Original Research

Night sweats have been attributed to tuberculosis, other acute and chronic febrile illnesses, menopause, pregnancy, hyperthyroidism, nocturnal hypoglycemia, other endocrine problems, neurologic diseases, sleep disorders (eg, sleep apnea and nightmares), malignancies, autoimmune diseases, coronary artery spasm, congestive heart failure, gastroesophageal reflux disease, psychiatric disorders, and certain medications. In 36 medical and surgical textbooks, night sweats were always discussed within sections covering specific diseases and never as a separate topic. References to the primary literature were never provided. We also searched Micromedix, a comprehensive source of information on medications, using “sweating” and “diaphoresis” as search terms.¹ Table W1 contains a comprehensive list of proposed causes of night sweats identified in our searches and accompanying references.

Only 2 epidemiologic studies of night sweats were found in the English language literature. Lea and Aber² interviewed 174 patients randomly selected from the inpatient units of a university hospital and found that 33% of nonobstetric patients and 60% of obstetric patients reported having had night sweats during the previous 3 months. Twenty-six percent of those with night sweats reported that their nighttime sweating was severe enough to require bathing and changing of bed linens. Reynolds,³ a gastroenterologist, queried 200 consecutive patients seen in his outpatient practice and found that 40% remembered experiencing night sweats at least once during the previous year. A total of 12% reported at least weekly night sweats. A review of the records of 750 patients at the Geriatric Continuity Clinic at the University of Oklahoma Family Medicine Center revealed that 10% reported having experienced night sweats during the previous month, when the question was asked as part of a standard review of systems questionnaire (J.W.M., unpublished data, 1999).

Our study was conducted in an effort to estimate the prevalence of night sweats in adult patients seen in primary care office settings, and to explore the associations of this symptom with demographic factors, physical characteristics, medical problems, and medications. We also sought to determine how distressing this symptom is to those who have it and to their sleep partners, whether patients are likely to report the symptom to their physicians, and what patients and their physicians think causes night sweats in individual cases.

Methods

Physician members of the Oklahoma Physicians Resource/Research Network (OKPRN) and the Texas Academy of Family Physicians Research Network (TAFP-Net) enrolled consecutive patients 18 years and older seen in their clinics during a 1-week period in the summer and a second 1-week period in the winter in the years 2000 and 2001. Patients who agreed to participate signed a consent form and then helped the nurse and physician complete a brief questionnaire on a preaddressed, stamped data collection card. For those who declined to participate, a card was generated containing the physician’s code number and the patient’s age and sex. Questions elicited demographic information; information about a selected set of medical conditions; medications, vitamins, herbs, and alcohol used regularly; and information about recent experiences with night sweats. Participating physicians were asked to check the questionnaires for accuracy and to record their opinions regarding the cause of the patients’ night sweats when they reported having had them. A laminated card with definitions of terms was provided to each physician.

“Night sweats” was defined as “sweating at night even when it isn’t excessively hot in your bedroom.” “Day sweats” was defined as “excessive sweating during the daytime.” “Pure night sweats” was defined as night sweats, but not day sweats, and “night and day sweats” as the combination of the 2. The time interval was specified as “during the last month.”

Completed questionnaires were mailed to the Oklahoma Center for Family Medicine Research for data entry and analysis. The data collection cards used by the Texas network included questions about race/ethnicity and panic attacks that were not included on the Oklahoma cards. Inadvertently, some of the Texas cards did not include the question about daytime sweating.

Statistix7 (Analytical Software, Tallahassee, Fla) was used for all statistical analyses. Medications were assigned to 1 of 47 categories according to their primary pharmacologic effects. Summary statistics were calculated for all participants and for the following subgroups: season (summer and winter), pattern of night sweats (excessive nighttime sweating only or night and day sweats), and age group. We anticipated that the majority of women with menopausal symptoms would be in the 41- to 55-year age group.

The two patterns of night sweats, “pure night sweats” and “night and day sweats,” were analyzed separately, and by sex and age. Logistic regression was used to identify the most significant predictors of night sweats while controlling for other variables. Variables were entered into the logistic models if they had a univariate association with the dependent variable at a P value of less than .05. They were then removed one at a time, in the order of largest to smallest P value, if they had a P value of greater than .01 after controlling for other variables. Conservative P values were chosen because of the large numbers of variables considered, in order to reduce the probability of type 1 errors. When appropriate, 95% confidence intervals were calculated.

Results

Study population

A total of 2267 patients of 31 different physicians participated in this study, including 1888 patients of 24 Oklahoma physicians and 379 patients of 7 Texas physicians. Their mean (standard deviation) age was 50.7 (18.8) years, with a range of 18 to 97 years. Sixty-nine percent were women. A total of 99% of Oklahoma patients and 93% of Texas patients seen during the study weeks agreed to participate in the study. Among Texas participants, 53% were Hispanic whites, 33% were non-Hispanic whites, 13% were African Americans, and 1% were categorized as other. On the basis of prior OKPRN studies, we suspect that approximately 90% of Oklahoma patients were non-Hispanic whites, but exact proportions were not determined for this study.

Prevalence of night sweats

The prevalence of pure night sweats, night and day sweats, and any night sweats are shown in Table 1. While the prevalence of night and day sweats was lower for older patients, severity tended to be greater. Severity and frequency were positively correlated for all categories of night sweats and for all subgroups of patients (overall Spearman coefficient = 0.33; P < .001). Overall, the frequencies of night sweats among those who reported the condition were: almost never, 18%; 1 to 3 nights per month, 38%; 1 to 3 nights per week, 27%; and 4 to 7 nights per week, 16%. Ten percent of both women and men with night sweats said that their night sweats were bothersome to others.

TABLE 1
Percentage of patients with pure night sweats and night and day

Frequency of reporting of night sweats

A minority of patients with night sweats (12%) had reported the symptom to their physicians. This was true even for those with severe night sweats (46%). Women younger than 70 years were more likely than men of the same age to have reported their night sweats to their physicians (15% vs. 6%; P < .001). The reverse was true for those 70 years and older (7% vs 13%; P =.08). Older patients with pure night sweats were more likely than younger patients to have reported them. After controlling for other variables, patients who were older (odds ratio [OR] = 1.03 per year of age; P < .001), those with night and day sweats (OR = 1.74; P =.0015), and those who reported that their night sweats bothered others (OR = 2.89; P =.001) were more likely to have reported the symptom to their physicians. Those who had reported their night sweats were also more likely to have hot flashes (OR = 2.98; P < .001) and to take estrogen (OR = 1.72; P =.003).

Factors associated with night sweats

The only variable associated with pure night sweats after controlling for all other variables was panic attacks. Variables associated with night and day sweats were younger age, greater body mass index, hot flashes, chronic infection, sleep disturbances, selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressants, “other” (non–SSRI, non-tricyclic) antidepressants, and xanthines.

For women, the only variable clearly associated with pure night sweats in the multivariate model was hot flashes. Panic attacks nearly reached significance (P =.026) and improved the regression model substantially (deviance reduced from 1446 to 87). Variables associated with night and day sweats were weight, sleep problems, hot flashes, antihistamines, SSRIs, and other (non–SSRI, nontricyclic) antidepressants.

For men, the only variable associated with pure night sweats after controlling for other variables was sleep problems. After exclusion of sleep problems and sedatives from the model on the assumption that they might be the result rather than the cause of night sweats, significant predictors were hot flashes (OR = 2.70; 95% confidence interval [CI], 1.35-5.40; P =.005) and regular use of multivitamins (OR = 1.87; 95% CI, 1.17-2.99; P =.009). Variables associated with night and day sweats included greater weight, hot flashes, and greater alcohol use. The ORs and CIs are shown in Table 1.

Interestingly, 32 men (5%) reported hot flashes, and those who did were more likely to report night sweats of both types. Men with hot flashes were evenly distributed across age categories. Their night sweats were more frequent, but not more severe, and they were more likely to bother others than those without hot flashes. Men with hot flashes were more likely to have told their physicians about their night sweats. After controlling for other variables, men with hot flashes were much more likely to have panic attacks (OR = 28.28; P < .001).

Patients 70 years and older made up 19.5% of our sample (N=429). The only factor associated with pure night sweats in the multivariate model was sleep disturbances (OR = 2.04; = 95% CI, 1.21-3.42; P =.007). Exclusion of sleep disturbances left no associated variables. Variables associated with night and day sweats were hot flashes (OR = 15.14; = 95% CI, 6.43-35.68; P < .001) and corticosteroids (OR = 5.45; 95% CI 1.58-18.86; P =.007).

Suspected causes

In cases where patients reported night sweats, only 19% of the patients and 18% of their physicians recorded opinions regarding causation. The suspected causes listed by patients and physicians were similar. Both groups listed menopause most frequently (48% and 44%, respectively). Other etiologies proposed were stress (12% and 8%) and medications (9% and 10%). Physicians listed diabetes as a possible cause in 11% of cases while only 4% of patients listed it. Other suspected causes included obesity, pregnancy, gastroesophageal reflux disease, sleep discomforts, and ambient temperature.

TABLE 2
Associations between independent variables and night sweats in men and women after using logistic regression modeling to control for all other variables

Discussion

As far as we know, this is the first systematic study of night sweats in a primary care population. It is exploratory in nature, and, because of its cross-sectional design, no firm conclusions can be drawn about causation.

Both pure night sweats and night and day sweats are extremely common, with a peak prevalence in men and women 41 to 55 years of age. In contrast to pure night sweats, night and day sweats are experienced infrequently by patients 70 years and older. The factors associated with pure night sweats are somewhat different than those associated with night and day sweats, suggesting different, though probably overlapping, sets of causes. The different associations seen for men and women, and for older and younger patients, are also noteworthy. Patients often fail to report night sweats to their primary care physician, even when frequent and severe, associated with sleep disturbances, or bothersome to others.

Because of the sampling method (ie, consecutive patients rather than a random sample of active patients), the prevalence estimates reflect the frequency at which physicians can expect to encounter patients with this symptom, rather than the prevalence of night sweats among active patients. Since patients with more symptoms probably see physicians more often, we assume we have overestimated the true prevalence of night sweats in the larger population. Participating physicians were also not selected randomly. It is impossible to know how this may have affected our results.

We were surprised that so few of our independent variables were associated with pure night sweats: only panic attacks (all patients), sleep disorders (men and older patients), and hot flashes (women). Factors not associated with pure night sweats included obesity; diabetes, insulin, or oral hypoglycemic agents; acute or chronic infections; gastroesophageal reflux disease; or thyroid medications. Pure night sweats were also not specifically associated with estrogen and progesterone, although they were associated with hot flashes. There was also no association of pure night sweats and alcohol consumption.

The fact that physicians and their patients could only speculate on a cause for night sweats in 1 out of 5 cases suggests a lack of familiarity with the multitude of suspected causes, a failure to detect certain common causes (eg, sleep disorders and panic attacks), or, most likely, that many common causes of night sweats have yet to be elucidated. If the last is correct, it may be an example of the bias in the primary and secondary clinical literature that occurs when clinical research is carried out primarily in the subspecialty clinics of academic medical centers.^4-7 Our findings speak to the need for greater support for primary care practice-based research.^8,9

In retrospect, the omission of the variable “panic attacks” from the Oklahoma cards was a mistake, since this variable was correlated with pure night sweats in women. It may have been more strongly associated with pure night sweats in men as well, if the number of respondents to this question had been larger. Also, some men complained of hot flashes, and when they did, they were more likely to have night sweats and panic attacks, suggesting that both hot flashes and night sweats in men should prompt physicians to ask additional questions about panic disorder. Although race was also omitted from the Oklahoma cards, this variable did not seem to be associated with differences in night sweats prevalence or association among those for whom this information was available.

The definition and description of night sweats used in this study were arbitrary and may have influenced the prevalence rates obtained. We attempted to exclude environmental temperature as a cause. Although the definitions provided clearly stated “within the last month,” the data collection cards did not specify a time interval. This may have resulted in some variation in interpretation.

The decisions that were made regarding logistic modeling strategies were conservative and may have excluded some important variables. However, with so many variables and no basis on which to judge a priori, we felt that a conservative approach was best. The decision to include in the models variables (eg, sleep problems and sedatives that might be considered consequences) rather than causes of night sweats, was also arbitrary and may have affected the results. An alternative explanation of the associations found between night sweats and sleep problems is that those who are unable to sleep for other reasons are more likely to notice excessive sweating than those who are asleep.

Future studies should more carefully examine factors found in this study to be associated with night sweats, such as panic attacks and sleep disorders, and other potential etiologic factors not considered, such as tobacco abuse, allergic diseases, migraines, congestive heart failure, and chronic lung disease. Given the high prevalence, future studies examining etiology should include appropriate control groups. Case-control and prospective studies should evaluate the natural history of both night sweats patterns and their association with quality and length of life. The potential value of night sweats as a clue to the early diagnosis of important under-recognized pathologies, such as sleep disorders and panic attacks, should be investigated. Finally, randomized trials of treatments to reduce the frequency, severity, and impact of night sweats should be undertaken once the potential causes have been better elucidated.

Acknowledgments

This research was made possible by a grant from the American Academy of Family Physicians Foundation. We would like to acknowledge the assistance of Lavonne Glover in preparing the manuscript and to the following practicing family physicians and their staff who made time in their busy schedules to collect the data: Nathan Boren, Jo Ann Carpenter, Stephen Cobb, Ed Farrow, Cary Fisher, Helen Franklin, Kurt Frantz, David Hadley, Terrill Hulson, Joe Jamison, Dee Legako, Migy Mathew, Tomas Owens, John Pittman, Mike Pontious, Paul Preslar, R. Scott Stewart, David Strickland, Clinton Strong, Terry Truong, Keith Underhill, Kyle Waugh, Dan Woiwode, Mike Woods, Rick Edwards, Bob C. Jones, Leah R. Mabry, Tom Mueller, Mike Ragsdale, Hugh Wilson, Frank D. Wright, and Samuel T. Coleridge.

Night sweats have been attributed to tuberculosis, other acute and chronic febrile illnesses, menopause, pregnancy, hyperthyroidism, nocturnal hypoglycemia, other endocrine problems, neurologic diseases, sleep disorders (eg, sleep apnea and nightmares), malignancies, autoimmune diseases, coronary artery spasm, congestive heart failure, gastroesophageal reflux disease, psychiatric disorders, and certain medications. In 36 medical and surgical textbooks, night sweats were always discussed within sections covering specific diseases and never as a separate topic. References to the primary literature were never provided. We also searched Micromedix, a comprehensive source of information on medications, using “sweating” and “diaphoresis” as search terms.¹ Table W1 contains a comprehensive list of proposed causes of night sweats identified in our searches and accompanying references.

Only 2 epidemiologic studies of night sweats were found in the English language literature. Lea and Aber² interviewed 174 patients randomly selected from the inpatient units of a university hospital and found that 33% of nonobstetric patients and 60% of obstetric patients reported having had night sweats during the previous 3 months. Twenty-six percent of those with night sweats reported that their nighttime sweating was severe enough to require bathing and changing of bed linens. Reynolds,³ a gastroenterologist, queried 200 consecutive patients seen in his outpatient practice and found that 40% remembered experiencing night sweats at least once during the previous year. A total of 12% reported at least weekly night sweats. A review of the records of 750 patients at the Geriatric Continuity Clinic at the University of Oklahoma Family Medicine Center revealed that 10% reported having experienced night sweats during the previous month, when the question was asked as part of a standard review of systems questionnaire (J.W.M., unpublished data, 1999).

Our study was conducted in an effort to estimate the prevalence of night sweats in adult patients seen in primary care office settings, and to explore the associations of this symptom with demographic factors, physical characteristics, medical problems, and medications. We also sought to determine how distressing this symptom is to those who have it and to their sleep partners, whether patients are likely to report the symptom to their physicians, and what patients and their physicians think causes night sweats in individual cases.

Methods

Physician members of the Oklahoma Physicians Resource/Research Network (OKPRN) and the Texas Academy of Family Physicians Research Network (TAFP-Net) enrolled consecutive patients 18 years and older seen in their clinics during a 1-week period in the summer and a second 1-week period in the winter in the years 2000 and 2001. Patients who agreed to participate signed a consent form and then helped the nurse and physician complete a brief questionnaire on a preaddressed, stamped data collection card. For those who declined to participate, a card was generated containing the physician’s code number and the patient’s age and sex. Questions elicited demographic information; information about a selected set of medical conditions; medications, vitamins, herbs, and alcohol used regularly; and information about recent experiences with night sweats. Participating physicians were asked to check the questionnaires for accuracy and to record their opinions regarding the cause of the patients’ night sweats when they reported having had them. A laminated card with definitions of terms was provided to each physician.

“Night sweats” was defined as “sweating at night even when it isn’t excessively hot in your bedroom.” “Day sweats” was defined as “excessive sweating during the daytime.” “Pure night sweats” was defined as night sweats, but not day sweats, and “night and day sweats” as the combination of the 2. The time interval was specified as “during the last month.”

Completed questionnaires were mailed to the Oklahoma Center for Family Medicine Research for data entry and analysis. The data collection cards used by the Texas network included questions about race/ethnicity and panic attacks that were not included on the Oklahoma cards. Inadvertently, some of the Texas cards did not include the question about daytime sweating.

Statistix7 (Analytical Software, Tallahassee, Fla) was used for all statistical analyses. Medications were assigned to 1 of 47 categories according to their primary pharmacologic effects. Summary statistics were calculated for all participants and for the following subgroups: season (summer and winter), pattern of night sweats (excessive nighttime sweating only or night and day sweats), and age group. We anticipated that the majority of women with menopausal symptoms would be in the 41- to 55-year age group.

The two patterns of night sweats, “pure night sweats” and “night and day sweats,” were analyzed separately, and by sex and age. Logistic regression was used to identify the most significant predictors of night sweats while controlling for other variables. Variables were entered into the logistic models if they had a univariate association with the dependent variable at a P value of less than .05. They were then removed one at a time, in the order of largest to smallest P value, if they had a P value of greater than .01 after controlling for other variables. Conservative P values were chosen because of the large numbers of variables considered, in order to reduce the probability of type 1 errors. When appropriate, 95% confidence intervals were calculated.

Results

Study population

A total of 2267 patients of 31 different physicians participated in this study, including 1888 patients of 24 Oklahoma physicians and 379 patients of 7 Texas physicians. Their mean (standard deviation) age was 50.7 (18.8) years, with a range of 18 to 97 years. Sixty-nine percent were women. A total of 99% of Oklahoma patients and 93% of Texas patients seen during the study weeks agreed to participate in the study. Among Texas participants, 53% were Hispanic whites, 33% were non-Hispanic whites, 13% were African Americans, and 1% were categorized as other. On the basis of prior OKPRN studies, we suspect that approximately 90% of Oklahoma patients were non-Hispanic whites, but exact proportions were not determined for this study.

Prevalence of night sweats

The prevalence of pure night sweats, night and day sweats, and any night sweats are shown in Table 1. While the prevalence of night and day sweats was lower for older patients, severity tended to be greater. Severity and frequency were positively correlated for all categories of night sweats and for all subgroups of patients (overall Spearman coefficient = 0.33; P < .001). Overall, the frequencies of night sweats among those who reported the condition were: almost never, 18%; 1 to 3 nights per month, 38%; 1 to 3 nights per week, 27%; and 4 to 7 nights per week, 16%. Ten percent of both women and men with night sweats said that their night sweats were bothersome to others.

TABLE 1
Percentage of patients with pure night sweats and night and day

Frequency of reporting of night sweats

A minority of patients with night sweats (12%) had reported the symptom to their physicians. This was true even for those with severe night sweats (46%). Women younger than 70 years were more likely than men of the same age to have reported their night sweats to their physicians (15% vs. 6%; P < .001). The reverse was true for those 70 years and older (7% vs 13%; P =.08). Older patients with pure night sweats were more likely than younger patients to have reported them. After controlling for other variables, patients who were older (odds ratio [OR] = 1.03 per year of age; P < .001), those with night and day sweats (OR = 1.74; P =.0015), and those who reported that their night sweats bothered others (OR = 2.89; P =.001) were more likely to have reported the symptom to their physicians. Those who had reported their night sweats were also more likely to have hot flashes (OR = 2.98; P < .001) and to take estrogen (OR = 1.72; P =.003).

Factors associated with night sweats

The only variable associated with pure night sweats after controlling for all other variables was panic attacks. Variables associated with night and day sweats were younger age, greater body mass index, hot flashes, chronic infection, sleep disturbances, selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressants, “other” (non–SSRI, non-tricyclic) antidepressants, and xanthines.

For women, the only variable clearly associated with pure night sweats in the multivariate model was hot flashes. Panic attacks nearly reached significance (P =.026) and improved the regression model substantially (deviance reduced from 1446 to 87). Variables associated with night and day sweats were weight, sleep problems, hot flashes, antihistamines, SSRIs, and other (non–SSRI, nontricyclic) antidepressants.

For men, the only variable associated with pure night sweats after controlling for other variables was sleep problems. After exclusion of sleep problems and sedatives from the model on the assumption that they might be the result rather than the cause of night sweats, significant predictors were hot flashes (OR = 2.70; 95% confidence interval [CI], 1.35-5.40; P =.005) and regular use of multivitamins (OR = 1.87; 95% CI, 1.17-2.99; P =.009). Variables associated with night and day sweats included greater weight, hot flashes, and greater alcohol use. The ORs and CIs are shown in Table 1.

Interestingly, 32 men (5%) reported hot flashes, and those who did were more likely to report night sweats of both types. Men with hot flashes were evenly distributed across age categories. Their night sweats were more frequent, but not more severe, and they were more likely to bother others than those without hot flashes. Men with hot flashes were more likely to have told their physicians about their night sweats. After controlling for other variables, men with hot flashes were much more likely to have panic attacks (OR = 28.28; P < .001).

Patients 70 years and older made up 19.5% of our sample (N=429). The only factor associated with pure night sweats in the multivariate model was sleep disturbances (OR = 2.04; = 95% CI, 1.21-3.42; P =.007). Exclusion of sleep disturbances left no associated variables. Variables associated with night and day sweats were hot flashes (OR = 15.14; = 95% CI, 6.43-35.68; P < .001) and corticosteroids (OR = 5.45; 95% CI 1.58-18.86; P =.007).

Suspected causes

In cases where patients reported night sweats, only 19% of the patients and 18% of their physicians recorded opinions regarding causation. The suspected causes listed by patients and physicians were similar. Both groups listed menopause most frequently (48% and 44%, respectively). Other etiologies proposed were stress (12% and 8%) and medications (9% and 10%). Physicians listed diabetes as a possible cause in 11% of cases while only 4% of patients listed it. Other suspected causes included obesity, pregnancy, gastroesophageal reflux disease, sleep discomforts, and ambient temperature.

TABLE 2
Associations between independent variables and night sweats in men and women after using logistic regression modeling to control for all other variables

Discussion

As far as we know, this is the first systematic study of night sweats in a primary care population. It is exploratory in nature, and, because of its cross-sectional design, no firm conclusions can be drawn about causation.

Both pure night sweats and night and day sweats are extremely common, with a peak prevalence in men and women 41 to 55 years of age. In contrast to pure night sweats, night and day sweats are experienced infrequently by patients 70 years and older. The factors associated with pure night sweats are somewhat different than those associated with night and day sweats, suggesting different, though probably overlapping, sets of causes. The different associations seen for men and women, and for older and younger patients, are also noteworthy. Patients often fail to report night sweats to their primary care physician, even when frequent and severe, associated with sleep disturbances, or bothersome to others.

Because of the sampling method (ie, consecutive patients rather than a random sample of active patients), the prevalence estimates reflect the frequency at which physicians can expect to encounter patients with this symptom, rather than the prevalence of night sweats among active patients. Since patients with more symptoms probably see physicians more often, we assume we have overestimated the true prevalence of night sweats in the larger population. Participating physicians were also not selected randomly. It is impossible to know how this may have affected our results.

We were surprised that so few of our independent variables were associated with pure night sweats: only panic attacks (all patients), sleep disorders (men and older patients), and hot flashes (women). Factors not associated with pure night sweats included obesity; diabetes, insulin, or oral hypoglycemic agents; acute or chronic infections; gastroesophageal reflux disease; or thyroid medications. Pure night sweats were also not specifically associated with estrogen and progesterone, although they were associated with hot flashes. There was also no association of pure night sweats and alcohol consumption.

The fact that physicians and their patients could only speculate on a cause for night sweats in 1 out of 5 cases suggests a lack of familiarity with the multitude of suspected causes, a failure to detect certain common causes (eg, sleep disorders and panic attacks), or, most likely, that many common causes of night sweats have yet to be elucidated. If the last is correct, it may be an example of the bias in the primary and secondary clinical literature that occurs when clinical research is carried out primarily in the subspecialty clinics of academic medical centers.^4-7 Our findings speak to the need for greater support for primary care practice-based research.^8,9

In retrospect, the omission of the variable “panic attacks” from the Oklahoma cards was a mistake, since this variable was correlated with pure night sweats in women. It may have been more strongly associated with pure night sweats in men as well, if the number of respondents to this question had been larger. Also, some men complained of hot flashes, and when they did, they were more likely to have night sweats and panic attacks, suggesting that both hot flashes and night sweats in men should prompt physicians to ask additional questions about panic disorder. Although race was also omitted from the Oklahoma cards, this variable did not seem to be associated with differences in night sweats prevalence or association among those for whom this information was available.

The definition and description of night sweats used in this study were arbitrary and may have influenced the prevalence rates obtained. We attempted to exclude environmental temperature as a cause. Although the definitions provided clearly stated “within the last month,” the data collection cards did not specify a time interval. This may have resulted in some variation in interpretation.

The decisions that were made regarding logistic modeling strategies were conservative and may have excluded some important variables. However, with so many variables and no basis on which to judge a priori, we felt that a conservative approach was best. The decision to include in the models variables (eg, sleep problems and sedatives that might be considered consequences) rather than causes of night sweats, was also arbitrary and may have affected the results. An alternative explanation of the associations found between night sweats and sleep problems is that those who are unable to sleep for other reasons are more likely to notice excessive sweating than those who are asleep.

