Original Research

Acute community-acquired lower respiratory tract infections (LRTIs) in adults include acute bronchitis, pneumonia, and infectious episodes in patients with asthma or chronic obstructive pulmonary disease (COPD). In acute bronchitis and exacerbations of COPD, the value of antibiotic therapy is doubtful; in pneumonia, however, it is widely accepted. Because distinguishing between these disease entities on clinical grounds alone is often impossible, deciding which patients would benefit from antibiotic treatment remains difficult.^1-6In the Netherlands, as in the United States and Great Britain, antibiotics are prescribed for patients with acute bronchitis approximately 80% of the time.^7-9

If a primary care physician (PCP) decides to treat LRTI with antibiotics, amoxicillin is the drug of first choice in the Netherlands.^10-13 However, amoxicillin is not effective in infections caused by atypical organisms such as Mycoplasma pneumoniae, Chlamydia pneumoniae, and Legionella pneumophila, which are responsible for 1% to 50% of cases of LRTI.^14-20 Roxithromycin and the newer macrolide antibiotics are recommended as drugs of choice for the empirical treatment of community-acquired pneumonia in low-risk patients in the United States and Canada^21-23 because those drugs cover both typical and atypical pathogens. Amoxicillin has long proved to be a reliable drug and one to which the resistance of common respiratory tract pathogens (Streptococcus pneumoniae and Haemophilus influenzae) in the Netherlands is low.^24-29

Community-based studies that evaluate treatment for LRTI are lacking. Also lacking are independent randomized controlled studies comparing amoxicillin with roxithromycin or other new macrolides for LRTI. Our double-blind randomized trial attempted to determine whether the preference for amoxicillin in the Netherlands is well founded. In the trial, patients with LRTI who in their PCP’s opinion needed antibiotic treatment were assigned to either amoxicillin or roxithromycin. We then compared the efficacy and safety of both drugs.

Methods

Eligibility criteria and baseline characteristics

Eligible study subjects were patients in the southern part of the Netherlands who presented with signs and symptoms of LRTI that their PCPs believed warranted antibiotic therapy. Table 1 lists the inclusion and exclusion criteria.

Baseline data (at day 1) were obtained to evaluate the comparability of prognostic factors between the intervention groups. The PCP performed an extensive medical history and physical examination. In addition, a sputum sample, oral washing, and nasopharyngeal swab were taken for bacteriologic examination. Venous blood samples were taken for blood chemistry, hematology, and serology (initial titers of the viral pathogens M pneumoniae and L pneumophila).

TABLE 1
CHECKLIST FOR PATIENT ELIGIBILITY

Interventions

Once the samples had been collected, patients were randomly assigned to oral treatment with either 500 mg amoxicillin 3 times daily for 10 days or 300 mg roxithromycin once daily for 10 days. A computer program using random permuted blocks of 6 prepared a randomization list for each participating center. Batches of drug packages, each provided with a unique trial code, had been sent in advance to the participating general practices. A double-dummy technique achieved blinding of patients, treating physicians, and investigators to the assigned medication. This was necessary because amoxicillin and roxithromycin have different dosing schedules (3 times a day versus once daily) and are not identical in appearance (capsule versus tablet). All capsules and tablets had identical appearance and taste. All patients received both forms of their assigned medication. Compliance with medication regimens was measured by Medical Event Monitoring Systems (MEMS), an electronic recording system that compiles the dosing history of ambulatory patients taking oral medication.³⁰

Chest X-Rays

Every patient underwent chest x-ray. The radiographs were reassessed for the presence or absence of infiltrate by a blinded independent senior radiologist. If the first and second radiologist disagreed, a third senior radiologist made a final assessment.

Follow-up

Follow-up consultations similar to the examination on day 1 took place on days 10 and 28. During treatment (days 1 through 10) and on days 21 and 27, follow-up was supplemented by a short diary in which patients recorded their symptoms and the times at which they resumed daily activities that they had abandoned or that had been impaired.

Outcomes measured

Efficacy was assessed by comparing the groups’ clinical response on day 10 (the primary outcome measure) and day 28 and their bacteriologic response on day 10. Satisfactory clinical response was defined in 4 ways: (1) decrease in symptoms of LRTI; (2) absence of symptoms of LRTI; (3) decrease in signs of LRTI; and (4) absence of signs of LRTI. All other outcomes were regarded as unsatisfactory responses.

Self-reported symptoms and time to resolution were compared between the 2 groups on days 1 through 10, 21, and 27. The percentage of patients who had abandoned daily activities or whose participation in daily activities had been impaired by illness was followed over time. Bacteriologic cure was defined as the absence of growth of a predominant bacterial pathogen (cultured at baseline) in a sputum sample taken on day 10.

We recorded patients’ compliance rates, frequency of adverse events, and acquired bacterial resistance. Compliance was defined as the number of doses taken divided by the number of doses prescribed.

Statistical analyses

The efficacy of amoxicillin and roxithromycin was evaluated using an intention-to-treat analysis. Differences were tested using a 2-sided chi-square test ( α= 0.05). Multiple logistic regression analysis was performed to analyze the effect of differences in baseline characteristics between the randomized groups. Differences in symptoms, time to resolution of symptoms, and time to resumption of abandoned and impaired daily activities were tested in life table analyses using the Gehan test. All statistical analyses were performed with Statistical Package for the Social Sciences software, version 8.0.

Results

Patient population

From January 1998 to April 1999, 25 PCPs from 15 practices recruited 196 patients aged 18 years to 89 years. Of these patients, 99 received amoxicillin and 97 received roxithromycin (Figure 1). The 2 groups’ demographic data, signs and symptoms, comorbidities, identified pathogens, and radiographic abnormalities were similar (Table 2). Multiple logistic regression analysis showed that none of the covariables altered the effects of the study medication.

TABLE 2
FINDINGS ON PRESENTATION

FIGURE 1
DISTRIBUTION OF PATIENTS FOR EFFICACY AND SAFETY ANALYSES

Clinical cure

Early Follow-Up. The rate of clinical cure, defined as the decrease in symptoms and signs at 10 days after randomization, was high and not significantly different between both groups. Using the stricter definition of clinical cure as the complete absence of symptoms and signs led to the same conclusion. Absolute cure rates using this strict definition were low (Table 3).

Physicians discontinued treatment with the study medication in 2 cases (1 amoxicillin and 1 roxithromycin) because of unsatisfactory clinical response. Both patients recovered rapidly after alternative antibiotic treatment. In one case, the patient discontinued amoxicillin after 8 days because of rash and urticaria and recovered quickly without further treatment.

Late Follow-Up. According to the physicians’ final assessments, the rate of clinical cure at 28 days was not significantly different between the 2 groups, although the percentage of patients who showed a decrease in symptoms was significantly higher in the amoxicillin group than in the roxithromycin group (Table 3). Again, cure rates were much lower when the strict definition of cure was used. Eleven patients in the amoxicillin group and 8 in the roxithromycin group were not clinically cured after 28 days. Of these patients, 10 (5 in each group) recovered shortly thereafter or did not consult their physician again for persisting symptoms of LRTI. Nine patients (6 in the amoxicillin group, 3 in the roxithromycin group) with exacerbation of COPD slowly returned to their baseline clinical situation. Four patients (3 in the amoxicillin group, 1 in the roxithromycin group) were found to have concomitant pulmonary cancer. Curative bilobectomy was performed in one of the patients. The others received palliative treatment.

TABLE 3
CLINICAL CURE RATE AT EARLY (10-DAY) AND LATE (28-DAY) FOLLOW-UP

Self-reported response over time

The time before resolution of symptoms according to the patients’ diaries was similar for patients treated with amoxicillin and those treated with roxithromycin (Figures 2A and 2B). The percentage of patients who had abandoned daily activities was followed over time. At baseline, more than half of the patients in the amoxicillin group and fewer than 40% in the roxithromycin group reported that they had abandoned daily activities. At day 10, this percentage had fallen to less than 20% in both groups and to less than 10% in both groups at day 28. Differences between the amoxicillin and roxithromycin groups were not significant.

Furthermore, the patients’ diaries revealed information about the time of impaired daily activities. The percentage of patients with impaired daily activities gradually decreased in both treatment groups from approximately 75% at baseline to 30% at day 10 and 20% at day 28.

FIGURE 2
TIME TO RESOLUTION OF SYMPTOMS AS DESCRIBED IN PATIENTS’ DIARIES

Subgroup analyses

The above analyses were repeated for a group of patients aged less than 65 years and a group aged 65 years and older. The trend in cure rates was the same. No differences were found between these age groups regarding the percentage of patients with satisfactory clinical response. Furthermore, the same analyses were performed for each of the clinical diagnoses made by the PCPs at baseline (ie, pneumonia, acute bronchitis, exacerbation of asthma or COPD, and unclassified LRTI). Overall, no significant differences were found between the amoxicillin and roxithromycin groups.

Bacteriologic evaluation

Pathogens were identified in 91 patients (46%). Viruses were most frequent, followed by H (Para) influenzae, S pneumoniae, and Moraxella catarrhalis (Table 4). Bacteriologic cure was achieved in 21 of the 23 patients (91%) in the amoxicillin group and in 23 of the 27 patients (85%) in the roxithromycin group (NS, Fisher’s exact test). In 9 patients of the amoxicillin group and 8 patients of the roxithromycin group, only the sample obtained after 10 days showed the growth of a predominant bacterial pathogen (superinfection).

TABLE 4
RESPIRATORY TRACT PATHOGENS ISOLATED

Safety and compliance

Thirty possible or probable adverse events were reported in 19 of 99 patients (19%) treated with amoxicillin: diarrhea (13), stomach ache (3), headache (3), and 11 other side effects, including nausea, vomiting, and rash, once each. In the roxithromycin group, 24 events were reported in 16 patients (16%): nausea (5), diarrhea (4), vomiting (4), rash (2), headache (2), and 7 others, including pruritus ani, dizziness, and mild bradycardia, once each.

Compliance with the medication regimen was high. Data from electronic monitoring were available for 160 patients (78 in the amoxicillin group, 82 in the roxithromycin group). The overall compliance rate for patients in both groups (ie, the number of doses taken divided by the number of doses prescribed) was 98%. In the amoxicillin group, the numbers of patients with less than 90% compliance in taking the tablets and capsules were 7 and 4, respectively. In the roxithromycin group, compliance in taking the tablets was at least 90% in all patients but compliance in taking the capsules was less than 90% in 6 patients.

Discussion

This community-based study shows that amoxicillin and roxithromycin are equally effective in the treatment of LRTI in the Netherlands. Clinical cure rates after 10 days of antibiotic treatment were approximately 90% in both study groups, although complete absence of symptoms was achieved in only a minority of cases. After 28 days of follow-up, cure rates remained high. The amoxicillin group had a significantly higher cure rate than the roxithromycin group as evidenced by the decrease in symptoms. However, this significant difference in favor of the amoxicillin group did not alter the PCPs’ overall conclusion after complete follow-up: that 90% of patients who received either drug had been effectively treated. Patients’ diary entries agreed with that impression.

The time to resolution of symptoms, the cumulative cure rate per day, and the influence of the illness on daily activities were not significantly different between patients treated with amoxicillin versus those given roxithromycin. Adverse events were mild and were divided evenly over both groups with the exception of diarrhea, which occurred more often in those taking amoxicillin.

In our study, complete absence of symptoms and signs after 28 days, as assessed by both physicians and patients, was achieved in only approximately half the patients. Complete remission of LRTI often takes more than 4 weeks.

Although LRTI is often managed in primary care, diagnostic and therapeutic decisions are usually based on the experiences of hospital-based specialists and on the results of trials conducted in hospital settings. Generalizing these results to primary care is of limited value, since disease in patients recruited for these studies is often at a later stage and more serious. In our trial, patients were recruited, diagnosed, and treated by PCPs in their natural setting, maintaining regular care as much as possible.

Nevertheless, generalization of our findings to everyday care may not be valid. To explore the degree of selection in our recruited patients, we compared the actual numbers of cases of LRTI in 3 practices (with a total of 9 PCPs and a total population of 13,269) with the numbers included in the present trial during 1 year of the inclusion period. Of the 463 presumably eligible patients, only 43 (9%) were actually included. This proportion is similar to that in a recent study of randomized controlled trials in primary care in which less than 10% of the eligible population were recruited for the trial.³¹ Included patients did not differ from other eligible patients with regard to age, clinical diagnosis, severity of illness, and need for antibiotic treatment (according to the PCPs).

Clinical studies, mostly in inpatient settings, on community-acquired pneumonia have identified causative pathogens in 50% to 69% of patients.^{14-17,21,23,32,33} Outpatient studies of acute bronchitis and LRTI have generally reported considerably lower percentages (16% to 44%).^19,20,34-36 In our study, pathogens that presumably caused LRTI were found in 46% of patients.

Because atypical pathogens were the presumptive causative agent in only 3 cases (2 M pneumoniae, 1 L pneumophila), the potential advantage of macrolide antibiotics over amoxicillin is minimal. Furthermore, bacterial resistance to macrolide antibiotics is believed to be considerable.^37,38 In Finland, bacterial resistance to erythromycin has been shown to rise quickly after an increase in the consumption of macrolide antibiotics.³⁹ In contrast to alarming reports in the literature,^{14,17,22,40,41} the low incidence of M pneumoniae and L pneumophila found in the current study supports the conservative approach (ie, amoxicillin or doxycycline) to treating community-acquired LRTI in the Netherlands.

M pneumoniae occurs at high rates in 4-year to 5-year cycles.⁴² This timing implies that the frequency of M pneumoniae might be higher if the same study were performed 1 year later. Because most M pneumoniae infections are self-limiting and clinical cure rates of macrolide antibiotics compared with those of placebo are the same,^43,44 however, this epidemiologic observation does not change the conclusions of the present study.

Compliance with medication was reliably measured and quantified by Medical Event Monitoring Systems. For both ethical and practical reasons, patients were informed about the monitoring mechanism. Their knowledge about the monitoring may have slightly increased compliance as compared with daily practice, although this assumption has not been confirmed in other studies.^45,46 Furthermore, compliance with antibiotic regimens is known to be greater than compliance with chronic medication regimens.^47,48

Conclusions

General practitioners frequently diagnose LRTI in general or pneumonia and acute bronchitis in particular, including infectious episodes in patients with asthma or COPD. In many cases, treatment with antibiotics follows. The results of our randomized controlled trial did not confirm the potentially greater value of roxithromycin, which is often recommended as the drug of choice for empirical treatment of community-acquired pneumonia, over amoxicillin. Because amoxicillin was as effective as roxithromycin, it remains a reliable first-choice antibiotic in the treatment of community-acquired LRTI.

Acknowledgments

The authors wish to thank the patients, general practitioners, and physicians’ assistants who participated in this study. They also thank Hans Verloop, director of the Vandra paper factory, Meer-Hoogstraten, Belgium, for donating cardboard boxes and Alexander Thissen, Josephine Asberg, and Ramon Ottenheijm for their assistance with the logistics of the study. The study was supported by a grant from the Research Institute for Extramural and Transmural Health Care, Maastricht.

Acute community-acquired lower respiratory tract infections (LRTIs) in adults include acute bronchitis, pneumonia, and infectious episodes in patients with asthma or chronic obstructive pulmonary disease (COPD). In acute bronchitis and exacerbations of COPD, the value of antibiotic therapy is doubtful; in pneumonia, however, it is widely accepted. Because distinguishing between these disease entities on clinical grounds alone is often impossible, deciding which patients would benefit from antibiotic treatment remains difficult.^1-6In the Netherlands, as in the United States and Great Britain, antibiotics are prescribed for patients with acute bronchitis approximately 80% of the time.^7-9

If a primary care physician (PCP) decides to treat LRTI with antibiotics, amoxicillin is the drug of first choice in the Netherlands.^10-13 However, amoxicillin is not effective in infections caused by atypical organisms such as Mycoplasma pneumoniae, Chlamydia pneumoniae, and Legionella pneumophila, which are responsible for 1% to 50% of cases of LRTI.^14-20 Roxithromycin and the newer macrolide antibiotics are recommended as drugs of choice for the empirical treatment of community-acquired pneumonia in low-risk patients in the United States and Canada^21-23 because those drugs cover both typical and atypical pathogens. Amoxicillin has long proved to be a reliable drug and one to which the resistance of common respiratory tract pathogens (Streptococcus pneumoniae and Haemophilus influenzae) in the Netherlands is low.^24-29

Community-based studies that evaluate treatment for LRTI are lacking. Also lacking are independent randomized controlled studies comparing amoxicillin with roxithromycin or other new macrolides for LRTI. Our double-blind randomized trial attempted to determine whether the preference for amoxicillin in the Netherlands is well founded. In the trial, patients with LRTI who in their PCP’s opinion needed antibiotic treatment were assigned to either amoxicillin or roxithromycin. We then compared the efficacy and safety of both drugs.

Methods

Eligibility criteria and baseline characteristics

Eligible study subjects were patients in the southern part of the Netherlands who presented with signs and symptoms of LRTI that their PCPs believed warranted antibiotic therapy. Table 1 lists the inclusion and exclusion criteria.

Baseline data (at day 1) were obtained to evaluate the comparability of prognostic factors between the intervention groups. The PCP performed an extensive medical history and physical examination. In addition, a sputum sample, oral washing, and nasopharyngeal swab were taken for bacteriologic examination. Venous blood samples were taken for blood chemistry, hematology, and serology (initial titers of the viral pathogens M pneumoniae and L pneumophila).

