.124,128,129 Resolution of germ-cell dysfunction may occur in men over time, but fertility remains poor for some. Women survivors treated with craniospinal or abdominal irradiation or with cyclophosphamide are at risk for ovarian failure and premature menopause and thus may be at increased risk for osteoporosis. If ovarian failure is suspected, measurement of follicle-stimulating hormone, luteinizing hormone, and serum estradiol and an evaluation by an endocrinologist should be considered.
ALL survivors should know that preliminary studies suggest that treatment is not associated with an increase in congenital malformations of their offspring. In a population-based prospective cohort study an increased rate of congenital defects was not found among 299 adult survivors.130
Ocular abnormalities
Ocular abnormalities in patients treated with CRT are common but generally asymptomatic. Two studies have evaluated the effect of CRT and systemic corticosteroids on the eyes. In a study of 82 ALL survivors who were a mean of 32 months after completion of therapy, 52% of the patients had posterior subcapsular cataracts (PSC) that were generally not visually significant and were not related to age at treatment or gender.131 Eighty-three percent of the 18 patients who had received CRT and systemic corticosteroids were noted to have asymptomatic ocular abnormalities after a median surveillance of 4.1 years.132 Optical densities of the lens were seen in 13 of the 18 of the survivors. There have been no published studies evaluating long-term survivors who received systemic corticosteroids without CRT. Periodic vision and cataract screening is recommended for ALL survivors treated with CRT and should be considered for all survivors of ALL until the risk of prolonged corticosteroid use in childhood is better understood.
Dental and periodontal disease
ALL survivors, especially those treated with CRT, are more likely to have problems with tooth development and be at risk for periodontal disease. In a large retrospective evaluation of dental records, 39.5% of ALL survivors had a dental abnormality, including root stunting (24.4%), microdontia (18.9%), or hypodontia (8.5%).133 Patients who were treated at an age younger than 8 years or who received CRT had more dental abnormalities than the other groups. Similar findings were seen in 2 smaller cross-sectional studies. Abnormal dental development occurred in 95% of all patients and 100% of patients aged 5 years or younger at diagnosis.134 Abnormalities included tooth agenesis, arrested tooth development, microdontia, and enamel dysplasia. Patients who received CRT and those treated at an age younger than 5 years had higher severity scores. Survivors did not have increased caries.135 However, patients younger than 5 years who were treated with cranial irradiation were found to have higher plaque and gingivitis scores, suggesting an increased risk of periodontal disease. A periodic dental and periodontal evaluation is recommended for survivors treated with CRT or at a young age.
Thyroid-related disorders
Following treatment with CRT, hypothyroidism infrequently occurs in ALL survivors through damage to the hypothalamic-pituitary-thyroid axis and/or the direct effect of radiation of the gland. Mohn and colleagues136 reported that 8 of 24 childhood ALL survivors who had received CRT (either 18 or 24 Gy) had either a low basal thyroid-stimulating hormone (TSH) or low peak TSH after thyrotropin-releasing hormone stimulation. Robison and colleagues137 reported that 10% of 175 ALL survivors who had been treated with either 18 or 24 Gy CRT or craniospinal radiation (CS-RT) therapy had a thyroid abnormality, including 5 children with primary hypothyroidism. Pasqualini and colleagues138 reported that 6 of 10 ALL survivors who received either CRT or CS-RT had subtle evidence of primary hypothyroidism. In contrast, 3 cross-sectional studies did not find evidence of primary hypothyroidism in 13, 31, and 64 patients, respectively.1,139-141 Littley and coworkers142 suggest that hypopituitarism is commonly underdiagnosed secondary to the subtle manifestations and insidious progression of disease. Radioactive scatter to the thyroid occurs with CRT in a dose-dependent fashion,143 and ALL survivors treated with either 18 or 24 Gy CRT are at risk for secondary hypothyroidism, thyroid nodules, and thyroid carcinoma.111 Periodic screening with TSH and free T-4 are recommended in ALL survivors treated with CRT. Further screening of the asymptomatic survivor with thyrotropin-releasing hormone stimulation test or ultrasound of the thyroid gland are costly and have not been prospectively studied.
Pulmonary late effects
ALL survivors may have an increased prevalence of mild, generally subclinical, restrictive pulmonary disease. In a small cross-sectional study of ALL survivors, Shaw and coworkers144 reported mild restrictive changes, with patients treated at a younger age at higher risk. Similarly, an analysis of 70 leukemia survivors found mild but significant decreases in forced vital capacity (FVC), forced expiratory volume in 1 second (FEV-1), total lung capacity (TLC), and transfer for carbon monoxide (DLCO).42 Cyclophosphamide, craniospinal irradiation, and a history of chest infections during treatment were independent variables associated with reductions in FEV-1, FVC, and TLC, while anthracyclines and craniospinal irradiation were associated with reductions in DLCO. ALL survivors also had impaired submaximal and maximal exercise capacity. These findings were further supported by analysis of a recent cross-sectional study of 128 patients a median of 7.6 years from therapy completion that reported an increased prevalence of subclinical restrictive pulmonary disease in ALL survivors.145 The long-term consequences and the possible role of smoking or other inhalant exposures need to be studied.
Liver dysfunction (Non-Hepatitis C)
During treatment with methotrexate (especially high-dose ranges) elevations of transaminases are common and generally transient. Two small longitudinal studies following ALL survivors for up to 7 years after completion of therapy did not report any patients with persistent transaminasemia, although Bessho and colleagues noted that 6 of 13 of their ALL survivors had elevated 2-hour postprandial bile acid levels, a more sensitive predictor of liver cirrhosis than transaminase level.146,147 Farrow and coworkers148 found that of 114 survivors who had ALT elevations greater than 5 times the upper limit of normal during therapy, only 17 (14.9%) had elevations persistently. Eight of these patients had chronic HCV infections. Of the remaining 9 patients, only 1 had a persistently elevated transaminase of greater than 2 times normal.
Although there are currently no data evaluating ALL survivors for long-term liver-related complications secondary to methotrexate, studies in patients with juvenile rheumatoid arthritis show that septal and portal fibrosis can occur with weekly low-dose methotrexate treatment of durations as short as 17 months.149 Obesity may be an associated risk factor for the development of cirrhosis in juvenile rheumatoid arthritis patients treated with methotrexate. Because of these potential risks, periodic measurement of ALT is recommended in follow-up of ALL survivors.
Urologic late effects
Cyclophosphamide is a long-recognized cause of hemorrhagic cystitis and a well-established bladder carcinogen. In a retrospective review150 of 314 children with ALL who were treated with cyclophosphamide between 1963 and 1973, 8% developed hemorrhagic cystitis. The frequency of diagnosis was not related to age or sex, but African American children were at higher risk. Cyclophosphamide-induced hemorrhagic cystitis generally presents during therapy, with children complaining of gross hematuria or irritative voiding complaints.151 Concurrent treatment with oral sodium 2-mercapatoethanesulfonate appears to markedly decrease the incidence of cyclophosphamide-induced hemorrhagic cystitis.152 In a nested case-control study of survivors of non–Hodgkin’s lymphoma, Travis and colleagues153 reported that there was a 2.4-fold increased risk of bladder cancer in patients treated with cumulative dosages of cyclophosphamide lower than 20 g. Because of the risk of chronic hemorrhagic cystitis and bladder cancer, ALL survivors treated with cyclophosphamide should have periodic screening urinalysis, and their review of systems should include voiding problems.
Alopecia
Alopecia is a bothersome late effect secondary to treatment with 24 Gy CRT for which there are no available treatments. In a retrospective study of 273 ALL survivors treated with CRT, 10% had alopecia.154
Acknowledgement
Dr Oeffinger received partial support for this work through the American Academy of Family Physicians Foundation Advanced Research Training Grant and the Robert Wood Johnson Foundation Generalist Physician Faculty Scholars Program.
We would like to thank Drs George Buchanan, Melissa Hudson, and Neyssa Marina for their critical review of this manuscript and Ms Laura Snell and Dr James Tysinger for their editing assistance.
1. MA, Ries LAG, Gurney JG, Ross JA. Leukemia. In: Ries LAG, Smith MA, Gurney JG, et al, eds. Cancer incidence and survival among children and adolescents: United States SEER Program 1975-1995, National Cancer Institute, SEER Program. Bethesda, MD: National Institutes of Health; 1999. NIH pub. no. 99-4649.
2. KC, Eshelman DA, Tomlinson GE, Buchanan GR. Programs for adult survivors of childhood cancer. J Clin Oncol 1998;16:2864-67.
3. DS. Transition to adult health care for adolescents and young adults with cancer. Cancer 1993;71:3411-14.
4. KC, Eshelman DA, Tomlinson GE, Buchanan GR, Foster BE. Grading of late effects in young adult survivors of childhood cancer followed in an ambulatory adult setting. Cancer 2000;88:1687-95.
5. H. The natural history of untreated acute leukemia. Ann NU Acad Sci 1954;60:322-58.
6. S, Diamond LK, Mercer RD, et al. Temporary remissions in acute leukemia in children produced by folic acid antagonist 4-aminopteroylglutamic acid (aminopterin). N Engl J Med 1948;238:787-93.
7. L, Gelber R, Cohen H, et al. Four-agent induction and intensive asparaginase therapy for treatment of childhood acute lymphoblastic leukemia. N Engl J Med 1986;315:657-63.
8. LL, Nesbit ME, Jr, Sather HN, Meadows AT, Ortega JA, Hammond GD. Factors associated with IQ scores in long-term survivors of childhood acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1984;6:115-21.
9. P, Waters B, Said J, Stevens M. Cognitive effects of cranial irradiation in leukaemia: a survey and meta-analysis. J Child Psychol Psychiatry 1988;29:839-52.
10. JM, Kornblith AB, Jones D, et al. A comparative study of the long term psychosocial functioning of childhood acute lymphoblastic leukemia survivors treated by intrathecal methotrexate with or without cranial radiation. Cancer 1998;82:208-18.
11. D, Reaman G, Bleyer W, et al. Successful prevention of central nervous (CNS) leukemia without cranial radiation in children with high risk acute lymphoblastic leukemia (ALL): a preliminary report. Proc Am Soc Clin Oncol 1989;8:828.-
12. W, Shuster J, Falletta J, et al. Clinical features and outcome in childhood T-cell leukemia-lymphoma according to stage of thymocyte differentiation: a Pediatric Onoclogy Group study. Blood 1988;72:1891-97.
13. CH, Behm FG, Singh B, et al. Heterogeneity of presenting features and their relation to treatment outcome in 120 children with T-cell acute lymphoblastic leukemia. Blood 1990;75:174-79.
14. M, Azuma E, Ido M, et al. Ten-year survey of the intellectual deficits in children with acute lymphoblastic leukemia receiving chemoimmunotherapy. Med Pediatr Oncol 1993;21:435-40.
15. DP, Urion DK, Tarbell NJ, Niemeyer C, Gelber R, Sallan SE. Late effects of central nervous system treatment of acute lymphoblastic leukemia in childhood are sex-dependent. Dev Med Child Neurol 1990;32:238-48.
16. AE, Aitken K, Eden OB. Computerized psychometry screening in long-term survivors of childhood acute lymphoblastic leukemia. Pediatr Hematol Oncol 1988;5:197-208.
17. H, Huk WJ, Ueberall MA, et al. CNS late effects after ALL therapy in childhood. Part I: Neuroradiological findings in long-term survivors of childhood ALL—an evaluation of the interferences between morphology and neuropsychological performance—the German Late Effects Working Group. Med Pediatr Oncol 1997;28:387-400.
18. JA, Kaleita TA, Noll RB, et al. CNS prophylaxis of childhood leukemia: what are the long-term neurological, neuropsychological, and behavioral effects? Neuropsychol Rev 1991;2:147-77.
19. JA, Waters BG, Cousens P, Stevens MM. Neuropsychological sequelae of central nervous system prophylaxis in survivors of childhood acute lymphoblastic leukemia. J Consult Clin Psychol 1989;57:251-56.
20. J, Horrocks J, Britton PG, Kernahan J. Attentional ability among survivors of leukaemia. Arch Dis Child 1999;80:318-23.
21. AS, Nesbit ME. Neuropsychologic (cognitive) disabilities in long-term survivors of childhood cancer. Pediatrician 1991;18:11-19.
22. RK, Kovnar E, Langston J, et al. Long-term survivors of leukemia treated in infancy: factors associated with neuropsychologic status. J Clin Oncol 1992;10:1095-102.
23. DP, Tarbell NJ, Fairclough D, et al. Cognitive sequelae of treatment in childhood acute lymphoblastic leukemia: cranial radiation requires an accomplice. J Clin Oncol 1995;13:2490-96.
24. CL, Varni JW, Katz ER. Cognitive functioning in long-term survivors of childhood leukemia: a prospective analysis. J Dev Behav Pediatr 1990;11:301-05.
25. M, Brouwers P, Valsecchi MG, Van Veldhuizen A, Huisman J. Association of 1800 cGy cranial irradiation with intellectual function in children with acute lymphoblastic leukaemia. Lancet 1994;344:224-27.
26. E, Anderson V, Godber T, Ekert H. Risk factors for intellectual and educational sequelae of cranial irradiation in childhood acute lymphoblastic leukaemia. Br J Cancer 1996;73:825-30.
27. V, Godber T, Smibert E, Ekert H. Neurobehavioural sequelae following cranial irradiation and chemotherapy in children: an analysis of risk factors. Pediatr Rehabil 1997;1:63-76.
28. Bleyer A. CNS chemoradiotherapy of childhood leukemia: the plot thickens but the ending bodes well. J Clin Oncol 1995;13:2480-82.
29. TA, Reaman GH, MacLean WE, Sather HN, Whitt JK. Neurodevelopmental outcome of infants with acute lymphoblastic leukemia: a Children’s Cancer Group report. Cancer 1999;85:1859-65.
30. RT, Madan-Swain A, Walco GA, et al. Cognitive and academic late effects among children previously treated for acute lymphocytic leukemia receiving chemotherapy as CNS prophylaxis. J Pediatr Psychol 1998;23:333-40.
31. L. Clinical neurological findings of children with acute lymphoblastic leukaemia at diagnosis and during treatment. Eur J Pediatr 1993;152:115-19.
32. HA, Schoemaker MM, Hofte M, et al. Fine motor and handwriting problems after treatment for childhood acute lymphoblastic leukemia. Med Pediatr Oncol 1996;27:551-55.
33. PG, Ciesielski KT, Hart BL, Benzel EC, Sanders JA. Evidence for cerebellar-frontal subsystem changes in children treated with intrathecal chemotherapy for leukemia. Arch Neurol 1998;55:1561-68.
34. R, Fears TR, Robison LL, et al. Educational attainment in long-term survivors of childhood acute lymphoblastic leukemia. JAMA 1994;272:1427-32.
35. P, Chen CH. Prevalence of obesity in children after therapy for acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1986;8:294-99.
36. I, Reilly JJ, Gibson BE, Donaldson MD. Patterns of obesity in boys and girls after treatment for acute lymphoblastic leukaemia. Arch Dis Child 1994;71:147-49.
37. MJ, Ochs JJ, Schriock EA, Carter M. A method of predicting adult height and obesity in long-term survivors of childhood acute lymphoblastic leukemia. J Clin Oncol 1992;10:128-33.
38. M, Didcock E, Davies HA, Ogilvy-Stuart AL, Wales JK, Shalet SM. High incidence of obesity in young adults after treatment of acute lymphoblastic leukemia in childhood. J Pediatr 1995;127:63-67.
39. Dongen-Melman JE, Hokken-Koelega AC, Hahlen K, De Groot A, Tromp CG, Egeler RM. Obesity after successful treatment of acute lymphoblastic leukemia in childhood. Pediatr Res 1995;38:86-90.
40. KK, Lanning M, Tapanainen P, Knip M. Long-term survivors of childhood cancer have an increased risk of manifesting the metabolic syndrome. J Clin Endocrinol Metab 1996;81:3051-55.
41. JT, Bell W, Webb DK, Gregory JW. Daily energy expenditure and physical activity in survivors of childhood malignancy. Pediatr Res 1998;43:607-13.
42. ME, Faragher EB, Jones PH, Woodcock A. Lung function and exercise capacity in survivors of childhood leukaemia. Med Pediatr Oncol 1995;24:222-30.
43. P, Gutjahr P, Stopfkuchen H. Physical performance in long-term survivors of acute leukaemia in childhood. Eur J Pediatr 1998;157:464-67.
44. MJ, Halton JM, Martin RF, Barr RD. Long-term gross motor performance following treatment for acute lymphoblastic leukemia. Med Pediatr Oncol 1998;3:86-90.
45. MJ, Halton JM, Barr RD. Limitation of ankle range of motion in survivors of acute lymphoblastic leukemia: a cross-sectional study. Med Pediatr Oncol 1999;32:279-82.
46. DS, Dietz WH, Srinivasan SR, Berenson GS. The relation of overweight to cardiovascular risk factors among children and adolescents: the Bogalusa Heart Study. Pediatrics 1999;103:1175-82.
47. M, Vanhala P, Kumpusalo E, Halonen P, Takala J. Relation between obesity from childhood to adulthood and the metabolic syndrome: population based study. BMJ 1998;317:319-21.
48. GS, Srinivasan SR, Bao W, et al. Association between multiple cardiovascular risk factors and atherosclerosis in children and young adults. N Engl J Med 1998;338:1650-56.
49. TC, Deedwania PC. The cardiovascular dysmetabolic syndrome. Am J Med 1998;105:77S-82S.
50. T, Bengtsson BA. Premature mortality due to cardiovascular disease in hypopituitism. Lancet 1990;336:285-88.
51. AS, Van’t Hoff W, Jones PJ, Clayton RN. The effect of hypopituitarism on life expectancy. J Clin Endocrinol Metab 1996;81:1169-72.
52. EM, Bulow B, Eskilsson J, Hagmar L. High incidence of cardiovascular disease and increased prevalence of cardiovascular risk factors in women with hypopituitarism not receiving growth hormone treatment: preliminary results. Growth Horm IGF Res 1999;9 (suppl):21-24.
53. MB. Effect of growth hormone on carbohydrate and lipid metabolism. Endocr Rev 1987;8:115-31.
54. FL, O’Neal D, Kamarudin N, Alford FP, Best JD. Growth hormone deficiency and cardiovascular risk. Baillieres Clin Endocrinol Metab 1998;12:199-216.
55. SA, Henderson A, Niththyananthan R, et al. The effects of short and long-term growth hormone replacement therapy in hypopituitary adults on lipid metabolism and carbohydrate tolerance. J Clin Endocrinol Metab 1995;80:356-63.
56. KA, Gray R, Anyaoku V, et al. Effects of four years’ treatment with biosynthetic human growth hormone (GH) on glucose homeostasis, insulin secretion and lipid metabolism in GH-deficient adults. Clin Endocrinol 1998;48:795-802.
57. D, Hew FL, Sikaris K, Ward G, Alford F, Best JD. Low density lipoprotein particle size in hypopituitary adults receiving conventional hormone replacement therapy. J Clin Endocrinol Metab 1996;81:2448-54.
58. Preventive Services Task Force. Guide to clinical preventive services. 2nd ed. Washington, DC: US Department of Health and Human Services; 1996.
59. PJ, Holen A, Glomstein A, et al. Long-term survival and quality of life in patients treated with a national ALL protocol 15-20 years earlier: IDM/HDM and late effects? Pediatr Hematol Oncol 1997;14:513-24.
60. AE. Posttraumatic distress in childhood cancer survivors and their parents. Med Pediatr Oncol 1998;1 (suppl):60-68.
61. LK, Chen E, Weiss R, et al. Comparison of psychologic outcome in adult survivors of childhood acute lymphoblastic leukemia versus sibling controls: a cooperative Children’s Cancer Group and National Institutes of Health study. J Clin Oncol 1997;15:547-56.
62. ML, Guo MD, Weiss R, et al. Smoking in adult survivors of childhood acute lymphoblastic leukemia. J Natl Cancer Inst 1998;90:219-25.
63. PB, Hough SF, Nel ED, van Riet FA, Beneke T, Wessels G. Bone mineral density in long-term survivors of childhood cancer. Int J Cancer Suppl 1998;11:44-7.
64. J, Hsieh K, Kalaitzoglou G, et al. Bone mineral density in young adult survivors of childhood cancer. J Pediatr Hematol Oncol 1998;20:241-45.
65. R, Brosnan P, Delpassand A, Zietz H, Klein MJ, Jaffe N. Osteopenia in young adult survivors of childhood cancer. Med Pediatr Oncol 1999;32:272-78.
66. V, Carlson ME, Roe TF, Ortega JA. Osteoporosis after cranial irradiation for acute lymphoblastic leukemia. J Pediatr 1990;117:238-44.
67. P, Komulainen J, Voutilainen R, et al. Reduced bone mineral density in long-term survivors of childhood acute lymphoblastic leukemia. J Pediatr Hematol Oncol 1998;20:234-40.
68. JT, Evans WD, Webb DK, Bell W, Gregory JW. Relative osteopenia after treatment for acute lymphoblastic leukemia. Pediatr Res 1999;45:544-51.
69. K, Holm K, Michaelsen KF, Hertz H, Muller J, Molgaard C. Bone mass after treatment for acute lymphoblastic leukemia in childhood. J Clin Oncol 1998;16:3752-60.
70. JJ, Kardos G, Roos JC, et al. Bone mineral density and markers of bone turnover in young adult survivors of childhood lymphoblastic leukaemia. Clin Endocrinol 1999;50:237-44.
71. BM, Rahim A, Mackie EM, Eden OB, Shalet SM. Clin Endocrinol 1998;48:777-783.
72. SA, Halton JM, Bradley C, Wu B, Barr RD. Bone and mineral abnormalities in childhood acute lymphoblastic leukemia: influence of disease, drugs and nutrition. Int J Cancer Suppl 1998;11:35-39.
73. B, Owens S, Okuyama T, Riggs S, Ferguson M, Litaker M. Effect of physical training and its cessation on percent fat and bone density of children with obesity. Obes Res 1999;7:208-14.
74. O, Kristinsson JO, Stefansson SO, Valdimarsson S, Sigurdsson G. Lean mass and physical activity as predictors of bone mineral density in 16-20-year old women. J Intern Med 1999;245:489-96.
75. I, van Croonenborg JJ, Kemper HC, Kostense PJ, Twisk JW. The effect of exercise training programs on bone mass: a meta-analysis of published controlled trials in pre- and postmenopausal women. Osteoporos Int 1999;9:1-12.
76. D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 1996;312:1254-59.
77. D, Sampietro-Colom L, Marshall D, Rico R, Granados A, Asua J. The effectiveness of bone density measurement and associated treatments for prevention of fractures: an international collaborative review. Int J Technol Assess Health Care 1998;14:237-54.
78. LL, Nesbit ME, Jr, Sather HN, Meadows AT, Ortega JA, Hammond GD. Height of children successfully treated for acute lymphoblastic leukemia: a report from the Late Effects Study Committee of Children’s Cancer Study Group. Med Pediatr Oncol 1985;13:14-21.
79. EA, Schell MJ, Carter M, Hustu O, Ochs JJ. Abnormal growth patterns and adult short stature in 115 long-term survivors of childhood leukemia. J Clin Oncol 1991;9:400-05.
80. M, Stanhope R, Chessells JM, Leiper AD. Impaired pubertal growth in acute lymphoblastic leukaemia. Arch Dis Child 1991;66:1403-07.
81. K, Dorffel W, Timme J, et al. Final height and puberty in 40 patients after antileukaemic treatment during childhood. Eur J Pediatr 1997;156:272-76.
82. P, Moell C, Cornu G, Malvaux P, Maes M. Subnormal growth during puberty in children treated for acute lymphoblastic leukemia. Pediatr Hematol Oncol 1992;9:217-22.
83. AC, van Doorn JW, Hahlen K, Stijnen T, de Muinck Keizer-Schrama SM, Drop SL. Long-term effects of treatment for acute lymphoblastic leukemia with and without cranial irradiation on growth and puberty: a comparative study. Pediatr Res 1993;33:577-82.
84. JA, Pollock BH, Jacaruso D, Morad A. Final attained height in patients successfully treated for childhood acute lymphoblastic leukemia. J Pediatr 1993;123:546-52.
85. AE, Adan L, Leverger G, Souberbielle JC, Schaison G, Brauner R. Growth hormone secretion, puberty and adult height after cranial irradiation with 18 Gy for leukaemia. Eur J Pediatr 1998;157:703-07.
86. J, Villaizan CJ, Garcia-Foncillas J, Azcona C, Salvador J, Sierrasesumaga L. Chemotherapy-induced growth hormone deficiency in children with cancer. Med Pediatr Oncol 1995;25:90-5.
87. J, Villaizan CJ, Garcia-Foncillas J, Salvador J, Sierrasesumaga L. Growth and growth hormone secretion in children with cancer treated with chemotherapy. J Pediatr 1997;131:105-12.
88. C, Mertens A, Walter A, et al. Final height after treatment for childhood acute lymphoblastic leukemia: comparison of no cranial irradiation with 1800 and 2400 centigrays of cranial irradiation. J Pediatr 1993;123:59-64.
89. A, Cacciari E, Rosito P, et al. Longitudinal growth and final height in long-term survivors of childhood leukaemia. Eur J Pediatr 1994;153:726-30.
90. TG, Byrne GC, Jones TW. Growth and growth hormone secretion after treatment for acute lymphoblastic leukemia in childhood 18-Gy versus 24-Gy cranial irradiation. J Pediatr Hematol Oncol 1995;17:167-71.
91. NH, Fisker S, Clausen N, Tuovinen V, Sindet-Pedersen S, Christiansen JS. Growth and endocrinological disorders up to 21 years after treatment for acute lymphoblastic leukemia in childhood. Med Pediatr Oncol 1998;30:351-56.
92. O’Halloran DJ, Tsatsoulis A, Whitehouse RW, Holmes SJ, Adams JE, Shalet SM. Increased bone density after recombinant human growth hormone (GH) therapy in adults with isolated GH deficiency. J Clin Endocrinol Metab 1993;76:1344-48.
93. F, Cuneo RC, Hesp R, Sonksen PH. The effects of treatment with recombinant human growth hormone on body composition and metabolism in adults with growth hormone deficiency. N Engl J Med 1989;321:1797-803.
94. P, Broman JE, Hetta J, et al. Quality of life in adults with growth hormone (GH) deficiency: response to treatment with recombinant human GH in a placebo-controlled 21-month trial. J Clin Endocrinol Metab 1995;80:3585-90.
95. SE, Colan SD, Gelber RD, Perez-Atayde AR, Sallan SE, Sanders SP. Late cardiac effects of doxorubicin therapy for acute lymphoblastic leukemia in childhood. N Engl J Med 1991;324:843-45.
96. SE, Lipsitz SR, Mone SM, et al. Female sex and drug dose as risk factors for late cardiotoxic effects of doxorubicin therapy for childhood cancer. N Engl J Med 1995;332:1738-43.
97. MA, Lipshultz SE. Epidemiology of anthracycline cardiotoxicity in children and adults. Semin Oncol 1998;25(suppl):72-85.
98. K, Levitt G, Bull C, Chessells J, Sullivan I. Anthracycline dose in childhood acute lymphoblastic leukemia: issues of early survival versus late cardiotoxicity. J Clin Oncol 1997;15:61-68.
99. K, Holm K, Lipsitz SR, et al. Relationship between cumulative anthracycline dose and late cardiotoxicity in childhood acute lymphoblastic leukemia. J Clin Oncol 1998;16:545-50.
100. LH. Ameliorating anthracycline cardiotoxicity in children with cancer: clinical trials with dexrazoxane. Semin Oncol 1998;25:86-92.
101. LJ, Graham T, Hurwitz R, et al. Guidelines for cardiac monitoring of children during and after anthracycline therapy: report of the Cardiology Committee of the Childrens Cancer Study Group. Pediatrics 1992;89:942-49.
102. for Disease Control and Prevention. Recommendations for prevention and control of hepatitis C virus (HCV) infection and HCV-related chronic disease. MMWR 1998;47:1-39.
103. M, Maggiore G, Silini E, Bono F, Vigano C. Hepatitis C virus infection in children treated for acute lymphoblastic leukemia. Blood 1994;84:2919-22.
104. SP, Ragusa R, Sciacca A, et al. Incidence and morbidity of infection by hepatitis C virus in children with acute lymphoblastic leukaemia. Eur J Pediatr 1994;153:271-75.
105. A, Testa M, Pontisso P, et al. Prevalence and natural history of hepatitis C infection in patients cured of childhood leukemia. Blood 1997;90:4628-33.
106. A, Alberti A. Hepatitis C virus serum markers and liver disease in children with leukemia. Leuk Lymphoma 1995;17:245-49.
107. S, Petris MG, Rossetti F, et al. Chronic hepatitis C virus infection after treatment for pediatric malignancy. Blood 1997;90:1315-20.
108. IM, Sanders J, Ruggiero F, Andrews T, Ungar D, Eyster ME. Chronic hepatitis C virus infections in leukemia survivors: prevalence, viral load, and severity of liver disease. Blood 1999;93:3672-77.
109. Dalton VM, Gelber RD, Li F, Donnelly MJ, Tarbell NJ, Sallan SE. Second malignancies in patients treated for childhood acute lymphoblastic leukemia. J Clin Oncol 1998;16:2848-53.
110. AW, Hancock ML, Pui CH, et al. Secondary brain tumors in children treated for acute lymphoblastic leukemia at St Jude Children’s Research Hospital. J Clin Oncol 1998;16:3761-67.
111. P, Straaten A, Gutjahr P. Secondary thyroid carcinoma after treatment for childhood cancer. Med Pediatr Oncol 1998;31:91-95.
112. Y, Leverger G, Carrere A, et al. Second thyroid neoplasms after prophylactic cranial irradiation for acute lymphoblastic leukemia. Am J Hematol 1998;59:91-94.
113. T, Ikuta H, Hibi S, Todo S. Second cutaneous neoplasms after acute lymphoblastic leukemia in childhood. Int J Hematol 1993;59:67-71.
114. J, Velasco-Benito JA, Pena-Penabad C, Armijo M. Basal cell carcioma in a girl after cobalt irradiation to the cranium for acute lymphoblastic leukemia: case report and literature review. Pediatr Dermatol 1996;13:54-57.
115. J, Philip P, Larsen SO, et al. Therapy-related myelodysplasia and acute myeloid leukemia: cytogenetic characteristics of 115 consecutive cases and risk in seven cohorts of patients treated intensively for malignant diseases in the Copenhagen series. Leukemia 1993;7:1975-86.
116. N, Shuster JJ, Bowman WP, et al. Intensive oral methotrexate protects against lymphoid marrow relapse in childhood B-precursor acute lymphoblastic leukemia. J Clin Oncol 1996;14:2803-11.
117. C, Hartmann JT, Kanz L, Bokemeyer C. Risk of secondary myeloid leukemia and myelodysplastic syndrome following standard-dose chemotherapy or high-dose chemotherapy with stem cell support in patients with potentially curable malignancies. J Cancer Res Clin Oncol 1998;124:207-14.
118. HM, Keating MJ. Therapy-related leukemia and myelodysplastic syndrome. Semin Oncol 1987;14:435-43.
119. MA, Rubinstein L, Anderson JR, et al. Secondary leukemia or myelodysplastic syndrome after treatment with epipodophyllotoxins. J Clin Oncol 1999;17:569-77.
120. MA, Rubinstein L, Cazenave L, et al. Report of the Cancer Therapy Evaluation Program monitoring plan for secondary acute myeloid leukemia following treatment with epipodophyllotoxins. J Natl Cancer Inst 1993;85:554-58.
121. CH, Relling MV, Rivera GK, et al. Epipodophyllotoxin-related acute myeloid leukemia: a study of 35 cases. Leukemia 1995;9:1990-96.
122. M, Akiyama Y, Koishi S, et al. Second malignancy following treatment of acute lymphoblastic leukemia in children. Int J Hematol 1998;67:397-401.
123. R, Clausen N, Siimes MA, et al. Reproduction following treatment for childhood leukemia: a population-based prospective cohort study of fertility and offspring. Med Pediatr Oncol 1991;19:459-66.
124. GA, Jenney ME. The reproductive system after childhood cancer. Br J Obstet Gynaecol 1998;105:946-53.
125. Wallace WH, Shalet SM, Tetlow LJ, Morris-Jones PH. Ovarian function following the treatment of childhood acute lymphoblastic leukaemia. Med Pediatr Oncol 1993;21:333-39.
126. MR, Robison LL, Nesbit ME, et al. Effects of radiation on ovarian function in long-term survivors of childhood acute lymphoblastic leukemia: a report from the Children’s Cancer Study Group. J Clin Oncol 1987;5:1759-65.
127. CA, Robison LL, Nesbit ME, et al. Effects of radiation on testicular function in long-term survivors of childhood acute lymphoblastic leukemia: a report from the Children’s Cancer Group. J Clin Oncol 1990;8:1981-87.
128. T, Kishi K, Imashuku S, et al. Testicular histology and function following long-term chemotherapy of acute leukemia in children and outcome of the patients who received testicular biopsy. Am J Pediatr Hematol Oncol 1986;8:288-93.
129. WH, Shalet SM, Lendon M, Morris-Jones PH. Male fertility in long-term survivors of childhood acute lymphoblastic leukaemia. Int J Androl 1991;14:312-19.
130. LB, Nicholson HS, Brasseux C, et al. Birth defects in offspring of adult survivors of childhood acute lymphoblastic leukemia: a Children’s Cancer Group/National Institutes of Health Report. Cancer 1996;78:169-76.
131. DL, Smith LE, Turner SJ, Gelber RD, Sallan SE. Ophthalmic evaluation of survivors of acute lymphoblastic leukemia. Ophthalmology 1988;95:151-55.
132. RG, Jr, Chauvenet AR, Smith TJ, Schwartz AC. Ophthalmic evaluation of long-term survivors of childhood acute lymphoblastic leukemia. Cancer 1986;58:963-68.
133. SC, Hopkins KP, Jones D, Crom D, Greenwald CA, Santana VM. Dental abnormalities in children treated for acute lymphoblastic leukemia. Leukemia 1997;11:792-96.
134. AL, Tarbell N, Valachovic RW, Gelber R, Schwenn M, Sallan S. Dentofacial development in long-term survivors of acute lymphoblastic leukemia: a comparison of three treatment modalities. Cancer 1990;66:2645-52.
135. AL, Waber DP, Sallan S, Tarbell NJ. The oral health of long-term survivors of acute lymphoblastic leukaemia: a comparison of three treatment modalities. Eur J Cancer B Oral Oncol 1995;31:250-52.
136. A, Chiarelli F, Di Marzio A, Impicciatore P, Marsico S, Angrilli F. Thyroid function in children treated for acute lymphoblastic leukemia. J Endocrinol Invest 1997;20:215-19.
137. LL, Nesbit ME, Sather HN, Meadows AT, Ortega JA, Hammond GD. Thyroid abnormalities in long-term survivors of childhood acute lymphoblastic leukemia. Pediatr Res 1985;19:266A.-
138. T, McCalla J, Berg S, et al. Subtle primary hypothyroidism in patients treated for acute lymphoblastic leukemia. Acta Endocrinol 1991;124:375-80.
139. CR, Miller JD, Guyda HJ, Esseltine DW, Chevalier LM, Freeman CR. Growth and development of long-term survivors of childhood acute lymphoblastic leukemia treated with and without prophylactic radiation of the central nervous system. Clin Invest Med 1985;8:307-14.
140. ML, Brecher ML, Glicksman AS, et al. Hypothalamic-pituitary function of children with acute lymphocytic leukemia after three forms of central nervous system prophylaxis: a retrospective study. Cancer 1986;57:1287-91.
141. EP, Leiper AD, Chessells JM. Thyroid function in children after treatment for acute lymphoblastic leukemia. Arch Dis Child 1988;64:631.-
142. MD, Shalet SM, Beardwell CG. Radiation and hypothalamic-pituitary function. Baillieres Clin Endocrinol Metab 1990;4:147-75.
143. F, Ohta K, Akanuma A, Sakata K. Dosimetry of radiation scattered to thyroid gland from prophylactic cranial irradiation for childhood leukemia. Pediatr Hematol Oncol 1994;11:47-53.
144. NJ, Tweeddale PM, Eden OB. Pulmonary function in childhood leukaemia survivors. Med Pediatr Oncol 1989;17:149-54.
145. K, Holm K, Olsen JH, Hertz H, Hesse B. Pulmonary function after treatment for acute lymphoblastic leukaemia in childhood. Br J Cancer 1998;78:21-27.
146. BL, Tanyer G, Poplack DG, et al. Transient acute hepatotoxicity of high-dose methotrexate therapy during childhood. NCI Monogr 1987;5:207-12.
147. F, Kinumaki H, Yokota S, Hayashi Y, Kobayashi M, Kamoshita S. Liver function studies in children with acute lymphocytic leukemia after cessation of therapy. Med Pediatr Oncol 1994;23:111-15.
148. AC, Buchanan GR, Zweiner RJ, Bowman WP, Winick NJ. Serum aminotransferase elevation during and following treatment of childhood acute lymphoblastic leukemia. J Clin Oncol 1997;15:1560-66.
149. PJ, Balistreri WF, Bove KE, Ballard ET, Passo MH. The relationship of hepatotoxic risk factors and liver histology in methotrexate therapy for juvenile rheumatoid arthritis. J Pediatr 1999;134:47-52.
150. HJ, Simone J, Aur RJA. Cyclophosphamide-induced hemorrhagic cystitis in children with leukemia. Cancer 1975;36:1572-76.
151. TJ, Benson RC. Cyclophosphamide-induced hemorrhagic cystitis: a review of 100 patients. Cancer 1988;61:451-57.
152. JM, Reed EC, Pippert GC, et al. Mesna compared with continuous bladder irrigation as uroprotection during high-dose chemotherapy and transplantation: a randomized trial. J Clin Oncol 1993;11:1306-10.
153. LB, Curtis RE, Glimelius B, et al. Bladder and kidney cancer following cyclophosphamide therapy for non-Hodgkin’s lymphoma. J Natl Cancer Inst 1995;87:524-30.
154. der Does-van den Berg A, de Vaan GAM, van Weerden JF, Hahlen K, van Weel-Sipman M, Veerman AJP. Late effects among long-term survivors of childhood acute leukemia in the Netherlands: a Dutch Childhood Leukemia Study Group report. Pediatr Res 1995;38:802-07.
Acute lymphoblastic leukemia (ALL), the most common childhood malignancy, accounts for almost one fourth of childhood cancers.1 The incidence of ALL has shown a moderate increase in the past 20 years. It is generally considered a cancer of younger children, with a peak incidence between the ages of 2 and 5 years. It is approximately 30% more common in boys than girls and approximately twice as common in white children as in black children. Improvements in ALL treatment during the past 20 years have increased the overall survival rate to approximately 80%. Thus, success in “curing” this childhood disease has resulted in a growing population of long-term survivors.
Since it is anticipated that the majority of long-term survivors of childhood ALL will seek health care from primary care physicians, it is important to understand the potential health problems that these patients may experience secondary to their cancer treatment.2-4 However, there are no articles in peer-reviewed family practice journals concerning the long-term follow-up of survivors of childhood ALL. Our clinical review briefly describes the evolution of the treatment for ALL, potential late effects of treatment, and recommendations for screening asymptomatic long-term survivors. Because this field of investigation is rapidly advancing and much of the available information is from cross-sectional and small cohort studies, these recommendations should not be viewed as a set of guidelines. Instead, our review is intended to contribute a foundation for primary care physicians providing longitudinal health care for ALL survivors while highlighting the areas needing further investigation. Also, because of the evolving changes in treatment protocols—and thus in potential late effects—it is essential to frequently communicate with our colleagues who specialize in the treatment of children with cancer.