Future studies should more carefully examine factors found in this study to be associated with night sweats, such as panic attacks and sleep disorders, and other potential etiologic factors not considered, such as tobacco abuse, allergic diseases, migraines, congestive heart failure, and chronic lung disease. Given the high prevalence, future studies examining etiology should include appropriate control groups. Case-control and prospective studies should evaluate the natural history of both night sweats patterns and their association with quality and length of life. The potential value of night sweats as a clue to the early diagnosis of important under-recognized pathologies, such as sleep disorders and panic attacks, should be investigated. Finally, randomized trials of treatments to reduce the frequency, severity, and impact of night sweats should be undertaken once the potential causes have been better elucidated.

Acknowledgments

This research was made possible by a grant from the American Academy of Family Physicians Foundation. We would like to acknowledge the assistance of Lavonne Glover in preparing the manuscript and to the following practicing family physicians and their staff who made time in their busy schedules to collect the data: Nathan Boren, Jo Ann Carpenter, Stephen Cobb, Ed Farrow, Cary Fisher, Helen Franklin, Kurt Frantz, David Hadley, Terrill Hulson, Joe Jamison, Dee Legako, Migy Mathew, Tomas Owens, John Pittman, Mike Pontious, Paul Preslar, R. Scott Stewart, David Strickland, Clinton Strong, Terry Truong, Keith Underhill, Kyle Waugh, Dan Woiwode, Mike Woods, Rick Edwards, Bob C. Jones, Leah R. Mabry, Tom Mueller, Mike Ragsdale, Hugh Wilson, Frank D. Wright, and Samuel T. Coleridge.

Night sweats have been attributed to tuberculosis, other acute and chronic febrile illnesses, menopause, pregnancy, hyperthyroidism, nocturnal hypoglycemia, other endocrine problems, neurologic diseases, sleep disorders (eg, sleep apnea and nightmares), malignancies, autoimmune diseases, coronary artery spasm, congestive heart failure, gastroesophageal reflux disease, psychiatric disorders, and certain medications. In 36 medical and surgical textbooks, night sweats were always discussed within sections covering specific diseases and never as a separate topic. References to the primary literature were never provided. We also searched Micromedix, a comprehensive source of information on medications, using “sweating” and “diaphoresis” as search terms.¹ Table W1 contains a comprehensive list of proposed causes of night sweats identified in our searches and accompanying references.

Only 2 epidemiologic studies of night sweats were found in the English language literature. Lea and Aber² interviewed 174 patients randomly selected from the inpatient units of a university hospital and found that 33% of nonobstetric patients and 60% of obstetric patients reported having had night sweats during the previous 3 months. Twenty-six percent of those with night sweats reported that their nighttime sweating was severe enough to require bathing and changing of bed linens. Reynolds,³ a gastroenterologist, queried 200 consecutive patients seen in his outpatient practice and found that 40% remembered experiencing night sweats at least once during the previous year. A total of 12% reported at least weekly night sweats. A review of the records of 750 patients at the Geriatric Continuity Clinic at the University of Oklahoma Family Medicine Center revealed that 10% reported having experienced night sweats during the previous month, when the question was asked as part of a standard review of systems questionnaire (J.W.M., unpublished data, 1999).

Our study was conducted in an effort to estimate the prevalence of night sweats in adult patients seen in primary care office settings, and to explore the associations of this symptom with demographic factors, physical characteristics, medical problems, and medications. We also sought to determine how distressing this symptom is to those who have it and to their sleep partners, whether patients are likely to report the symptom to their physicians, and what patients and their physicians think causes night sweats in individual cases.

Methods

Physician members of the Oklahoma Physicians Resource/Research Network (OKPRN) and the Texas Academy of Family Physicians Research Network (TAFP-Net) enrolled consecutive patients 18 years and older seen in their clinics during a 1-week period in the summer and a second 1-week period in the winter in the years 2000 and 2001. Patients who agreed to participate signed a consent form and then helped the nurse and physician complete a brief questionnaire on a preaddressed, stamped data collection card. For those who declined to participate, a card was generated containing the physician’s code number and the patient’s age and sex. Questions elicited demographic information; information about a selected set of medical conditions; medications, vitamins, herbs, and alcohol used regularly; and information about recent experiences with night sweats. Participating physicians were asked to check the questionnaires for accuracy and to record their opinions regarding the cause of the patients’ night sweats when they reported having had them. A laminated card with definitions of terms was provided to each physician.

“Night sweats” was defined as “sweating at night even when it isn’t excessively hot in your bedroom.” “Day sweats” was defined as “excessive sweating during the daytime.” “Pure night sweats” was defined as night sweats, but not day sweats, and “night and day sweats” as the combination of the 2. The time interval was specified as “during the last month.”

Completed questionnaires were mailed to the Oklahoma Center for Family Medicine Research for data entry and analysis. The data collection cards used by the Texas network included questions about race/ethnicity and panic attacks that were not included on the Oklahoma cards. Inadvertently, some of the Texas cards did not include the question about daytime sweating.

Statistix7 (Analytical Software, Tallahassee, Fla) was used for all statistical analyses. Medications were assigned to 1 of 47 categories according to their primary pharmacologic effects. Summary statistics were calculated for all participants and for the following subgroups: season (summer and winter), pattern of night sweats (excessive nighttime sweating only or night and day sweats), and age group. We anticipated that the majority of women with menopausal symptoms would be in the 41- to 55-year age group.

The two patterns of night sweats, “pure night sweats” and “night and day sweats,” were analyzed separately, and by sex and age. Logistic regression was used to identify the most significant predictors of night sweats while controlling for other variables. Variables were entered into the logistic models if they had a univariate association with the dependent variable at a P value of less than .05. They were then removed one at a time, in the order of largest to smallest P value, if they had a P value of greater than .01 after controlling for other variables. Conservative P values were chosen because of the large numbers of variables considered, in order to reduce the probability of type 1 errors. When appropriate, 95% confidence intervals were calculated.

Results

Study population

A total of 2267 patients of 31 different physicians participated in this study, including 1888 patients of 24 Oklahoma physicians and 379 patients of 7 Texas physicians. Their mean (standard deviation) age was 50.7 (18.8) years, with a range of 18 to 97 years. Sixty-nine percent were women. A total of 99% of Oklahoma patients and 93% of Texas patients seen during the study weeks agreed to participate in the study. Among Texas participants, 53% were Hispanic whites, 33% were non-Hispanic whites, 13% were African Americans, and 1% were categorized as other. On the basis of prior OKPRN studies, we suspect that approximately 90% of Oklahoma patients were non-Hispanic whites, but exact proportions were not determined for this study.

Prevalence of night sweats

The prevalence of pure night sweats, night and day sweats, and any night sweats are shown in Table 1. While the prevalence of night and day sweats was lower for older patients, severity tended to be greater. Severity and frequency were positively correlated for all categories of night sweats and for all subgroups of patients (overall Spearman coefficient = 0.33; P < .001). Overall, the frequencies of night sweats among those who reported the condition were: almost never, 18%; 1 to 3 nights per month, 38%; 1 to 3 nights per week, 27%; and 4 to 7 nights per week, 16%. Ten percent of both women and men with night sweats said that their night sweats were bothersome to others.

TABLE 1
Percentage of patients with pure night sweats and night and day

Frequency of reporting of night sweats

A minority of patients with night sweats (12%) had reported the symptom to their physicians. This was true even for those with severe night sweats (46%). Women younger than 70 years were more likely than men of the same age to have reported their night sweats to their physicians (15% vs. 6%; P < .001). The reverse was true for those 70 years and older (7% vs 13%; P =.08). Older patients with pure night sweats were more likely than younger patients to have reported them. After controlling for other variables, patients who were older (odds ratio [OR] = 1.03 per year of age; P < .001), those with night and day sweats (OR = 1.74; P =.0015), and those who reported that their night sweats bothered others (OR = 2.89; P =.001) were more likely to have reported the symptom to their physicians. Those who had reported their night sweats were also more likely to have hot flashes (OR = 2.98; P < .001) and to take estrogen (OR = 1.72; P =.003).

Factors associated with night sweats

The only variable associated with pure night sweats after controlling for all other variables was panic attacks. Variables associated with night and day sweats were younger age, greater body mass index, hot flashes, chronic infection, sleep disturbances, selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressants, “other” (non–SSRI, non-tricyclic) antidepressants, and xanthines.

For women, the only variable clearly associated with pure night sweats in the multivariate model was hot flashes. Panic attacks nearly reached significance (P =.026) and improved the regression model substantially (deviance reduced from 1446 to 87). Variables associated with night and day sweats were weight, sleep problems, hot flashes, antihistamines, SSRIs, and other (non–SSRI, nontricyclic) antidepressants.

For men, the only variable associated with pure night sweats after controlling for other variables was sleep problems. After exclusion of sleep problems and sedatives from the model on the assumption that they might be the result rather than the cause of night sweats, significant predictors were hot flashes (OR = 2.70; 95% confidence interval [CI], 1.35-5.40; P =.005) and regular use of multivitamins (OR = 1.87; 95% CI, 1.17-2.99; P =.009). Variables associated with night and day sweats included greater weight, hot flashes, and greater alcohol use. The ORs and CIs are shown in Table 1.

Interestingly, 32 men (5%) reported hot flashes, and those who did were more likely to report night sweats of both types. Men with hot flashes were evenly distributed across age categories. Their night sweats were more frequent, but not more severe, and they were more likely to bother others than those without hot flashes. Men with hot flashes were more likely to have told their physicians about their night sweats. After controlling for other variables, men with hot flashes were much more likely to have panic attacks (OR = 28.28; P < .001).

Patients 70 years and older made up 19.5% of our sample (N=429). The only factor associated with pure night sweats in the multivariate model was sleep disturbances (OR = 2.04; = 95% CI, 1.21-3.42; P =.007). Exclusion of sleep disturbances left no associated variables. Variables associated with night and day sweats were hot flashes (OR = 15.14; = 95% CI, 6.43-35.68; P < .001) and corticosteroids (OR = 5.45; 95% CI 1.58-18.86; P =.007).

Suspected causes

In cases where patients reported night sweats, only 19% of the patients and 18% of their physicians recorded opinions regarding causation. The suspected causes listed by patients and physicians were similar. Both groups listed menopause most frequently (48% and 44%, respectively). Other etiologies proposed were stress (12% and 8%) and medications (9% and 10%). Physicians listed diabetes as a possible cause in 11% of cases while only 4% of patients listed it. Other suspected causes included obesity, pregnancy, gastroesophageal reflux disease, sleep discomforts, and ambient temperature.

TABLE 2
Associations between independent variables and night sweats in men and women after using logistic regression modeling to control for all other variables

Discussion

As far as we know, this is the first systematic study of night sweats in a primary care population. It is exploratory in nature, and, because of its cross-sectional design, no firm conclusions can be drawn about causation.

Both pure night sweats and night and day sweats are extremely common, with a peak prevalence in men and women 41 to 55 years of age. In contrast to pure night sweats, night and day sweats are experienced infrequently by patients 70 years and older. The factors associated with pure night sweats are somewhat different than those associated with night and day sweats, suggesting different, though probably overlapping, sets of causes. The different associations seen for men and women, and for older and younger patients, are also noteworthy. Patients often fail to report night sweats to their primary care physician, even when frequent and severe, associated with sleep disturbances, or bothersome to others.

Because of the sampling method (ie, consecutive patients rather than a random sample of active patients), the prevalence estimates reflect the frequency at which physicians can expect to encounter patients with this symptom, rather than the prevalence of night sweats among active patients. Since patients with more symptoms probably see physicians more often, we assume we have overestimated the true prevalence of night sweats in the larger population. Participating physicians were also not selected randomly. It is impossible to know how this may have affected our results.

We were surprised that so few of our independent variables were associated with pure night sweats: only panic attacks (all patients), sleep disorders (men and older patients), and hot flashes (women). Factors not associated with pure night sweats included obesity; diabetes, insulin, or oral hypoglycemic agents; acute or chronic infections; gastroesophageal reflux disease; or thyroid medications. Pure night sweats were also not specifically associated with estrogen and progesterone, although they were associated with hot flashes. There was also no association of pure night sweats and alcohol consumption.

The fact that physicians and their patients could only speculate on a cause for night sweats in 1 out of 5 cases suggests a lack of familiarity with the multitude of suspected causes, a failure to detect certain common causes (eg, sleep disorders and panic attacks), or, most likely, that many common causes of night sweats have yet to be elucidated. If the last is correct, it may be an example of the bias in the primary and secondary clinical literature that occurs when clinical research is carried out primarily in the subspecialty clinics of academic medical centers.^4-7 Our findings speak to the need for greater support for primary care practice-based research.^8,9

In retrospect, the omission of the variable “panic attacks” from the Oklahoma cards was a mistake, since this variable was correlated with pure night sweats in women. It may have been more strongly associated with pure night sweats in men as well, if the number of respondents to this question had been larger. Also, some men complained of hot flashes, and when they did, they were more likely to have night sweats and panic attacks, suggesting that both hot flashes and night sweats in men should prompt physicians to ask additional questions about panic disorder. Although race was also omitted from the Oklahoma cards, this variable did not seem to be associated with differences in night sweats prevalence or association among those for whom this information was available.

The definition and description of night sweats used in this study were arbitrary and may have influenced the prevalence rates obtained. We attempted to exclude environmental temperature as a cause. Although the definitions provided clearly stated “within the last month,” the data collection cards did not specify a time interval. This may have resulted in some variation in interpretation.

The decisions that were made regarding logistic modeling strategies were conservative and may have excluded some important variables. However, with so many variables and no basis on which to judge a priori, we felt that a conservative approach was best. The decision to include in the models variables (eg, sleep problems and sedatives that might be considered consequences) rather than causes of night sweats, was also arbitrary and may have affected the results. An alternative explanation of the associations found between night sweats and sleep problems is that those who are unable to sleep for other reasons are more likely to notice excessive sweating than those who are asleep.

Future studies should more carefully examine factors found in this study to be associated with night sweats, such as panic attacks and sleep disorders, and other potential etiologic factors not considered, such as tobacco abuse, allergic diseases, migraines, congestive heart failure, and chronic lung disease. Given the high prevalence, future studies examining etiology should include appropriate control groups. Case-control and prospective studies should evaluate the natural history of both night sweats patterns and their association with quality and length of life. The potential value of night sweats as a clue to the early diagnosis of important under-recognized pathologies, such as sleep disorders and panic attacks, should be investigated. Finally, randomized trials of treatments to reduce the frequency, severity, and impact of night sweats should be undertaken once the potential causes have been better elucidated.

Acknowledgments

This research was made possible by a grant from the American Academy of Family Physicians Foundation. We would like to acknowledge the assistance of Lavonne Glover in preparing the manuscript and to the following practicing family physicians and their staff who made time in their busy schedules to collect the data: Nathan Boren, Jo Ann Carpenter, Stephen Cobb, Ed Farrow, Cary Fisher, Helen Franklin, Kurt Frantz, David Hadley, Terrill Hulson, Joe Jamison, Dee Legako, Migy Mathew, Tomas Owens, John Pittman, Mike Pontious, Paul Preslar, R. Scott Stewart, David Strickland, Clinton Strong, Terry Truong, Keith Underhill, Kyle Waugh, Dan Woiwode, Mike Woods, Rick Edwards, Bob C. Jones, Leah R. Mabry, Tom Mueller, Mike Ragsdale, Hugh Wilson, Frank D. Wright, and Samuel T. Coleridge.

From 1% to 4% of full-term infants are readmitted to the hospital for jaundice in the first week of life, representing as many as 109,000 admissions¹ Delayed diagnosis of jaundice puts babies at risk for kernicterus, which had virtually disappeared in the United States but is now on the rise. There are anecdotal reports of 22 full-term infants born in the early 1990s who developed kernicterus after discharge from the hospital within 48 hours of birth.¹ The Joint Commission on Accreditation of Healthcare Organizations (JCAHO) recently issued a Sentinel Event Alert recommending that organizations take steps to raise awareness among neonatal caregivers of the potential for kernicterus and its risk factors by reviewing their current patient care processes with regard to the identification and management of hyperbilirubinemia in newborns and by identifying risk reduction strategies that could enhance the effectiveness of these processes.²

The JCAHO alert cites the American Academy of Pediatrics (AAP) Practice Parameter for Management of Hyperbilirubinemia in the Healthy Term Newborn, which is based on available data and expert consensus, as an example of a guideline for identifying at-risk newborns and their diagnosis and treatment. The AAP guideline suggests checking for jaundice by blanching the skin with digital pressure to reveal its underlying color. The guideline states that clinical assessment must be done in a well-lighted room and suggests that as the bilirubin level rises, the extent of caudal progression may be helpful in quantifying the degree of jaundice.³

The AAP jaundice guideline suggests that the use of an icterometer (transcutaneous jaundice meter) may be helpful in the clinical assessment of jaundice.³ A variety of instruments have been tested in different patient populations.^4-8 A potential role for such devices is their use by parents. The Ingram icterometer (Cascade Health Care Products, Salem, Ore.) is particularly promising because of its low cost ($17) and simplicity.⁵ It is a simple handheld device, made of clear plastic, on which are painted 5 transverse stripes of precise and graded hue. The stripes and spaces between them are 3/16 inch wide and are numbered from 1 (lightest in color) to 5 (darkest). When the icterometer is used, the painted side is pressed against the tip of the infant’s nose until the skin becomes blanched. The yellow color of the blanched skin can then be matched with the yellow stripes on the instrument, and a jaundice score assigned.

The purpose of my study was to determine whether mothers can accurately assess the presence and severity of jaundice in their newborns, both visually and with an icterometer, after hospital discharge. Maternal assessments were compared with bilirubin levels and home health nurse assessments to determine their accuracy. Serum bilirubin levels were used as the reference standard. Maternal comfort with the examination techniques was also assessed.

Methods

This study was approved by the Ramsey (now Regions) Hospital institutional review board. Mothers who gave birth at Regions Hospital in St. Paul, Minn., participated in the study. Mothers on the postpartum ward were invited to participate, but were excluded if they were not proficient in reading English, did not have a telephone, or lived more than 10 miles from the hospital. Infants were excluded if they were in the intensive care nursery, were not discharged on the same day as the mother, or if they received phototherapy. Mothers were advised to follow their health care providers’ instructions about timing for the first follow-up visit, and any provider instructions regarding jaundice.

After obtaining consent, the author or a study assistant showed the mothers how to examine their infants for jaundice by 2 methods. Each mother was instructed to examine her baby in a well-lighted room. First, the mother was shown how to look for jaundice by digitally blanching the skin on the cheek. The mother then documented whether she saw any underlying yellow color on her baby. Next, the mother was shown how to determine the caudal progression of the jaundice and to draw a horizontal line on an illustration of a baby corresponding to where the jaundice ended. The distance from the top of the infant’s head to the line drawn by the mother was used to determine the caudal progression. The mother was then shown how to use the Ingram icterometer and obtain a reading from the baby’s nose. Each mother was given an icterometer and a study booklet to document her examination for a total of 7 days, beginning the day after discharge from the hospital. The study booklet also contained some demographic questions, and questions about the mother’s comfort level with both methods of jaundice assessment. The mother was instructed to return the booklet and icterometer by mail when completed. The mother was sent a $25 gift certificate when the study materials were returned.

Within 7 days of discharge, a home health nurse visited each mother and infant in the home. The nurses were trained in the same methods of clinically assessing jaundice, and they assessed each infant by visually determining the caudal progression and by use of the icterometer. The nurse did not share the results of her examination with the mother. The nurse obtained bilirubin levels from all infants and notified the infants’ health care providers of any bilirubin levels higher than 14 mg/dL.

Standard descriptive statistics were calculated for all variables. Categorical relationships were assessed using kappa and chi-square statistics, as appropriate. All analyses were performed using Statistical Package for Social Sciences for Windows, version 10.0.5.

Results

A total 113 of 177 mothers returned their study packets. Home health nurses visited 96 of the 113 mothers; the other 17 mothers were not visited because they declined the visits or could not be located. Although all babies were to have serum bilirubin levels determined whether or not they appeared jaundiced, only 90 of the 96 infants had the blood test. For the other 6 infants, either insufficient blood was drawn or the mother refused the test. On the day of the nurse’s visits, mothers documented in their study booklets the caudal progression of jaundice (for 56 infants) and icterometer readings (for 55 infants).

The educational levels of the mothers were as follows: 15% completed grade school or less; 40% completed high school; and 45% completed college. The mothers reported being from the following racial and ethnic groups: white, 59%; Hispanic, 16%; black, 14%; Asian, 8%; and other, 3%. A total of 53% of the women were primiparous, 84% completed examination forms for their babies for all 7 days, and 53% assessed their infants as being jaundiced during the study.

On the day of the nurse’s visit, there was moderate agreement between the nurses and the mothers about the presence of jaundice in the infants (= 0.50; P < .001). For those infants with jaundice, there was little agreement on the extent of caudal progression between the nurses and the mothers (correlation = 0.36; P > 0.1), but there was moderate agreement between their icterometer readings (correlation = 0.58; P < .05).

The total serum bilirubin results ranged from 0.8 mg/dL to 18.8 mg/dL, with a mean of 7.4 mg/dL. The mean bilirubin level of infants thought to be jaundiced by their mothers was 11.3 mg/dL, while the mean bilirubin of infants not thought to be jaundiced was 4.8 mg/dL (P < .001).

The mothers’ icterometer readings and determinations of jaundice to the nipple line or below it are compared with bilirubin levels in (Table 1). (Table 2) summarizes the diagnostic accuracy of jaundice extending to the nipple line or below it, and for icterometer readings of 2.5, in identifying bilirubin levels of 12 mg/dL and 17 mg/dL. A bilirubin level of 12 mg/dL is the level at which the AAP guideline suggests considering phototherapy for infants aged 24 to 47 hours, and 17 mg/dL is the level at which phototherapy should be considered for infants older than 72 hours.³

The mothers of the 3 infants with bilirubin levels 17 mg/dL recognized that their infants were jaundiced and determined that the jaundice extended below the nipple line. The icterometer readings obtained by the mothers were 2.5, 3, and 3.5. The corresponding icterometer readings by the nurses were 4.5, 3.5 and 3.

The study booklet contained 6 questions about the mothers’ reactions to the study. Almost all of the mothers (98%) responded that the method for checking for caudal progression of jaundice was explained clearly, and even more (99%) felt the use of the icterometer was explained clearly. A total of 69% of the mothers felt it was “very easy” or “easy” to check for caudal progression, and 80% felt it was “very easy” or “easy” to use the icterometer. Forty-six percent of the mothers reported that checking their babies for jaundice made them “very worried” or “somewhat worried” about their babies’ health. Mothers with less education were significantly more likely to report being worried than mothers with higher education levels (P < .05). However, 93% of the mothers reported that checking their babies for jaundice made them “very reassured” or “somewhat reassured” about their babies’ health.

TABLE 1
Maternal assessment of jaundice, by caudal progression and icterometer readings, compared with serum bilirubin levels

TABLE 2
Diagnostic accuracy of maternal visual assessment of jaundice and of the Ingram icterometer

Discussion

The ability of mothers to detect and respond to jaundice in their newborns after discharge from the hospital has not been previously studied. Opinions about the value of parental education regarding jaundice vary markedly. The AAP recommends that all mothers be able to recognize signs of jaundice before discharge.⁹ Others are skeptical that such education will be helpful: “Experience suggests that asking mothers to observe infants for the development of jaundice is not satisfactory. Despite such instructions, it is difficult for many parents to recognize significant jaundice.”¹⁰

Several studies have documented that jaundice is first seen in the face and progresses caudally to the trunk and extremities.^11-13 These studies also found good correlation between serum bilirubin levels and the advancement of dermal icterus. In a previous study, parents were able to accurately assess the caudal progression of jaundice while their babies were in the hospital.¹⁴ However, the bilirubin levels in that study were relatively low, reflecting the brief hospital stay of most of the infants. In contrast, a recent study concluded that the clinical examination for jaundice by nurses and physicians had poor reliability and only moderate correlation with bilirubin levels.¹⁵ The authors did conclude, however, that finding no jaundice below the nipple line reliably predicted that an infant would have a bilirubin concentration of less than 12.0 mg/dL. In this study, finding no jaundice below the nipple line reliably predicted that an infant would have a bilirubin concentration of less than 17.0 mg/dL.

Because of the relatively small number of infants having bilirubin levels high enough to require potential intervention, the measures of diagnostic accuracy in the tables should be interpreted with caution. However, the results of my study confirm several prior reports that restricting bilirubin testing to infants with icterometer readings 2.5 would have safely eliminated many unnecessary tests.^6,14,16 Although most of the infants in my study were white, the efficacy of the icterometer has also been documented in Asian and black newborns.¹⁷

Previous studies have shown that neonatal jaundice and its treatments are associated with an increased risk of maternal behaviors consistent with the vulnerable child syndrome.^18,19 This syndrome was originally described in 1964 in children whose parents believed that their child had suffered a “close call,” and thereafter perceived the child as vulnerable to serious injury or accident.¹⁸ Frequent blood tests to monitor bilirubin levels, supplementation or replacement of breast milk with formula, the physical separation of the mother and infant because of phototherapy, and prolonged hospitalization may create the impression that the infant is seriously ill, despite reassurances from medical personnel. Therefore, the mothers were asked whether the study itself served as a source of anxiety. Almost half of the mothers in this study reported that checking their babies for jaundice made them very or somewhat worried about their babies’ health. Some of the women must have felt ambivalent, however, because almost all of them (93%) also reported that checking their babies for jaundice made them very or somewhat reassured about their babies’ health. Most of the 48 comments written by the mothers in the study booklets were very positive.

Conclusions

One of the strategies recommended by the JCAHO to reduce the risk of kernicterus is to provide parents with adequate educational materials about newborn infants that include information about jaundice.² The message given to parents should be consistent, and should reassure mothers that most jaundiced infants are basically healthy. My study results suggest that it may also be useful for parents to be shown how to visually assess jaundice or to be given an Ingram icterometer to monitor their infants for jaundice after hospital discharge. Further study is needed to determine the optimal method of parental education about newborn jaundice.

Acknowledgments

This study was funded by a grant from the Ramsey Foundation. The author thanks Laura Lantz, Pamela Ristau, Kim Stone, Annette Swain, Mary Jo Feely, and the nurses at Integrated Home Care for their assistance with this project.