TABLE 1
CHECKLIST FOR PATIENT ELIGIBILITY

Interventions

Once the samples had been collected, patients were randomly assigned to oral treatment with either 500 mg amoxicillin 3 times daily for 10 days or 300 mg roxithromycin once daily for 10 days. A computer program using random permuted blocks of 6 prepared a randomization list for each participating center. Batches of drug packages, each provided with a unique trial code, had been sent in advance to the participating general practices. A double-dummy technique achieved blinding of patients, treating physicians, and investigators to the assigned medication. This was necessary because amoxicillin and roxithromycin have different dosing schedules (3 times a day versus once daily) and are not identical in appearance (capsule versus tablet). All capsules and tablets had identical appearance and taste. All patients received both forms of their assigned medication. Compliance with medication regimens was measured by Medical Event Monitoring Systems (MEMS), an electronic recording system that compiles the dosing history of ambulatory patients taking oral medication.³⁰

Chest X-Rays

Every patient underwent chest x-ray. The radiographs were reassessed for the presence or absence of infiltrate by a blinded independent senior radiologist. If the first and second radiologist disagreed, a third senior radiologist made a final assessment.

Follow-up

Follow-up consultations similar to the examination on day 1 took place on days 10 and 28. During treatment (days 1 through 10) and on days 21 and 27, follow-up was supplemented by a short diary in which patients recorded their symptoms and the times at which they resumed daily activities that they had abandoned or that had been impaired.

Outcomes measured

Efficacy was assessed by comparing the groups’ clinical response on day 10 (the primary outcome measure) and day 28 and their bacteriologic response on day 10. Satisfactory clinical response was defined in 4 ways: (1) decrease in symptoms of LRTI; (2) absence of symptoms of LRTI; (3) decrease in signs of LRTI; and (4) absence of signs of LRTI. All other outcomes were regarded as unsatisfactory responses.

Self-reported symptoms and time to resolution were compared between the 2 groups on days 1 through 10, 21, and 27. The percentage of patients who had abandoned daily activities or whose participation in daily activities had been impaired by illness was followed over time. Bacteriologic cure was defined as the absence of growth of a predominant bacterial pathogen (cultured at baseline) in a sputum sample taken on day 10.

We recorded patients’ compliance rates, frequency of adverse events, and acquired bacterial resistance. Compliance was defined as the number of doses taken divided by the number of doses prescribed.

Statistical analyses

The efficacy of amoxicillin and roxithromycin was evaluated using an intention-to-treat analysis. Differences were tested using a 2-sided chi-square test ( α= 0.05). Multiple logistic regression analysis was performed to analyze the effect of differences in baseline characteristics between the randomized groups. Differences in symptoms, time to resolution of symptoms, and time to resumption of abandoned and impaired daily activities were tested in life table analyses using the Gehan test. All statistical analyses were performed with Statistical Package for the Social Sciences software, version 8.0.

Results

Patient population

From January 1998 to April 1999, 25 PCPs from 15 practices recruited 196 patients aged 18 years to 89 years. Of these patients, 99 received amoxicillin and 97 received roxithromycin (Figure 1). The 2 groups’ demographic data, signs and symptoms, comorbidities, identified pathogens, and radiographic abnormalities were similar (Table 2). Multiple logistic regression analysis showed that none of the covariables altered the effects of the study medication.

TABLE 2
FINDINGS ON PRESENTATION

FIGURE 1
DISTRIBUTION OF PATIENTS FOR EFFICACY AND SAFETY ANALYSES

Clinical cure

Early Follow-Up. The rate of clinical cure, defined as the decrease in symptoms and signs at 10 days after randomization, was high and not significantly different between both groups. Using the stricter definition of clinical cure as the complete absence of symptoms and signs led to the same conclusion. Absolute cure rates using this strict definition were low (Table 3).

Physicians discontinued treatment with the study medication in 2 cases (1 amoxicillin and 1 roxithromycin) because of unsatisfactory clinical response. Both patients recovered rapidly after alternative antibiotic treatment. In one case, the patient discontinued amoxicillin after 8 days because of rash and urticaria and recovered quickly without further treatment.

Late Follow-Up. According to the physicians’ final assessments, the rate of clinical cure at 28 days was not significantly different between the 2 groups, although the percentage of patients who showed a decrease in symptoms was significantly higher in the amoxicillin group than in the roxithromycin group (Table 3). Again, cure rates were much lower when the strict definition of cure was used. Eleven patients in the amoxicillin group and 8 in the roxithromycin group were not clinically cured after 28 days. Of these patients, 10 (5 in each group) recovered shortly thereafter or did not consult their physician again for persisting symptoms of LRTI. Nine patients (6 in the amoxicillin group, 3 in the roxithromycin group) with exacerbation of COPD slowly returned to their baseline clinical situation. Four patients (3 in the amoxicillin group, 1 in the roxithromycin group) were found to have concomitant pulmonary cancer. Curative bilobectomy was performed in one of the patients. The others received palliative treatment.

TABLE 3
CLINICAL CURE RATE AT EARLY (10-DAY) AND LATE (28-DAY) FOLLOW-UP

Self-reported response over time

The time before resolution of symptoms according to the patients’ diaries was similar for patients treated with amoxicillin and those treated with roxithromycin (Figures 2A and 2B). The percentage of patients who had abandoned daily activities was followed over time. At baseline, more than half of the patients in the amoxicillin group and fewer than 40% in the roxithromycin group reported that they had abandoned daily activities. At day 10, this percentage had fallen to less than 20% in both groups and to less than 10% in both groups at day 28. Differences between the amoxicillin and roxithromycin groups were not significant.

Furthermore, the patients’ diaries revealed information about the time of impaired daily activities. The percentage of patients with impaired daily activities gradually decreased in both treatment groups from approximately 75% at baseline to 30% at day 10 and 20% at day 28.

FIGURE 2
TIME TO RESOLUTION OF SYMPTOMS AS DESCRIBED IN PATIENTS’ DIARIES

Subgroup analyses

The above analyses were repeated for a group of patients aged less than 65 years and a group aged 65 years and older. The trend in cure rates was the same. No differences were found between these age groups regarding the percentage of patients with satisfactory clinical response. Furthermore, the same analyses were performed for each of the clinical diagnoses made by the PCPs at baseline (ie, pneumonia, acute bronchitis, exacerbation of asthma or COPD, and unclassified LRTI). Overall, no significant differences were found between the amoxicillin and roxithromycin groups.

Bacteriologic evaluation

Pathogens were identified in 91 patients (46%). Viruses were most frequent, followed by H (Para) influenzae, S pneumoniae, and Moraxella catarrhalis (Table 4). Bacteriologic cure was achieved in 21 of the 23 patients (91%) in the amoxicillin group and in 23 of the 27 patients (85%) in the roxithromycin group (NS, Fisher’s exact test). In 9 patients of the amoxicillin group and 8 patients of the roxithromycin group, only the sample obtained after 10 days showed the growth of a predominant bacterial pathogen (superinfection).

TABLE 4
RESPIRATORY TRACT PATHOGENS ISOLATED

Safety and compliance

Thirty possible or probable adverse events were reported in 19 of 99 patients (19%) treated with amoxicillin: diarrhea (13), stomach ache (3), headache (3), and 11 other side effects, including nausea, vomiting, and rash, once each. In the roxithromycin group, 24 events were reported in 16 patients (16%): nausea (5), diarrhea (4), vomiting (4), rash (2), headache (2), and 7 others, including pruritus ani, dizziness, and mild bradycardia, once each.

Compliance with the medication regimen was high. Data from electronic monitoring were available for 160 patients (78 in the amoxicillin group, 82 in the roxithromycin group). The overall compliance rate for patients in both groups (ie, the number of doses taken divided by the number of doses prescribed) was 98%. In the amoxicillin group, the numbers of patients with less than 90% compliance in taking the tablets and capsules were 7 and 4, respectively. In the roxithromycin group, compliance in taking the tablets was at least 90% in all patients but compliance in taking the capsules was less than 90% in 6 patients.

Discussion

This community-based study shows that amoxicillin and roxithromycin are equally effective in the treatment of LRTI in the Netherlands. Clinical cure rates after 10 days of antibiotic treatment were approximately 90% in both study groups, although complete absence of symptoms was achieved in only a minority of cases. After 28 days of follow-up, cure rates remained high. The amoxicillin group had a significantly higher cure rate than the roxithromycin group as evidenced by the decrease in symptoms. However, this significant difference in favor of the amoxicillin group did not alter the PCPs’ overall conclusion after complete follow-up: that 90% of patients who received either drug had been effectively treated. Patients’ diary entries agreed with that impression.

The time to resolution of symptoms, the cumulative cure rate per day, and the influence of the illness on daily activities were not significantly different between patients treated with amoxicillin versus those given roxithromycin. Adverse events were mild and were divided evenly over both groups with the exception of diarrhea, which occurred more often in those taking amoxicillin.

In our study, complete absence of symptoms and signs after 28 days, as assessed by both physicians and patients, was achieved in only approximately half the patients. Complete remission of LRTI often takes more than 4 weeks.

Although LRTI is often managed in primary care, diagnostic and therapeutic decisions are usually based on the experiences of hospital-based specialists and on the results of trials conducted in hospital settings. Generalizing these results to primary care is of limited value, since disease in patients recruited for these studies is often at a later stage and more serious. In our trial, patients were recruited, diagnosed, and treated by PCPs in their natural setting, maintaining regular care as much as possible.

Nevertheless, generalization of our findings to everyday care may not be valid. To explore the degree of selection in our recruited patients, we compared the actual numbers of cases of LRTI in 3 practices (with a total of 9 PCPs and a total population of 13,269) with the numbers included in the present trial during 1 year of the inclusion period. Of the 463 presumably eligible patients, only 43 (9%) were actually included. This proportion is similar to that in a recent study of randomized controlled trials in primary care in which less than 10% of the eligible population were recruited for the trial.³¹ Included patients did not differ from other eligible patients with regard to age, clinical diagnosis, severity of illness, and need for antibiotic treatment (according to the PCPs).

Clinical studies, mostly in inpatient settings, on community-acquired pneumonia have identified causative pathogens in 50% to 69% of patients.^{14-17,21,23,32,33} Outpatient studies of acute bronchitis and LRTI have generally reported considerably lower percentages (16% to 44%).^19,20,34-36 In our study, pathogens that presumably caused LRTI were found in 46% of patients.

Because atypical pathogens were the presumptive causative agent in only 3 cases (2 M pneumoniae, 1 L pneumophila), the potential advantage of macrolide antibiotics over amoxicillin is minimal. Furthermore, bacterial resistance to macrolide antibiotics is believed to be considerable.^37,38 In Finland, bacterial resistance to erythromycin has been shown to rise quickly after an increase in the consumption of macrolide antibiotics.³⁹ In contrast to alarming reports in the literature,^{14,17,22,40,41} the low incidence of M pneumoniae and L pneumophila found in the current study supports the conservative approach (ie, amoxicillin or doxycycline) to treating community-acquired LRTI in the Netherlands.

M pneumoniae occurs at high rates in 4-year to 5-year cycles.⁴² This timing implies that the frequency of M pneumoniae might be higher if the same study were performed 1 year later. Because most M pneumoniae infections are self-limiting and clinical cure rates of macrolide antibiotics compared with those of placebo are the same,^43,44 however, this epidemiologic observation does not change the conclusions of the present study.

Compliance with medication was reliably measured and quantified by Medical Event Monitoring Systems. For both ethical and practical reasons, patients were informed about the monitoring mechanism. Their knowledge about the monitoring may have slightly increased compliance as compared with daily practice, although this assumption has not been confirmed in other studies.^45,46 Furthermore, compliance with antibiotic regimens is known to be greater than compliance with chronic medication regimens.^47,48

Conclusions

General practitioners frequently diagnose LRTI in general or pneumonia and acute bronchitis in particular, including infectious episodes in patients with asthma or COPD. In many cases, treatment with antibiotics follows. The results of our randomized controlled trial did not confirm the potentially greater value of roxithromycin, which is often recommended as the drug of choice for empirical treatment of community-acquired pneumonia, over amoxicillin. Because amoxicillin was as effective as roxithromycin, it remains a reliable first-choice antibiotic in the treatment of community-acquired LRTI.

Acknowledgments

The authors wish to thank the patients, general practitioners, and physicians’ assistants who participated in this study. They also thank Hans Verloop, director of the Vandra paper factory, Meer-Hoogstraten, Belgium, for donating cardboard boxes and Alexander Thissen, Josephine Asberg, and Ramon Ottenheijm for their assistance with the logistics of the study. The study was supported by a grant from the Research Institute for Extramural and Transmural Health Care, Maastricht.

Acute community-acquired lower respiratory tract infections (LRTIs) in adults include acute bronchitis, pneumonia, and infectious episodes in patients with asthma or chronic obstructive pulmonary disease (COPD). In acute bronchitis and exacerbations of COPD, the value of antibiotic therapy is doubtful; in pneumonia, however, it is widely accepted. Because distinguishing between these disease entities on clinical grounds alone is often impossible, deciding which patients would benefit from antibiotic treatment remains difficult.^1-6In the Netherlands, as in the United States and Great Britain, antibiotics are prescribed for patients with acute bronchitis approximately 80% of the time.^7-9

If a primary care physician (PCP) decides to treat LRTI with antibiotics, amoxicillin is the drug of first choice in the Netherlands.^10-13 However, amoxicillin is not effective in infections caused by atypical organisms such as Mycoplasma pneumoniae, Chlamydia pneumoniae, and Legionella pneumophila, which are responsible for 1% to 50% of cases of LRTI.^14-20 Roxithromycin and the newer macrolide antibiotics are recommended as drugs of choice for the empirical treatment of community-acquired pneumonia in low-risk patients in the United States and Canada^21-23 because those drugs cover both typical and atypical pathogens. Amoxicillin has long proved to be a reliable drug and one to which the resistance of common respiratory tract pathogens (Streptococcus pneumoniae and Haemophilus influenzae) in the Netherlands is low.^24-29

Community-based studies that evaluate treatment for LRTI are lacking. Also lacking are independent randomized controlled studies comparing amoxicillin with roxithromycin or other new macrolides for LRTI. Our double-blind randomized trial attempted to determine whether the preference for amoxicillin in the Netherlands is well founded. In the trial, patients with LRTI who in their PCP’s opinion needed antibiotic treatment were assigned to either amoxicillin or roxithromycin. We then compared the efficacy and safety of both drugs.

Methods

Eligibility criteria and baseline characteristics

Eligible study subjects were patients in the southern part of the Netherlands who presented with signs and symptoms of LRTI that their PCPs believed warranted antibiotic therapy. Table 1 lists the inclusion and exclusion criteria.

Baseline data (at day 1) were obtained to evaluate the comparability of prognostic factors between the intervention groups. The PCP performed an extensive medical history and physical examination. In addition, a sputum sample, oral washing, and nasopharyngeal swab were taken for bacteriologic examination. Venous blood samples were taken for blood chemistry, hematology, and serology (initial titers of the viral pathogens M pneumoniae and L pneumophila).

TABLE 1
CHECKLIST FOR PATIENT ELIGIBILITY

Interventions

Once the samples had been collected, patients were randomly assigned to oral treatment with either 500 mg amoxicillin 3 times daily for 10 days or 300 mg roxithromycin once daily for 10 days. A computer program using random permuted blocks of 6 prepared a randomization list for each participating center. Batches of drug packages, each provided with a unique trial code, had been sent in advance to the participating general practices. A double-dummy technique achieved blinding of patients, treating physicians, and investigators to the assigned medication. This was necessary because amoxicillin and roxithromycin have different dosing schedules (3 times a day versus once daily) and are not identical in appearance (capsule versus tablet). All capsules and tablets had identical appearance and taste. All patients received both forms of their assigned medication. Compliance with medication regimens was measured by Medical Event Monitoring Systems (MEMS), an electronic recording system that compiles the dosing history of ambulatory patients taking oral medication.³⁰

Chest X-Rays

Every patient underwent chest x-ray. The radiographs were reassessed for the presence or absence of infiltrate by a blinded independent senior radiologist. If the first and second radiologist disagreed, a third senior radiologist made a final assessment.

Follow-up

Follow-up consultations similar to the examination on day 1 took place on days 10 and 28. During treatment (days 1 through 10) and on days 21 and 27, follow-up was supplemented by a short diary in which patients recorded their symptoms and the times at which they resumed daily activities that they had abandoned or that had been impaired.

Outcomes measured

Efficacy was assessed by comparing the groups’ clinical response on day 10 (the primary outcome measure) and day 28 and their bacteriologic response on day 10. Satisfactory clinical response was defined in 4 ways: (1) decrease in symptoms of LRTI; (2) absence of symptoms of LRTI; (3) decrease in signs of LRTI; and (4) absence of signs of LRTI. All other outcomes were regarded as unsatisfactory responses.

Self-reported symptoms and time to resolution were compared between the 2 groups on days 1 through 10, 21, and 27. The percentage of patients who had abandoned daily activities or whose participation in daily activities had been impaired by illness was followed over time. Bacteriologic cure was defined as the absence of growth of a predominant bacterial pathogen (cultured at baseline) in a sputum sample taken on day 10.

We recorded patients’ compliance rates, frequency of adverse events, and acquired bacterial resistance. Compliance was defined as the number of doses taken divided by the number of doses prescribed.

Statistical analyses

The efficacy of amoxicillin and roxithromycin was evaluated using an intention-to-treat analysis. Differences were tested using a 2-sided chi-square test ( α= 0.05). Multiple logistic regression analysis was performed to analyze the effect of differences in baseline characteristics between the randomized groups. Differences in symptoms, time to resolution of symptoms, and time to resumption of abandoned and impaired daily activities were tested in life table analyses using the Gehan test. All statistical analyses were performed with Statistical Package for the Social Sciences software, version 8.0.

Results

Patient population

From January 1998 to April 1999, 25 PCPs from 15 practices recruited 196 patients aged 18 years to 89 years. Of these patients, 99 received amoxicillin and 97 received roxithromycin (Figure 1). The 2 groups’ demographic data, signs and symptoms, comorbidities, identified pathogens, and radiographic abnormalities were similar (Table 2). Multiple logistic regression analysis showed that none of the covariables altered the effects of the study medication.