Evolution of treatment for childhood all
During the 1940s childhood leukemias had a uniformly rapid fatal course over a short period of time, thus the designation of the term “acute.”5 In the late 1940s, Farber and colleagues6 found that aminopterin (a folic acid antagonist) could induce temporary remissions in leukemia. This discovery opened the era of clinical investigation into the uses of combined chemotherapy in treating childhood ALL Figure 1. The use of antimetabolite therapy for prolonged periods started in the late 1950s and early 1960s and suggested that it was possible for children to have an extended period of remission and possibly be cured. The addition of anthracyclines such as daunorubicin in the 1970s and the discovery that the enzyme L-asparaginase was useful in ALL therapy for depleting cells of the essential amino acid L-asparagine further boosted the ability to induce and sustain remission.7
A significant factor in morbidity and mortality from childhood ALL was the development of leukemia within the central nervous system (CNS). Left untreated, more than half the children with ALL developed leukemia in the CNS, even when bone marrow remission was sustained. In most patients, CNS relapse was followed by bone marrow relapse. Prophylactic radiation to the head and spine, introduced in the early 1970s, significantly decreased the incidence of CNS leukemia and resulted in significant advancement in long-term survival. However, in the early 1980s—as a consequence of the appreciation of neurodevelopmental delays and cognitive dysfunction secondary to relatively higher-dose (24 Gy) cranial irradiation (CRT), different methods of CNS treatment and prophylaxis evolved, either using lower-dose CRT (18 Gy), intensification of systemic methotrexate (MTX) dosaging, or intrathecal medications.8-11
Current treatment regimens divide therapy into remission induction, consolidation and CNS prophylaxis, and maintenance or continuous treatment. Induction chemotherapy (aimed at an initial reduction in blast cell percentage in the bone marrow to 5% or lower) consists of a 1-month schedule of vincristine, prednisone, and L-asparaginase alone or with other agents. Following induction, a consolidation phase consisting of an intensified period of treatment combines the use of antimetabolites and other agents with intrathecal chemotherapy for CNS prophylaxis. Maintenance therapy continues for a period of approximately 2 years and relies heavily on the use of methotrexate and 6-mercaptopurine. During the past 2 decades, recognized differences in the phenotype of the leukemic cells have resulted in protocol modifications to improve outcome and reduce toxicity. Increasingly, the T-cell phenotype of childhood ALL has been treated more effectively with intensified regimens that include cyclophosphamide, cytarabine, and anthracylines.12,13
Late effects of treatment for childhood all
A late effect is defined as any chronic or late occurring physical or psychosocial outcome persisting or developing more than 5 years after diagnosis of the cancer. In this section we describe potential late effects in order from more common or serious health problems to less common or serious ones Table 1. Many of these late effects may have long asymptomatic intervals before end-stage disease or serious health outcomes, such as survivors with hepatitis C who develop cirrhosis or those with a late-onset cardiomyopathy who present in congestive heart failure. Included in each section is a discussion about the screening tests commonly used in long-term follow-up programs that include asymptomatic survivors4Table 2. It should be stressed that the value of most of these tests has not been studied in this population in a prospective or a well-designed retrospective manner with adequate sample sizes, which limits the strength of the recommendations. Clinicians should be selective in ordering tests and providing preventive services and should actively incorporate the patient’s concerns and fears when arriving at an individualized decision on whether to perform a test. Figure 2 is a compilation of information pertinent to the follow-up of a survivor of childhood ALL, provided as a single-page template for clinical use.
Because bone marrow transplantation (BMT) is a relatively new therapy affecting a much smaller number of ALL survivors, our review does not include the late effects related to total body irradiation and BMT.
Cognitive dysfunction and performance at school and work
As described in the section on the evolution of treatment, 24 Gy CRT is associated with cognitive dysfunction. A meta-analysis of more than 30 retrospective and prospective studies established that 24 Gy CRT in combination with MTX resulted in a mean decrease of 10 points in full-scale intelligence quotient (IQ).9 Verbal scores were affected more than performance IQ, and changes were noted to be progressive. Although more than half the patients had mild to moderate learning problems, the outcomes were highly variable, and some patients experienced 20- to 30-point losses, while others had no discernable changes.9,14 Deficits have been noted in measures of visual-spatial abilities, attention-concentration, nonverbal memory, and somatosensory functioning.8-10,15-20 Studies have also shown that girls and patients treated with CRT before the age of 4 years are at significantly higher risk. Neuropathologic changes resulting from 24 Gy CRT include leukoencephalopathy, mineralizing microangiopathy, subacute necrotizing leukomyelopathy, and intracerebral calcifications, commonly with subsequent cerebral atrophy and microcephally.21,22
Treatment with 18 Gy CRT in combination with chemotherapy also affects cognition, though not as profoundly as with 24 Gy CRT. In a retrospective study of children with ALL, randomized by risk group to receive either 18 Gy CRT with chemotherapy or chemotherapy alone, 66 survivors were subsequently tested using several cognitive measures.23 Girls who were treated with CRT/chemotherapy had a mean IQ 9 points lower than those treated with chemotherapy alone. All patients had impairments in verbal coding and short-term memory regardless of CRT use or MTX dose, suggesting that another agent such as glucocorticoids may be responsible. Other small prospective and retrospective studies have found a mild decrease in full-scale IQ in patients treated with 18 Gy CRT/chemotherapy, although subanalysis generally showed that changes were only significant for girls and patients treated at a younger age.24-27
Recent studies suggest that neurodevelopmental outcomes for survivors treated with chemotherapy alone are generally positive.28 An analysis of 30 survivors whose condition was diagnosed before the age of 12 months showed no decrease in 6 cognitive and motor indices and no sex differences.29 Though full-scale IQ was normal, Brown and colleagues30 reported that girls had significantly decreased nonverbal scores in a study of 47 ALL survivors. Fine motor disturbances and manual dexterity difficulties, which may compound learning difficulties, have been seen in 25% to 33% of ALL survivors evaluated in 2 small cross-sectional studies.31,32 Changes in cerebellar-frontal subsystems that correlate with neuropsychological deficits have also been seen in ALL patients treated with chemotherapy alone.33
The Children’s Cancer Group investigated the impact of treatment on scholastic performance of 593 adult survivors, compared with 409 sibling controls.34 Patients treated with 24 Gy CRT were more likely to enter special education or learning-disabled programs, with relative risks of 4.1 and 5.3, respectively. Previous treatment with 18 Gy CRT had less impact, with a relative risk of 4.0 to enter a special education program but no increased risk of entering a learning-disabled program. Patients treated with CRT (18 or 24 Gy) were just as likely to enter gifted and talented programs as their sibling controls. In general, survivors were as likely to finish high school and enter college as controls, but those treated with 24 Gy or treated before the age of 6 years were less likely to enter college. There were no sex differences in educational achievements.
There are no studies that explore problems in job acquisition, promotion, and retention for ALL survivors with evidence of cognitive dysfunction. Abstract thinking abilities in higher-level decision making may be problematic for some ALL survivors, particularly those treated with 24 Gy CRT. Further study is warranted, particularly in evaluating methods to assist at-risk survivors in developing job skills and applying for a job.
Obesity, physical inactivity, and risk of premature cardiovascular disease
Several retrospective cohort and cross-sectional studies have shown an increased incidence and prevalence of obesity in ALL survivors. Early studies suggested that the resulting obesity was secondary to CRT, with 38% to 57% of the survivors having a body mass index (BMI) >2 standard deviations (SDs) above the norm at the time of attainment of final height.35-38 Two recent cross-sectional studies suggest that the increased prevalence of obesity may be due to other factors. Van Dongen-Melman and coworkers39 compared the weight gain and BMI of 113 ALL survivors who had received CRT/chemotherapy or chemotherapy alone and found that children treated with a combination of prednisone and dexamethasone had the highest prevalence of obesity (44%).39 Talvensaari and colleagues40 evaluated 50 childhood cancer survivors with a median age of 18 years (including 28 ALL patients) and found an increased prevalence of obesity in survivors that was not associated with CRT.
Obesity in ALL survivors may be due in part to reduced physical activity. In a small cross-sectional study with sibling controls, ALL survivors had decreased activity levels and total daily energy expenditures that correlated with their percentage of body fat.41 Maximal and submaximal exercise capacity were reduced in another cross-sectional study.42 Similarly, in a study of 53 ALL survivors with a longer interval from ALL diagnosis (mean=10.5 years), 25% and 31%, respectively, were unable to reach normal maximal oxygen uptake and normal oxygen uptake at the anaerobic threshold.43
Changes in gross motor skills may also affect the physical activity level of ALL survivors. Balance, strength, running speed and agility, and hand grip strength were decreased in a cohort of 36 ALL survivors with a median age of 9.3 years.44 In a follow-up of this cohort, Wright and coworkers45 reported that the ALL survivors had significantly less active and passive dorsiflexion range of motion of the ankle than did controls. Younger age at diagnosis and female sex were significant predictors, while treatment with CRT did not increase risk. These studies suggest that ALL survivors should be assessed for gross motor deficits that might alter exercise choices.
In the general population, obesity and physical inactivity are risk factors for cardiovascular disease. Obesity (an especially important risk factor during young adulthood) enhances the development of hypertension, dyslipidemia, and insulin resistance.46-48 Because the median age of ALL survivors is still relatively young, there are no cohort or case-control studies evaluating the treatment-related risk of premature onset of coronary artery disease. Talvensaari and coworkers40 reported that 50 childhood cancer survivors (including 28 ALL survivors) had an increased risk of fasting hyperinsulinemia and reduced high-density lipoprotein (HDL) cholesterol compared with 50 age- and sex-matched controls. Eight of the cancer survivors with reduced spontaneous growth hormone (GH) secretion (4/8 had received CRT) had obesity, hyperinsulinemia, and reduced HDL cholesterol, fitting the criteria for cardiac dysmetabolic syndrome, a clustering of metabolic problems associated with a markedly increased risk of cardiovascular disease.49
Studies of noncancer populations may shed light on the cardiovascular risk of ALL survivors with GH deficiency. Hypopituitarism with GH deficiency in adults is associated with increased vascular mortality.50-52 Adults with GH deficiency also have an increased prevalence of dyslipidemia53,54 and insulin resistance,55 that may improve with GH therapy.56,57
Counseling on the benefits of proper diet and exercise is an important component of long-term care for ALL survivors. Periodic analysis of lipoproteins has not been prospectively studied in ALL survivors, but the US Preventive Services Task Force states that adolescents and young adults who have major risk factors for cardiovascular disease should be screened.58
Psychosocial well-being of all survivors
The long-term psychosocial welfare of ALL survivors is complex. A population-based sibling-matched control study of 93 ALL survivors who were at least 15 years postdiagnosis showed no difference in quality of life or mental health.59 Similarly, no differences were found in symptoms of anxiety and posttraumatic stress in 130 leukemia survivors and 155 controls.60 In contrast, a large cooperative study of the Children’s Cancer Group and the National Institutes of Health evaluated 580 adult survivors and 396 sibling controls and reported that survivors had greater negative mood and reported more tension, depression, anger, and confusion.61 Female, minority, and unemployed survivors reported the highest total mood disturbance. Issues related to late effects, especially cognitive dysfunction, obesity, and physical inactivity, may have an impact on the mental health of survivors.
Few data are available on the risk behavior of ALL survivors. In a cohort study of 592 young adult ALL survivors and 409 sibling controls, Tao and colleagues62 reported that ALL survivors were less likely to start smoking, but once they started they were no more likely to quit than their siblings. Fourteen percent of the ALL survivors were smokers. Although no prospective studies have evaluated the effect of smoking on the incidence and severity of late effects of ALL treatment, it will have an impact on survivors with cardiovascular risk factors, restrictive pulmonary disease, and osteopenia. Counseling on smoking cessation is imperative in the long-term health care of ALL survivors.
Osteopenia and osteoporosis
Several well-designed small to medium-size cross-sectional studies of childhood cancer survivors63-65 and ALL survivors66-71 with median ages at evaluation ranging from 12 to 25 years consistently showed reduction in bone mineral density, bone mass content (BMC), and/or age-adjusted bone mass. Age at diagnosis, interval since treatment, sex, and cumulative dosages of MTX and corticosteroids have not been consistently associated with reduction in bone mass. In contrast, CRT has consistently been identified as a risk factor, although the 3 studies that evaluated GH status showed variation in the relationship of GH deficiency and reduced bone mass.69-71 Impairment of peak bone mass is likely multifactorial in etiology, with predisposing risk factors including altered bone metabolism at the time of onset of leukemia, interference in bone metabolism by corticosteroids and MTX, and impaired bone growth and skeletal maturation caused by pituitary dysfunction/GH deficiency. In an ongoing prospective cohort study, Atkinson and coworkers72 reported that by 6 months of therapy for ALL, 64% of the children had a reduction from baseline measures of BMC, and by the end of 2 years of therapy 83% were osteopenic. Hypomagnesemia due to renal wasting of magnesium after treatment with high-dose corticosteroids and/or aminoglycosides was associated with the progression in changes and may be a key factor in the alteration of bone metabolism.
Reduction in peak bone mass in young adults is a significant risk factor for developing osteoporosis and subsequent fracture, and measures to prevent or reverse bone loss are important. Exercise increases bone density in obese children73 and young adults74 and has recently been shown by meta-analysis75 to prevent or reverse almost 1% of bone loss per year in pre- and postmenopausal women. With ALL survivors likely to be less physically active,41-43 it is essential to counsel them on the benefits of exercise in preventing cardiovascular disease and osteoporosis and help them develop an exercise plan. Additionally, counseling on calcium intake and avoidance of smoking is important. Though bone densitometry has not been an effective screening test for the general population, it has value in high-risk groups.76,77 Prospective randomized trials are needed to evaluate the usefulness and frequency of screening.
GH deficiency
Cross-sectional and longitudinal studies have consistently shown that patients treated with 24 Gy CRT have a decrease in median height of approximately 1 to 1.5 SD score, or 5 to 10 cm.37,78-84 Treatment with 18 Gy CRT85 or chemotherapy alone86,87 affect the final height to a lesser degree. Sklar and coworkers88 reported a change in final height SD score of -0.65 for patients treated with 18 Gy CRT and -0.49 for those treated with chemotherapy alone. Girls and patients treated at a younger age (<5 years) have the greatest growth reduction.37,78,88,89 These changes are thought to be secondary to GH deficiency, resulting in a blunted pubertal growth spurt. The greater the deficiency, the more profound the impairment of growth.90 Brennan and colleagues71 reported a median decrement in final height of 2.1 SD in patients with severe GH deficiency. Treatment with GH in these patients usually results in near normalization of final height.
Though GH therapy is generally stopped when children reach their final height or by the age of 18 years, deficiency persists. In a small cross-sectional study of 30 ALL survivors, 9 of 15 patients who received 24 Gy CRT (median age=21.4 years) were GH deficient.91 In another cross-sectional analysis of the GH status of 32 ALL survivors (median age=23 years), 21 of 32 were GH deficient, including 9 who were severely deficient.71 The consequences of GH deficiency in adulthood are not well understood. Small studies suggest that GH replacement may improve bone mineral density,92 body composition,93 and quality of life.94
Late onset anthracycline-induced cardiomyopathy
Anthracyclines (notably daunorubicin and doxorubicin) are often used during the induction phase of treatment, with some protocols using moderate to high dosages (Ž350 mg/m2) for high-risk patients. In the past 10 years it has become apparent that childhood cancer patients treated with an anthracycline are at increased risk for developing late-onset cardiomyopathy.95-97 Classically, anthracycline-induced cardiomyopathy is characterized by elevated afterload followed by the development of a dilated thin-walled left ventricle. Over time this can lead to a stiff and poorly compliant left ventricle. Most patients are asymptomatic, but longitudinal studies suggest that a significant proportion will experience progressive changes and may develop congestive heart failure.96,97
Lipshultz and coworkers95 assessed the cardiac status of 115 ALL survivors treated with doxorubicin and found that 65% of those treated with 228 mg/m2 or more had increased left ventricular afterload.95 In a follow-up study, Lipshultz and colleagues96 reported that female sex, younger age at treatment, higher rate of administration of doxorubicin, and cumulative dose of doxorubicin were independent risk factors for the development of altered left ventricular function. Two recent cross-sectional studies suggest that the risk of left ventricular dysfunction is uncommon in children who received cumulative doses less than 300 mg per m2.98,99 In patients treated with cumulative doses less than 270 mg per m2, Sorensen and coworkers98 did not find that female sex and younger age at treatment were risk factors. However, because late cardiac abnormalities were seen in survivors who received only 90 mg per m2, there might be no absolute level below which cardiotoxicity can be prevented.
Because of the concerns about cardiotoxicity, most recent protocols limit anthracycline doses to less than 300 mg per m2, and the use of cardioprotectants such as dexrazoxane in children is under investigation.100 Primary care physicians who provide follow-up care for adult survivors should communicate with oncologists at the treating institution, obtain information about the cumulative dosage of anthracyclines, and discuss long-term screening. Because patients with anthracycline-induced cardiomyopathies generally have a prolonged asymptomatic interval before becoming symptomatic, interval screening is recommended. Optimal timing and testing modality for screening have not been prospectively studied. It is currently recommended that patients who received 300 mg/m2 or more of an anthracycline have a screening echocardiogram every 2 to 3 years to evaluate left ventricular function and shortening fraction.101 It is also important to question patients regarding symptoms of congestive heart failure and to aggressively evaluate them if present.
Hepatitis C
Because most ALL patients receive blood products during therapy, those treated before adequate blood donor screening for hepatitis C was initiated in the early 1990s are at risk for chronic liver disease.102 The prevalence of circulating hepatitis C virus (HCV) ribonucleic acid (RNA) in ALL patients treated in Italy before 1990 ranges from 23% to 49%.103-105 The natural history of ALL survivors with hepatitis C is not well understood. In an Italian study, only 4% of the 56 HCV-RNA seropositive patients had persistently elevated alanine aminotransferase (ALT) over the course of follow-up (mean=17 years).106 For a median of 14 years, 81 survivors of various childhood cancers who were HCV-RNA seropositive were followed, and none showed progression to liver failure.107 In contrast, Paul and coworkers108 reported that 12% of 75 leukemia survivors were anti-HCV positive, 6 of 9 had liver biopsies that showed at least moderate portal inflammation, and half had bridging fibrosis. The Centers for Disease Control and Prevention102 recommend universal screening with anti-HCV for all patients who received blood products before July 1992.
Second malignant neoplasms
Second malignant neoplasms (SMN) are rare in ALL survivors. Thirteen SMNs were diagnosed a median of 6.7 years from ALL diagnosis in a cohort study of 1597 ALL survivors and were associated with the use of radiation (8/13, CNS or head and neck) or chemotherapy (3/13, hematopoietic).109 The cumulative incidence of brain tumors at 20 years in a cohort of 1612 patients was only 1.39%, and more than half of these tumors were either low-grade or benign.110 CNS tumors did not occur in patients treated with chemotherapy alone. Thyroid tumors (predominantly papillary carcinoma) can rarely occur after treatment with cranial or craniospinal irradiation.111,112 Cases of basal cell carcinoma along the spinal axis have also been reported in patients treated with craniospinal irradiation.113,114
Therapy-related acute myelogenous leukemia (t-AML) has been seen following treatment of several childhood cancers, such as ALL and Hodgkin’s and non– Hodgkin’s lymphoma. Cohort studies have shown that agents with leukemogenic potential include alklyating agents and epidophyllotoxin chemotherapy.115-121 Most t-AMLs occur within 8 years of treatment, although cases occurring up to 13 years have been reported.115 Myelodysplasia (especially pancytopenia) generally precedes t-AML. The risk of t-AML following treatment for ALL has been small in 2 cohort studies.109,122 However, because precancerous states (myelodysplastic changes or myelodysplastic syndrome) are usually antecedent to t-AML and early diagnosis may improve outcomes, most institutions recommend obtaining a complete blood count (CBC) with a platelet count and a white blood cell differential in the routine follow-up of ALL survivors who have been treated with an alkylating agent, such as cyclophosphamide, or an epidophyllotoxin, such as etoposide. How long and how frequently a CBC should be obtained in follow-up of an ALL survivor have not been established.
Fertility and reproduction
Most antimetabolite-based treatment protocols for ALL do not affect long-term fertility for men or women.123,124 Craniospinal and abdominal irradiation have been associated with infertility in both sexes but are no longer used for ALL.125-127 Cyclophosphamide (an alkylating agent commonly used in earlier protocols but currently limited to high-risk patients) is also associated with infertility in a dosedependent fashion in both sexes.124,128,129 Resolution of germ-cell dysfunction may occur in men over time, but fertility remains poor for some. Women survivors treated with craniospinal or abdominal irradiation or with cyclophosphamide are at risk for ovarian failure and premature menopause and thus may be at increased risk for osteoporosis. If ovarian failure is suspected, measurement of follicle-stimulating hormone, luteinizing hormone, and serum estradiol and an evaluation by an endocrinologist should be considered.
ALL survivors should know that preliminary studies suggest that treatment is not associated with an increase in congenital malformations of their offspring. In a population-based prospective cohort study an increased rate of congenital defects was not found among 299 adult survivors.130
Ocular abnormalities
Ocular abnormalities in patients treated with CRT are common but generally asymptomatic. Two studies have evaluated the effect of CRT and systemic corticosteroids on the eyes. In a study of 82 ALL survivors who were a mean of 32 months after completion of therapy, 52% of the patients had posterior subcapsular cataracts (PSC) that were generally not visually significant and were not related to age at treatment or gender.131 Eighty-three percent of the 18 patients who had received CRT and systemic corticosteroids were noted to have asymptomatic ocular abnormalities after a median surveillance of 4.1 years.132 Optical densities of the lens were seen in 13 of the 18 of the survivors. There have been no published studies evaluating long-term survivors who received systemic corticosteroids without CRT. Periodic vision and cataract screening is recommended for ALL survivors treated with CRT and should be considered for all survivors of ALL until the risk of prolonged corticosteroid use in childhood is better understood.
Dental and periodontal disease
ALL survivors, especially those treated with CRT, are more likely to have problems with tooth development and be at risk for periodontal disease. In a large retrospective evaluation of dental records, 39.5% of ALL survivors had a dental abnormality, including root stunting (24.4%), microdontia (18.9%), or hypodontia (8.5%).133 Patients who were treated at an age younger than 8 years or who received CRT had more dental abnormalities than the other groups. Similar findings were seen in 2 smaller cross-sectional studies. Abnormal dental development occurred in 95% of all patients and 100% of patients aged 5 years or younger at diagnosis.134 Abnormalities included tooth agenesis, arrested tooth development, microdontia, and enamel dysplasia. Patients who received CRT and those treated at an age younger than 5 years had higher severity scores. Survivors did not have increased caries.135 However, patients younger than 5 years who were treated with cranial irradiation were found to have higher plaque and gingivitis scores, suggesting an increased risk of periodontal disease. A periodic dental and periodontal evaluation is recommended for survivors treated with CRT or at a young age.
Thyroid-related disorders
Following treatment with CRT, hypothyroidism infrequently occurs in ALL survivors through damage to the hypothalamic-pituitary-thyroid axis and/or the direct effect of radiation of the gland. Mohn and colleagues136 reported that 8 of 24 childhood ALL survivors who had received CRT (either 18 or 24 Gy) had either a low basal thyroid-stimulating hormone (TSH) or low peak TSH after thyrotropin-releasing hormone stimulation. Robison and colleagues137 reported that 10% of 175 ALL survivors who had been treated with either 18 or 24 Gy CRT or craniospinal radiation (CS-RT) therapy had a thyroid abnormality, including 5 children with primary hypothyroidism. Pasqualini and colleagues138 reported that 6 of 10 ALL survivors who received either CRT or CS-RT had subtle evidence of primary hypothyroidism. In contrast, 3 cross-sectional studies did not find evidence of primary hypothyroidism in 13, 31, and 64 patients, respectively.1,139-141 Littley and coworkers142 suggest that hypopituitarism is commonly underdiagnosed secondary to the subtle manifestations and insidious progression of disease. Radioactive scatter to the thyroid occurs with CRT in a dose-dependent fashion,143 and ALL survivors treated with either 18 or 24 Gy CRT are at risk for secondary hypothyroidism, thyroid nodules, and thyroid carcinoma.111 Periodic screening with TSH and free T-4 are recommended in ALL survivors treated with CRT. Further screening of the asymptomatic survivor with thyrotropin-releasing hormone stimulation test or ultrasound of the thyroid gland are costly and have not been prospectively studied.
Pulmonary late effects
ALL survivors may have an increased prevalence of mild, generally subclinical, restrictive pulmonary disease. In a small cross-sectional study of ALL survivors, Shaw and coworkers144 reported mild restrictive changes, with patients treated at a younger age at higher risk. Similarly, an analysis of 70 leukemia survivors found mild but significant decreases in forced vital capacity (FVC), forced expiratory volume in 1 second (FEV-1), total lung capacity (TLC), and transfer for carbon monoxide (DLCO).42 Cyclophosphamide, craniospinal irradiation, and a history of chest infections during treatment were independent variables associated with reductions in FEV-1, FVC, and TLC, while anthracyclines and craniospinal irradiation were associated with reductions in DLCO. ALL survivors also had impaired submaximal and maximal exercise capacity. These findings were further supported by analysis of a recent cross-sectional study of 128 patients a median of 7.6 years from therapy completion that reported an increased prevalence of subclinical restrictive pulmonary disease in ALL survivors.145 The long-term consequences and the possible role of smoking or other inhalant exposures need to be studied.
Liver dysfunction (Non-Hepatitis C)
During treatment with methotrexate (especially high-dose ranges) elevations of transaminases are common and generally transient. Two small longitudinal studies following ALL survivors for up to 7 years after completion of therapy did not report any patients with persistent transaminasemia, although Bessho and colleagues noted that 6 of 13 of their ALL survivors had elevated 2-hour postprandial bile acid levels, a more sensitive predictor of liver cirrhosis than transaminase level.146,147 Farrow and coworkers148 found that of 114 survivors who had ALT elevations greater than 5 times the upper limit of normal during therapy, only 17 (14.9%) had elevations persistently. Eight of these patients had chronic HCV infections. Of the remaining 9 patients, only 1 had a persistently elevated transaminase of greater than 2 times normal.
Although there are currently no data evaluating ALL survivors for long-term liver-related complications secondary to methotrexate, studies in patients with juvenile rheumatoid arthritis show that septal and portal fibrosis can occur with weekly low-dose methotrexate treatment of durations as short as 17 months.149 Obesity may be an associated risk factor for the development of cirrhosis in juvenile rheumatoid arthritis patients treated with methotrexate. Because of these potential risks, periodic measurement of ALT is recommended in follow-up of ALL survivors.
Urologic late effects
Cyclophosphamide is a long-recognized cause of hemorrhagic cystitis and a well-established bladder carcinogen. In a retrospective review150 of 314 children with ALL who were treated with cyclophosphamide between 1963 and 1973, 8% developed hemorrhagic cystitis. The frequency of diagnosis was not related to age or sex, but African American children were at higher risk. Cyclophosphamide-induced hemorrhagic cystitis generally presents during therapy, with children complaining of gross hematuria or irritative voiding complaints.151 Concurrent treatment with oral sodium 2-mercapatoethanesulfonate appears to markedly decrease the incidence of cyclophosphamide-induced hemorrhagic cystitis.152 In a nested case-control study of survivors of non–Hodgkin’s lymphoma, Travis and colleagues153 reported that there was a 2.4-fold increased risk of bladder cancer in patients treated with cumulative dosages of cyclophosphamide lower than 20 g. Because of the risk of chronic hemorrhagic cystitis and bladder cancer, ALL survivors treated with cyclophosphamide should have periodic screening urinalysis, and their review of systems should include voiding problems.
Alopecia
Alopecia is a bothersome late effect secondary to treatment with 24 Gy CRT for which there are no available treatments. In a retrospective study of 273 ALL survivors treated with CRT, 10% had alopecia.154
Acknowledgement
Dr Oeffinger received partial support for this work through the American Academy of Family Physicians Foundation Advanced Research Training Grant and the Robert Wood Johnson Foundation Generalist Physician Faculty Scholars Program.
We would like to thank Drs George Buchanan, Melissa Hudson, and Neyssa Marina for their critical review of this manuscript and Ms Laura Snell and Dr James Tysinger for their editing assistance.
Acute lymphoblastic leukemia (ALL), the most common childhood malignancy, accounts for almost one fourth of childhood cancers.1 The incidence of ALL has shown a moderate increase in the past 20 years. It is generally considered a cancer of younger children, with a peak incidence between the ages of 2 and 5 years. It is approximately 30% more common in boys than girls and approximately twice as common in white children as in black children. Improvements in ALL treatment during the past 20 years have increased the overall survival rate to approximately 80%. Thus, success in “curing” this childhood disease has resulted in a growing population of long-term survivors.
Since it is anticipated that the majority of long-term survivors of childhood ALL will seek health care from primary care physicians, it is important to understand the potential health problems that these patients may experience secondary to their cancer treatment.2-4 However, there are no articles in peer-reviewed family practice journals concerning the long-term follow-up of survivors of childhood ALL. Our clinical review briefly describes the evolution of the treatment for ALL, potential late effects of treatment, and recommendations for screening asymptomatic long-term survivors. Because this field of investigation is rapidly advancing and much of the available information is from cross-sectional and small cohort studies, these recommendations should not be viewed as a set of guidelines. Instead, our review is intended to contribute a foundation for primary care physicians providing longitudinal health care for ALL survivors while highlighting the areas needing further investigation. Also, because of the evolving changes in treatment protocols—and thus in potential late effects—it is essential to frequently communicate with our colleagues who specialize in the treatment of children with cancer.
Evolution of treatment for childhood all
During the 1940s childhood leukemias had a uniformly rapid fatal course over a short period of time, thus the designation of the term “acute.”5 In the late 1940s, Farber and colleagues6 found that aminopterin (a folic acid antagonist) could induce temporary remissions in leukemia. This discovery opened the era of clinical investigation into the uses of combined chemotherapy in treating childhood ALL Figure 1. The use of antimetabolite therapy for prolonged periods started in the late 1950s and early 1960s and suggested that it was possible for children to have an extended period of remission and possibly be cured. The addition of anthracyclines such as daunorubicin in the 1970s and the discovery that the enzyme L-asparaginase was useful in ALL therapy for depleting cells of the essential amino acid L-asparagine further boosted the ability to induce and sustain remission.7
A significant factor in morbidity and mortality from childhood ALL was the development of leukemia within the central nervous system (CNS). Left untreated, more than half the children with ALL developed leukemia in the CNS, even when bone marrow remission was sustained. In most patients, CNS relapse was followed by bone marrow relapse. Prophylactic radiation to the head and spine, introduced in the early 1970s, significantly decreased the incidence of CNS leukemia and resulted in significant advancement in long-term survival. However, in the early 1980s—as a consequence of the appreciation of neurodevelopmental delays and cognitive dysfunction secondary to relatively higher-dose (24 Gy) cranial irradiation (CRT), different methods of CNS treatment and prophylaxis evolved, either using lower-dose CRT (18 Gy), intensification of systemic methotrexate (MTX) dosaging, or intrathecal medications.8-11
Current treatment regimens divide therapy into remission induction, consolidation and CNS prophylaxis, and maintenance or continuous treatment. Induction chemotherapy (aimed at an initial reduction in blast cell percentage in the bone marrow to 5% or lower) consists of a 1-month schedule of vincristine, prednisone, and L-asparaginase alone or with other agents. Following induction, a consolidation phase consisting of an intensified period of treatment combines the use of antimetabolites and other agents with intrathecal chemotherapy for CNS prophylaxis. Maintenance therapy continues for a period of approximately 2 years and relies heavily on the use of methotrexate and 6-mercaptopurine. During the past 2 decades, recognized differences in the phenotype of the leukemic cells have resulted in protocol modifications to improve outcome and reduce toxicity. Increasingly, the T-cell phenotype of childhood ALL has been treated more effectively with intensified regimens that include cyclophosphamide, cytarabine, and anthracylines.12,13
Late effects of treatment for childhood all
A late effect is defined as any chronic or late occurring physical or psychosocial outcome persisting or developing more than 5 years after diagnosis of the cancer. In this section we describe potential late effects in order from more common or serious health problems to less common or serious ones Table 1. Many of these late effects may have long asymptomatic intervals before end-stage disease or serious health outcomes, such as survivors with hepatitis C who develop cirrhosis or those with a late-onset cardiomyopathy who present in congestive heart failure. Included in each section is a discussion about the screening tests commonly used in long-term follow-up programs that include asymptomatic survivors4Table 2. It should be stressed that the value of most of these tests has not been studied in this population in a prospective or a well-designed retrospective manner with adequate sample sizes, which limits the strength of the recommendations. Clinicians should be selective in ordering tests and providing preventive services and should actively incorporate the patient’s concerns and fears when arriving at an individualized decision on whether to perform a test. Figure 2 is a compilation of information pertinent to the follow-up of a survivor of childhood ALL, provided as a single-page template for clinical use.
Because bone marrow transplantation (BMT) is a relatively new therapy affecting a much smaller number of ALL survivors, our review does not include the late effects related to total body irradiation and BMT.
Cognitive dysfunction and performance at school and work
As described in the section on the evolution of treatment, 24 Gy CRT is associated with cognitive dysfunction. A meta-analysis of more than 30 retrospective and prospective studies established that 24 Gy CRT in combination with MTX resulted in a mean decrease of 10 points in full-scale intelligence quotient (IQ).9 Verbal scores were affected more than performance IQ, and changes were noted to be progressive. Although more than half the patients had mild to moderate learning problems, the outcomes were highly variable, and some patients experienced 20- to 30-point losses, while others had no discernable changes.9,14 Deficits have been noted in measures of visual-spatial abilities, attention-concentration, nonverbal memory, and somatosensory functioning.8-10,15-20 Studies have also shown that girls and patients treated with CRT before the age of 4 years are at significantly higher risk. Neuropathologic changes resulting from 24 Gy CRT include leukoencephalopathy, mineralizing microangiopathy, subacute necrotizing leukomyelopathy, and intracerebral calcifications, commonly with subsequent cerebral atrophy and microcephally.21,22
Treatment with 18 Gy CRT in combination with chemotherapy also affects cognition, though not as profoundly as with 24 Gy CRT. In a retrospective study of children with ALL, randomized by risk group to receive either 18 Gy CRT with chemotherapy or chemotherapy alone, 66 survivors were subsequently tested using several cognitive measures.23 Girls who were treated with CRT/chemotherapy had a mean IQ 9 points lower than those treated with chemotherapy alone. All patients had impairments in verbal coding and short-term memory regardless of CRT use or MTX dose, suggesting that another agent such as glucocorticoids may be responsible. Other small prospective and retrospective studies have found a mild decrease in full-scale IQ in patients treated with 18 Gy CRT/chemotherapy, although subanalysis generally showed that changes were only significant for girls and patients treated at a younger age.24-27
Recent studies suggest that neurodevelopmental outcomes for survivors treated with chemotherapy alone are generally positive.28 An analysis of 30 survivors whose condition was diagnosed before the age of 12 months showed no decrease in 6 cognitive and motor indices and no sex differences.29 Though full-scale IQ was normal, Brown and colleagues30 reported that girls had significantly decreased nonverbal scores in a study of 47 ALL survivors. Fine motor disturbances and manual dexterity difficulties, which may compound learning difficulties, have been seen in 25% to 33% of ALL survivors evaluated in 2 small cross-sectional studies.31,32 Changes in cerebellar-frontal subsystems that correlate with neuropsychological deficits have also been seen in ALL patients treated with chemotherapy alone.33
The Children’s Cancer Group investigated the impact of treatment on scholastic performance of 593 adult survivors, compared with 409 sibling controls.34 Patients treated with 24 Gy CRT were more likely to enter special education or learning-disabled programs, with relative risks of 4.1 and 5.3, respectively. Previous treatment with 18 Gy CRT had less impact, with a relative risk of 4.0 to enter a special education program but no increased risk of entering a learning-disabled program. Patients treated with CRT (18 or 24 Gy) were just as likely to enter gifted and talented programs as their sibling controls. In general, survivors were as likely to finish high school and enter college as controls, but those treated with 24 Gy or treated before the age of 6 years were less likely to enter college. There were no sex differences in educational achievements.
There are no studies that explore problems in job acquisition, promotion, and retention for ALL survivors with evidence of cognitive dysfunction. Abstract thinking abilities in higher-level decision making may be problematic for some ALL survivors, particularly those treated with 24 Gy CRT. Further study is warranted, particularly in evaluating methods to assist at-risk survivors in developing job skills and applying for a job.
Obesity, physical inactivity, and risk of premature cardiovascular disease
Several retrospective cohort and cross-sectional studies have shown an increased incidence and prevalence of obesity in ALL survivors. Early studies suggested that the resulting obesity was secondary to CRT, with 38% to 57% of the survivors having a body mass index (BMI) >2 standard deviations (SDs) above the norm at the time of attainment of final height.35-38 Two recent cross-sectional studies suggest that the increased prevalence of obesity may be due to other factors. Van Dongen-Melman and coworkers39 compared the weight gain and BMI of 113 ALL survivors who had received CRT/chemotherapy or chemotherapy alone and found that children treated with a combination of prednisone and dexamethasone had the highest prevalence of obesity (44%).39 Talvensaari and colleagues40 evaluated 50 childhood cancer survivors with a median age of 18 years (including 28 ALL patients) and found an increased prevalence of obesity in survivors that was not associated with CRT.
Obesity in ALL survivors may be due in part to reduced physical activity. In a small cross-sectional study with sibling controls, ALL survivors had decreased activity levels and total daily energy expenditures that correlated with their percentage of body fat.41 Maximal and submaximal exercise capacity were reduced in another cross-sectional study.42 Similarly, in a study of 53 ALL survivors with a longer interval from ALL diagnosis (mean=10.5 years), 25% and 31%, respectively, were unable to reach normal maximal oxygen uptake and normal oxygen uptake at the anaerobic threshold.43
Changes in gross motor skills may also affect the physical activity level of ALL survivors. Balance, strength, running speed and agility, and hand grip strength were decreased in a cohort of 36 ALL survivors with a median age of 9.3 years.44 In a follow-up of this cohort, Wright and coworkers45 reported that the ALL survivors had significantly less active and passive dorsiflexion range of motion of the ankle than did controls. Younger age at diagnosis and female sex were significant predictors, while treatment with CRT did not increase risk. These studies suggest that ALL survivors should be assessed for gross motor deficits that might alter exercise choices.
In the general population, obesity and physical inactivity are risk factors for cardiovascular disease. Obesity (an especially important risk factor during young adulthood) enhances the development of hypertension, dyslipidemia, and insulin resistance.46-48 Because the median age of ALL survivors is still relatively young, there are no cohort or case-control studies evaluating the treatment-related risk of premature onset of coronary artery disease. Talvensaari and coworkers40 reported that 50 childhood cancer survivors (including 28 ALL survivors) had an increased risk of fasting hyperinsulinemia and reduced high-density lipoprotein (HDL) cholesterol compared with 50 age- and sex-matched controls. Eight of the cancer survivors with reduced spontaneous growth hormone (GH) secretion (4/8 had received CRT) had obesity, hyperinsulinemia, and reduced HDL cholesterol, fitting the criteria for cardiac dysmetabolic syndrome, a clustering of metabolic problems associated with a markedly increased risk of cardiovascular disease.49
Studies of noncancer populations may shed light on the cardiovascular risk of ALL survivors with GH deficiency. Hypopituitarism with GH deficiency in adults is associated with increased vascular mortality.50-52 Adults with GH deficiency also have an increased prevalence of dyslipidemia53,54 and insulin resistance,55 that may improve with GH therapy.56,57
Counseling on the benefits of proper diet and exercise is an important component of long-term care for ALL survivors. Periodic analysis of lipoproteins has not been prospectively studied in ALL survivors, but the US Preventive Services Task Force states that adolescents and young adults who have major risk factors for cardiovascular disease should be screened.58
Psychosocial well-being of all survivors
The long-term psychosocial welfare of ALL survivors is complex. A population-based sibling-matched control study of 93 ALL survivors who were at least 15 years postdiagnosis showed no difference in quality of life or mental health.59 Similarly, no differences were found in symptoms of anxiety and posttraumatic stress in 130 leukemia survivors and 155 controls.60 In contrast, a large cooperative study of the Children’s Cancer Group and the National Institutes of Health evaluated 580 adult survivors and 396 sibling controls and reported that survivors had greater negative mood and reported more tension, depression, anger, and confusion.61 Female, minority, and unemployed survivors reported the highest total mood disturbance. Issues related to late effects, especially cognitive dysfunction, obesity, and physical inactivity, may have an impact on the mental health of survivors.
Few data are available on the risk behavior of ALL survivors. In a cohort study of 592 young adult ALL survivors and 409 sibling controls, Tao and colleagues62 reported that ALL survivors were less likely to start smoking, but once they started they were no more likely to quit than their siblings. Fourteen percent of the ALL survivors were smokers. Although no prospective studies have evaluated the effect of smoking on the incidence and severity of late effects of ALL treatment, it will have an impact on survivors with cardiovascular risk factors, restrictive pulmonary disease, and osteopenia. Counseling on smoking cessation is imperative in the long-term health care of ALL survivors.
Osteopenia and osteoporosis
Several well-designed small to medium-size cross-sectional studies of childhood cancer survivors63-65 and ALL survivors66-71 with median ages at evaluation ranging from 12 to 25 years consistently showed reduction in bone mineral density, bone mass content (BMC), and/or age-adjusted bone mass. Age at diagnosis, interval since treatment, sex, and cumulative dosages of MTX and corticosteroids have not been consistently associated with reduction in bone mass. In contrast, CRT has consistently been identified as a risk factor, although the 3 studies that evaluated GH status showed variation in the relationship of GH deficiency and reduced bone mass.69-71 Impairment of peak bone mass is likely multifactorial in etiology, with predisposing risk factors including altered bone metabolism at the time of onset of leukemia, interference in bone metabolism by corticosteroids and MTX, and impaired bone growth and skeletal maturation caused by pituitary dysfunction/GH deficiency. In an ongoing prospective cohort study, Atkinson and coworkers72 reported that by 6 months of therapy for ALL, 64% of the children had a reduction from baseline measures of BMC, and by the end of 2 years of therapy 83% were osteopenic. Hypomagnesemia due to renal wasting of magnesium after treatment with high-dose corticosteroids and/or aminoglycosides was associated with the progression in changes and may be a key factor in the alteration of bone metabolism.