From 1% to 4% of full-term infants are readmitted to the hospital for jaundice in the first week of life, representing as many as 109,000 admissions¹ Delayed diagnosis of jaundice puts babies at risk for kernicterus, which had virtually disappeared in the United States but is now on the rise. There are anecdotal reports of 22 full-term infants born in the early 1990s who developed kernicterus after discharge from the hospital within 48 hours of birth.¹ The Joint Commission on Accreditation of Healthcare Organizations (JCAHO) recently issued a Sentinel Event Alert recommending that organizations take steps to raise awareness among neonatal caregivers of the potential for kernicterus and its risk factors by reviewing their current patient care processes with regard to the identification and management of hyperbilirubinemia in newborns and by identifying risk reduction strategies that could enhance the effectiveness of these processes.²

The JCAHO alert cites the American Academy of Pediatrics (AAP) Practice Parameter for Management of Hyperbilirubinemia in the Healthy Term Newborn, which is based on available data and expert consensus, as an example of a guideline for identifying at-risk newborns and their diagnosis and treatment. The AAP guideline suggests checking for jaundice by blanching the skin with digital pressure to reveal its underlying color. The guideline states that clinical assessment must be done in a well-lighted room and suggests that as the bilirubin level rises, the extent of caudal progression may be helpful in quantifying the degree of jaundice.³

The AAP jaundice guideline suggests that the use of an icterometer (transcutaneous jaundice meter) may be helpful in the clinical assessment of jaundice.³ A variety of instruments have been tested in different patient populations.^4-8 A potential role for such devices is their use by parents. The Ingram icterometer (Cascade Health Care Products, Salem, Ore.) is particularly promising because of its low cost ($17) and simplicity.⁵ It is a simple handheld device, made of clear plastic, on which are painted 5 transverse stripes of precise and graded hue. The stripes and spaces between them are 3/16 inch wide and are numbered from 1 (lightest in color) to 5 (darkest). When the icterometer is used, the painted side is pressed against the tip of the infant’s nose until the skin becomes blanched. The yellow color of the blanched skin can then be matched with the yellow stripes on the instrument, and a jaundice score assigned.

The purpose of my study was to determine whether mothers can accurately assess the presence and severity of jaundice in their newborns, both visually and with an icterometer, after hospital discharge. Maternal assessments were compared with bilirubin levels and home health nurse assessments to determine their accuracy. Serum bilirubin levels were used as the reference standard. Maternal comfort with the examination techniques was also assessed.

Methods

This study was approved by the Ramsey (now Regions) Hospital institutional review board. Mothers who gave birth at Regions Hospital in St. Paul, Minn., participated in the study. Mothers on the postpartum ward were invited to participate, but were excluded if they were not proficient in reading English, did not have a telephone, or lived more than 10 miles from the hospital. Infants were excluded if they were in the intensive care nursery, were not discharged on the same day as the mother, or if they received phototherapy. Mothers were advised to follow their health care providers’ instructions about timing for the first follow-up visit, and any provider instructions regarding jaundice.

After obtaining consent, the author or a study assistant showed the mothers how to examine their infants for jaundice by 2 methods. Each mother was instructed to examine her baby in a well-lighted room. First, the mother was shown how to look for jaundice by digitally blanching the skin on the cheek. The mother then documented whether she saw any underlying yellow color on her baby. Next, the mother was shown how to determine the caudal progression of the jaundice and to draw a horizontal line on an illustration of a baby corresponding to where the jaundice ended. The distance from the top of the infant’s head to the line drawn by the mother was used to determine the caudal progression. The mother was then shown how to use the Ingram icterometer and obtain a reading from the baby’s nose. Each mother was given an icterometer and a study booklet to document her examination for a total of 7 days, beginning the day after discharge from the hospital. The study booklet also contained some demographic questions, and questions about the mother’s comfort level with both methods of jaundice assessment. The mother was instructed to return the booklet and icterometer by mail when completed. The mother was sent a $25 gift certificate when the study materials were returned.

Within 7 days of discharge, a home health nurse visited each mother and infant in the home. The nurses were trained in the same methods of clinically assessing jaundice, and they assessed each infant by visually determining the caudal progression and by use of the icterometer. The nurse did not share the results of her examination with the mother. The nurse obtained bilirubin levels from all infants and notified the infants’ health care providers of any bilirubin levels higher than 14 mg/dL.

Standard descriptive statistics were calculated for all variables. Categorical relationships were assessed using kappa and chi-square statistics, as appropriate. All analyses were performed using Statistical Package for Social Sciences for Windows, version 10.0.5.

Results

A total 113 of 177 mothers returned their study packets. Home health nurses visited 96 of the 113 mothers; the other 17 mothers were not visited because they declined the visits or could not be located. Although all babies were to have serum bilirubin levels determined whether or not they appeared jaundiced, only 90 of the 96 infants had the blood test. For the other 6 infants, either insufficient blood was drawn or the mother refused the test. On the day of the nurse’s visits, mothers documented in their study booklets the caudal progression of jaundice (for 56 infants) and icterometer readings (for 55 infants).

The educational levels of the mothers were as follows: 15% completed grade school or less; 40% completed high school; and 45% completed college. The mothers reported being from the following racial and ethnic groups: white, 59%; Hispanic, 16%; black, 14%; Asian, 8%; and other, 3%. A total of 53% of the women were primiparous, 84% completed examination forms for their babies for all 7 days, and 53% assessed their infants as being jaundiced during the study.

On the day of the nurse’s visit, there was moderate agreement between the nurses and the mothers about the presence of jaundice in the infants (= 0.50; P < .001). For those infants with jaundice, there was little agreement on the extent of caudal progression between the nurses and the mothers (correlation = 0.36; P > 0.1), but there was moderate agreement between their icterometer readings (correlation = 0.58; P < .05).

The total serum bilirubin results ranged from 0.8 mg/dL to 18.8 mg/dL, with a mean of 7.4 mg/dL. The mean bilirubin level of infants thought to be jaundiced by their mothers was 11.3 mg/dL, while the mean bilirubin of infants not thought to be jaundiced was 4.8 mg/dL (P < .001).

The mothers’ icterometer readings and determinations of jaundice to the nipple line or below it are compared with bilirubin levels in (Table 1). (Table 2) summarizes the diagnostic accuracy of jaundice extending to the nipple line or below it, and for icterometer readings of 2.5, in identifying bilirubin levels of 12 mg/dL and 17 mg/dL. A bilirubin level of 12 mg/dL is the level at which the AAP guideline suggests considering phototherapy for infants aged 24 to 47 hours, and 17 mg/dL is the level at which phototherapy should be considered for infants older than 72 hours.³

The mothers of the 3 infants with bilirubin levels 17 mg/dL recognized that their infants were jaundiced and determined that the jaundice extended below the nipple line. The icterometer readings obtained by the mothers were 2.5, 3, and 3.5. The corresponding icterometer readings by the nurses were 4.5, 3.5 and 3.

The study booklet contained 6 questions about the mothers’ reactions to the study. Almost all of the mothers (98%) responded that the method for checking for caudal progression of jaundice was explained clearly, and even more (99%) felt the use of the icterometer was explained clearly. A total of 69% of the mothers felt it was “very easy” or “easy” to check for caudal progression, and 80% felt it was “very easy” or “easy” to use the icterometer. Forty-six percent of the mothers reported that checking their babies for jaundice made them “very worried” or “somewhat worried” about their babies’ health. Mothers with less education were significantly more likely to report being worried than mothers with higher education levels (P < .05). However, 93% of the mothers reported that checking their babies for jaundice made them “very reassured” or “somewhat reassured” about their babies’ health.

TABLE 1
Maternal assessment of jaundice, by caudal progression and icterometer readings, compared with serum bilirubin levels

TABLE 2
Diagnostic accuracy of maternal visual assessment of jaundice and of the Ingram icterometer

Discussion

The ability of mothers to detect and respond to jaundice in their newborns after discharge from the hospital has not been previously studied. Opinions about the value of parental education regarding jaundice vary markedly. The AAP recommends that all mothers be able to recognize signs of jaundice before discharge.⁹ Others are skeptical that such education will be helpful: “Experience suggests that asking mothers to observe infants for the development of jaundice is not satisfactory. Despite such instructions, it is difficult for many parents to recognize significant jaundice.”¹⁰

Several studies have documented that jaundice is first seen in the face and progresses caudally to the trunk and extremities.^11-13 These studies also found good correlation between serum bilirubin levels and the advancement of dermal icterus. In a previous study, parents were able to accurately assess the caudal progression of jaundice while their babies were in the hospital.¹⁴ However, the bilirubin levels in that study were relatively low, reflecting the brief hospital stay of most of the infants. In contrast, a recent study concluded that the clinical examination for jaundice by nurses and physicians had poor reliability and only moderate correlation with bilirubin levels.¹⁵ The authors did conclude, however, that finding no jaundice below the nipple line reliably predicted that an infant would have a bilirubin concentration of less than 12.0 mg/dL. In this study, finding no jaundice below the nipple line reliably predicted that an infant would have a bilirubin concentration of less than 17.0 mg/dL.

Because of the relatively small number of infants having bilirubin levels high enough to require potential intervention, the measures of diagnostic accuracy in the tables should be interpreted with caution. However, the results of my study confirm several prior reports that restricting bilirubin testing to infants with icterometer readings 2.5 would have safely eliminated many unnecessary tests.^6,14,16 Although most of the infants in my study were white, the efficacy of the icterometer has also been documented in Asian and black newborns.¹⁷

Previous studies have shown that neonatal jaundice and its treatments are associated with an increased risk of maternal behaviors consistent with the vulnerable child syndrome.^18,19 This syndrome was originally described in 1964 in children whose parents believed that their child had suffered a “close call,” and thereafter perceived the child as vulnerable to serious injury or accident.¹⁸ Frequent blood tests to monitor bilirubin levels, supplementation or replacement of breast milk with formula, the physical separation of the mother and infant because of phototherapy, and prolonged hospitalization may create the impression that the infant is seriously ill, despite reassurances from medical personnel. Therefore, the mothers were asked whether the study itself served as a source of anxiety. Almost half of the mothers in this study reported that checking their babies for jaundice made them very or somewhat worried about their babies’ health. Some of the women must have felt ambivalent, however, because almost all of them (93%) also reported that checking their babies for jaundice made them very or somewhat reassured about their babies’ health. Most of the 48 comments written by the mothers in the study booklets were very positive.

Conclusions

One of the strategies recommended by the JCAHO to reduce the risk of kernicterus is to provide parents with adequate educational materials about newborn infants that include information about jaundice.² The message given to parents should be consistent, and should reassure mothers that most jaundiced infants are basically healthy. My study results suggest that it may also be useful for parents to be shown how to visually assess jaundice or to be given an Ingram icterometer to monitor their infants for jaundice after hospital discharge. Further study is needed to determine the optimal method of parental education about newborn jaundice.

Acknowledgments

This study was funded by a grant from the Ramsey Foundation. The author thanks Laura Lantz, Pamela Ristau, Kim Stone, Annette Swain, Mary Jo Feely, and the nurses at Integrated Home Care for their assistance with this project.

From 1% to 4% of full-term infants are readmitted to the hospital for jaundice in the first week of life, representing as many as 109,000 admissions¹ Delayed diagnosis of jaundice puts babies at risk for kernicterus, which had virtually disappeared in the United States but is now on the rise. There are anecdotal reports of 22 full-term infants born in the early 1990s who developed kernicterus after discharge from the hospital within 48 hours of birth.¹ The Joint Commission on Accreditation of Healthcare Organizations (JCAHO) recently issued a Sentinel Event Alert recommending that organizations take steps to raise awareness among neonatal caregivers of the potential for kernicterus and its risk factors by reviewing their current patient care processes with regard to the identification and management of hyperbilirubinemia in newborns and by identifying risk reduction strategies that could enhance the effectiveness of these processes.²

The JCAHO alert cites the American Academy of Pediatrics (AAP) Practice Parameter for Management of Hyperbilirubinemia in the Healthy Term Newborn, which is based on available data and expert consensus, as an example of a guideline for identifying at-risk newborns and their diagnosis and treatment. The AAP guideline suggests checking for jaundice by blanching the skin with digital pressure to reveal its underlying color. The guideline states that clinical assessment must be done in a well-lighted room and suggests that as the bilirubin level rises, the extent of caudal progression may be helpful in quantifying the degree of jaundice.³

The AAP jaundice guideline suggests that the use of an icterometer (transcutaneous jaundice meter) may be helpful in the clinical assessment of jaundice.³ A variety of instruments have been tested in different patient populations.^4-8 A potential role for such devices is their use by parents. The Ingram icterometer (Cascade Health Care Products, Salem, Ore.) is particularly promising because of its low cost ($17) and simplicity.⁵ It is a simple handheld device, made of clear plastic, on which are painted 5 transverse stripes of precise and graded hue. The stripes and spaces between them are 3/16 inch wide and are numbered from 1 (lightest in color) to 5 (darkest). When the icterometer is used, the painted side is pressed against the tip of the infant’s nose until the skin becomes blanched. The yellow color of the blanched skin can then be matched with the yellow stripes on the instrument, and a jaundice score assigned.

The purpose of my study was to determine whether mothers can accurately assess the presence and severity of jaundice in their newborns, both visually and with an icterometer, after hospital discharge. Maternal assessments were compared with bilirubin levels and home health nurse assessments to determine their accuracy. Serum bilirubin levels were used as the reference standard. Maternal comfort with the examination techniques was also assessed.

Methods

This study was approved by the Ramsey (now Regions) Hospital institutional review board. Mothers who gave birth at Regions Hospital in St. Paul, Minn., participated in the study. Mothers on the postpartum ward were invited to participate, but were excluded if they were not proficient in reading English, did not have a telephone, or lived more than 10 miles from the hospital. Infants were excluded if they were in the intensive care nursery, were not discharged on the same day as the mother, or if they received phototherapy. Mothers were advised to follow their health care providers’ instructions about timing for the first follow-up visit, and any provider instructions regarding jaundice.

After obtaining consent, the author or a study assistant showed the mothers how to examine their infants for jaundice by 2 methods. Each mother was instructed to examine her baby in a well-lighted room. First, the mother was shown how to look for jaundice by digitally blanching the skin on the cheek. The mother then documented whether she saw any underlying yellow color on her baby. Next, the mother was shown how to determine the caudal progression of the jaundice and to draw a horizontal line on an illustration of a baby corresponding to where the jaundice ended. The distance from the top of the infant’s head to the line drawn by the mother was used to determine the caudal progression. The mother was then shown how to use the Ingram icterometer and obtain a reading from the baby’s nose. Each mother was given an icterometer and a study booklet to document her examination for a total of 7 days, beginning the day after discharge from the hospital. The study booklet also contained some demographic questions, and questions about the mother’s comfort level with both methods of jaundice assessment. The mother was instructed to return the booklet and icterometer by mail when completed. The mother was sent a $25 gift certificate when the study materials were returned.

Within 7 days of discharge, a home health nurse visited each mother and infant in the home. The nurses were trained in the same methods of clinically assessing jaundice, and they assessed each infant by visually determining the caudal progression and by use of the icterometer. The nurse did not share the results of her examination with the mother. The nurse obtained bilirubin levels from all infants and notified the infants’ health care providers of any bilirubin levels higher than 14 mg/dL.

Standard descriptive statistics were calculated for all variables. Categorical relationships were assessed using kappa and chi-square statistics, as appropriate. All analyses were performed using Statistical Package for Social Sciences for Windows, version 10.0.5.

Results

A total 113 of 177 mothers returned their study packets. Home health nurses visited 96 of the 113 mothers; the other 17 mothers were not visited because they declined the visits or could not be located. Although all babies were to have serum bilirubin levels determined whether or not they appeared jaundiced, only 90 of the 96 infants had the blood test. For the other 6 infants, either insufficient blood was drawn or the mother refused the test. On the day of the nurse’s visits, mothers documented in their study booklets the caudal progression of jaundice (for 56 infants) and icterometer readings (for 55 infants).

The educational levels of the mothers were as follows: 15% completed grade school or less; 40% completed high school; and 45% completed college. The mothers reported being from the following racial and ethnic groups: white, 59%; Hispanic, 16%; black, 14%; Asian, 8%; and other, 3%. A total of 53% of the women were primiparous, 84% completed examination forms for their babies for all 7 days, and 53% assessed their infants as being jaundiced during the study.

On the day of the nurse’s visit, there was moderate agreement between the nurses and the mothers about the presence of jaundice in the infants (= 0.50; P < .001). For those infants with jaundice, there was little agreement on the extent of caudal progression between the nurses and the mothers (correlation = 0.36; P > 0.1), but there was moderate agreement between their icterometer readings (correlation = 0.58; P < .05).

The total serum bilirubin results ranged from 0.8 mg/dL to 18.8 mg/dL, with a mean of 7.4 mg/dL. The mean bilirubin level of infants thought to be jaundiced by their mothers was 11.3 mg/dL, while the mean bilirubin of infants not thought to be jaundiced was 4.8 mg/dL (P < .001).

The mothers’ icterometer readings and determinations of jaundice to the nipple line or below it are compared with bilirubin levels in (Table 1). (Table 2) summarizes the diagnostic accuracy of jaundice extending to the nipple line or below it, and for icterometer readings of 2.5, in identifying bilirubin levels of 12 mg/dL and 17 mg/dL. A bilirubin level of 12 mg/dL is the level at which the AAP guideline suggests considering phototherapy for infants aged 24 to 47 hours, and 17 mg/dL is the level at which phototherapy should be considered for infants older than 72 hours.³

The mothers of the 3 infants with bilirubin levels 17 mg/dL recognized that their infants were jaundiced and determined that the jaundice extended below the nipple line. The icterometer readings obtained by the mothers were 2.5, 3, and 3.5. The corresponding icterometer readings by the nurses were 4.5, 3.5 and 3.

The study booklet contained 6 questions about the mothers’ reactions to the study. Almost all of the mothers (98%) responded that the method for checking for caudal progression of jaundice was explained clearly, and even more (99%) felt the use of the icterometer was explained clearly. A total of 69% of the mothers felt it was “very easy” or “easy” to check for caudal progression, and 80% felt it was “very easy” or “easy” to use the icterometer. Forty-six percent of the mothers reported that checking their babies for jaundice made them “very worried” or “somewhat worried” about their babies’ health. Mothers with less education were significantly more likely to report being worried than mothers with higher education levels (P < .05). However, 93% of the mothers reported that checking their babies for jaundice made them “very reassured” or “somewhat reassured” about their babies’ health.

TABLE 1
Maternal assessment of jaundice, by caudal progression and icterometer readings, compared with serum bilirubin levels

TABLE 2
Diagnostic accuracy of maternal visual assessment of jaundice and of the Ingram icterometer

Discussion

The ability of mothers to detect and respond to jaundice in their newborns after discharge from the hospital has not been previously studied. Opinions about the value of parental education regarding jaundice vary markedly. The AAP recommends that all mothers be able to recognize signs of jaundice before discharge.⁹ Others are skeptical that such education will be helpful: “Experience suggests that asking mothers to observe infants for the development of jaundice is not satisfactory. Despite such instructions, it is difficult for many parents to recognize significant jaundice.”¹⁰

Several studies have documented that jaundice is first seen in the face and progresses caudally to the trunk and extremities.^11-13 These studies also found good correlation between serum bilirubin levels and the advancement of dermal icterus. In a previous study, parents were able to accurately assess the caudal progression of jaundice while their babies were in the hospital.¹⁴ However, the bilirubin levels in that study were relatively low, reflecting the brief hospital stay of most of the infants. In contrast, a recent study concluded that the clinical examination for jaundice by nurses and physicians had poor reliability and only moderate correlation with bilirubin levels.¹⁵ The authors did conclude, however, that finding no jaundice below the nipple line reliably predicted that an infant would have a bilirubin concentration of less than 12.0 mg/dL. In this study, finding no jaundice below the nipple line reliably predicted that an infant would have a bilirubin concentration of less than 17.0 mg/dL.

Because of the relatively small number of infants having bilirubin levels high enough to require potential intervention, the measures of diagnostic accuracy in the tables should be interpreted with caution. However, the results of my study confirm several prior reports that restricting bilirubin testing to infants with icterometer readings 2.5 would have safely eliminated many unnecessary tests.^6,14,16 Although most of the infants in my study were white, the efficacy of the icterometer has also been documented in Asian and black newborns.¹⁷

Previous studies have shown that neonatal jaundice and its treatments are associated with an increased risk of maternal behaviors consistent with the vulnerable child syndrome.^18,19 This syndrome was originally described in 1964 in children whose parents believed that their child had suffered a “close call,” and thereafter perceived the child as vulnerable to serious injury or accident.¹⁸ Frequent blood tests to monitor bilirubin levels, supplementation or replacement of breast milk with formula, the physical separation of the mother and infant because of phototherapy, and prolonged hospitalization may create the impression that the infant is seriously ill, despite reassurances from medical personnel. Therefore, the mothers were asked whether the study itself served as a source of anxiety. Almost half of the mothers in this study reported that checking their babies for jaundice made them very or somewhat worried about their babies’ health. Some of the women must have felt ambivalent, however, because almost all of them (93%) also reported that checking their babies for jaundice made them very or somewhat reassured about their babies’ health. Most of the 48 comments written by the mothers in the study booklets were very positive.

Conclusions

One of the strategies recommended by the JCAHO to reduce the risk of kernicterus is to provide parents with adequate educational materials about newborn infants that include information about jaundice.² The message given to parents should be consistent, and should reassure mothers that most jaundiced infants are basically healthy. My study results suggest that it may also be useful for parents to be shown how to visually assess jaundice or to be given an Ingram icterometer to monitor their infants for jaundice after hospital discharge. Further study is needed to determine the optimal method of parental education about newborn jaundice.

Acknowledgments

This study was funded by a grant from the Ramsey Foundation. The author thanks Laura Lantz, Pamela Ristau, Kim Stone, Annette Swain, Mary Jo Feely, and the nurses at Integrated Home Care for their assistance with this project.

Family physicians are strongly encouraged to support and promote breastfeeding, the optimal form of infant nutrition.¹ Despite its known benefits (fewer infant infections^2-6 and decreased maternal risks of premenopausal breast cancer⁷ and post-menopausal hip fractures⁸), only 64% of mothers initiated breastfeeding in 1998⁹ and only 29% of mothers fed their 6-month-old infant by breast, well below the Healthy People 2010 goal of 50% breastfeeding at 6 months.¹⁰ Clearly, determining the factors that influence breastfeeding beyond the early postpartum period would be beneficial.

Returning to work is a consistent risk factor for weaning.^11-14 The impact of early bottle-feeding on the duration of breastfeeding has been studied with less consistent results.^15,20 Insufficient milk supply is a common subjective reason given for termination.^15,19,21,22 Older women and those with a higher level of education are at less risk of early breastfeeding termination.^{9,11,15,16,21,23,24}

Few investigators have described how breastfeeding patterns may affect breastfeeding duration. Little is known about the effects of timing, frequency, and duration of individual breastfeedings, or the roles of breast pain and infection, sleep, and manual expression on early weaning. We studied women who indicated their intent to breastfeed prenatally to identify demographic factors and breastfeeding patterns associated with weaning in the first 12 weeks postpartum.

Methods

Population

We interviewed breastfeeding women by telephone at 3, 6, 9, and 12 weeks postpartum to investigate lactation mastitis risk factors and predictors of weaning. Pregnant women intending to breastfeed were recruited from 2 geographic sites between June 1994 and January 1998. In suburban Detroit, Michigan, women attending orientation at a freestanding birthing center were asked to participate. In Omaha, Nebraska, women at a single large company were recruited when applying for maternity leave.

Data collection

During the computer-assisted interview, subjects were asked to recall each of the previous 3 weeks. The initial interview, which collected demographic information, typically lasted 15 to 20 minutes; subsequent interviews were shorter. The survey addressed breastfeeding practices and recent health events. Exclusive breastfeeders were women who fed their infants only by breast. We did not collect information on pacifiers; therefore, exclusively breastfed infants may have also received pacifiers. Women who manually expressed or used a device to assist in expression were classified as “pumping” their breasts. Respondents were asked if they had bottle-fed the infant; they were not asked about bottle contents or volume.

Subjects were queried on potential difficulties including breast or nipple pain while nursing, nipple cracks, and mastitis (diagnosed by a health care provider), as well as other health problems and behaviors. Subjects who had stopped breastfeeding in the previous 3 weeks were asked when and why, given a list of possible explanations and an open-ended opportunity. Respondents could provide multiple reasons for termination.

Data analysis

Kaplan-Meier estimates describe the distribution of weaning times for the 2 sites. A log-rank test was used to assess group differences. Relationships between demographic factors and time of weaning were assessed by Cox regression analysis. Discrete survival analysis was used to determine whether variables measured on a weekly basis were related to breastfeeding cessation. Hazard ratios describe the association of the exposures between women who stopped breastfeeding at a given time and those who continued. Because breastfeeding cessation was a rare event in later weeks of the study, as were certain clinical or behavioral breastfeeding factors, weeks 4-12 were collapsed into a single interval. Two variables, number of daily feedings and duration of each feeding, were examined only in the first 3 weeks because the information was often missing beyond 3 weeks. All analyses were performed using the Statistical Package for the Social Sciences.²⁵

Results

Description of subjects

A total of 1057 women agreed to be contacted. Of those, 946 (89.5%) participated in at least 1 interview. Of the 111 women who did not participate, 11 refused and 100 could not be located. Six hundred fifty-eight (69.6%) women completed all 4 interviews. The 56 women who entered the study at week 6 because they could not be reached for the first interview were similar in all factors to women who entered earlier. Of the 946, 711 (75.2%) were from Michigan and 235 (24.8%) were from Nebraska.

Subjects from Michigan were significantly more likely than those from Nebraska to be older than 30 years (52.0% vs 38.3%), have at least a bachelor’s degree (62.9% vs 48.5%), have 3 or more children (38.5% vs 19.6%), and have had a vaginal delivery (99.6% vs 77.0%) (Table W1).* The groups were similar in race, household income, and marital status.