TABLE 2
FINDINGS ON PRESENTATION

FIGURE 1
DISTRIBUTION OF PATIENTS FOR EFFICACY AND SAFETY ANALYSES

Clinical cure

Early Follow-Up. The rate of clinical cure, defined as the decrease in symptoms and signs at 10 days after randomization, was high and not significantly different between both groups. Using the stricter definition of clinical cure as the complete absence of symptoms and signs led to the same conclusion. Absolute cure rates using this strict definition were low (Table 3).

Physicians discontinued treatment with the study medication in 2 cases (1 amoxicillin and 1 roxithromycin) because of unsatisfactory clinical response. Both patients recovered rapidly after alternative antibiotic treatment. In one case, the patient discontinued amoxicillin after 8 days because of rash and urticaria and recovered quickly without further treatment.

Late Follow-Up. According to the physicians’ final assessments, the rate of clinical cure at 28 days was not significantly different between the 2 groups, although the percentage of patients who showed a decrease in symptoms was significantly higher in the amoxicillin group than in the roxithromycin group (Table 3). Again, cure rates were much lower when the strict definition of cure was used. Eleven patients in the amoxicillin group and 8 in the roxithromycin group were not clinically cured after 28 days. Of these patients, 10 (5 in each group) recovered shortly thereafter or did not consult their physician again for persisting symptoms of LRTI. Nine patients (6 in the amoxicillin group, 3 in the roxithromycin group) with exacerbation of COPD slowly returned to their baseline clinical situation. Four patients (3 in the amoxicillin group, 1 in the roxithromycin group) were found to have concomitant pulmonary cancer. Curative bilobectomy was performed in one of the patients. The others received palliative treatment.

TABLE 3
CLINICAL CURE RATE AT EARLY (10-DAY) AND LATE (28-DAY) FOLLOW-UP

Self-reported response over time

The time before resolution of symptoms according to the patients’ diaries was similar for patients treated with amoxicillin and those treated with roxithromycin (Figures 2A and 2B). The percentage of patients who had abandoned daily activities was followed over time. At baseline, more than half of the patients in the amoxicillin group and fewer than 40% in the roxithromycin group reported that they had abandoned daily activities. At day 10, this percentage had fallen to less than 20% in both groups and to less than 10% in both groups at day 28. Differences between the amoxicillin and roxithromycin groups were not significant.

Furthermore, the patients’ diaries revealed information about the time of impaired daily activities. The percentage of patients with impaired daily activities gradually decreased in both treatment groups from approximately 75% at baseline to 30% at day 10 and 20% at day 28.

FIGURE 2
TIME TO RESOLUTION OF SYMPTOMS AS DESCRIBED IN PATIENTS’ DIARIES

Subgroup analyses

The above analyses were repeated for a group of patients aged less than 65 years and a group aged 65 years and older. The trend in cure rates was the same. No differences were found between these age groups regarding the percentage of patients with satisfactory clinical response. Furthermore, the same analyses were performed for each of the clinical diagnoses made by the PCPs at baseline (ie, pneumonia, acute bronchitis, exacerbation of asthma or COPD, and unclassified LRTI). Overall, no significant differences were found between the amoxicillin and roxithromycin groups.

Bacteriologic evaluation

Pathogens were identified in 91 patients (46%). Viruses were most frequent, followed by H (Para) influenzae, S pneumoniae, and Moraxella catarrhalis (Table 4). Bacteriologic cure was achieved in 21 of the 23 patients (91%) in the amoxicillin group and in 23 of the 27 patients (85%) in the roxithromycin group (NS, Fisher’s exact test). In 9 patients of the amoxicillin group and 8 patients of the roxithromycin group, only the sample obtained after 10 days showed the growth of a predominant bacterial pathogen (superinfection).

TABLE 4
RESPIRATORY TRACT PATHOGENS ISOLATED

Safety and compliance

Thirty possible or probable adverse events were reported in 19 of 99 patients (19%) treated with amoxicillin: diarrhea (13), stomach ache (3), headache (3), and 11 other side effects, including nausea, vomiting, and rash, once each. In the roxithromycin group, 24 events were reported in 16 patients (16%): nausea (5), diarrhea (4), vomiting (4), rash (2), headache (2), and 7 others, including pruritus ani, dizziness, and mild bradycardia, once each.

Compliance with the medication regimen was high. Data from electronic monitoring were available for 160 patients (78 in the amoxicillin group, 82 in the roxithromycin group). The overall compliance rate for patients in both groups (ie, the number of doses taken divided by the number of doses prescribed) was 98%. In the amoxicillin group, the numbers of patients with less than 90% compliance in taking the tablets and capsules were 7 and 4, respectively. In the roxithromycin group, compliance in taking the tablets was at least 90% in all patients but compliance in taking the capsules was less than 90% in 6 patients.

Discussion

This community-based study shows that amoxicillin and roxithromycin are equally effective in the treatment of LRTI in the Netherlands. Clinical cure rates after 10 days of antibiotic treatment were approximately 90% in both study groups, although complete absence of symptoms was achieved in only a minority of cases. After 28 days of follow-up, cure rates remained high. The amoxicillin group had a significantly higher cure rate than the roxithromycin group as evidenced by the decrease in symptoms. However, this significant difference in favor of the amoxicillin group did not alter the PCPs’ overall conclusion after complete follow-up: that 90% of patients who received either drug had been effectively treated. Patients’ diary entries agreed with that impression.

The time to resolution of symptoms, the cumulative cure rate per day, and the influence of the illness on daily activities were not significantly different between patients treated with amoxicillin versus those given roxithromycin. Adverse events were mild and were divided evenly over both groups with the exception of diarrhea, which occurred more often in those taking amoxicillin.

In our study, complete absence of symptoms and signs after 28 days, as assessed by both physicians and patients, was achieved in only approximately half the patients. Complete remission of LRTI often takes more than 4 weeks.

Although LRTI is often managed in primary care, diagnostic and therapeutic decisions are usually based on the experiences of hospital-based specialists and on the results of trials conducted in hospital settings. Generalizing these results to primary care is of limited value, since disease in patients recruited for these studies is often at a later stage and more serious. In our trial, patients were recruited, diagnosed, and treated by PCPs in their natural setting, maintaining regular care as much as possible.

Nevertheless, generalization of our findings to everyday care may not be valid. To explore the degree of selection in our recruited patients, we compared the actual numbers of cases of LRTI in 3 practices (with a total of 9 PCPs and a total population of 13,269) with the numbers included in the present trial during 1 year of the inclusion period. Of the 463 presumably eligible patients, only 43 (9%) were actually included. This proportion is similar to that in a recent study of randomized controlled trials in primary care in which less than 10% of the eligible population were recruited for the trial.³¹ Included patients did not differ from other eligible patients with regard to age, clinical diagnosis, severity of illness, and need for antibiotic treatment (according to the PCPs).

Clinical studies, mostly in inpatient settings, on community-acquired pneumonia have identified causative pathogens in 50% to 69% of patients.^{14-17,21,23,32,33} Outpatient studies of acute bronchitis and LRTI have generally reported considerably lower percentages (16% to 44%).^19,20,34-36 In our study, pathogens that presumably caused LRTI were found in 46% of patients.

Because atypical pathogens were the presumptive causative agent in only 3 cases (2 M pneumoniae, 1 L pneumophila), the potential advantage of macrolide antibiotics over amoxicillin is minimal. Furthermore, bacterial resistance to macrolide antibiotics is believed to be considerable.^37,38 In Finland, bacterial resistance to erythromycin has been shown to rise quickly after an increase in the consumption of macrolide antibiotics.³⁹ In contrast to alarming reports in the literature,^{14,17,22,40,41} the low incidence of M pneumoniae and L pneumophila found in the current study supports the conservative approach (ie, amoxicillin or doxycycline) to treating community-acquired LRTI in the Netherlands.

M pneumoniae occurs at high rates in 4-year to 5-year cycles.⁴² This timing implies that the frequency of M pneumoniae might be higher if the same study were performed 1 year later. Because most M pneumoniae infections are self-limiting and clinical cure rates of macrolide antibiotics compared with those of placebo are the same,^43,44 however, this epidemiologic observation does not change the conclusions of the present study.

Compliance with medication was reliably measured and quantified by Medical Event Monitoring Systems. For both ethical and practical reasons, patients were informed about the monitoring mechanism. Their knowledge about the monitoring may have slightly increased compliance as compared with daily practice, although this assumption has not been confirmed in other studies.^45,46 Furthermore, compliance with antibiotic regimens is known to be greater than compliance with chronic medication regimens.^47,48

Conclusions

General practitioners frequently diagnose LRTI in general or pneumonia and acute bronchitis in particular, including infectious episodes in patients with asthma or COPD. In many cases, treatment with antibiotics follows. The results of our randomized controlled trial did not confirm the potentially greater value of roxithromycin, which is often recommended as the drug of choice for empirical treatment of community-acquired pneumonia, over amoxicillin. Because amoxicillin was as effective as roxithromycin, it remains a reliable first-choice antibiotic in the treatment of community-acquired LRTI.

Acknowledgments

The authors wish to thank the patients, general practitioners, and physicians’ assistants who participated in this study. They also thank Hans Verloop, director of the Vandra paper factory, Meer-Hoogstraten, Belgium, for donating cardboard boxes and Alexander Thissen, Josephine Asberg, and Ramon Ottenheijm for their assistance with the logistics of the study. The study was supported by a grant from the Research Institute for Extramural and Transmural Health Care, Maastricht.

Antibiotics continue to be commonly used to treat the common cold^1-3 despite longstanding doubts about their efficacy^4,5 or ability to prevent complications.⁶ Upper respiratory tract infection (URTI) is the most common reason for a new consultation in family practice and the second most common reason for the prescribing of an antibiotic.⁷ Reported prescription rates for antibiotics for treating the common cold range from 17% to 60% in the United Kingdom and United States and 78% in New Zealand.^1,8 Ineffective but widespread use of antibiotics is not only a poor use of health care funds but also a cause of morbidity (from adverse effects) and the development of resistant strains of bacteria.^8-10

A promising technique to reduce antibiotic use is the delayed prescription. The only published randomized controlled trials of delayed prescription use examined its effect for the treatment of sore throat, acute childhood otitis media, and cough.^11-13 In the sore throat study, antibiotics were used by 99% of a group given antibiotics, by 13% of a group not offered any, and 31% of a group given a prescription to be taken after 3 days if symptoms persisted. The authors of the otitis media study noted a 66% reduction of antibiotic use in the delayed-prescription group, who had more symptoms, signs, and sleepless nights than the “take-now” group. In the study with acute cough, the use of antibiotics was reduced by 55% in the group with delayed prescriptions. Our study, undertaken in winter 2000, tested the use of a delayed prescription versus instructions to take antibiotics immediately in patients presenting to general practitioners with URTIs.

Methods

The 15 family physicians (FPs) who recruited patients for this study were selected primarily from a group who had reported in a previous study that they frequently gave delayed prescriptions to patients.¹⁴ Ethical approval was given by the Auckland Ethics Committee.

Inclusion and exclusion criteria

Patients of any age were eligible if they presented to their FP with a new case of the common cold and either the FP thought the patient wanted antibiotics or the patient stated that desire. For young children, the parents indicated whether or not they wanted antibiotics. FPs were provided with the diagnostic criteria for URTI from the International Classification of Health Problems in Primary Care (ICHPPC-2), which defines an URTI as including the presence of acute inflammation of the nasal or pharyngeal mucosa in the absence of other specifically defined respiratory infection.¹⁵

Patients were excluded if they had suspected streptococcal tonsillitis, sinusitis, bronchitis, or pneumonia. Also excluded were patients with lower respiratory signs, those who needed an x-ray, those with a past history of rheumatic fever, and those who had experienced a serious illness or any antibiotic treatment in the previous 2 weeks. Throat cultures were not required. Eligible patients were invited to participate and signed an informed consent form. Ideally, the offer to join the study was to be made to consecutive patients, but this did not occur in all practices.

Interventions

The intervention group was given a prescription for antibiotics with instructions to fill it after 3 days if symptoms failed to improve. The control group received a prescription with instructions to start taking the antibiotic medication immediately. General practitioners prescribed any antibiotic that they considered most appropriate. In both groups, patients were advised to return to see their doctor if symptoms worsened.

Data collection

At recruitment, the patient’s temperature was taken and the list of symptoms was recorded in duplicate. The patient was asked to take his or her temperature daily with a digital thermometer (Assess Diagnostics Medical Industries Australia Pty. Ltd., 148-152 Regent St., Redfern NSW 2016, Australia) that was provided. Patients were given symptom checklists to complete daily for 10 days after the visit. Symptoms listed were dry cough, night cough, sneezing, sore throat, pain on inspiration, pain when coughing, hoarse voice, headache, staying home from work or unable to do normal daily tasks, unwell, diarrhea, vomiting, and nausea without vomiting. Patients were instructed to record whether they had a runny nose with clear secretions (“clear runny nose”), stuffy (blocked) nose, or runny nose with dark secretions (“colored runny nose”). Patients further checked off whether they had clear sputum only in the morning, colored sputum in the morning, clear sputum all day, or colored sputum all day.

A point was allocated for each symptom. The maximum possible score was 15.¹⁶ A study assistant telephoned all participants on day 3, day 7, and day 10 to ask about their temperature and symptoms. At the end of the study, the research assistant asked participants about their level of satisfaction with the consultation, using the questions and scoring system devised by Little et al.¹¹ Although no data were collected about revisit rates, data were collected about the patient’s intention to visit a physician for the next cold.

Outcomes and analysis

The outcomes were antibiotic use, symptom scores, and the responses to satisfaction-related questions asked at the end of the study. Outcomes of intervention and control groups were compared on an intention-to-treat basis.¹⁷ Because of the repeated measures, the temperatures and summary scores of symptoms were determined with the general linear mixed model that uses Statistical Analysis System (SAS, Cary, N.C.), version 8, for Windows. Chi-square determinations and the Mantel-Haenszel odds ratio were performed for discrete variables using Statistical Package for the Social Sciences (SPSS), version 10, for Windows. When the final data point for continuous variables was missing, the last recorded value was analyzed as the current value. For discrete values, worst-case and best-case scenarios were performed. The sample size of 212 patients was based on a reduction from 60% of antibiotics consumed immediately to 40% in the delayed-prescription group (alpha 0.05, beta 0.2).

Allocation and masking

The unit of randomization was the patient. N.K., who was not a recruiter, generated the allocation schedule with Excel 97. Letters containing instructions for the intervention strategy pertaining to each patient or allocating the patient to the control group were placed in opaque envelopes and sealed. The study number was written on the outside of the envelope according to the randomization schedule. The envelopes were then given to the research assistant, who placed them in a large brown envelope with the consent forms and information sheets for recruiting family physicians. The recruiters opened each envelope immediately after recruitment of each patient.

Patients were told only that they would be given 1 of 2 sets of instruction about taking antibiotics for their colds. Participants read an information sheet and then completed a consent form. Thus, patients were blind to what the other group would take. The research assistant asked the participants not to tell her which instructions they had been given for taking antibiotics. If both types of blinding had been followed correctly, this study could be described as double blinded. However, because we cannot confirm the effectiveness of blinding the research assistant, we prefer to call this study single blinded. One copy of the allocation schedule was kept in the office of N.K.; another was kept by the departmental secretary.

Results

The Figure shows the trial profile summarizing participant flow. The baseline characteristics of the patients in both groups were similar (Table 1).

Patients in the delayed prescription group were less likely to use antibiotics (48%, 95% CI, 35%-60%) than those in the “take antibiotics now” group (89%, 95% CI, 76%-94%). The odds ratio for not using antibiotics was 0.12 (95% CI, 0.05 to 0.29) using intention-to-treat analysis. By antibiotic use, we mean that the patients consumed at least 1 dose of the antibiotic medication.

Table 2 shows the outcomes for temperature and symptom score using an intention-to-treat model. The general linear model for repeated measures found average temperature significantly higher (by 0.2°C) in the immediate antibiotic use group (P = .039) and no significant difference for the symptom score (P = .29). Reanalyzing with only collected data (without intention to treat) found no significant differences from the intention-to-treat analysis. The power to detect a difference in symptom score of 30% is 80% for an alpha of 0.05, assuming that the study gives measures of variation of the symptom score that are close to the real values. There were no significant adverse effects from taking antibiotics or not. Patients’ beliefs and intentions were not affected by the interventions (Table 3).

TABLE 1
BASELINE CHARACTERISTICS AND SYMPTOMS OF THE 2 GROUPS

TABLE 2
OUTCOMES AT BASELINE AND ON DAYS 3, 7, AND 10

TABLE 3
SATISFACTION, ATTITUDES, AND BELIEFS

FIGURE
PROGRESS OF PATIENTS THROUGH THE TRIAL

Discussion

We believe that this is the first published randomized controlled trial of delayed prescriptions for antibiotics for the common cold. Asking patients to wait for 3 days before taking their medication reduced consumption of antibiotics from 89% to 48% (P = .0001). The 41% reduction is smaller than that found in the study by Little and colleagues¹¹ of 1% in the take-now group and 69% in the delayed-prescription group. Patients in the UK study returned to the office in 3 days to pick up their prescription, whereas the New Zealand group received the prescription with instructions to wait 3 days before filling it. If the third day had occurred on a weekend, the patients would have had to seek assistance from an after-hours clinic, thereby incurring a direct patient charge.

Our study assessed only the effect of delayed prescriptions, whereas the study by Little and colleagues tested the combined effect of a delayed prescription and the barrier of having to return to the clinic to obtain the prescription. Furthermore, our approach may be more acceptable to a wider group of doctors and patients, although at the expense of a higher consumption rate.