Reduction in peak bone mass in young adults is a significant risk factor for developing osteoporosis and subsequent fracture, and measures to prevent or reverse bone loss are important. Exercise increases bone density in obese children73 and young adults74 and has recently been shown by meta-analysis75 to prevent or reverse almost 1% of bone loss per year in pre- and postmenopausal women. With ALL survivors likely to be less physically active,41-43 it is essential to counsel them on the benefits of exercise in preventing cardiovascular disease and osteoporosis and help them develop an exercise plan. Additionally, counseling on calcium intake and avoidance of smoking is important. Though bone densitometry has not been an effective screening test for the general population, it has value in high-risk groups.76,77 Prospective randomized trials are needed to evaluate the usefulness and frequency of screening.
GH deficiency
Cross-sectional and longitudinal studies have consistently shown that patients treated with 24 Gy CRT have a decrease in median height of approximately 1 to 1.5 SD score, or 5 to 10 cm.37,78-84 Treatment with 18 Gy CRT85 or chemotherapy alone86,87 affect the final height to a lesser degree. Sklar and coworkers88 reported a change in final height SD score of -0.65 for patients treated with 18 Gy CRT and -0.49 for those treated with chemotherapy alone. Girls and patients treated at a younger age (<5 years) have the greatest growth reduction.37,78,88,89 These changes are thought to be secondary to GH deficiency, resulting in a blunted pubertal growth spurt. The greater the deficiency, the more profound the impairment of growth.90 Brennan and colleagues71 reported a median decrement in final height of 2.1 SD in patients with severe GH deficiency. Treatment with GH in these patients usually results in near normalization of final height.
Though GH therapy is generally stopped when children reach their final height or by the age of 18 years, deficiency persists. In a small cross-sectional study of 30 ALL survivors, 9 of 15 patients who received 24 Gy CRT (median age=21.4 years) were GH deficient.91 In another cross-sectional analysis of the GH status of 32 ALL survivors (median age=23 years), 21 of 32 were GH deficient, including 9 who were severely deficient.71 The consequences of GH deficiency in adulthood are not well understood. Small studies suggest that GH replacement may improve bone mineral density,92 body composition,93 and quality of life.94
Late onset anthracycline-induced cardiomyopathy
Anthracyclines (notably daunorubicin and doxorubicin) are often used during the induction phase of treatment, with some protocols using moderate to high dosages (Ž350 mg/m2) for high-risk patients. In the past 10 years it has become apparent that childhood cancer patients treated with an anthracycline are at increased risk for developing late-onset cardiomyopathy.95-97 Classically, anthracycline-induced cardiomyopathy is characterized by elevated afterload followed by the development of a dilated thin-walled left ventricle. Over time this can lead to a stiff and poorly compliant left ventricle. Most patients are asymptomatic, but longitudinal studies suggest that a significant proportion will experience progressive changes and may develop congestive heart failure.96,97
Lipshultz and coworkers95 assessed the cardiac status of 115 ALL survivors treated with doxorubicin and found that 65% of those treated with 228 mg/m2 or more had increased left ventricular afterload.95 In a follow-up study, Lipshultz and colleagues96 reported that female sex, younger age at treatment, higher rate of administration of doxorubicin, and cumulative dose of doxorubicin were independent risk factors for the development of altered left ventricular function. Two recent cross-sectional studies suggest that the risk of left ventricular dysfunction is uncommon in children who received cumulative doses less than 300 mg per m2.98,99 In patients treated with cumulative doses less than 270 mg per m2, Sorensen and coworkers98 did not find that female sex and younger age at treatment were risk factors. However, because late cardiac abnormalities were seen in survivors who received only 90 mg per m2, there might be no absolute level below which cardiotoxicity can be prevented.
Because of the concerns about cardiotoxicity, most recent protocols limit anthracycline doses to less than 300 mg per m2, and the use of cardioprotectants such as dexrazoxane in children is under investigation.100 Primary care physicians who provide follow-up care for adult survivors should communicate with oncologists at the treating institution, obtain information about the cumulative dosage of anthracyclines, and discuss long-term screening. Because patients with anthracycline-induced cardiomyopathies generally have a prolonged asymptomatic interval before becoming symptomatic, interval screening is recommended. Optimal timing and testing modality for screening have not been prospectively studied. It is currently recommended that patients who received 300 mg/m2 or more of an anthracycline have a screening echocardiogram every 2 to 3 years to evaluate left ventricular function and shortening fraction.101 It is also important to question patients regarding symptoms of congestive heart failure and to aggressively evaluate them if present.
Hepatitis C
Because most ALL patients receive blood products during therapy, those treated before adequate blood donor screening for hepatitis C was initiated in the early 1990s are at risk for chronic liver disease.102 The prevalence of circulating hepatitis C virus (HCV) ribonucleic acid (RNA) in ALL patients treated in Italy before 1990 ranges from 23% to 49%.103-105 The natural history of ALL survivors with hepatitis C is not well understood. In an Italian study, only 4% of the 56 HCV-RNA seropositive patients had persistently elevated alanine aminotransferase (ALT) over the course of follow-up (mean=17 years).106 For a median of 14 years, 81 survivors of various childhood cancers who were HCV-RNA seropositive were followed, and none showed progression to liver failure.107 In contrast, Paul and coworkers108 reported that 12% of 75 leukemia survivors were anti-HCV positive, 6 of 9 had liver biopsies that showed at least moderate portal inflammation, and half had bridging fibrosis. The Centers for Disease Control and Prevention102 recommend universal screening with anti-HCV for all patients who received blood products before July 1992.
Second malignant neoplasms
Second malignant neoplasms (SMN) are rare in ALL survivors. Thirteen SMNs were diagnosed a median of 6.7 years from ALL diagnosis in a cohort study of 1597 ALL survivors and were associated with the use of radiation (8/13, CNS or head and neck) or chemotherapy (3/13, hematopoietic).109 The cumulative incidence of brain tumors at 20 years in a cohort of 1612 patients was only 1.39%, and more than half of these tumors were either low-grade or benign.110 CNS tumors did not occur in patients treated with chemotherapy alone. Thyroid tumors (predominantly papillary carcinoma) can rarely occur after treatment with cranial or craniospinal irradiation.111,112 Cases of basal cell carcinoma along the spinal axis have also been reported in patients treated with craniospinal irradiation.113,114
Therapy-related acute myelogenous leukemia (t-AML) has been seen following treatment of several childhood cancers, such as ALL and Hodgkin’s and non– Hodgkin’s lymphoma. Cohort studies have shown that agents with leukemogenic potential include alklyating agents and epidophyllotoxin chemotherapy.115-121 Most t-AMLs occur within 8 years of treatment, although cases occurring up to 13 years have been reported.115 Myelodysplasia (especially pancytopenia) generally precedes t-AML. The risk of t-AML following treatment for ALL has been small in 2 cohort studies.109,122 However, because precancerous states (myelodysplastic changes or myelodysplastic syndrome) are usually antecedent to t-AML and early diagnosis may improve outcomes, most institutions recommend obtaining a complete blood count (CBC) with a platelet count and a white blood cell differential in the routine follow-up of ALL survivors who have been treated with an alkylating agent, such as cyclophosphamide, or an epidophyllotoxin, such as etoposide. How long and how frequently a CBC should be obtained in follow-up of an ALL survivor have not been established.
Fertility and reproduction
Most antimetabolite-based treatment protocols for ALL do not affect long-term fertility for men or women.123,124 Craniospinal and abdominal irradiation have been associated with infertility in both sexes but are no longer used for ALL.125-127 Cyclophosphamide (an alkylating agent commonly used in earlier protocols but currently limited to high-risk patients) is also associated with infertility in a dosedependent fashion in both sexes.124,128,129 Resolution of germ-cell dysfunction may occur in men over time, but fertility remains poor for some. Women survivors treated with craniospinal or abdominal irradiation or with cyclophosphamide are at risk for ovarian failure and premature menopause and thus may be at increased risk for osteoporosis. If ovarian failure is suspected, measurement of follicle-stimulating hormone, luteinizing hormone, and serum estradiol and an evaluation by an endocrinologist should be considered.
ALL survivors should know that preliminary studies suggest that treatment is not associated with an increase in congenital malformations of their offspring. In a population-based prospective cohort study an increased rate of congenital defects was not found among 299 adult survivors.130
Ocular abnormalities
Ocular abnormalities in patients treated with CRT are common but generally asymptomatic. Two studies have evaluated the effect of CRT and systemic corticosteroids on the eyes. In a study of 82 ALL survivors who were a mean of 32 months after completion of therapy, 52% of the patients had posterior subcapsular cataracts (PSC) that were generally not visually significant and were not related to age at treatment or gender.131 Eighty-three percent of the 18 patients who had received CRT and systemic corticosteroids were noted to have asymptomatic ocular abnormalities after a median surveillance of 4.1 years.132 Optical densities of the lens were seen in 13 of the 18 of the survivors. There have been no published studies evaluating long-term survivors who received systemic corticosteroids without CRT. Periodic vision and cataract screening is recommended for ALL survivors treated with CRT and should be considered for all survivors of ALL until the risk of prolonged corticosteroid use in childhood is better understood.
Dental and periodontal disease
ALL survivors, especially those treated with CRT, are more likely to have problems with tooth development and be at risk for periodontal disease. In a large retrospective evaluation of dental records, 39.5% of ALL survivors had a dental abnormality, including root stunting (24.4%), microdontia (18.9%), or hypodontia (8.5%).133 Patients who were treated at an age younger than 8 years or who received CRT had more dental abnormalities than the other groups. Similar findings were seen in 2 smaller cross-sectional studies. Abnormal dental development occurred in 95% of all patients and 100% of patients aged 5 years or younger at diagnosis.134 Abnormalities included tooth agenesis, arrested tooth development, microdontia, and enamel dysplasia. Patients who received CRT and those treated at an age younger than 5 years had higher severity scores. Survivors did not have increased caries.135 However, patients younger than 5 years who were treated with cranial irradiation were found to have higher plaque and gingivitis scores, suggesting an increased risk of periodontal disease. A periodic dental and periodontal evaluation is recommended for survivors treated with CRT or at a young age.
Thyroid-related disorders
Following treatment with CRT, hypothyroidism infrequently occurs in ALL survivors through damage to the hypothalamic-pituitary-thyroid axis and/or the direct effect of radiation of the gland. Mohn and colleagues136 reported that 8 of 24 childhood ALL survivors who had received CRT (either 18 or 24 Gy) had either a low basal thyroid-stimulating hormone (TSH) or low peak TSH after thyrotropin-releasing hormone stimulation. Robison and colleagues137 reported that 10% of 175 ALL survivors who had been treated with either 18 or 24 Gy CRT or craniospinal radiation (CS-RT) therapy had a thyroid abnormality, including 5 children with primary hypothyroidism. Pasqualini and colleagues138 reported that 6 of 10 ALL survivors who received either CRT or CS-RT had subtle evidence of primary hypothyroidism. In contrast, 3 cross-sectional studies did not find evidence of primary hypothyroidism in 13, 31, and 64 patients, respectively.1,139-141 Littley and coworkers142 suggest that hypopituitarism is commonly underdiagnosed secondary to the subtle manifestations and insidious progression of disease. Radioactive scatter to the thyroid occurs with CRT in a dose-dependent fashion,143 and ALL survivors treated with either 18 or 24 Gy CRT are at risk for secondary hypothyroidism, thyroid nodules, and thyroid carcinoma.111 Periodic screening with TSH and free T-4 are recommended in ALL survivors treated with CRT. Further screening of the asymptomatic survivor with thyrotropin-releasing hormone stimulation test or ultrasound of the thyroid gland are costly and have not been prospectively studied.
Pulmonary late effects
ALL survivors may have an increased prevalence of mild, generally subclinical, restrictive pulmonary disease. In a small cross-sectional study of ALL survivors, Shaw and coworkers144 reported mild restrictive changes, with patients treated at a younger age at higher risk. Similarly, an analysis of 70 leukemia survivors found mild but significant decreases in forced vital capacity (FVC), forced expiratory volume in 1 second (FEV-1), total lung capacity (TLC), and transfer for carbon monoxide (DLCO).42 Cyclophosphamide, craniospinal irradiation, and a history of chest infections during treatment were independent variables associated with reductions in FEV-1, FVC, and TLC, while anthracyclines and craniospinal irradiation were associated with reductions in DLCO. ALL survivors also had impaired submaximal and maximal exercise capacity. These findings were further supported by analysis of a recent cross-sectional study of 128 patients a median of 7.6 years from therapy completion that reported an increased prevalence of subclinical restrictive pulmonary disease in ALL survivors.145 The long-term consequences and the possible role of smoking or other inhalant exposures need to be studied.
Liver dysfunction (Non-Hepatitis C)
During treatment with methotrexate (especially high-dose ranges) elevations of transaminases are common and generally transient. Two small longitudinal studies following ALL survivors for up to 7 years after completion of therapy did not report any patients with persistent transaminasemia, although Bessho and colleagues noted that 6 of 13 of their ALL survivors had elevated 2-hour postprandial bile acid levels, a more sensitive predictor of liver cirrhosis than transaminase level.146,147 Farrow and coworkers148 found that of 114 survivors who had ALT elevations greater than 5 times the upper limit of normal during therapy, only 17 (14.9%) had elevations persistently. Eight of these patients had chronic HCV infections. Of the remaining 9 patients, only 1 had a persistently elevated transaminase of greater than 2 times normal.
Although there are currently no data evaluating ALL survivors for long-term liver-related complications secondary to methotrexate, studies in patients with juvenile rheumatoid arthritis show that septal and portal fibrosis can occur with weekly low-dose methotrexate treatment of durations as short as 17 months.149 Obesity may be an associated risk factor for the development of cirrhosis in juvenile rheumatoid arthritis patients treated with methotrexate. Because of these potential risks, periodic measurement of ALT is recommended in follow-up of ALL survivors.
Urologic late effects
Cyclophosphamide is a long-recognized cause of hemorrhagic cystitis and a well-established bladder carcinogen. In a retrospective review150 of 314 children with ALL who were treated with cyclophosphamide between 1963 and 1973, 8% developed hemorrhagic cystitis. The frequency of diagnosis was not related to age or sex, but African American children were at higher risk. Cyclophosphamide-induced hemorrhagic cystitis generally presents during therapy, with children complaining of gross hematuria or irritative voiding complaints.151 Concurrent treatment with oral sodium 2-mercapatoethanesulfonate appears to markedly decrease the incidence of cyclophosphamide-induced hemorrhagic cystitis.152 In a nested case-control study of survivors of non–Hodgkin’s lymphoma, Travis and colleagues153 reported that there was a 2.4-fold increased risk of bladder cancer in patients treated with cumulative dosages of cyclophosphamide lower than 20 g. Because of the risk of chronic hemorrhagic cystitis and bladder cancer, ALL survivors treated with cyclophosphamide should have periodic screening urinalysis, and their review of systems should include voiding problems.
Alopecia
Alopecia is a bothersome late effect secondary to treatment with 24 Gy CRT for which there are no available treatments. In a retrospective study of 273 ALL survivors treated with CRT, 10% had alopecia.154
Acknowledgement
Dr Oeffinger received partial support for this work through the American Academy of Family Physicians Foundation Advanced Research Training Grant and the Robert Wood Johnson Foundation Generalist Physician Faculty Scholars Program.
We would like to thank Drs George Buchanan, Melissa Hudson, and Neyssa Marina for their critical review of this manuscript and Ms Laura Snell and Dr James Tysinger for their editing assistance.
1. MA, Ries LAG, Gurney JG, Ross JA. Leukemia. In: Ries LAG, Smith MA, Gurney JG, et al, eds. Cancer incidence and survival among children and adolescents: United States SEER Program 1975-1995, National Cancer Institute, SEER Program. Bethesda, MD: National Institutes of Health; 1999. NIH pub. no. 99-4649.
2. KC, Eshelman DA, Tomlinson GE, Buchanan GR. Programs for adult survivors of childhood cancer. J Clin Oncol 1998;16:2864-67.
3. DS. Transition to adult health care for adolescents and young adults with cancer. Cancer 1993;71:3411-14.
4. KC, Eshelman DA, Tomlinson GE, Buchanan GR, Foster BE. Grading of late effects in young adult survivors of childhood cancer followed in an ambulatory adult setting. Cancer 2000;88:1687-95.
5. H. The natural history of untreated acute leukemia. Ann NU Acad Sci 1954;60:322-58.
6. S, Diamond LK, Mercer RD, et al. Temporary remissions in acute leukemia in children produced by folic acid antagonist 4-aminopteroylglutamic acid (aminopterin). N Engl J Med 1948;238:787-93.
7. L, Gelber R, Cohen H, et al. Four-agent induction and intensive asparaginase therapy for treatment of childhood acute lymphoblastic leukemia. N Engl J Med 1986;315:657-63.
8. LL, Nesbit ME, Jr, Sather HN, Meadows AT, Ortega JA, Hammond GD. Factors associated with IQ scores in long-term survivors of childhood acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1984;6:115-21.
9. P, Waters B, Said J, Stevens M. Cognitive effects of cranial irradiation in leukaemia: a survey and meta-analysis. J Child Psychol Psychiatry 1988;29:839-52.
10. JM, Kornblith AB, Jones D, et al. A comparative study of the long term psychosocial functioning of childhood acute lymphoblastic leukemia survivors treated by intrathecal methotrexate with or without cranial radiation. Cancer 1998;82:208-18.
11. D, Reaman G, Bleyer W, et al. Successful prevention of central nervous (CNS) leukemia without cranial radiation in children with high risk acute lymphoblastic leukemia (ALL): a preliminary report. Proc Am Soc Clin Oncol 1989;8:828.-
12. W, Shuster J, Falletta J, et al. Clinical features and outcome in childhood T-cell leukemia-lymphoma according to stage of thymocyte differentiation: a Pediatric Onoclogy Group study. Blood 1988;72:1891-97.
13. CH, Behm FG, Singh B, et al. Heterogeneity of presenting features and their relation to treatment outcome in 120 children with T-cell acute lymphoblastic leukemia. Blood 1990;75:174-79.
14. M, Azuma E, Ido M, et al. Ten-year survey of the intellectual deficits in children with acute lymphoblastic leukemia receiving chemoimmunotherapy. Med Pediatr Oncol 1993;21:435-40.
15. DP, Urion DK, Tarbell NJ, Niemeyer C, Gelber R, Sallan SE. Late effects of central nervous system treatment of acute lymphoblastic leukemia in childhood are sex-dependent. Dev Med Child Neurol 1990;32:238-48.
16. AE, Aitken K, Eden OB. Computerized psychometry screening in long-term survivors of childhood acute lymphoblastic leukemia. Pediatr Hematol Oncol 1988;5:197-208.
17. H, Huk WJ, Ueberall MA, et al. CNS late effects after ALL therapy in childhood. Part I: Neuroradiological findings in long-term survivors of childhood ALL—an evaluation of the interferences between morphology and neuropsychological performance—the German Late Effects Working Group. Med Pediatr Oncol 1997;28:387-400.
18. JA, Kaleita TA, Noll RB, et al. CNS prophylaxis of childhood leukemia: what are the long-term neurological, neuropsychological, and behavioral effects? Neuropsychol Rev 1991;2:147-77.
19. JA, Waters BG, Cousens P, Stevens MM. Neuropsychological sequelae of central nervous system prophylaxis in survivors of childhood acute lymphoblastic leukemia. J Consult Clin Psychol 1989;57:251-56.
20. J, Horrocks J, Britton PG, Kernahan J. Attentional ability among survivors of leukaemia. Arch Dis Child 1999;80:318-23.
21. AS, Nesbit ME. Neuropsychologic (cognitive) disabilities in long-term survivors of childhood cancer. Pediatrician 1991;18:11-19.
22. RK, Kovnar E, Langston J, et al. Long-term survivors of leukemia treated in infancy: factors associated with neuropsychologic status. J Clin Oncol 1992;10:1095-102.
23. DP, Tarbell NJ, Fairclough D, et al. Cognitive sequelae of treatment in childhood acute lymphoblastic leukemia: cranial radiation requires an accomplice. J Clin Oncol 1995;13:2490-96.
24. CL, Varni JW, Katz ER. Cognitive functioning in long-term survivors of childhood leukemia: a prospective analysis. J Dev Behav Pediatr 1990;11:301-05.
25. M, Brouwers P, Valsecchi MG, Van Veldhuizen A, Huisman J. Association of 1800 cGy cranial irradiation with intellectual function in children with acute lymphoblastic leukaemia. Lancet 1994;344:224-27.
26. E, Anderson V, Godber T, Ekert H. Risk factors for intellectual and educational sequelae of cranial irradiation in childhood acute lymphoblastic leukaemia. Br J Cancer 1996;73:825-30.
27. V, Godber T, Smibert E, Ekert H. Neurobehavioural sequelae following cranial irradiation and chemotherapy in children: an analysis of risk factors. Pediatr Rehabil 1997;1:63-76.
28. Bleyer A. CNS chemoradiotherapy of childhood leukemia: the plot thickens but the ending bodes well. J Clin Oncol 1995;13:2480-82.
29. TA, Reaman GH, MacLean WE, Sather HN, Whitt JK. Neurodevelopmental outcome of infants with acute lymphoblastic leukemia: a Children’s Cancer Group report. Cancer 1999;85:1859-65.
30. RT, Madan-Swain A, Walco GA, et al. Cognitive and academic late effects among children previously treated for acute lymphocytic leukemia receiving chemotherapy as CNS prophylaxis. J Pediatr Psychol 1998;23:333-40.
31. L. Clinical neurological findings of children with acute lymphoblastic leukaemia at diagnosis and during treatment. Eur J Pediatr 1993;152:115-19.
32. HA, Schoemaker MM, Hofte M, et al. Fine motor and handwriting problems after treatment for childhood acute lymphoblastic leukemia. Med Pediatr Oncol 1996;27:551-55.
33. PG, Ciesielski KT, Hart BL, Benzel EC, Sanders JA. Evidence for cerebellar-frontal subsystem changes in children treated with intrathecal chemotherapy for leukemia. Arch Neurol 1998;55:1561-68.
34. R, Fears TR, Robison LL, et al. Educational attainment in long-term survivors of childhood acute lymphoblastic leukemia. JAMA 1994;272:1427-32.
35. P, Chen CH. Prevalence of obesity in children after therapy for acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1986;8:294-99.
36. I, Reilly JJ, Gibson BE, Donaldson MD. Patterns of obesity in boys and girls after treatment for acute lymphoblastic leukaemia. Arch Dis Child 1994;71:147-49.
37. MJ, Ochs JJ, Schriock EA, Carter M. A method of predicting adult height and obesity in long-term survivors of childhood acute lymphoblastic leukemia. J Clin Oncol 1992;10:128-33.
38. M, Didcock E, Davies HA, Ogilvy-Stuart AL, Wales JK, Shalet SM. High incidence of obesity in young adults after treatment of acute lymphoblastic leukemia in childhood. J Pediatr 1995;127:63-67.
39. Dongen-Melman JE, Hokken-Koelega AC, Hahlen K, De Groot A, Tromp CG, Egeler RM. Obesity after successful treatment of acute lymphoblastic leukemia in childhood. Pediatr Res 1995;38:86-90.
40. KK, Lanning M, Tapanainen P, Knip M. Long-term survivors of childhood cancer have an increased risk of manifesting the metabolic syndrome. J Clin Endocrinol Metab 1996;81:3051-55.
41. JT, Bell W, Webb DK, Gregory JW. Daily energy expenditure and physical activity in survivors of childhood malignancy. Pediatr Res 1998;43:607-13.
42. ME, Faragher EB, Jones PH, Woodcock A. Lung function and exercise capacity in survivors of childhood leukaemia. Med Pediatr Oncol 1995;24:222-30.
43. P, Gutjahr P, Stopfkuchen H. Physical performance in long-term survivors of acute leukaemia in childhood. Eur J Pediatr 1998;157:464-67.
44. MJ, Halton JM, Martin RF, Barr RD. Long-term gross motor performance following treatment for acute lymphoblastic leukemia. Med Pediatr Oncol 1998;3:86-90.
45. MJ, Halton JM, Barr RD. Limitation of ankle range of motion in survivors of acute lymphoblastic leukemia: a cross-sectional study. Med Pediatr Oncol 1999;32:279-82.
46. DS, Dietz WH, Srinivasan SR, Berenson GS. The relation of overweight to cardiovascular risk factors among children and adolescents: the Bogalusa Heart Study. Pediatrics 1999;103:1175-82.
47. M, Vanhala P, Kumpusalo E, Halonen P, Takala J. Relation between obesity from childhood to adulthood and the metabolic syndrome: population based study. BMJ 1998;317:319-21.
48. GS, Srinivasan SR, Bao W, et al. Association between multiple cardiovascular risk factors and atherosclerosis in children and young adults. N Engl J Med 1998;338:1650-56.
49. TC, Deedwania PC. The cardiovascular dysmetabolic syndrome. Am J Med 1998;105:77S-82S.
50. T, Bengtsson BA. Premature mortality due to cardiovascular disease in hypopituitism. Lancet 1990;336:285-88.
51. AS, Van’t Hoff W, Jones PJ, Clayton RN. The effect of hypopituitarism on life expectancy. J Clin Endocrinol Metab 1996;81:1169-72.
52. EM, Bulow B, Eskilsson J, Hagmar L. High incidence of cardiovascular disease and increased prevalence of cardiovascular risk factors in women with hypopituitarism not receiving growth hormone treatment: preliminary results. Growth Horm IGF Res 1999;9 (suppl):21-24.
53. MB. Effect of growth hormone on carbohydrate and lipid metabolism. Endocr Rev 1987;8:115-31.
54. FL, O’Neal D, Kamarudin N, Alford FP, Best JD. Growth hormone deficiency and cardiovascular risk. Baillieres Clin Endocrinol Metab 1998;12:199-216.
55. SA, Henderson A, Niththyananthan R, et al. The effects of short and long-term growth hormone replacement therapy in hypopituitary adults on lipid metabolism and carbohydrate tolerance. J Clin Endocrinol Metab 1995;80:356-63.
56. KA, Gray R, Anyaoku V, et al. Effects of four years’ treatment with biosynthetic human growth hormone (GH) on glucose homeostasis, insulin secretion and lipid metabolism in GH-deficient adults. Clin Endocrinol 1998;48:795-802.
57. D, Hew FL, Sikaris K, Ward G, Alford F, Best JD. Low density lipoprotein particle size in hypopituitary adults receiving conventional hormone replacement therapy. J Clin Endocrinol Metab 1996;81:2448-54.
58. Preventive Services Task Force. Guide to clinical preventive services. 2nd ed. Washington, DC: US Department of Health and Human Services; 1996.
59. PJ, Holen A, Glomstein A, et al. Long-term survival and quality of life in patients treated with a national ALL protocol 15-20 years earlier: IDM/HDM and late effects? Pediatr Hematol Oncol 1997;14:513-24.
60. AE. Posttraumatic distress in childhood cancer survivors and their parents. Med Pediatr Oncol 1998;1 (suppl):60-68.
61. LK, Chen E, Weiss R, et al. Comparison of psychologic outcome in adult survivors of childhood acute lymphoblastic leukemia versus sibling controls: a cooperative Children’s Cancer Group and National Institutes of Health study. J Clin Oncol 1997;15:547-56.
62. ML, Guo MD, Weiss R, et al. Smoking in adult survivors of childhood acute lymphoblastic leukemia. J Natl Cancer Inst 1998;90:219-25.
63. PB, Hough SF, Nel ED, van Riet FA, Beneke T, Wessels G. Bone mineral density in long-term survivors of childhood cancer. Int J Cancer Suppl 1998;11:44-7.
64. J, Hsieh K, Kalaitzoglou G, et al. Bone mineral density in young adult survivors of childhood cancer. J Pediatr Hematol Oncol 1998;20:241-45.
65. R, Brosnan P, Delpassand A, Zietz H, Klein MJ, Jaffe N. Osteopenia in young adult survivors of childhood cancer. Med Pediatr Oncol 1999;32:272-78.
66. V, Carlson ME, Roe TF, Ortega JA. Osteoporosis after cranial irradiation for acute lymphoblastic leukemia. J Pediatr 1990;117:238-44.
67. P, Komulainen J, Voutilainen R, et al. Reduced bone mineral density in long-term survivors of childhood acute lymphoblastic leukemia. J Pediatr Hematol Oncol 1998;20:234-40.
68. JT, Evans WD, Webb DK, Bell W, Gregory JW. Relative osteopenia after treatment for acute lymphoblastic leukemia. Pediatr Res 1999;45:544-51.
69. K, Holm K, Michaelsen KF, Hertz H, Muller J, Molgaard C. Bone mass after treatment for acute lymphoblastic leukemia in childhood. J Clin Oncol 1998;16:3752-60.
70. JJ, Kardos G, Roos JC, et al. Bone mineral density and markers of bone turnover in young adult survivors of childhood lymphoblastic leukaemia. Clin Endocrinol 1999;50:237-44.
71. BM, Rahim A, Mackie EM, Eden OB, Shalet SM. Clin Endocrinol 1998;48:777-783.
72. SA, Halton JM, Bradley C, Wu B, Barr RD. Bone and mineral abnormalities in childhood acute lymphoblastic leukemia: influence of disease, drugs and nutrition. Int J Cancer Suppl 1998;11:35-39.
73. B, Owens S, Okuyama T, Riggs S, Ferguson M, Litaker M. Effect of physical training and its cessation on percent fat and bone density of children with obesity. Obes Res 1999;7:208-14.
74. O, Kristinsson JO, Stefansson SO, Valdimarsson S, Sigurdsson G. Lean mass and physical activity as predictors of bone mineral density in 16-20-year old women. J Intern Med 1999;245:489-96.
75. I, van Croonenborg JJ, Kemper HC, Kostense PJ, Twisk JW. The effect of exercise training programs on bone mass: a meta-analysis of published controlled trials in pre- and postmenopausal women. Osteoporos Int 1999;9:1-12.
76. D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 1996;312:1254-59.
77. D, Sampietro-Colom L, Marshall D, Rico R, Granados A, Asua J. The effectiveness of bone density measurement and associated treatments for prevention of fractures: an international collaborative review. Int J Technol Assess Health Care 1998;14:237-54.
78. LL, Nesbit ME, Jr, Sather HN, Meadows AT, Ortega JA, Hammond GD. Height of children successfully treated for acute lymphoblastic leukemia: a report from the Late Effects Study Committee of Children’s Cancer Study Group. Med Pediatr Oncol 1985;13:14-21.
79. EA, Schell MJ, Carter M, Hustu O, Ochs JJ. Abnormal growth patterns and adult short stature in 115 long-term survivors of childhood leukemia. J Clin Oncol 1991;9:400-05.
80. M, Stanhope R, Chessells JM, Leiper AD. Impaired pubertal growth in acute lymphoblastic leukaemia. Arch Dis Child 1991;66:1403-07.
81. K, Dorffel W, Timme J, et al. Final height and puberty in 40 patients after antileukaemic treatment during childhood. Eur J Pediatr 1997;156:272-76.
82. P, Moell C, Cornu G, Malvaux P, Maes M. Subnormal growth during puberty in children treated for acute lymphoblastic leukemia. Pediatr Hematol Oncol 1992;9:217-22.
83. AC, van Doorn JW, Hahlen K, Stijnen T, de Muinck Keizer-Schrama SM, Drop SL. Long-term effects of treatment for acute lymphoblastic leukemia with and without cranial irradiation on growth and puberty: a comparative study. Pediatr Res 1993;33:577-82.
84. JA, Pollock BH, Jacaruso D, Morad A. Final attained height in patients successfully treated for childhood acute lymphoblastic leukemia. J Pediatr 1993;123:546-52.
85. AE, Adan L, Leverger G, Souberbielle JC, Schaison G, Brauner R. Growth hormone secretion, puberty and adult height after cranial irradiation with 18 Gy for leukaemia. Eur J Pediatr 1998;157:703-07.
86. J, Villaizan CJ, Garcia-Foncillas J, Azcona C, Salvador J, Sierrasesumaga L. Chemotherapy-induced growth hormone deficiency in children with cancer. Med Pediatr Oncol 1995;25:90-5.
87. J, Villaizan CJ, Garcia-Foncillas J, Salvador J, Sierrasesumaga L. Growth and growth hormone secretion in children with cancer treated with chemotherapy. J Pediatr 1997;131:105-12.
88. C, Mertens A, Walter A, et al. Final height after treatment for childhood acute lymphoblastic leukemia: comparison of no cranial irradiation with 1800 and 2400 centigrays of cranial irradiation. J Pediatr 1993;123:59-64.
89. A, Cacciari E, Rosito P, et al. Longitudinal growth and final height in long-term survivors of childhood leukaemia. Eur J Pediatr 1994;153:726-30.
90. TG, Byrne GC, Jones TW. Growth and growth hormone secretion after treatment for acute lymphoblastic leukemia in childhood 18-Gy versus 24-Gy cranial irradiation. J Pediatr Hematol Oncol 1995;17:167-71.
91. NH, Fisker S, Clausen N, Tuovinen V, Sindet-Pedersen S, Christiansen JS. Growth and endocrinological disorders up to 21 years after treatment for acute lymphoblastic leukemia in childhood. Med Pediatr Oncol 1998;30:351-56.
92. O’Halloran DJ, Tsatsoulis A, Whitehouse RW, Holmes SJ, Adams JE, Shalet SM. Increased bone density after recombinant human growth hormone (GH) therapy in adults with isolated GH deficiency. J Clin Endocrinol Metab 1993;76:1344-48.
93. F, Cuneo RC, Hesp R, Sonksen PH. The effects of treatment with recombinant human growth hormone on body composition and metabolism in adults with growth hormone deficiency. N Engl J Med 1989;321:1797-803.
94. P, Broman JE, Hetta J, et al. Quality of life in adults with growth hormone (GH) deficiency: response to treatment with recombinant human GH in a placebo-controlled 21-month trial. J Clin Endocrinol Metab 1995;80:3585-90.
95. SE, Colan SD, Gelber RD, Perez-Atayde AR, Sallan SE, Sanders SP. Late cardiac effects of doxorubicin therapy for acute lymphoblastic leukemia in childhood. N Engl J Med 1991;324:843-45.
96. SE, Lipsitz SR, Mone SM, et al. Female sex and drug dose as risk factors for late cardiotoxic effects of doxorubicin therapy for childhood cancer. N Engl J Med 1995;332:1738-43.
97. MA, Lipshultz SE. Epidemiology of anthracycline cardiotoxicity in children and adults. Semin Oncol 1998;25(suppl):72-85.
98. K, Levitt G, Bull C, Chessells J, Sullivan I. Anthracycline dose in childhood acute lymphoblastic leukemia: issues of early survival versus late cardiotoxicity. J Clin Oncol 1997;15:61-68.
99. K, Holm K, Lipsitz SR, et al. Relationship between cumulative anthracycline dose and late cardiotoxicity in childhood acute lymphoblastic leukemia. J Clin Oncol 1998;16:545-50.
100. LH. Ameliorating anthracycline cardiotoxicity in children with cancer: clinical trials with dexrazoxane. Semin Oncol 1998;25:86-92.
101. LJ, Graham T, Hurwitz R, et al. Guidelines for cardiac monitoring of children during and after anthracycline therapy: report of the Cardiology Committee of the Childrens Cancer Study Group. Pediatrics 1992;89:942-49.
102. for Disease Control and Prevention. Recommendations for prevention and control of hepatitis C virus (HCV) infection and HCV-related chronic disease. MMWR 1998;47:1-39.
103. M, Maggiore G, Silini E, Bono F, Vigano C. Hepatitis C virus infection in children treated for acute lymphoblastic leukemia. Blood 1994;84:2919-22.
104. SP, Ragusa R, Sciacca A, et al. Incidence and morbidity of infection by hepatitis C virus in children with acute lymphoblastic leukaemia. Eur J Pediatr 1994;153:271-75.
105. A, Testa M, Pontisso P, et al. Prevalence and natural history of hepatitis C infection in patients cured of childhood leukemia. Blood 1997;90:4628-33.
106. A, Alberti A. Hepatitis C virus serum markers and liver disease in children with leukemia. Leuk Lymphoma 1995;17:245-49.
107. S, Petris MG, Rossetti F, et al. Chronic hepatitis C virus infection after treatment for pediatric malignancy. Blood 1997;90:1315-20.
108. IM, Sanders J, Ruggiero F, Andrews T, Ungar D, Eyster ME. Chronic hepatitis C virus infections in leukemia survivors: prevalence, viral load, and severity of liver disease. Blood 1999;93:3672-77.
109. Dalton VM, Gelber RD, Li F, Donnelly MJ, Tarbell NJ, Sallan SE. Second malignancies in patients treated for childhood acute lymphoblastic leukemia. J Clin Oncol 1998;16:2848-53.
110. AW, Hancock ML, Pui CH, et al. Secondary brain tumors in children treated for acute lymphoblastic leukemia at St Jude Children’s Research Hospital. J Clin Oncol 1998;16:3761-67.
111. P, Straaten A, Gutjahr P. Secondary thyroid carcinoma after treatment for childhood cancer. Med Pediatr Oncol 1998;31:91-95.
112. Y, Leverger G, Carrere A, et al. Second thyroid neoplasms after prophylactic cranial irradiation for acute lymphoblastic leukemia. Am J Hematol 1998;59:91-94.
113. T, Ikuta H, Hibi S, Todo S. Second cutaneous neoplasms after acute lymphoblastic leukemia in childhood. Int J Hematol 1993;59:67-71.
114. J, Velasco-Benito JA, Pena-Penabad C, Armijo M. Basal cell carcioma in a girl after cobalt irradiation to the cranium for acute lymphoblastic leukemia: case report and literature review. Pediatr Dermatol 1996;13:54-57.
115. J, Philip P, Larsen SO, et al. Therapy-related myelodysplasia and acute myeloid leukemia: cytogenetic characteristics of 115 consecutive cases and risk in seven cohorts of patients treated intensively for malignant diseases in the Copenhagen series. Leukemia 1993;7:1975-86.
116. N, Shuster JJ, Bowman WP, et al. Intensive oral methotrexate protects against lymphoid marrow relapse in childhood B-precursor acute lymphoblastic leukemia. J Clin Oncol 1996;14:2803-11.
117. C, Hartmann JT, Kanz L, Bokemeyer C. Risk of secondary myeloid leukemia and myelodysplastic syndrome following standard-dose chemotherapy or high-dose chemotherapy with stem cell support in patients with potentially curable malignancies. J Cancer Res Clin Oncol 1998;124:207-14.
118. HM, Keating MJ. Therapy-related leukemia and myelodysplastic syndrome. Semin Oncol 1987;14:435-43.
119. MA, Rubinstein L, Anderson JR, et al. Secondary leukemia or myelodysplastic syndrome after treatment with epipodophyllotoxins. J Clin Oncol 1999;17:569-77.
120. MA, Rubinstein L, Cazenave L, et al. Report of the Cancer Therapy Evaluation Program monitoring plan for secondary acute myeloid leukemia following treatment with epipodophyllotoxins. J Natl Cancer Inst 1993;85:554-58.
121. CH, Relling MV, Rivera GK, et al. Epipodophyllotoxin-related acute myeloid leukemia: a study of 35 cases. Leukemia 1995;9:1990-96.
122. M, Akiyama Y, Koishi S, et al. Second malignancy following treatment of acute lymphoblastic leukemia in children. Int J Hematol 1998;67:397-401.
123. R, Clausen N, Siimes MA, et al. Reproduction following treatment for childhood leukemia: a population-based prospective cohort study of fertility and offspring. Med Pediatr Oncol 1991;19:459-66.
124. GA, Jenney ME. The reproductive system after childhood cancer. Br J Obstet Gynaecol 1998;105:946-53.
125. Wallace WH, Shalet SM, Tetlow LJ, Morris-Jones PH. Ovarian function following the treatment of childhood acute lymphoblastic leukaemia. Med Pediatr Oncol 1993;21:333-39.
126. MR, Robison LL, Nesbit ME, et al. Effects of radiation on ovarian function in long-term survivors of childhood acute lymphoblastic leukemia: a report from the Children’s Cancer Study Group. J Clin Oncol 1987;5:1759-65.
127. CA, Robison LL, Nesbit ME, et al. Effects of radiation on testicular function in long-term survivors of childhood acute lymphoblastic leukemia: a report from the Children’s Cancer Group. J Clin Oncol 1990;8:1981-87.