Demographic factors

A total of 673 women (71.1%) continued breastfeeding until 12 weeks; 28% were exclusive breastfeeders. Michigan women were more likely to breastfeed at weeks 2 through 12 than their Nebraskan counterparts (P < .0001, Figure). A college degree was associated with 40% less weaning (Table 1). Age and annual household income were directly related to continued breastfeeding at both sites. Number of children in the household was not associated with termination. Previous breastfeeding experience showed a nonsignificant but consistent trend toward lower weaning risk.

TABLE 1
Relationships of demographics and other characteristics with time to weaning, by site

FIGURE
Probability of breastfeeding, by site, by postpartum week

Clinical and behavioral factors

Because time to weaning differed significantly by site, the survival analyses of clinical and behavioral factors were performed separately for Michigan and Nebraska and controlled for education, age, and previous breastfeeding experience.

During the first 3 weeks, Michigan women with mastitis were nearly 6 times more likely than Michigan women without mastitis to stop breast-feeding in the week of diagnosis (Table 2). Women from Nebraska showed nonsignificant results in the same direction in weeks 4 to 12. (No women from Nebraska with mastitis terminated during weeks 1 through 3.) Although nipple sores and cracks were not associated with weaning, breast pain was associated with weaning. For each day of pain in the first 3 weeks, there was a 10% increase in risk of cessation among Michigan women and a 26% increase among Nebraska women. The association between pain and weaning in weeks 4 through 12 is less clear. In these later weeks, women who reported pain were unexpectedly 75% to 80% more likely to continue breastfeeding than women who did not report pain, yet for Nebraska women the number of days with pain remained significantly associated with breastfeeding cessation.

Subjective depression and breastfeeding cessation were not related. The association between daily sleep and weaning varied by site. During weeks 4 through 12, Michigan women with more daily sleep were less likely to terminate. An opposite, but marginally significant trend, was observed for Nebraska women. Weaning was not associated with outside household help. Nonvaginal birth was not associated with weaning for Nebraska women. (There were only 2 cesarean sections in the Michigan group.)

Michigan women who expressed breast milk during the first 3 weeks were twice as likely to stop breastfeeding as those who did not pump. During the same period, Michigan women who used a bottle for some feedings were 9 times more likely to wean than nonbottle users. Respondents in Nebraska showed similar nonsignificant trends in the first 3 weeks. By contrast, during weeks 4 through 12, both Nebraska and Michigan women who pumped were about 75% less likely to wean, while women who used a bottle for some feedings were 98% less likely to stop breastfeeding.

Breast milk expression increased gradually over time, from 30% of women pumping an average of 3 times per day in the first 3 weeks to 45% of women pumping 5 times per day in the last 3 weeks. To determine if pumping and bottle-feeding had an effect independent of pain or mastitis on weaning in the first 3 weeks, we performed additional analyses controlling for pain, cracks and sores, and mastitis in the same week. The results were similar to those presented in Table 2. Michigan women who pumped were 3 times more likely to wean than those who did not pump (hazard ratio [HR] = 3.0, 95% confidence interval [CI], 1.3 - 6.7), while for Nebraska women there was no association between pumping and weaning (HR = 0.6, 95% CI, 0.3 - 1.5). Bottle-feeding was again significantly associated with weaning in weeks 1 through 3 for Michigan women (HR = 10.9, 95% CI, 4.5 - 26.7) and not associated in Nebraskans (HR = 0.8, 95% CI, 0.4 - 2.0).

Duration and frequency of feedings were investigated as weaning risk factors. There appeared to be a threshold for both variables during the first 3 weeks in Michigan women. Michigan women who breastfed less than 10 minutes per feeding were nearly 5 times more likely to stop breastfeeding than women who breastfed longer. Michigan women who breastfed 6 or fewer times per day were 8 times more likely to stop than those who breastfed more often. Results for Nebraska women fell in the same direction but were not statistically significant.

TABLE 2
Relationships of clinical and behavioral factors to breastfeeding cessation in the same week, adjusted for mother’s age, education, and previous breastfeeding experience

Subjective factors

At each interview, women who had stopped breastfeeding in the previous 3 weeks were asked why they had made that decision. Most women (75%) provided only one reason. At the first interview, insufficient milk supply (37.3%) and breast pain or mastitis (32.9%) were the most common reasons for termination (Table 3). Insufficient milk supply was the reason most often given (35.0%) during weeks 4 through 6. At both weeks 9 and 12, return to work was the reason given most often (53.1% and 58.3%, respectively).

TABLE 3
Percentage of women citing given reason for termination of breastfeeding

Discussion

Mastitis, pain, and days with pain in the first 3 weeks were important clinical factors associated with breastfeeding cessation in this cohort of women who prenatally self-identified as intending to breastfeed. Women who intend to breastfeed should be counseled regarding these possible complications, their temporary nature, prevention, and treatment. Mastitis is not an indication for breastfeeding termination; in fact, increased feedings and milk expression are considered treatment.^26,27 Women who reported pain the first 3 weeks were more likely to stop breastfeeding than women who reported pain after the first 3 weeks. It is difficult to explain this finding; perhaps there are women who have pain during their entire breastfeeding career and yet continue to breastfeed because they are more pain-tolerant, have less severe or frequent pain than those who wean, or are more committed to breastfeeding.

Other clinical factors investigated were depression and daily sleep hours. Weaning was not associated with subjective depression. However, subjects did not undergo formal psychological testing as in the study that reported an association.²⁴ The relationship between daily sleep hours and termination was not consistent, and likely not clinically significant.

The demographic risk factors related to breast-feeding termination in our study are similar to those previously reported,^{14,15,20,21,23,24} namely, younger maternal age and lower educational level. Investigations of parity have been inconsistent.^16,28 We found no association of weaning with parity. Prior breastfeeding experience has been reported as improving breastfeeding rates^15,28; our results are consistent with those findings, but not significantly so. All subjects had access to prenatal breastfeeding education and postnatal breastfeeding support, which may have diminished the differences between women with breastfeeding experience and those without experience.²⁰

Michigan and Nebraska women who pumped or bottle-fed during weeks 4 through 12 were significantly less likely to terminate breastfeeding. In contrast, Michigan women who pumped or bottle-fed during the first 3 weeks postpartum were more likely to terminate even after controlling for pain and mastitis. A commitment to exclusive breastfeeding may be necessary in the early postpartum period for long-term success.^15,19 To our knowledge, the seemingly protective effect associated with pumping and bottle-feeding after the first 3 weeks has not been previously reported.

Breastfeeding 6 or fewer times per day and feedings of 10 minutes or less were associated with termination during the first 3 weeks. Other studies also indicate that the ratio of breast to bottle feedings is important for long-term success. Feinstein and colleagues¹⁵ found that more than one daily bottle of formula supplementation was associated with shorter breastfeeding duration, which was minimized if there were 7 or more breastfeedings per day. Another study found no weaning difference between women who offered their infant only one bottle daily during weeks 2 through 6 and a bottle-avoiding group.¹⁷

The most frequent reasons given for termination were similar to those reported by others, namely, insufficient milk supply and return to work.^11-15,21,22 Insufficient milk supply was a more common reason in the first few weeks after birth; return to work became an increasingly common reason after week 6.

We were unable to examine the role of pacifiers or smoking in breastfeedng termination because pacifier information was not collected and there were too few smokers for meaningful analysis. Smoking has been consistently reported as associated with early cessation.^15,20,29,30 Although pacifier use does not appear to be directly related,^31,32 it has been proposed as a marker for breastfeeding problems. The homogeneity of the sample limits our ability to make generalizations regarding other populations, such as women of color. However, the large sample size and the similarity of termination risk factors between 2 different populations of women lend confidence to our conclusions. As we did not assess mothers’ intentions, some of the variables found associated with termination might be intentional activities of weaning rather than risk factors for termination. The significant difference in termination risk between the sites also may be related to mothers’ intentions or level of commitment. The Michigan women may have intended to breastfeed longer from the outset. The Michigan recruitment site was an alternative birthing center. Women being delivered there may be more persistent in their breast-feeding efforts. Both sites provided access to breast-feeding support personnel, but the Michigan women, as a group, may have been more motivated to continue.

Our results provide clinically useful information. Additional support may be needed for younger and less educated women. Special efforts should be made for early diagnosis and treatment of mastitis and breast pain, particularly during the first 3 weeks. Exclusive breastfeeding without bottle supplementation should be recommended for the first 3 weeks, with at least 7 feedings per day. Each feeding should preferably last more than 10 minutes.

These results should also reassure breastfeeding women and their providers regarding the use of bottles. Bottle-feeding after 3 weeks does not appear to jeopardize breastfeeding success up to 12 weeks and may even improve it.

* Table W1 appears on the JFP Web site at www.jfponline.com.

Acknowledgments

This study was supported by National Institutes of Health grant #30866.

Family physicians are strongly encouraged to support and promote breastfeeding, the optimal form of infant nutrition.¹ Despite its known benefits (fewer infant infections^2-6 and decreased maternal risks of premenopausal breast cancer⁷ and post-menopausal hip fractures⁸), only 64% of mothers initiated breastfeeding in 1998⁹ and only 29% of mothers fed their 6-month-old infant by breast, well below the Healthy People 2010 goal of 50% breastfeeding at 6 months.¹⁰ Clearly, determining the factors that influence breastfeeding beyond the early postpartum period would be beneficial.

Returning to work is a consistent risk factor for weaning.^11-14 The impact of early bottle-feeding on the duration of breastfeeding has been studied with less consistent results.^15,20 Insufficient milk supply is a common subjective reason given for termination.^15,19,21,22 Older women and those with a higher level of education are at less risk of early breastfeeding termination.^{9,11,15,16,21,23,24}

Few investigators have described how breastfeeding patterns may affect breastfeeding duration. Little is known about the effects of timing, frequency, and duration of individual breastfeedings, or the roles of breast pain and infection, sleep, and manual expression on early weaning. We studied women who indicated their intent to breastfeed prenatally to identify demographic factors and breastfeeding patterns associated with weaning in the first 12 weeks postpartum.

Methods

Population

We interviewed breastfeeding women by telephone at 3, 6, 9, and 12 weeks postpartum to investigate lactation mastitis risk factors and predictors of weaning. Pregnant women intending to breastfeed were recruited from 2 geographic sites between June 1994 and January 1998. In suburban Detroit, Michigan, women attending orientation at a freestanding birthing center were asked to participate. In Omaha, Nebraska, women at a single large company were recruited when applying for maternity leave.

Data collection

During the computer-assisted interview, subjects were asked to recall each of the previous 3 weeks. The initial interview, which collected demographic information, typically lasted 15 to 20 minutes; subsequent interviews were shorter. The survey addressed breastfeeding practices and recent health events. Exclusive breastfeeders were women who fed their infants only by breast. We did not collect information on pacifiers; therefore, exclusively breastfed infants may have also received pacifiers. Women who manually expressed or used a device to assist in expression were classified as “pumping” their breasts. Respondents were asked if they had bottle-fed the infant; they were not asked about bottle contents or volume.

Subjects were queried on potential difficulties including breast or nipple pain while nursing, nipple cracks, and mastitis (diagnosed by a health care provider), as well as other health problems and behaviors. Subjects who had stopped breastfeeding in the previous 3 weeks were asked when and why, given a list of possible explanations and an open-ended opportunity. Respondents could provide multiple reasons for termination.

Data analysis

Kaplan-Meier estimates describe the distribution of weaning times for the 2 sites. A log-rank test was used to assess group differences. Relationships between demographic factors and time of weaning were assessed by Cox regression analysis. Discrete survival analysis was used to determine whether variables measured on a weekly basis were related to breastfeeding cessation. Hazard ratios describe the association of the exposures between women who stopped breastfeeding at a given time and those who continued. Because breastfeeding cessation was a rare event in later weeks of the study, as were certain clinical or behavioral breastfeeding factors, weeks 4-12 were collapsed into a single interval. Two variables, number of daily feedings and duration of each feeding, were examined only in the first 3 weeks because the information was often missing beyond 3 weeks. All analyses were performed using the Statistical Package for the Social Sciences.²⁵

Results

Description of subjects

A total of 1057 women agreed to be contacted. Of those, 946 (89.5%) participated in at least 1 interview. Of the 111 women who did not participate, 11 refused and 100 could not be located. Six hundred fifty-eight (69.6%) women completed all 4 interviews. The 56 women who entered the study at week 6 because they could not be reached for the first interview were similar in all factors to women who entered earlier. Of the 946, 711 (75.2%) were from Michigan and 235 (24.8%) were from Nebraska.

Subjects from Michigan were significantly more likely than those from Nebraska to be older than 30 years (52.0% vs 38.3%), have at least a bachelor’s degree (62.9% vs 48.5%), have 3 or more children (38.5% vs 19.6%), and have had a vaginal delivery (99.6% vs 77.0%) (Table W1).* The groups were similar in race, household income, and marital status.

Demographic factors

A total of 673 women (71.1%) continued breastfeeding until 12 weeks; 28% were exclusive breastfeeders. Michigan women were more likely to breastfeed at weeks 2 through 12 than their Nebraskan counterparts (P < .0001, Figure). A college degree was associated with 40% less weaning (Table 1). Age and annual household income were directly related to continued breastfeeding at both sites. Number of children in the household was not associated with termination. Previous breastfeeding experience showed a nonsignificant but consistent trend toward lower weaning risk.

TABLE 1
Relationships of demographics and other characteristics with time to weaning, by site

FIGURE
Probability of breastfeeding, by site, by postpartum week

Clinical and behavioral factors

Because time to weaning differed significantly by site, the survival analyses of clinical and behavioral factors were performed separately for Michigan and Nebraska and controlled for education, age, and previous breastfeeding experience.

During the first 3 weeks, Michigan women with mastitis were nearly 6 times more likely than Michigan women without mastitis to stop breast-feeding in the week of diagnosis (Table 2). Women from Nebraska showed nonsignificant results in the same direction in weeks 4 to 12. (No women from Nebraska with mastitis terminated during weeks 1 through 3.) Although nipple sores and cracks were not associated with weaning, breast pain was associated with weaning. For each day of pain in the first 3 weeks, there was a 10% increase in risk of cessation among Michigan women and a 26% increase among Nebraska women. The association between pain and weaning in weeks 4 through 12 is less clear. In these later weeks, women who reported pain were unexpectedly 75% to 80% more likely to continue breastfeeding than women who did not report pain, yet for Nebraska women the number of days with pain remained significantly associated with breastfeeding cessation.

Subjective depression and breastfeeding cessation were not related. The association between daily sleep and weaning varied by site. During weeks 4 through 12, Michigan women with more daily sleep were less likely to terminate. An opposite, but marginally significant trend, was observed for Nebraska women. Weaning was not associated with outside household help. Nonvaginal birth was not associated with weaning for Nebraska women. (There were only 2 cesarean sections in the Michigan group.)

Michigan women who expressed breast milk during the first 3 weeks were twice as likely to stop breastfeeding as those who did not pump. During the same period, Michigan women who used a bottle for some feedings were 9 times more likely to wean than nonbottle users. Respondents in Nebraska showed similar nonsignificant trends in the first 3 weeks. By contrast, during weeks 4 through 12, both Nebraska and Michigan women who pumped were about 75% less likely to wean, while women who used a bottle for some feedings were 98% less likely to stop breastfeeding.

Breast milk expression increased gradually over time, from 30% of women pumping an average of 3 times per day in the first 3 weeks to 45% of women pumping 5 times per day in the last 3 weeks. To determine if pumping and bottle-feeding had an effect independent of pain or mastitis on weaning in the first 3 weeks, we performed additional analyses controlling for pain, cracks and sores, and mastitis in the same week. The results were similar to those presented in Table 2. Michigan women who pumped were 3 times more likely to wean than those who did not pump (hazard ratio [HR] = 3.0, 95% confidence interval [CI], 1.3 - 6.7), while for Nebraska women there was no association between pumping and weaning (HR = 0.6, 95% CI, 0.3 - 1.5). Bottle-feeding was again significantly associated with weaning in weeks 1 through 3 for Michigan women (HR = 10.9, 95% CI, 4.5 - 26.7) and not associated in Nebraskans (HR = 0.8, 95% CI, 0.4 - 2.0).

Duration and frequency of feedings were investigated as weaning risk factors. There appeared to be a threshold for both variables during the first 3 weeks in Michigan women. Michigan women who breastfed less than 10 minutes per feeding were nearly 5 times more likely to stop breastfeeding than women who breastfed longer. Michigan women who breastfed 6 or fewer times per day were 8 times more likely to stop than those who breastfed more often. Results for Nebraska women fell in the same direction but were not statistically significant.

TABLE 2
Relationships of clinical and behavioral factors to breastfeeding cessation in the same week, adjusted for mother’s age, education, and previous breastfeeding experience

Subjective factors

At each interview, women who had stopped breastfeeding in the previous 3 weeks were asked why they had made that decision. Most women (75%) provided only one reason. At the first interview, insufficient milk supply (37.3%) and breast pain or mastitis (32.9%) were the most common reasons for termination (Table 3). Insufficient milk supply was the reason most often given (35.0%) during weeks 4 through 6. At both weeks 9 and 12, return to work was the reason given most often (53.1% and 58.3%, respectively).

TABLE 3
Percentage of women citing given reason for termination of breastfeeding

Discussion

Mastitis, pain, and days with pain in the first 3 weeks were important clinical factors associated with breastfeeding cessation in this cohort of women who prenatally self-identified as intending to breastfeed. Women who intend to breastfeed should be counseled regarding these possible complications, their temporary nature, prevention, and treatment. Mastitis is not an indication for breastfeeding termination; in fact, increased feedings and milk expression are considered treatment.^26,27 Women who reported pain the first 3 weeks were more likely to stop breastfeeding than women who reported pain after the first 3 weeks. It is difficult to explain this finding; perhaps there are women who have pain during their entire breastfeeding career and yet continue to breastfeed because they are more pain-tolerant, have less severe or frequent pain than those who wean, or are more committed to breastfeeding.

Other clinical factors investigated were depression and daily sleep hours. Weaning was not associated with subjective depression. However, subjects did not undergo formal psychological testing as in the study that reported an association.²⁴ The relationship between daily sleep hours and termination was not consistent, and likely not clinically significant.

The demographic risk factors related to breast-feeding termination in our study are similar to those previously reported,^{14,15,20,21,23,24} namely, younger maternal age and lower educational level. Investigations of parity have been inconsistent.^16,28 We found no association of weaning with parity. Prior breastfeeding experience has been reported as improving breastfeeding rates^15,28; our results are consistent with those findings, but not significantly so. All subjects had access to prenatal breastfeeding education and postnatal breastfeeding support, which may have diminished the differences between women with breastfeeding experience and those without experience.²⁰

Michigan and Nebraska women who pumped or bottle-fed during weeks 4 through 12 were significantly less likely to terminate breastfeeding. In contrast, Michigan women who pumped or bottle-fed during the first 3 weeks postpartum were more likely to terminate even after controlling for pain and mastitis. A commitment to exclusive breastfeeding may be necessary in the early postpartum period for long-term success.^15,19 To our knowledge, the seemingly protective effect associated with pumping and bottle-feeding after the first 3 weeks has not been previously reported.

Breastfeeding 6 or fewer times per day and feedings of 10 minutes or less were associated with termination during the first 3 weeks. Other studies also indicate that the ratio of breast to bottle feedings is important for long-term success. Feinstein and colleagues¹⁵ found that more than one daily bottle of formula supplementation was associated with shorter breastfeeding duration, which was minimized if there were 7 or more breastfeedings per day. Another study found no weaning difference between women who offered their infant only one bottle daily during weeks 2 through 6 and a bottle-avoiding group.¹⁷

The most frequent reasons given for termination were similar to those reported by others, namely, insufficient milk supply and return to work.^11-15,21,22 Insufficient milk supply was a more common reason in the first few weeks after birth; return to work became an increasingly common reason after week 6.

We were unable to examine the role of pacifiers or smoking in breastfeedng termination because pacifier information was not collected and there were too few smokers for meaningful analysis. Smoking has been consistently reported as associated with early cessation.^15,20,29,30 Although pacifier use does not appear to be directly related,^31,32 it has been proposed as a marker for breastfeeding problems. The homogeneity of the sample limits our ability to make generalizations regarding other populations, such as women of color. However, the large sample size and the similarity of termination risk factors between 2 different populations of women lend confidence to our conclusions. As we did not assess mothers’ intentions, some of the variables found associated with termination might be intentional activities of weaning rather than risk factors for termination. The significant difference in termination risk between the sites also may be related to mothers’ intentions or level of commitment. The Michigan women may have intended to breastfeed longer from the outset. The Michigan recruitment site was an alternative birthing center. Women being delivered there may be more persistent in their breast-feeding efforts. Both sites provided access to breast-feeding support personnel, but the Michigan women, as a group, may have been more motivated to continue.

Our results provide clinically useful information. Additional support may be needed for younger and less educated women. Special efforts should be made for early diagnosis and treatment of mastitis and breast pain, particularly during the first 3 weeks. Exclusive breastfeeding without bottle supplementation should be recommended for the first 3 weeks, with at least 7 feedings per day. Each feeding should preferably last more than 10 minutes.

These results should also reassure breastfeeding women and their providers regarding the use of bottles. Bottle-feeding after 3 weeks does not appear to jeopardize breastfeeding success up to 12 weeks and may even improve it.

* Table W1 appears on the JFP Web site at www.jfponline.com.

Acknowledgments

This study was supported by National Institutes of Health grant #30866.

Family physicians are strongly encouraged to support and promote breastfeeding, the optimal form of infant nutrition.¹ Despite its known benefits (fewer infant infections^2-6 and decreased maternal risks of premenopausal breast cancer⁷ and post-menopausal hip fractures⁸), only 64% of mothers initiated breastfeeding in 1998⁹ and only 29% of mothers fed their 6-month-old infant by breast, well below the Healthy People 2010 goal of 50% breastfeeding at 6 months.¹⁰ Clearly, determining the factors that influence breastfeeding beyond the early postpartum period would be beneficial.

Returning to work is a consistent risk factor for weaning.^11-14 The impact of early bottle-feeding on the duration of breastfeeding has been studied with less consistent results.^15,20 Insufficient milk supply is a common subjective reason given for termination.^15,19,21,22 Older women and those with a higher level of education are at less risk of early breastfeeding termination.^{9,11,15,16,21,23,24}

Few investigators have described how breastfeeding patterns may affect breastfeeding duration. Little is known about the effects of timing, frequency, and duration of individual breastfeedings, or the roles of breast pain and infection, sleep, and manual expression on early weaning. We studied women who indicated their intent to breastfeed prenatally to identify demographic factors and breastfeeding patterns associated with weaning in the first 12 weeks postpartum.

Methods

Population

We interviewed breastfeeding women by telephone at 3, 6, 9, and 12 weeks postpartum to investigate lactation mastitis risk factors and predictors of weaning. Pregnant women intending to breastfeed were recruited from 2 geographic sites between June 1994 and January 1998. In suburban Detroit, Michigan, women attending orientation at a freestanding birthing center were asked to participate. In Omaha, Nebraska, women at a single large company were recruited when applying for maternity leave.

Data collection

During the computer-assisted interview, subjects were asked to recall each of the previous 3 weeks. The initial interview, which collected demographic information, typically lasted 15 to 20 minutes; subsequent interviews were shorter. The survey addressed breastfeeding practices and recent health events. Exclusive breastfeeders were women who fed their infants only by breast. We did not collect information on pacifiers; therefore, exclusively breastfed infants may have also received pacifiers. Women who manually expressed or used a device to assist in expression were classified as “pumping” their breasts. Respondents were asked if they had bottle-fed the infant; they were not asked about bottle contents or volume.

Subjects were queried on potential difficulties including breast or nipple pain while nursing, nipple cracks, and mastitis (diagnosed by a health care provider), as well as other health problems and behaviors. Subjects who had stopped breastfeeding in the previous 3 weeks were asked when and why, given a list of possible explanations and an open-ended opportunity. Respondents could provide multiple reasons for termination.

Data analysis

Kaplan-Meier estimates describe the distribution of weaning times for the 2 sites. A log-rank test was used to assess group differences. Relationships between demographic factors and time of weaning were assessed by Cox regression analysis. Discrete survival analysis was used to determine whether variables measured on a weekly basis were related to breastfeeding cessation. Hazard ratios describe the association of the exposures between women who stopped breastfeeding at a given time and those who continued. Because breastfeeding cessation was a rare event in later weeks of the study, as were certain clinical or behavioral breastfeeding factors, weeks 4-12 were collapsed into a single interval. Two variables, number of daily feedings and duration of each feeding, were examined only in the first 3 weeks because the information was often missing beyond 3 weeks. All analyses were performed using the Statistical Package for the Social Sciences.²⁵

Results

Description of subjects

A total of 1057 women agreed to be contacted. Of those, 946 (89.5%) participated in at least 1 interview. Of the 111 women who did not participate, 11 refused and 100 could not be located. Six hundred fifty-eight (69.6%) women completed all 4 interviews. The 56 women who entered the study at week 6 because they could not be reached for the first interview were similar in all factors to women who entered earlier. Of the 946, 711 (75.2%) were from Michigan and 235 (24.8%) were from Nebraska.

Subjects from Michigan were significantly more likely than those from Nebraska to be older than 30 years (52.0% vs 38.3%), have at least a bachelor’s degree (62.9% vs 48.5%), have 3 or more children (38.5% vs 19.6%), and have had a vaginal delivery (99.6% vs 77.0%) (Table W1).* The groups were similar in race, household income, and marital status.

Demographic factors

A total of 673 women (71.1%) continued breastfeeding until 12 weeks; 28% were exclusive breastfeeders. Michigan women were more likely to breastfeed at weeks 2 through 12 than their Nebraskan counterparts (P < .0001, Figure). A college degree was associated with 40% less weaning (Table 1). Age and annual household income were directly related to continued breastfeeding at both sites. Number of children in the household was not associated with termination. Previous breastfeeding experience showed a nonsignificant but consistent trend toward lower weaning risk.

TABLE 1
Relationships of demographics and other characteristics with time to weaning, by site

FIGURE
Probability of breastfeeding, by site, by postpartum week

Clinical and behavioral factors

Because time to weaning differed significantly by site, the survival analyses of clinical and behavioral factors were performed separately for Michigan and Nebraska and controlled for education, age, and previous breastfeeding experience.