The external validity (generalizability) of this study is difficult to assess. As with the study by Little and colleagues,¹¹ the FPs had different rates of recruitment. One investigator in the current study (B.A.) kept a list of all patients who presented to him with symptoms of the common cold. Of the 44 who were potentially eligible, 4 refused to be part of the study and 10 had other medical problems (eg, heart transplant, previous lung removal) that would have made inclusion potentially hazardous. Thus, 88% of those who had a common cold and were eligible may have participated in the study.

We do not know how many patients were excluded or refused to participate; the recruiting physicians did not supply this information as requested. There was no systematic difference in symptom scores for patients of the different recruiting doctors. As with the study by Little and colleagues, the doctors found themselves too busy to enroll patients. Such problems are always an issue in general practice research.¹⁸ Little and colleagues checked the internal validity of their telephone information; therefore, we did not repeat this. In an earlier study,¹⁴ the recruiting family physicians’ preference for using delayed prescriptions may have made them more supportive of the delayed prescription than of the immediate prescription. This issue cannot be resolved, since we needed doctors who would prescribe either a delayed prescription or an immediate prescription in order to recruit enough patients.

The strength of this study lies in the blinded nature of the intervention delivery to the patient, the analysis by intention to treat, and the study’s originality. Our intervention had no impact on patients’ satisfaction, concerns, or the likelihood of seeing a doctor for next illness (Table 3). In contrast, Little¹¹ found that antibiotic use predicted future consultations for sore throat and the belief that antibiotics were effective for sore throat.^11,19 The differences may relate to the different patient symptoms and geographical differences (common cold in New Zealand versus sore throat in the UK) or the fact that all patients in our study left with a prescription. Another possible reason is that the patients knew they were participating in a study, whereas in the Little study, the instructions were more vague.¹¹

Doctors often misinterpret patient expectations. Improving communications between patient and doctor may be central to reducing patients’ demand for antibiotics. Britten makes the claim that “all the misunderstandings were associated with lack of patients’ participation in the consultation in terms of voicing of expectations and preferences or the voicing of responses to doctors’ decisions and actions.”²⁰ The need for delayed prescriptions had been highlighted as a solution. We know that the common cold presents no great diagnostic dilemma but can produce enormous treatment dilemmas.²¹ Barry believes that by changing doctors’ views and helping patients to explain what they want from the office visit may lead to changes in treatment patterns.²² We concur with Little that unless patients are very ill, general practitioners should consider exploring their concerns, explaining the natural history of their illness, and avoiding or delaying prescribing antibiotics.¹¹

We were pleased to see a reduction of antibiotics consumed (89% to 48%). However, 48% still represents a high proportion of patients who consumed antibiotics for an illness that is most unlikely to respond to those drugs. More placebo-controlled randomized trials of antibiotics for respiratory tract infections in the primary care setting are needed. We suggest that FPs clarify patients’ expectation for antibiotics and not prescribe them unless the patient insists. For patients who expect to take antibiotics and cannot be persuaded otherwise, a delayed prescription may be the first step in educating them that these medicines are not routinely required.

Conclusions

Delayed prescriptions are a safe and effective means of reducing antibiotic use in patients with the common cold who want antibiotics. The additional barrier of asking the patient to pick up the prescription from the office if symptoms persist after 3 days may reduce antibiotic use even further. When the patient demands a prescription, delaying its delivery has the potential to provide gentle education that antibiotics are an unnecessary treatment.

Acknowledgments

The authors acknowledge the support of the Health Research Council for funding the study, Tania Milne for data collection, and Alistair Stewart for statistical advice. The authors further thank the participating family physicians and patients for their contribution.

Antibiotics continue to be commonly used to treat the common cold^1-3 despite longstanding doubts about their efficacy^4,5 or ability to prevent complications.⁶ Upper respiratory tract infection (URTI) is the most common reason for a new consultation in family practice and the second most common reason for the prescribing of an antibiotic.⁷ Reported prescription rates for antibiotics for treating the common cold range from 17% to 60% in the United Kingdom and United States and 78% in New Zealand.^1,8 Ineffective but widespread use of antibiotics is not only a poor use of health care funds but also a cause of morbidity (from adverse effects) and the development of resistant strains of bacteria.^8-10

A promising technique to reduce antibiotic use is the delayed prescription. The only published randomized controlled trials of delayed prescription use examined its effect for the treatment of sore throat, acute childhood otitis media, and cough.^11-13 In the sore throat study, antibiotics were used by 99% of a group given antibiotics, by 13% of a group not offered any, and 31% of a group given a prescription to be taken after 3 days if symptoms persisted. The authors of the otitis media study noted a 66% reduction of antibiotic use in the delayed-prescription group, who had more symptoms, signs, and sleepless nights than the “take-now” group. In the study with acute cough, the use of antibiotics was reduced by 55% in the group with delayed prescriptions. Our study, undertaken in winter 2000, tested the use of a delayed prescription versus instructions to take antibiotics immediately in patients presenting to general practitioners with URTIs.

Methods

The 15 family physicians (FPs) who recruited patients for this study were selected primarily from a group who had reported in a previous study that they frequently gave delayed prescriptions to patients.¹⁴ Ethical approval was given by the Auckland Ethics Committee.

Inclusion and exclusion criteria

Patients of any age were eligible if they presented to their FP with a new case of the common cold and either the FP thought the patient wanted antibiotics or the patient stated that desire. For young children, the parents indicated whether or not they wanted antibiotics. FPs were provided with the diagnostic criteria for URTI from the International Classification of Health Problems in Primary Care (ICHPPC-2), which defines an URTI as including the presence of acute inflammation of the nasal or pharyngeal mucosa in the absence of other specifically defined respiratory infection.¹⁵

Patients were excluded if they had suspected streptococcal tonsillitis, sinusitis, bronchitis, or pneumonia. Also excluded were patients with lower respiratory signs, those who needed an x-ray, those with a past history of rheumatic fever, and those who had experienced a serious illness or any antibiotic treatment in the previous 2 weeks. Throat cultures were not required. Eligible patients were invited to participate and signed an informed consent form. Ideally, the offer to join the study was to be made to consecutive patients, but this did not occur in all practices.

Interventions

The intervention group was given a prescription for antibiotics with instructions to fill it after 3 days if symptoms failed to improve. The control group received a prescription with instructions to start taking the antibiotic medication immediately. General practitioners prescribed any antibiotic that they considered most appropriate. In both groups, patients were advised to return to see their doctor if symptoms worsened.

Data collection

At recruitment, the patient’s temperature was taken and the list of symptoms was recorded in duplicate. The patient was asked to take his or her temperature daily with a digital thermometer (Assess Diagnostics Medical Industries Australia Pty. Ltd., 148-152 Regent St., Redfern NSW 2016, Australia) that was provided. Patients were given symptom checklists to complete daily for 10 days after the visit. Symptoms listed were dry cough, night cough, sneezing, sore throat, pain on inspiration, pain when coughing, hoarse voice, headache, staying home from work or unable to do normal daily tasks, unwell, diarrhea, vomiting, and nausea without vomiting. Patients were instructed to record whether they had a runny nose with clear secretions (“clear runny nose”), stuffy (blocked) nose, or runny nose with dark secretions (“colored runny nose”). Patients further checked off whether they had clear sputum only in the morning, colored sputum in the morning, clear sputum all day, or colored sputum all day.

A point was allocated for each symptom. The maximum possible score was 15.¹⁶ A study assistant telephoned all participants on day 3, day 7, and day 10 to ask about their temperature and symptoms. At the end of the study, the research assistant asked participants about their level of satisfaction with the consultation, using the questions and scoring system devised by Little et al.¹¹ Although no data were collected about revisit rates, data were collected about the patient’s intention to visit a physician for the next cold.

Outcomes and analysis

The outcomes were antibiotic use, symptom scores, and the responses to satisfaction-related questions asked at the end of the study. Outcomes of intervention and control groups were compared on an intention-to-treat basis.¹⁷ Because of the repeated measures, the temperatures and summary scores of symptoms were determined with the general linear mixed model that uses Statistical Analysis System (SAS, Cary, N.C.), version 8, for Windows. Chi-square determinations and the Mantel-Haenszel odds ratio were performed for discrete variables using Statistical Package for the Social Sciences (SPSS), version 10, for Windows. When the final data point for continuous variables was missing, the last recorded value was analyzed as the current value. For discrete values, worst-case and best-case scenarios were performed. The sample size of 212 patients was based on a reduction from 60% of antibiotics consumed immediately to 40% in the delayed-prescription group (alpha 0.05, beta 0.2).

Allocation and masking

The unit of randomization was the patient. N.K., who was not a recruiter, generated the allocation schedule with Excel 97. Letters containing instructions for the intervention strategy pertaining to each patient or allocating the patient to the control group were placed in opaque envelopes and sealed. The study number was written on the outside of the envelope according to the randomization schedule. The envelopes were then given to the research assistant, who placed them in a large brown envelope with the consent forms and information sheets for recruiting family physicians. The recruiters opened each envelope immediately after recruitment of each patient.

Patients were told only that they would be given 1 of 2 sets of instruction about taking antibiotics for their colds. Participants read an information sheet and then completed a consent form. Thus, patients were blind to what the other group would take. The research assistant asked the participants not to tell her which instructions they had been given for taking antibiotics. If both types of blinding had been followed correctly, this study could be described as double blinded. However, because we cannot confirm the effectiveness of blinding the research assistant, we prefer to call this study single blinded. One copy of the allocation schedule was kept in the office of N.K.; another was kept by the departmental secretary.

Results

The Figure shows the trial profile summarizing participant flow. The baseline characteristics of the patients in both groups were similar (Table 1).

Patients in the delayed prescription group were less likely to use antibiotics (48%, 95% CI, 35%-60%) than those in the “take antibiotics now” group (89%, 95% CI, 76%-94%). The odds ratio for not using antibiotics was 0.12 (95% CI, 0.05 to 0.29) using intention-to-treat analysis. By antibiotic use, we mean that the patients consumed at least 1 dose of the antibiotic medication.

Table 2 shows the outcomes for temperature and symptom score using an intention-to-treat model. The general linear model for repeated measures found average temperature significantly higher (by 0.2°C) in the immediate antibiotic use group (P = .039) and no significant difference for the symptom score (P = .29). Reanalyzing with only collected data (without intention to treat) found no significant differences from the intention-to-treat analysis. The power to detect a difference in symptom score of 30% is 80% for an alpha of 0.05, assuming that the study gives measures of variation of the symptom score that are close to the real values. There were no significant adverse effects from taking antibiotics or not. Patients’ beliefs and intentions were not affected by the interventions (Table 3).

TABLE 1
BASELINE CHARACTERISTICS AND SYMPTOMS OF THE 2 GROUPS

TABLE 2
OUTCOMES AT BASELINE AND ON DAYS 3, 7, AND 10

TABLE 3
SATISFACTION, ATTITUDES, AND BELIEFS

FIGURE
PROGRESS OF PATIENTS THROUGH THE TRIAL

Discussion

We believe that this is the first published randomized controlled trial of delayed prescriptions for antibiotics for the common cold. Asking patients to wait for 3 days before taking their medication reduced consumption of antibiotics from 89% to 48% (P = .0001). The 41% reduction is smaller than that found in the study by Little and colleagues¹¹ of 1% in the take-now group and 69% in the delayed-prescription group. Patients in the UK study returned to the office in 3 days to pick up their prescription, whereas the New Zealand group received the prescription with instructions to wait 3 days before filling it. If the third day had occurred on a weekend, the patients would have had to seek assistance from an after-hours clinic, thereby incurring a direct patient charge.

Our study assessed only the effect of delayed prescriptions, whereas the study by Little and colleagues tested the combined effect of a delayed prescription and the barrier of having to return to the clinic to obtain the prescription. Furthermore, our approach may be more acceptable to a wider group of doctors and patients, although at the expense of a higher consumption rate.

The external validity (generalizability) of this study is difficult to assess. As with the study by Little and colleagues,¹¹ the FPs had different rates of recruitment. One investigator in the current study (B.A.) kept a list of all patients who presented to him with symptoms of the common cold. Of the 44 who were potentially eligible, 4 refused to be part of the study and 10 had other medical problems (eg, heart transplant, previous lung removal) that would have made inclusion potentially hazardous. Thus, 88% of those who had a common cold and were eligible may have participated in the study.

We do not know how many patients were excluded or refused to participate; the recruiting physicians did not supply this information as requested. There was no systematic difference in symptom scores for patients of the different recruiting doctors. As with the study by Little and colleagues, the doctors found themselves too busy to enroll patients. Such problems are always an issue in general practice research.¹⁸ Little and colleagues checked the internal validity of their telephone information; therefore, we did not repeat this. In an earlier study,¹⁴ the recruiting family physicians’ preference for using delayed prescriptions may have made them more supportive of the delayed prescription than of the immediate prescription. This issue cannot be resolved, since we needed doctors who would prescribe either a delayed prescription or an immediate prescription in order to recruit enough patients.

The strength of this study lies in the blinded nature of the intervention delivery to the patient, the analysis by intention to treat, and the study’s originality. Our intervention had no impact on patients’ satisfaction, concerns, or the likelihood of seeing a doctor for next illness (Table 3). In contrast, Little¹¹ found that antibiotic use predicted future consultations for sore throat and the belief that antibiotics were effective for sore throat.^11,19 The differences may relate to the different patient symptoms and geographical differences (common cold in New Zealand versus sore throat in the UK) or the fact that all patients in our study left with a prescription. Another possible reason is that the patients knew they were participating in a study, whereas in the Little study, the instructions were more vague.¹¹

Doctors often misinterpret patient expectations. Improving communications between patient and doctor may be central to reducing patients’ demand for antibiotics. Britten makes the claim that “all the misunderstandings were associated with lack of patients’ participation in the consultation in terms of voicing of expectations and preferences or the voicing of responses to doctors’ decisions and actions.”²⁰ The need for delayed prescriptions had been highlighted as a solution. We know that the common cold presents no great diagnostic dilemma but can produce enormous treatment dilemmas.²¹ Barry believes that by changing doctors’ views and helping patients to explain what they want from the office visit may lead to changes in treatment patterns.²² We concur with Little that unless patients are very ill, general practitioners should consider exploring their concerns, explaining the natural history of their illness, and avoiding or delaying prescribing antibiotics.¹¹

We were pleased to see a reduction of antibiotics consumed (89% to 48%). However, 48% still represents a high proportion of patients who consumed antibiotics for an illness that is most unlikely to respond to those drugs. More placebo-controlled randomized trials of antibiotics for respiratory tract infections in the primary care setting are needed. We suggest that FPs clarify patients’ expectation for antibiotics and not prescribe them unless the patient insists. For patients who expect to take antibiotics and cannot be persuaded otherwise, a delayed prescription may be the first step in educating them that these medicines are not routinely required.

Conclusions

Delayed prescriptions are a safe and effective means of reducing antibiotic use in patients with the common cold who want antibiotics. The additional barrier of asking the patient to pick up the prescription from the office if symptoms persist after 3 days may reduce antibiotic use even further. When the patient demands a prescription, delaying its delivery has the potential to provide gentle education that antibiotics are an unnecessary treatment.

Acknowledgments

The authors acknowledge the support of the Health Research Council for funding the study, Tania Milne for data collection, and Alistair Stewart for statistical advice. The authors further thank the participating family physicians and patients for their contribution.

Antibiotics continue to be commonly used to treat the common cold^1-3 despite longstanding doubts about their efficacy^4,5 or ability to prevent complications.⁶ Upper respiratory tract infection (URTI) is the most common reason for a new consultation in family practice and the second most common reason for the prescribing of an antibiotic.⁷ Reported prescription rates for antibiotics for treating the common cold range from 17% to 60% in the United Kingdom and United States and 78% in New Zealand.^1,8 Ineffective but widespread use of antibiotics is not only a poor use of health care funds but also a cause of morbidity (from adverse effects) and the development of resistant strains of bacteria.^8-10

A promising technique to reduce antibiotic use is the delayed prescription. The only published randomized controlled trials of delayed prescription use examined its effect for the treatment of sore throat, acute childhood otitis media, and cough.^11-13 In the sore throat study, antibiotics were used by 99% of a group given antibiotics, by 13% of a group not offered any, and 31% of a group given a prescription to be taken after 3 days if symptoms persisted. The authors of the otitis media study noted a 66% reduction of antibiotic use in the delayed-prescription group, who had more symptoms, signs, and sleepless nights than the “take-now” group. In the study with acute cough, the use of antibiotics was reduced by 55% in the group with delayed prescriptions. Our study, undertaken in winter 2000, tested the use of a delayed prescription versus instructions to take antibiotics immediately in patients presenting to general practitioners with URTIs.

Methods

The 15 family physicians (FPs) who recruited patients for this study were selected primarily from a group who had reported in a previous study that they frequently gave delayed prescriptions to patients.¹⁴ Ethical approval was given by the Auckland Ethics Committee.

Inclusion and exclusion criteria

Patients of any age were eligible if they presented to their FP with a new case of the common cold and either the FP thought the patient wanted antibiotics or the patient stated that desire. For young children, the parents indicated whether or not they wanted antibiotics. FPs were provided with the diagnostic criteria for URTI from the International Classification of Health Problems in Primary Care (ICHPPC-2), which defines an URTI as including the presence of acute inflammation of the nasal or pharyngeal mucosa in the absence of other specifically defined respiratory infection.¹⁵

Patients were excluded if they had suspected streptococcal tonsillitis, sinusitis, bronchitis, or pneumonia. Also excluded were patients with lower respiratory signs, those who needed an x-ray, those with a past history of rheumatic fever, and those who had experienced a serious illness or any antibiotic treatment in the previous 2 weeks. Throat cultures were not required. Eligible patients were invited to participate and signed an informed consent form. Ideally, the offer to join the study was to be made to consecutive patients, but this did not occur in all practices.