128. T, Kishi K, Imashuku S, et al. Testicular histology and function following long-term chemotherapy of acute leukemia in children and outcome of the patients who received testicular biopsy. Am J Pediatr Hematol Oncol 1986;8:288-93.
129. WH, Shalet SM, Lendon M, Morris-Jones PH. Male fertility in long-term survivors of childhood acute lymphoblastic leukaemia. Int J Androl 1991;14:312-19.
130. LB, Nicholson HS, Brasseux C, et al. Birth defects in offspring of adult survivors of childhood acute lymphoblastic leukemia: a Children’s Cancer Group/National Institutes of Health Report. Cancer 1996;78:169-76.
131. DL, Smith LE, Turner SJ, Gelber RD, Sallan SE. Ophthalmic evaluation of survivors of acute lymphoblastic leukemia. Ophthalmology 1988;95:151-55.
132. RG, Jr, Chauvenet AR, Smith TJ, Schwartz AC. Ophthalmic evaluation of long-term survivors of childhood acute lymphoblastic leukemia. Cancer 1986;58:963-68.
133. SC, Hopkins KP, Jones D, Crom D, Greenwald CA, Santana VM. Dental abnormalities in children treated for acute lymphoblastic leukemia. Leukemia 1997;11:792-96.
134. AL, Tarbell N, Valachovic RW, Gelber R, Schwenn M, Sallan S. Dentofacial development in long-term survivors of acute lymphoblastic leukemia: a comparison of three treatment modalities. Cancer 1990;66:2645-52.
135. AL, Waber DP, Sallan S, Tarbell NJ. The oral health of long-term survivors of acute lymphoblastic leukaemia: a comparison of three treatment modalities. Eur J Cancer B Oral Oncol 1995;31:250-52.
136. A, Chiarelli F, Di Marzio A, Impicciatore P, Marsico S, Angrilli F. Thyroid function in children treated for acute lymphoblastic leukemia. J Endocrinol Invest 1997;20:215-19.
137. LL, Nesbit ME, Sather HN, Meadows AT, Ortega JA, Hammond GD. Thyroid abnormalities in long-term survivors of childhood acute lymphoblastic leukemia. Pediatr Res 1985;19:266A.-
138. T, McCalla J, Berg S, et al. Subtle primary hypothyroidism in patients treated for acute lymphoblastic leukemia. Acta Endocrinol 1991;124:375-80.
139. CR, Miller JD, Guyda HJ, Esseltine DW, Chevalier LM, Freeman CR. Growth and development of long-term survivors of childhood acute lymphoblastic leukemia treated with and without prophylactic radiation of the central nervous system. Clin Invest Med 1985;8:307-14.
140. ML, Brecher ML, Glicksman AS, et al. Hypothalamic-pituitary function of children with acute lymphocytic leukemia after three forms of central nervous system prophylaxis: a retrospective study. Cancer 1986;57:1287-91.
141. EP, Leiper AD, Chessells JM. Thyroid function in children after treatment for acute lymphoblastic leukemia. Arch Dis Child 1988;64:631.-
142. MD, Shalet SM, Beardwell CG. Radiation and hypothalamic-pituitary function. Baillieres Clin Endocrinol Metab 1990;4:147-75.
143. F, Ohta K, Akanuma A, Sakata K. Dosimetry of radiation scattered to thyroid gland from prophylactic cranial irradiation for childhood leukemia. Pediatr Hematol Oncol 1994;11:47-53.
144. NJ, Tweeddale PM, Eden OB. Pulmonary function in childhood leukaemia survivors. Med Pediatr Oncol 1989;17:149-54.
145. K, Holm K, Olsen JH, Hertz H, Hesse B. Pulmonary function after treatment for acute lymphoblastic leukaemia in childhood. Br J Cancer 1998;78:21-27.
146. BL, Tanyer G, Poplack DG, et al. Transient acute hepatotoxicity of high-dose methotrexate therapy during childhood. NCI Monogr 1987;5:207-12.
147. F, Kinumaki H, Yokota S, Hayashi Y, Kobayashi M, Kamoshita S. Liver function studies in children with acute lymphocytic leukemia after cessation of therapy. Med Pediatr Oncol 1994;23:111-15.
148. AC, Buchanan GR, Zweiner RJ, Bowman WP, Winick NJ. Serum aminotransferase elevation during and following treatment of childhood acute lymphoblastic leukemia. J Clin Oncol 1997;15:1560-66.
149. PJ, Balistreri WF, Bove KE, Ballard ET, Passo MH. The relationship of hepatotoxic risk factors and liver histology in methotrexate therapy for juvenile rheumatoid arthritis. J Pediatr 1999;134:47-52.
150. HJ, Simone J, Aur RJA. Cyclophosphamide-induced hemorrhagic cystitis in children with leukemia. Cancer 1975;36:1572-76.
151. TJ, Benson RC. Cyclophosphamide-induced hemorrhagic cystitis: a review of 100 patients. Cancer 1988;61:451-57.
152. JM, Reed EC, Pippert GC, et al. Mesna compared with continuous bladder irrigation as uroprotection during high-dose chemotherapy and transplantation: a randomized trial. J Clin Oncol 1993;11:1306-10.
153. LB, Curtis RE, Glimelius B, et al. Bladder and kidney cancer following cyclophosphamide therapy for non-Hodgkin’s lymphoma. J Natl Cancer Inst 1995;87:524-30.
154. der Does-van den Berg A, de Vaan GAM, van Weerden JF, Hahlen K, van Weel-Sipman M, Veerman AJP. Late effects among long-term survivors of childhood acute leukemia in the Netherlands: a Dutch Childhood Leukemia Study Group report. Pediatr Res 1995;38:802-07.
1. MA, Ries LAG, Gurney JG, Ross JA. Leukemia. In: Ries LAG, Smith MA, Gurney JG, et al, eds. Cancer incidence and survival among children and adolescents: United States SEER Program 1975-1995, National Cancer Institute, SEER Program. Bethesda, MD: National Institutes of Health; 1999. NIH pub. no. 99-4649.
2. KC, Eshelman DA, Tomlinson GE, Buchanan GR. Programs for adult survivors of childhood cancer. J Clin Oncol 1998;16:2864-67.
3. DS. Transition to adult health care for adolescents and young adults with cancer. Cancer 1993;71:3411-14.
4. KC, Eshelman DA, Tomlinson GE, Buchanan GR, Foster BE. Grading of late effects in young adult survivors of childhood cancer followed in an ambulatory adult setting. Cancer 2000;88:1687-95.
5. H. The natural history of untreated acute leukemia. Ann NU Acad Sci 1954;60:322-58.
6. S, Diamond LK, Mercer RD, et al. Temporary remissions in acute leukemia in children produced by folic acid antagonist 4-aminopteroylglutamic acid (aminopterin). N Engl J Med 1948;238:787-93.
7. L, Gelber R, Cohen H, et al. Four-agent induction and intensive asparaginase therapy for treatment of childhood acute lymphoblastic leukemia. N Engl J Med 1986;315:657-63.
8. LL, Nesbit ME, Jr, Sather HN, Meadows AT, Ortega JA, Hammond GD. Factors associated with IQ scores in long-term survivors of childhood acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1984;6:115-21.
9. P, Waters B, Said J, Stevens M. Cognitive effects of cranial irradiation in leukaemia: a survey and meta-analysis. J Child Psychol Psychiatry 1988;29:839-52.
10. JM, Kornblith AB, Jones D, et al. A comparative study of the long term psychosocial functioning of childhood acute lymphoblastic leukemia survivors treated by intrathecal methotrexate with or without cranial radiation. Cancer 1998;82:208-18.
11. D, Reaman G, Bleyer W, et al. Successful prevention of central nervous (CNS) leukemia without cranial radiation in children with high risk acute lymphoblastic leukemia (ALL): a preliminary report. Proc Am Soc Clin Oncol 1989;8:828.-
12. W, Shuster J, Falletta J, et al. Clinical features and outcome in childhood T-cell leukemia-lymphoma according to stage of thymocyte differentiation: a Pediatric Onoclogy Group study. Blood 1988;72:1891-97.
13. CH, Behm FG, Singh B, et al. Heterogeneity of presenting features and their relation to treatment outcome in 120 children with T-cell acute lymphoblastic leukemia. Blood 1990;75:174-79.
14. M, Azuma E, Ido M, et al. Ten-year survey of the intellectual deficits in children with acute lymphoblastic leukemia receiving chemoimmunotherapy. Med Pediatr Oncol 1993;21:435-40.
15. DP, Urion DK, Tarbell NJ, Niemeyer C, Gelber R, Sallan SE. Late effects of central nervous system treatment of acute lymphoblastic leukemia in childhood are sex-dependent. Dev Med Child Neurol 1990;32:238-48.
16. AE, Aitken K, Eden OB. Computerized psychometry screening in long-term survivors of childhood acute lymphoblastic leukemia. Pediatr Hematol Oncol 1988;5:197-208.
17. H, Huk WJ, Ueberall MA, et al. CNS late effects after ALL therapy in childhood. Part I: Neuroradiological findings in long-term survivors of childhood ALL—an evaluation of the interferences between morphology and neuropsychological performance—the German Late Effects Working Group. Med Pediatr Oncol 1997;28:387-400.
18. JA, Kaleita TA, Noll RB, et al. CNS prophylaxis of childhood leukemia: what are the long-term neurological, neuropsychological, and behavioral effects? Neuropsychol Rev 1991;2:147-77.
19. JA, Waters BG, Cousens P, Stevens MM. Neuropsychological sequelae of central nervous system prophylaxis in survivors of childhood acute lymphoblastic leukemia. J Consult Clin Psychol 1989;57:251-56.
20. J, Horrocks J, Britton PG, Kernahan J. Attentional ability among survivors of leukaemia. Arch Dis Child 1999;80:318-23.
21. AS, Nesbit ME. Neuropsychologic (cognitive) disabilities in long-term survivors of childhood cancer. Pediatrician 1991;18:11-19.
22. RK, Kovnar E, Langston J, et al. Long-term survivors of leukemia treated in infancy: factors associated with neuropsychologic status. J Clin Oncol 1992;10:1095-102.
23. DP, Tarbell NJ, Fairclough D, et al. Cognitive sequelae of treatment in childhood acute lymphoblastic leukemia: cranial radiation requires an accomplice. J Clin Oncol 1995;13:2490-96.
24. CL, Varni JW, Katz ER. Cognitive functioning in long-term survivors of childhood leukemia: a prospective analysis. J Dev Behav Pediatr 1990;11:301-05.
25. M, Brouwers P, Valsecchi MG, Van Veldhuizen A, Huisman J. Association of 1800 cGy cranial irradiation with intellectual function in children with acute lymphoblastic leukaemia. Lancet 1994;344:224-27.
26. E, Anderson V, Godber T, Ekert H. Risk factors for intellectual and educational sequelae of cranial irradiation in childhood acute lymphoblastic leukaemia. Br J Cancer 1996;73:825-30.
27. V, Godber T, Smibert E, Ekert H. Neurobehavioural sequelae following cranial irradiation and chemotherapy in children: an analysis of risk factors. Pediatr Rehabil 1997;1:63-76.
28. Bleyer A. CNS chemoradiotherapy of childhood leukemia: the plot thickens but the ending bodes well. J Clin Oncol 1995;13:2480-82.
29. TA, Reaman GH, MacLean WE, Sather HN, Whitt JK. Neurodevelopmental outcome of infants with acute lymphoblastic leukemia: a Children’s Cancer Group report. Cancer 1999;85:1859-65.
30. RT, Madan-Swain A, Walco GA, et al. Cognitive and academic late effects among children previously treated for acute lymphocytic leukemia receiving chemotherapy as CNS prophylaxis. J Pediatr Psychol 1998;23:333-40.
31. L. Clinical neurological findings of children with acute lymphoblastic leukaemia at diagnosis and during treatment. Eur J Pediatr 1993;152:115-19.
32. HA, Schoemaker MM, Hofte M, et al. Fine motor and handwriting problems after treatment for childhood acute lymphoblastic leukemia. Med Pediatr Oncol 1996;27:551-55.
33. PG, Ciesielski KT, Hart BL, Benzel EC, Sanders JA. Evidence for cerebellar-frontal subsystem changes in children treated with intrathecal chemotherapy for leukemia. Arch Neurol 1998;55:1561-68.
34. R, Fears TR, Robison LL, et al. Educational attainment in long-term survivors of childhood acute lymphoblastic leukemia. JAMA 1994;272:1427-32.
35. P, Chen CH. Prevalence of obesity in children after therapy for acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1986;8:294-99.
36. I, Reilly JJ, Gibson BE, Donaldson MD. Patterns of obesity in boys and girls after treatment for acute lymphoblastic leukaemia. Arch Dis Child 1994;71:147-49.
37. MJ, Ochs JJ, Schriock EA, Carter M. A method of predicting adult height and obesity in long-term survivors of childhood acute lymphoblastic leukemia. J Clin Oncol 1992;10:128-33.
38. M, Didcock E, Davies HA, Ogilvy-Stuart AL, Wales JK, Shalet SM. High incidence of obesity in young adults after treatment of acute lymphoblastic leukemia in childhood. J Pediatr 1995;127:63-67.
39. Dongen-Melman JE, Hokken-Koelega AC, Hahlen K, De Groot A, Tromp CG, Egeler RM. Obesity after successful treatment of acute lymphoblastic leukemia in childhood. Pediatr Res 1995;38:86-90.
40. KK, Lanning M, Tapanainen P, Knip M. Long-term survivors of childhood cancer have an increased risk of manifesting the metabolic syndrome. J Clin Endocrinol Metab 1996;81:3051-55.
41. JT, Bell W, Webb DK, Gregory JW. Daily energy expenditure and physical activity in survivors of childhood malignancy. Pediatr Res 1998;43:607-13.
42. ME, Faragher EB, Jones PH, Woodcock A. Lung function and exercise capacity in survivors of childhood leukaemia. Med Pediatr Oncol 1995;24:222-30.
43. P, Gutjahr P, Stopfkuchen H. Physical performance in long-term survivors of acute leukaemia in childhood. Eur J Pediatr 1998;157:464-67.
44. MJ, Halton JM, Martin RF, Barr RD. Long-term gross motor performance following treatment for acute lymphoblastic leukemia. Med Pediatr Oncol 1998;3:86-90.
45. MJ, Halton JM, Barr RD. Limitation of ankle range of motion in survivors of acute lymphoblastic leukemia: a cross-sectional study. Med Pediatr Oncol 1999;32:279-82.
46. DS, Dietz WH, Srinivasan SR, Berenson GS. The relation of overweight to cardiovascular risk factors among children and adolescents: the Bogalusa Heart Study. Pediatrics 1999;103:1175-82.
47. M, Vanhala P, Kumpusalo E, Halonen P, Takala J. Relation between obesity from childhood to adulthood and the metabolic syndrome: population based study. BMJ 1998;317:319-21.
48. GS, Srinivasan SR, Bao W, et al. Association between multiple cardiovascular risk factors and atherosclerosis in children and young adults. N Engl J Med 1998;338:1650-56.
49. TC, Deedwania PC. The cardiovascular dysmetabolic syndrome. Am J Med 1998;105:77S-82S.
50. T, Bengtsson BA. Premature mortality due to cardiovascular disease in hypopituitism. Lancet 1990;336:285-88.
51. AS, Van’t Hoff W, Jones PJ, Clayton RN. The effect of hypopituitarism on life expectancy. J Clin Endocrinol Metab 1996;81:1169-72.
52. EM, Bulow B, Eskilsson J, Hagmar L. High incidence of cardiovascular disease and increased prevalence of cardiovascular risk factors in women with hypopituitarism not receiving growth hormone treatment: preliminary results. Growth Horm IGF Res 1999;9 (suppl):21-24.
53. MB. Effect of growth hormone on carbohydrate and lipid metabolism. Endocr Rev 1987;8:115-31.
54. FL, O’Neal D, Kamarudin N, Alford FP, Best JD. Growth hormone deficiency and cardiovascular risk. Baillieres Clin Endocrinol Metab 1998;12:199-216.
55. SA, Henderson A, Niththyananthan R, et al. The effects of short and long-term growth hormone replacement therapy in hypopituitary adults on lipid metabolism and carbohydrate tolerance. J Clin Endocrinol Metab 1995;80:356-63.
56. KA, Gray R, Anyaoku V, et al. Effects of four years’ treatment with biosynthetic human growth hormone (GH) on glucose homeostasis, insulin secretion and lipid metabolism in GH-deficient adults. Clin Endocrinol 1998;48:795-802.
57. D, Hew FL, Sikaris K, Ward G, Alford F, Best JD. Low density lipoprotein particle size in hypopituitary adults receiving conventional hormone replacement therapy. J Clin Endocrinol Metab 1996;81:2448-54.
58. Preventive Services Task Force. Guide to clinical preventive services. 2nd ed. Washington, DC: US Department of Health and Human Services; 1996.
59. PJ, Holen A, Glomstein A, et al. Long-term survival and quality of life in patients treated with a national ALL protocol 15-20 years earlier: IDM/HDM and late effects? Pediatr Hematol Oncol 1997;14:513-24.
60. AE. Posttraumatic distress in childhood cancer survivors and their parents. Med Pediatr Oncol 1998;1 (suppl):60-68.
61. LK, Chen E, Weiss R, et al. Comparison of psychologic outcome in adult survivors of childhood acute lymphoblastic leukemia versus sibling controls: a cooperative Children’s Cancer Group and National Institutes of Health study. J Clin Oncol 1997;15:547-56.
62. ML, Guo MD, Weiss R, et al. Smoking in adult survivors of childhood acute lymphoblastic leukemia. J Natl Cancer Inst 1998;90:219-25.
63. PB, Hough SF, Nel ED, van Riet FA, Beneke T, Wessels G. Bone mineral density in long-term survivors of childhood cancer. Int J Cancer Suppl 1998;11:44-7.
64. J, Hsieh K, Kalaitzoglou G, et al. Bone mineral density in young adult survivors of childhood cancer. J Pediatr Hematol Oncol 1998;20:241-45.
65. R, Brosnan P, Delpassand A, Zietz H, Klein MJ, Jaffe N. Osteopenia in young adult survivors of childhood cancer. Med Pediatr Oncol 1999;32:272-78.
66. V, Carlson ME, Roe TF, Ortega JA. Osteoporosis after cranial irradiation for acute lymphoblastic leukemia. J Pediatr 1990;117:238-44.
67. P, Komulainen J, Voutilainen R, et al. Reduced bone mineral density in long-term survivors of childhood acute lymphoblastic leukemia. J Pediatr Hematol Oncol 1998;20:234-40.
68. JT, Evans WD, Webb DK, Bell W, Gregory JW. Relative osteopenia after treatment for acute lymphoblastic leukemia. Pediatr Res 1999;45:544-51.
69. K, Holm K, Michaelsen KF, Hertz H, Muller J, Molgaard C. Bone mass after treatment for acute lymphoblastic leukemia in childhood. J Clin Oncol 1998;16:3752-60.
70. JJ, Kardos G, Roos JC, et al. Bone mineral density and markers of bone turnover in young adult survivors of childhood lymphoblastic leukaemia. Clin Endocrinol 1999;50:237-44.
71. BM, Rahim A, Mackie EM, Eden OB, Shalet SM. Clin Endocrinol 1998;48:777-783.
72. SA, Halton JM, Bradley C, Wu B, Barr RD. Bone and mineral abnormalities in childhood acute lymphoblastic leukemia: influence of disease, drugs and nutrition. Int J Cancer Suppl 1998;11:35-39.
73. B, Owens S, Okuyama T, Riggs S, Ferguson M, Litaker M. Effect of physical training and its cessation on percent fat and bone density of children with obesity. Obes Res 1999;7:208-14.
74. O, Kristinsson JO, Stefansson SO, Valdimarsson S, Sigurdsson G. Lean mass and physical activity as predictors of bone mineral density in 16-20-year old women. J Intern Med 1999;245:489-96.
75. I, van Croonenborg JJ, Kemper HC, Kostense PJ, Twisk JW. The effect of exercise training programs on bone mass: a meta-analysis of published controlled trials in pre- and postmenopausal women. Osteoporos Int 1999;9:1-12.
76. D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 1996;312:1254-59.
77. D, Sampietro-Colom L, Marshall D, Rico R, Granados A, Asua J. The effectiveness of bone density measurement and associated treatments for prevention of fractures: an international collaborative review. Int J Technol Assess Health Care 1998;14:237-54.
78. LL, Nesbit ME, Jr, Sather HN, Meadows AT, Ortega JA, Hammond GD. Height of children successfully treated for acute lymphoblastic leukemia: a report from the Late Effects Study Committee of Children’s Cancer Study Group. Med Pediatr Oncol 1985;13:14-21.
79. EA, Schell MJ, Carter M, Hustu O, Ochs JJ. Abnormal growth patterns and adult short stature in 115 long-term survivors of childhood leukemia. J Clin Oncol 1991;9:400-05.
80. M, Stanhope R, Chessells JM, Leiper AD. Impaired pubertal growth in acute lymphoblastic leukaemia. Arch Dis Child 1991;66:1403-07.
81. K, Dorffel W, Timme J, et al. Final height and puberty in 40 patients after antileukaemic treatment during childhood. Eur J Pediatr 1997;156:272-76.
82. P, Moell C, Cornu G, Malvaux P, Maes M. Subnormal growth during puberty in children treated for acute lymphoblastic leukemia. Pediatr Hematol Oncol 1992;9:217-22.
83. AC, van Doorn JW, Hahlen K, Stijnen T, de Muinck Keizer-Schrama SM, Drop SL. Long-term effects of treatment for acute lymphoblastic leukemia with and without cranial irradiation on growth and puberty: a comparative study. Pediatr Res 1993;33:577-82.
84. JA, Pollock BH, Jacaruso D, Morad A. Final attained height in patients successfully treated for childhood acute lymphoblastic leukemia. J Pediatr 1993;123:546-52.
85. AE, Adan L, Leverger G, Souberbielle JC, Schaison G, Brauner R. Growth hormone secretion, puberty and adult height after cranial irradiation with 18 Gy for leukaemia. Eur J Pediatr 1998;157:703-07.
86. J, Villaizan CJ, Garcia-Foncillas J, Azcona C, Salvador J, Sierrasesumaga L. Chemotherapy-induced growth hormone deficiency in children with cancer. Med Pediatr Oncol 1995;25:90-5.
87. J, Villaizan CJ, Garcia-Foncillas J, Salvador J, Sierrasesumaga L. Growth and growth hormone secretion in children with cancer treated with chemotherapy. J Pediatr 1997;131:105-12.
88. C, Mertens A, Walter A, et al. Final height after treatment for childhood acute lymphoblastic leukemia: comparison of no cranial irradiation with 1800 and 2400 centigrays of cranial irradiation. J Pediatr 1993;123:59-64.
89. A, Cacciari E, Rosito P, et al. Longitudinal growth and final height in long-term survivors of childhood leukaemia. Eur J Pediatr 1994;153:726-30.
90. TG, Byrne GC, Jones TW. Growth and growth hormone secretion after treatment for acute lymphoblastic leukemia in childhood 18-Gy versus 24-Gy cranial irradiation. J Pediatr Hematol Oncol 1995;17:167-71.
91. NH, Fisker S, Clausen N, Tuovinen V, Sindet-Pedersen S, Christiansen JS. Growth and endocrinological disorders up to 21 years after treatment for acute lymphoblastic leukemia in childhood. Med Pediatr Oncol 1998;30:351-56.
92. O’Halloran DJ, Tsatsoulis A, Whitehouse RW, Holmes SJ, Adams JE, Shalet SM. Increased bone density after recombinant human growth hormone (GH) therapy in adults with isolated GH deficiency. J Clin Endocrinol Metab 1993;76:1344-48.
93. F, Cuneo RC, Hesp R, Sonksen PH. The effects of treatment with recombinant human growth hormone on body composition and metabolism in adults with growth hormone deficiency. N Engl J Med 1989;321:1797-803.
94. P, Broman JE, Hetta J, et al. Quality of life in adults with growth hormone (GH) deficiency: response to treatment with recombinant human GH in a placebo-controlled 21-month trial. J Clin Endocrinol Metab 1995;80:3585-90.
95. SE, Colan SD, Gelber RD, Perez-Atayde AR, Sallan SE, Sanders SP. Late cardiac effects of doxorubicin therapy for acute lymphoblastic leukemia in childhood. N Engl J Med 1991;324:843-45.
96. SE, Lipsitz SR, Mone SM, et al. Female sex and drug dose as risk factors for late cardiotoxic effects of doxorubicin therapy for childhood cancer. N Engl J Med 1995;332:1738-43.
97. MA, Lipshultz SE. Epidemiology of anthracycline cardiotoxicity in children and adults. Semin Oncol 1998;25(suppl):72-85.
98. K, Levitt G, Bull C, Chessells J, Sullivan I. Anthracycline dose in childhood acute lymphoblastic leukemia: issues of early survival versus late cardiotoxicity. J Clin Oncol 1997;15:61-68.
99. K, Holm K, Lipsitz SR, et al. Relationship between cumulative anthracycline dose and late cardiotoxicity in childhood acute lymphoblastic leukemia. J Clin Oncol 1998;16:545-50.
100. LH. Ameliorating anthracycline cardiotoxicity in children with cancer: clinical trials with dexrazoxane. Semin Oncol 1998;25:86-92.
101. LJ, Graham T, Hurwitz R, et al. Guidelines for cardiac monitoring of children during and after anthracycline therapy: report of the Cardiology Committee of the Childrens Cancer Study Group. Pediatrics 1992;89:942-49.
102. for Disease Control and Prevention. Recommendations for prevention and control of hepatitis C virus (HCV) infection and HCV-related chronic disease. MMWR 1998;47:1-39.
103. M, Maggiore G, Silini E, Bono F, Vigano C. Hepatitis C virus infection in children treated for acute lymphoblastic leukemia. Blood 1994;84:2919-22.
104. SP, Ragusa R, Sciacca A, et al. Incidence and morbidity of infection by hepatitis C virus in children with acute lymphoblastic leukaemia. Eur J Pediatr 1994;153:271-75.
105. A, Testa M, Pontisso P, et al. Prevalence and natural history of hepatitis C infection in patients cured of childhood leukemia. Blood 1997;90:4628-33.
106. A, Alberti A. Hepatitis C virus serum markers and liver disease in children with leukemia. Leuk Lymphoma 1995;17:245-49.
107. S, Petris MG, Rossetti F, et al. Chronic hepatitis C virus infection after treatment for pediatric malignancy. Blood 1997;90:1315-20.
108. IM, Sanders J, Ruggiero F, Andrews T, Ungar D, Eyster ME. Chronic hepatitis C virus infections in leukemia survivors: prevalence, viral load, and severity of liver disease. Blood 1999;93:3672-77.
109. Dalton VM, Gelber RD, Li F, Donnelly MJ, Tarbell NJ, Sallan SE. Second malignancies in patients treated for childhood acute lymphoblastic leukemia. J Clin Oncol 1998;16:2848-53.
110. AW, Hancock ML, Pui CH, et al. Secondary brain tumors in children treated for acute lymphoblastic leukemia at St Jude Children’s Research Hospital. J Clin Oncol 1998;16:3761-67.
111. P, Straaten A, Gutjahr P. Secondary thyroid carcinoma after treatment for childhood cancer. Med Pediatr Oncol 1998;31:91-95.
112. Y, Leverger G, Carrere A, et al. Second thyroid neoplasms after prophylactic cranial irradiation for acute lymphoblastic leukemia. Am J Hematol 1998;59:91-94.
113. T, Ikuta H, Hibi S, Todo S. Second cutaneous neoplasms after acute lymphoblastic leukemia in childhood. Int J Hematol 1993;59:67-71.
114. J, Velasco-Benito JA, Pena-Penabad C, Armijo M. Basal cell carcioma in a girl after cobalt irradiation to the cranium for acute lymphoblastic leukemia: case report and literature review. Pediatr Dermatol 1996;13:54-57.
115. J, Philip P, Larsen SO, et al. Therapy-related myelodysplasia and acute myeloid leukemia: cytogenetic characteristics of 115 consecutive cases and risk in seven cohorts of patients treated intensively for malignant diseases in the Copenhagen series. Leukemia 1993;7:1975-86.
116. N, Shuster JJ, Bowman WP, et al. Intensive oral methotrexate protects against lymphoid marrow relapse in childhood B-precursor acute lymphoblastic leukemia. J Clin Oncol 1996;14:2803-11.
117. C, Hartmann JT, Kanz L, Bokemeyer C. Risk of secondary myeloid leukemia and myelodysplastic syndrome following standard-dose chemotherapy or high-dose chemotherapy with stem cell support in patients with potentially curable malignancies. J Cancer Res Clin Oncol 1998;124:207-14.
118. HM, Keating MJ. Therapy-related leukemia and myelodysplastic syndrome. Semin Oncol 1987;14:435-43.
119. MA, Rubinstein L, Anderson JR, et al. Secondary leukemia or myelodysplastic syndrome after treatment with epipodophyllotoxins. J Clin Oncol 1999;17:569-77.
120. MA, Rubinstein L, Cazenave L, et al. Report of the Cancer Therapy Evaluation Program monitoring plan for secondary acute myeloid leukemia following treatment with epipodophyllotoxins. J Natl Cancer Inst 1993;85:554-58.
121. CH, Relling MV, Rivera GK, et al. Epipodophyllotoxin-related acute myeloid leukemia: a study of 35 cases. Leukemia 1995;9:1990-96.
122. M, Akiyama Y, Koishi S, et al. Second malignancy following treatment of acute lymphoblastic leukemia in children. Int J Hematol 1998;67:397-401.
123. R, Clausen N, Siimes MA, et al. Reproduction following treatment for childhood leukemia: a population-based prospective cohort study of fertility and offspring. Med Pediatr Oncol 1991;19:459-66.
124. GA, Jenney ME. The reproductive system after childhood cancer. Br J Obstet Gynaecol 1998;105:946-53.
125. Wallace WH, Shalet SM, Tetlow LJ, Morris-Jones PH. Ovarian function following the treatment of childhood acute lymphoblastic leukaemia. Med Pediatr Oncol 1993;21:333-39.
126. MR, Robison LL, Nesbit ME, et al. Effects of radiation on ovarian function in long-term survivors of childhood acute lymphoblastic leukemia: a report from the Children’s Cancer Study Group. J Clin Oncol 1987;5:1759-65.
127. CA, Robison LL, Nesbit ME, et al. Effects of radiation on testicular function in long-term survivors of childhood acute lymphoblastic leukemia: a report from the Children’s Cancer Group. J Clin Oncol 1990;8:1981-87.
128. T, Kishi K, Imashuku S, et al. Testicular histology and function following long-term chemotherapy of acute leukemia in children and outcome of the patients who received testicular biopsy. Am J Pediatr Hematol Oncol 1986;8:288-93.
129. WH, Shalet SM, Lendon M, Morris-Jones PH. Male fertility in long-term survivors of childhood acute lymphoblastic leukaemia. Int J Androl 1991;14:312-19.
130. LB, Nicholson HS, Brasseux C, et al. Birth defects in offspring of adult survivors of childhood acute lymphoblastic leukemia: a Children’s Cancer Group/National Institutes of Health Report. Cancer 1996;78:169-76.
131. DL, Smith LE, Turner SJ, Gelber RD, Sallan SE. Ophthalmic evaluation of survivors of acute lymphoblastic leukemia. Ophthalmology 1988;95:151-55.
132. RG, Jr, Chauvenet AR, Smith TJ, Schwartz AC. Ophthalmic evaluation of long-term survivors of childhood acute lymphoblastic leukemia. Cancer 1986;58:963-68.
133. SC, Hopkins KP, Jones D, Crom D, Greenwald CA, Santana VM. Dental abnormalities in children treated for acute lymphoblastic leukemia. Leukemia 1997;11:792-96.
134. AL, Tarbell N, Valachovic RW, Gelber R, Schwenn M, Sallan S. Dentofacial development in long-term survivors of acute lymphoblastic leukemia: a comparison of three treatment modalities. Cancer 1990;66:2645-52.
135. AL, Waber DP, Sallan S, Tarbell NJ. The oral health of long-term survivors of acute lymphoblastic leukaemia: a comparison of three treatment modalities. Eur J Cancer B Oral Oncol 1995;31:250-52.
136. A, Chiarelli F, Di Marzio A, Impicciatore P, Marsico S, Angrilli F. Thyroid function in children treated for acute lymphoblastic leukemia. J Endocrinol Invest 1997;20:215-19.
137. LL, Nesbit ME, Sather HN, Meadows AT, Ortega JA, Hammond GD. Thyroid abnormalities in long-term survivors of childhood acute lymphoblastic leukemia. Pediatr Res 1985;19:266A.-
138. T, McCalla J, Berg S, et al. Subtle primary hypothyroidism in patients treated for acute lymphoblastic leukemia. Acta Endocrinol 1991;124:375-80.
139. CR, Miller JD, Guyda HJ, Esseltine DW, Chevalier LM, Freeman CR. Growth and development of long-term survivors of childhood acute lymphoblastic leukemia treated with and without prophylactic radiation of the central nervous system. Clin Invest Med 1985;8:307-14.
140. ML, Brecher ML, Glicksman AS, et al. Hypothalamic-pituitary function of children with acute lymphocytic leukemia after three forms of central nervous system prophylaxis: a retrospective study. Cancer 1986;57:1287-91.
141. EP, Leiper AD, Chessells JM. Thyroid function in children after treatment for acute lymphoblastic leukemia. Arch Dis Child 1988;64:631.-
142. MD, Shalet SM, Beardwell CG. Radiation and hypothalamic-pituitary function. Baillieres Clin Endocrinol Metab 1990;4:147-75.
143. F, Ohta K, Akanuma A, Sakata K. Dosimetry of radiation scattered to thyroid gland from prophylactic cranial irradiation for childhood leukemia. Pediatr Hematol Oncol 1994;11:47-53.
144. NJ, Tweeddale PM, Eden OB. Pulmonary function in childhood leukaemia survivors. Med Pediatr Oncol 1989;17:149-54.
145. K, Holm K, Olsen JH, Hertz H, Hesse B. Pulmonary function after treatment for acute lymphoblastic leukaemia in childhood. Br J Cancer 1998;78:21-27.
146. BL, Tanyer G, Poplack DG, et al. Transient acute hepatotoxicity of high-dose methotrexate therapy during childhood. NCI Monogr 1987;5:207-12.
147. F, Kinumaki H, Yokota S, Hayashi Y, Kobayashi M, Kamoshita S. Liver function studies in children with acute lymphocytic leukemia after cessation of therapy. Med Pediatr Oncol 1994;23:111-15.
148. AC, Buchanan GR, Zweiner RJ, Bowman WP, Winick NJ. Serum aminotransferase elevation during and following treatment of childhood acute lymphoblastic leukemia. J Clin Oncol 1997;15:1560-66.
149. PJ, Balistreri WF, Bove KE, Ballard ET, Passo MH. The relationship of hepatotoxic risk factors and liver histology in methotrexate therapy for juvenile rheumatoid arthritis. J Pediatr 1999;134:47-52.
150. HJ, Simone J, Aur RJA. Cyclophosphamide-induced hemorrhagic cystitis in children with leukemia. Cancer 1975;36:1572-76.
151. TJ, Benson RC. Cyclophosphamide-induced hemorrhagic cystitis: a review of 100 patients. Cancer 1988;61:451-57.
152. JM, Reed EC, Pippert GC, et al. Mesna compared with continuous bladder irrigation as uroprotection during high-dose chemotherapy and transplantation: a randomized trial. J Clin Oncol 1993;11:1306-10.
153. LB, Curtis RE, Glimelius B, et al. Bladder and kidney cancer following cyclophosphamide therapy for non-Hodgkin’s lymphoma. J Natl Cancer Inst 1995;87:524-30.
154. der Does-van den Berg A, de Vaan GAM, van Weerden JF, Hahlen K, van Weel-Sipman M, Veerman AJP. Late effects among long-term survivors of childhood acute leukemia in the Netherlands: a Dutch Childhood Leukemia Study Group report. Pediatr Res 1995;38:802-07.
The Educational Value of Consumer-Targeted Prescription Drug Print Advertising
METHODS: We collected advertisements appearing in 18 popular magazines from 1989 through 1998. Two coders independently evaluated 320 advertisements encompassing 101 drug brands to determine if information appeared about specific aspects of the medical conditions for which the drug was promoted and about the treatment (mean k reliability=0.91). We employed basic descriptive statistics using the advertisement as the unit of analysis and cross-tabulations using the brand as the unit of analysis.
RESULTS: Virtually all the advertisements gave the name of the condition treated by the promoted drug, and a majority provided information about the symptoms of that condition. However, few reported details about the condition’s precursors or its prevalence; attempts to clarify misconceptions about the condition were also rare. The advertisements seldom provided information about the drug’s mechanism of action, its success rate, treatment duration, alternative treatments, and behavioral changes that could enhance the health of affected patients.
CONCLUSIONS: Informative advertisements were identified, but most of the promotions provided only a minimal amount of information. Strategies for improving the educational value of DTC advertisements are considered.
The appropriateness of direct-to-consumer (DTC) prescription drug advertising has been a topic of heated discussion in recent years.1,2 This debate is likely to intensify as the drug industry’s advertising expenditures continue to increase3,4 and the marketing channels employed expand to include television, which may not lend itself to highly informative messages, and the Internet, which is difficult to regulate.5 Ironically, DTC advertising proponents and opponents have most often based their positions on assumptions about the informational value of such promotions.6 However, empirical data on this issue are sparse.
The opponents of DTC advertising argue that the objectives of drug promotion and health education are inherently at odds.7 It has been noted that the aim of DTC advertising is to increase sales, not optimize health care.8,9 It has been presumed that the profit motive discourages the provision of complete and accurate information about pharmaceuticals.10 Reports have documented violations of the public trust11 and the dissemination of misleading information to health care professionals.12,13 If even highly trained physicians are prone to influence by commercial rather than scientific sources of pharmaceutical information,14 how can consumers protect themselves from misleading claims? To make matters worse, the public may be ill equipped to fully understand such advertising, no matter how accurate its content might be.7,15
Defenders of DTC advertising counter these arguments by suggesting that selling and educating are not necessarily incompatible goals16 and that it is in the industry’s self-interest to be truthful with consumers.17 Although instances of misleading DTC advertising can be identified, such cases are said to be rare.5 Proponents of DTC advertising typically attribute greater sophistication to the public than critics do,16 rejecting the notion that consumers cannot comprehend such advertising.18 In any regard, misunderstandings can be corrected by the physician, in whom the power to prescribe ultimately resides.1 Although many physicians oppose DTC advertising, a majority feel that such advertisements might help patients become better informed about drugs,19 conjecture that has empirical support.20
We have attempted to contribute to the debate on the value of DTC advertising by describing its content. In a recent JFP article,21 we documented the types of appeals and inducements used in these promotions. In this article we describe the scope of the educational efforts represented in advertisements appearing in consumer publications from 1989 through 1998. After identifying basic issues that most DTC advertisements should address to have educational merit, we examined the extent to which such promotions provide consumers with information about medical conditions, including their symptoms, precursors, and prevalence. It has been argued that DTC advertising increases the public’s awareness of these aspects of diseases and other health conditions. This kind of patient education is said to help people to take responsibility for their health by teaching them how to recognize disease and motivating them to seek medical attention for conditions that might otherwise be left undiagnosed and untreated.1,22-24
We then examined the extent to which DTC advertising incorporates information about treatments for those conditions. It has been said that consumer-targeted prescription drug advertising provides the public with valuable information about available treatments that leads to a better match between patients’ needs and available drugs.25 Specifically, we examined the extent to which DTC advertising offers details about promoted drugs’ mechanisms of action, the duration and success of advertised treatments, behavioral modifications that can improve the patient’s health independently or in concert with drug treatment, and competing treatments.
Methods
Sampling Procedure
Major popular magazines were ranked in terms of the average number of advertising pages sold from 1989 through 1996 (the last year for which data were available) and then grouped into an established set of categories.26 The highest-ranked publication within each of 13 categories was selected for inclusion in our sample. To these magazines we added 5 publications with audiences composed of narrower segments of our population defined by ethnicity, age, and sexual orientation. Thus, we garnered our sample of advertisements from 18 diverse publications Table 1. All the advertisements promoting a specific, named prescription drug that appeared in these magazines from January 1989 through December 1998 were photocopied. Advertisements for the same brand often differed in nonsubstantive ways and were thus coded as a single case based on detailed rules that are available from the authors. After aggregating essentially identical advertisements, 320 remained for analysis covering 101 distinct brands.
Medical Conditions Classification
Each advertisement was grouped in one of 14 medical condition categories based on the indication of the marketed drug. The numbers of advertisements and brands within each of these categories were as follows: allergies (46 advertisements, 8 brands); cancer (2, 2); cardiovascular disease (36, 10); dermatologic conditions (37, 12); diabetes (9, 4); gastrointestinal/nutritional disorders (17, 7); human immunodeficiency virus (HIV)/acquired immune deficiency syndrome (AIDS; 33, 11); infections, non-HIV (16, 6); musculoskeletal disorders (17, 7); obstetric/gynecologic conditions (45, 10); psychiatric/neurologic disorders (17, 7); respiratory disorders (3, 3); tobacco addiction (23, 6); and urologic conditions (19, 8).