During the first 3 weeks, Michigan women with mastitis were nearly 6 times more likely than Michigan women without mastitis to stop breast-feeding in the week of diagnosis (Table 2). Women from Nebraska showed nonsignificant results in the same direction in weeks 4 to 12. (No women from Nebraska with mastitis terminated during weeks 1 through 3.) Although nipple sores and cracks were not associated with weaning, breast pain was associated with weaning. For each day of pain in the first 3 weeks, there was a 10% increase in risk of cessation among Michigan women and a 26% increase among Nebraska women. The association between pain and weaning in weeks 4 through 12 is less clear. In these later weeks, women who reported pain were unexpectedly 75% to 80% more likely to continue breastfeeding than women who did not report pain, yet for Nebraska women the number of days with pain remained significantly associated with breastfeeding cessation.

Subjective depression and breastfeeding cessation were not related. The association between daily sleep and weaning varied by site. During weeks 4 through 12, Michigan women with more daily sleep were less likely to terminate. An opposite, but marginally significant trend, was observed for Nebraska women. Weaning was not associated with outside household help. Nonvaginal birth was not associated with weaning for Nebraska women. (There were only 2 cesarean sections in the Michigan group.)

Michigan women who expressed breast milk during the first 3 weeks were twice as likely to stop breastfeeding as those who did not pump. During the same period, Michigan women who used a bottle for some feedings were 9 times more likely to wean than nonbottle users. Respondents in Nebraska showed similar nonsignificant trends in the first 3 weeks. By contrast, during weeks 4 through 12, both Nebraska and Michigan women who pumped were about 75% less likely to wean, while women who used a bottle for some feedings were 98% less likely to stop breastfeeding.

Breast milk expression increased gradually over time, from 30% of women pumping an average of 3 times per day in the first 3 weeks to 45% of women pumping 5 times per day in the last 3 weeks. To determine if pumping and bottle-feeding had an effect independent of pain or mastitis on weaning in the first 3 weeks, we performed additional analyses controlling for pain, cracks and sores, and mastitis in the same week. The results were similar to those presented in Table 2. Michigan women who pumped were 3 times more likely to wean than those who did not pump (hazard ratio [HR] = 3.0, 95% confidence interval [CI], 1.3 - 6.7), while for Nebraska women there was no association between pumping and weaning (HR = 0.6, 95% CI, 0.3 - 1.5). Bottle-feeding was again significantly associated with weaning in weeks 1 through 3 for Michigan women (HR = 10.9, 95% CI, 4.5 - 26.7) and not associated in Nebraskans (HR = 0.8, 95% CI, 0.4 - 2.0).

Duration and frequency of feedings were investigated as weaning risk factors. There appeared to be a threshold for both variables during the first 3 weeks in Michigan women. Michigan women who breastfed less than 10 minutes per feeding were nearly 5 times more likely to stop breastfeeding than women who breastfed longer. Michigan women who breastfed 6 or fewer times per day were 8 times more likely to stop than those who breastfed more often. Results for Nebraska women fell in the same direction but were not statistically significant.

TABLE 2
Relationships of clinical and behavioral factors to breastfeeding cessation in the same week, adjusted for mother’s age, education, and previous breastfeeding experience

Subjective factors

At each interview, women who had stopped breastfeeding in the previous 3 weeks were asked why they had made that decision. Most women (75%) provided only one reason. At the first interview, insufficient milk supply (37.3%) and breast pain or mastitis (32.9%) were the most common reasons for termination (Table 3). Insufficient milk supply was the reason most often given (35.0%) during weeks 4 through 6. At both weeks 9 and 12, return to work was the reason given most often (53.1% and 58.3%, respectively).

TABLE 3
Percentage of women citing given reason for termination of breastfeeding

Discussion

Mastitis, pain, and days with pain in the first 3 weeks were important clinical factors associated with breastfeeding cessation in this cohort of women who prenatally self-identified as intending to breastfeed. Women who intend to breastfeed should be counseled regarding these possible complications, their temporary nature, prevention, and treatment. Mastitis is not an indication for breastfeeding termination; in fact, increased feedings and milk expression are considered treatment.^26,27 Women who reported pain the first 3 weeks were more likely to stop breastfeeding than women who reported pain after the first 3 weeks. It is difficult to explain this finding; perhaps there are women who have pain during their entire breastfeeding career and yet continue to breastfeed because they are more pain-tolerant, have less severe or frequent pain than those who wean, or are more committed to breastfeeding.

Other clinical factors investigated were depression and daily sleep hours. Weaning was not associated with subjective depression. However, subjects did not undergo formal psychological testing as in the study that reported an association.²⁴ The relationship between daily sleep hours and termination was not consistent, and likely not clinically significant.

The demographic risk factors related to breast-feeding termination in our study are similar to those previously reported,^{14,15,20,21,23,24} namely, younger maternal age and lower educational level. Investigations of parity have been inconsistent.^16,28 We found no association of weaning with parity. Prior breastfeeding experience has been reported as improving breastfeeding rates^15,28; our results are consistent with those findings, but not significantly so. All subjects had access to prenatal breastfeeding education and postnatal breastfeeding support, which may have diminished the differences between women with breastfeeding experience and those without experience.²⁰

Michigan and Nebraska women who pumped or bottle-fed during weeks 4 through 12 were significantly less likely to terminate breastfeeding. In contrast, Michigan women who pumped or bottle-fed during the first 3 weeks postpartum were more likely to terminate even after controlling for pain and mastitis. A commitment to exclusive breastfeeding may be necessary in the early postpartum period for long-term success.^15,19 To our knowledge, the seemingly protective effect associated with pumping and bottle-feeding after the first 3 weeks has not been previously reported.

Breastfeeding 6 or fewer times per day and feedings of 10 minutes or less were associated with termination during the first 3 weeks. Other studies also indicate that the ratio of breast to bottle feedings is important for long-term success. Feinstein and colleagues¹⁵ found that more than one daily bottle of formula supplementation was associated with shorter breastfeeding duration, which was minimized if there were 7 or more breastfeedings per day. Another study found no weaning difference between women who offered their infant only one bottle daily during weeks 2 through 6 and a bottle-avoiding group.¹⁷

The most frequent reasons given for termination were similar to those reported by others, namely, insufficient milk supply and return to work.^11-15,21,22 Insufficient milk supply was a more common reason in the first few weeks after birth; return to work became an increasingly common reason after week 6.

We were unable to examine the role of pacifiers or smoking in breastfeedng termination because pacifier information was not collected and there were too few smokers for meaningful analysis. Smoking has been consistently reported as associated with early cessation.^15,20,29,30 Although pacifier use does not appear to be directly related,^31,32 it has been proposed as a marker for breastfeeding problems. The homogeneity of the sample limits our ability to make generalizations regarding other populations, such as women of color. However, the large sample size and the similarity of termination risk factors between 2 different populations of women lend confidence to our conclusions. As we did not assess mothers’ intentions, some of the variables found associated with termination might be intentional activities of weaning rather than risk factors for termination. The significant difference in termination risk between the sites also may be related to mothers’ intentions or level of commitment. The Michigan women may have intended to breastfeed longer from the outset. The Michigan recruitment site was an alternative birthing center. Women being delivered there may be more persistent in their breast-feeding efforts. Both sites provided access to breast-feeding support personnel, but the Michigan women, as a group, may have been more motivated to continue.

Our results provide clinically useful information. Additional support may be needed for younger and less educated women. Special efforts should be made for early diagnosis and treatment of mastitis and breast pain, particularly during the first 3 weeks. Exclusive breastfeeding without bottle supplementation should be recommended for the first 3 weeks, with at least 7 feedings per day. Each feeding should preferably last more than 10 minutes.

These results should also reassure breastfeeding women and their providers regarding the use of bottles. Bottle-feeding after 3 weeks does not appear to jeopardize breastfeeding success up to 12 weeks and may even improve it.

* Table W1 appears on the JFP Web site at www.jfponline.com.

Acknowledgments

This study was supported by National Institutes of Health grant #30866.

One alternative therapy for osteoarthritis (OA) is Sadenosylmethionine (SAMe), a naturally occurring sulphur-containing physiologic compound synthesized from amino acid L-methionine and adenosine triphosphate (ATP).^1,2 Although scientists are not certain how it works to control pain, SAMe plays a key role in 3 major pathways: transmethylation, transsulfuration, and aminopropylation.² SAMe was introduced in the United States in 1999 as a dietary supplement to promote joint health, mobility, and joint comfort. On the basis of a 1987 review of 12 clinical studies involving more than 20,000 patients, SAMe has been touted as “the prototype of a new class of safe drugs for the treatment of osteoarthritis.”³ However, the majority of the patients in those studies (97%) were enrolled in a single open field trial.

Although systematic reviews have demonstrated the benefit of other alternative strategies for OA, such as glucosamine and chondroitin,^4,5 there has been no systematic review of SAMe for OA. Because individual studies of SAMe vary in their sample sizes and report conflicting results, we conducted a meta-analysis to assess the efficacy of SAMe for OA as compared with that of placebo or NSAIDs. We also examined whether study quality, drug dosage, or length of treatment is associated with the effect, and we identified needs for future research.

Methods

Literature search and data sources

We conducted computerized searches using the term “arthritis” and all synonyms for SAMe: “S-Adenosylmethionine,” “Ademetionine,” “S-adenosyl-L-methionine,” “Adenosyl-l-methionine,” “Samyr,” “Gumbaral,” “Sammy,” and “SAM-e.” Results were then combined into the optimally sensitive search strategy for retrieving all clinical trials.^6,7 All languages were included. Our database search included MEDLINE (1966- September 2000), EMBASE (1987-2000), CAMPAIN (Complementary and Alternative Medicine and Pain), Science Citation Index, International Pharmaceutical Abstracts, The Cochrane Complementary Medicine Field Registry, National Institutes of Health Office of Dietary Supplements Database, and Micromedix. We also hand searched the 3 journals with the highest impact factors for rheumatology (Arthritis and Rheumatism, British Journal of Rheumatology, and Journal of Rheumatology, 1985-1999),⁸ English-language journals from which we had already retrieved articles, and complementary medicine journals (inception to 1999). In addition, we examined bibliographies from retrieved articles, books, and Web sites related to SAMe and contacted manufacturers of SAMe for previously unidentified research studies.

Inclusion criteria

Criteria for inclusion were established a priori. Studies had to include a sample of patients with a diagnosis of OA; be a randomized controlled trial; compare SAMe with placebo or NSAID; and report data for at least 1 of the outcome variables: pain, functional limitation, and adverse effects. Two raters independently screened studies to determine whether they met the inclusion criteria and agreed in their assessments.

Quality assessment and data extraction

Two raters independently rated study quality of the English studies using the 5-point Jadad scale⁹ that assesses random allocation, double-blinding, and the reporting of withdrawals and dropouts. An additional rating item concerned concealed allocation. Only 1 of the 2 raters assessed the quality of the 4 non-English articles. Two reviewers also independently extracted descriptive information and outcomes that reflected pain, functional impairment, and adverse effects. Any differences in ratings and data extraction were discussed and a consensus was reached.

For pain and functional impairment we computed the difference in the average response between treatment groups and control groups, standardized to account for differences in the measurement scale across studies. The result is a difference effect size (ES) with a positive ES favoring SAMe. We also applied a correction factor¹⁰ that adjusts for the positive bias in the ES estimate for small samples. For the binary outcome of adverse effects, we computed the odds ratio (OR) for the individual trials.¹¹ An OR of less than 1 indicated that treatment with SAMe was more effective than the control.

Heterogeneity in the strategy to measure pain was expected. Either individual studies pooled several pain items (eg, day pain and resting pain) that were rated using a 4- or 5-point rating scale or Visual Analog Scale (VAS), or studies used a single-item VAS. Functional limitation reflects stiffness, swelling, and joint mobility as rated by the physician according to the degree of joint movement (eg, flexion, extension, abduction, adduction, and rotation). In some studies, this score also included a pain item. Adverse effects refer to patient reports of nonspecific gastrointestinal complaints, mucocutaneous symptoms, and central nervous systems disturbances. Finally, a pooled dropout rate because of side effects was computed across studies as a measure of the tolerability of SAMe.

Statistical analysis

Outcomes for each subject measured at multiple time points tend to be correlated, which introduces dependency between corresponding ESs. To avoid this dependency, we computed the ES for the end-of-treatment only, rather than for all time points. Although dependency is also a concern when results are reported for more than one outcome within a study,^12-14 we did not control for this. Following the test for homogeneity or consistency within the set of ESs using the Q statistic with α = .10,11we computed the weighted mean ES with 95% confidence intervals (CI) across studies for each outcome, weighting for sample size (the inverse of the variance). The choice of a fixed-effects model was dependent on the finding of homogeneity of results.

To assess sensitivity of the results, we examined the relationship of the ES to the dosage of SAMe, length of treatment, and study quality rating. Subgroup analyses examined differences related to the location of the OA to estimate the robustness of results. Finally, we assessed potential publication bias informally by using the funnel plot of ES by precision, and statistically through the rank correlation between the standardized ES and standardized study variance.¹⁵

Results

Description of studies

Twenty studies were identified through our search and 11 of them^16-26 met the inclusion criteria (Table). We excluded one duplicate study₂₇and one study whose sample included persons with rheumatoid arthritis.²⁸ Other excluded studies compared the routes of administration of SAMe,²⁹ compared SAMe plus ketoprofen with ketoprofen alone,³⁰ or were not randomized controlled trials.^31-34 Four of the included studies^18,20,21,25 were published in Italian; the others were published in English. The majority of studies (7 of 11) were conducted in Italy.

Quality assessment

Percent agreement between raters for the items on the Jadad scale averaged 87.5%. Following discussion, the raters reached consensus for all items. Using Jadad’s criteria, all studies were rated of high quality (score 3), although only 2 studies^16,23 included a description of the method of randomization. None of the studies addressed allocation concealment.

Study characteristics

Ten of the 11 studies used a parallel groups design including one with 3 arms¹⁹; the 11th one²⁵ used a crossover design (Table W1).* The SAMe dosage in 6 studies was 1200 mg per day orally^{18,19,22-24,26}; 3 studies used 600 mg per day orally^17,21,25; and one used 400 mg per day intravenously.²⁰ In one study¹⁶ the dosage varied. Duration of treatment ranged from 10 days to 84 days; a duration of 28 or 30 days was used in 8 of the studies. A variety of NSAIDs served as active comparators and 2 studies^16,19 used placebo. The studies involved 1442 subjects with a mean age of 60.3 years, of whom 70.1% were women. Mean duration of OA was 5.7 years, ranging from 2.6 years to 9.1 years. In 5 studies, the majority of subjects had OA of the knee; across all studies 54.2% of the subjects had OA of the knee.

TABLE
Characteristics of studies included in meta-analysis

Analysis of outcomes

Pain. Twelve ESs from 7 studies^{16,18-20,22,23,25} were computed for pain, ranging from -.501 to +.794. Because of borderline heterogeneity of the results for SAMe versus placebo (Q[2] = 5.41; P= .067), a more conservative random effects model was used to compute the mean ES of .223 (P= .352; 95% CI, -.247 to .693). Homogeneity was present for SAMe versus NSAIDs (Q[8] = 9.31, P= .317) and on the basis of a fixed effects model, the weighted mean ES was .122 (P= .057; 95% CI, -.029 to .273). Among the studies of SAMe versus NSAIDs, effect size was not related to study quality (P= .32), length of intervention (P= .31), or dosage of SAMe (P= .97). Finally, there was no evidence of publication bias according to the funnel P lot (Figure W1)* or the rank order correlation (P= .297) for studies of SAMe versus NSAIDs.

Functional limitation. Six studies^17-20,24,26 contributed 10 effect sizes for functional limitation. The length of the intervention phase was 28 days to 42 days for all 6 studies. Only one study¹⁹ compared SAMe with placebo (ES = .309; P= .002; 95% CI, .098 - .519). Among the studies comparing SAMe with NSAIDs, there was homogeneity of results (Q[8] = 2.53; P= .96) with a weighted mean ES of .025 (95% CI, -.127 to .176), indicating no difference between SAMe and NSAIDs with respect to functional limitation. There was no relationship of ES to study quality (P = .30), length of treatment (P= .71), or dosage of SAMe (P= .48). Both the funnel plot (Figure W2)** and the rank correlation of standardized ES and variance (P= .097) suggested no evidence of publication bias with respect to the functional limitation outcome for SAMe versus NSAIDs.

Adverse effects. Two studies^16,19 reported adverse effects when comparing SAMe with placebo. Results were homogenous (Q[2] = 2.035; P= .362), with a pooled OR of 1.37 (95% CI, .81 - 2.32). Among the studies comparing SAMe with NSAIDs results also were homogeneous (Q[6] = 4.41; P =.622), with a pooled OR of .424 (95% CI, .294 - .611). Again, the effect size was not related to quality of study (P= .409), length of treatment (P= .367), or dosage of SAMe (P= .341). That is, those treated with SAMe were 58% less likely to experience side effects than those treated with NSAIDs. Further, this was independent of study quality, dosage of SAMe, or the length of the intervention.

As an additional indication of tolerability we compared the overall dropout rates due to side effects. The dropout rate was highest (6.9%) among those treated with NSAIDs, followed by those receiving placebo (5.0%). The dropout rate for SAMe users was lowest at 2.6%. The only significant difference was between those treated with SAMe and with NSAIDs (P= .001).

Discussion

Results of this meta-analysis indicate that SAMe has a comparable effect to that of NSAIDs in reducing pain and functional limitation. In addition, there was significantly less likelihood of patients reporting adverse effects with the use of SAMe. When SAMe is compared with placebo, however, there is no differential effect on pain according to 2 studies, although there is minimal improved functional limitation according to one study. This improvement corresponds to a 15% decrease in functional limitation in the SAMe group as compared with placebo. The likelihood of adverse effects was similar in the 2 groups. Given the combined sample sizes in this meta-analysis, there was a more than 90% power to detect a moderate difference between groups at a .05 level of significance.

Several reporting issues were noted during the extraction of study data. Some researchers did not adequately describe study dropouts and how they were handled. Sample characteristics may have been reported for the initial sample, but there was no mention of the characteristics of the final sample, so that bias in subject loss could not be assessed in any studies that did not use intention-to-treat analysis. Some authors reported intervention results on the basis of the location of the OA, but only reported characteristics (age, sex, duration of disease) for the full sample. This precluded examining the relationship of intervention effect size to demographic characteristics. Finally, because not all authors provided complete descriptive statistics, we based the computation of the ES for one study on post-test scores only, rather than on the change from baseline, a strategy that could underestimate the ES. This potential underestimation occurred in a study with one of the larger sample sizes that, in turn, would carry more weight in the analysis.

Limitations

Potential limitations must also be noted in our analysis. First, in 6 of the studies, the SAMe dosage of 1200 mg per day exceeded the dosing recommendations for SAMe. These recommendations include 800 mg per day for 2 weeks followed by 400 mg per day as a maintenance dose, or to increase from 200 mg per day to 1200 mg per day over a 19-day period followed by 400 mg per day thereafter.³⁵ Dosage was not related to the ES, however, in studies comparing SAMe with NSAIDs. Second, most studies used a short intervention (28 to 30 days). It may be that NSAIDs are more effective in the long run, that a longer treatment period is needed for patients to realize the effect of SAMe, or that there are more adverse side effects with SAMe over time. It is not yet clear how effective SAMe is over time. Those studies that did have an intervention longer than 30 days^18,22 did not compare SAMe with ibuprofen. In general, concomitant medications for treatment of OA were not permitted, but 3 studies^24-26 failed to provide this information. Finally, most of the studies looked at OA of the knee and/or hip, so generalizability of the results to other locations of OA is limited. Although we included subgroup analyses by location of OA, statistical power for subgroup analysis was low because of the smaller number of subjects for whom data were available.

Conclusions

Although SAMe appears to offer pain relief and improve functional limitations associated with OA without the side effects of NSAIDs, it must be remembered that SAMe is not considered a drug in the United States and is therefore not subject to federal regulations. (In contrast, Samyr is a prescription drug in Italy and is available in 200 mg and 400 mg doses.) Recent testing by ConsumerLab.com of over-the-counter brands of SAMe in the United States found, on average, that for 6 of the 13 brands tested, less than half the amount of SAMe stated on the label was actually present.³⁶ Patients who use SAMe in the United States may fail to experience relief because of this dose inconsistency.

We offer several suggestions for further research. First, the long-term effectiveness of SAMe for the treatment of OA has not been investigated in a randomized controlled trial. Since OA is the most prevalent form of arthritis, the long-term effectiveness of SAMe should be assessed in this manner. Second, given that SAMe has been shown to decrease depression,¹ it seems prudent to use multivariate techniques to examine both depression and OA outcomes (pain and functional limitation) to determine whether the effect of SAMe is directly on the joint or indirectly mediated through depression. Perhaps in the short term SAMe does decrease pain through decreasing depressive symptoms, but in the long term the effectiveness related to pain may diminish. Third, whether SAMe treats the symptoms of the disease or alters the course of the disease by increasing the production of new cartilage, as suggested by animal models, has not been investigated. Finally, can use of SAMe enhance the effectiveness of other nonpharmacologic modalities? These questions should all be investigated before we can make a determination about the efficacy and safety of SAMe for the treatment of OA.

Acknowledgments

This research was supported by grant #5-P50-AT00084-02 from the National Center for Complementary and Alternative Medicine, National Institutes of Health.

One alternative therapy for osteoarthritis (OA) is Sadenosylmethionine (SAMe), a naturally occurring sulphur-containing physiologic compound synthesized from amino acid L-methionine and adenosine triphosphate (ATP).^1,2 Although scientists are not certain how it works to control pain, SAMe plays a key role in 3 major pathways: transmethylation, transsulfuration, and aminopropylation.² SAMe was introduced in the United States in 1999 as a dietary supplement to promote joint health, mobility, and joint comfort. On the basis of a 1987 review of 12 clinical studies involving more than 20,000 patients, SAMe has been touted as “the prototype of a new class of safe drugs for the treatment of osteoarthritis.”³ However, the majority of the patients in those studies (97%) were enrolled in a single open field trial.

Although systematic reviews have demonstrated the benefit of other alternative strategies for OA, such as glucosamine and chondroitin,^4,5 there has been no systematic review of SAMe for OA. Because individual studies of SAMe vary in their sample sizes and report conflicting results, we conducted a meta-analysis to assess the efficacy of SAMe for OA as compared with that of placebo or NSAIDs. We also examined whether study quality, drug dosage, or length of treatment is associated with the effect, and we identified needs for future research.

Methods

Literature search and data sources

We conducted computerized searches using the term “arthritis” and all synonyms for SAMe: “S-Adenosylmethionine,” “Ademetionine,” “S-adenosyl-L-methionine,” “Adenosyl-l-methionine,” “Samyr,” “Gumbaral,” “Sammy,” and “SAM-e.” Results were then combined into the optimally sensitive search strategy for retrieving all clinical trials.^6,7 All languages were included. Our database search included MEDLINE (1966- September 2000), EMBASE (1987-2000), CAMPAIN (Complementary and Alternative Medicine and Pain), Science Citation Index, International Pharmaceutical Abstracts, The Cochrane Complementary Medicine Field Registry, National Institutes of Health Office of Dietary Supplements Database, and Micromedix. We also hand searched the 3 journals with the highest impact factors for rheumatology (Arthritis and Rheumatism, British Journal of Rheumatology, and Journal of Rheumatology, 1985-1999),⁸ English-language journals from which we had already retrieved articles, and complementary medicine journals (inception to 1999). In addition, we examined bibliographies from retrieved articles, books, and Web sites related to SAMe and contacted manufacturers of SAMe for previously unidentified research studies.

Inclusion criteria

Criteria for inclusion were established a priori. Studies had to include a sample of patients with a diagnosis of OA; be a randomized controlled trial; compare SAMe with placebo or NSAID; and report data for at least 1 of the outcome variables: pain, functional limitation, and adverse effects. Two raters independently screened studies to determine whether they met the inclusion criteria and agreed in their assessments.

Quality assessment and data extraction

Two raters independently rated study quality of the English studies using the 5-point Jadad scale⁹ that assesses random allocation, double-blinding, and the reporting of withdrawals and dropouts. An additional rating item concerned concealed allocation. Only 1 of the 2 raters assessed the quality of the 4 non-English articles. Two reviewers also independently extracted descriptive information and outcomes that reflected pain, functional impairment, and adverse effects. Any differences in ratings and data extraction were discussed and a consensus was reached.

For pain and functional impairment we computed the difference in the average response between treatment groups and control groups, standardized to account for differences in the measurement scale across studies. The result is a difference effect size (ES) with a positive ES favoring SAMe. We also applied a correction factor¹⁰ that adjusts for the positive bias in the ES estimate for small samples. For the binary outcome of adverse effects, we computed the odds ratio (OR) for the individual trials.¹¹ An OR of less than 1 indicated that treatment with SAMe was more effective than the control.

Heterogeneity in the strategy to measure pain was expected. Either individual studies pooled several pain items (eg, day pain and resting pain) that were rated using a 4- or 5-point rating scale or Visual Analog Scale (VAS), or studies used a single-item VAS. Functional limitation reflects stiffness, swelling, and joint mobility as rated by the physician according to the degree of joint movement (eg, flexion, extension, abduction, adduction, and rotation). In some studies, this score also included a pain item. Adverse effects refer to patient reports of nonspecific gastrointestinal complaints, mucocutaneous symptoms, and central nervous systems disturbances. Finally, a pooled dropout rate because of side effects was computed across studies as a measure of the tolerability of SAMe.

Statistical analysis

Outcomes for each subject measured at multiple time points tend to be correlated, which introduces dependency between corresponding ESs. To avoid this dependency, we computed the ES for the end-of-treatment only, rather than for all time points. Although dependency is also a concern when results are reported for more than one outcome within a study,^12-14 we did not control for this. Following the test for homogeneity or consistency within the set of ESs using the Q statistic with α = .10,11we computed the weighted mean ES with 95% confidence intervals (CI) across studies for each outcome, weighting for sample size (the inverse of the variance). The choice of a fixed-effects model was dependent on the finding of homogeneity of results.