Interventions

The intervention group was given a prescription for antibiotics with instructions to fill it after 3 days if symptoms failed to improve. The control group received a prescription with instructions to start taking the antibiotic medication immediately. General practitioners prescribed any antibiotic that they considered most appropriate. In both groups, patients were advised to return to see their doctor if symptoms worsened.

Data collection

At recruitment, the patient’s temperature was taken and the list of symptoms was recorded in duplicate. The patient was asked to take his or her temperature daily with a digital thermometer (Assess Diagnostics Medical Industries Australia Pty. Ltd., 148-152 Regent St., Redfern NSW 2016, Australia) that was provided. Patients were given symptom checklists to complete daily for 10 days after the visit. Symptoms listed were dry cough, night cough, sneezing, sore throat, pain on inspiration, pain when coughing, hoarse voice, headache, staying home from work or unable to do normal daily tasks, unwell, diarrhea, vomiting, and nausea without vomiting. Patients were instructed to record whether they had a runny nose with clear secretions (“clear runny nose”), stuffy (blocked) nose, or runny nose with dark secretions (“colored runny nose”). Patients further checked off whether they had clear sputum only in the morning, colored sputum in the morning, clear sputum all day, or colored sputum all day.

A point was allocated for each symptom. The maximum possible score was 15.¹⁶ A study assistant telephoned all participants on day 3, day 7, and day 10 to ask about their temperature and symptoms. At the end of the study, the research assistant asked participants about their level of satisfaction with the consultation, using the questions and scoring system devised by Little et al.¹¹ Although no data were collected about revisit rates, data were collected about the patient’s intention to visit a physician for the next cold.

Outcomes and analysis

The outcomes were antibiotic use, symptom scores, and the responses to satisfaction-related questions asked at the end of the study. Outcomes of intervention and control groups were compared on an intention-to-treat basis.¹⁷ Because of the repeated measures, the temperatures and summary scores of symptoms were determined with the general linear mixed model that uses Statistical Analysis System (SAS, Cary, N.C.), version 8, for Windows. Chi-square determinations and the Mantel-Haenszel odds ratio were performed for discrete variables using Statistical Package for the Social Sciences (SPSS), version 10, for Windows. When the final data point for continuous variables was missing, the last recorded value was analyzed as the current value. For discrete values, worst-case and best-case scenarios were performed. The sample size of 212 patients was based on a reduction from 60% of antibiotics consumed immediately to 40% in the delayed-prescription group (alpha 0.05, beta 0.2).

Allocation and masking

The unit of randomization was the patient. N.K., who was not a recruiter, generated the allocation schedule with Excel 97. Letters containing instructions for the intervention strategy pertaining to each patient or allocating the patient to the control group were placed in opaque envelopes and sealed. The study number was written on the outside of the envelope according to the randomization schedule. The envelopes were then given to the research assistant, who placed them in a large brown envelope with the consent forms and information sheets for recruiting family physicians. The recruiters opened each envelope immediately after recruitment of each patient.

Patients were told only that they would be given 1 of 2 sets of instruction about taking antibiotics for their colds. Participants read an information sheet and then completed a consent form. Thus, patients were blind to what the other group would take. The research assistant asked the participants not to tell her which instructions they had been given for taking antibiotics. If both types of blinding had been followed correctly, this study could be described as double blinded. However, because we cannot confirm the effectiveness of blinding the research assistant, we prefer to call this study single blinded. One copy of the allocation schedule was kept in the office of N.K.; another was kept by the departmental secretary.

Results

The Figure shows the trial profile summarizing participant flow. The baseline characteristics of the patients in both groups were similar (Table 1).

Patients in the delayed prescription group were less likely to use antibiotics (48%, 95% CI, 35%-60%) than those in the “take antibiotics now” group (89%, 95% CI, 76%-94%). The odds ratio for not using antibiotics was 0.12 (95% CI, 0.05 to 0.29) using intention-to-treat analysis. By antibiotic use, we mean that the patients consumed at least 1 dose of the antibiotic medication.

Table 2 shows the outcomes for temperature and symptom score using an intention-to-treat model. The general linear model for repeated measures found average temperature significantly higher (by 0.2°C) in the immediate antibiotic use group (P = .039) and no significant difference for the symptom score (P = .29). Reanalyzing with only collected data (without intention to treat) found no significant differences from the intention-to-treat analysis. The power to detect a difference in symptom score of 30% is 80% for an alpha of 0.05, assuming that the study gives measures of variation of the symptom score that are close to the real values. There were no significant adverse effects from taking antibiotics or not. Patients’ beliefs and intentions were not affected by the interventions (Table 3).

TABLE 1
BASELINE CHARACTERISTICS AND SYMPTOMS OF THE 2 GROUPS

TABLE 2
OUTCOMES AT BASELINE AND ON DAYS 3, 7, AND 10

TABLE 3
SATISFACTION, ATTITUDES, AND BELIEFS

FIGURE
PROGRESS OF PATIENTS THROUGH THE TRIAL

Discussion

We believe that this is the first published randomized controlled trial of delayed prescriptions for antibiotics for the common cold. Asking patients to wait for 3 days before taking their medication reduced consumption of antibiotics from 89% to 48% (P = .0001). The 41% reduction is smaller than that found in the study by Little and colleagues¹¹ of 1% in the take-now group and 69% in the delayed-prescription group. Patients in the UK study returned to the office in 3 days to pick up their prescription, whereas the New Zealand group received the prescription with instructions to wait 3 days before filling it. If the third day had occurred on a weekend, the patients would have had to seek assistance from an after-hours clinic, thereby incurring a direct patient charge.

Our study assessed only the effect of delayed prescriptions, whereas the study by Little and colleagues tested the combined effect of a delayed prescription and the barrier of having to return to the clinic to obtain the prescription. Furthermore, our approach may be more acceptable to a wider group of doctors and patients, although at the expense of a higher consumption rate.

The external validity (generalizability) of this study is difficult to assess. As with the study by Little and colleagues,¹¹ the FPs had different rates of recruitment. One investigator in the current study (B.A.) kept a list of all patients who presented to him with symptoms of the common cold. Of the 44 who were potentially eligible, 4 refused to be part of the study and 10 had other medical problems (eg, heart transplant, previous lung removal) that would have made inclusion potentially hazardous. Thus, 88% of those who had a common cold and were eligible may have participated in the study.

We do not know how many patients were excluded or refused to participate; the recruiting physicians did not supply this information as requested. There was no systematic difference in symptom scores for patients of the different recruiting doctors. As with the study by Little and colleagues, the doctors found themselves too busy to enroll patients. Such problems are always an issue in general practice research.¹⁸ Little and colleagues checked the internal validity of their telephone information; therefore, we did not repeat this. In an earlier study,¹⁴ the recruiting family physicians’ preference for using delayed prescriptions may have made them more supportive of the delayed prescription than of the immediate prescription. This issue cannot be resolved, since we needed doctors who would prescribe either a delayed prescription or an immediate prescription in order to recruit enough patients.

The strength of this study lies in the blinded nature of the intervention delivery to the patient, the analysis by intention to treat, and the study’s originality. Our intervention had no impact on patients’ satisfaction, concerns, or the likelihood of seeing a doctor for next illness (Table 3). In contrast, Little¹¹ found that antibiotic use predicted future consultations for sore throat and the belief that antibiotics were effective for sore throat.^11,19 The differences may relate to the different patient symptoms and geographical differences (common cold in New Zealand versus sore throat in the UK) or the fact that all patients in our study left with a prescription. Another possible reason is that the patients knew they were participating in a study, whereas in the Little study, the instructions were more vague.¹¹

Doctors often misinterpret patient expectations. Improving communications between patient and doctor may be central to reducing patients’ demand for antibiotics. Britten makes the claim that “all the misunderstandings were associated with lack of patients’ participation in the consultation in terms of voicing of expectations and preferences or the voicing of responses to doctors’ decisions and actions.”²⁰ The need for delayed prescriptions had been highlighted as a solution. We know that the common cold presents no great diagnostic dilemma but can produce enormous treatment dilemmas.²¹ Barry believes that by changing doctors’ views and helping patients to explain what they want from the office visit may lead to changes in treatment patterns.²² We concur with Little that unless patients are very ill, general practitioners should consider exploring their concerns, explaining the natural history of their illness, and avoiding or delaying prescribing antibiotics.¹¹

We were pleased to see a reduction of antibiotics consumed (89% to 48%). However, 48% still represents a high proportion of patients who consumed antibiotics for an illness that is most unlikely to respond to those drugs. More placebo-controlled randomized trials of antibiotics for respiratory tract infections in the primary care setting are needed. We suggest that FPs clarify patients’ expectation for antibiotics and not prescribe them unless the patient insists. For patients who expect to take antibiotics and cannot be persuaded otherwise, a delayed prescription may be the first step in educating them that these medicines are not routinely required.

Conclusions

Delayed prescriptions are a safe and effective means of reducing antibiotic use in patients with the common cold who want antibiotics. The additional barrier of asking the patient to pick up the prescription from the office if symptoms persist after 3 days may reduce antibiotic use even further. When the patient demands a prescription, delaying its delivery has the potential to provide gentle education that antibiotics are an unnecessary treatment.

Acknowledgments

The authors acknowledge the support of the Health Research Council for funding the study, Tania Milne for data collection, and Alistair Stewart for statistical advice. The authors further thank the participating family physicians and patients for their contribution.

Interest is growing in the use of nonpharmacologic methods for the prevention and management of hypertension. Specifically, the effect of dietary fiber on the incidence and treatment of hypertension has been explored. Epidemiologic studies show that the amount of dietary fiber ingested is inversely related to the incidence of hypertension as well as to systolic blood pressure (SBP) and diastolic blood pressure (DBP) in both hypertensive and normotensive patients.^1–5 The results obtained from clinical trials, however, are inconsistent; some report modest blood pressure reductions after increased fiber intake,^6–12 while others fail to demonstrate any effect of dietary fiber on blood pressure.^13–16 Some animal trials^17,18 and human trials^19,20 have shown a consistent lowering of blood pressure upon consumption of larger amounts of soluble fiber, suggesting that the antihypertensive effects of fiber may be caused by the soluble fraction and that these effects may be contingent upon the intake of a sufficiently large quantity.

Hypertension often occurs in association with obesity, impaired glucose tolerance, and dyslipidemia. Hyperinsulinemia and insulin resistance are thought to be key pathogenic links among these disturbances.^21–23 Studies show that soluble fiber from oats reduces both postprandial blood glucose and insulin levels.^24–27 Therefore, we conducted the following pilot trial to investigate the antihypertensive and insulin-modifying effects of oat cereal supplementation in a population of mild and borderline hyperinsulinemic and hypertensive men and women.

Methods

Study protocol

The study participants in this 6-week, randomized, parallel-group, active-controlled pilot trial were recruited by means of local community screenings and mass media advertising. The study protocol was reviewed and approved by the University of Minnesota Institutional Review Board. All participants provided informed consent before official enrollment in the study. One hundred nine men and women aged 20 to 70 were screened for eligibility with a physical exam, medical history, and chemistry and lipid profile (see Table 1 for exclusion criteria). Only generally healthy, untreated hypertensives with average SBP of 130 to 160 mm Hg and DBP of 85 to 100 mm Hg and with at least 1 reading greater than 140/90 as well as moderately elevated levels of fasting insulin (≥10μU/mL) were considered for enrollment. Participants were determined to be eligible only after 2 sets of hypertensive (SBP > 130, DBP > 85) baseline blood pressure readings had been taken 7 days apart and only if all inclusion criteria were fulfilled.

Ultimately, 22 men and women were randomized to either an oat cereal treatment group (standardized to 5.52 g/day beta-glucan) or a low fiber cereal control group (<1 g/day total fiber). Four of these individuals (1 in the treatment and 3 in the control group) discontinued participation because of time constraints. Eighteen healthy, nonsmoking men and women aged 27 to 59 years (44 ± 18; mean, SD) completed the trial. Cereal treatments were isocaloric. Participants were instructed to consume all their cereal (treatment, 137 g; control, 146 g) daily for the next 6 weeks but were allowed to prepare and consume the cereal however and whenever they wished.

Cereal compliance was determined by participant self-report in a daily cereal calendar. In addition, dietary intake was reviewed both at baseline and at the end of the 6-week intervention, using 3-day food records. Side effect data were gathered from participants at baseline and the end of the intervention. Side effects were assessed via a questionnaire consisting of 21 items relating to potential side effects from increased fiber intake (eg, loose stools, flatulence) or hypertension (eg, headaches, dizziness). Participants reported the frequency at which they experienced these side effects on a scale ranging from “never” to “very frequently” (event occurring once or more per day). Each response was assigned a numerical value. Prestudy and post study averages were used in analyses.

Blood Pressure, Plasma Lipid Concentrations, Glucose Metabolism, and Insulin Sensitivity

Blood pressure was measured weekly for each participant for the duration of the study. Each participant reported to the Hypertension and Cholesterol Research Clinic located at the University of Minnesota Medical School at approximately the same time for each blood pressure reading. All readings were obtained in the morning after participants had rested quietly, seated, for at least 5 minutes in an examination room. An examiner who was blinded to the treatment groups took readings on the right arm using a mercury column sphygmomanometer (Korotkoff phase V for DBP). Standard cuff size was used unless upper arm circumference exceeded 31 cm, in which case the examiner used a large cuff with 15 x 35-cm bladders. Measurements were repeated 4 times in 2-minute intervals. The mean of the last 3 readings was calculated and used in analyses.

To determine plasma lipid concentrations (total, high-density lipoprotein [HDL], and low-density lipoprotein [LDL] cholesterol and triglycerides), pretreatment and posttreatment blood samples were drawn. A 75-g, 3-hour oral glucose-tolerance test (OGTT) was administered before and after treatment to assess participants’ glucose tolerance and insulin response. Whole blood sampling occurred at -30, 0, 30, 60, 90, 120, 150, and 180 minutes. A measure of insulin sensitivity was assessed within 48 hours after the OGTT by means of the modified frequently sampled intravenous glucose tolerance test (FSIGT).²⁸ The glucose and insulin data derived from this test were used to calculate the insulin sensitivity index (SI) employing the minimal-model method developed by Bergman.²⁹

Statistical methods

Reported results are expressed in terms of means ± SD or means SE. Student’s t test for independent samples was used to compare the 2 treatment groups at baseline and to compare mean change scores between the 2 groups. Additionally, area-under-the-curve analyses were performed to compare OGTT insulin curves. All analyses were performed on data from an intent-to-treat population, which included all randomized participants. Statistical tests were 2 sided, performed at the 5% level of significance, and conducted with Statistical Analysis System software (SAS Institute, Cary, N.C.).

Results

No statistically significant differences in baseline characteristics occurred between the groups, although this comparison is limited by the small sample size Table 2. LDL cholesterol and total cholesterol levels and blood pressure were somewhat higher in the treatment group. The blood pressure measurements in the treatment group resulted in an average SBP of 143 ± 3.7 mm Hg before intervention and 135 ± 2.6 mm Hg after intervention (an average of the last 2 study visits, P < .01) Table 3. No significant change in SBP was observed in the control group. A significant difference between the treatment and control groups was observed for the change in SBP (P < .02). DBP dropped from 93 ± 1.9 mm Hg to 87 ± 2.2 mm Hg after the oat fiber intervention (P = .02), with no significant change in the control group (P = .94). A borderline significant trend was noted for the change scores of DBP between groups (P = .055).

Changes in fasting insulin, insulin sensitivity (S_I), and insulin curves derived from the oral glucose tolerance tests were assessed. Fasting insulin values Table 3 were taken from the OGTT (preglucose infusion values). Neither the control group (P = 1.00) nor the treatment group (P = .753) showed a significant change in fasting insulin levels. The Bergman minimal model method was used to estimate insulin sensitivity and showed no significant change in either group. Area-under-the-curve analysis of the insulin data derived from the OGTTs before and after treatment with oat cereal (Figure 1 and Figure 2) suggested a trend toward significance in terms of less insulin required to clear a glucose load (top of graphs, P = .093), with no significant changes in the control group (bottom).

Total cholesterol concentrations dropped 16.2 ± 6.3 mg/dL in the oat cereal group (P = .030), with a slight (nonsignificant) increase in the control group (P = .48). Additionally, a comparison of the changes in total cholesterol between the 2 groups revealed a significant mean difference of 21.1 ± 9.1 mg/dL (P = .035). LDL cholesterol was also reduced significantly after the oat cereal intervention by 15.8 ± 5.9 mg/dL (P = .025). The nonsignificant increase in LDL cholesterol in the control group (P = .231) combined with the significant reduction in the treatment group resulted in a significant difference between the groups after intervention (P < .015). Neither group experienced significant changes in HDL cholesterol or triglyceride concentrations.

An analysis of the side effect data showed no significant difference in the occurrence of side effects between groups. There was an overall decrease in the frequency of dietary fiber-related and hypertension-related side effects in both groups, with a more substantial reduction occurring in the oat cereal group (P = .11). Total body weight did not change significantly in either group. Additionally, both groups were very compliant (approximately 90%) in terms of cereal consumption Table 3.

Discussion

The results of this pilot study suggest that the inclusion of oats into the standard American diet of people with borderline or mild hypertension may reduce both SBP and DBP. In persons consuming 5.52 g/day of beta-glucan soluble fiber from oat cereal for 6 weeks, we found a statistically and clinically significant decrease in both SBP and DBP (7.5 mm Hg and 5.5 mm Hg, respectively) and a trend toward improved OGTT-determined insulin sensitivity. These findings warrant a large-scale clinical trial to explore further the relationship between whole-grain oat consumption and blood pressure, especially considering the limitations of this pilot study.