Advertisement Coding
Two coders independently analyzed each advertisement. Our goal for this classification effort was to determine whether each advertisement contained 5 kinds of information pertaining to the focal medical condition and 6 types of drug treatment information Table 2.
Statistical Analyses
We examined each of the 11 education codes listed in Table 2 and computed 3 more general measures. We created a condition information index by summing the number of medical condition codes present within each advertisement (theoretical range=0-5). Likewise, a treatment information index was computed by summing the number of treatment codes present within each advertisement (theoretical range=0-6). A combined education index was created by adding the condition and treatment information variables (theoretical range=0-11).
The reliability of coding was assessed with the k statistic, a measure of inter-rater agreement with a range of 0 to 1.0.27 The mean k value for the variables reported was 0.91 (range=0.88-1.0). Landis and Koch28 suggest that k values in this range can be considered “almost perfect.”
We employed 2 units of analyses. For our descriptive analyses, the advertisement served as the unit of analysis (N=320). We relied on the brand (N=101) as the unit of analysis when computing cross-tabulations and one-way analyses of variance. Using the advertisement as the unit for inferential statistical analyses would have violated the assumption of independence of observations,29 because most advertisements for a particular brand represent alterations of earlier promotions for that brand. The advertisements for each brand were aggregated into a single case by weighting; each advertisement was assigned a weight of 1/n, with n representing the number of advertisements for that brand. When a significant chi-square value was obtained for a cross-tabulation, the source of significance within the table was determined by examining the cell-adjusted standardized residuals using a P <.05 criterion.30
Results
The Figure shows the percentage of advertisements for which the 5 medical condition information codes and 6 treatment information codes were present. These results have been ordered in the Figure on the basis of frequency of occurrence.
Information About Medical Conditions
The medical condition name was provided in virtually all of the advertisements (95%). This information was excluded in 1 “teaser ad” that simply showed a bottle of tablets for a forthcoming treatment, in 7 advertisements alerting patients of a new formulation of a drug already prescribed for them, and in 9 advertisements notifying patients of an alternative delivery system for an established treatment. Sixty percent of the advertisements described at least one symptom of the condition treated by the promoted drug or explicitly stated that the condition could be a “silent disease”. Among those advertisements that omitted symptom information, 54% were targeted to individuals who had already been diagnosed with the condition and were presumably aware of its symptoms. The remaining advertisements tended to be for conditions with well-known symptoms (eg, pregnancy, impotence, and tobacco addiction). Information about a precursor to the condition such as a cause or risk factor was provided in 27% of the advertisements. A total of 6% of advertisements provided precursor information by means of a set of diagnostic questions, such as a quiz that readers could use to assess their risk of being or becoming affected by the condition (data not diagramed). Information about condition prevalence (12% of advertisements) and clarifications about a condition-related misconception (9%) was rarely provided.
Information About Treatments
The Figure also displays the percentage of advertisements reporting each of the 6 types of treatment information. Most often present was information about a drug’s mechanism of action, found in 36% of the advertisements. An acknowledgement of the existence of one or more competing treatments was offered in 29% of the advertisements. Supportive behaviors such as changes in diet, physical activity, and sleeping patterns were reported in 24% of the advertisements. In our judgment, relevant supportive behaviors are often recommended by physicians for most of the conditions addressed in these advertisements. Readers were given information about time to onset of action in 20% of the advertisements and typical required treatment duration in 11%. A success rate estimate was rarely reported (9%).
Three of these treatment information codes were significantly associated with the medical conditions variable (all P values=.02). Supportive behaviors that could be used alone or to enhance the effectiveness of therapy were more likely to be found in diabetes and tobacco cessation treatments (75% and 67% of brands, respectively). Time to onset of action information was especially likely to be offered in advertisements for dermatological and urological problems (58% and 50%, respectively). Information about treatment duration was most often found in advertisements for infection and dermatological treatments (50% and 33%, respectively).
Indexes
For summative purposes we examined the condition information, treatment information, and overall education indexes for each of the 14 medical condition categories, using the drug brand as the unit of analysis. The average brand provided only 2.1 of the 5 types of information about the condition for which the drug was marketed. Only 1.2 of the 6 types of treatment information we coded was provided in the average brand-weighted advertisement. Summing across the condition and treatment information variables, the average number of educational codes present for these 101 brands was only 3.2 out of a possible score of 11 (range=1.0-7.3).
Next we categorized the brands under the medical condition category for which each was promoted Table 3. Despite weak statistical power, a significant difference across the 14 medical conditions was found for 2 of the 3 indexes, the condition information index (F [13,87]=3.05, P=.001) and the combined education index (F [13,87]=2.39, P=.008). On the basis of post hoc comparisons, both effects could be attributed to the greater provision of information in promotions for urologic brands than in HIV/AIDS brand advertisements.
Information Formats
Information about conditions and treatments was typically provided in narrative form. Tabular data or charts (eg, bar, pie, or column charts) were rarely used to provide information about the drug (2% of advertisements) or the medical condition (<1%). Diagrams and pictures were also rarely used to provide information about the effectiveness of the treatment (2% of advertisements) or the nature of the condition (7%).
Other Sources of Information
One view of DTC advertising is that it simply provides an introduction to a treatment and that more detailed patient education is best provided in other ways, such as with brochures and videotaped presentations or through the Internet. An explicit offer of printed or audiovisual information was provided in 35% of the advertisements, and a toll-free information telephone number was found in 73%. Increasingly, the Internet is being used as a means through which interested readers can obtain more information about advertised products. In 1996, 1997, and 1998, the percentages of advertisements that provided a Web site address were 14%, 33%, and 57%, respectively. Most telephone numbers were numeric (55%), but some of these and all Web addresses referenced a benefit of use (eg, 1-800-5-SHUT-EYE for a treatment of insomnia), the treated body part (eg, www.kidsears.com for a treatment of ear infections), the medical condition treated (eg, 1-800-66-ANGINA), the brand name of the drug (eg, www.atrovent.com), or the drug company’s name (eg, 1-800-GLAXO-RX).
Discussion
A time may come when DTC advertising is recommended for its educational value, but that day is not yet at hand. The advertisements that have appeared during the past decade have been superficial in their coverage of medical conditions and their treatments. Although most provide the name of the treated medical condition and its associated symptoms, the large majority do not inform potential patients about such basic matters as the risk factors for the condition or its prevalence. Likewise, these advertisements seldom educate patients about the mechanism of action by which the drug treats a particular condition, its success in doing so, alternative treatments, and behavioral changes that could augment or supplant treatment. To their credit, most of these promotions do offer the reader alternative ways of learning more about their condition and the advertised drug.
The medical community should exert pressure on the drug industry to incorporate more information about conditions and treatments in its advertising. It is in the industry’s best interest to do so. Advertising that incorporates quality health and drug information will have greater credibility, and deservedly so. Also, consumers are less likely to be influenced by messages that appear to be more promotional than informational.31 Thus, providing complete and accurate information is the right thing to do and may even enhance the effectiveness of DTC advertising. If such information is not provided voluntarily by the industry in future advertising, the medical establishment should lobby for regulation.
These results also highlight an important opportunity for professional organizations to contribute to consumer education on prescription drugs. According to the National Institute for Health Care Management, the most successfully promoted drugs fall into 5 categories: antidepressants, cholesterol-lowering agents, gastric acid reducers, oral antihistamines, and antihypertensives.32 Organizations such as the American Academy of Family Physicians and the American College of Physicians already produce patient informational materials, but these efforts could be intensified so physicians would have a ready source of “counter-detailing.” Patients requesting drugs for which the indications are questionable could be given a handout and told, “This is what my professional society has to say about _________. This information is produced by the best experts in the field and provides a more balanced view than what you will find in profit-motivated advertisements. Look it over, and let’s talk about this at our next visit.”
If the industry were to be prodded into taking a more educational stance in its consumer-targeted prescription drug advertising, the impact of this shift on physician-patient interactions would need to be investigated. A more educated patient may take less time to treat and counsel, might show greater adherence to treatment regimens, and could assume greater responsibility for his or her health. On the other hand, educational promotions may lead to requests from more determined patients for drugs that are not medically indicated, requiring time-consuming re-education by their physicians.
Limitations
Our study has limitations. We have evaluated the educational qualities of these advertisements by imposing a common set of standards for each promotion. A stronger approach would have been to convene a panel of experts for each of the many conditions treated by the promoted drugs to identify the specific details that consumers need to know about the condition and its treatment. Resource constraints prevented us from developing expert-based educational standards for each condition or disease. However, it is reasonable to expect the drug industry to provide to consumers basic information about the treatments being promoted and the conditions these drugs address. We acknowledge that particular content analytic codes may have been irrelevant to certain conditions. For instance, precursor information does not need to be provided when the causes of the condition are obvious (eg, advertisements for contraceptive drugs); mention of supportive lifestyle changes should not be demanded when helpful behavioral changes do not exist (eg, advertisements for hair loss treatments); and references to the presence of competing treatments should not be expected in the rare event that such treatments are nonexistent. Such exceptions aside, we believe that the 11 codes assessed are both fundamental and relevant to the vast majority of the conditions covered by these advertisements.
Also, with this preliminary investigation we sought only to assess the extent to which the industry is making an effort to provide information about medical conditions and treatments for those conditions. We did not examine the educational quality of these efforts, including the completeness and accuracy of information provided about conditions and treatments. Such assessments will require input from medical and pharmaceutical experts selected for their specialized knowledge.
Conclusions
We acknowledge that instances of informative advertising can readily be found in our sample of advertisements. Thus, although DTC advertising in general is not serving the information needs of consumers, there are companies and individuals within the industry who are motivated to treat their advertising as vehicles for effective health promotion through quality education. Billions of dollars will be spent in the next few years on consumer-targeted prescription drug promotions.4 Consumers will be very receptive to those advertisements that address their personal health needs and concerns33; many will talk to their primary care physicians as a result of these promotions.34 The drug industry thus has a tremendous opportunity to silence its critics and improve the public’s health by providing objective medical and drug information.35
Acknowledgments
We wish to acknowledge the contributions made by Ronald Emerick, Robert LaGreca, Love Lord, and Sarah Shaw in the collection and coding of advertisements and by 2 anonymous peer reviewers for their thoughtful evaluations.
Related Resources:
- U.S. Food & Drug Administration, Center for Drug Evaluation and Research, Division of Drug Marketing, Advertising and Communications www.fda.gov/cder/ddmac/
- Public Citizen Health Research Group “Drugs” page, listing numerous government documents that describe recommendations for drug use, sales and marketing. http://www.citizen.org/hrg/publications/drugs.htm
- AARP Executive Summary: Are Consumers Well Informed About Prescription Drugs? The Impact of Printed Direct-to-Consumer Advertising http://research.aarp.org/health/2000_04_advertising_1.html
- drkoop.com-Perspective: ‘Direct-to-Consumer’ Advertising of Prescription Drugs is More Harmful than Helpful to Consumers www.drkoop.com/news/focus/november/dtc_no.html
1. Holmer AF. Direct-to-consumer prescription drug advertising builds bridges between patients and physicians. JAMA 1999;281:380-82.
2. Hollon MF. Direct-to-consumer marketing of prescription drugs: creating consumer demand. JAMA 1999;281:382-84.
3. Barrett A. Are drug ads a cure-all? Business Week March 30, 1998;59-60.
4. Growth seen in ads for direct-to-consumer drugs. AMA News April 27, 1998;16.-
5. Maguire P. How direct-to-consumer advertising is putting the squeeze on physicians. ACP-ASIM Observer 1999;19:1, 24-25.
6. Pierpaoli PG. ASHP’s position on direct-to-consumer advertising of prescription drug products. Am J Hosp Pharm 1986;43:1763-65.
7. Cohen EP. Direct-to-the-public advertisement of prescription drugs. N Engl J Med 1988;318:373.-
8. Committee on Drugs. Prescription drug advertising direct to the consumer. Pediatrics 1991;88:174-75.
9. Bradley LR, Zito JM. Direct-to-consumer prescription drug advertising. Med Care 1997;35:86-92.
10. T’hoen E. Direct-to-consumer advertising: For better profits or for better health? Am J Health-Syst Pharm 1998;55:594-97.
11. Wall Street Journal August 22, 1997; 168:B5. Schering-Plough is told to halt Claritin TV ads.
12. Wilkes MS, Doblin BH, Shapiro MF. Pharmaceutical advertisements in leading medical journals: experts’ assessments. Ann Intern Med 1992;116:912-19.
13. Stryer D, Bero LA. Characteristics of materials distributed by drug companies: an evaluation of appropriateness. JGIM 1996;11:575-83.
14. Avorn J, Chen M, Hartley R. Scientific versus commercial sources of influence on the prescribing behavior of physicians. Am J Med 1982;73:4-8.
15. Schommer JC, Doucette WR, Mehta BH. Rote learning after exposure to a direct-to-consumer television advertisement for a prescription drug. Clin Ther 1998;20:617-32.
16. Ingram RA. Some comments on direct-to-consumer advertising. J Pharm Marketing Manage 1992;7:67-74.
17. Wind Y. Pharmaceutical advertising: a business school perspective. Arch Fam Med 1994;3:321-23.
18. Rubin PH. What the FDA doesn’t want you to know. Am Enterprise 1991;2:18-20.
19. Lipsky MS, Taylor CA. The opinions and experiences of family physicians regarding direct-to-consumer advertising. J Fam Pract 1997;45:495-99.
20. Peyrot M, Alperstein NM, Van Doren D, Poli LG. Direct-to-consumer ads can influence behavior: advertising increases consumer knowledge and prescription drug requests. Marketing Health Services 1998;18:26-32.
21. Bell RA, Kravitz RL, Wilkes MS. Direct-to-consumer prescription drug advertising, 1989-1998: a content analysis of conditions, targets, inducements and appeals. J Fam Pract 2000;49:329-35.
22. Keith A. The benefits of pharmaceutical promotion: an economic and health perspective. J Pharm Marketing Manage 1992;7:121-33.
23. Whyte J. Direct consumer advertising of prescription drugs. JAMA 1993;268:146,150.-
24. Reynolds WJ. Trends in advertising pharmaceuticals: a publisher’s perspective. J Pharm Marketing Manage 1992;7:5-22.
25. Masson A, Rubin PH. Matching prescription drugs and consumers. N Engl J Med 1985;313:513-15.
26. Katz WA. Magazines for libraries. New Providence, NJ: R.R. Bowker; 1997.
27. Cohen J. A coefficient of agreement for nominal scales: educational and psychological measurement 1960;20:37-46.
28. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159-74.
29. Hays WL. Statistics. New York, NY: Holt, Rinehart, and Winson, 1981.
30. Everitt BS. The analysis of contingency tables. London, England: Chapman and Hall; 1980.
31. Petty RE, Cacioppo JT. Communication and persuasion. New York, NY: Springer-Verlag; 1986.
32. National Institute for Health Care Management, Research and Education Foundation. Factors affecting the growth of prescription drug expenditures. July 9, 1999. Available at www.nihcm.org.
33. Bell RA, Kravitz RL, Wilkes M. Direct-to-consumer prescription drug advertising and the public. JGIM 1999;13:651-57.
34. Bell RA, Wilkes MS, Kravitz RL. Advertising-induced prescription drug requests: patients’ anticipated reactions to a physician who refuses. J Fam Pract 1999;48:446-52.
35. Wilkes MS, Bell RA, Kravitz RL. Consumer-directed prescription drug advertising: trends, impact, and implications. Health Aff 2000;19:110-28.
METHODS: We collected advertisements appearing in 18 popular magazines from 1989 through 1998. Two coders independently evaluated 320 advertisements encompassing 101 drug brands to determine if information appeared about specific aspects of the medical conditions for which the drug was promoted and about the treatment (mean k reliability=0.91). We employed basic descriptive statistics using the advertisement as the unit of analysis and cross-tabulations using the brand as the unit of analysis.
RESULTS: Virtually all the advertisements gave the name of the condition treated by the promoted drug, and a majority provided information about the symptoms of that condition. However, few reported details about the condition’s precursors or its prevalence; attempts to clarify misconceptions about the condition were also rare. The advertisements seldom provided information about the drug’s mechanism of action, its success rate, treatment duration, alternative treatments, and behavioral changes that could enhance the health of affected patients.
CONCLUSIONS: Informative advertisements were identified, but most of the promotions provided only a minimal amount of information. Strategies for improving the educational value of DTC advertisements are considered.
The appropriateness of direct-to-consumer (DTC) prescription drug advertising has been a topic of heated discussion in recent years.1,2 This debate is likely to intensify as the drug industry’s advertising expenditures continue to increase3,4 and the marketing channels employed expand to include television, which may not lend itself to highly informative messages, and the Internet, which is difficult to regulate.5 Ironically, DTC advertising proponents and opponents have most often based their positions on assumptions about the informational value of such promotions.6 However, empirical data on this issue are sparse.
The opponents of DTC advertising argue that the objectives of drug promotion and health education are inherently at odds.7 It has been noted that the aim of DTC advertising is to increase sales, not optimize health care.8,9 It has been presumed that the profit motive discourages the provision of complete and accurate information about pharmaceuticals.10 Reports have documented violations of the public trust11 and the dissemination of misleading information to health care professionals.12,13 If even highly trained physicians are prone to influence by commercial rather than scientific sources of pharmaceutical information,14 how can consumers protect themselves from misleading claims? To make matters worse, the public may be ill equipped to fully understand such advertising, no matter how accurate its content might be.7,15
Defenders of DTC advertising counter these arguments by suggesting that selling and educating are not necessarily incompatible goals16 and that it is in the industry’s self-interest to be truthful with consumers.17 Although instances of misleading DTC advertising can be identified, such cases are said to be rare.5 Proponents of DTC advertising typically attribute greater sophistication to the public than critics do,16 rejecting the notion that consumers cannot comprehend such advertising.18 In any regard, misunderstandings can be corrected by the physician, in whom the power to prescribe ultimately resides.1 Although many physicians oppose DTC advertising, a majority feel that such advertisements might help patients become better informed about drugs,19 conjecture that has empirical support.20
We have attempted to contribute to the debate on the value of DTC advertising by describing its content. In a recent JFP article,21 we documented the types of appeals and inducements used in these promotions. In this article we describe the scope of the educational efforts represented in advertisements appearing in consumer publications from 1989 through 1998. After identifying basic issues that most DTC advertisements should address to have educational merit, we examined the extent to which such promotions provide consumers with information about medical conditions, including their symptoms, precursors, and prevalence. It has been argued that DTC advertising increases the public’s awareness of these aspects of diseases and other health conditions. This kind of patient education is said to help people to take responsibility for their health by teaching them how to recognize disease and motivating them to seek medical attention for conditions that might otherwise be left undiagnosed and untreated.1,22-24
We then examined the extent to which DTC advertising incorporates information about treatments for those conditions. It has been said that consumer-targeted prescription drug advertising provides the public with valuable information about available treatments that leads to a better match between patients’ needs and available drugs.25 Specifically, we examined the extent to which DTC advertising offers details about promoted drugs’ mechanisms of action, the duration and success of advertised treatments, behavioral modifications that can improve the patient’s health independently or in concert with drug treatment, and competing treatments.
Methods
Sampling Procedure
Major popular magazines were ranked in terms of the average number of advertising pages sold from 1989 through 1996 (the last year for which data were available) and then grouped into an established set of categories.26 The highest-ranked publication within each of 13 categories was selected for inclusion in our sample. To these magazines we added 5 publications with audiences composed of narrower segments of our population defined by ethnicity, age, and sexual orientation. Thus, we garnered our sample of advertisements from 18 diverse publications Table 1. All the advertisements promoting a specific, named prescription drug that appeared in these magazines from January 1989 through December 1998 were photocopied. Advertisements for the same brand often differed in nonsubstantive ways and were thus coded as a single case based on detailed rules that are available from the authors. After aggregating essentially identical advertisements, 320 remained for analysis covering 101 distinct brands.
Medical Conditions Classification
Each advertisement was grouped in one of 14 medical condition categories based on the indication of the marketed drug. The numbers of advertisements and brands within each of these categories were as follows: allergies (46 advertisements, 8 brands); cancer (2, 2); cardiovascular disease (36, 10); dermatologic conditions (37, 12); diabetes (9, 4); gastrointestinal/nutritional disorders (17, 7); human immunodeficiency virus (HIV)/acquired immune deficiency syndrome (AIDS; 33, 11); infections, non-HIV (16, 6); musculoskeletal disorders (17, 7); obstetric/gynecologic conditions (45, 10); psychiatric/neurologic disorders (17, 7); respiratory disorders (3, 3); tobacco addiction (23, 6); and urologic conditions (19, 8).
Advertisement Coding
Two coders independently analyzed each advertisement. Our goal for this classification effort was to determine whether each advertisement contained 5 kinds of information pertaining to the focal medical condition and 6 types of drug treatment information Table 2.
Statistical Analyses
We examined each of the 11 education codes listed in Table 2 and computed 3 more general measures. We created a condition information index by summing the number of medical condition codes present within each advertisement (theoretical range=0-5). Likewise, a treatment information index was computed by summing the number of treatment codes present within each advertisement (theoretical range=0-6). A combined education index was created by adding the condition and treatment information variables (theoretical range=0-11).
The reliability of coding was assessed with the k statistic, a measure of inter-rater agreement with a range of 0 to 1.0.27 The mean k value for the variables reported was 0.91 (range=0.88-1.0). Landis and Koch28 suggest that k values in this range can be considered “almost perfect.”
We employed 2 units of analyses. For our descriptive analyses, the advertisement served as the unit of analysis (N=320). We relied on the brand (N=101) as the unit of analysis when computing cross-tabulations and one-way analyses of variance. Using the advertisement as the unit for inferential statistical analyses would have violated the assumption of independence of observations,29 because most advertisements for a particular brand represent alterations of earlier promotions for that brand. The advertisements for each brand were aggregated into a single case by weighting; each advertisement was assigned a weight of 1/n, with n representing the number of advertisements for that brand. When a significant chi-square value was obtained for a cross-tabulation, the source of significance within the table was determined by examining the cell-adjusted standardized residuals using a P <.05 criterion.30
Results
The Figure shows the percentage of advertisements for which the 5 medical condition information codes and 6 treatment information codes were present. These results have been ordered in the Figure on the basis of frequency of occurrence.
Information About Medical Conditions
The medical condition name was provided in virtually all of the advertisements (95%). This information was excluded in 1 “teaser ad” that simply showed a bottle of tablets for a forthcoming treatment, in 7 advertisements alerting patients of a new formulation of a drug already prescribed for them, and in 9 advertisements notifying patients of an alternative delivery system for an established treatment. Sixty percent of the advertisements described at least one symptom of the condition treated by the promoted drug or explicitly stated that the condition could be a “silent disease”. Among those advertisements that omitted symptom information, 54% were targeted to individuals who had already been diagnosed with the condition and were presumably aware of its symptoms. The remaining advertisements tended to be for conditions with well-known symptoms (eg, pregnancy, impotence, and tobacco addiction). Information about a precursor to the condition such as a cause or risk factor was provided in 27% of the advertisements. A total of 6% of advertisements provided precursor information by means of a set of diagnostic questions, such as a quiz that readers could use to assess their risk of being or becoming affected by the condition (data not diagramed). Information about condition prevalence (12% of advertisements) and clarifications about a condition-related misconception (9%) was rarely provided.
Information About Treatments
The Figure also displays the percentage of advertisements reporting each of the 6 types of treatment information. Most often present was information about a drug’s mechanism of action, found in 36% of the advertisements. An acknowledgement of the existence of one or more competing treatments was offered in 29% of the advertisements. Supportive behaviors such as changes in diet, physical activity, and sleeping patterns were reported in 24% of the advertisements. In our judgment, relevant supportive behaviors are often recommended by physicians for most of the conditions addressed in these advertisements. Readers were given information about time to onset of action in 20% of the advertisements and typical required treatment duration in 11%. A success rate estimate was rarely reported (9%).
Three of these treatment information codes were significantly associated with the medical conditions variable (all P values=.02). Supportive behaviors that could be used alone or to enhance the effectiveness of therapy were more likely to be found in diabetes and tobacco cessation treatments (75% and 67% of brands, respectively). Time to onset of action information was especially likely to be offered in advertisements for dermatological and urological problems (58% and 50%, respectively). Information about treatment duration was most often found in advertisements for infection and dermatological treatments (50% and 33%, respectively).
Indexes
For summative purposes we examined the condition information, treatment information, and overall education indexes for each of the 14 medical condition categories, using the drug brand as the unit of analysis. The average brand provided only 2.1 of the 5 types of information about the condition for which the drug was marketed. Only 1.2 of the 6 types of treatment information we coded was provided in the average brand-weighted advertisement. Summing across the condition and treatment information variables, the average number of educational codes present for these 101 brands was only 3.2 out of a possible score of 11 (range=1.0-7.3).
Next we categorized the brands under the medical condition category for which each was promoted Table 3. Despite weak statistical power, a significant difference across the 14 medical conditions was found for 2 of the 3 indexes, the condition information index (F [13,87]=3.05, P=.001) and the combined education index (F [13,87]=2.39, P=.008). On the basis of post hoc comparisons, both effects could be attributed to the greater provision of information in promotions for urologic brands than in HIV/AIDS brand advertisements.
Information Formats
Information about conditions and treatments was typically provided in narrative form. Tabular data or charts (eg, bar, pie, or column charts) were rarely used to provide information about the drug (2% of advertisements) or the medical condition (<1%). Diagrams and pictures were also rarely used to provide information about the effectiveness of the treatment (2% of advertisements) or the nature of the condition (7%).
Other Sources of Information
One view of DTC advertising is that it simply provides an introduction to a treatment and that more detailed patient education is best provided in other ways, such as with brochures and videotaped presentations or through the Internet. An explicit offer of printed or audiovisual information was provided in 35% of the advertisements, and a toll-free information telephone number was found in 73%. Increasingly, the Internet is being used as a means through which interested readers can obtain more information about advertised products. In 1996, 1997, and 1998, the percentages of advertisements that provided a Web site address were 14%, 33%, and 57%, respectively. Most telephone numbers were numeric (55%), but some of these and all Web addresses referenced a benefit of use (eg, 1-800-5-SHUT-EYE for a treatment of insomnia), the treated body part (eg, www.kidsears.com for a treatment of ear infections), the medical condition treated (eg, 1-800-66-ANGINA), the brand name of the drug (eg, www.atrovent.com), or the drug company’s name (eg, 1-800-GLAXO-RX).
Discussion
A time may come when DTC advertising is recommended for its educational value, but that day is not yet at hand. The advertisements that have appeared during the past decade have been superficial in their coverage of medical conditions and their treatments. Although most provide the name of the treated medical condition and its associated symptoms, the large majority do not inform potential patients about such basic matters as the risk factors for the condition or its prevalence. Likewise, these advertisements seldom educate patients about the mechanism of action by which the drug treats a particular condition, its success in doing so, alternative treatments, and behavioral changes that could augment or supplant treatment. To their credit, most of these promotions do offer the reader alternative ways of learning more about their condition and the advertised drug.
The medical community should exert pressure on the drug industry to incorporate more information about conditions and treatments in its advertising. It is in the industry’s best interest to do so. Advertising that incorporates quality health and drug information will have greater credibility, and deservedly so. Also, consumers are less likely to be influenced by messages that appear to be more promotional than informational.31 Thus, providing complete and accurate information is the right thing to do and may even enhance the effectiveness of DTC advertising. If such information is not provided voluntarily by the industry in future advertising, the medical establishment should lobby for regulation.
These results also highlight an important opportunity for professional organizations to contribute to consumer education on prescription drugs. According to the National Institute for Health Care Management, the most successfully promoted drugs fall into 5 categories: antidepressants, cholesterol-lowering agents, gastric acid reducers, oral antihistamines, and antihypertensives.32 Organizations such as the American Academy of Family Physicians and the American College of Physicians already produce patient informational materials, but these efforts could be intensified so physicians would have a ready source of “counter-detailing.” Patients requesting drugs for which the indications are questionable could be given a handout and told, “This is what my professional society has to say about _________. This information is produced by the best experts in the field and provides a more balanced view than what you will find in profit-motivated advertisements. Look it over, and let’s talk about this at our next visit.”
If the industry were to be prodded into taking a more educational stance in its consumer-targeted prescription drug advertising, the impact of this shift on physician-patient interactions would need to be investigated. A more educated patient may take less time to treat and counsel, might show greater adherence to treatment regimens, and could assume greater responsibility for his or her health. On the other hand, educational promotions may lead to requests from more determined patients for drugs that are not medically indicated, requiring time-consuming re-education by their physicians.
Limitations
Our study has limitations. We have evaluated the educational qualities of these advertisements by imposing a common set of standards for each promotion. A stronger approach would have been to convene a panel of experts for each of the many conditions treated by the promoted drugs to identify the specific details that consumers need to know about the condition and its treatment. Resource constraints prevented us from developing expert-based educational standards for each condition or disease. However, it is reasonable to expect the drug industry to provide to consumers basic information about the treatments being promoted and the conditions these drugs address. We acknowledge that particular content analytic codes may have been irrelevant to certain conditions. For instance, precursor information does not need to be provided when the causes of the condition are obvious (eg, advertisements for contraceptive drugs); mention of supportive lifestyle changes should not be demanded when helpful behavioral changes do not exist (eg, advertisements for hair loss treatments); and references to the presence of competing treatments should not be expected in the rare event that such treatments are nonexistent. Such exceptions aside, we believe that the 11 codes assessed are both fundamental and relevant to the vast majority of the conditions covered by these advertisements.
Also, with this preliminary investigation we sought only to assess the extent to which the industry is making an effort to provide information about medical conditions and treatments for those conditions. We did not examine the educational quality of these efforts, including the completeness and accuracy of information provided about conditions and treatments. Such assessments will require input from medical and pharmaceutical experts selected for their specialized knowledge.
Conclusions
We acknowledge that instances of informative advertising can readily be found in our sample of advertisements. Thus, although DTC advertising in general is not serving the information needs of consumers, there are companies and individuals within the industry who are motivated to treat their advertising as vehicles for effective health promotion through quality education. Billions of dollars will be spent in the next few years on consumer-targeted prescription drug promotions.4 Consumers will be very receptive to those advertisements that address their personal health needs and concerns33; many will talk to their primary care physicians as a result of these promotions.34 The drug industry thus has a tremendous opportunity to silence its critics and improve the public’s health by providing objective medical and drug information.35
Acknowledgments
We wish to acknowledge the contributions made by Ronald Emerick, Robert LaGreca, Love Lord, and Sarah Shaw in the collection and coding of advertisements and by 2 anonymous peer reviewers for their thoughtful evaluations.
Related Resources:
- U.S. Food & Drug Administration, Center for Drug Evaluation and Research, Division of Drug Marketing, Advertising and Communications www.fda.gov/cder/ddmac/
- Public Citizen Health Research Group “Drugs” page, listing numerous government documents that describe recommendations for drug use, sales and marketing. http://www.citizen.org/hrg/publications/drugs.htm
- AARP Executive Summary: Are Consumers Well Informed About Prescription Drugs? The Impact of Printed Direct-to-Consumer Advertising http://research.aarp.org/health/2000_04_advertising_1.html
- drkoop.com-Perspective: ‘Direct-to-Consumer’ Advertising of Prescription Drugs is More Harmful than Helpful to Consumers www.drkoop.com/news/focus/november/dtc_no.html
METHODS: We collected advertisements appearing in 18 popular magazines from 1989 through 1998. Two coders independently evaluated 320 advertisements encompassing 101 drug brands to determine if information appeared about specific aspects of the medical conditions for which the drug was promoted and about the treatment (mean k reliability=0.91). We employed basic descriptive statistics using the advertisement as the unit of analysis and cross-tabulations using the brand as the unit of analysis.
RESULTS: Virtually all the advertisements gave the name of the condition treated by the promoted drug, and a majority provided information about the symptoms of that condition. However, few reported details about the condition’s precursors or its prevalence; attempts to clarify misconceptions about the condition were also rare. The advertisements seldom provided information about the drug’s mechanism of action, its success rate, treatment duration, alternative treatments, and behavioral changes that could enhance the health of affected patients.
CONCLUSIONS: Informative advertisements were identified, but most of the promotions provided only a minimal amount of information. Strategies for improving the educational value of DTC advertisements are considered.
The appropriateness of direct-to-consumer (DTC) prescription drug advertising has been a topic of heated discussion in recent years.1,2 This debate is likely to intensify as the drug industry’s advertising expenditures continue to increase3,4 and the marketing channels employed expand to include television, which may not lend itself to highly informative messages, and the Internet, which is difficult to regulate.5 Ironically, DTC advertising proponents and opponents have most often based their positions on assumptions about the informational value of such promotions.6 However, empirical data on this issue are sparse.
The opponents of DTC advertising argue that the objectives of drug promotion and health education are inherently at odds.7 It has been noted that the aim of DTC advertising is to increase sales, not optimize health care.8,9 It has been presumed that the profit motive discourages the provision of complete and accurate information about pharmaceuticals.10 Reports have documented violations of the public trust11 and the dissemination of misleading information to health care professionals.12,13 If even highly trained physicians are prone to influence by commercial rather than scientific sources of pharmaceutical information,14 how can consumers protect themselves from misleading claims? To make matters worse, the public may be ill equipped to fully understand such advertising, no matter how accurate its content might be.7,15
Defenders of DTC advertising counter these arguments by suggesting that selling and educating are not necessarily incompatible goals16 and that it is in the industry’s self-interest to be truthful with consumers.17 Although instances of misleading DTC advertising can be identified, such cases are said to be rare.5 Proponents of DTC advertising typically attribute greater sophistication to the public than critics do,16 rejecting the notion that consumers cannot comprehend such advertising.18 In any regard, misunderstandings can be corrected by the physician, in whom the power to prescribe ultimately resides.1 Although many physicians oppose DTC advertising, a majority feel that such advertisements might help patients become better informed about drugs,19 conjecture that has empirical support.20
We have attempted to contribute to the debate on the value of DTC advertising by describing its content. In a recent JFP article,21 we documented the types of appeals and inducements used in these promotions. In this article we describe the scope of the educational efforts represented in advertisements appearing in consumer publications from 1989 through 1998. After identifying basic issues that most DTC advertisements should address to have educational merit, we examined the extent to which such promotions provide consumers with information about medical conditions, including their symptoms, precursors, and prevalence. It has been argued that DTC advertising increases the public’s awareness of these aspects of diseases and other health conditions. This kind of patient education is said to help people to take responsibility for their health by teaching them how to recognize disease and motivating them to seek medical attention for conditions that might otherwise be left undiagnosed and untreated.1,22-24
We then examined the extent to which DTC advertising incorporates information about treatments for those conditions. It has been said that consumer-targeted prescription drug advertising provides the public with valuable information about available treatments that leads to a better match between patients’ needs and available drugs.25 Specifically, we examined the extent to which DTC advertising offers details about promoted drugs’ mechanisms of action, the duration and success of advertised treatments, behavioral modifications that can improve the patient’s health independently or in concert with drug treatment, and competing treatments.
Methods
Sampling Procedure
Major popular magazines were ranked in terms of the average number of advertising pages sold from 1989 through 1996 (the last year for which data were available) and then grouped into an established set of categories.26 The highest-ranked publication within each of 13 categories was selected for inclusion in our sample. To these magazines we added 5 publications with audiences composed of narrower segments of our population defined by ethnicity, age, and sexual orientation. Thus, we garnered our sample of advertisements from 18 diverse publications Table 1. All the advertisements promoting a specific, named prescription drug that appeared in these magazines from January 1989 through December 1998 were photocopied. Advertisements for the same brand often differed in nonsubstantive ways and were thus coded as a single case based on detailed rules that are available from the authors. After aggregating essentially identical advertisements, 320 remained for analysis covering 101 distinct brands.
Medical Conditions Classification
Each advertisement was grouped in one of 14 medical condition categories based on the indication of the marketed drug. The numbers of advertisements and brands within each of these categories were as follows: allergies (46 advertisements, 8 brands); cancer (2, 2); cardiovascular disease (36, 10); dermatologic conditions (37, 12); diabetes (9, 4); gastrointestinal/nutritional disorders (17, 7); human immunodeficiency virus (HIV)/acquired immune deficiency syndrome (AIDS; 33, 11); infections, non-HIV (16, 6); musculoskeletal disorders (17, 7); obstetric/gynecologic conditions (45, 10); psychiatric/neurologic disorders (17, 7); respiratory disorders (3, 3); tobacco addiction (23, 6); and urologic conditions (19, 8).
Advertisement Coding
Two coders independently analyzed each advertisement. Our goal for this classification effort was to determine whether each advertisement contained 5 kinds of information pertaining to the focal medical condition and 6 types of drug treatment information Table 2.
Statistical Analyses
We examined each of the 11 education codes listed in Table 2 and computed 3 more general measures. We created a condition information index by summing the number of medical condition codes present within each advertisement (theoretical range=0-5). Likewise, a treatment information index was computed by summing the number of treatment codes present within each advertisement (theoretical range=0-6). A combined education index was created by adding the condition and treatment information variables (theoretical range=0-11).
The reliability of coding was assessed with the k statistic, a measure of inter-rater agreement with a range of 0 to 1.0.27 The mean k value for the variables reported was 0.91 (range=0.88-1.0). Landis and Koch28 suggest that k values in this range can be considered “almost perfect.”
We employed 2 units of analyses. For our descriptive analyses, the advertisement served as the unit of analysis (N=320). We relied on the brand (N=101) as the unit of analysis when computing cross-tabulations and one-way analyses of variance. Using the advertisement as the unit for inferential statistical analyses would have violated the assumption of independence of observations,29 because most advertisements for a particular brand represent alterations of earlier promotions for that brand. The advertisements for each brand were aggregated into a single case by weighting; each advertisement was assigned a weight of 1/n, with n representing the number of advertisements for that brand. When a significant chi-square value was obtained for a cross-tabulation, the source of significance within the table was determined by examining the cell-adjusted standardized residuals using a P <.05 criterion.30
Results
The Figure shows the percentage of advertisements for which the 5 medical condition information codes and 6 treatment information codes were present. These results have been ordered in the Figure on the basis of frequency of occurrence.
Information About Medical Conditions
The medical condition name was provided in virtually all of the advertisements (95%). This information was excluded in 1 “teaser ad” that simply showed a bottle of tablets for a forthcoming treatment, in 7 advertisements alerting patients of a new formulation of a drug already prescribed for them, and in 9 advertisements notifying patients of an alternative delivery system for an established treatment. Sixty percent of the advertisements described at least one symptom of the condition treated by the promoted drug or explicitly stated that the condition could be a “silent disease”. Among those advertisements that omitted symptom information, 54% were targeted to individuals who had already been diagnosed with the condition and were presumably aware of its symptoms. The remaining advertisements tended to be for conditions with well-known symptoms (eg, pregnancy, impotence, and tobacco addiction). Information about a precursor to the condition such as a cause or risk factor was provided in 27% of the advertisements. A total of 6% of advertisements provided precursor information by means of a set of diagnostic questions, such as a quiz that readers could use to assess their risk of being or becoming affected by the condition (data not diagramed). Information about condition prevalence (12% of advertisements) and clarifications about a condition-related misconception (9%) was rarely provided.
Information About Treatments
The Figure also displays the percentage of advertisements reporting each of the 6 types of treatment information. Most often present was information about a drug’s mechanism of action, found in 36% of the advertisements. An acknowledgement of the existence of one or more competing treatments was offered in 29% of the advertisements. Supportive behaviors such as changes in diet, physical activity, and sleeping patterns were reported in 24% of the advertisements. In our judgment, relevant supportive behaviors are often recommended by physicians for most of the conditions addressed in these advertisements. Readers were given information about time to onset of action in 20% of the advertisements and typical required treatment duration in 11%. A success rate estimate was rarely reported (9%).
Three of these treatment information codes were significantly associated with the medical conditions variable (all P values=.02). Supportive behaviors that could be used alone or to enhance the effectiveness of therapy were more likely to be found in diabetes and tobacco cessation treatments (75% and 67% of brands, respectively). Time to onset of action information was especially likely to be offered in advertisements for dermatological and urological problems (58% and 50%, respectively). Information about treatment duration was most often found in advertisements for infection and dermatological treatments (50% and 33%, respectively).
Indexes
For summative purposes we examined the condition information, treatment information, and overall education indexes for each of the 14 medical condition categories, using the drug brand as the unit of analysis. The average brand provided only 2.1 of the 5 types of information about the condition for which the drug was marketed. Only 1.2 of the 6 types of treatment information we coded was provided in the average brand-weighted advertisement. Summing across the condition and treatment information variables, the average number of educational codes present for these 101 brands was only 3.2 out of a possible score of 11 (range=1.0-7.3).