To assess sensitivity of the results, we examined the relationship of the ES to the dosage of SAMe, length of treatment, and study quality rating. Subgroup analyses examined differences related to the location of the OA to estimate the robustness of results. Finally, we assessed potential publication bias informally by using the funnel plot of ES by precision, and statistically through the rank correlation between the standardized ES and standardized study variance.¹⁵

Results

Description of studies

Twenty studies were identified through our search and 11 of them^16-26 met the inclusion criteria (Table). We excluded one duplicate study₂₇and one study whose sample included persons with rheumatoid arthritis.²⁸ Other excluded studies compared the routes of administration of SAMe,²⁹ compared SAMe plus ketoprofen with ketoprofen alone,³⁰ or were not randomized controlled trials.^31-34 Four of the included studies^18,20,21,25 were published in Italian; the others were published in English. The majority of studies (7 of 11) were conducted in Italy.

Quality assessment

Percent agreement between raters for the items on the Jadad scale averaged 87.5%. Following discussion, the raters reached consensus for all items. Using Jadad’s criteria, all studies were rated of high quality (score 3), although only 2 studies^16,23 included a description of the method of randomization. None of the studies addressed allocation concealment.

Study characteristics

Ten of the 11 studies used a parallel groups design including one with 3 arms¹⁹; the 11th one²⁵ used a crossover design (Table W1).* The SAMe dosage in 6 studies was 1200 mg per day orally^{18,19,22-24,26}; 3 studies used 600 mg per day orally^17,21,25; and one used 400 mg per day intravenously.²⁰ In one study¹⁶ the dosage varied. Duration of treatment ranged from 10 days to 84 days; a duration of 28 or 30 days was used in 8 of the studies. A variety of NSAIDs served as active comparators and 2 studies^16,19 used placebo. The studies involved 1442 subjects with a mean age of 60.3 years, of whom 70.1% were women. Mean duration of OA was 5.7 years, ranging from 2.6 years to 9.1 years. In 5 studies, the majority of subjects had OA of the knee; across all studies 54.2% of the subjects had OA of the knee.

TABLE
Characteristics of studies included in meta-analysis

Analysis of outcomes

Pain. Twelve ESs from 7 studies^{16,18-20,22,23,25} were computed for pain, ranging from -.501 to +.794. Because of borderline heterogeneity of the results for SAMe versus placebo (Q[2] = 5.41; P= .067), a more conservative random effects model was used to compute the mean ES of .223 (P= .352; 95% CI, -.247 to .693). Homogeneity was present for SAMe versus NSAIDs (Q[8] = 9.31, P= .317) and on the basis of a fixed effects model, the weighted mean ES was .122 (P= .057; 95% CI, -.029 to .273). Among the studies of SAMe versus NSAIDs, effect size was not related to study quality (P= .32), length of intervention (P= .31), or dosage of SAMe (P= .97). Finally, there was no evidence of publication bias according to the funnel P lot (Figure W1)* or the rank order correlation (P= .297) for studies of SAMe versus NSAIDs.

Functional limitation. Six studies^17-20,24,26 contributed 10 effect sizes for functional limitation. The length of the intervention phase was 28 days to 42 days for all 6 studies. Only one study¹⁹ compared SAMe with placebo (ES = .309; P= .002; 95% CI, .098 - .519). Among the studies comparing SAMe with NSAIDs, there was homogeneity of results (Q[8] = 2.53; P= .96) with a weighted mean ES of .025 (95% CI, -.127 to .176), indicating no difference between SAMe and NSAIDs with respect to functional limitation. There was no relationship of ES to study quality (P = .30), length of treatment (P= .71), or dosage of SAMe (P= .48). Both the funnel plot (Figure W2)** and the rank correlation of standardized ES and variance (P= .097) suggested no evidence of publication bias with respect to the functional limitation outcome for SAMe versus NSAIDs.

Adverse effects. Two studies^16,19 reported adverse effects when comparing SAMe with placebo. Results were homogenous (Q[2] = 2.035; P= .362), with a pooled OR of 1.37 (95% CI, .81 - 2.32). Among the studies comparing SAMe with NSAIDs results also were homogeneous (Q[6] = 4.41; P =.622), with a pooled OR of .424 (95% CI, .294 - .611). Again, the effect size was not related to quality of study (P= .409), length of treatment (P= .367), or dosage of SAMe (P= .341). That is, those treated with SAMe were 58% less likely to experience side effects than those treated with NSAIDs. Further, this was independent of study quality, dosage of SAMe, or the length of the intervention.

As an additional indication of tolerability we compared the overall dropout rates due to side effects. The dropout rate was highest (6.9%) among those treated with NSAIDs, followed by those receiving placebo (5.0%). The dropout rate for SAMe users was lowest at 2.6%. The only significant difference was between those treated with SAMe and with NSAIDs (P= .001).

Discussion

Results of this meta-analysis indicate that SAMe has a comparable effect to that of NSAIDs in reducing pain and functional limitation. In addition, there was significantly less likelihood of patients reporting adverse effects with the use of SAMe. When SAMe is compared with placebo, however, there is no differential effect on pain according to 2 studies, although there is minimal improved functional limitation according to one study. This improvement corresponds to a 15% decrease in functional limitation in the SAMe group as compared with placebo. The likelihood of adverse effects was similar in the 2 groups. Given the combined sample sizes in this meta-analysis, there was a more than 90% power to detect a moderate difference between groups at a .05 level of significance.

Several reporting issues were noted during the extraction of study data. Some researchers did not adequately describe study dropouts and how they were handled. Sample characteristics may have been reported for the initial sample, but there was no mention of the characteristics of the final sample, so that bias in subject loss could not be assessed in any studies that did not use intention-to-treat analysis. Some authors reported intervention results on the basis of the location of the OA, but only reported characteristics (age, sex, duration of disease) for the full sample. This precluded examining the relationship of intervention effect size to demographic characteristics. Finally, because not all authors provided complete descriptive statistics, we based the computation of the ES for one study on post-test scores only, rather than on the change from baseline, a strategy that could underestimate the ES. This potential underestimation occurred in a study with one of the larger sample sizes that, in turn, would carry more weight in the analysis.

Limitations

Potential limitations must also be noted in our analysis. First, in 6 of the studies, the SAMe dosage of 1200 mg per day exceeded the dosing recommendations for SAMe. These recommendations include 800 mg per day for 2 weeks followed by 400 mg per day as a maintenance dose, or to increase from 200 mg per day to 1200 mg per day over a 19-day period followed by 400 mg per day thereafter.³⁵ Dosage was not related to the ES, however, in studies comparing SAMe with NSAIDs. Second, most studies used a short intervention (28 to 30 days). It may be that NSAIDs are more effective in the long run, that a longer treatment period is needed for patients to realize the effect of SAMe, or that there are more adverse side effects with SAMe over time. It is not yet clear how effective SAMe is over time. Those studies that did have an intervention longer than 30 days^18,22 did not compare SAMe with ibuprofen. In general, concomitant medications for treatment of OA were not permitted, but 3 studies^24-26 failed to provide this information. Finally, most of the studies looked at OA of the knee and/or hip, so generalizability of the results to other locations of OA is limited. Although we included subgroup analyses by location of OA, statistical power for subgroup analysis was low because of the smaller number of subjects for whom data were available.

Conclusions

Although SAMe appears to offer pain relief and improve functional limitations associated with OA without the side effects of NSAIDs, it must be remembered that SAMe is not considered a drug in the United States and is therefore not subject to federal regulations. (In contrast, Samyr is a prescription drug in Italy and is available in 200 mg and 400 mg doses.) Recent testing by ConsumerLab.com of over-the-counter brands of SAMe in the United States found, on average, that for 6 of the 13 brands tested, less than half the amount of SAMe stated on the label was actually present.³⁶ Patients who use SAMe in the United States may fail to experience relief because of this dose inconsistency.

We offer several suggestions for further research. First, the long-term effectiveness of SAMe for the treatment of OA has not been investigated in a randomized controlled trial. Since OA is the most prevalent form of arthritis, the long-term effectiveness of SAMe should be assessed in this manner. Second, given that SAMe has been shown to decrease depression,¹ it seems prudent to use multivariate techniques to examine both depression and OA outcomes (pain and functional limitation) to determine whether the effect of SAMe is directly on the joint or indirectly mediated through depression. Perhaps in the short term SAMe does decrease pain through decreasing depressive symptoms, but in the long term the effectiveness related to pain may diminish. Third, whether SAMe treats the symptoms of the disease or alters the course of the disease by increasing the production of new cartilage, as suggested by animal models, has not been investigated. Finally, can use of SAMe enhance the effectiveness of other nonpharmacologic modalities? These questions should all be investigated before we can make a determination about the efficacy and safety of SAMe for the treatment of OA.

Acknowledgments

This research was supported by grant #5-P50-AT00084-02 from the National Center for Complementary and Alternative Medicine, National Institutes of Health.

One alternative therapy for osteoarthritis (OA) is Sadenosylmethionine (SAMe), a naturally occurring sulphur-containing physiologic compound synthesized from amino acid L-methionine and adenosine triphosphate (ATP).^1,2 Although scientists are not certain how it works to control pain, SAMe plays a key role in 3 major pathways: transmethylation, transsulfuration, and aminopropylation.² SAMe was introduced in the United States in 1999 as a dietary supplement to promote joint health, mobility, and joint comfort. On the basis of a 1987 review of 12 clinical studies involving more than 20,000 patients, SAMe has been touted as “the prototype of a new class of safe drugs for the treatment of osteoarthritis.”³ However, the majority of the patients in those studies (97%) were enrolled in a single open field trial.

Although systematic reviews have demonstrated the benefit of other alternative strategies for OA, such as glucosamine and chondroitin,^4,5 there has been no systematic review of SAMe for OA. Because individual studies of SAMe vary in their sample sizes and report conflicting results, we conducted a meta-analysis to assess the efficacy of SAMe for OA as compared with that of placebo or NSAIDs. We also examined whether study quality, drug dosage, or length of treatment is associated with the effect, and we identified needs for future research.

Methods

Literature search and data sources

We conducted computerized searches using the term “arthritis” and all synonyms for SAMe: “S-Adenosylmethionine,” “Ademetionine,” “S-adenosyl-L-methionine,” “Adenosyl-l-methionine,” “Samyr,” “Gumbaral,” “Sammy,” and “SAM-e.” Results were then combined into the optimally sensitive search strategy for retrieving all clinical trials.^6,7 All languages were included. Our database search included MEDLINE (1966- September 2000), EMBASE (1987-2000), CAMPAIN (Complementary and Alternative Medicine and Pain), Science Citation Index, International Pharmaceutical Abstracts, The Cochrane Complementary Medicine Field Registry, National Institutes of Health Office of Dietary Supplements Database, and Micromedix. We also hand searched the 3 journals with the highest impact factors for rheumatology (Arthritis and Rheumatism, British Journal of Rheumatology, and Journal of Rheumatology, 1985-1999),⁸ English-language journals from which we had already retrieved articles, and complementary medicine journals (inception to 1999). In addition, we examined bibliographies from retrieved articles, books, and Web sites related to SAMe and contacted manufacturers of SAMe for previously unidentified research studies.

Inclusion criteria

Criteria for inclusion were established a priori. Studies had to include a sample of patients with a diagnosis of OA; be a randomized controlled trial; compare SAMe with placebo or NSAID; and report data for at least 1 of the outcome variables: pain, functional limitation, and adverse effects. Two raters independently screened studies to determine whether they met the inclusion criteria and agreed in their assessments.

Quality assessment and data extraction

Two raters independently rated study quality of the English studies using the 5-point Jadad scale⁹ that assesses random allocation, double-blinding, and the reporting of withdrawals and dropouts. An additional rating item concerned concealed allocation. Only 1 of the 2 raters assessed the quality of the 4 non-English articles. Two reviewers also independently extracted descriptive information and outcomes that reflected pain, functional impairment, and adverse effects. Any differences in ratings and data extraction were discussed and a consensus was reached.

For pain and functional impairment we computed the difference in the average response between treatment groups and control groups, standardized to account for differences in the measurement scale across studies. The result is a difference effect size (ES) with a positive ES favoring SAMe. We also applied a correction factor¹⁰ that adjusts for the positive bias in the ES estimate for small samples. For the binary outcome of adverse effects, we computed the odds ratio (OR) for the individual trials.¹¹ An OR of less than 1 indicated that treatment with SAMe was more effective than the control.

Heterogeneity in the strategy to measure pain was expected. Either individual studies pooled several pain items (eg, day pain and resting pain) that were rated using a 4- or 5-point rating scale or Visual Analog Scale (VAS), or studies used a single-item VAS. Functional limitation reflects stiffness, swelling, and joint mobility as rated by the physician according to the degree of joint movement (eg, flexion, extension, abduction, adduction, and rotation). In some studies, this score also included a pain item. Adverse effects refer to patient reports of nonspecific gastrointestinal complaints, mucocutaneous symptoms, and central nervous systems disturbances. Finally, a pooled dropout rate because of side effects was computed across studies as a measure of the tolerability of SAMe.

Statistical analysis

Outcomes for each subject measured at multiple time points tend to be correlated, which introduces dependency between corresponding ESs. To avoid this dependency, we computed the ES for the end-of-treatment only, rather than for all time points. Although dependency is also a concern when results are reported for more than one outcome within a study,^12-14 we did not control for this. Following the test for homogeneity or consistency within the set of ESs using the Q statistic with α = .10,11we computed the weighted mean ES with 95% confidence intervals (CI) across studies for each outcome, weighting for sample size (the inverse of the variance). The choice of a fixed-effects model was dependent on the finding of homogeneity of results.

To assess sensitivity of the results, we examined the relationship of the ES to the dosage of SAMe, length of treatment, and study quality rating. Subgroup analyses examined differences related to the location of the OA to estimate the robustness of results. Finally, we assessed potential publication bias informally by using the funnel plot of ES by precision, and statistically through the rank correlation between the standardized ES and standardized study variance.¹⁵

Results

Description of studies

Twenty studies were identified through our search and 11 of them^16-26 met the inclusion criteria (Table). We excluded one duplicate study₂₇and one study whose sample included persons with rheumatoid arthritis.²⁸ Other excluded studies compared the routes of administration of SAMe,²⁹ compared SAMe plus ketoprofen with ketoprofen alone,³⁰ or were not randomized controlled trials.^31-34 Four of the included studies^18,20,21,25 were published in Italian; the others were published in English. The majority of studies (7 of 11) were conducted in Italy.

Quality assessment

Percent agreement between raters for the items on the Jadad scale averaged 87.5%. Following discussion, the raters reached consensus for all items. Using Jadad’s criteria, all studies were rated of high quality (score 3), although only 2 studies^16,23 included a description of the method of randomization. None of the studies addressed allocation concealment.

Study characteristics

Ten of the 11 studies used a parallel groups design including one with 3 arms¹⁹; the 11th one²⁵ used a crossover design (Table W1).* The SAMe dosage in 6 studies was 1200 mg per day orally^{18,19,22-24,26}; 3 studies used 600 mg per day orally^17,21,25; and one used 400 mg per day intravenously.²⁰ In one study¹⁶ the dosage varied. Duration of treatment ranged from 10 days to 84 days; a duration of 28 or 30 days was used in 8 of the studies. A variety of NSAIDs served as active comparators and 2 studies^16,19 used placebo. The studies involved 1442 subjects with a mean age of 60.3 years, of whom 70.1% were women. Mean duration of OA was 5.7 years, ranging from 2.6 years to 9.1 years. In 5 studies, the majority of subjects had OA of the knee; across all studies 54.2% of the subjects had OA of the knee.

TABLE
Characteristics of studies included in meta-analysis

Analysis of outcomes

Pain. Twelve ESs from 7 studies^{16,18-20,22,23,25} were computed for pain, ranging from -.501 to +.794. Because of borderline heterogeneity of the results for SAMe versus placebo (Q[2] = 5.41; P= .067), a more conservative random effects model was used to compute the mean ES of .223 (P= .352; 95% CI, -.247 to .693). Homogeneity was present for SAMe versus NSAIDs (Q[8] = 9.31, P= .317) and on the basis of a fixed effects model, the weighted mean ES was .122 (P= .057; 95% CI, -.029 to .273). Among the studies of SAMe versus NSAIDs, effect size was not related to study quality (P= .32), length of intervention (P= .31), or dosage of SAMe (P= .97). Finally, there was no evidence of publication bias according to the funnel P lot (Figure W1)* or the rank order correlation (P= .297) for studies of SAMe versus NSAIDs.

Functional limitation. Six studies^17-20,24,26 contributed 10 effect sizes for functional limitation. The length of the intervention phase was 28 days to 42 days for all 6 studies. Only one study¹⁹ compared SAMe with placebo (ES = .309; P= .002; 95% CI, .098 - .519). Among the studies comparing SAMe with NSAIDs, there was homogeneity of results (Q[8] = 2.53; P= .96) with a weighted mean ES of .025 (95% CI, -.127 to .176), indicating no difference between SAMe and NSAIDs with respect to functional limitation. There was no relationship of ES to study quality (P = .30), length of treatment (P= .71), or dosage of SAMe (P= .48). Both the funnel plot (Figure W2)** and the rank correlation of standardized ES and variance (P= .097) suggested no evidence of publication bias with respect to the functional limitation outcome for SAMe versus NSAIDs.

Adverse effects. Two studies^16,19 reported adverse effects when comparing SAMe with placebo. Results were homogenous (Q[2] = 2.035; P= .362), with a pooled OR of 1.37 (95% CI, .81 - 2.32). Among the studies comparing SAMe with NSAIDs results also were homogeneous (Q[6] = 4.41; P =.622), with a pooled OR of .424 (95% CI, .294 - .611). Again, the effect size was not related to quality of study (P= .409), length of treatment (P= .367), or dosage of SAMe (P= .341). That is, those treated with SAMe were 58% less likely to experience side effects than those treated with NSAIDs. Further, this was independent of study quality, dosage of SAMe, or the length of the intervention.

As an additional indication of tolerability we compared the overall dropout rates due to side effects. The dropout rate was highest (6.9%) among those treated with NSAIDs, followed by those receiving placebo (5.0%). The dropout rate for SAMe users was lowest at 2.6%. The only significant difference was between those treated with SAMe and with NSAIDs (P= .001).

Discussion

Results of this meta-analysis indicate that SAMe has a comparable effect to that of NSAIDs in reducing pain and functional limitation. In addition, there was significantly less likelihood of patients reporting adverse effects with the use of SAMe. When SAMe is compared with placebo, however, there is no differential effect on pain according to 2 studies, although there is minimal improved functional limitation according to one study. This improvement corresponds to a 15% decrease in functional limitation in the SAMe group as compared with placebo. The likelihood of adverse effects was similar in the 2 groups. Given the combined sample sizes in this meta-analysis, there was a more than 90% power to detect a moderate difference between groups at a .05 level of significance.

Several reporting issues were noted during the extraction of study data. Some researchers did not adequately describe study dropouts and how they were handled. Sample characteristics may have been reported for the initial sample, but there was no mention of the characteristics of the final sample, so that bias in subject loss could not be assessed in any studies that did not use intention-to-treat analysis. Some authors reported intervention results on the basis of the location of the OA, but only reported characteristics (age, sex, duration of disease) for the full sample. This precluded examining the relationship of intervention effect size to demographic characteristics. Finally, because not all authors provided complete descriptive statistics, we based the computation of the ES for one study on post-test scores only, rather than on the change from baseline, a strategy that could underestimate the ES. This potential underestimation occurred in a study with one of the larger sample sizes that, in turn, would carry more weight in the analysis.

Limitations

Potential limitations must also be noted in our analysis. First, in 6 of the studies, the SAMe dosage of 1200 mg per day exceeded the dosing recommendations for SAMe. These recommendations include 800 mg per day for 2 weeks followed by 400 mg per day as a maintenance dose, or to increase from 200 mg per day to 1200 mg per day over a 19-day period followed by 400 mg per day thereafter.³⁵ Dosage was not related to the ES, however, in studies comparing SAMe with NSAIDs. Second, most studies used a short intervention (28 to 30 days). It may be that NSAIDs are more effective in the long run, that a longer treatment period is needed for patients to realize the effect of SAMe, or that there are more adverse side effects with SAMe over time. It is not yet clear how effective SAMe is over time. Those studies that did have an intervention longer than 30 days^18,22 did not compare SAMe with ibuprofen. In general, concomitant medications for treatment of OA were not permitted, but 3 studies^24-26 failed to provide this information. Finally, most of the studies looked at OA of the knee and/or hip, so generalizability of the results to other locations of OA is limited. Although we included subgroup analyses by location of OA, statistical power for subgroup analysis was low because of the smaller number of subjects for whom data were available.

Conclusions

Although SAMe appears to offer pain relief and improve functional limitations associated with OA without the side effects of NSAIDs, it must be remembered that SAMe is not considered a drug in the United States and is therefore not subject to federal regulations. (In contrast, Samyr is a prescription drug in Italy and is available in 200 mg and 400 mg doses.) Recent testing by ConsumerLab.com of over-the-counter brands of SAMe in the United States found, on average, that for 6 of the 13 brands tested, less than half the amount of SAMe stated on the label was actually present.³⁶ Patients who use SAMe in the United States may fail to experience relief because of this dose inconsistency.

We offer several suggestions for further research. First, the long-term effectiveness of SAMe for the treatment of OA has not been investigated in a randomized controlled trial. Since OA is the most prevalent form of arthritis, the long-term effectiveness of SAMe should be assessed in this manner. Second, given that SAMe has been shown to decrease depression,¹ it seems prudent to use multivariate techniques to examine both depression and OA outcomes (pain and functional limitation) to determine whether the effect of SAMe is directly on the joint or indirectly mediated through depression. Perhaps in the short term SAMe does decrease pain through decreasing depressive symptoms, but in the long term the effectiveness related to pain may diminish. Third, whether SAMe treats the symptoms of the disease or alters the course of the disease by increasing the production of new cartilage, as suggested by animal models, has not been investigated. Finally, can use of SAMe enhance the effectiveness of other nonpharmacologic modalities? These questions should all be investigated before we can make a determination about the efficacy and safety of SAMe for the treatment of OA.

Acknowledgments

This research was supported by grant #5-P50-AT00084-02 from the National Center for Complementary and Alternative Medicine, National Institutes of Health.

Breastfeeding is associated with lower rates of infant mortality and morbidity,^1-6 a reduced rate of sudden infant death syndrome (SIDS),^7,8 delayed resumption of fertility,⁹ and reduced health care cost.^10,11 The American Academy of Family Physicians has issued a policy statement supporting breastfeeding as the optimal form of nutrition for infants¹² and the American Academy of Pediatrics recommends that infants should be breastfed for at least 12 months.¹³ Therefore, it is important to understand the factors associated with reduced breastfeeding. In previous studies, the factors associated with reduced breastfeeding included maternal employment,¹⁴ child care attendance,¹⁵ maternal smoking,^14,16,17 and demographic factors.^16,18,19

Several recent studies have also identified an association between non-nutritive sucking (eg, pacifiers) and reduced breastfeeding^20-35 that is consistent with the World Health Organization (and UNICEF) recommendation that pacifiers not be used by breastfeeding infants.³⁶ Cross-sectional investigations in Sweden,^20-22 Brazil,²³ New Zealand,²⁴ England,²⁵ Greece,²⁶ and Sweden and Norway²⁷ found strong associations between pacifier use and reduced breastfeeding (either less exclusive breastfeeding, shorter duration of breastfeeding, or breastfeeding problems), with only one²⁶ not reporting statistically significant findings.

Of particular interest were several longitudinal studies in Brazil (2 studies), Sweden, Italy, and the United States. In Brazil, one found that pacifier users had an adjusted relative risk of 2.87 for weaning,²⁸ and the other an adjusted odds ratio of 2.5 for the cessation of breastfeeding associated with pacifier use.²⁹

Hörnell and colleagues³⁰ and Aarts and colleagues³¹ reported longitudinal data from 506 mothers’ daily infant feeding practices in Uppsala, Sweden. All mothers had at least one previous child breastfed at least 4 months and were planning to breastfeed the study child for at least 6 months. Thumb sucking was not associated with the breastfeeding pattern, but infants using a pacifier frequently had approximately 1 less breastfeeding session and 15 minutes to 30 minutes less total breastfeeding time per day than those not using a pacifier at 2, 4, 8, and 12-week follow-up points. Cross-sectional and survival analyses of breastfeeding at 4 months compared with non-nutritive sucking at 1 month showed no significant relationship with thumb sucking, but a significant relationship with pacifier use, with increasing frequency of pacifier use related to a decline in breastfeeding duration. Riva and coworkers³² studied 1601 women in Italy and showed that pacifier use was associated with an elevated hazard ratio of 1.18 (95% confidence interval [CI], 1.04-1.34) for breastfeeding cessation in adjusted analyses.

In the only published US study, Howard and colleagues³³ reported on the effects of early pacifier use on breastfeeding duration among 265 infants in the Rochester, New York area on the basis of maternal telephone interviews at 2, 6, 12, and 24 weeks and every 90 days thereafter until the breastfeeding ended. Results were adjusted for factors such as maternal age, breastfeeding goals, and plans to work. Pacifier introduction by 6 weeks was significantly associated with shortened duration of some breastfeeding (hazard ratio [HR] = 1.61; 95% CI, 1.19-2.19; P = .002), as was a plan to return to work (HR = 1.42). Digit sucking was not examined and interactions were not assessed.

We found only one prospective study³¹ that considered the effects of both pacifier and digit sucking, and one study that considered the effects of pacifier and plans to return to work³³ on breastfeeding duration. However, no studies simultaneously looked at the effects of maternal employment or child care, pacifier use, digit sucking, and any potential interactions, although they have been shown to be individually associated with cessation of breastfeeding. Thus, the purpose of this study was to assess the associations of non-nutritive sucking (pacifiers and fingers) with cessation of breastfeeding, while considering child care attendance, from birth to age 6 months, using a longitudinal study design in a sample of children in the United States.

Methods

The data were collected as part of a larger, prospective study of a birth cohort assessing fluoride exposures longitudinally and relationships with dental caries and dental fluorosis.^16,37-43 Mothers were recruited at the time of their infants’ births at 8 hospitals in eastern Iowa from March 1992 to February 1995, using appropriate informed consent procedures approved by the Institutional Review Board. The recruitment questionnaire assessed household smoking patterns during pregnancy, whether women planned to breastfeed, and other demographic factors.