As with all small-scale trials, this one lacked sufficient power to detect true changes in both primary and secondary outcome variables. It is possible that regression to the mean explains at least part of the treatment effect, since participants in the oats group began the study with higher SBP, DBP, and LDL cholesterol levels than controls. In addition, it is possible that the reported blood pressure changes could have been caused by “other” undetected dietary change made by members of the oats group. Future trials will need to collect and analyze dietary data carefully; feeding trials should be considered. Such dietary analyses may indicate that certain micronutrients partially explain the hypotensive effects of whole-grain oat consumption. The DASH trial and others have consistently demonstrated that diets rich in certain micronutrients can reduce blood pressure.^30,31

Soluble fiber-rich oat cereals may affect blood pressure by modulating changes in insulin metabolism. The mechanism of action is thought to involve the slowed absorption of macronutrients from the gut, resulting in a flattening of the postprandial glycemic curve.²⁹ These lower postprandial blood glucose levels elicit a lower insulin response to accommodate its clearance from the plasma. This process may lead to improved insulin sensitivity if the lower circulating insulin levels lead eventually to upregulation of the insulin receptors in peripheral tissues. A recent animal trial demonstrated that soluble fiber feeding improved insulin sensitivity by increasing skeletal muscle plasma membrane GLUT-4 content.³² Findings in this pilot suggest that over time, oat ingestion may reduce the amount of insulin needed to clear a glucose load. However, the study was underpowered to detect significant differences in more sensitive measures of insulin resistance. The causal mechanistic relationship among whole grain oat consumption, blood pressure, and insulin resistance might be best studied using a long-term feeding study design.

Alternate mechanisms, such as attenuation in endothelial function, may have affected blood pressure responses in this study.³³ Drugs specific to endothelial cell receptors mediating vasodilation are known to lower blood pressure.³⁴ Moreover, plasma cholesterol reductions are associated with improvements in endothelium-mediated vasodilation.^35,36 In addition, preliminary evidence in animals supports a direct relationship between changes in plasma cholesterol concentrations and blood pressure.³⁷ In the present study, plasma cholesterol levels were significantly reduced in participants who ingested whole grain oat-based cereals compared to a more refined grain wheat, corn, and rice control. Thus, it is possible that the blood pressure reduction observed in the subjects consuming oats resulted in part from improved endothelial function due to a drop in plasma cholesterol. Additional research is needed to fully investigate this pathway.

From a practical standpoint, improvements in SBP and DBP such as those observed in this study would be a useful contribution to the clinical management of hypertension. The cereal feeding intervention was well tolerated. Participants were very compliant for the 6-week treatment period. Substantial improvements in blood lipids could serve as an added incentive for patients to maintain long-term compliance with feeding recommendations.^18,19 Since treatment of hypertension is a lifelong process for most patients, future studies would need to assess the effectiveness of oat cereals to maintain blood pressure benefits over a longer time. Such studies may need to consider dietary options such as soluble fiber-rich fruits in addition to cereal consumption in efforts to deliver the desired quantity of soluble fiber. Future trials will have to investigate the antihypertensive effect of whole oats in other populations, such as people with diabetes, and to study not only surrogate endpoints such as blood pressure but also patient-oriented outcomes such as mortality and morbidity.

Interest is growing in the use of nonpharmacologic methods for the prevention and management of hypertension. Specifically, the effect of dietary fiber on the incidence and treatment of hypertension has been explored. Epidemiologic studies show that the amount of dietary fiber ingested is inversely related to the incidence of hypertension as well as to systolic blood pressure (SBP) and diastolic blood pressure (DBP) in both hypertensive and normotensive patients.^1–5 The results obtained from clinical trials, however, are inconsistent; some report modest blood pressure reductions after increased fiber intake,^6–12 while others fail to demonstrate any effect of dietary fiber on blood pressure.^13–16 Some animal trials^17,18 and human trials^19,20 have shown a consistent lowering of blood pressure upon consumption of larger amounts of soluble fiber, suggesting that the antihypertensive effects of fiber may be caused by the soluble fraction and that these effects may be contingent upon the intake of a sufficiently large quantity.

Hypertension often occurs in association with obesity, impaired glucose tolerance, and dyslipidemia. Hyperinsulinemia and insulin resistance are thought to be key pathogenic links among these disturbances.^21–23 Studies show that soluble fiber from oats reduces both postprandial blood glucose and insulin levels.^24–27 Therefore, we conducted the following pilot trial to investigate the antihypertensive and insulin-modifying effects of oat cereal supplementation in a population of mild and borderline hyperinsulinemic and hypertensive men and women.

Methods

Study protocol

The study participants in this 6-week, randomized, parallel-group, active-controlled pilot trial were recruited by means of local community screenings and mass media advertising. The study protocol was reviewed and approved by the University of Minnesota Institutional Review Board. All participants provided informed consent before official enrollment in the study. One hundred nine men and women aged 20 to 70 were screened for eligibility with a physical exam, medical history, and chemistry and lipid profile (see Table 1 for exclusion criteria). Only generally healthy, untreated hypertensives with average SBP of 130 to 160 mm Hg and DBP of 85 to 100 mm Hg and with at least 1 reading greater than 140/90 as well as moderately elevated levels of fasting insulin (≥10μU/mL) were considered for enrollment. Participants were determined to be eligible only after 2 sets of hypertensive (SBP > 130, DBP > 85) baseline blood pressure readings had been taken 7 days apart and only if all inclusion criteria were fulfilled.

Ultimately, 22 men and women were randomized to either an oat cereal treatment group (standardized to 5.52 g/day beta-glucan) or a low fiber cereal control group (<1 g/day total fiber). Four of these individuals (1 in the treatment and 3 in the control group) discontinued participation because of time constraints. Eighteen healthy, nonsmoking men and women aged 27 to 59 years (44 ± 18; mean, SD) completed the trial. Cereal treatments were isocaloric. Participants were instructed to consume all their cereal (treatment, 137 g; control, 146 g) daily for the next 6 weeks but were allowed to prepare and consume the cereal however and whenever they wished.

Cereal compliance was determined by participant self-report in a daily cereal calendar. In addition, dietary intake was reviewed both at baseline and at the end of the 6-week intervention, using 3-day food records. Side effect data were gathered from participants at baseline and the end of the intervention. Side effects were assessed via a questionnaire consisting of 21 items relating to potential side effects from increased fiber intake (eg, loose stools, flatulence) or hypertension (eg, headaches, dizziness). Participants reported the frequency at which they experienced these side effects on a scale ranging from “never” to “very frequently” (event occurring once or more per day). Each response was assigned a numerical value. Prestudy and post study averages were used in analyses.

Blood Pressure, Plasma Lipid Concentrations, Glucose Metabolism, and Insulin Sensitivity

Blood pressure was measured weekly for each participant for the duration of the study. Each participant reported to the Hypertension and Cholesterol Research Clinic located at the University of Minnesota Medical School at approximately the same time for each blood pressure reading. All readings were obtained in the morning after participants had rested quietly, seated, for at least 5 minutes in an examination room. An examiner who was blinded to the treatment groups took readings on the right arm using a mercury column sphygmomanometer (Korotkoff phase V for DBP). Standard cuff size was used unless upper arm circumference exceeded 31 cm, in which case the examiner used a large cuff with 15 x 35-cm bladders. Measurements were repeated 4 times in 2-minute intervals. The mean of the last 3 readings was calculated and used in analyses.

To determine plasma lipid concentrations (total, high-density lipoprotein [HDL], and low-density lipoprotein [LDL] cholesterol and triglycerides), pretreatment and posttreatment blood samples were drawn. A 75-g, 3-hour oral glucose-tolerance test (OGTT) was administered before and after treatment to assess participants’ glucose tolerance and insulin response. Whole blood sampling occurred at -30, 0, 30, 60, 90, 120, 150, and 180 minutes. A measure of insulin sensitivity was assessed within 48 hours after the OGTT by means of the modified frequently sampled intravenous glucose tolerance test (FSIGT).²⁸ The glucose and insulin data derived from this test were used to calculate the insulin sensitivity index (SI) employing the minimal-model method developed by Bergman.²⁹

Statistical methods

Reported results are expressed in terms of means ± SD or means SE. Student’s t test for independent samples was used to compare the 2 treatment groups at baseline and to compare mean change scores between the 2 groups. Additionally, area-under-the-curve analyses were performed to compare OGTT insulin curves. All analyses were performed on data from an intent-to-treat population, which included all randomized participants. Statistical tests were 2 sided, performed at the 5% level of significance, and conducted with Statistical Analysis System software (SAS Institute, Cary, N.C.).

Results

No statistically significant differences in baseline characteristics occurred between the groups, although this comparison is limited by the small sample size Table 2. LDL cholesterol and total cholesterol levels and blood pressure were somewhat higher in the treatment group. The blood pressure measurements in the treatment group resulted in an average SBP of 143 ± 3.7 mm Hg before intervention and 135 ± 2.6 mm Hg after intervention (an average of the last 2 study visits, P < .01) Table 3. No significant change in SBP was observed in the control group. A significant difference between the treatment and control groups was observed for the change in SBP (P < .02). DBP dropped from 93 ± 1.9 mm Hg to 87 ± 2.2 mm Hg after the oat fiber intervention (P = .02), with no significant change in the control group (P = .94). A borderline significant trend was noted for the change scores of DBP between groups (P = .055).

Changes in fasting insulin, insulin sensitivity (S_I), and insulin curves derived from the oral glucose tolerance tests were assessed. Fasting insulin values Table 3 were taken from the OGTT (preglucose infusion values). Neither the control group (P = 1.00) nor the treatment group (P = .753) showed a significant change in fasting insulin levels. The Bergman minimal model method was used to estimate insulin sensitivity and showed no significant change in either group. Area-under-the-curve analysis of the insulin data derived from the OGTTs before and after treatment with oat cereal (Figure 1 and Figure 2) suggested a trend toward significance in terms of less insulin required to clear a glucose load (top of graphs, P = .093), with no significant changes in the control group (bottom).

Total cholesterol concentrations dropped 16.2 ± 6.3 mg/dL in the oat cereal group (P = .030), with a slight (nonsignificant) increase in the control group (P = .48). Additionally, a comparison of the changes in total cholesterol between the 2 groups revealed a significant mean difference of 21.1 ± 9.1 mg/dL (P = .035). LDL cholesterol was also reduced significantly after the oat cereal intervention by 15.8 ± 5.9 mg/dL (P = .025). The nonsignificant increase in LDL cholesterol in the control group (P = .231) combined with the significant reduction in the treatment group resulted in a significant difference between the groups after intervention (P < .015). Neither group experienced significant changes in HDL cholesterol or triglyceride concentrations.

An analysis of the side effect data showed no significant difference in the occurrence of side effects between groups. There was an overall decrease in the frequency of dietary fiber-related and hypertension-related side effects in both groups, with a more substantial reduction occurring in the oat cereal group (P = .11). Total body weight did not change significantly in either group. Additionally, both groups were very compliant (approximately 90%) in terms of cereal consumption Table 3.

Discussion

The results of this pilot study suggest that the inclusion of oats into the standard American diet of people with borderline or mild hypertension may reduce both SBP and DBP. In persons consuming 5.52 g/day of beta-glucan soluble fiber from oat cereal for 6 weeks, we found a statistically and clinically significant decrease in both SBP and DBP (7.5 mm Hg and 5.5 mm Hg, respectively) and a trend toward improved OGTT-determined insulin sensitivity. These findings warrant a large-scale clinical trial to explore further the relationship between whole-grain oat consumption and blood pressure, especially considering the limitations of this pilot study.

As with all small-scale trials, this one lacked sufficient power to detect true changes in both primary and secondary outcome variables. It is possible that regression to the mean explains at least part of the treatment effect, since participants in the oats group began the study with higher SBP, DBP, and LDL cholesterol levels than controls. In addition, it is possible that the reported blood pressure changes could have been caused by “other” undetected dietary change made by members of the oats group. Future trials will need to collect and analyze dietary data carefully; feeding trials should be considered. Such dietary analyses may indicate that certain micronutrients partially explain the hypotensive effects of whole-grain oat consumption. The DASH trial and others have consistently demonstrated that diets rich in certain micronutrients can reduce blood pressure.^30,31

Soluble fiber-rich oat cereals may affect blood pressure by modulating changes in insulin metabolism. The mechanism of action is thought to involve the slowed absorption of macronutrients from the gut, resulting in a flattening of the postprandial glycemic curve.²⁹ These lower postprandial blood glucose levels elicit a lower insulin response to accommodate its clearance from the plasma. This process may lead to improved insulin sensitivity if the lower circulating insulin levels lead eventually to upregulation of the insulin receptors in peripheral tissues. A recent animal trial demonstrated that soluble fiber feeding improved insulin sensitivity by increasing skeletal muscle plasma membrane GLUT-4 content.³² Findings in this pilot suggest that over time, oat ingestion may reduce the amount of insulin needed to clear a glucose load. However, the study was underpowered to detect significant differences in more sensitive measures of insulin resistance. The causal mechanistic relationship among whole grain oat consumption, blood pressure, and insulin resistance might be best studied using a long-term feeding study design.

Alternate mechanisms, such as attenuation in endothelial function, may have affected blood pressure responses in this study.³³ Drugs specific to endothelial cell receptors mediating vasodilation are known to lower blood pressure.³⁴ Moreover, plasma cholesterol reductions are associated with improvements in endothelium-mediated vasodilation.^35,36 In addition, preliminary evidence in animals supports a direct relationship between changes in plasma cholesterol concentrations and blood pressure.³⁷ In the present study, plasma cholesterol levels were significantly reduced in participants who ingested whole grain oat-based cereals compared to a more refined grain wheat, corn, and rice control. Thus, it is possible that the blood pressure reduction observed in the subjects consuming oats resulted in part from improved endothelial function due to a drop in plasma cholesterol. Additional research is needed to fully investigate this pathway.

From a practical standpoint, improvements in SBP and DBP such as those observed in this study would be a useful contribution to the clinical management of hypertension. The cereal feeding intervention was well tolerated. Participants were very compliant for the 6-week treatment period. Substantial improvements in blood lipids could serve as an added incentive for patients to maintain long-term compliance with feeding recommendations.^18,19 Since treatment of hypertension is a lifelong process for most patients, future studies would need to assess the effectiveness of oat cereals to maintain blood pressure benefits over a longer time. Such studies may need to consider dietary options such as soluble fiber-rich fruits in addition to cereal consumption in efforts to deliver the desired quantity of soluble fiber. Future trials will have to investigate the antihypertensive effect of whole oats in other populations, such as people with diabetes, and to study not only surrogate endpoints such as blood pressure but also patient-oriented outcomes such as mortality and morbidity.

Interest is growing in the use of nonpharmacologic methods for the prevention and management of hypertension. Specifically, the effect of dietary fiber on the incidence and treatment of hypertension has been explored. Epidemiologic studies show that the amount of dietary fiber ingested is inversely related to the incidence of hypertension as well as to systolic blood pressure (SBP) and diastolic blood pressure (DBP) in both hypertensive and normotensive patients.^1–5 The results obtained from clinical trials, however, are inconsistent; some report modest blood pressure reductions after increased fiber intake,^6–12 while others fail to demonstrate any effect of dietary fiber on blood pressure.^13–16 Some animal trials^17,18 and human trials^19,20 have shown a consistent lowering of blood pressure upon consumption of larger amounts of soluble fiber, suggesting that the antihypertensive effects of fiber may be caused by the soluble fraction and that these effects may be contingent upon the intake of a sufficiently large quantity.

Hypertension often occurs in association with obesity, impaired glucose tolerance, and dyslipidemia. Hyperinsulinemia and insulin resistance are thought to be key pathogenic links among these disturbances.^21–23 Studies show that soluble fiber from oats reduces both postprandial blood glucose and insulin levels.^24–27 Therefore, we conducted the following pilot trial to investigate the antihypertensive and insulin-modifying effects of oat cereal supplementation in a population of mild and borderline hyperinsulinemic and hypertensive men and women.

Methods

Study protocol

The study participants in this 6-week, randomized, parallel-group, active-controlled pilot trial were recruited by means of local community screenings and mass media advertising. The study protocol was reviewed and approved by the University of Minnesota Institutional Review Board. All participants provided informed consent before official enrollment in the study. One hundred nine men and women aged 20 to 70 were screened for eligibility with a physical exam, medical history, and chemistry and lipid profile (see Table 1 for exclusion criteria). Only generally healthy, untreated hypertensives with average SBP of 130 to 160 mm Hg and DBP of 85 to 100 mm Hg and with at least 1 reading greater than 140/90 as well as moderately elevated levels of fasting insulin (≥10μU/mL) were considered for enrollment. Participants were determined to be eligible only after 2 sets of hypertensive (SBP > 130, DBP > 85) baseline blood pressure readings had been taken 7 days apart and only if all inclusion criteria were fulfilled.

Ultimately, 22 men and women were randomized to either an oat cereal treatment group (standardized to 5.52 g/day beta-glucan) or a low fiber cereal control group (<1 g/day total fiber). Four of these individuals (1 in the treatment and 3 in the control group) discontinued participation because of time constraints. Eighteen healthy, nonsmoking men and women aged 27 to 59 years (44 ± 18; mean, SD) completed the trial. Cereal treatments were isocaloric. Participants were instructed to consume all their cereal (treatment, 137 g; control, 146 g) daily for the next 6 weeks but were allowed to prepare and consume the cereal however and whenever they wished.

Cereal compliance was determined by participant self-report in a daily cereal calendar. In addition, dietary intake was reviewed both at baseline and at the end of the 6-week intervention, using 3-day food records. Side effect data were gathered from participants at baseline and the end of the intervention. Side effects were assessed via a questionnaire consisting of 21 items relating to potential side effects from increased fiber intake (eg, loose stools, flatulence) or hypertension (eg, headaches, dizziness). Participants reported the frequency at which they experienced these side effects on a scale ranging from “never” to “very frequently” (event occurring once or more per day). Each response was assigned a numerical value. Prestudy and post study averages were used in analyses.