Next we categorized the brands under the medical condition category for which each was promoted Table 3. Despite weak statistical power, a significant difference across the 14 medical conditions was found for 2 of the 3 indexes, the condition information index (F [13,87]=3.05, P=.001) and the combined education index (F [13,87]=2.39, P=.008). On the basis of post hoc comparisons, both effects could be attributed to the greater provision of information in promotions for urologic brands than in HIV/AIDS brand advertisements.
Information Formats
Information about conditions and treatments was typically provided in narrative form. Tabular data or charts (eg, bar, pie, or column charts) were rarely used to provide information about the drug (2% of advertisements) or the medical condition (<1%). Diagrams and pictures were also rarely used to provide information about the effectiveness of the treatment (2% of advertisements) or the nature of the condition (7%).
Other Sources of Information
One view of DTC advertising is that it simply provides an introduction to a treatment and that more detailed patient education is best provided in other ways, such as with brochures and videotaped presentations or through the Internet. An explicit offer of printed or audiovisual information was provided in 35% of the advertisements, and a toll-free information telephone number was found in 73%. Increasingly, the Internet is being used as a means through which interested readers can obtain more information about advertised products. In 1996, 1997, and 1998, the percentages of advertisements that provided a Web site address were 14%, 33%, and 57%, respectively. Most telephone numbers were numeric (55%), but some of these and all Web addresses referenced a benefit of use (eg, 1-800-5-SHUT-EYE for a treatment of insomnia), the treated body part (eg, www.kidsears.com for a treatment of ear infections), the medical condition treated (eg, 1-800-66-ANGINA), the brand name of the drug (eg, www.atrovent.com), or the drug company’s name (eg, 1-800-GLAXO-RX).
Discussion
A time may come when DTC advertising is recommended for its educational value, but that day is not yet at hand. The advertisements that have appeared during the past decade have been superficial in their coverage of medical conditions and their treatments. Although most provide the name of the treated medical condition and its associated symptoms, the large majority do not inform potential patients about such basic matters as the risk factors for the condition or its prevalence. Likewise, these advertisements seldom educate patients about the mechanism of action by which the drug treats a particular condition, its success in doing so, alternative treatments, and behavioral changes that could augment or supplant treatment. To their credit, most of these promotions do offer the reader alternative ways of learning more about their condition and the advertised drug.
The medical community should exert pressure on the drug industry to incorporate more information about conditions and treatments in its advertising. It is in the industry’s best interest to do so. Advertising that incorporates quality health and drug information will have greater credibility, and deservedly so. Also, consumers are less likely to be influenced by messages that appear to be more promotional than informational.31 Thus, providing complete and accurate information is the right thing to do and may even enhance the effectiveness of DTC advertising. If such information is not provided voluntarily by the industry in future advertising, the medical establishment should lobby for regulation.
These results also highlight an important opportunity for professional organizations to contribute to consumer education on prescription drugs. According to the National Institute for Health Care Management, the most successfully promoted drugs fall into 5 categories: antidepressants, cholesterol-lowering agents, gastric acid reducers, oral antihistamines, and antihypertensives.32 Organizations such as the American Academy of Family Physicians and the American College of Physicians already produce patient informational materials, but these efforts could be intensified so physicians would have a ready source of “counter-detailing.” Patients requesting drugs for which the indications are questionable could be given a handout and told, “This is what my professional society has to say about _________. This information is produced by the best experts in the field and provides a more balanced view than what you will find in profit-motivated advertisements. Look it over, and let’s talk about this at our next visit.”
If the industry were to be prodded into taking a more educational stance in its consumer-targeted prescription drug advertising, the impact of this shift on physician-patient interactions would need to be investigated. A more educated patient may take less time to treat and counsel, might show greater adherence to treatment regimens, and could assume greater responsibility for his or her health. On the other hand, educational promotions may lead to requests from more determined patients for drugs that are not medically indicated, requiring time-consuming re-education by their physicians.
Limitations
Our study has limitations. We have evaluated the educational qualities of these advertisements by imposing a common set of standards for each promotion. A stronger approach would have been to convene a panel of experts for each of the many conditions treated by the promoted drugs to identify the specific details that consumers need to know about the condition and its treatment. Resource constraints prevented us from developing expert-based educational standards for each condition or disease. However, it is reasonable to expect the drug industry to provide to consumers basic information about the treatments being promoted and the conditions these drugs address. We acknowledge that particular content analytic codes may have been irrelevant to certain conditions. For instance, precursor information does not need to be provided when the causes of the condition are obvious (eg, advertisements for contraceptive drugs); mention of supportive lifestyle changes should not be demanded when helpful behavioral changes do not exist (eg, advertisements for hair loss treatments); and references to the presence of competing treatments should not be expected in the rare event that such treatments are nonexistent. Such exceptions aside, we believe that the 11 codes assessed are both fundamental and relevant to the vast majority of the conditions covered by these advertisements.
Also, with this preliminary investigation we sought only to assess the extent to which the industry is making an effort to provide information about medical conditions and treatments for those conditions. We did not examine the educational quality of these efforts, including the completeness and accuracy of information provided about conditions and treatments. Such assessments will require input from medical and pharmaceutical experts selected for their specialized knowledge.
Conclusions
We acknowledge that instances of informative advertising can readily be found in our sample of advertisements. Thus, although DTC advertising in general is not serving the information needs of consumers, there are companies and individuals within the industry who are motivated to treat their advertising as vehicles for effective health promotion through quality education. Billions of dollars will be spent in the next few years on consumer-targeted prescription drug promotions.4 Consumers will be very receptive to those advertisements that address their personal health needs and concerns33; many will talk to their primary care physicians as a result of these promotions.34 The drug industry thus has a tremendous opportunity to silence its critics and improve the public’s health by providing objective medical and drug information.35
Acknowledgments
We wish to acknowledge the contributions made by Ronald Emerick, Robert LaGreca, Love Lord, and Sarah Shaw in the collection and coding of advertisements and by 2 anonymous peer reviewers for their thoughtful evaluations.
Related Resources:
- U.S. Food & Drug Administration, Center for Drug Evaluation and Research, Division of Drug Marketing, Advertising and Communications www.fda.gov/cder/ddmac/
- Public Citizen Health Research Group “Drugs” page, listing numerous government documents that describe recommendations for drug use, sales and marketing. http://www.citizen.org/hrg/publications/drugs.htm
- AARP Executive Summary: Are Consumers Well Informed About Prescription Drugs? The Impact of Printed Direct-to-Consumer Advertising http://research.aarp.org/health/2000_04_advertising_1.html
- drkoop.com-Perspective: ‘Direct-to-Consumer’ Advertising of Prescription Drugs is More Harmful than Helpful to Consumers www.drkoop.com/news/focus/november/dtc_no.html
1. Holmer AF. Direct-to-consumer prescription drug advertising builds bridges between patients and physicians. JAMA 1999;281:380-82.
2. Hollon MF. Direct-to-consumer marketing of prescription drugs: creating consumer demand. JAMA 1999;281:382-84.
3. Barrett A. Are drug ads a cure-all? Business Week March 30, 1998;59-60.
4. Growth seen in ads for direct-to-consumer drugs. AMA News April 27, 1998;16.-
5. Maguire P. How direct-to-consumer advertising is putting the squeeze on physicians. ACP-ASIM Observer 1999;19:1, 24-25.
6. Pierpaoli PG. ASHP’s position on direct-to-consumer advertising of prescription drug products. Am J Hosp Pharm 1986;43:1763-65.
7. Cohen EP. Direct-to-the-public advertisement of prescription drugs. N Engl J Med 1988;318:373.-
8. Committee on Drugs. Prescription drug advertising direct to the consumer. Pediatrics 1991;88:174-75.
9. Bradley LR, Zito JM. Direct-to-consumer prescription drug advertising. Med Care 1997;35:86-92.
10. T’hoen E. Direct-to-consumer advertising: For better profits or for better health? Am J Health-Syst Pharm 1998;55:594-97.
11. Wall Street Journal August 22, 1997; 168:B5. Schering-Plough is told to halt Claritin TV ads.
12. Wilkes MS, Doblin BH, Shapiro MF. Pharmaceutical advertisements in leading medical journals: experts’ assessments. Ann Intern Med 1992;116:912-19.
13. Stryer D, Bero LA. Characteristics of materials distributed by drug companies: an evaluation of appropriateness. JGIM 1996;11:575-83.
14. Avorn J, Chen M, Hartley R. Scientific versus commercial sources of influence on the prescribing behavior of physicians. Am J Med 1982;73:4-8.
15. Schommer JC, Doucette WR, Mehta BH. Rote learning after exposure to a direct-to-consumer television advertisement for a prescription drug. Clin Ther 1998;20:617-32.
16. Ingram RA. Some comments on direct-to-consumer advertising. J Pharm Marketing Manage 1992;7:67-74.
17. Wind Y. Pharmaceutical advertising: a business school perspective. Arch Fam Med 1994;3:321-23.
18. Rubin PH. What the FDA doesn’t want you to know. Am Enterprise 1991;2:18-20.
19. Lipsky MS, Taylor CA. The opinions and experiences of family physicians regarding direct-to-consumer advertising. J Fam Pract 1997;45:495-99.
20. Peyrot M, Alperstein NM, Van Doren D, Poli LG. Direct-to-consumer ads can influence behavior: advertising increases consumer knowledge and prescription drug requests. Marketing Health Services 1998;18:26-32.
21. Bell RA, Kravitz RL, Wilkes MS. Direct-to-consumer prescription drug advertising, 1989-1998: a content analysis of conditions, targets, inducements and appeals. J Fam Pract 2000;49:329-35.
22. Keith A. The benefits of pharmaceutical promotion: an economic and health perspective. J Pharm Marketing Manage 1992;7:121-33.
23. Whyte J. Direct consumer advertising of prescription drugs. JAMA 1993;268:146,150.-
24. Reynolds WJ. Trends in advertising pharmaceuticals: a publisher’s perspective. J Pharm Marketing Manage 1992;7:5-22.
25. Masson A, Rubin PH. Matching prescription drugs and consumers. N Engl J Med 1985;313:513-15.
26. Katz WA. Magazines for libraries. New Providence, NJ: R.R. Bowker; 1997.
27. Cohen J. A coefficient of agreement for nominal scales: educational and psychological measurement 1960;20:37-46.
28. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159-74.
29. Hays WL. Statistics. New York, NY: Holt, Rinehart, and Winson, 1981.
30. Everitt BS. The analysis of contingency tables. London, England: Chapman and Hall; 1980.
31. Petty RE, Cacioppo JT. Communication and persuasion. New York, NY: Springer-Verlag; 1986.
32. National Institute for Health Care Management, Research and Education Foundation. Factors affecting the growth of prescription drug expenditures. July 9, 1999. Available at www.nihcm.org.
33. Bell RA, Kravitz RL, Wilkes M. Direct-to-consumer prescription drug advertising and the public. JGIM 1999;13:651-57.
34. Bell RA, Wilkes MS, Kravitz RL. Advertising-induced prescription drug requests: patients’ anticipated reactions to a physician who refuses. J Fam Pract 1999;48:446-52.
35. Wilkes MS, Bell RA, Kravitz RL. Consumer-directed prescription drug advertising: trends, impact, and implications. Health Aff 2000;19:110-28.
1. Holmer AF. Direct-to-consumer prescription drug advertising builds bridges between patients and physicians. JAMA 1999;281:380-82.
2. Hollon MF. Direct-to-consumer marketing of prescription drugs: creating consumer demand. JAMA 1999;281:382-84.
3. Barrett A. Are drug ads a cure-all? Business Week March 30, 1998;59-60.
4. Growth seen in ads for direct-to-consumer drugs. AMA News April 27, 1998;16.-
5. Maguire P. How direct-to-consumer advertising is putting the squeeze on physicians. ACP-ASIM Observer 1999;19:1, 24-25.
6. Pierpaoli PG. ASHP’s position on direct-to-consumer advertising of prescription drug products. Am J Hosp Pharm 1986;43:1763-65.
7. Cohen EP. Direct-to-the-public advertisement of prescription drugs. N Engl J Med 1988;318:373.-
8. Committee on Drugs. Prescription drug advertising direct to the consumer. Pediatrics 1991;88:174-75.
9. Bradley LR, Zito JM. Direct-to-consumer prescription drug advertising. Med Care 1997;35:86-92.
10. T’hoen E. Direct-to-consumer advertising: For better profits or for better health? Am J Health-Syst Pharm 1998;55:594-97.
11. Wall Street Journal August 22, 1997; 168:B5. Schering-Plough is told to halt Claritin TV ads.
12. Wilkes MS, Doblin BH, Shapiro MF. Pharmaceutical advertisements in leading medical journals: experts’ assessments. Ann Intern Med 1992;116:912-19.
13. Stryer D, Bero LA. Characteristics of materials distributed by drug companies: an evaluation of appropriateness. JGIM 1996;11:575-83.
14. Avorn J, Chen M, Hartley R. Scientific versus commercial sources of influence on the prescribing behavior of physicians. Am J Med 1982;73:4-8.
15. Schommer JC, Doucette WR, Mehta BH. Rote learning after exposure to a direct-to-consumer television advertisement for a prescription drug. Clin Ther 1998;20:617-32.
16. Ingram RA. Some comments on direct-to-consumer advertising. J Pharm Marketing Manage 1992;7:67-74.
17. Wind Y. Pharmaceutical advertising: a business school perspective. Arch Fam Med 1994;3:321-23.
18. Rubin PH. What the FDA doesn’t want you to know. Am Enterprise 1991;2:18-20.
19. Lipsky MS, Taylor CA. The opinions and experiences of family physicians regarding direct-to-consumer advertising. J Fam Pract 1997;45:495-99.
20. Peyrot M, Alperstein NM, Van Doren D, Poli LG. Direct-to-consumer ads can influence behavior: advertising increases consumer knowledge and prescription drug requests. Marketing Health Services 1998;18:26-32.
21. Bell RA, Kravitz RL, Wilkes MS. Direct-to-consumer prescription drug advertising, 1989-1998: a content analysis of conditions, targets, inducements and appeals. J Fam Pract 2000;49:329-35.
22. Keith A. The benefits of pharmaceutical promotion: an economic and health perspective. J Pharm Marketing Manage 1992;7:121-33.
23. Whyte J. Direct consumer advertising of prescription drugs. JAMA 1993;268:146,150.-
24. Reynolds WJ. Trends in advertising pharmaceuticals: a publisher’s perspective. J Pharm Marketing Manage 1992;7:5-22.
25. Masson A, Rubin PH. Matching prescription drugs and consumers. N Engl J Med 1985;313:513-15.
26. Katz WA. Magazines for libraries. New Providence, NJ: R.R. Bowker; 1997.
27. Cohen J. A coefficient of agreement for nominal scales: educational and psychological measurement 1960;20:37-46.
28. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159-74.
29. Hays WL. Statistics. New York, NY: Holt, Rinehart, and Winson, 1981.
30. Everitt BS. The analysis of contingency tables. London, England: Chapman and Hall; 1980.
31. Petty RE, Cacioppo JT. Communication and persuasion. New York, NY: Springer-Verlag; 1986.
32. National Institute for Health Care Management, Research and Education Foundation. Factors affecting the growth of prescription drug expenditures. July 9, 1999. Available at www.nihcm.org.
33. Bell RA, Kravitz RL, Wilkes M. Direct-to-consumer prescription drug advertising and the public. JGIM 1999;13:651-57.
34. Bell RA, Wilkes MS, Kravitz RL. Advertising-induced prescription drug requests: patients’ anticipated reactions to a physician who refuses. J Fam Pract 1999;48:446-52.
35. Wilkes MS, Bell RA, Kravitz RL. Consumer-directed prescription drug advertising: trends, impact, and implications. Health Aff 2000;19:110-28.
Can Granuloma Annulare Evolve Into Cutaneous Sarcoidosis?
Universal Newborn Hearing Screening
METHODS: We used DPOAE to screen newborn infants from February 1997 to March 1999.
RESULTS: Of 1002 infants, 111 failed the initial screen (11.1%). When screening was repeated, only 2 infants failed. One infant failed the second screen and a tympanogram. He was treated and he passed a third use of DPOAE. An additional infant failed the repeat screen but passed the tympanogram. That infant was referred on for auditory brain response testing.
CONCLUSIONS: DPOAE testing can be accomplished easily in a normal newborn nursery with an acceptable false-positive rate when a two-stage approach is used. The cost for each test was $19.88. The cost to find the 1 infant with sensory neural hearing loss was $22,114.
In 1993 a consensus statement from the National Institutes of Health (NIH)1 recommended universal newborn hearing screening by the age of 3 months and also stated that otoacoustic emission might be the technology used for screening. These recommendations were based on the following: (1) the incidence of hearing loss is 1 to 6 per 1000; (2) only one half of the infants with hearing loss are discovered with high-risk screening; (3) the current average age at diagnosis of hearing loss is 2.5 years; and (4) early identification and treatment by the age of 6 months will improve outcomes.1 The 1994 position statement from the Joint Committee on Infant Hearing2 reiterated this recommendation.
Bess and Paradise3 and Paradise4 succinctly enumerated the difficulties with universal screening, including high false-positive rates, overall expense, patient acceptance, and feasibility. Since 1993, research and publications5-9 have supported universal screening; now the cost of identifying an infant with hearing loss is less than the cost of identifying an infant with phenylketonuria.10 Also, studies by Appuzzo and Yoshinaga-Itano11 and Yoshinaga-Itano and colleagues12 showed early identification and intervention improves language and social development, and today 37 states have more than 1 hospital providing universal hearing screening programs.13 A variety of screening devices are used, including automated auditory brainstem responses (AABR), transient evoked otoacoustic emission, and distortion product otoacoustic emissions (DPOAE).
Methods
Study Group
Well infants hospitalized from 24 to 72 hours were screened at ages 6 to 72 hours. Data was obtained from a sample of 1002 infants screened at Rapid City Regional Hospital between February 1997 and March 1999. During that period, newborn infants receiving care from family physicians and a pediatrician in the regular nursery were screened. Specific statistics on the age of each infant at screening were not kept.
The screening team consisted of a family practice physician, a registered nurse, and 2 family practice residents. All members of the team performed hearing screening.
We explained the hearing screen procedure, methods, risks, and benefits to the parents of the patient, and obtained informed consent. The follow-up plan for failed hearing screens was also outlined.
Equipment
The GSI 60 computer-based DPOAE instrument (Grason-Stadler, Inc, Milford, NH) was used to perform screenings. It generates paired frequencies F1 (65 dB) and F2 (55 dB). These frequencies travel through the middle ear to the cochlea, where a third tone is generated at the outer hair cell level. Normal cochlear stimulation in this manner produces a DPOAE at a specific frequency predicted by the formula 2F1-F2. During distortion product measurement, the frequency range of 2000 to 4000 Hz was selected for distortion product frequencies.
Procedure
The initial DPOAE measure was obtained on the morning of hospital discharge, with most infants tested between the ages of 12 and 72 hours. Average actual DPOAE screen time was 20 minutes. Passing the test was defined as an emission signal reproducible at 3 frequencies on 2 separate tests. The distortion product was required to be 5 decibels above the noise floor to be considered a pass.
Infants who failed the DPOAE measure were rescreened using DPOAE at 8 weeks. All second-stage screen failures were screened with tympanometry and DPOAE and were referred to their private physician or an ear, nose, and throat specialist. Infants that failed DPOAE and passed tympanometry were to be referred for diagnostic auditory brain response testing (Figure).
Results
Of the 1002 infants screened, 11.1% (111) failed the initial screening (Table 1). Seventy-nine infants were returned to our clinic for retesting 1 to 3 months after their initial evaluation. Two infants (0.2% of the total) failed the second screening. One infant failed the second screen and the tympanogram. He was treated and he passed the third DPOAE. An additional infant failed the repeat screen but passed the tympanogram. She was referred for auditory brain response testing and was found to have mild sensory neural hearing loss. A total of 32 infants were lost to follow-up despite frequent attempts to contact parents by telephone and mail.
Discussion
Recent advances in technology have provided the means to screen newborns and infants for hearing deficits. The 2 most commonly used technologies are AABRs and otoacoustic emissions. Otoacoustic emissions were first described by Kemp in 1978. The cochlear basal membrane vibrates when stimulated by sound, and this vibration sends a retrograde wave back through the cochlear fluid that ultimately vibrates the eardrum, producing a sound wave that can be detected by a microphone at the external ear. DPOAE are produced by stimulating the cochlea with a series of 2 specific frequencies, resulting in a single predictable frequency response.14 Transient evoked otoacoustic emissions, an alternative to DPOAE, are evoked by a click that results in the emission of several frequencies at the same time. AABRs are also used in screening programs, with the advantage of testing both the cochlea and retrocochlear functions. However, the majority of infants with hearing loss have cochlear deficits.15 Otoacoustic emission testing can be done with the infant awake, feeding, or sucking on a pacifier. AABR requires the infant to be asleep. Currently, it is unknown which screening test or combination of tests is best. A few studies have compared various screening methods, but no consensus has been reached.16,17 Twenty-two states currently mandate some form of newborn universal hearing screening.18
We looked at the feasibility of universal infant hearing screening and whether it meets the criteria for screening tests discussed by Frame and coworkers.19
Disease Recognition
Is there an identifiable disease? Yes. The NIH consensus statement1 identified the risk of hearing loss at 1 per 1000 births. Other studies have found a base rate of from 2 per 1000 to as high as 5 to 9.75 per 1000 in high-risk infants.5-8,20 Without universal screening, the average age at which a deaf child is identified is 2.5 years.1 Typically, these children are tested because of a delay in language and speech skills. The goal of any screening program is to identify and effectively treat a disease in the asymptomatic stage.19 Unfortunately, not all infant hearing loss is identifiable at birth. Approximately 20% to 30% of children develop their deafness in the first few months of life.1 Because of this, health care systems must be vigilant and have a low threshold for repeat screening in older infants. Assuming a hearing loss incidence of 1 per 1000, a birthrate of 4 million per year, and the ability to detect 80% of affected newborns, 3200 infants per year could be diagnosed with hearing loss.
Early Identification
Does early identification and intervention while the infant is asymptomatic improve outcomes? The experts who developed the NIH consensus statement1 stated that early intervention probably does improve outcomes. When this statement was made, however, there had not been adequate studies showing improved outcomes with intervention by the age of 6 months. Since then there have been 2 studies showing significant improvement in expressive and receptive language skills in infants diagnosed and treated before the age of 6 months. Apuzzo and Yoshinaga-Itano11 showed significant improvement in language skills in study of a subset of infants diagnosed before the age of 2 months. Their study included 63 infants, and only those with severe hearing loss (23 infants) had a statistically significant improvement in outcomes. Infants with profound, mild, and mild-to-moderate hearing loss did not have significant improvement. A later study by Yoshinaga and colleagues12 revealed significant improvement in all categories of hearing loss when diagnosed and treated before the age of 6 months. Of the 150 infants in that study, 72 were diagnosed before the age of 6 months, and 78 were diagnosed after the age of 6 months. The early-identified infants had significantly better language skills regardless of the degree of hearing loss, socioeconomic class, and level of cognitive skills. Early diagnosis needs to be coupled with an effective intervention program.1 The intervention should be multidisciplinary and include a physician experienced in otologic disorders, an audiologist with experience in infants and hearing augmentation, a speech and language pathologist, a sign language specialist, and family support services.2
Sensitivity and Specificity
What are the false-negative and false-positive rates of the test? In Rhode Island, universal screening began in 1991 using otoacoustic emission testing. In an article by Vohr and coworkers9 only 5 infants with sensorineural hearing loss had passed the initial screen (false-negatives). Of the 47,257 infants screened there were 106 true-positives, so the sensitivity was 95%. Mehl and Thompson8 reported on the Colorado screening program that used either otoacoustic emissions or AABR. There were no false-negatives, and 41,796 infants were screened. At the other end of the spectrum, Lutman and coworkers21 found a false-negative rate of 20% in high-risk infants screened with otoacoustic testing. The majority of these infants were from the newborn intensive care unit and thus were at higher risk for central nervous system disease (eg, kernicterus, cytomegalovirus) and associated hearing loss. Otoacoustic emissions are normal in infants with hearing loss secondary to central lesions. This may account for some of the false-negative tests in that study.15,22 Infants with known central nervous system disease should probably be evaluated with auditory brain response testing. There has not been a controlled long-term study specifically evaluating the actual false-negative rate in a universal screening program. In the Rhode Island and Colorado groups the false-negative rate appears to be very low. Our study was not designed to determine a false-negative rate. The large number of required subjects (50,000)21 and the length of time required to do a meaningful assessment of the false-negative rate were beyond the scope of our study.
There is a consistently high false-positive rate for infants tested in the newborn nursery—between 3.5% and 10%.5-9,20 This is in part because of matter (vernix) in the newborn ear, the background noise in the nursery, and fluid in the middle ear. The fluid and vernix clear spontaneously during the first few days of life. As a result, most studies repeat the failed newborn screens at the age of 2 to 8 weeks. Infants who fail the second screen are referred for diagnostic auditory brain response testing. The overall failure rate after the second screen was 1% or even less in the referenced studies.59,20 The actual rate of sensorineural hearing loss in these studies is 1 to 2 per 1000. Thus, for every 1000 infants tested, approximately 10 were referred for diagnostic auditory brain testing. The rate of true sensorineural hearing loss was 1 to 2 per 1000. Therefore, 5 to 10 diagnostic auditory brain response tests were done to find each infant with hearing loss. The positive predictive value (PPV) for infants who failed the second screen was 16% in the Rhode Island study and between 5% and 19% in Colorado.8,9 In comparison, the PPV for hypothyroidism screening is 3% and is 80% for phenylketonuria screening.8
Cost-Effectiveness
It appears universal screening will be cost effective, but no studies have been done to prove this. In our study the cost was just higher than $22,000 dollars to find 1 low-risk infant with hearing loss (Tables 2 and 3). Costs in other studies have ranged from $4000 to $17,500 per child identified with hearing loss.6-8 The significant factor when comparing these costs is the actual incidence of hearing loss. Studies with lower costs per identified child had a higher incidence of hearing loss. For example, if our study had found 2 children with sensorineural hearing loss, the cost per child diagnosed would have been just higher than $11,000. This cost of hearing screens can then be compared with the cost of identifying infants with phenylketonuria disease (~$10,000) and hypothyroidism (~$40,000) per child given the diagnosis.8,10 There is a clear cost benefit with phenylketonuria and hyperthyroid screening, but a cost benefit has not been proved with universal hearing screening. The potential for cost savings relies on fewer children requiring special education and fewer adults on long-term disability. The state of South Dakota provides an in-residence school for the deaf at no cost for state residents but charges approximately $15,000 per year for patients who are residents of other states. In-residence schooling in other states can cost as much as $30,000 per year.10 This is approximately 4 times the cost of regular schooling. If only a portion of the early-identified infants attended 12 years of regular school, universal screening would be cost-beneficial. A study comparing schooling costs in early-identified infants versus late-identified infants has not been done.
Is it feasible for rural states to implement a universal hearing screening program? Our study found infants could be successfully screened at a specified time rather than waiting for the infant to be perfectly quiet. Therefore, otoacoustic emission testing lends itself to being taken on the road, and the same equipment and personnel can be used to test infants at a number of hospitals within a 100-mile radius. The software is user friendly, and the cost of otoacoustic emission devices has dropped to approximately $7000. Wyoming began implementation of a universal hearing screening program in 1994, and as of 1998 that program has been fully implemented. In 1998, 95% of all infants born in Wyoming were successfully screened.23 Universal hearing screening can be done, and has been done in rural states and communities.
Conclusions
Testing only high-risk newborn infants results, at best, in early identification of only half of the infants with hearing loss.1 Universal screening is the only effective way to identify the majority of newborn infants with hearing loss. Early identification and treatment has been shown to significantly improve expressive and receptive language skills in infants and children with hearing loss. For this reason, it is clearly worthwhile to pursue universal screening. However, long-term patient outcomes and cost benefits should be studied.
Acknowledgments
We would like to thank the Children’s Miracle Network at Rapid City Regional Hospital, which provided funding of $13,562.50 to purchase the otoacoustic testing equipment.
We also thank Douglas A. Bright, MD, program director, Rapid City Regional Hospital Family Practice Residency, for his guidance and support with our project.
1. National Institutes of Health. Early identification of hearing loss in infants and young children. NIH consensus statement 1993;11:1-24.
2. American Academy of Pediatrics Joint Committee on Infant Hearing 1994 position statement. Pediatrics 1995;95:152-56.
3. Bess FH, Paradise JL. Universal screening for infant hearing impairment: not simple, not risk free, not necessarily beneficial, and not presently justified. Pediatrics 1994;93:330-34.
4. Paradise J. Universal hearing screening: should we leap before we look? Pediatrics 1999;103:670-72.
5. Huynh MT, Pollack RA, Cunningham RA. Universal newborn hearing screening: feasibility in a community hospital. J Fam Pract 1996;42:487-90.
6. Mason JA, Hermann KR. Universal infant hearing screening by automated brainstem response measurement. Pediatrics 1998;101:221-28.
7. Maxon AB, White KR, Behrens TR, Vohr BR. Referral rates and cost efficiency in a universal newborn hearing screening program using transient evoked otoacoustic emissions. J Am Acad Audiol 1995;6:271-77.
8. Mehl AL, Thompson V. Newborn hearing screening: the great omission. Pediatrics 1998;101:E4.-
9. Vohr BR, Carty LM, Moore PE, Letourneau K. The Rhode Island hearing assessment program: experience with statewide hearing screening (1993-1996). J Pediatrics 1998;133:353-57.
10. Johnson JL, et al. Implementing a statewide system of services for infants and toddlers with hearing disabilities. Semin Hearing 1993;14:105-19.
11. Apuzzo ML, Yoshinaga-Itano C. Early identification of infants with significant hearing loss and the Minnesota Child Development Inventory. Semin Hearing 1995;124-39.
12. Yoshinaga-Itano C, Sedey AL, Coulter BA, Mehl AL. Language of early-and later-identified children with hearing loss. Pediatrics 1998;102:1161-71.
13. Cundall J. Universal newborn hearing screening in Tennessee. Tennessee Med 1997;370-71.
14. Osterhammel PA, Rasmussen AN. Distortion product otoacoustic emissions basic properties and clinical aspects. Hearing J 1992;45:38-41.
15. Psarommatis IM, Tsakanikos MD, Kontorgianni AD, Ntouniadakis DE, Apostolopoulos NK. Profound hearing loss and presence of click-evoked otoacoustic emissions in the neonate. Int J Pediatr Otorhinolaryngol 1997;39:237-43.
16. Ochi A, Yasuhara A, Kobayashi Y. Comparison of distortion product otoacoustic emissions with auditory brain-stem response for clinical use in neonatal intensive care unit. electroencephalography and clinical neuropsyciology 1998;108:577-83.
17. Doyle KJ, Burggraaff B, Fujikawa S, Kim J. Newborn hearing screening by otoacoustic emissions and automated auditory brainstem response. Int J Pediatr Otorhinolaryngol 1997;41:111-19.
18. Time. January 1, 2000;117.-
19. Frame PS, Carlson SJ. A critical review of periodic health screening using specific screening criteria part 1: selected diseases of the respiratory, cardiovascular, and central nervous system. J Fam Pract 1975;2:29-36.
20. White KR, Vohr BR, Behrens TR. Universal newborn hearing screening using transient evoked otoacoustic emissions: results of the Rhode Island Hearing Assessment Project. Semin Hearing 1993;14:18-29.
21. Lutman ME, Davis AC, Fortnum HM, Wood S. Field sensitivity of targeted neonatal hearing screening by transient evoked otoacoustic emissions. Ear Hear 1997;18:265-76.
22. Fowler KB, et al. Newborn hearing screening: will children with hearing loss caused by congenital cytomegalovirus infection be missed? J Pediatr 1999;135:60-64.
23. Personal conversation with Nancy Pajka. January 2000.
METHODS: We used DPOAE to screen newborn infants from February 1997 to March 1999.
RESULTS: Of 1002 infants, 111 failed the initial screen (11.1%). When screening was repeated, only 2 infants failed. One infant failed the second screen and a tympanogram. He was treated and he passed a third use of DPOAE. An additional infant failed the repeat screen but passed the tympanogram. That infant was referred on for auditory brain response testing.
CONCLUSIONS: DPOAE testing can be accomplished easily in a normal newborn nursery with an acceptable false-positive rate when a two-stage approach is used. The cost for each test was $19.88. The cost to find the 1 infant with sensory neural hearing loss was $22,114.
In 1993 a consensus statement from the National Institutes of Health (NIH)1 recommended universal newborn hearing screening by the age of 3 months and also stated that otoacoustic emission might be the technology used for screening. These recommendations were based on the following: (1) the incidence of hearing loss is 1 to 6 per 1000; (2) only one half of the infants with hearing loss are discovered with high-risk screening; (3) the current average age at diagnosis of hearing loss is 2.5 years; and (4) early identification and treatment by the age of 6 months will improve outcomes.1 The 1994 position statement from the Joint Committee on Infant Hearing2 reiterated this recommendation.
Bess and Paradise3 and Paradise4 succinctly enumerated the difficulties with universal screening, including high false-positive rates, overall expense, patient acceptance, and feasibility. Since 1993, research and publications5-9 have supported universal screening; now the cost of identifying an infant with hearing loss is less than the cost of identifying an infant with phenylketonuria.10 Also, studies by Appuzzo and Yoshinaga-Itano11 and Yoshinaga-Itano and colleagues12 showed early identification and intervention improves language and social development, and today 37 states have more than 1 hospital providing universal hearing screening programs.13 A variety of screening devices are used, including automated auditory brainstem responses (AABR), transient evoked otoacoustic emission, and distortion product otoacoustic emissions (DPOAE).
Methods
Study Group
Well infants hospitalized from 24 to 72 hours were screened at ages 6 to 72 hours. Data was obtained from a sample of 1002 infants screened at Rapid City Regional Hospital between February 1997 and March 1999. During that period, newborn infants receiving care from family physicians and a pediatrician in the regular nursery were screened. Specific statistics on the age of each infant at screening were not kept.
The screening team consisted of a family practice physician, a registered nurse, and 2 family practice residents. All members of the team performed hearing screening.
We explained the hearing screen procedure, methods, risks, and benefits to the parents of the patient, and obtained informed consent. The follow-up plan for failed hearing screens was also outlined.
Equipment
The GSI 60 computer-based DPOAE instrument (Grason-Stadler, Inc, Milford, NH) was used to perform screenings. It generates paired frequencies F1 (65 dB) and F2 (55 dB). These frequencies travel through the middle ear to the cochlea, where a third tone is generated at the outer hair cell level. Normal cochlear stimulation in this manner produces a DPOAE at a specific frequency predicted by the formula 2F1-F2. During distortion product measurement, the frequency range of 2000 to 4000 Hz was selected for distortion product frequencies.
Procedure
The initial DPOAE measure was obtained on the morning of hospital discharge, with most infants tested between the ages of 12 and 72 hours. Average actual DPOAE screen time was 20 minutes. Passing the test was defined as an emission signal reproducible at 3 frequencies on 2 separate tests. The distortion product was required to be 5 decibels above the noise floor to be considered a pass.
Infants who failed the DPOAE measure were rescreened using DPOAE at 8 weeks. All second-stage screen failures were screened with tympanometry and DPOAE and were referred to their private physician or an ear, nose, and throat specialist. Infants that failed DPOAE and passed tympanometry were to be referred for diagnostic auditory brain response testing (Figure).
Results
Of the 1002 infants screened, 11.1% (111) failed the initial screening (Table 1). Seventy-nine infants were returned to our clinic for retesting 1 to 3 months after their initial evaluation. Two infants (0.2% of the total) failed the second screening. One infant failed the second screen and the tympanogram. He was treated and he passed the third DPOAE. An additional infant failed the repeat screen but passed the tympanogram. She was referred for auditory brain response testing and was found to have mild sensory neural hearing loss. A total of 32 infants were lost to follow-up despite frequent attempts to contact parents by telephone and mail.
Discussion
Recent advances in technology have provided the means to screen newborns and infants for hearing deficits. The 2 most commonly used technologies are AABRs and otoacoustic emissions. Otoacoustic emissions were first described by Kemp in 1978. The cochlear basal membrane vibrates when stimulated by sound, and this vibration sends a retrograde wave back through the cochlear fluid that ultimately vibrates the eardrum, producing a sound wave that can be detected by a microphone at the external ear. DPOAE are produced by stimulating the cochlea with a series of 2 specific frequencies, resulting in a single predictable frequency response.14 Transient evoked otoacoustic emissions, an alternative to DPOAE, are evoked by a click that results in the emission of several frequencies at the same time. AABRs are also used in screening programs, with the advantage of testing both the cochlea and retrocochlear functions. However, the majority of infants with hearing loss have cochlear deficits.15 Otoacoustic emission testing can be done with the infant awake, feeding, or sucking on a pacifier. AABR requires the infant to be asleep. Currently, it is unknown which screening test or combination of tests is best. A few studies have compared various screening methods, but no consensus has been reached.16,17 Twenty-two states currently mandate some form of newborn universal hearing screening.18
We looked at the feasibility of universal infant hearing screening and whether it meets the criteria for screening tests discussed by Frame and coworkers.19
Disease Recognition
Is there an identifiable disease? Yes. The NIH consensus statement1 identified the risk of hearing loss at 1 per 1000 births. Other studies have found a base rate of from 2 per 1000 to as high as 5 to 9.75 per 1000 in high-risk infants.5-8,20 Without universal screening, the average age at which a deaf child is identified is 2.5 years.1 Typically, these children are tested because of a delay in language and speech skills. The goal of any screening program is to identify and effectively treat a disease in the asymptomatic stage.19 Unfortunately, not all infant hearing loss is identifiable at birth. Approximately 20% to 30% of children develop their deafness in the first few months of life.1 Because of this, health care systems must be vigilant and have a low threshold for repeat screening in older infants. Assuming a hearing loss incidence of 1 per 1000, a birthrate of 4 million per year, and the ability to detect 80% of affected newborns, 3200 infants per year could be diagnosed with hearing loss.
Early Identification
Does early identification and intervention while the infant is asymptomatic improve outcomes? The experts who developed the NIH consensus statement1 stated that early intervention probably does improve outcomes. When this statement was made, however, there had not been adequate studies showing improved outcomes with intervention by the age of 6 months. Since then there have been 2 studies showing significant improvement in expressive and receptive language skills in infants diagnosed and treated before the age of 6 months. Apuzzo and Yoshinaga-Itano11 showed significant improvement in language skills in study of a subset of infants diagnosed before the age of 2 months. Their study included 63 infants, and only those with severe hearing loss (23 infants) had a statistically significant improvement in outcomes. Infants with profound, mild, and mild-to-moderate hearing loss did not have significant improvement. A later study by Yoshinaga and colleagues12 revealed significant improvement in all categories of hearing loss when diagnosed and treated before the age of 6 months. Of the 150 infants in that study, 72 were diagnosed before the age of 6 months, and 78 were diagnosed after the age of 6 months. The early-identified infants had significantly better language skills regardless of the degree of hearing loss, socioeconomic class, and level of cognitive skills. Early diagnosis needs to be coupled with an effective intervention program.1 The intervention should be multidisciplinary and include a physician experienced in otologic disorders, an audiologist with experience in infants and hearing augmentation, a speech and language pathologist, a sign language specialist, and family support services.2
Sensitivity and Specificity
What are the false-negative and false-positive rates of the test? In Rhode Island, universal screening began in 1991 using otoacoustic emission testing. In an article by Vohr and coworkers9 only 5 infants with sensorineural hearing loss had passed the initial screen (false-negatives). Of the 47,257 infants screened there were 106 true-positives, so the sensitivity was 95%. Mehl and Thompson8 reported on the Colorado screening program that used either otoacoustic emissions or AABR. There were no false-negatives, and 41,796 infants were screened. At the other end of the spectrum, Lutman and coworkers21 found a false-negative rate of 20% in high-risk infants screened with otoacoustic testing. The majority of these infants were from the newborn intensive care unit and thus were at higher risk for central nervous system disease (eg, kernicterus, cytomegalovirus) and associated hearing loss. Otoacoustic emissions are normal in infants with hearing loss secondary to central lesions. This may account for some of the false-negative tests in that study.15,22 Infants with known central nervous system disease should probably be evaluated with auditory brain response testing. There has not been a controlled long-term study specifically evaluating the actual false-negative rate in a universal screening program. In the Rhode Island and Colorado groups the false-negative rate appears to be very low. Our study was not designed to determine a false-negative rate. The large number of required subjects (50,000)21 and the length of time required to do a meaningful assessment of the false-negative rate were beyond the scope of our study.