Information regarding infants’ weight, feeding practices (breastfeeding vs bottle-feeding), non-nutritive sucking (pacifier use and sucking thumb or fingers), child care attendance (number of full or half days), maternal smoking, otitis media experience, and antibiotic use was collected by mailed questionnaire sent at 6 weeks, 3 months, and 6 months of age. Each questionnaire concerned the preceding time period. Nonrespondents received follow-up mailings after 3 weeks and telephone follow-up after 6 weeks. Direct validation of responses was not conducted, but subjects were contacted by mail or telephone, when necessary, to clarify or correct responses. Data were double-entered and verified.

Breastfeeding and bottle-feeding practices for each period were summarized in 3 ways: (1) exclusive breastfeeding, (2) any breastfeeding, and (3) mostly bottle-feeding (defined as at least 75% of estimated total calories based on body weight from formula, milk, or juice). These definitions generally correspond to those proposed by Labbok and Krasovec⁴⁴ of full, almost exclusive breastfeeding, and low, partial breastfeeding, respectively.

Time until cessation of all breastfeeding was modeled using the Cox proportional hazard regression model⁴⁵ against 3 main factors of interest: pacifier use (yes/no), digit sucking (yes/no), and child care attendance (total number of child care days). Since no information was collected regarding maternal employment, we considered child care attendance as a proxy. Pacifier use and digit sucking were coded “yes” if the child started using the pacifier or sucking on the digit, respectively, any time during the first 6 weeks of life. Main effects, 2-way interactions among these variables, and nonlinear effects of child care days were tested while adjusting for maternal and paternal age and education, family income, breastfeeding plans, maternal smoking, infant’s sex, and infant antibiotic use. We used the likelihood ratio test to assess significance at an alpha level of 0.05, and the statistical analyses were conducted using PROC PHREG in SAS software.⁴⁶

Results

The number of mothers who were successfully recruited and who provided at least one subsequent completed questionnaire was 1387. There were 1236 (89%) respondents at 6 weeks, 1196 (86%) at 3 months, and 1048 (76%) at 6 months.

Table 1 summarizes the study sample at baseline recruitment. Approximately two thirds of mothers and fathers had at least some college education; 76% had family income of at least $20,000; 95% were white; 43% of the infants were the mother’s first-born child; and 65% of the mothers said they were planning to breastfeed their infants.

Table 2 summarizes the breastfeeding practices of the cohort by presenting the percentages of infants at each time point with different feeding practices. Approximately 46% reported some breastfeeding on the 6-week questionnaire, declining to 36% at 3 months and 27% at 6 months. Only 16% of the infants were exclusively breastfed at 6 weeks, dropping to 1% by 6 months. A high percentage of infants were mostly bottle fed at each of the 3 corresponding time periods.

Table 2 also summarizes the patterns of non-nutritive sucking across the infant ages. A high percentage of the infants practiced some form of non-nutritive sucking during each period (86.3%, 92.0%, and 86.3% at 6 weeks, 3 months, and 6 months, respectively). From the 6-week to 6-month responses, pacifier use declined from 81% to 59%, while digit sucking increased from 50% to 83% and then declined to 76%. Table 3 summarizes child care attendance during the 6 months, with half days and full days of child care added together. Thirty-four percent of the infants attended some child care, with approximately 12% receiving more than 25 full days of child care by the age of 6 months or the time of censor/failure, where censor in this case is loss to follow-up prior to reaching 6 months of age.

We next analyzed the data using Cox regression, an analysis method designed for longitudinal data on event times, such as time until death. The outcome variable was time until cessation of all breastfeeding. The median failure time (cessation) was 72 days (95% CI, 68-78) with interquartile range from 53 to 192 days. Seventy-four percent had ceased breastfeeding by 6 months and 26% were censored because of continued breastfeeding at 6 months when analysis ended or earlier loss to follow-up.

Table 4 reports the relative hazard ratios and 95% confidence intervals at various levels of child care, pacifier use, and digit sucking, while adjusting for the other potential confounders considered (see Methods section). The baseline category (or reference cell) is a child with no child care and no non-nutritive sucking. We see from Table 4 that the estimated risk of breastfeeding cessation is the highest, with a value of 1.88 (95% CI, 1.36-2.62), for a child who sucks on both pacifier and digit at no child care days. This hazard ratio drops to 1.52 (95% CI, 1.03-2.25) at 15 child care days and then becomes nonsignificant at 30 and 60 child care days.

Our results in Table 4 also show that pacifier use at zero child care days has a significant effect in that a child who sucks only on a pacifier has a 67% increase in the hazard of cessation of breastfeeding, compared with a child with no non-nutritive sucking. At higher levels of child care days, this effect changes and becomes a protective effect, although this effect was not significant at 15 child care days, was significant at 30 child care days, and was borderline significant at 60 child care days. Finally, the effect of digit sucking and child care by themselves tended not to be significant at the 0.05 level, with the one exception at 15 child care days where there is a significant effect of 1.41 (95% CI, 1.02-1.96).

Discussion

Our findings concerning pacifiers are generally consistent with several recent studies that have demonstrated associations between pacifier use and reduced breastfeeding, including the few reported longitudinal studies. However, these other studies did not control for child care attendance. We found that the effect of pacifier use changed with increasing number of child care days. For example, in the absence of child care, children who sucked on a pacifier were about 1.7 times as likely to cease breastfeeding than children who did not use a pacifier. For 15 days to 60 days of child care, the hazard ratios were less than 1.0, with results statistically significant at only 30 days.

Furthermore, our analyses showed the joint effect of pacifier use and digit sucking at various child care days. We found a significant reduction in breastfeeding for children who use both pacifier and a digit by the age of 6 weeks. But this joint non-nutritive effect reduces to being nonsignificant with 30 or more child care days. Although we found that digit sucking and child care days by themselves had little effect on cessation of breastfeeding, it was important to consider them because these variables significantly interacted with pacifier use.

Our study found that for infants who did not attend child care, pacifier use significantly increased the odds of breastfeeding cessation, as did the combination of pacifier use and digit sucking. However, digit sucking with no pacifier use in the absence of child care did not increase the odds of breastfeeding cessation. In contrast, for infants who attended child care for 30 days in the first 6 months of life, pacifier use alone appeared to be somewhat protective in maintaining breastfeeding, while digit sucking, either alone or in combination with pacifier, increased the odds of breastfeeding cessation, with significance at 15 days. It is possible that pacifiers were used sparingly in child care, whereas digits were available and more widely used, so that non-nutritive sucking interference with breastfeeding was more strongly influenced by digit sucking. Alternatively, it is possible that mothers who placed their infants in child care early in life used pacifiers differently than mothers who did not. That is, for non-child care infants, pacifiers may have been part of a planned strategy to wean from breastfeeding, whereas for children in child care, pacifiers may have been part of a planned strategy to encourage sucking behavior and comfort children until the mother was available for breastfeeding. In such a scenario, digit sucking was less under parental control, particularly at child care, so that it may have interfered with breastfeeding despite parental planning or desires.

Limitations

There are several limitations when considering our study’s findings. The study group was not a probability sample fully representative of a defined population. It was of generally high socioeconomic status and, representative of Iowa, had little minority inclusion. Respondents were more educated than nonrespondents.³⁹ Although response rates were generally favorable, approximately 100 to 300 did not respond at a given time point, resulting in censoring of 26% of the cases. Data on breastfeeding, sucking, childcare, and so forth were collected at 3 discrete time points and not on a more frequent, daily, weekly, or monthly basis. Although recall bias was limited by the short-term nature of recall with 6-week and 3-month intervals, it could have an effect on results. Since so few infants exclusively breastfed, any breastfeeding was the only suitable dependent variable. No maternal employment data were collected and quantification of pacifier use was not included.

Only our study and that of Howard and colleagues³³ reported results from the United States. The statistical analyses by Howard and colleagues concerning pacifier use adjusted for a number of factors, including plans to return to work, family and paternal preferences for breastfeeding, and breastfeeding goal. Our study adjusted for plans to breastfeed and demographic factors while assessing the effects of pacifier use, digit sucking, and number of child care days. However, neither study specifically assessed reasons for use of the pacifier, in particular, in relation to work and child care requirements. So pacifiers could have been used to facilitate weaning, thus resulting in the association with reduced breastfeeding. Also, there may be other confounding differences between those using pacifiers and those who did not.

Although decisions by mothers to return to work, or for other reasons, have their infants attend child care were not generally associated with reductions in breastfeeding, our results suggest that child care has an important impact on determining the relationships between non-nutritive sucking behaviors and cessation of breastfeeding. It has been suggested that infants’ abilities to easily and successfully breastfeed are adversely affected by non-nutritive sucking, resulting in reductions in the frequency and consistency of the breastfeeding sessions. Our data support the concept. However, it is important to acknowledge that decisions to stop breastfeeding (often prior to return to work) may have preceded and led to the increase in non-nutritive sucking, rather than sucking leading to cessation of breastfeeding. That is, after the decision has been made to stop breastfeeding, a pacifier may be introduced to ease the transition to bottle feeding.

Additional studies involving in-depth interviews concerning initial and subsequent breastfeeding, employment, and child care plans would be warranted to address this question further. In addition, more controlled studies to determine whether there is any biological relationship between non-nutritive sucking and breastfeeding difficulties are warranted. Clearly, the social, biological, and economic factors involved in decisions to initiate and cease breastfeeding are complex and will require more study, both in the United States and throughout the world.

Acknowledgments

Our study was supported in part by National Institutes of Health grants #RO1-DE09551 and #P30-DE10126 and the University of Iowa’s Obermann Center for Advanced Study. We thank the staff of the Iowa Fluoride Study for their assistance in implementing the study, and Tina Craig for manuscript preparation.

Breastfeeding is associated with lower rates of infant mortality and morbidity,^1-6 a reduced rate of sudden infant death syndrome (SIDS),^7,8 delayed resumption of fertility,⁹ and reduced health care cost.^10,11 The American Academy of Family Physicians has issued a policy statement supporting breastfeeding as the optimal form of nutrition for infants¹² and the American Academy of Pediatrics recommends that infants should be breastfed for at least 12 months.¹³ Therefore, it is important to understand the factors associated with reduced breastfeeding. In previous studies, the factors associated with reduced breastfeeding included maternal employment,¹⁴ child care attendance,¹⁵ maternal smoking,^14,16,17 and demographic factors.^16,18,19

Several recent studies have also identified an association between non-nutritive sucking (eg, pacifiers) and reduced breastfeeding^20-35 that is consistent with the World Health Organization (and UNICEF) recommendation that pacifiers not be used by breastfeeding infants.³⁶ Cross-sectional investigations in Sweden,^20-22 Brazil,²³ New Zealand,²⁴ England,²⁵ Greece,²⁶ and Sweden and Norway²⁷ found strong associations between pacifier use and reduced breastfeeding (either less exclusive breastfeeding, shorter duration of breastfeeding, or breastfeeding problems), with only one²⁶ not reporting statistically significant findings.

Of particular interest were several longitudinal studies in Brazil (2 studies), Sweden, Italy, and the United States. In Brazil, one found that pacifier users had an adjusted relative risk of 2.87 for weaning,²⁸ and the other an adjusted odds ratio of 2.5 for the cessation of breastfeeding associated with pacifier use.²⁹

Hörnell and colleagues³⁰ and Aarts and colleagues³¹ reported longitudinal data from 506 mothers’ daily infant feeding practices in Uppsala, Sweden. All mothers had at least one previous child breastfed at least 4 months and were planning to breastfeed the study child for at least 6 months. Thumb sucking was not associated with the breastfeeding pattern, but infants using a pacifier frequently had approximately 1 less breastfeeding session and 15 minutes to 30 minutes less total breastfeeding time per day than those not using a pacifier at 2, 4, 8, and 12-week follow-up points. Cross-sectional and survival analyses of breastfeeding at 4 months compared with non-nutritive sucking at 1 month showed no significant relationship with thumb sucking, but a significant relationship with pacifier use, with increasing frequency of pacifier use related to a decline in breastfeeding duration. Riva and coworkers³² studied 1601 women in Italy and showed that pacifier use was associated with an elevated hazard ratio of 1.18 (95% confidence interval [CI], 1.04-1.34) for breastfeeding cessation in adjusted analyses.

In the only published US study, Howard and colleagues³³ reported on the effects of early pacifier use on breastfeeding duration among 265 infants in the Rochester, New York area on the basis of maternal telephone interviews at 2, 6, 12, and 24 weeks and every 90 days thereafter until the breastfeeding ended. Results were adjusted for factors such as maternal age, breastfeeding goals, and plans to work. Pacifier introduction by 6 weeks was significantly associated with shortened duration of some breastfeeding (hazard ratio [HR] = 1.61; 95% CI, 1.19-2.19; P = .002), as was a plan to return to work (HR = 1.42). Digit sucking was not examined and interactions were not assessed.

We found only one prospective study³¹ that considered the effects of both pacifier and digit sucking, and one study that considered the effects of pacifier and plans to return to work³³ on breastfeeding duration. However, no studies simultaneously looked at the effects of maternal employment or child care, pacifier use, digit sucking, and any potential interactions, although they have been shown to be individually associated with cessation of breastfeeding. Thus, the purpose of this study was to assess the associations of non-nutritive sucking (pacifiers and fingers) with cessation of breastfeeding, while considering child care attendance, from birth to age 6 months, using a longitudinal study design in a sample of children in the United States.

Methods

The data were collected as part of a larger, prospective study of a birth cohort assessing fluoride exposures longitudinally and relationships with dental caries and dental fluorosis.^16,37-43 Mothers were recruited at the time of their infants’ births at 8 hospitals in eastern Iowa from March 1992 to February 1995, using appropriate informed consent procedures approved by the Institutional Review Board. The recruitment questionnaire assessed household smoking patterns during pregnancy, whether women planned to breastfeed, and other demographic factors.

Information regarding infants’ weight, feeding practices (breastfeeding vs bottle-feeding), non-nutritive sucking (pacifier use and sucking thumb or fingers), child care attendance (number of full or half days), maternal smoking, otitis media experience, and antibiotic use was collected by mailed questionnaire sent at 6 weeks, 3 months, and 6 months of age. Each questionnaire concerned the preceding time period. Nonrespondents received follow-up mailings after 3 weeks and telephone follow-up after 6 weeks. Direct validation of responses was not conducted, but subjects were contacted by mail or telephone, when necessary, to clarify or correct responses. Data were double-entered and verified.

Breastfeeding and bottle-feeding practices for each period were summarized in 3 ways: (1) exclusive breastfeeding, (2) any breastfeeding, and (3) mostly bottle-feeding (defined as at least 75% of estimated total calories based on body weight from formula, milk, or juice). These definitions generally correspond to those proposed by Labbok and Krasovec⁴⁴ of full, almost exclusive breastfeeding, and low, partial breastfeeding, respectively.

Time until cessation of all breastfeeding was modeled using the Cox proportional hazard regression model⁴⁵ against 3 main factors of interest: pacifier use (yes/no), digit sucking (yes/no), and child care attendance (total number of child care days). Since no information was collected regarding maternal employment, we considered child care attendance as a proxy. Pacifier use and digit sucking were coded “yes” if the child started using the pacifier or sucking on the digit, respectively, any time during the first 6 weeks of life. Main effects, 2-way interactions among these variables, and nonlinear effects of child care days were tested while adjusting for maternal and paternal age and education, family income, breastfeeding plans, maternal smoking, infant’s sex, and infant antibiotic use. We used the likelihood ratio test to assess significance at an alpha level of 0.05, and the statistical analyses were conducted using PROC PHREG in SAS software.⁴⁶

Results

The number of mothers who were successfully recruited and who provided at least one subsequent completed questionnaire was 1387. There were 1236 (89%) respondents at 6 weeks, 1196 (86%) at 3 months, and 1048 (76%) at 6 months.

Table 1 summarizes the study sample at baseline recruitment. Approximately two thirds of mothers and fathers had at least some college education; 76% had family income of at least $20,000; 95% were white; 43% of the infants were the mother’s first-born child; and 65% of the mothers said they were planning to breastfeed their infants.

Table 2 summarizes the breastfeeding practices of the cohort by presenting the percentages of infants at each time point with different feeding practices. Approximately 46% reported some breastfeeding on the 6-week questionnaire, declining to 36% at 3 months and 27% at 6 months. Only 16% of the infants were exclusively breastfed at 6 weeks, dropping to 1% by 6 months. A high percentage of infants were mostly bottle fed at each of the 3 corresponding time periods.

Table 2 also summarizes the patterns of non-nutritive sucking across the infant ages. A high percentage of the infants practiced some form of non-nutritive sucking during each period (86.3%, 92.0%, and 86.3% at 6 weeks, 3 months, and 6 months, respectively). From the 6-week to 6-month responses, pacifier use declined from 81% to 59%, while digit sucking increased from 50% to 83% and then declined to 76%. Table 3 summarizes child care attendance during the 6 months, with half days and full days of child care added together. Thirty-four percent of the infants attended some child care, with approximately 12% receiving more than 25 full days of child care by the age of 6 months or the time of censor/failure, where censor in this case is loss to follow-up prior to reaching 6 months of age.

We next analyzed the data using Cox regression, an analysis method designed for longitudinal data on event times, such as time until death. The outcome variable was time until cessation of all breastfeeding. The median failure time (cessation) was 72 days (95% CI, 68-78) with interquartile range from 53 to 192 days. Seventy-four percent had ceased breastfeeding by 6 months and 26% were censored because of continued breastfeeding at 6 months when analysis ended or earlier loss to follow-up.

Table 4 reports the relative hazard ratios and 95% confidence intervals at various levels of child care, pacifier use, and digit sucking, while adjusting for the other potential confounders considered (see Methods section). The baseline category (or reference cell) is a child with no child care and no non-nutritive sucking. We see from Table 4 that the estimated risk of breastfeeding cessation is the highest, with a value of 1.88 (95% CI, 1.36-2.62), for a child who sucks on both pacifier and digit at no child care days. This hazard ratio drops to 1.52 (95% CI, 1.03-2.25) at 15 child care days and then becomes nonsignificant at 30 and 60 child care days.

Our results in Table 4 also show that pacifier use at zero child care days has a significant effect in that a child who sucks only on a pacifier has a 67% increase in the hazard of cessation of breastfeeding, compared with a child with no non-nutritive sucking. At higher levels of child care days, this effect changes and becomes a protective effect, although this effect was not significant at 15 child care days, was significant at 30 child care days, and was borderline significant at 60 child care days. Finally, the effect of digit sucking and child care by themselves tended not to be significant at the 0.05 level, with the one exception at 15 child care days where there is a significant effect of 1.41 (95% CI, 1.02-1.96).

Discussion

Our findings concerning pacifiers are generally consistent with several recent studies that have demonstrated associations between pacifier use and reduced breastfeeding, including the few reported longitudinal studies. However, these other studies did not control for child care attendance. We found that the effect of pacifier use changed with increasing number of child care days. For example, in the absence of child care, children who sucked on a pacifier were about 1.7 times as likely to cease breastfeeding than children who did not use a pacifier. For 15 days to 60 days of child care, the hazard ratios were less than 1.0, with results statistically significant at only 30 days.

Furthermore, our analyses showed the joint effect of pacifier use and digit sucking at various child care days. We found a significant reduction in breastfeeding for children who use both pacifier and a digit by the age of 6 weeks. But this joint non-nutritive effect reduces to being nonsignificant with 30 or more child care days. Although we found that digit sucking and child care days by themselves had little effect on cessation of breastfeeding, it was important to consider them because these variables significantly interacted with pacifier use.

Our study found that for infants who did not attend child care, pacifier use significantly increased the odds of breastfeeding cessation, as did the combination of pacifier use and digit sucking. However, digit sucking with no pacifier use in the absence of child care did not increase the odds of breastfeeding cessation. In contrast, for infants who attended child care for 30 days in the first 6 months of life, pacifier use alone appeared to be somewhat protective in maintaining breastfeeding, while digit sucking, either alone or in combination with pacifier, increased the odds of breastfeeding cessation, with significance at 15 days. It is possible that pacifiers were used sparingly in child care, whereas digits were available and more widely used, so that non-nutritive sucking interference with breastfeeding was more strongly influenced by digit sucking. Alternatively, it is possible that mothers who placed their infants in child care early in life used pacifiers differently than mothers who did not. That is, for non-child care infants, pacifiers may have been part of a planned strategy to wean from breastfeeding, whereas for children in child care, pacifiers may have been part of a planned strategy to encourage sucking behavior and comfort children until the mother was available for breastfeeding. In such a scenario, digit sucking was less under parental control, particularly at child care, so that it may have interfered with breastfeeding despite parental planning or desires.

Limitations

There are several limitations when considering our study’s findings. The study group was not a probability sample fully representative of a defined population. It was of generally high socioeconomic status and, representative of Iowa, had little minority inclusion. Respondents were more educated than nonrespondents.³⁹ Although response rates were generally favorable, approximately 100 to 300 did not respond at a given time point, resulting in censoring of 26% of the cases. Data on breastfeeding, sucking, childcare, and so forth were collected at 3 discrete time points and not on a more frequent, daily, weekly, or monthly basis. Although recall bias was limited by the short-term nature of recall with 6-week and 3-month intervals, it could have an effect on results. Since so few infants exclusively breastfed, any breastfeeding was the only suitable dependent variable. No maternal employment data were collected and quantification of pacifier use was not included.

Only our study and that of Howard and colleagues³³ reported results from the United States. The statistical analyses by Howard and colleagues concerning pacifier use adjusted for a number of factors, including plans to return to work, family and paternal preferences for breastfeeding, and breastfeeding goal. Our study adjusted for plans to breastfeed and demographic factors while assessing the effects of pacifier use, digit sucking, and number of child care days. However, neither study specifically assessed reasons for use of the pacifier, in particular, in relation to work and child care requirements. So pacifiers could have been used to facilitate weaning, thus resulting in the association with reduced breastfeeding. Also, there may be other confounding differences between those using pacifiers and those who did not.

Although decisions by mothers to return to work, or for other reasons, have their infants attend child care were not generally associated with reductions in breastfeeding, our results suggest that child care has an important impact on determining the relationships between non-nutritive sucking behaviors and cessation of breastfeeding. It has been suggested that infants’ abilities to easily and successfully breastfeed are adversely affected by non-nutritive sucking, resulting in reductions in the frequency and consistency of the breastfeeding sessions. Our data support the concept. However, it is important to acknowledge that decisions to stop breastfeeding (often prior to return to work) may have preceded and led to the increase in non-nutritive sucking, rather than sucking leading to cessation of breastfeeding. That is, after the decision has been made to stop breastfeeding, a pacifier may be introduced to ease the transition to bottle feeding.

Additional studies involving in-depth interviews concerning initial and subsequent breastfeeding, employment, and child care plans would be warranted to address this question further. In addition, more controlled studies to determine whether there is any biological relationship between non-nutritive sucking and breastfeeding difficulties are warranted. Clearly, the social, biological, and economic factors involved in decisions to initiate and cease breastfeeding are complex and will require more study, both in the United States and throughout the world.

Acknowledgments

Our study was supported in part by National Institutes of Health grants #RO1-DE09551 and #P30-DE10126 and the University of Iowa’s Obermann Center for Advanced Study. We thank the staff of the Iowa Fluoride Study for their assistance in implementing the study, and Tina Craig for manuscript preparation.

Breastfeeding is associated with lower rates of infant mortality and morbidity,^1-6 a reduced rate of sudden infant death syndrome (SIDS),^7,8 delayed resumption of fertility,⁹ and reduced health care cost.^10,11 The American Academy of Family Physicians has issued a policy statement supporting breastfeeding as the optimal form of nutrition for infants¹² and the American Academy of Pediatrics recommends that infants should be breastfed for at least 12 months.¹³ Therefore, it is important to understand the factors associated with reduced breastfeeding. In previous studies, the factors associated with reduced breastfeeding included maternal employment,¹⁴ child care attendance,¹⁵ maternal smoking,^14,16,17 and demographic factors.^16,18,19

Several recent studies have also identified an association between non-nutritive sucking (eg, pacifiers) and reduced breastfeeding^20-35 that is consistent with the World Health Organization (and UNICEF) recommendation that pacifiers not be used by breastfeeding infants.³⁶ Cross-sectional investigations in Sweden,^20-22 Brazil,²³ New Zealand,²⁴ England,²⁵ Greece,²⁶ and Sweden and Norway²⁷ found strong associations between pacifier use and reduced breastfeeding (either less exclusive breastfeeding, shorter duration of breastfeeding, or breastfeeding problems), with only one²⁶ not reporting statistically significant findings.

Of particular interest were several longitudinal studies in Brazil (2 studies), Sweden, Italy, and the United States. In Brazil, one found that pacifier users had an adjusted relative risk of 2.87 for weaning,²⁸ and the other an adjusted odds ratio of 2.5 for the cessation of breastfeeding associated with pacifier use.²⁹

Hörnell and colleagues³⁰ and Aarts and colleagues³¹ reported longitudinal data from 506 mothers’ daily infant feeding practices in Uppsala, Sweden. All mothers had at least one previous child breastfed at least 4 months and were planning to breastfeed the study child for at least 6 months. Thumb sucking was not associated with the breastfeeding pattern, but infants using a pacifier frequently had approximately 1 less breastfeeding session and 15 minutes to 30 minutes less total breastfeeding time per day than those not using a pacifier at 2, 4, 8, and 12-week follow-up points. Cross-sectional and survival analyses of breastfeeding at 4 months compared with non-nutritive sucking at 1 month showed no significant relationship with thumb sucking, but a significant relationship with pacifier use, with increasing frequency of pacifier use related to a decline in breastfeeding duration. Riva and coworkers³² studied 1601 women in Italy and showed that pacifier use was associated with an elevated hazard ratio of 1.18 (95% confidence interval [CI], 1.04-1.34) for breastfeeding cessation in adjusted analyses.