Blood Pressure, Plasma Lipid Concentrations, Glucose Metabolism, and Insulin Sensitivity

Blood pressure was measured weekly for each participant for the duration of the study. Each participant reported to the Hypertension and Cholesterol Research Clinic located at the University of Minnesota Medical School at approximately the same time for each blood pressure reading. All readings were obtained in the morning after participants had rested quietly, seated, for at least 5 minutes in an examination room. An examiner who was blinded to the treatment groups took readings on the right arm using a mercury column sphygmomanometer (Korotkoff phase V for DBP). Standard cuff size was used unless upper arm circumference exceeded 31 cm, in which case the examiner used a large cuff with 15 x 35-cm bladders. Measurements were repeated 4 times in 2-minute intervals. The mean of the last 3 readings was calculated and used in analyses.

To determine plasma lipid concentrations (total, high-density lipoprotein [HDL], and low-density lipoprotein [LDL] cholesterol and triglycerides), pretreatment and posttreatment blood samples were drawn. A 75-g, 3-hour oral glucose-tolerance test (OGTT) was administered before and after treatment to assess participants’ glucose tolerance and insulin response. Whole blood sampling occurred at -30, 0, 30, 60, 90, 120, 150, and 180 minutes. A measure of insulin sensitivity was assessed within 48 hours after the OGTT by means of the modified frequently sampled intravenous glucose tolerance test (FSIGT).²⁸ The glucose and insulin data derived from this test were used to calculate the insulin sensitivity index (SI) employing the minimal-model method developed by Bergman.²⁹

Statistical methods

Reported results are expressed in terms of means ± SD or means SE. Student’s t test for independent samples was used to compare the 2 treatment groups at baseline and to compare mean change scores between the 2 groups. Additionally, area-under-the-curve analyses were performed to compare OGTT insulin curves. All analyses were performed on data from an intent-to-treat population, which included all randomized participants. Statistical tests were 2 sided, performed at the 5% level of significance, and conducted with Statistical Analysis System software (SAS Institute, Cary, N.C.).

Results

No statistically significant differences in baseline characteristics occurred between the groups, although this comparison is limited by the small sample size Table 2. LDL cholesterol and total cholesterol levels and blood pressure were somewhat higher in the treatment group. The blood pressure measurements in the treatment group resulted in an average SBP of 143 ± 3.7 mm Hg before intervention and 135 ± 2.6 mm Hg after intervention (an average of the last 2 study visits, P < .01) Table 3. No significant change in SBP was observed in the control group. A significant difference between the treatment and control groups was observed for the change in SBP (P < .02). DBP dropped from 93 ± 1.9 mm Hg to 87 ± 2.2 mm Hg after the oat fiber intervention (P = .02), with no significant change in the control group (P = .94). A borderline significant trend was noted for the change scores of DBP between groups (P = .055).

Changes in fasting insulin, insulin sensitivity (S_I), and insulin curves derived from the oral glucose tolerance tests were assessed. Fasting insulin values Table 3 were taken from the OGTT (preglucose infusion values). Neither the control group (P = 1.00) nor the treatment group (P = .753) showed a significant change in fasting insulin levels. The Bergman minimal model method was used to estimate insulin sensitivity and showed no significant change in either group. Area-under-the-curve analysis of the insulin data derived from the OGTTs before and after treatment with oat cereal (Figure 1 and Figure 2) suggested a trend toward significance in terms of less insulin required to clear a glucose load (top of graphs, P = .093), with no significant changes in the control group (bottom).

Total cholesterol concentrations dropped 16.2 ± 6.3 mg/dL in the oat cereal group (P = .030), with a slight (nonsignificant) increase in the control group (P = .48). Additionally, a comparison of the changes in total cholesterol between the 2 groups revealed a significant mean difference of 21.1 ± 9.1 mg/dL (P = .035). LDL cholesterol was also reduced significantly after the oat cereal intervention by 15.8 ± 5.9 mg/dL (P = .025). The nonsignificant increase in LDL cholesterol in the control group (P = .231) combined with the significant reduction in the treatment group resulted in a significant difference between the groups after intervention (P < .015). Neither group experienced significant changes in HDL cholesterol or triglyceride concentrations.

An analysis of the side effect data showed no significant difference in the occurrence of side effects between groups. There was an overall decrease in the frequency of dietary fiber-related and hypertension-related side effects in both groups, with a more substantial reduction occurring in the oat cereal group (P = .11). Total body weight did not change significantly in either group. Additionally, both groups were very compliant (approximately 90%) in terms of cereal consumption Table 3.

Discussion

The results of this pilot study suggest that the inclusion of oats into the standard American diet of people with borderline or mild hypertension may reduce both SBP and DBP. In persons consuming 5.52 g/day of beta-glucan soluble fiber from oat cereal for 6 weeks, we found a statistically and clinically significant decrease in both SBP and DBP (7.5 mm Hg and 5.5 mm Hg, respectively) and a trend toward improved OGTT-determined insulin sensitivity. These findings warrant a large-scale clinical trial to explore further the relationship between whole-grain oat consumption and blood pressure, especially considering the limitations of this pilot study.

As with all small-scale trials, this one lacked sufficient power to detect true changes in both primary and secondary outcome variables. It is possible that regression to the mean explains at least part of the treatment effect, since participants in the oats group began the study with higher SBP, DBP, and LDL cholesterol levels than controls. In addition, it is possible that the reported blood pressure changes could have been caused by “other” undetected dietary change made by members of the oats group. Future trials will need to collect and analyze dietary data carefully; feeding trials should be considered. Such dietary analyses may indicate that certain micronutrients partially explain the hypotensive effects of whole-grain oat consumption. The DASH trial and others have consistently demonstrated that diets rich in certain micronutrients can reduce blood pressure.^30,31

Soluble fiber-rich oat cereals may affect blood pressure by modulating changes in insulin metabolism. The mechanism of action is thought to involve the slowed absorption of macronutrients from the gut, resulting in a flattening of the postprandial glycemic curve.²⁹ These lower postprandial blood glucose levels elicit a lower insulin response to accommodate its clearance from the plasma. This process may lead to improved insulin sensitivity if the lower circulating insulin levels lead eventually to upregulation of the insulin receptors in peripheral tissues. A recent animal trial demonstrated that soluble fiber feeding improved insulin sensitivity by increasing skeletal muscle plasma membrane GLUT-4 content.³² Findings in this pilot suggest that over time, oat ingestion may reduce the amount of insulin needed to clear a glucose load. However, the study was underpowered to detect significant differences in more sensitive measures of insulin resistance. The causal mechanistic relationship among whole grain oat consumption, blood pressure, and insulin resistance might be best studied using a long-term feeding study design.

Alternate mechanisms, such as attenuation in endothelial function, may have affected blood pressure responses in this study.³³ Drugs specific to endothelial cell receptors mediating vasodilation are known to lower blood pressure.³⁴ Moreover, plasma cholesterol reductions are associated with improvements in endothelium-mediated vasodilation.^35,36 In addition, preliminary evidence in animals supports a direct relationship between changes in plasma cholesterol concentrations and blood pressure.³⁷ In the present study, plasma cholesterol levels were significantly reduced in participants who ingested whole grain oat-based cereals compared to a more refined grain wheat, corn, and rice control. Thus, it is possible that the blood pressure reduction observed in the subjects consuming oats resulted in part from improved endothelial function due to a drop in plasma cholesterol. Additional research is needed to fully investigate this pathway.

From a practical standpoint, improvements in SBP and DBP such as those observed in this study would be a useful contribution to the clinical management of hypertension. The cereal feeding intervention was well tolerated. Participants were very compliant for the 6-week treatment period. Substantial improvements in blood lipids could serve as an added incentive for patients to maintain long-term compliance with feeding recommendations.^18,19 Since treatment of hypertension is a lifelong process for most patients, future studies would need to assess the effectiveness of oat cereals to maintain blood pressure benefits over a longer time. Such studies may need to consider dietary options such as soluble fiber-rich fruits in addition to cereal consumption in efforts to deliver the desired quantity of soluble fiber. Future trials will have to investigate the antihypertensive effect of whole oats in other populations, such as people with diabetes, and to study not only surrogate endpoints such as blood pressure but also patient-oriented outcomes such as mortality and morbidity.

For more than 50 years academic medicine has held a privileged position in American society. Medical schools receive significant state and federal support from a variety of sources, including the National Institutes of Health, Public Health Services programs specifically developed to support medical education, the National Science Foundation, Medicare, and Medicaid.^1-4 In return, academic medical centers have provided training to medical students and residents and have made significant contributions to medical research and clinical care.^1,3 Recently, however, concern has been voiced about whether academic health centers have fulfilled important components of their tacit social contract with the American public, caused in part by changes in medical education financing, trends toward a competition model of health care delivery, and the erosion of trust between health care providers and patients.^1,3,5-8

Foreman⁹ suggested changes to medical school education that would help academic health centers fulfill their reciprocal social obligation to improve the public’s health. His recommendations included integrating behavioral and population-based sciences, providing students with learning experiences in community settings where they have the opportunity to work with committed mentors, and developing a critical mass of community-based faculty who are dedicated to addressing the various needs of underserved communities and providing them with the necessary support to continue their community-based efforts. Some academic health centers, including the University of New Mexico School of Medicine, The Johns Hopkins University School of Medicine, the University of Washington School of Medicine, and the Medical College of Pennsylvania/Hahnemann School of Medicine in Philadelphia have begun to implement some of Foreman’s suggestions to strengthen social responsiveness.^3,10-12

Within academic medical centers, departments of family medicine have pioneered placing medical students in community-based settings. Of the 124 medical schools that participated in the annual Liaison Committee on Medical Education survey in 1996, 69% of family practice clerkships had a community-based placement, compared with 40% for internal medicine and 25% for pediatric clerkships.¹³

Family practice residency programs have also striven to respond to the needs of their surrounding communities. In 1999 the Strategic Planning Working Group of the Academic Family Medicine Organization and the Association of Family Practice Residency Directors developed the following list of competencies for family practice residents to acquire during training: (1) family practice residents should understand Community-Oriented Primary Care (COPC) and the practice of population-based medicine; (2) family practice residencies should model COPC or population-based interventions within their practices; and (3) family practice graduates should be capable of recognizing community health needs, developing interventions, and assessing the outcomes.¹⁴ Several family practice residency programs, such as the one at Montefiore Medical Center in New York City have worked to address their communities’ concerns by implementing COPC.¹⁵

Less information is available on the involvement in community activities of individual family medicine professionals, which include faculty medical doctors (MDs), nonfaculty MDs, doctors of philosophy (PhDs), and master’s degree-prepared department members. The objective of our study was to determine the nature and type of community involvement of professionals in departments of family medicine. We also assessed community activities, challenges and incentives to those activities, and desired resources for working in the community. Insights into these topics increase our understanding of how personnel in academic health centers are attempting to meet the challenge of responding to the health care needs of their surrounding communities while they maintain a commitment to the traditional missions of education, research, and clinical service.

Methods

A pilot survey was sent to 25 members of the Society of Teachers of Family Medicine (STFM). Minor revisions were made according to respondents’ feedback, resulting in a 24-item, structured survey that we mailed to a national random sample of physician and nonphysician active members of STFM (N = 770). The first section of the questionnaire asked respondents a series of demographic and descriptive questions, including participants’ age, sex, ethnicity, professional effort (full or part time), length of time in their current department, and the year they completed residency or a doctoral degree. Additional information was collected on a variety of topics, including type of community-based involvement, reasons for that involvement, challenges to community-based involvement, and support or resources desired from their departments. A list of community-based activities was provided on the questionnaire, as was one write-in option Table 1. Although all activities were community-oriented, not all activities were conducted in the community.

Surveys were distributed in 2 mailings over a 6-month period with the second mailing going only to nonrespondents. Descriptive statistics consisting of percentages for categorical variables and medians for continuous, non-normally distributed variables were calculated. Univariate analyses were accomplished with the chi-square test or, in the case of non-normally distributed variables (age, years in the department, percentage of professional time spent on community-based activities), with the nonparametric Wilcoxon rank sum test.

Multiple logistic regression was used to examine the relationship between binary outcome variables and multiple explanatory variables. The logistic regression outcomes we considered were the individual types of community involvement, barriers to community involvement, and support desired. The candidate explanatory variables were chosen a priori: age, sex, degree (4 categories: master’s degree [reference group], MD degree, PhD degree, and both MD and PhD degrees), and years in department. Following the structure of the survey, analyses of barriers and desired support were restricted to those who had some type of community-based involvement in the previous year. A backward selection stepwise technique was used to build the models. Explanatory variable effects are shown as odds ratios (ORs) with 95% confidence intervals (CIs). For all analyses we used the Stata 6.0 statistical software package.¹⁶

Results

A total of 446 usable surveys were returned (58% response rate). Of these, 3 were blank and therefore unusable. Demographic characteristics indicated that respondents were representative of active STFM membership and national family medicine department faculty as reported by the Association of American Medical Colleges¹⁷Table 2.

Ninety-five percent of respondents had participated in a community-based project within the previous 12 months. Projects represented a continuum of involvement with community members. Nevertheless, much of the community-based activity was traditional in nature and included precepting medical students and residents in the community, providing clinical services at community-based sites, and conducting educational presentations in the community Table 1. When we considered only activities actually taking place within the community and excluded education about the community that took place elsewhere (the second, third, and fourth items under the heading “Any Education” in Table 1), 92% of respondents had been involved in a community-based project in the previous 12 months.

Faculty participated in community projects for several reasons, the most prevalent being personal interest or satisfaction (77%). Respondents identified insufficient time as the biggest barrier to involvement in community-based activities and noted sufficient release time as the most important form of support or resources they desired from their departments Table 3. Respondents’ academic institutions were most likely to serve urban communities (60%), followed by suburban (33%), small town (20%), and rural (16%) communities.

The association between types of community involvement and respondents’ sex, age, and professional degree was examined with logistic regression analysis. Even when controlling for degree, more men than women reported educating faculty on topics regarding community-based education and how to precept medical students or residents in community sites (OR = 2.01; 95% CI, 1.20 - 3.37; P = .008) and providing clinical care at community-based sites (OR = 1.73; 95% CI, 1.14 - 2.61; P = .009). The longer a respondent had been a member of a department, the more likely he or she was to report having served as a board, committee, or council member of a community health organization, even after controlling for age (for each 5-year interval spent in their department: OR = 1.23; 95% CI, 1.03 - 1.47; P = .023). Not surprisingly, MDs were 5.27 times more likely to report that they had precepted medical students or residents at community-based sites (95% CI, 1.29 - 21.46; P = .02) and provided medical care at community-based sites (OR = 5.35; 95% CI, 1.08 - 26.47; P = .04) than non-MD respondents. MD and PhD respondents, however, were less likely than those without such degrees to work with community members to develop and implement a research project to meet a community-identified health concern (PhDs: OR = 0.17; 95% CI, 0.04-0.84; P = .03; MDs: OR = 0.28; 95% CI, 0.07-1.09; P = .07).

We also analyzed the type of community served to determine its effect on participation in community activities. Institutions serving rural communities were more likely to have designed a community health curriculum (51% vs 36%; P = .023 by Fisher’s exact test) and to have evaluated a community-based project or program (32% vs 18%; P = .010 by Fisher’s exact test). Those serving a small town were also more likely to have evaluated a community-based project or program (30% versus 18%; P = .026 by Fisher’s exact test). Those serving urban communities were more likely to have taught students to work in a community site (58% vs 48%; P = .052 by Fisher’s exact test), to have designed a community health curriculum (43% vs 31%; P = .010 by Fisher’s exact test), and to have educated faculty on community-based education (27% vs 17%; P = .021 by Fisher’s exact test). Neither community served nor community activity, however, is mutually exclusive.

Some of the barriers to community-based activities and desired support for such work were also associated with respondents’ sex, age, and number of years in the current department. Women were 2.41 times more likely than men to report a lack of technical assistance as a barrier to community-based projects (95% CI, 1.41-4.13; P = .001). However, women were only 1.56 times (95% CI, 0.99-2.46; P = .054) more likely than men to desire technical support from their department. Men were 1.57 times (95% CI, 0.99-2.48; P = .054) more likely than women to desire help in forming relationships with the community. Increased age was associated with a decreased desire for sufficient release or protected time for community-based work. For each decade increase in age, there was a 28% reduction in the perceived need for sufficient release or protected time (OR = 0.72; 95% CI, 0.55-0.95; P = .02). Similarly, respondents who had been in their departments longer were less likely to report a need for faculty development regarding community-based activities (OR = 0.95 for each year [a 5% reduction for each additional year]; 95% CI, 0.91-0.99; P = .009).

Discussion

Advocates of community health have challenged academic institutions to more and better involvement in teaching and researching community health and providing service in the community. However, there are almost no data describing the status quo. Our study of 446 health providers who demographically mirror current STFM members and family medicine department faculty establishes a baseline of current activities. The findings support some of our beliefs, call others into question, and raise a number of specific areas for further study.

First, our results indicate that significant numbers of family medicine personnel are participating in a variety of community-based activities. Ninety-five percent of those responding reported having participated in a community education, service, or research project in the past year; 92% performed those activities in the community itself. The activities included precepting medical students and residents, providing clinical services at community-based sites, and making educational presentations in the community. Although this finding does not obviate the need for more and better services, it does suggest that faculty are fulfilling their responsibilities in this area. Less than half of our respondents participated in research, however, a finding that merits further investigation.

Second, this group of physicians and other family and community medicine personnel reported personal interest and satisfaction as the primary motivation (77%)for participating in community projects. This finding supports attempts to motivate community involvement as a personally rewarding experience. Other motivating factors were health of the community and importance to medical student and resident education.

Predictably, the most commonly perceived barrier to community service project participation was a lack of time. More release time was the most desired form of department support for surmounting that barrier. However, we found no data about release time and service. Bland and Schmitz¹⁸ have suggested that dedicating 40% of effort to research is necessary for adequate research productivity. If community service is a mission of a medical school, it seems that protecting time for community service projects would also be necessary. Further research is needed to ascertain whether schools offer faculty protected time for community service and, if so, how much is necessary or optimal.