There is a consistently high false-positive rate for infants tested in the newborn nursery—between 3.5% and 10%.5-9,20 This is in part because of matter (vernix) in the newborn ear, the background noise in the nursery, and fluid in the middle ear. The fluid and vernix clear spontaneously during the first few days of life. As a result, most studies repeat the failed newborn screens at the age of 2 to 8 weeks. Infants who fail the second screen are referred for diagnostic auditory brain response testing. The overall failure rate after the second screen was 1% or even less in the referenced studies.59,20 The actual rate of sensorineural hearing loss in these studies is 1 to 2 per 1000. Thus, for every 1000 infants tested, approximately 10 were referred for diagnostic auditory brain testing. The rate of true sensorineural hearing loss was 1 to 2 per 1000. Therefore, 5 to 10 diagnostic auditory brain response tests were done to find each infant with hearing loss. The positive predictive value (PPV) for infants who failed the second screen was 16% in the Rhode Island study and between 5% and 19% in Colorado.8,9 In comparison, the PPV for hypothyroidism screening is 3% and is 80% for phenylketonuria screening.8
Cost-Effectiveness
It appears universal screening will be cost effective, but no studies have been done to prove this. In our study the cost was just higher than $22,000 dollars to find 1 low-risk infant with hearing loss (Tables 2 and 3). Costs in other studies have ranged from $4000 to $17,500 per child identified with hearing loss.6-8 The significant factor when comparing these costs is the actual incidence of hearing loss. Studies with lower costs per identified child had a higher incidence of hearing loss. For example, if our study had found 2 children with sensorineural hearing loss, the cost per child diagnosed would have been just higher than $11,000. This cost of hearing screens can then be compared with the cost of identifying infants with phenylketonuria disease (~$10,000) and hypothyroidism (~$40,000) per child given the diagnosis.8,10 There is a clear cost benefit with phenylketonuria and hyperthyroid screening, but a cost benefit has not been proved with universal hearing screening. The potential for cost savings relies on fewer children requiring special education and fewer adults on long-term disability. The state of South Dakota provides an in-residence school for the deaf at no cost for state residents but charges approximately $15,000 per year for patients who are residents of other states. In-residence schooling in other states can cost as much as $30,000 per year.10 This is approximately 4 times the cost of regular schooling. If only a portion of the early-identified infants attended 12 years of regular school, universal screening would be cost-beneficial. A study comparing schooling costs in early-identified infants versus late-identified infants has not been done.
Is it feasible for rural states to implement a universal hearing screening program? Our study found infants could be successfully screened at a specified time rather than waiting for the infant to be perfectly quiet. Therefore, otoacoustic emission testing lends itself to being taken on the road, and the same equipment and personnel can be used to test infants at a number of hospitals within a 100-mile radius. The software is user friendly, and the cost of otoacoustic emission devices has dropped to approximately $7000. Wyoming began implementation of a universal hearing screening program in 1994, and as of 1998 that program has been fully implemented. In 1998, 95% of all infants born in Wyoming were successfully screened.23 Universal hearing screening can be done, and has been done in rural states and communities.
Conclusions
Testing only high-risk newborn infants results, at best, in early identification of only half of the infants with hearing loss.1 Universal screening is the only effective way to identify the majority of newborn infants with hearing loss. Early identification and treatment has been shown to significantly improve expressive and receptive language skills in infants and children with hearing loss. For this reason, it is clearly worthwhile to pursue universal screening. However, long-term patient outcomes and cost benefits should be studied.
Acknowledgments
We would like to thank the Children’s Miracle Network at Rapid City Regional Hospital, which provided funding of $13,562.50 to purchase the otoacoustic testing equipment.
We also thank Douglas A. Bright, MD, program director, Rapid City Regional Hospital Family Practice Residency, for his guidance and support with our project.
METHODS: We used DPOAE to screen newborn infants from February 1997 to March 1999.
RESULTS: Of 1002 infants, 111 failed the initial screen (11.1%). When screening was repeated, only 2 infants failed. One infant failed the second screen and a tympanogram. He was treated and he passed a third use of DPOAE. An additional infant failed the repeat screen but passed the tympanogram. That infant was referred on for auditory brain response testing.
CONCLUSIONS: DPOAE testing can be accomplished easily in a normal newborn nursery with an acceptable false-positive rate when a two-stage approach is used. The cost for each test was $19.88. The cost to find the 1 infant with sensory neural hearing loss was $22,114.
In 1993 a consensus statement from the National Institutes of Health (NIH)1 recommended universal newborn hearing screening by the age of 3 months and also stated that otoacoustic emission might be the technology used for screening. These recommendations were based on the following: (1) the incidence of hearing loss is 1 to 6 per 1000; (2) only one half of the infants with hearing loss are discovered with high-risk screening; (3) the current average age at diagnosis of hearing loss is 2.5 years; and (4) early identification and treatment by the age of 6 months will improve outcomes.1 The 1994 position statement from the Joint Committee on Infant Hearing2 reiterated this recommendation.
Bess and Paradise3 and Paradise4 succinctly enumerated the difficulties with universal screening, including high false-positive rates, overall expense, patient acceptance, and feasibility. Since 1993, research and publications5-9 have supported universal screening; now the cost of identifying an infant with hearing loss is less than the cost of identifying an infant with phenylketonuria.10 Also, studies by Appuzzo and Yoshinaga-Itano11 and Yoshinaga-Itano and colleagues12 showed early identification and intervention improves language and social development, and today 37 states have more than 1 hospital providing universal hearing screening programs.13 A variety of screening devices are used, including automated auditory brainstem responses (AABR), transient evoked otoacoustic emission, and distortion product otoacoustic emissions (DPOAE).
Methods
Study Group
Well infants hospitalized from 24 to 72 hours were screened at ages 6 to 72 hours. Data was obtained from a sample of 1002 infants screened at Rapid City Regional Hospital between February 1997 and March 1999. During that period, newborn infants receiving care from family physicians and a pediatrician in the regular nursery were screened. Specific statistics on the age of each infant at screening were not kept.
The screening team consisted of a family practice physician, a registered nurse, and 2 family practice residents. All members of the team performed hearing screening.
We explained the hearing screen procedure, methods, risks, and benefits to the parents of the patient, and obtained informed consent. The follow-up plan for failed hearing screens was also outlined.
Equipment
The GSI 60 computer-based DPOAE instrument (Grason-Stadler, Inc, Milford, NH) was used to perform screenings. It generates paired frequencies F1 (65 dB) and F2 (55 dB). These frequencies travel through the middle ear to the cochlea, where a third tone is generated at the outer hair cell level. Normal cochlear stimulation in this manner produces a DPOAE at a specific frequency predicted by the formula 2F1-F2. During distortion product measurement, the frequency range of 2000 to 4000 Hz was selected for distortion product frequencies.
Procedure
The initial DPOAE measure was obtained on the morning of hospital discharge, with most infants tested between the ages of 12 and 72 hours. Average actual DPOAE screen time was 20 minutes. Passing the test was defined as an emission signal reproducible at 3 frequencies on 2 separate tests. The distortion product was required to be 5 decibels above the noise floor to be considered a pass.
Infants who failed the DPOAE measure were rescreened using DPOAE at 8 weeks. All second-stage screen failures were screened with tympanometry and DPOAE and were referred to their private physician or an ear, nose, and throat specialist. Infants that failed DPOAE and passed tympanometry were to be referred for diagnostic auditory brain response testing (Figure).
Results
Of the 1002 infants screened, 11.1% (111) failed the initial screening (Table 1). Seventy-nine infants were returned to our clinic for retesting 1 to 3 months after their initial evaluation. Two infants (0.2% of the total) failed the second screening. One infant failed the second screen and the tympanogram. He was treated and he passed the third DPOAE. An additional infant failed the repeat screen but passed the tympanogram. She was referred for auditory brain response testing and was found to have mild sensory neural hearing loss. A total of 32 infants were lost to follow-up despite frequent attempts to contact parents by telephone and mail.
Discussion
Recent advances in technology have provided the means to screen newborns and infants for hearing deficits. The 2 most commonly used technologies are AABRs and otoacoustic emissions. Otoacoustic emissions were first described by Kemp in 1978. The cochlear basal membrane vibrates when stimulated by sound, and this vibration sends a retrograde wave back through the cochlear fluid that ultimately vibrates the eardrum, producing a sound wave that can be detected by a microphone at the external ear. DPOAE are produced by stimulating the cochlea with a series of 2 specific frequencies, resulting in a single predictable frequency response.14 Transient evoked otoacoustic emissions, an alternative to DPOAE, are evoked by a click that results in the emission of several frequencies at the same time. AABRs are also used in screening programs, with the advantage of testing both the cochlea and retrocochlear functions. However, the majority of infants with hearing loss have cochlear deficits.15 Otoacoustic emission testing can be done with the infant awake, feeding, or sucking on a pacifier. AABR requires the infant to be asleep. Currently, it is unknown which screening test or combination of tests is best. A few studies have compared various screening methods, but no consensus has been reached.16,17 Twenty-two states currently mandate some form of newborn universal hearing screening.18
We looked at the feasibility of universal infant hearing screening and whether it meets the criteria for screening tests discussed by Frame and coworkers.19
Disease Recognition
Is there an identifiable disease? Yes. The NIH consensus statement1 identified the risk of hearing loss at 1 per 1000 births. Other studies have found a base rate of from 2 per 1000 to as high as 5 to 9.75 per 1000 in high-risk infants.5-8,20 Without universal screening, the average age at which a deaf child is identified is 2.5 years.1 Typically, these children are tested because of a delay in language and speech skills. The goal of any screening program is to identify and effectively treat a disease in the asymptomatic stage.19 Unfortunately, not all infant hearing loss is identifiable at birth. Approximately 20% to 30% of children develop their deafness in the first few months of life.1 Because of this, health care systems must be vigilant and have a low threshold for repeat screening in older infants. Assuming a hearing loss incidence of 1 per 1000, a birthrate of 4 million per year, and the ability to detect 80% of affected newborns, 3200 infants per year could be diagnosed with hearing loss.
Early Identification
Does early identification and intervention while the infant is asymptomatic improve outcomes? The experts who developed the NIH consensus statement1 stated that early intervention probably does improve outcomes. When this statement was made, however, there had not been adequate studies showing improved outcomes with intervention by the age of 6 months. Since then there have been 2 studies showing significant improvement in expressive and receptive language skills in infants diagnosed and treated before the age of 6 months. Apuzzo and Yoshinaga-Itano11 showed significant improvement in language skills in study of a subset of infants diagnosed before the age of 2 months. Their study included 63 infants, and only those with severe hearing loss (23 infants) had a statistically significant improvement in outcomes. Infants with profound, mild, and mild-to-moderate hearing loss did not have significant improvement. A later study by Yoshinaga and colleagues12 revealed significant improvement in all categories of hearing loss when diagnosed and treated before the age of 6 months. Of the 150 infants in that study, 72 were diagnosed before the age of 6 months, and 78 were diagnosed after the age of 6 months. The early-identified infants had significantly better language skills regardless of the degree of hearing loss, socioeconomic class, and level of cognitive skills. Early diagnosis needs to be coupled with an effective intervention program.1 The intervention should be multidisciplinary and include a physician experienced in otologic disorders, an audiologist with experience in infants and hearing augmentation, a speech and language pathologist, a sign language specialist, and family support services.2
Sensitivity and Specificity
What are the false-negative and false-positive rates of the test? In Rhode Island, universal screening began in 1991 using otoacoustic emission testing. In an article by Vohr and coworkers9 only 5 infants with sensorineural hearing loss had passed the initial screen (false-negatives). Of the 47,257 infants screened there were 106 true-positives, so the sensitivity was 95%. Mehl and Thompson8 reported on the Colorado screening program that used either otoacoustic emissions or AABR. There were no false-negatives, and 41,796 infants were screened. At the other end of the spectrum, Lutman and coworkers21 found a false-negative rate of 20% in high-risk infants screened with otoacoustic testing. The majority of these infants were from the newborn intensive care unit and thus were at higher risk for central nervous system disease (eg, kernicterus, cytomegalovirus) and associated hearing loss. Otoacoustic emissions are normal in infants with hearing loss secondary to central lesions. This may account for some of the false-negative tests in that study.15,22 Infants with known central nervous system disease should probably be evaluated with auditory brain response testing. There has not been a controlled long-term study specifically evaluating the actual false-negative rate in a universal screening program. In the Rhode Island and Colorado groups the false-negative rate appears to be very low. Our study was not designed to determine a false-negative rate. The large number of required subjects (50,000)21 and the length of time required to do a meaningful assessment of the false-negative rate were beyond the scope of our study.
There is a consistently high false-positive rate for infants tested in the newborn nursery—between 3.5% and 10%.5-9,20 This is in part because of matter (vernix) in the newborn ear, the background noise in the nursery, and fluid in the middle ear. The fluid and vernix clear spontaneously during the first few days of life. As a result, most studies repeat the failed newborn screens at the age of 2 to 8 weeks. Infants who fail the second screen are referred for diagnostic auditory brain response testing. The overall failure rate after the second screen was 1% or even less in the referenced studies.59,20 The actual rate of sensorineural hearing loss in these studies is 1 to 2 per 1000. Thus, for every 1000 infants tested, approximately 10 were referred for diagnostic auditory brain testing. The rate of true sensorineural hearing loss was 1 to 2 per 1000. Therefore, 5 to 10 diagnostic auditory brain response tests were done to find each infant with hearing loss. The positive predictive value (PPV) for infants who failed the second screen was 16% in the Rhode Island study and between 5% and 19% in Colorado.8,9 In comparison, the PPV for hypothyroidism screening is 3% and is 80% for phenylketonuria screening.8
Cost-Effectiveness
It appears universal screening will be cost effective, but no studies have been done to prove this. In our study the cost was just higher than $22,000 dollars to find 1 low-risk infant with hearing loss (Tables 2 and 3). Costs in other studies have ranged from $4000 to $17,500 per child identified with hearing loss.6-8 The significant factor when comparing these costs is the actual incidence of hearing loss. Studies with lower costs per identified child had a higher incidence of hearing loss. For example, if our study had found 2 children with sensorineural hearing loss, the cost per child diagnosed would have been just higher than $11,000. This cost of hearing screens can then be compared with the cost of identifying infants with phenylketonuria disease (~$10,000) and hypothyroidism (~$40,000) per child given the diagnosis.8,10 There is a clear cost benefit with phenylketonuria and hyperthyroid screening, but a cost benefit has not been proved with universal hearing screening. The potential for cost savings relies on fewer children requiring special education and fewer adults on long-term disability. The state of South Dakota provides an in-residence school for the deaf at no cost for state residents but charges approximately $15,000 per year for patients who are residents of other states. In-residence schooling in other states can cost as much as $30,000 per year.10 This is approximately 4 times the cost of regular schooling. If only a portion of the early-identified infants attended 12 years of regular school, universal screening would be cost-beneficial. A study comparing schooling costs in early-identified infants versus late-identified infants has not been done.
Is it feasible for rural states to implement a universal hearing screening program? Our study found infants could be successfully screened at a specified time rather than waiting for the infant to be perfectly quiet. Therefore, otoacoustic emission testing lends itself to being taken on the road, and the same equipment and personnel can be used to test infants at a number of hospitals within a 100-mile radius. The software is user friendly, and the cost of otoacoustic emission devices has dropped to approximately $7000. Wyoming began implementation of a universal hearing screening program in 1994, and as of 1998 that program has been fully implemented. In 1998, 95% of all infants born in Wyoming were successfully screened.23 Universal hearing screening can be done, and has been done in rural states and communities.
Conclusions
Testing only high-risk newborn infants results, at best, in early identification of only half of the infants with hearing loss.1 Universal screening is the only effective way to identify the majority of newborn infants with hearing loss. Early identification and treatment has been shown to significantly improve expressive and receptive language skills in infants and children with hearing loss. For this reason, it is clearly worthwhile to pursue universal screening. However, long-term patient outcomes and cost benefits should be studied.
Acknowledgments
We would like to thank the Children’s Miracle Network at Rapid City Regional Hospital, which provided funding of $13,562.50 to purchase the otoacoustic testing equipment.
We also thank Douglas A. Bright, MD, program director, Rapid City Regional Hospital Family Practice Residency, for his guidance and support with our project.
1. National Institutes of Health. Early identification of hearing loss in infants and young children. NIH consensus statement 1993;11:1-24.
2. American Academy of Pediatrics Joint Committee on Infant Hearing 1994 position statement. Pediatrics 1995;95:152-56.
3. Bess FH, Paradise JL. Universal screening for infant hearing impairment: not simple, not risk free, not necessarily beneficial, and not presently justified. Pediatrics 1994;93:330-34.
4. Paradise J. Universal hearing screening: should we leap before we look? Pediatrics 1999;103:670-72.
5. Huynh MT, Pollack RA, Cunningham RA. Universal newborn hearing screening: feasibility in a community hospital. J Fam Pract 1996;42:487-90.
6. Mason JA, Hermann KR. Universal infant hearing screening by automated brainstem response measurement. Pediatrics 1998;101:221-28.
7. Maxon AB, White KR, Behrens TR, Vohr BR. Referral rates and cost efficiency in a universal newborn hearing screening program using transient evoked otoacoustic emissions. J Am Acad Audiol 1995;6:271-77.
8. Mehl AL, Thompson V. Newborn hearing screening: the great omission. Pediatrics 1998;101:E4.-
9. Vohr BR, Carty LM, Moore PE, Letourneau K. The Rhode Island hearing assessment program: experience with statewide hearing screening (1993-1996). J Pediatrics 1998;133:353-57.
10. Johnson JL, et al. Implementing a statewide system of services for infants and toddlers with hearing disabilities. Semin Hearing 1993;14:105-19.
11. Apuzzo ML, Yoshinaga-Itano C. Early identification of infants with significant hearing loss and the Minnesota Child Development Inventory. Semin Hearing 1995;124-39.
12. Yoshinaga-Itano C, Sedey AL, Coulter BA, Mehl AL. Language of early-and later-identified children with hearing loss. Pediatrics 1998;102:1161-71.
13. Cundall J. Universal newborn hearing screening in Tennessee. Tennessee Med 1997;370-71.
14. Osterhammel PA, Rasmussen AN. Distortion product otoacoustic emissions basic properties and clinical aspects. Hearing J 1992;45:38-41.
15. Psarommatis IM, Tsakanikos MD, Kontorgianni AD, Ntouniadakis DE, Apostolopoulos NK. Profound hearing loss and presence of click-evoked otoacoustic emissions in the neonate. Int J Pediatr Otorhinolaryngol 1997;39:237-43.
16. Ochi A, Yasuhara A, Kobayashi Y. Comparison of distortion product otoacoustic emissions with auditory brain-stem response for clinical use in neonatal intensive care unit. electroencephalography and clinical neuropsyciology 1998;108:577-83.
17. Doyle KJ, Burggraaff B, Fujikawa S, Kim J. Newborn hearing screening by otoacoustic emissions and automated auditory brainstem response. Int J Pediatr Otorhinolaryngol 1997;41:111-19.
18. Time. January 1, 2000;117.-
19. Frame PS, Carlson SJ. A critical review of periodic health screening using specific screening criteria part 1: selected diseases of the respiratory, cardiovascular, and central nervous system. J Fam Pract 1975;2:29-36.
20. White KR, Vohr BR, Behrens TR. Universal newborn hearing screening using transient evoked otoacoustic emissions: results of the Rhode Island Hearing Assessment Project. Semin Hearing 1993;14:18-29.
21. Lutman ME, Davis AC, Fortnum HM, Wood S. Field sensitivity of targeted neonatal hearing screening by transient evoked otoacoustic emissions. Ear Hear 1997;18:265-76.
22. Fowler KB, et al. Newborn hearing screening: will children with hearing loss caused by congenital cytomegalovirus infection be missed? J Pediatr 1999;135:60-64.
23. Personal conversation with Nancy Pajka. January 2000.
1. National Institutes of Health. Early identification of hearing loss in infants and young children. NIH consensus statement 1993;11:1-24.
2. American Academy of Pediatrics Joint Committee on Infant Hearing 1994 position statement. Pediatrics 1995;95:152-56.
3. Bess FH, Paradise JL. Universal screening for infant hearing impairment: not simple, not risk free, not necessarily beneficial, and not presently justified. Pediatrics 1994;93:330-34.
4. Paradise J. Universal hearing screening: should we leap before we look? Pediatrics 1999;103:670-72.
5. Huynh MT, Pollack RA, Cunningham RA. Universal newborn hearing screening: feasibility in a community hospital. J Fam Pract 1996;42:487-90.
6. Mason JA, Hermann KR. Universal infant hearing screening by automated brainstem response measurement. Pediatrics 1998;101:221-28.
7. Maxon AB, White KR, Behrens TR, Vohr BR. Referral rates and cost efficiency in a universal newborn hearing screening program using transient evoked otoacoustic emissions. J Am Acad Audiol 1995;6:271-77.
8. Mehl AL, Thompson V. Newborn hearing screening: the great omission. Pediatrics 1998;101:E4.-
9. Vohr BR, Carty LM, Moore PE, Letourneau K. The Rhode Island hearing assessment program: experience with statewide hearing screening (1993-1996). J Pediatrics 1998;133:353-57.
10. Johnson JL, et al. Implementing a statewide system of services for infants and toddlers with hearing disabilities. Semin Hearing 1993;14:105-19.
11. Apuzzo ML, Yoshinaga-Itano C. Early identification of infants with significant hearing loss and the Minnesota Child Development Inventory. Semin Hearing 1995;124-39.
12. Yoshinaga-Itano C, Sedey AL, Coulter BA, Mehl AL. Language of early-and later-identified children with hearing loss. Pediatrics 1998;102:1161-71.
13. Cundall J. Universal newborn hearing screening in Tennessee. Tennessee Med 1997;370-71.
14. Osterhammel PA, Rasmussen AN. Distortion product otoacoustic emissions basic properties and clinical aspects. Hearing J 1992;45:38-41.
15. Psarommatis IM, Tsakanikos MD, Kontorgianni AD, Ntouniadakis DE, Apostolopoulos NK. Profound hearing loss and presence of click-evoked otoacoustic emissions in the neonate. Int J Pediatr Otorhinolaryngol 1997;39:237-43.
16. Ochi A, Yasuhara A, Kobayashi Y. Comparison of distortion product otoacoustic emissions with auditory brain-stem response for clinical use in neonatal intensive care unit. electroencephalography and clinical neuropsyciology 1998;108:577-83.
17. Doyle KJ, Burggraaff B, Fujikawa S, Kim J. Newborn hearing screening by otoacoustic emissions and automated auditory brainstem response. Int J Pediatr Otorhinolaryngol 1997;41:111-19.
18. Time. January 1, 2000;117.-
19. Frame PS, Carlson SJ. A critical review of periodic health screening using specific screening criteria part 1: selected diseases of the respiratory, cardiovascular, and central nervous system. J Fam Pract 1975;2:29-36.
20. White KR, Vohr BR, Behrens TR. Universal newborn hearing screening using transient evoked otoacoustic emissions: results of the Rhode Island Hearing Assessment Project. Semin Hearing 1993;14:18-29.
21. Lutman ME, Davis AC, Fortnum HM, Wood S. Field sensitivity of targeted neonatal hearing screening by transient evoked otoacoustic emissions. Ear Hear 1997;18:265-76.
22. Fowler KB, et al. Newborn hearing screening: will children with hearing loss caused by congenital cytomegalovirus infection be missed? J Pediatr 1999;135:60-64.
23. Personal conversation with Nancy Pajka. January 2000.
The Efficacy of Liquid-Based Cervical Cytology Using Direct-to-Vial Sample Collection
METHODS: A total of 1004 nonpregnant women aged 18 years or older with an intact cervix had Pap tests collected with an Ayre spatula and cytobrush, and the sample was placed in a preservative solution. The specimens were processed as thin layer Pap tests according to the manufacturer’s specifications. Another group of 2110 women with a similar patient profile had a CPT collected immediately preceding the initiation of the trial. The subjects in each group consisted of an equal percentage of women presenting for a routine Pap test or a colposcopy examination. We compared the distributions of diagnostic categories between the groups using a chi-square test.
RESULTS: A significantly greater percentage of satisfactory Pap tests were obtained using LBCC (84.0%) compared with the CPT (60.5%, P <.001). Fewer satisfactory but limited by (SBLB, 14.8%) and unsatisfactory (1.2%) Pap tests were reported using LBCC compared with the CPT (35.7% and 3.8%, respectively, c2=170.7, P <.001). A significantly greater percentage of low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL) Pap test results were reported using LBCC (7.4% and 3.7%, respectively) compared with the CPT (1.7% and 1.7%, respectively, c2=74.4, P & .001). The predictive value of a positive LBCC test (93.9%) was similar to that for a positive CPT (87.8%) when compared with histology results.
CONCLUSIONS: Compared with the CPT, LBCC detected a significantly greater percentage of satisfactory Pap tests and significantly reduced the number of unsatisfactory and SBLB tests. Four times the percentage of LSIL and twice the percentage of HSIL Pap test results were obtained using LBCC compared with the CPT. These findings demonstrate that LBCC significantly improves the adequacy of Pap tests and may increase the rate of detection of cervical neoplasia compared with the CPT.
The Papanicolaou (Pap) test is one of the most common and effective cancer screening tests used by primary care clinicians. However, a recent study conducted through the Agency for Health Care Policy and Research1 estimated that the true sensitivity of the Pap test is only 51%. Two thirds of the false-negative results occur because of sampling error. The first liquid-based cervical cytology (LBCC) test was approved for clinical use by the United States Food and Drug Administration in 1996.2 With LBCC, standard cervical cytology sampling devices are used to obtain a sample from the cervix. The sampling devices are then rinsed in a buffered alcohol transport and preservative solution instead of immediately transferring the cells to a glass slide. Sampling error is reduced, since all the cells are transferred into solution, in contrast with the conventional Pap test (CPT), in which only 20% of the cells are actually transferred to the glass slide.3 The cytologic specimen in solution is then sent to the laboratory where most of the debris, leukocytes, and erythrocytes are filtered from the specimen and a thin layer Pap test is prepared using a special processor. The cells are dispersed evenly with relatively little clumping or overlapping or obscuring debris (Figures 1A and B). The resulting clean smear is circular and contains approximately half the number of representative cells of a CPT. Residual cervical cells in solution that are not initially used for the Pap test may be tested at a later date for lower genital tract pathogens, such as carcinogenic human papillomavirus.4,5
Liquid-based thin layer cervical cytology has performed favorably in studies using split sample collection techniques.6-9 With the split sample technique, the cytologic specimen is first transferred to a glass slide for the CPT before transferring the remaining cervical cells to the liquid transport fluid. Evaluation of the LBCC Pap test using this nonrandom study design may be biased, since it is based solely on cells normally discarded. The CPT slides receive the first and perhaps better cellular sample; thus, the quality of that specimen may be vastly different than that available for LBCC. Currently, there are minimal data available in which LBCC was evaluated as it was intended to be used, for a primary cytologic specimen.10,11 Thus, the full diagnostic potential of this new Pap test technique has not been fully realized. The purpose of our trial was to determine the rates of specimen adequacy and cervical cytologic and histologically confirmed diagnoses obtained with LBCC using a direct-to-vial collection technique and compare these results with those obtained using the CPT.
Methods
Subjects
In our clinical study one group of women was enrolled prospectively, and a second group was considered retrospectively. The first group (LBCC) consisted of 1004 nonpregnant women aged 18 years or older with an intact cervix. The second group consisted of 2110 women with a similar patient profile who had a CPT collected at the same clinics during the 2 years immediately preceding the initiation of the trial.
To ensure an enriched sample of women with cervical neoplasia, approximately 20% of the subjects in both cohorts were included from women presenting consecutively for a colposcopy examination following an abnormal CPT result. The remaining 80% of the subjects represented a general screening population of women presenting for a routine Pap test. All subjects were seen at 1 of 3 clinics: the Family Medicine Center, the Obstetrics and Gynecology Clinic, or the Comprehensive Cancer Center at the Medical College of Georgia in Augusta.
Equipment and Materials
Samples for CPTs and LBCC (ThinPrep Pap Test, Cytyc Corporation, Boxborough, Mass) were collected using a plastic Ayre spatula and cytobrush (Medscand, North Hollywood, Fla). Cervical cells from the second group were processed by CPT methods. Cervical cells from the prospectively enrolled group were preserved and transferred in LBCC solution (PreservCyt, Cytyc Corporation). Cervical samples for the LBCC test were processed using the ThinPrep 2000 Processor (Cytyc Corporation).12
Study Design
Eligible prospectively enrolled women presenting consecutively for a routine Pap test or colposcopy examination at any of the 3 designated clinics signed an internal review board-approved informed consent form before participating. After properly visualizing the cervix, a plastic Ayre spatula was used to collect an ectocervical sample. Another sample was collected using a cytobrush rotated 90° to 180° within the endocervical canal. The spatula and cytobrush were then immediately rinsed vigorously in the LBCC solution. The container was sealed, labeled, and sent to the cytology laboratory for processing according to the manufacturer’s instructions.12 Cytotechnicians or pathologists in the Department of Pathology, Medical College of Georgia (MCG), evaluated all cytology and histology specimens. Cervical cytology was reported using the Bethesda System.
Computerized records from the MCG Cytology Laboratory were used to establish the second group. CPT records with similar patient profiles and clinic affiliation were selected chronologically beginning the day before the initiation of the trial and working backward in time until 2110 records were obtained (approximately 2 years). Cervical specimens were collected previously from this group using the same types of collection devices and technique as that for the LBCC group. However, immediately following sample collection these endocervical and ectocervical cells were plated on a glass slide, fixed, and sent to the cytology laboratory for Pap staining and diagnosis.
Statistical Analysis
The proportion of subjects in each of the Pap test adequacy groups (satisfactory, unsatisfactory, and satisfactory but limited by [SBLB]) were compared between the sample and control groups for the total study group, the colposcopy and general screening cohorts, and the 3 age groups (18-30 years, 31-40 years, 41 years or older). The chi-square statistic was used to compare the proportions unless the tables included cells with expected counts less than 5; for those cases we used the Fisher exact test as extended to tables larger than 2×2. A summary comparison controlling for age group was made using the Cohran-Mantel-Haenszel chi-square statistic. Average age was compared between the groups using the Student t test. At the given sample size the power was calculated to be 80% or greater for detecting a difference in positive rates of 3.5% between the LBCC and the CPT.
Results
Subjects in both groups were comparable by type of clinical visit. Consequently, 1004 LBCC subjects and 2110 CPT subjects were included in the data analysis. The routine Pap test screening group consisted of 796 and 1585 subjects in the LBCC and CPT groups, respectively, representing 79.0% of each population. Approximately 21.0% of the LBCC and CPT subjects (208 and 425, respectively) were represented by women presenting for colposcopy examination following an abnormal Pap test result. Subjects were also stratified into 1 of 3 age ranges for analytic purposes. Thus, 423 and 602 subjects were aged between 18 and 30 years; 287 and 523 subjects were aged 31 to 40 years; and 294 and 885 subjects were aged 41 years or older for the LBCC and CPT populations, respectively. A statistically significant difference between the age distributions was noted (P <.001) with a greater proportion of younger subjects in the LBCC group (mean age=35.0 years) than in the CPT cohort (mean age=39.8 years).
A greater percentage of satisfactory Pap tests were obtained when using LBCC (84.0%) compared with the CPT (60.5%), P ".001 (Figure 1). Also, significantly fewer unsatisfactory and SBLB Pap tests were reported using LBCC (1.2% and 14.8%) compared with the CPT (3.8% and 35.7%). For the colposcopy group, significantly more satisfactory Pap tests were reported for LBCC compared with the CPT (81.7% and 55.3%, respectively; P <.001). Similar results were also noted for the screening group. These significant results were maintained when the data were adjusted for the 3 age groups (Table 1).
The frequencies of cytologic diagnoses for LBCC and the CPT are seen in Table 2. The LBCC method detected a significantly greater percentage of women with low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL) Pap tests results (7.4% and 3.7%, respectively) compared with the CPT (1.7% and 1.7%, respectively). Similar significant results were noted when the LBCC and CPT groups were stratified into the screening and colposcopy populations. When both groups were adjusted for age, a greater percentage of LSIL and HSIL diagnoses were noted for LBCC in all 3 age ranges compared with the CPT (Table 4). The classification of atypical squamous cells of undetermined significance (ASCUS) cases was not modified by LBCC compared with the CPT. Because of the limited number of squamous cell carcinoma diagnoses reported, a comparison between the 2 groups was not possible.
The frequency of histologic diagnoses in the 2 groups closely paralleled Pap test diagnoses in the same populations. The LBCC and CPT groups had 5.4% and 1.0% cervical intraepithelial neoplasia 1 (CIN 1) histologic diagnoses, respectively, while 5.9% and 1.7% of the women had histologic diagnoses of CIN 2 and 3, respectively. These results were also maintained consistently across all 3 age ranges. The predictive value of a positive LBCC test (93.9%) was similar to that for the predictive value of a positive CPT (87.8%) when based on available histology results (Table 5). The sensitivity and specificity were equivalent for each test. Of all women with an LSIL or more severe Pap test result in each group, 1.9% of women in the CPT group and 6.5% of women in the LBCC group had cervical biopsies indicating CIN 1 or more severe disease.
Discussion
Pap test specimen adequacy results reflect whether the most likely site for neoplasia of the cervix—the transformation zone—has been sampled properly. LBCC provided significantly greater rates of satisfactory specimen adequacy than the CPT. Also, compared with the CPT, LBCC effectively reduced the number of SBLB and unsatisfactory specimen adequacy reports. Our findings are consistent with those of studies published previously that used split sample collection techniques.6,7 The improvement of Pap test adequacy using LBCC may be because a clearer inspection of pertinent normal and abnormal cells is achieved by eliminating or minimizing obscuring inflammatory cells, red blood cells, other debris, and clumped or overlapped cells.13 A significant decrease in unsatisfactory Pap tests reduces the extra cost, time, and patient and clinician inconvenience incurred from the necessity of repeating a nonrepresentative Pap test. These advantages may be of particular importance to patients, clinicians, and third-party payers.
Also, in regard to specimen adequacy, LBCC performed better than the CPT for women of all age ranges. This is especially important for older women who typically have a greater frequency of unsatisfactory Pap tests because their active transformation zones are positioned deeply within the endocervical canal and are often inaccessible to comprehensive sampling. In addition, a thin atrophic epithelium (commonly seen in the older age group) is more easily traumatized during sampling, which causes potentially obscuring bleeding. Yet, LBCC reduced the rate of unsatisfactory Pap tests by approximately 50% for women older than 40 years. Other than the advantage of filtering unwanted cells, the difference may be explained by the fact that 80% of cervical cells are discarded with the collection devices and not transferred to a glass slide using the CPT method.3 In contrast, nearly all cells are transferred to solution with LBCC. Thus, the larger specimen obtained with the sampling method used in LBCC increases the availability of a typically limited number of endocervical cells retrieved from post-treatment and postmenopausal women who are routinely more difficult to sample.
The ability of the Pap test to detect cytologic changes consistent with neoplasia, when present, is critically important. In our trial, the LBCC method detected a greater percentage of women with LSIL and HSIL compared with the CPT. Approximately 4 times the number of women with LSIL and twice the number of cases of HSIL were detected by LBCC compared with the CPT. These results replicate or surpass those found in studies that evaluated LBCC using a split sample method.6-9 Twice the rate of neoplasia detection by LBCC has been reported in studies using the split sample method. Our results, based on a direct-to-vial study design, may portray the true potential for LBCC to detect cervical neoplasia. A cleaner monolayer Pap test with a more comprehensive cellular specimen likely accounts for these remarkable differences. Similarly, the reduced number of unsatisfactory and SBLB Pap tests using LBCC may contribute to the greater yield of SILs. Of note, LBCC did not detect a significantly greater rate of ASCUS Pap test results; this may create problematic management decisions for some clinicians.14 The ability of LBCC to detect a greater percentage of women with SILs was maintained for the general screening population, the colposcopy cohort, and across the 3 age ranges. LBCC may be better able to detect women with HSIL, a lesion unlikely to resolve spontaneously. Women with HSIL (Figures 2A and B) and a comparable histologic diagnosis deserve prompt therapy for this true cancer precursor. Although many women with LSIL have disease that may regress to normal, as many as 20% to 40% of these women will have histologically confirmed CIN 2 or 3 that deserves treatment as well.15
Limitations
Our study is subject to several limitations. Most important, the observed differences in Pap test performance may have been because of actual differences of cervical neoplasia prevalences between the 2 groups, rather than a difference between Pap tests. Our subjects were selected from comparable patient populations by a study design that other researchers10,11,16-20 have used to evaluate the direct-to-vial technique. Other than sampling a population of women with LBCC and then having them return for a CPT 1 month later, which risks compliance failure and spectrum of cervical disease changes or acquisition of new disease, there is no other study design to evaluate a direct-to-vial technique. One may question whether our increased rate of neoplasia detection demonstrated by LBCC is due to a substantial number of LBCC false-positive results.21 Yet, because the positive predictive values for the LBCC and CPT were similar and there were 3 times as many women with biopsy results of CIN 1 or greater in the LBCC group compared with the CPT group when only women with LSIL or greater Pap test results were considered, the increased rates of SIL detection using LBCC appear to represent an increased detection of true disease and not false-positive results, further validating our findings.21 The equivalent results for LBCC and CPT specificity (99%) indicate that the increased rate of SIL detection by LBCC was not due to a reduced specificity. Also, the favorable test results demonstrated for LBCC are not entirely dissimilar from those of other studies that used a split sample technique, a more biased study design for LBCC.6-9 Studies by Vassilakos and colleagues16 and Papillo and coworkers10 demonstrated that a liquid-based thin layer cytologic report of LSIL or more severe disease correlated better with histologic results of CIN 1 or greater (80.5% and 80.2%, respectively) than CPTs (71.7% and 72.2%, respectively). Finally, because of the limited number of women with cervical caner in this study population, we were unable to determine if there were any differences between LBCC and the CPT in their ability to detect cervical cancer.
Conclusions
Compared with the CPT, LBCC detected a significantly greater percentage of satisfactory Pap tests and significantly reduced the number of unsatisfactory and SBLB tests. These findings demonstrate that LBCC significantly improves the adequacy of Pap tests and may increase the rate of detection of cervical neoplasia compared with CPT. Further study is necessary and warranted, since failure to detect cancer in a timely fashion affects ultimate cure rates, medical costs, quality of life, and perhaps medicolegal expenses. Although liquid-based thin layer cervical cytology is rapidly replacing the glass slide method throughout the United States, additional studies are also necessary to determine whether LBCC reduces the incidence of cervical cancer.
Acknowledgments
We appreciate the assistance of Cytyc Corporation in providing the supplies and grant support necessary for completion of this study.
We would like to thank the faculty and residents in the Family Medicine Center and OB/GYN Clinic who assisted with specimen collection, Jim Best for processing and interpreting cytologic specimens, Aisha Lavin for data management assistance, and April Dean for manuscript preparation.
1. Agency for Health Care Policy and Research. Evidence report/technology assessment no. 5: evaluation of cervical cytology. Rockville, Md: Agency for Health Care Policy and Research; 1999. US Department of Health and Human Services AHCPR publication no. 99-E010.
2. Lee KR, Ashfaq R, Birdsong GG, Corkill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and a fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol 1997;90:278-84.
3. Hutchinson ML, Agarwal P, Denault T, Berger B, Cibas ES. A new look at cervical cytology: ThinPrep multicenter trial results. Acta Cytol 1992;36:499-504.
4. Ferris DG, Wright TC, Jr, Litaker MS, et al. Comparison of two tests for detecting carcinogenic HPV in women with Papanicolaou smear reports of ASCUS and LSIL. J Fam Pract 1998;46:136-41.
5. Ferenczy A, Franco E, Arseneau J, Wright TC, Richart RM. Diagnostic performance of hybrid capture human papillomavirus deoxyribonucleic acid assay combined with liquid-based cytologic study. Am J Obstet Gynecol 1996;175:651-56.
6. Roberts JM, Gurley AM, Thurloe JK, Bowditch R, Laverty CRA. Evaluation of the ThinPrep Pap test as an adjunct to the conventional Pap smear. Med J Aust 1997;167:466-69.
7. Hutchinson ML, Isenstein LM, Goodman A, et al. Homogeneous sampling accounts for the increased diagnostic accuracy using the ThinPrep(tm) Processor. Anat Path 1994;101:215-19.
8. Corkhill M, Knapp D, Hutchinson ML. Improved accuracy for cervical cytology with the ThinPrep method and the endocervical brush-spatula collection procedure. J Lower Genital Tract Dis 1998;2:12-16.
9. KR, Ashfaq R, Birdsong GG, Corkhill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and a fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol 1997;90:278-84.
10. Papillo JL, Zarka MA, St John TL. Evaluation of the ThinPrep Pap test in clinical practice: a seven-month, 16,314-case experience in northern Vermont. Acta Cytol 1998;42:203-08.