In the only published US study, Howard and colleagues³³ reported on the effects of early pacifier use on breastfeeding duration among 265 infants in the Rochester, New York area on the basis of maternal telephone interviews at 2, 6, 12, and 24 weeks and every 90 days thereafter until the breastfeeding ended. Results were adjusted for factors such as maternal age, breastfeeding goals, and plans to work. Pacifier introduction by 6 weeks was significantly associated with shortened duration of some breastfeeding (hazard ratio [HR] = 1.61; 95% CI, 1.19-2.19; P = .002), as was a plan to return to work (HR = 1.42). Digit sucking was not examined and interactions were not assessed.

We found only one prospective study³¹ that considered the effects of both pacifier and digit sucking, and one study that considered the effects of pacifier and plans to return to work³³ on breastfeeding duration. However, no studies simultaneously looked at the effects of maternal employment or child care, pacifier use, digit sucking, and any potential interactions, although they have been shown to be individually associated with cessation of breastfeeding. Thus, the purpose of this study was to assess the associations of non-nutritive sucking (pacifiers and fingers) with cessation of breastfeeding, while considering child care attendance, from birth to age 6 months, using a longitudinal study design in a sample of children in the United States.

Methods

The data were collected as part of a larger, prospective study of a birth cohort assessing fluoride exposures longitudinally and relationships with dental caries and dental fluorosis.^16,37-43 Mothers were recruited at the time of their infants’ births at 8 hospitals in eastern Iowa from March 1992 to February 1995, using appropriate informed consent procedures approved by the Institutional Review Board. The recruitment questionnaire assessed household smoking patterns during pregnancy, whether women planned to breastfeed, and other demographic factors.

Information regarding infants’ weight, feeding practices (breastfeeding vs bottle-feeding), non-nutritive sucking (pacifier use and sucking thumb or fingers), child care attendance (number of full or half days), maternal smoking, otitis media experience, and antibiotic use was collected by mailed questionnaire sent at 6 weeks, 3 months, and 6 months of age. Each questionnaire concerned the preceding time period. Nonrespondents received follow-up mailings after 3 weeks and telephone follow-up after 6 weeks. Direct validation of responses was not conducted, but subjects were contacted by mail or telephone, when necessary, to clarify or correct responses. Data were double-entered and verified.

Breastfeeding and bottle-feeding practices for each period were summarized in 3 ways: (1) exclusive breastfeeding, (2) any breastfeeding, and (3) mostly bottle-feeding (defined as at least 75% of estimated total calories based on body weight from formula, milk, or juice). These definitions generally correspond to those proposed by Labbok and Krasovec⁴⁴ of full, almost exclusive breastfeeding, and low, partial breastfeeding, respectively.

Time until cessation of all breastfeeding was modeled using the Cox proportional hazard regression model⁴⁵ against 3 main factors of interest: pacifier use (yes/no), digit sucking (yes/no), and child care attendance (total number of child care days). Since no information was collected regarding maternal employment, we considered child care attendance as a proxy. Pacifier use and digit sucking were coded “yes” if the child started using the pacifier or sucking on the digit, respectively, any time during the first 6 weeks of life. Main effects, 2-way interactions among these variables, and nonlinear effects of child care days were tested while adjusting for maternal and paternal age and education, family income, breastfeeding plans, maternal smoking, infant’s sex, and infant antibiotic use. We used the likelihood ratio test to assess significance at an alpha level of 0.05, and the statistical analyses were conducted using PROC PHREG in SAS software.⁴⁶

Results

The number of mothers who were successfully recruited and who provided at least one subsequent completed questionnaire was 1387. There were 1236 (89%) respondents at 6 weeks, 1196 (86%) at 3 months, and 1048 (76%) at 6 months.

Table 1 summarizes the study sample at baseline recruitment. Approximately two thirds of mothers and fathers had at least some college education; 76% had family income of at least $20,000; 95% were white; 43% of the infants were the mother’s first-born child; and 65% of the mothers said they were planning to breastfeed their infants.

Table 2 summarizes the breastfeeding practices of the cohort by presenting the percentages of infants at each time point with different feeding practices. Approximately 46% reported some breastfeeding on the 6-week questionnaire, declining to 36% at 3 months and 27% at 6 months. Only 16% of the infants were exclusively breastfed at 6 weeks, dropping to 1% by 6 months. A high percentage of infants were mostly bottle fed at each of the 3 corresponding time periods.

Table 2 also summarizes the patterns of non-nutritive sucking across the infant ages. A high percentage of the infants practiced some form of non-nutritive sucking during each period (86.3%, 92.0%, and 86.3% at 6 weeks, 3 months, and 6 months, respectively). From the 6-week to 6-month responses, pacifier use declined from 81% to 59%, while digit sucking increased from 50% to 83% and then declined to 76%. Table 3 summarizes child care attendance during the 6 months, with half days and full days of child care added together. Thirty-four percent of the infants attended some child care, with approximately 12% receiving more than 25 full days of child care by the age of 6 months or the time of censor/failure, where censor in this case is loss to follow-up prior to reaching 6 months of age.

We next analyzed the data using Cox regression, an analysis method designed for longitudinal data on event times, such as time until death. The outcome variable was time until cessation of all breastfeeding. The median failure time (cessation) was 72 days (95% CI, 68-78) with interquartile range from 53 to 192 days. Seventy-four percent had ceased breastfeeding by 6 months and 26% were censored because of continued breastfeeding at 6 months when analysis ended or earlier loss to follow-up.

Table 4 reports the relative hazard ratios and 95% confidence intervals at various levels of child care, pacifier use, and digit sucking, while adjusting for the other potential confounders considered (see Methods section). The baseline category (or reference cell) is a child with no child care and no non-nutritive sucking. We see from Table 4 that the estimated risk of breastfeeding cessation is the highest, with a value of 1.88 (95% CI, 1.36-2.62), for a child who sucks on both pacifier and digit at no child care days. This hazard ratio drops to 1.52 (95% CI, 1.03-2.25) at 15 child care days and then becomes nonsignificant at 30 and 60 child care days.

Our results in Table 4 also show that pacifier use at zero child care days has a significant effect in that a child who sucks only on a pacifier has a 67% increase in the hazard of cessation of breastfeeding, compared with a child with no non-nutritive sucking. At higher levels of child care days, this effect changes and becomes a protective effect, although this effect was not significant at 15 child care days, was significant at 30 child care days, and was borderline significant at 60 child care days. Finally, the effect of digit sucking and child care by themselves tended not to be significant at the 0.05 level, with the one exception at 15 child care days where there is a significant effect of 1.41 (95% CI, 1.02-1.96).

Discussion

Our findings concerning pacifiers are generally consistent with several recent studies that have demonstrated associations between pacifier use and reduced breastfeeding, including the few reported longitudinal studies. However, these other studies did not control for child care attendance. We found that the effect of pacifier use changed with increasing number of child care days. For example, in the absence of child care, children who sucked on a pacifier were about 1.7 times as likely to cease breastfeeding than children who did not use a pacifier. For 15 days to 60 days of child care, the hazard ratios were less than 1.0, with results statistically significant at only 30 days.

Furthermore, our analyses showed the joint effect of pacifier use and digit sucking at various child care days. We found a significant reduction in breastfeeding for children who use both pacifier and a digit by the age of 6 weeks. But this joint non-nutritive effect reduces to being nonsignificant with 30 or more child care days. Although we found that digit sucking and child care days by themselves had little effect on cessation of breastfeeding, it was important to consider them because these variables significantly interacted with pacifier use.

Our study found that for infants who did not attend child care, pacifier use significantly increased the odds of breastfeeding cessation, as did the combination of pacifier use and digit sucking. However, digit sucking with no pacifier use in the absence of child care did not increase the odds of breastfeeding cessation. In contrast, for infants who attended child care for 30 days in the first 6 months of life, pacifier use alone appeared to be somewhat protective in maintaining breastfeeding, while digit sucking, either alone or in combination with pacifier, increased the odds of breastfeeding cessation, with significance at 15 days. It is possible that pacifiers were used sparingly in child care, whereas digits were available and more widely used, so that non-nutritive sucking interference with breastfeeding was more strongly influenced by digit sucking. Alternatively, it is possible that mothers who placed their infants in child care early in life used pacifiers differently than mothers who did not. That is, for non-child care infants, pacifiers may have been part of a planned strategy to wean from breastfeeding, whereas for children in child care, pacifiers may have been part of a planned strategy to encourage sucking behavior and comfort children until the mother was available for breastfeeding. In such a scenario, digit sucking was less under parental control, particularly at child care, so that it may have interfered with breastfeeding despite parental planning or desires.

Limitations

There are several limitations when considering our study’s findings. The study group was not a probability sample fully representative of a defined population. It was of generally high socioeconomic status and, representative of Iowa, had little minority inclusion. Respondents were more educated than nonrespondents.³⁹ Although response rates were generally favorable, approximately 100 to 300 did not respond at a given time point, resulting in censoring of 26% of the cases. Data on breastfeeding, sucking, childcare, and so forth were collected at 3 discrete time points and not on a more frequent, daily, weekly, or monthly basis. Although recall bias was limited by the short-term nature of recall with 6-week and 3-month intervals, it could have an effect on results. Since so few infants exclusively breastfed, any breastfeeding was the only suitable dependent variable. No maternal employment data were collected and quantification of pacifier use was not included.

Only our study and that of Howard and colleagues³³ reported results from the United States. The statistical analyses by Howard and colleagues concerning pacifier use adjusted for a number of factors, including plans to return to work, family and paternal preferences for breastfeeding, and breastfeeding goal. Our study adjusted for plans to breastfeed and demographic factors while assessing the effects of pacifier use, digit sucking, and number of child care days. However, neither study specifically assessed reasons for use of the pacifier, in particular, in relation to work and child care requirements. So pacifiers could have been used to facilitate weaning, thus resulting in the association with reduced breastfeeding. Also, there may be other confounding differences between those using pacifiers and those who did not.

Although decisions by mothers to return to work, or for other reasons, have their infants attend child care were not generally associated with reductions in breastfeeding, our results suggest that child care has an important impact on determining the relationships between non-nutritive sucking behaviors and cessation of breastfeeding. It has been suggested that infants’ abilities to easily and successfully breastfeed are adversely affected by non-nutritive sucking, resulting in reductions in the frequency and consistency of the breastfeeding sessions. Our data support the concept. However, it is important to acknowledge that decisions to stop breastfeeding (often prior to return to work) may have preceded and led to the increase in non-nutritive sucking, rather than sucking leading to cessation of breastfeeding. That is, after the decision has been made to stop breastfeeding, a pacifier may be introduced to ease the transition to bottle feeding.

Additional studies involving in-depth interviews concerning initial and subsequent breastfeeding, employment, and child care plans would be warranted to address this question further. In addition, more controlled studies to determine whether there is any biological relationship between non-nutritive sucking and breastfeeding difficulties are warranted. Clearly, the social, biological, and economic factors involved in decisions to initiate and cease breastfeeding are complex and will require more study, both in the United States and throughout the world.

Acknowledgments

Our study was supported in part by National Institutes of Health grants #RO1-DE09551 and #P30-DE10126 and the University of Iowa’s Obermann Center for Advanced Study. We thank the staff of the Iowa Fluoride Study for their assistance in implementing the study, and Tina Craig for manuscript preparation.

Recent studies have emphasized the importance of tight metabolic control in combination with state-of-the-art management of other risk factors to prevent macrovascular and microvascular complications in patients with type 2 diabetes mellitus.^1-5 Guidelines for diabetes care recommend systematic monitoring of patients’ health status, including metabolic control, cardiovascular risk factors, and desired outcome of care.^6-8

The formulation of clinical guidelines alone, however, is insufficient to improve actual care.^9,10 Strategies to reinforce the guidelines in daily practice include monitoring the patient’s clinical condition over a given period of time, feedback to the clinician about the outcome, audit of clinical performance, academic detailing by peers, and evidence-based guidelines.^10-12 Monitoring and feedback with systematic follow-up of relevant treatment targets enhanced a proactive approach to patients,¹³ which is a key factor for successful diabetes care.¹⁴ As large numbers of patients with type 2 diabetes are treated in family practice, it is important that target-specific monitoring fit into the overall primary care function of family practice and that it answer the needs, demands, and expectations of patients.

Since 1985, the Nijmegen University Department of Family Practice has been developing a computer-assisted practice network, the Nijmegen Academic Research Network CMR/NMP, to study chronic diseases.^15,16 The objectives of this network are to support care for patients with chronic diseases and to create an optimal setting for clinical research under family practice conditions. This paper analyzes the outcome of diabetes care in the CMR/NMP 7 years after the introduction of an audit-enhanced monitoring system (AEMS).¹⁷

The aims were to assess (1) the outcome of care compared with external guideline criteria and the results of clinical trials, and (2) the relationship of outcome to process of care measures and to patient-related and practice-related factors.

Methods

Study population

Data were collected at the 10 family practices in the CMR/NMP, with 25 family physicians and a patient list of approximately 46,000 in 1999.¹⁶ All patients meeting World Health Organization criteria for the diagnosis of type 2 diabetes mellitus and under treatment by a family physician in 1993 and 1999 were included in the AEMS.^15,18 Patients who were treated with insulin within 1 year of diagnosis and who continued to take it were considered to have type 1 diabetes mellitus. All other patients were regarded as type 2, regardless of current treatment. For this study we included all type 2 diabetes patients under treatment by their family physician in 1993 and 1999. Patients who had died or who had moved to another area or been admitted to a residential nursing home before the end of the year were excluded, as were those who had been newly diagnosed during the year.

Audit-enhanced monitoring system

Since 1989, data have been collected on all type 2 diabetes patients at the time of diagnosis and during all regular (quarterly) diabetes-related outpatient visits. In 1992, a structured annual review, based on guidelines from the Dutch College of Family Physicians,¹⁹ was added. Starting in 1992, monitoring has consisted of the assessment of (1) compliance with 3 monthly control visits and an annual review visit; (2) glycemic control (ie, fasting blood glucose and Hb A_1c); (3) diabetes-related complications (ie, retinopathy, creatinine clearance, and foot problems); (4) cardiovascular risk factors (ie, smoking behavior, blood pressure, and lipid profile); (5) cardiovascular morbidity (ie, myocardial infarction, angina pectoris, heart failure, peripheral vascular disease, transient ischemic attack, or cerebrovascular accident). In addition, all reasons for dropping out, including cause of death, were recorded. Morbidity and causes of death were defined as in the International Classification of Health Problems in Primary Care.

To facilitate data collection, a computerized Research Registration System (RRS) was developed. The system was integrated into a standard Dutch electronic record system for family practice (Promedico, Euroned). The RRS generates templates for recording data at the quarterly or annual diabetes control visits into the patient’s electronic record. Templates guide the delivery of care and a reminder system is integrated into the RRS. Office assistants contact patients who do not come in for visits at regular intervals, both those (< 1%) who usually do not come in and those who are supposed to but fail to do so.

Family physicians sent the RRS data files to the University Department of Family Practice, where they were processed into a feedback report on process of care and outcome of care measures on 3 levels: (1) total study population; (2) practice population; (3) individual patient. Process and outcome measures were compared with external criteria based on guidelines from the Dutch College of Family Medicine and with average performance at the other practices. Feedback items were selected in consultation with the participating physicians. In this way, feedback corresponded with daily practice needs. During the project, the feedback was gradually extended from process to outcome measures. The feedback was standard to all practices.

Feedback was discussed at University Department of Family Medicine meetings, which maintained uniform registration and safeguarded the progress of the project. The feedback was also sent to every practice and participating GP. This report contained practice-level as well as physician-level data. The Figure demonstrates one way in which data are presented at the meetings and shows the percentage of patients who attended their annual diabetes control visit in the year studied.

Targets for care

Targets for care consisted of 2 elements: process and outcome measures. The key marker for process of care was compliance to the annual diabetes control visit. Key markers for desired outcome of care were (1) Hb A_1c < 8.5%,¹⁹ (2) blood pressure less than 160/90 mm Hg (revised to 150/85 mm Hg in 1999),^8,19 and (3) lipids in accordance with Dutch guidelines for general practice⁸: (a) cholesterol < 5 mmol (192 mg/dL) for patients with cardiovascular morbidity; (b) cholesterol/HDL ratio < 5.0 in smokers without cardiovascular morbidity; and (c) cholesterol/HDL < 6.0 in nonsmokers without cardiovascular morbidity. These guidelines for lipid-lowering therapy are based on sex, a life expectancy of at least 5 years, smoking status, presence of cardiovascular morbidity, total cholesterol levels, high-density lipoprotein (HDL) cholesterol levels, and triglyceride levels. If even 1 of these variables is absent, the potential value of lipid lowering cannot be determined.⁸

Analysis

Cross-sectional analysis was performed on the outcome of diabetes care in patients with type 2 diabetes who were treated by their family physicians in 1993 and 1999. The comparison was based on all patients who had been treated for the full calendar year in 1993 and 1999; therefore, it was based on a dynamic population. Process and outcome measures are compared using the chi-squared, unpaired t, or Mann–Whitney test, as appropriate. Results are expressed as means plus or minus standard deviations or as proportions. Multilevel analysis was performed to assess factors that contributed to the variance in compliance with the annual review and the desired glycemic level (Hb A_1c < 8.5%).

Results

In 1993, 540 type 2 diabetes patients (prevalence 1.3%) were included in the AEMS. Of these, 51 had been newly diagnosed (incidence 1.2/1000); 37 had been treated by a specialist (7%); and 20 did not participate (4%). Excluding the 108 patients in the latter 3 categories left a total of 432 patients for analysis. In 1999, 851 patients were included (prevalence 1.9%). Of these, 138 had been newly diagnosed (incidence 3.0/1000); 88 had been treated by a specialist (10%); and 31 did not participate (4%). Excluding the 257 patients in those 3 categories left 594 for analysis. Table 1 shows the baseline characteristics of patients included in the analysis.

Annual review was attended by 73% of patients in 1993 and 84% of patients in 1999 (Table 2). Increased compliance was achieved at all the practices, although differences between practices remained in 1999 (Figure). Univariate analysis showed that compliance with the annual review in 1999 was related to the practice (P = .001) but not to patient factors such as sex, age, duration of diabetes, therapy regimen, or cardiovascular morbidity, even after adjusting for blood glucose levels. Patients who did not attend their annual diabetes control visit had statistically significantly higher fasting blood glucose levels than patients who did comply (8.9 mmol/L [160 mg/dL] vs 8.2 mmol [147 mg/dL], P = .03). In 1993, 59% of patients had visited an ophthalmologist in the previous 2 years versus 80% in 1999.

In 1993, Hb A_1c was measured in 51% of patients with a mean of 8.2%. In 1999, compliance in measurement of Hb A_1c improved to 82%, with a mean Hb A_1c level of 7.1% (P = .0001, Table 3). The percentages of patients with an Hb A_1c level of more than 8.5% decreased from 41% to 13% (P = .001). These outcomes were associated with changes in treatment (P = .001): a decrease in patients treated with diet only (22% in 1993 vs 13% in 1999) and with oral hypoglycemic monotherapy (45% in 1993 vs 37% in 1999); an increase in patients treated with combination therapy using 2 or more oral hypoglycemic agents (22% in 1993 vs 31% in 1999); and an increase in insulin therapy (11% in 1993 vs 19% in 1999). Univariate analysis showed that poor glycemic control (Hb A_1c > 8.5%) in 1999 was related to the therapy regimen (P = .001) but not to sex, age, duration of diabetes, cardiovascular morbidity, or practice. The glycemic control in patients treated with combination therapy or insulin was poorer than in patients treated with diet only or oral hypoglycemic monotherapy, probably reflecting the fact that patients with less severe disease are managed with single agents and diet.

Compliance with measurement of blood pressure improved from 72% to 83% during the study period (Table 3). However, the percentage of patients with a systolic blood pressure below 150 mm Hg or a diastolic blood pressure below 85 mm Hg did not change between 1993 and 1999 whether patients were hypertensive or not. In hypertensive patients with type 2 diabetes, the mean diastolic blood pressure decreased from 88 mm Hg to 85 mm Hg (P = .004), but mean systolic blood pressure did not change.

The mean cholesterol level was lower in 1999 than in 1993 (6.2 vs 5.4 mmol/L; 238 mg/dL vs 207 mg/dL, P = .0001), as was the mean triglyceride level (2.54 mmol/L vs 2.07 mmol/L; 221 mg/dL vs 180 mg/dL, P = .0003). In both years, data regarding which patients could be considered for lipid-lowering therapy were available for 63% and 82%, respectively. In 1993, a far higher proportion of patients had failed to reach lipid target levels than was the case in 1999 (48% vs 28%, respectively, P = .001).

Multilevel analysis showed that paying an annual diabetes control visit (a process outcome) was related to the practice (intraclass correlation coefficient [ICC] = 0.29) but not to patient factors. Reaching the glycemic target level of Hb A_1c < 8.5%, however, was not related to practice factors (ICC = 0.003).

TABLE 1
Chacteristics of type 2 diabetes patients under family physician care in 1993 and 1999

TABLE 2
Process of care for type 2 diabetes patients under family physician care, 1993 vs 1999

TABLE 3
Outcomes of care for type 2 diabetes patients under family physician care, 1993 vs 1999

FIGURE
Percentage of patients with annual review (target = 75%) in 1999 (n=594)

Discussion

During 7 years of structured audit-enhanced monitoring of patients with type 2 diabetes in an academic family practice research network, the intermediate measures of diabetes care improved. In particular, the mean Hb A1c of 7.1% can be seen as a measure of good quality of care. The number of patients treated according to Dutch family practice guidelines (a process of care outcome) also increased.⁸

While our data were collected during normal daily care (effectiveness), the findings come close to the outcome of care under ideal trial conditions (efficacy).²¹ In the UK Prospective Diabetes Study (UKPDS), the median Hb A_1c level for all newly diagnosed patients in the group with intensive blood glucose control over 10 years reached a comparable level of 7.0%.¹ Thus, the outcome of our study approaches that achieved under trial conditions. When we analyzed patients without outcome data as poorly controlled (worst-case scenario), Hb A_1c was less than 8.5% in 28%.

The trend of improvement in glycemic control could have been a result of improved overall diabetes care in the Netherlands during the study period. Data about the outcome of diabetes care in the family medicine setting in the Netherlands during the study period are scarce and, when available, are derived from other research networks. In these networks a mean Hb A_1c of 7.0% to 7.6% was reached.²² Yet indicators from other studies suggest that our results were far better than outcome from usual care. Recently published data on such outcomes in family medicine in the Netherlands showed that Hb A_1c, blood pressure, and lipids were measured in less than 30% of patients.^23,24 Outcomes from usual care as reported in research studies appear to be strongly biased by selection and probably cannot serve as a valid reference value.

The disappointing effect on the percentage of patients who reached the target blood pressure could have resulted from evaluating the data prematurely. When the study began, the primary objective was to improve glycemic control. Shortly after the publication of the Scandinavian Simvastatin Survival Study (4S)³ and the UKPDS,¹² the guidelines of the Dutch College of Family Physicians were changed⁸ and more attention was paid to blood pressure and lipid control. This new approach was discussed with the participating family physicians. Consequently, the target for blood pressure was revised from 160/90 mm Hg to 150/85 mm Hg and lipid-lowering therapy was tailored to each patient’s cardiovascular risk profile. The 1999 outcome with respect to blood pressure and lipid control was measured only 1 year after these changes had been announced. Nevertheless, mean diastolic blood pressure in hypertensive patients and total cholesterol and triglyceride levels decreased significantly, and more patients reached target levels for lipids in 1999 than in 1993.

Our outcome was reached through enhanced compliance to guidelines. Therefore, the outcome in 1999 was based on a larger percentage of available patients. Because the AEMS studied a dynamic group of patients, the study groups in 1993 and 1999 were not identical. Theoretically, improvement in outcome could have been reached by including more easily manageable patients. However, no patient factors such as sex, age, duration of diabetes, treatment modality, or cardiovascular morbidity were related to compliance with annual review. The higher fasting blood glucose levels in patients who were noncompliant with annual review probably reflected under-treatment rather than more severe illness status. Therefore, we are confident that the findings reflect improved overall diabetes care.

The data on process measures in this study compare favorably with those of multipractice audits of diabetes care in the United Kingdom.^25-29 The high prevalence rate of 2.0% (exclusively patients with type 2 diabetes) supports the validity of our data.^25-27 Among our patients, 96% had been seen at least once during the previous year. In the large studies by Khunti and Bennett, only 85% had been seen during the previous year.^25,26 The mean annual compliance rate of nearly 85% with Hb A_1c and blood pressure measurements in our study was high. In particular, compliance with lipid control^25,26,28 and funduscopy^26,28 was better in our study. In 2 longitudinal studies that used an organized care system in which feedback was provided to the participating family physicians, compliance rates in process measures of up to 75% were reported.^30,31

The outcomes of this study were achieved in an academic family practice research network, with specific facilities for the proactive supervision of patients with chronic diseases. These results cannot and should not be generalized to “routine” family practice. Monitoring and feedback in routine family practice are in themselves insufficient to improve the quality of care.¹⁰ Care assessment should preferably take a more comprehensive approach in which evidence-based goals for care are formulated, care is improved to reach those goals, and care is measured to see whether those goals have been achieved.¹¹ Our academic network provides this comprehensive approach.

The electronic Research Registration System played an important role in the audit-enhanced monitoring. In the pilot phase of the project, paper records were used. Although using paper records had clear disadvantages, one could expect to achieve similar results using such records in combination with a central electronic data bank.

Conclusions

Outcomes of diabetes care in our family research setting were comparable with those reported in randomized controlled trials. Therefore, it is possible for the management of diabetes in family practice to be efficacious. This finding should encourage more efforts by physicians in family practice to bridge the gap between efficacy and effectiveness.

Important differences remained in achieved process measures between the academic family practices. While the outcome of diabetes care in the network was favorable, the outcome of treatment was unsatisfactory in a substantial number of patients. Further implementation strategies must be developed. The differences in achieved process measures were probably unrelated to socioeconomic differences between the practice populations, since the practice pairs 1/3 and 7/10 served comparable communities and had different levels of compliance.

Our study demonstrated that a high quality of diabetes care in family practice can be achieved. Audit-enhanced monitoring, which will provide the greatest benefit to the most patients with type 2 diabetes mellitus, should be implemented as part of a quality improvement system.

ACKNOWLEDGMENTS

The authors wish to thank the family physicians and practice nurses for their continuing support and data collection.