11. Diaz-Rosario LA, Kabawat SE. Performance of a fluid-based, thin-layer Papanicolaou smear method in the clinical setting of an independent laboratory and an outpatient screening population in New England. Arch Pathol Lab Med 1999;123:817-21.
12. Bolick DR, Hellman DJ. Laboratory implementation and efficacy assessment of the ThinPrep cervical cancer screening system. Acta Cytol 1998;42:209-13.
13. Linder J, Zahniser D. The ThinPrep Pap test: a review of clinical studies. Acta Cytol 1997;41:30-38.
14. Ferris DG, Wright TC, Litaker MS, et al. Triage of women with ASCUS and LSIL on Pap smear reports: management by repeat Pap smear, HPV DNA testing, or colposcopy? J Fam Pract 1998;46:125-34.
15. Kurman RJ, Henson DE, Herbst AL, Noller KL, Schiffmann MH. Interim guidelines for management of abnormal cervical cytology. JAMA 1994;271:1866-69.
16. Vassilakos P, Schwartz D, de Marval F, et al. Biopsy-based comparison of liquid-based, thin-layer preparations to conventional Pap smears. J Reprod Med 2000;45:11-16.
17. Vassilakos P, Griffin S, Megevand E, Campara A. CytoRich liquid-based cervical cytologic test screening results in a routine cytopathology service. Acta Cytol 1998;42:198-202.
18. Vassilakos P, Saurel J, Rondez R. Direct to vial use of the AutoCyte PREP liquid-based preparation for cervical-vaginal specimens in three European laboratories. Acta Cytol 1999;43:65-68.
19. Ashfaq R, Gibbons D, Vela C, Saboorian MH, Iliya F. Thin Prep Pap test accuracy for glandular disease. Acta Cytol 1999;43:81-85.
20. Weintraub J, Morabia A. Efficacy of a liquid-based thin layer method for cervical cancer screening in a population with a low incidence of cervical cancer. Diagn Cytopathol 2000;22:52-59.
21. Sawaya GF, Grimes DA. New technologies in cervical cytology screening: a word of caution. Obstet Gynecol 1999;94:307-10.
METHODS: A total of 1004 nonpregnant women aged 18 years or older with an intact cervix had Pap tests collected with an Ayre spatula and cytobrush, and the sample was placed in a preservative solution. The specimens were processed as thin layer Pap tests according to the manufacturer’s specifications. Another group of 2110 women with a similar patient profile had a CPT collected immediately preceding the initiation of the trial. The subjects in each group consisted of an equal percentage of women presenting for a routine Pap test or a colposcopy examination. We compared the distributions of diagnostic categories between the groups using a chi-square test.
RESULTS: A significantly greater percentage of satisfactory Pap tests were obtained using LBCC (84.0%) compared with the CPT (60.5%, P <.001). Fewer satisfactory but limited by (SBLB, 14.8%) and unsatisfactory (1.2%) Pap tests were reported using LBCC compared with the CPT (35.7% and 3.8%, respectively, c2=170.7, P <.001). A significantly greater percentage of low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL) Pap test results were reported using LBCC (7.4% and 3.7%, respectively) compared with the CPT (1.7% and 1.7%, respectively, c2=74.4, P & .001). The predictive value of a positive LBCC test (93.9%) was similar to that for a positive CPT (87.8%) when compared with histology results.
CONCLUSIONS: Compared with the CPT, LBCC detected a significantly greater percentage of satisfactory Pap tests and significantly reduced the number of unsatisfactory and SBLB tests. Four times the percentage of LSIL and twice the percentage of HSIL Pap test results were obtained using LBCC compared with the CPT. These findings demonstrate that LBCC significantly improves the adequacy of Pap tests and may increase the rate of detection of cervical neoplasia compared with the CPT.
The Papanicolaou (Pap) test is one of the most common and effective cancer screening tests used by primary care clinicians. However, a recent study conducted through the Agency for Health Care Policy and Research1 estimated that the true sensitivity of the Pap test is only 51%. Two thirds of the false-negative results occur because of sampling error. The first liquid-based cervical cytology (LBCC) test was approved for clinical use by the United States Food and Drug Administration in 1996.2 With LBCC, standard cervical cytology sampling devices are used to obtain a sample from the cervix. The sampling devices are then rinsed in a buffered alcohol transport and preservative solution instead of immediately transferring the cells to a glass slide. Sampling error is reduced, since all the cells are transferred into solution, in contrast with the conventional Pap test (CPT), in which only 20% of the cells are actually transferred to the glass slide.3 The cytologic specimen in solution is then sent to the laboratory where most of the debris, leukocytes, and erythrocytes are filtered from the specimen and a thin layer Pap test is prepared using a special processor. The cells are dispersed evenly with relatively little clumping or overlapping or obscuring debris (Figures 1A and B). The resulting clean smear is circular and contains approximately half the number of representative cells of a CPT. Residual cervical cells in solution that are not initially used for the Pap test may be tested at a later date for lower genital tract pathogens, such as carcinogenic human papillomavirus.4,5
Liquid-based thin layer cervical cytology has performed favorably in studies using split sample collection techniques.6-9 With the split sample technique, the cytologic specimen is first transferred to a glass slide for the CPT before transferring the remaining cervical cells to the liquid transport fluid. Evaluation of the LBCC Pap test using this nonrandom study design may be biased, since it is based solely on cells normally discarded. The CPT slides receive the first and perhaps better cellular sample; thus, the quality of that specimen may be vastly different than that available for LBCC. Currently, there are minimal data available in which LBCC was evaluated as it was intended to be used, for a primary cytologic specimen.10,11 Thus, the full diagnostic potential of this new Pap test technique has not been fully realized. The purpose of our trial was to determine the rates of specimen adequacy and cervical cytologic and histologically confirmed diagnoses obtained with LBCC using a direct-to-vial collection technique and compare these results with those obtained using the CPT.
Methods
Subjects
In our clinical study one group of women was enrolled prospectively, and a second group was considered retrospectively. The first group (LBCC) consisted of 1004 nonpregnant women aged 18 years or older with an intact cervix. The second group consisted of 2110 women with a similar patient profile who had a CPT collected at the same clinics during the 2 years immediately preceding the initiation of the trial.
To ensure an enriched sample of women with cervical neoplasia, approximately 20% of the subjects in both cohorts were included from women presenting consecutively for a colposcopy examination following an abnormal CPT result. The remaining 80% of the subjects represented a general screening population of women presenting for a routine Pap test. All subjects were seen at 1 of 3 clinics: the Family Medicine Center, the Obstetrics and Gynecology Clinic, or the Comprehensive Cancer Center at the Medical College of Georgia in Augusta.
Equipment and Materials
Samples for CPTs and LBCC (ThinPrep Pap Test, Cytyc Corporation, Boxborough, Mass) were collected using a plastic Ayre spatula and cytobrush (Medscand, North Hollywood, Fla). Cervical cells from the second group were processed by CPT methods. Cervical cells from the prospectively enrolled group were preserved and transferred in LBCC solution (PreservCyt, Cytyc Corporation). Cervical samples for the LBCC test were processed using the ThinPrep 2000 Processor (Cytyc Corporation).12
Study Design
Eligible prospectively enrolled women presenting consecutively for a routine Pap test or colposcopy examination at any of the 3 designated clinics signed an internal review board-approved informed consent form before participating. After properly visualizing the cervix, a plastic Ayre spatula was used to collect an ectocervical sample. Another sample was collected using a cytobrush rotated 90° to 180° within the endocervical canal. The spatula and cytobrush were then immediately rinsed vigorously in the LBCC solution. The container was sealed, labeled, and sent to the cytology laboratory for processing according to the manufacturer’s instructions.12 Cytotechnicians or pathologists in the Department of Pathology, Medical College of Georgia (MCG), evaluated all cytology and histology specimens. Cervical cytology was reported using the Bethesda System.
Computerized records from the MCG Cytology Laboratory were used to establish the second group. CPT records with similar patient profiles and clinic affiliation were selected chronologically beginning the day before the initiation of the trial and working backward in time until 2110 records were obtained (approximately 2 years). Cervical specimens were collected previously from this group using the same types of collection devices and technique as that for the LBCC group. However, immediately following sample collection these endocervical and ectocervical cells were plated on a glass slide, fixed, and sent to the cytology laboratory for Pap staining and diagnosis.
Statistical Analysis
The proportion of subjects in each of the Pap test adequacy groups (satisfactory, unsatisfactory, and satisfactory but limited by [SBLB]) were compared between the sample and control groups for the total study group, the colposcopy and general screening cohorts, and the 3 age groups (18-30 years, 31-40 years, 41 years or older). The chi-square statistic was used to compare the proportions unless the tables included cells with expected counts less than 5; for those cases we used the Fisher exact test as extended to tables larger than 2×2. A summary comparison controlling for age group was made using the Cohran-Mantel-Haenszel chi-square statistic. Average age was compared between the groups using the Student t test. At the given sample size the power was calculated to be 80% or greater for detecting a difference in positive rates of 3.5% between the LBCC and the CPT.
Results
Subjects in both groups were comparable by type of clinical visit. Consequently, 1004 LBCC subjects and 2110 CPT subjects were included in the data analysis. The routine Pap test screening group consisted of 796 and 1585 subjects in the LBCC and CPT groups, respectively, representing 79.0% of each population. Approximately 21.0% of the LBCC and CPT subjects (208 and 425, respectively) were represented by women presenting for colposcopy examination following an abnormal Pap test result. Subjects were also stratified into 1 of 3 age ranges for analytic purposes. Thus, 423 and 602 subjects were aged between 18 and 30 years; 287 and 523 subjects were aged 31 to 40 years; and 294 and 885 subjects were aged 41 years or older for the LBCC and CPT populations, respectively. A statistically significant difference between the age distributions was noted (P <.001) with a greater proportion of younger subjects in the LBCC group (mean age=35.0 years) than in the CPT cohort (mean age=39.8 years).
A greater percentage of satisfactory Pap tests were obtained when using LBCC (84.0%) compared with the CPT (60.5%), P ".001 (Figure 1). Also, significantly fewer unsatisfactory and SBLB Pap tests were reported using LBCC (1.2% and 14.8%) compared with the CPT (3.8% and 35.7%). For the colposcopy group, significantly more satisfactory Pap tests were reported for LBCC compared with the CPT (81.7% and 55.3%, respectively; P <.001). Similar results were also noted for the screening group. These significant results were maintained when the data were adjusted for the 3 age groups (Table 1).
The frequencies of cytologic diagnoses for LBCC and the CPT are seen in Table 2. The LBCC method detected a significantly greater percentage of women with low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL) Pap tests results (7.4% and 3.7%, respectively) compared with the CPT (1.7% and 1.7%, respectively). Similar significant results were noted when the LBCC and CPT groups were stratified into the screening and colposcopy populations. When both groups were adjusted for age, a greater percentage of LSIL and HSIL diagnoses were noted for LBCC in all 3 age ranges compared with the CPT (Table 4). The classification of atypical squamous cells of undetermined significance (ASCUS) cases was not modified by LBCC compared with the CPT. Because of the limited number of squamous cell carcinoma diagnoses reported, a comparison between the 2 groups was not possible.
The frequency of histologic diagnoses in the 2 groups closely paralleled Pap test diagnoses in the same populations. The LBCC and CPT groups had 5.4% and 1.0% cervical intraepithelial neoplasia 1 (CIN 1) histologic diagnoses, respectively, while 5.9% and 1.7% of the women had histologic diagnoses of CIN 2 and 3, respectively. These results were also maintained consistently across all 3 age ranges. The predictive value of a positive LBCC test (93.9%) was similar to that for the predictive value of a positive CPT (87.8%) when based on available histology results (Table 5). The sensitivity and specificity were equivalent for each test. Of all women with an LSIL or more severe Pap test result in each group, 1.9% of women in the CPT group and 6.5% of women in the LBCC group had cervical biopsies indicating CIN 1 or more severe disease.
Discussion
Pap test specimen adequacy results reflect whether the most likely site for neoplasia of the cervix—the transformation zone—has been sampled properly. LBCC provided significantly greater rates of satisfactory specimen adequacy than the CPT. Also, compared with the CPT, LBCC effectively reduced the number of SBLB and unsatisfactory specimen adequacy reports. Our findings are consistent with those of studies published previously that used split sample collection techniques.6,7 The improvement of Pap test adequacy using LBCC may be because a clearer inspection of pertinent normal and abnormal cells is achieved by eliminating or minimizing obscuring inflammatory cells, red blood cells, other debris, and clumped or overlapped cells.13 A significant decrease in unsatisfactory Pap tests reduces the extra cost, time, and patient and clinician inconvenience incurred from the necessity of repeating a nonrepresentative Pap test. These advantages may be of particular importance to patients, clinicians, and third-party payers.
Also, in regard to specimen adequacy, LBCC performed better than the CPT for women of all age ranges. This is especially important for older women who typically have a greater frequency of unsatisfactory Pap tests because their active transformation zones are positioned deeply within the endocervical canal and are often inaccessible to comprehensive sampling. In addition, a thin atrophic epithelium (commonly seen in the older age group) is more easily traumatized during sampling, which causes potentially obscuring bleeding. Yet, LBCC reduced the rate of unsatisfactory Pap tests by approximately 50% for women older than 40 years. Other than the advantage of filtering unwanted cells, the difference may be explained by the fact that 80% of cervical cells are discarded with the collection devices and not transferred to a glass slide using the CPT method.3 In contrast, nearly all cells are transferred to solution with LBCC. Thus, the larger specimen obtained with the sampling method used in LBCC increases the availability of a typically limited number of endocervical cells retrieved from post-treatment and postmenopausal women who are routinely more difficult to sample.
The ability of the Pap test to detect cytologic changes consistent with neoplasia, when present, is critically important. In our trial, the LBCC method detected a greater percentage of women with LSIL and HSIL compared with the CPT. Approximately 4 times the number of women with LSIL and twice the number of cases of HSIL were detected by LBCC compared with the CPT. These results replicate or surpass those found in studies that evaluated LBCC using a split sample method.6-9 Twice the rate of neoplasia detection by LBCC has been reported in studies using the split sample method. Our results, based on a direct-to-vial study design, may portray the true potential for LBCC to detect cervical neoplasia. A cleaner monolayer Pap test with a more comprehensive cellular specimen likely accounts for these remarkable differences. Similarly, the reduced number of unsatisfactory and SBLB Pap tests using LBCC may contribute to the greater yield of SILs. Of note, LBCC did not detect a significantly greater rate of ASCUS Pap test results; this may create problematic management decisions for some clinicians.14 The ability of LBCC to detect a greater percentage of women with SILs was maintained for the general screening population, the colposcopy cohort, and across the 3 age ranges. LBCC may be better able to detect women with HSIL, a lesion unlikely to resolve spontaneously. Women with HSIL (Figures 2A and B) and a comparable histologic diagnosis deserve prompt therapy for this true cancer precursor. Although many women with LSIL have disease that may regress to normal, as many as 20% to 40% of these women will have histologically confirmed CIN 2 or 3 that deserves treatment as well.15
Limitations
Our study is subject to several limitations. Most important, the observed differences in Pap test performance may have been because of actual differences of cervical neoplasia prevalences between the 2 groups, rather than a difference between Pap tests. Our subjects were selected from comparable patient populations by a study design that other researchers10,11,16-20 have used to evaluate the direct-to-vial technique. Other than sampling a population of women with LBCC and then having them return for a CPT 1 month later, which risks compliance failure and spectrum of cervical disease changes or acquisition of new disease, there is no other study design to evaluate a direct-to-vial technique. One may question whether our increased rate of neoplasia detection demonstrated by LBCC is due to a substantial number of LBCC false-positive results.21 Yet, because the positive predictive values for the LBCC and CPT were similar and there were 3 times as many women with biopsy results of CIN 1 or greater in the LBCC group compared with the CPT group when only women with LSIL or greater Pap test results were considered, the increased rates of SIL detection using LBCC appear to represent an increased detection of true disease and not false-positive results, further validating our findings.21 The equivalent results for LBCC and CPT specificity (99%) indicate that the increased rate of SIL detection by LBCC was not due to a reduced specificity. Also, the favorable test results demonstrated for LBCC are not entirely dissimilar from those of other studies that used a split sample technique, a more biased study design for LBCC.6-9 Studies by Vassilakos and colleagues16 and Papillo and coworkers10 demonstrated that a liquid-based thin layer cytologic report of LSIL or more severe disease correlated better with histologic results of CIN 1 or greater (80.5% and 80.2%, respectively) than CPTs (71.7% and 72.2%, respectively). Finally, because of the limited number of women with cervical caner in this study population, we were unable to determine if there were any differences between LBCC and the CPT in their ability to detect cervical cancer.
Conclusions
Compared with the CPT, LBCC detected a significantly greater percentage of satisfactory Pap tests and significantly reduced the number of unsatisfactory and SBLB tests. These findings demonstrate that LBCC significantly improves the adequacy of Pap tests and may increase the rate of detection of cervical neoplasia compared with CPT. Further study is necessary and warranted, since failure to detect cancer in a timely fashion affects ultimate cure rates, medical costs, quality of life, and perhaps medicolegal expenses. Although liquid-based thin layer cervical cytology is rapidly replacing the glass slide method throughout the United States, additional studies are also necessary to determine whether LBCC reduces the incidence of cervical cancer.
Acknowledgments
We appreciate the assistance of Cytyc Corporation in providing the supplies and grant support necessary for completion of this study.
We would like to thank the faculty and residents in the Family Medicine Center and OB/GYN Clinic who assisted with specimen collection, Jim Best for processing and interpreting cytologic specimens, Aisha Lavin for data management assistance, and April Dean for manuscript preparation.
METHODS: A total of 1004 nonpregnant women aged 18 years or older with an intact cervix had Pap tests collected with an Ayre spatula and cytobrush, and the sample was placed in a preservative solution. The specimens were processed as thin layer Pap tests according to the manufacturer’s specifications. Another group of 2110 women with a similar patient profile had a CPT collected immediately preceding the initiation of the trial. The subjects in each group consisted of an equal percentage of women presenting for a routine Pap test or a colposcopy examination. We compared the distributions of diagnostic categories between the groups using a chi-square test.
RESULTS: A significantly greater percentage of satisfactory Pap tests were obtained using LBCC (84.0%) compared with the CPT (60.5%, P <.001). Fewer satisfactory but limited by (SBLB, 14.8%) and unsatisfactory (1.2%) Pap tests were reported using LBCC compared with the CPT (35.7% and 3.8%, respectively, c2=170.7, P <.001). A significantly greater percentage of low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL) Pap test results were reported using LBCC (7.4% and 3.7%, respectively) compared with the CPT (1.7% and 1.7%, respectively, c2=74.4, P & .001). The predictive value of a positive LBCC test (93.9%) was similar to that for a positive CPT (87.8%) when compared with histology results.
CONCLUSIONS: Compared with the CPT, LBCC detected a significantly greater percentage of satisfactory Pap tests and significantly reduced the number of unsatisfactory and SBLB tests. Four times the percentage of LSIL and twice the percentage of HSIL Pap test results were obtained using LBCC compared with the CPT. These findings demonstrate that LBCC significantly improves the adequacy of Pap tests and may increase the rate of detection of cervical neoplasia compared with the CPT.
The Papanicolaou (Pap) test is one of the most common and effective cancer screening tests used by primary care clinicians. However, a recent study conducted through the Agency for Health Care Policy and Research1 estimated that the true sensitivity of the Pap test is only 51%. Two thirds of the false-negative results occur because of sampling error. The first liquid-based cervical cytology (LBCC) test was approved for clinical use by the United States Food and Drug Administration in 1996.2 With LBCC, standard cervical cytology sampling devices are used to obtain a sample from the cervix. The sampling devices are then rinsed in a buffered alcohol transport and preservative solution instead of immediately transferring the cells to a glass slide. Sampling error is reduced, since all the cells are transferred into solution, in contrast with the conventional Pap test (CPT), in which only 20% of the cells are actually transferred to the glass slide.3 The cytologic specimen in solution is then sent to the laboratory where most of the debris, leukocytes, and erythrocytes are filtered from the specimen and a thin layer Pap test is prepared using a special processor. The cells are dispersed evenly with relatively little clumping or overlapping or obscuring debris (Figures 1A and B). The resulting clean smear is circular and contains approximately half the number of representative cells of a CPT. Residual cervical cells in solution that are not initially used for the Pap test may be tested at a later date for lower genital tract pathogens, such as carcinogenic human papillomavirus.4,5
Liquid-based thin layer cervical cytology has performed favorably in studies using split sample collection techniques.6-9 With the split sample technique, the cytologic specimen is first transferred to a glass slide for the CPT before transferring the remaining cervical cells to the liquid transport fluid. Evaluation of the LBCC Pap test using this nonrandom study design may be biased, since it is based solely on cells normally discarded. The CPT slides receive the first and perhaps better cellular sample; thus, the quality of that specimen may be vastly different than that available for LBCC. Currently, there are minimal data available in which LBCC was evaluated as it was intended to be used, for a primary cytologic specimen.10,11 Thus, the full diagnostic potential of this new Pap test technique has not been fully realized. The purpose of our trial was to determine the rates of specimen adequacy and cervical cytologic and histologically confirmed diagnoses obtained with LBCC using a direct-to-vial collection technique and compare these results with those obtained using the CPT.
Methods
Subjects
In our clinical study one group of women was enrolled prospectively, and a second group was considered retrospectively. The first group (LBCC) consisted of 1004 nonpregnant women aged 18 years or older with an intact cervix. The second group consisted of 2110 women with a similar patient profile who had a CPT collected at the same clinics during the 2 years immediately preceding the initiation of the trial.
To ensure an enriched sample of women with cervical neoplasia, approximately 20% of the subjects in both cohorts were included from women presenting consecutively for a colposcopy examination following an abnormal CPT result. The remaining 80% of the subjects represented a general screening population of women presenting for a routine Pap test. All subjects were seen at 1 of 3 clinics: the Family Medicine Center, the Obstetrics and Gynecology Clinic, or the Comprehensive Cancer Center at the Medical College of Georgia in Augusta.
Equipment and Materials
Samples for CPTs and LBCC (ThinPrep Pap Test, Cytyc Corporation, Boxborough, Mass) were collected using a plastic Ayre spatula and cytobrush (Medscand, North Hollywood, Fla). Cervical cells from the second group were processed by CPT methods. Cervical cells from the prospectively enrolled group were preserved and transferred in LBCC solution (PreservCyt, Cytyc Corporation). Cervical samples for the LBCC test were processed using the ThinPrep 2000 Processor (Cytyc Corporation).12
Study Design
Eligible prospectively enrolled women presenting consecutively for a routine Pap test or colposcopy examination at any of the 3 designated clinics signed an internal review board-approved informed consent form before participating. After properly visualizing the cervix, a plastic Ayre spatula was used to collect an ectocervical sample. Another sample was collected using a cytobrush rotated 90° to 180° within the endocervical canal. The spatula and cytobrush were then immediately rinsed vigorously in the LBCC solution. The container was sealed, labeled, and sent to the cytology laboratory for processing according to the manufacturer’s instructions.12 Cytotechnicians or pathologists in the Department of Pathology, Medical College of Georgia (MCG), evaluated all cytology and histology specimens. Cervical cytology was reported using the Bethesda System.
Computerized records from the MCG Cytology Laboratory were used to establish the second group. CPT records with similar patient profiles and clinic affiliation were selected chronologically beginning the day before the initiation of the trial and working backward in time until 2110 records were obtained (approximately 2 years). Cervical specimens were collected previously from this group using the same types of collection devices and technique as that for the LBCC group. However, immediately following sample collection these endocervical and ectocervical cells were plated on a glass slide, fixed, and sent to the cytology laboratory for Pap staining and diagnosis.
Statistical Analysis
The proportion of subjects in each of the Pap test adequacy groups (satisfactory, unsatisfactory, and satisfactory but limited by [SBLB]) were compared between the sample and control groups for the total study group, the colposcopy and general screening cohorts, and the 3 age groups (18-30 years, 31-40 years, 41 years or older). The chi-square statistic was used to compare the proportions unless the tables included cells with expected counts less than 5; for those cases we used the Fisher exact test as extended to tables larger than 2×2. A summary comparison controlling for age group was made using the Cohran-Mantel-Haenszel chi-square statistic. Average age was compared between the groups using the Student t test. At the given sample size the power was calculated to be 80% or greater for detecting a difference in positive rates of 3.5% between the LBCC and the CPT.
Results
Subjects in both groups were comparable by type of clinical visit. Consequently, 1004 LBCC subjects and 2110 CPT subjects were included in the data analysis. The routine Pap test screening group consisted of 796 and 1585 subjects in the LBCC and CPT groups, respectively, representing 79.0% of each population. Approximately 21.0% of the LBCC and CPT subjects (208 and 425, respectively) were represented by women presenting for colposcopy examination following an abnormal Pap test result. Subjects were also stratified into 1 of 3 age ranges for analytic purposes. Thus, 423 and 602 subjects were aged between 18 and 30 years; 287 and 523 subjects were aged 31 to 40 years; and 294 and 885 subjects were aged 41 years or older for the LBCC and CPT populations, respectively. A statistically significant difference between the age distributions was noted (P <.001) with a greater proportion of younger subjects in the LBCC group (mean age=35.0 years) than in the CPT cohort (mean age=39.8 years).
A greater percentage of satisfactory Pap tests were obtained when using LBCC (84.0%) compared with the CPT (60.5%), P ".001 (Figure 1). Also, significantly fewer unsatisfactory and SBLB Pap tests were reported using LBCC (1.2% and 14.8%) compared with the CPT (3.8% and 35.7%). For the colposcopy group, significantly more satisfactory Pap tests were reported for LBCC compared with the CPT (81.7% and 55.3%, respectively; P <.001). Similar results were also noted for the screening group. These significant results were maintained when the data were adjusted for the 3 age groups (Table 1).
The frequencies of cytologic diagnoses for LBCC and the CPT are seen in Table 2. The LBCC method detected a significantly greater percentage of women with low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL) Pap tests results (7.4% and 3.7%, respectively) compared with the CPT (1.7% and 1.7%, respectively). Similar significant results were noted when the LBCC and CPT groups were stratified into the screening and colposcopy populations. When both groups were adjusted for age, a greater percentage of LSIL and HSIL diagnoses were noted for LBCC in all 3 age ranges compared with the CPT (Table 4). The classification of atypical squamous cells of undetermined significance (ASCUS) cases was not modified by LBCC compared with the CPT. Because of the limited number of squamous cell carcinoma diagnoses reported, a comparison between the 2 groups was not possible.
The frequency of histologic diagnoses in the 2 groups closely paralleled Pap test diagnoses in the same populations. The LBCC and CPT groups had 5.4% and 1.0% cervical intraepithelial neoplasia 1 (CIN 1) histologic diagnoses, respectively, while 5.9% and 1.7% of the women had histologic diagnoses of CIN 2 and 3, respectively. These results were also maintained consistently across all 3 age ranges. The predictive value of a positive LBCC test (93.9%) was similar to that for the predictive value of a positive CPT (87.8%) when based on available histology results (Table 5). The sensitivity and specificity were equivalent for each test. Of all women with an LSIL or more severe Pap test result in each group, 1.9% of women in the CPT group and 6.5% of women in the LBCC group had cervical biopsies indicating CIN 1 or more severe disease.
Discussion
Pap test specimen adequacy results reflect whether the most likely site for neoplasia of the cervix—the transformation zone—has been sampled properly. LBCC provided significantly greater rates of satisfactory specimen adequacy than the CPT. Also, compared with the CPT, LBCC effectively reduced the number of SBLB and unsatisfactory specimen adequacy reports. Our findings are consistent with those of studies published previously that used split sample collection techniques.6,7 The improvement of Pap test adequacy using LBCC may be because a clearer inspection of pertinent normal and abnormal cells is achieved by eliminating or minimizing obscuring inflammatory cells, red blood cells, other debris, and clumped or overlapped cells.13 A significant decrease in unsatisfactory Pap tests reduces the extra cost, time, and patient and clinician inconvenience incurred from the necessity of repeating a nonrepresentative Pap test. These advantages may be of particular importance to patients, clinicians, and third-party payers.
Also, in regard to specimen adequacy, LBCC performed better than the CPT for women of all age ranges. This is especially important for older women who typically have a greater frequency of unsatisfactory Pap tests because their active transformation zones are positioned deeply within the endocervical canal and are often inaccessible to comprehensive sampling. In addition, a thin atrophic epithelium (commonly seen in the older age group) is more easily traumatized during sampling, which causes potentially obscuring bleeding. Yet, LBCC reduced the rate of unsatisfactory Pap tests by approximately 50% for women older than 40 years. Other than the advantage of filtering unwanted cells, the difference may be explained by the fact that 80% of cervical cells are discarded with the collection devices and not transferred to a glass slide using the CPT method.3 In contrast, nearly all cells are transferred to solution with LBCC. Thus, the larger specimen obtained with the sampling method used in LBCC increases the availability of a typically limited number of endocervical cells retrieved from post-treatment and postmenopausal women who are routinely more difficult to sample.
The ability of the Pap test to detect cytologic changes consistent with neoplasia, when present, is critically important. In our trial, the LBCC method detected a greater percentage of women with LSIL and HSIL compared with the CPT. Approximately 4 times the number of women with LSIL and twice the number of cases of HSIL were detected by LBCC compared with the CPT. These results replicate or surpass those found in studies that evaluated LBCC using a split sample method.6-9 Twice the rate of neoplasia detection by LBCC has been reported in studies using the split sample method. Our results, based on a direct-to-vial study design, may portray the true potential for LBCC to detect cervical neoplasia. A cleaner monolayer Pap test with a more comprehensive cellular specimen likely accounts for these remarkable differences. Similarly, the reduced number of unsatisfactory and SBLB Pap tests using LBCC may contribute to the greater yield of SILs. Of note, LBCC did not detect a significantly greater rate of ASCUS Pap test results; this may create problematic management decisions for some clinicians.14 The ability of LBCC to detect a greater percentage of women with SILs was maintained for the general screening population, the colposcopy cohort, and across the 3 age ranges. LBCC may be better able to detect women with HSIL, a lesion unlikely to resolve spontaneously. Women with HSIL (Figures 2A and B) and a comparable histologic diagnosis deserve prompt therapy for this true cancer precursor. Although many women with LSIL have disease that may regress to normal, as many as 20% to 40% of these women will have histologically confirmed CIN 2 or 3 that deserves treatment as well.15
Limitations
Our study is subject to several limitations. Most important, the observed differences in Pap test performance may have been because of actual differences of cervical neoplasia prevalences between the 2 groups, rather than a difference between Pap tests. Our subjects were selected from comparable patient populations by a study design that other researchers10,11,16-20 have used to evaluate the direct-to-vial technique. Other than sampling a population of women with LBCC and then having them return for a CPT 1 month later, which risks compliance failure and spectrum of cervical disease changes or acquisition of new disease, there is no other study design to evaluate a direct-to-vial technique. One may question whether our increased rate of neoplasia detection demonstrated by LBCC is due to a substantial number of LBCC false-positive results.21 Yet, because the positive predictive values for the LBCC and CPT were similar and there were 3 times as many women with biopsy results of CIN 1 or greater in the LBCC group compared with the CPT group when only women with LSIL or greater Pap test results were considered, the increased rates of SIL detection using LBCC appear to represent an increased detection of true disease and not false-positive results, further validating our findings.21 The equivalent results for LBCC and CPT specificity (99%) indicate that the increased rate of SIL detection by LBCC was not due to a reduced specificity. Also, the favorable test results demonstrated for LBCC are not entirely dissimilar from those of other studies that used a split sample technique, a more biased study design for LBCC.6-9 Studies by Vassilakos and colleagues16 and Papillo and coworkers10 demonstrated that a liquid-based thin layer cytologic report of LSIL or more severe disease correlated better with histologic results of CIN 1 or greater (80.5% and 80.2%, respectively) than CPTs (71.7% and 72.2%, respectively). Finally, because of the limited number of women with cervical caner in this study population, we were unable to determine if there were any differences between LBCC and the CPT in their ability to detect cervical cancer.
Conclusions
Compared with the CPT, LBCC detected a significantly greater percentage of satisfactory Pap tests and significantly reduced the number of unsatisfactory and SBLB tests. These findings demonstrate that LBCC significantly improves the adequacy of Pap tests and may increase the rate of detection of cervical neoplasia compared with CPT. Further study is necessary and warranted, since failure to detect cancer in a timely fashion affects ultimate cure rates, medical costs, quality of life, and perhaps medicolegal expenses. Although liquid-based thin layer cervical cytology is rapidly replacing the glass slide method throughout the United States, additional studies are also necessary to determine whether LBCC reduces the incidence of cervical cancer.
Acknowledgments
We appreciate the assistance of Cytyc Corporation in providing the supplies and grant support necessary for completion of this study.
We would like to thank the faculty and residents in the Family Medicine Center and OB/GYN Clinic who assisted with specimen collection, Jim Best for processing and interpreting cytologic specimens, Aisha Lavin for data management assistance, and April Dean for manuscript preparation.
1. Agency for Health Care Policy and Research. Evidence report/technology assessment no. 5: evaluation of cervical cytology. Rockville, Md: Agency for Health Care Policy and Research; 1999. US Department of Health and Human Services AHCPR publication no. 99-E010.
2. Lee KR, Ashfaq R, Birdsong GG, Corkill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and a fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol 1997;90:278-84.
3. Hutchinson ML, Agarwal P, Denault T, Berger B, Cibas ES. A new look at cervical cytology: ThinPrep multicenter trial results. Acta Cytol 1992;36:499-504.
4. Ferris DG, Wright TC, Jr, Litaker MS, et al. Comparison of two tests for detecting carcinogenic HPV in women with Papanicolaou smear reports of ASCUS and LSIL. J Fam Pract 1998;46:136-41.
5. Ferenczy A, Franco E, Arseneau J, Wright TC, Richart RM. Diagnostic performance of hybrid capture human papillomavirus deoxyribonucleic acid assay combined with liquid-based cytologic study. Am J Obstet Gynecol 1996;175:651-56.
6. Roberts JM, Gurley AM, Thurloe JK, Bowditch R, Laverty CRA. Evaluation of the ThinPrep Pap test as an adjunct to the conventional Pap smear. Med J Aust 1997;167:466-69.
7. Hutchinson ML, Isenstein LM, Goodman A, et al. Homogeneous sampling accounts for the increased diagnostic accuracy using the ThinPrep(tm) Processor. Anat Path 1994;101:215-19.
8. Corkhill M, Knapp D, Hutchinson ML. Improved accuracy for cervical cytology with the ThinPrep method and the endocervical brush-spatula collection procedure. J Lower Genital Tract Dis 1998;2:12-16.
9. KR, Ashfaq R, Birdsong GG, Corkhill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and a fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol 1997;90:278-84.
10. Papillo JL, Zarka MA, St John TL. Evaluation of the ThinPrep Pap test in clinical practice: a seven-month, 16,314-case experience in northern Vermont. Acta Cytol 1998;42:203-08.
11. Diaz-Rosario LA, Kabawat SE. Performance of a fluid-based, thin-layer Papanicolaou smear method in the clinical setting of an independent laboratory and an outpatient screening population in New England. Arch Pathol Lab Med 1999;123:817-21.
12. Bolick DR, Hellman DJ. Laboratory implementation and efficacy assessment of the ThinPrep cervical cancer screening system. Acta Cytol 1998;42:209-13.
13. Linder J, Zahniser D. The ThinPrep Pap test: a review of clinical studies. Acta Cytol 1997;41:30-38.
14. Ferris DG, Wright TC, Litaker MS, et al. Triage of women with ASCUS and LSIL on Pap smear reports: management by repeat Pap smear, HPV DNA testing, or colposcopy? J Fam Pract 1998;46:125-34.
15. Kurman RJ, Henson DE, Herbst AL, Noller KL, Schiffmann MH. Interim guidelines for management of abnormal cervical cytology. JAMA 1994;271:1866-69.
16. Vassilakos P, Schwartz D, de Marval F, et al. Biopsy-based comparison of liquid-based, thin-layer preparations to conventional Pap smears. J Reprod Med 2000;45:11-16.
17. Vassilakos P, Griffin S, Megevand E, Campara A. CytoRich liquid-based cervical cytologic test screening results in a routine cytopathology service. Acta Cytol 1998;42:198-202.
18. Vassilakos P, Saurel J, Rondez R. Direct to vial use of the AutoCyte PREP liquid-based preparation for cervical-vaginal specimens in three European laboratories. Acta Cytol 1999;43:65-68.
19. Ashfaq R, Gibbons D, Vela C, Saboorian MH, Iliya F. Thin Prep Pap test accuracy for glandular disease. Acta Cytol 1999;43:81-85.
20. Weintraub J, Morabia A. Efficacy of a liquid-based thin layer method for cervical cancer screening in a population with a low incidence of cervical cancer. Diagn Cytopathol 2000;22:52-59.
21. Sawaya GF, Grimes DA. New technologies in cervical cytology screening: a word of caution. Obstet Gynecol 1999;94:307-10.
1. Agency for Health Care Policy and Research. Evidence report/technology assessment no. 5: evaluation of cervical cytology. Rockville, Md: Agency for Health Care Policy and Research; 1999. US Department of Health and Human Services AHCPR publication no. 99-E010.
2. Lee KR, Ashfaq R, Birdsong GG, Corkill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and a fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol 1997;90:278-84.
3. Hutchinson ML, Agarwal P, Denault T, Berger B, Cibas ES. A new look at cervical cytology: ThinPrep multicenter trial results. Acta Cytol 1992;36:499-504.
4. Ferris DG, Wright TC, Jr, Litaker MS, et al. Comparison of two tests for detecting carcinogenic HPV in women with Papanicolaou smear reports of ASCUS and LSIL. J Fam Pract 1998;46:136-41.
5. Ferenczy A, Franco E, Arseneau J, Wright TC, Richart RM. Diagnostic performance of hybrid capture human papillomavirus deoxyribonucleic acid assay combined with liquid-based cytologic study. Am J Obstet Gynecol 1996;175:651-56.
6. Roberts JM, Gurley AM, Thurloe JK, Bowditch R, Laverty CRA. Evaluation of the ThinPrep Pap test as an adjunct to the conventional Pap smear. Med J Aust 1997;167:466-69.
7. Hutchinson ML, Isenstein LM, Goodman A, et al. Homogeneous sampling accounts for the increased diagnostic accuracy using the ThinPrep(tm) Processor. Anat Path 1994;101:215-19.
8. Corkhill M, Knapp D, Hutchinson ML. Improved accuracy for cervical cytology with the ThinPrep method and the endocervical brush-spatula collection procedure. J Lower Genital Tract Dis 1998;2:12-16.
9. KR, Ashfaq R, Birdsong GG, Corkhill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and a fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol 1997;90:278-84.
10. Papillo JL, Zarka MA, St John TL. Evaluation of the ThinPrep Pap test in clinical practice: a seven-month, 16,314-case experience in northern Vermont. Acta Cytol 1998;42:203-08.
11. Diaz-Rosario LA, Kabawat SE. Performance of a fluid-based, thin-layer Papanicolaou smear method in the clinical setting of an independent laboratory and an outpatient screening population in New England. Arch Pathol Lab Med 1999;123:817-21.
12. Bolick DR, Hellman DJ. Laboratory implementation and efficacy assessment of the ThinPrep cervical cancer screening system. Acta Cytol 1998;42:209-13.
13. Linder J, Zahniser D. The ThinPrep Pap test: a review of clinical studies. Acta Cytol 1997;41:30-38.
14. Ferris DG, Wright TC, Litaker MS, et al. Triage of women with ASCUS and LSIL on Pap smear reports: management by repeat Pap smear, HPV DNA testing, or colposcopy? J Fam Pract 1998;46:125-34.
15. Kurman RJ, Henson DE, Herbst AL, Noller KL, Schiffmann MH. Interim guidelines for management of abnormal cervical cytology. JAMA 1994;271:1866-69.
16. Vassilakos P, Schwartz D, de Marval F, et al. Biopsy-based comparison of liquid-based, thin-layer preparations to conventional Pap smears. J Reprod Med 2000;45:11-16.
17. Vassilakos P, Griffin S, Megevand E, Campara A. CytoRich liquid-based cervical cytologic test screening results in a routine cytopathology service. Acta Cytol 1998;42:198-202.
18. Vassilakos P, Saurel J, Rondez R. Direct to vial use of the AutoCyte PREP liquid-based preparation for cervical-vaginal specimens in three European laboratories. Acta Cytol 1999;43:65-68.
19. Ashfaq R, Gibbons D, Vela C, Saboorian MH, Iliya F. Thin Prep Pap test accuracy for glandular disease. Acta Cytol 1999;43:81-85.
20. Weintraub J, Morabia A. Efficacy of a liquid-based thin layer method for cervical cancer screening in a population with a low incidence of cervical cancer. Diagn Cytopathol 2000;22:52-59.
21. Sawaya GF, Grimes DA. New technologies in cervical cytology screening: a word of caution. Obstet Gynecol 1999;94:307-10.