Participation patterns, perceived barriers, and desired resources varied by age, sex, educational background, and academic rank. These factors are often interrelated and individual effects are difficult to segregate. Greater experience and time with an organization may be associated with higher status (rank), which in turn may lead to greater access to monetary and other resources, more protected time, and greater ability to allocate one’s own time. There are still more male family physicians than female, and more men have higher faculty rank. These factors may affect our findings that men were significantly more often involved in teaching other faculty about community-based education and providing care at a community or school clinic.

Among respondents who desired technical assistance, women were nearly twice as likely as men to report lack of technical assistance as a barrier (61% versus 34%) but desired technical assistance only slightly more often than men. Men reported desiring help in forming relationships with community members more often than women did; the difference approached, but did not achieve, significance. The findings are intriguing, but speculation about their implications would be based on stereotypes. Certainly further investigation is desirable.

Controlling for age, the longer the respondents had been members of their departments, the more often they participated on community health boards, committees, and organizations. Respondents who have been in their departments longer may be better established in their careers and in the community, resulting in more frequent invitations to these activities. Other explanations could include changes accompanying life stages, such as concern for assisting younger generations.

Length of employment in a particular department correlated with less reported need for faculty development around community-based activities. Since we did not attempt to ascertain respondents’ levels of expertise, we cannot interpret this finding. However, it cannot be assumed that long experience and lack of reported need necessarily reflect a high skill level.

Older respondents were less likely to desire release or supported time for their community activities. The perceived need for more time diminished by 28% with each decade of life. It may be that they have already garnered sufficient support and protected time in their institutions.

That MDs were significantly more likely than non-MDs to have precepted students and residents at community sites reflects the requirements of medical education accrediting bodies. The reason for the prevalence of research by respondents with master’s degrees and not those with terminal degrees is not known, although we surmise that at least some may have been hired specifically to conduct research. More study of the role of this small subset of respondents is warranted.

We did not examine differences in practice environments and their effect on community-based activities. University-, military-, and community-based practices have different goals, incentives, and disincentives, as do managed care and fee-for-service organizations. Furthermore, the traditional patterns of these organizations may be changing in response to interest in performance measures.¹⁹ This is another important area for investigation.

Limitations

This is a descriptive, not a definitive, study. The 58% response rate to the survey may limit the generalizability of our findings. Individuals who are involved or interested in community projects may have been more likely to return the survey, resulting in an overestimation of involvement in community-based activities. Although we do not have demographic or community involvement information about nonrespondents, our sample is demographically similar to active STFM membership and national family medicine department faculty. We provided examples of community-based activities; however, individual interpretations of what constitutes such an activity may differ. Using exploratory analyses increased the likelihood that a significant result would occur by chance. Thus, marginally significant results require further study, and those with P values between .01 and .05 should be considered hypothesis generating.

Conclusions

This descriptive study helps establish a baseline for better understanding academic physicians’ current participation in community-based activities. Although the scope of this study is narrow, it suggests that most academic faculty are providing community service and education and are deriving satisfaction from doing so.

Our results also raise a number of questions for further study. Is there enough appropriate research being done within communities to address its health needs? Should women and younger faculty receive additional support in establishing community-based activities, and if so, what kind? If women perceive technical barriers more often, why do they not report a desire for technical assistance more often? Is the difference between men and women in ease of forming community partnerships meaningful? The answers to these questions will provide a richer understanding of the ability of an academic health center to respond to the health care needs of their surrounding communities.

For more than 50 years academic medicine has held a privileged position in American society. Medical schools receive significant state and federal support from a variety of sources, including the National Institutes of Health, Public Health Services programs specifically developed to support medical education, the National Science Foundation, Medicare, and Medicaid.^1-4 In return, academic medical centers have provided training to medical students and residents and have made significant contributions to medical research and clinical care.^1,3 Recently, however, concern has been voiced about whether academic health centers have fulfilled important components of their tacit social contract with the American public, caused in part by changes in medical education financing, trends toward a competition model of health care delivery, and the erosion of trust between health care providers and patients.^1,3,5-8

Foreman⁹ suggested changes to medical school education that would help academic health centers fulfill their reciprocal social obligation to improve the public’s health. His recommendations included integrating behavioral and population-based sciences, providing students with learning experiences in community settings where they have the opportunity to work with committed mentors, and developing a critical mass of community-based faculty who are dedicated to addressing the various needs of underserved communities and providing them with the necessary support to continue their community-based efforts. Some academic health centers, including the University of New Mexico School of Medicine, The Johns Hopkins University School of Medicine, the University of Washington School of Medicine, and the Medical College of Pennsylvania/Hahnemann School of Medicine in Philadelphia have begun to implement some of Foreman’s suggestions to strengthen social responsiveness.^3,10-12

Within academic medical centers, departments of family medicine have pioneered placing medical students in community-based settings. Of the 124 medical schools that participated in the annual Liaison Committee on Medical Education survey in 1996, 69% of family practice clerkships had a community-based placement, compared with 40% for internal medicine and 25% for pediatric clerkships.¹³

Family practice residency programs have also striven to respond to the needs of their surrounding communities. In 1999 the Strategic Planning Working Group of the Academic Family Medicine Organization and the Association of Family Practice Residency Directors developed the following list of competencies for family practice residents to acquire during training: (1) family practice residents should understand Community-Oriented Primary Care (COPC) and the practice of population-based medicine; (2) family practice residencies should model COPC or population-based interventions within their practices; and (3) family practice graduates should be capable of recognizing community health needs, developing interventions, and assessing the outcomes.¹⁴ Several family practice residency programs, such as the one at Montefiore Medical Center in New York City have worked to address their communities’ concerns by implementing COPC.¹⁵

Less information is available on the involvement in community activities of individual family medicine professionals, which include faculty medical doctors (MDs), nonfaculty MDs, doctors of philosophy (PhDs), and master’s degree-prepared department members. The objective of our study was to determine the nature and type of community involvement of professionals in departments of family medicine. We also assessed community activities, challenges and incentives to those activities, and desired resources for working in the community. Insights into these topics increase our understanding of how personnel in academic health centers are attempting to meet the challenge of responding to the health care needs of their surrounding communities while they maintain a commitment to the traditional missions of education, research, and clinical service.

Methods

A pilot survey was sent to 25 members of the Society of Teachers of Family Medicine (STFM). Minor revisions were made according to respondents’ feedback, resulting in a 24-item, structured survey that we mailed to a national random sample of physician and nonphysician active members of STFM (N = 770). The first section of the questionnaire asked respondents a series of demographic and descriptive questions, including participants’ age, sex, ethnicity, professional effort (full or part time), length of time in their current department, and the year they completed residency or a doctoral degree. Additional information was collected on a variety of topics, including type of community-based involvement, reasons for that involvement, challenges to community-based involvement, and support or resources desired from their departments. A list of community-based activities was provided on the questionnaire, as was one write-in option Table 1. Although all activities were community-oriented, not all activities were conducted in the community.

Surveys were distributed in 2 mailings over a 6-month period with the second mailing going only to nonrespondents. Descriptive statistics consisting of percentages for categorical variables and medians for continuous, non-normally distributed variables were calculated. Univariate analyses were accomplished with the chi-square test or, in the case of non-normally distributed variables (age, years in the department, percentage of professional time spent on community-based activities), with the nonparametric Wilcoxon rank sum test.

Multiple logistic regression was used to examine the relationship between binary outcome variables and multiple explanatory variables. The logistic regression outcomes we considered were the individual types of community involvement, barriers to community involvement, and support desired. The candidate explanatory variables were chosen a priori: age, sex, degree (4 categories: master’s degree [reference group], MD degree, PhD degree, and both MD and PhD degrees), and years in department. Following the structure of the survey, analyses of barriers and desired support were restricted to those who had some type of community-based involvement in the previous year. A backward selection stepwise technique was used to build the models. Explanatory variable effects are shown as odds ratios (ORs) with 95% confidence intervals (CIs). For all analyses we used the Stata 6.0 statistical software package.¹⁶

Results

A total of 446 usable surveys were returned (58% response rate). Of these, 3 were blank and therefore unusable. Demographic characteristics indicated that respondents were representative of active STFM membership and national family medicine department faculty as reported by the Association of American Medical Colleges¹⁷Table 2.

Ninety-five percent of respondents had participated in a community-based project within the previous 12 months. Projects represented a continuum of involvement with community members. Nevertheless, much of the community-based activity was traditional in nature and included precepting medical students and residents in the community, providing clinical services at community-based sites, and conducting educational presentations in the community Table 1. When we considered only activities actually taking place within the community and excluded education about the community that took place elsewhere (the second, third, and fourth items under the heading “Any Education” in Table 1), 92% of respondents had been involved in a community-based project in the previous 12 months.

Faculty participated in community projects for several reasons, the most prevalent being personal interest or satisfaction (77%). Respondents identified insufficient time as the biggest barrier to involvement in community-based activities and noted sufficient release time as the most important form of support or resources they desired from their departments Table 3. Respondents’ academic institutions were most likely to serve urban communities (60%), followed by suburban (33%), small town (20%), and rural (16%) communities.

The association between types of community involvement and respondents’ sex, age, and professional degree was examined with logistic regression analysis. Even when controlling for degree, more men than women reported educating faculty on topics regarding community-based education and how to precept medical students or residents in community sites (OR = 2.01; 95% CI, 1.20 - 3.37; P = .008) and providing clinical care at community-based sites (OR = 1.73; 95% CI, 1.14 - 2.61; P = .009). The longer a respondent had been a member of a department, the more likely he or she was to report having served as a board, committee, or council member of a community health organization, even after controlling for age (for each 5-year interval spent in their department: OR = 1.23; 95% CI, 1.03 - 1.47; P = .023). Not surprisingly, MDs were 5.27 times more likely to report that they had precepted medical students or residents at community-based sites (95% CI, 1.29 - 21.46; P = .02) and provided medical care at community-based sites (OR = 5.35; 95% CI, 1.08 - 26.47; P = .04) than non-MD respondents. MD and PhD respondents, however, were less likely than those without such degrees to work with community members to develop and implement a research project to meet a community-identified health concern (PhDs: OR = 0.17; 95% CI, 0.04-0.84; P = .03; MDs: OR = 0.28; 95% CI, 0.07-1.09; P = .07).

We also analyzed the type of community served to determine its effect on participation in community activities. Institutions serving rural communities were more likely to have designed a community health curriculum (51% vs 36%; P = .023 by Fisher’s exact test) and to have evaluated a community-based project or program (32% vs 18%; P = .010 by Fisher’s exact test). Those serving a small town were also more likely to have evaluated a community-based project or program (30% versus 18%; P = .026 by Fisher’s exact test). Those serving urban communities were more likely to have taught students to work in a community site (58% vs 48%; P = .052 by Fisher’s exact test), to have designed a community health curriculum (43% vs 31%; P = .010 by Fisher’s exact test), and to have educated faculty on community-based education (27% vs 17%; P = .021 by Fisher’s exact test). Neither community served nor community activity, however, is mutually exclusive.

Some of the barriers to community-based activities and desired support for such work were also associated with respondents’ sex, age, and number of years in the current department. Women were 2.41 times more likely than men to report a lack of technical assistance as a barrier to community-based projects (95% CI, 1.41-4.13; P = .001). However, women were only 1.56 times (95% CI, 0.99-2.46; P = .054) more likely than men to desire technical support from their department. Men were 1.57 times (95% CI, 0.99-2.48; P = .054) more likely than women to desire help in forming relationships with the community. Increased age was associated with a decreased desire for sufficient release or protected time for community-based work. For each decade increase in age, there was a 28% reduction in the perceived need for sufficient release or protected time (OR = 0.72; 95% CI, 0.55-0.95; P = .02). Similarly, respondents who had been in their departments longer were less likely to report a need for faculty development regarding community-based activities (OR = 0.95 for each year [a 5% reduction for each additional year]; 95% CI, 0.91-0.99; P = .009).

Discussion

Advocates of community health have challenged academic institutions to more and better involvement in teaching and researching community health and providing service in the community. However, there are almost no data describing the status quo. Our study of 446 health providers who demographically mirror current STFM members and family medicine department faculty establishes a baseline of current activities. The findings support some of our beliefs, call others into question, and raise a number of specific areas for further study.

First, our results indicate that significant numbers of family medicine personnel are participating in a variety of community-based activities. Ninety-five percent of those responding reported having participated in a community education, service, or research project in the past year; 92% performed those activities in the community itself. The activities included precepting medical students and residents, providing clinical services at community-based sites, and making educational presentations in the community. Although this finding does not obviate the need for more and better services, it does suggest that faculty are fulfilling their responsibilities in this area. Less than half of our respondents participated in research, however, a finding that merits further investigation.

Second, this group of physicians and other family and community medicine personnel reported personal interest and satisfaction as the primary motivation (77%)for participating in community projects. This finding supports attempts to motivate community involvement as a personally rewarding experience. Other motivating factors were health of the community and importance to medical student and resident education.

Predictably, the most commonly perceived barrier to community service project participation was a lack of time. More release time was the most desired form of department support for surmounting that barrier. However, we found no data about release time and service. Bland and Schmitz¹⁸ have suggested that dedicating 40% of effort to research is necessary for adequate research productivity. If community service is a mission of a medical school, it seems that protecting time for community service projects would also be necessary. Further research is needed to ascertain whether schools offer faculty protected time for community service and, if so, how much is necessary or optimal.

Participation patterns, perceived barriers, and desired resources varied by age, sex, educational background, and academic rank. These factors are often interrelated and individual effects are difficult to segregate. Greater experience and time with an organization may be associated with higher status (rank), which in turn may lead to greater access to monetary and other resources, more protected time, and greater ability to allocate one’s own time. There are still more male family physicians than female, and more men have higher faculty rank. These factors may affect our findings that men were significantly more often involved in teaching other faculty about community-based education and providing care at a community or school clinic.

Among respondents who desired technical assistance, women were nearly twice as likely as men to report lack of technical assistance as a barrier (61% versus 34%) but desired technical assistance only slightly more often than men. Men reported desiring help in forming relationships with community members more often than women did; the difference approached, but did not achieve, significance. The findings are intriguing, but speculation about their implications would be based on stereotypes. Certainly further investigation is desirable.

Controlling for age, the longer the respondents had been members of their departments, the more often they participated on community health boards, committees, and organizations. Respondents who have been in their departments longer may be better established in their careers and in the community, resulting in more frequent invitations to these activities. Other explanations could include changes accompanying life stages, such as concern for assisting younger generations.

Length of employment in a particular department correlated with less reported need for faculty development around community-based activities. Since we did not attempt to ascertain respondents’ levels of expertise, we cannot interpret this finding. However, it cannot be assumed that long experience and lack of reported need necessarily reflect a high skill level.

Older respondents were less likely to desire release or supported time for their community activities. The perceived need for more time diminished by 28% with each decade of life. It may be that they have already garnered sufficient support and protected time in their institutions.

That MDs were significantly more likely than non-MDs to have precepted students and residents at community sites reflects the requirements of medical education accrediting bodies. The reason for the prevalence of research by respondents with master’s degrees and not those with terminal degrees is not known, although we surmise that at least some may have been hired specifically to conduct research. More study of the role of this small subset of respondents is warranted.

We did not examine differences in practice environments and their effect on community-based activities. University-, military-, and community-based practices have different goals, incentives, and disincentives, as do managed care and fee-for-service organizations. Furthermore, the traditional patterns of these organizations may be changing in response to interest in performance measures.¹⁹ This is another important area for investigation.

Limitations

This is a descriptive, not a definitive, study. The 58% response rate to the survey may limit the generalizability of our findings. Individuals who are involved or interested in community projects may have been more likely to return the survey, resulting in an overestimation of involvement in community-based activities. Although we do not have demographic or community involvement information about nonrespondents, our sample is demographically similar to active STFM membership and national family medicine department faculty. We provided examples of community-based activities; however, individual interpretations of what constitutes such an activity may differ. Using exploratory analyses increased the likelihood that a significant result would occur by chance. Thus, marginally significant results require further study, and those with P values between .01 and .05 should be considered hypothesis generating.

Conclusions

This descriptive study helps establish a baseline for better understanding academic physicians’ current participation in community-based activities. Although the scope of this study is narrow, it suggests that most academic faculty are providing community service and education and are deriving satisfaction from doing so.

Our results also raise a number of questions for further study. Is there enough appropriate research being done within communities to address its health needs? Should women and younger faculty receive additional support in establishing community-based activities, and if so, what kind? If women perceive technical barriers more often, why do they not report a desire for technical assistance more often? Is the difference between men and women in ease of forming community partnerships meaningful? The answers to these questions will provide a richer understanding of the ability of an academic health center to respond to the health care needs of their surrounding communities.

For more than 50 years academic medicine has held a privileged position in American society. Medical schools receive significant state and federal support from a variety of sources, including the National Institutes of Health, Public Health Services programs specifically developed to support medical education, the National Science Foundation, Medicare, and Medicaid.^1-4 In return, academic medical centers have provided training to medical students and residents and have made significant contributions to medical research and clinical care.^1,3 Recently, however, concern has been voiced about whether academic health centers have fulfilled important components of their tacit social contract with the American public, caused in part by changes in medical education financing, trends toward a competition model of health care delivery, and the erosion of trust between health care providers and patients.^1,3,5-8

Foreman⁹ suggested changes to medical school education that would help academic health centers fulfill their reciprocal social obligation to improve the public’s health. His recommendations included integrating behavioral and population-based sciences, providing students with learning experiences in community settings where they have the opportunity to work with committed mentors, and developing a critical mass of community-based faculty who are dedicated to addressing the various needs of underserved communities and providing them with the necessary support to continue their community-based efforts. Some academic health centers, including the University of New Mexico School of Medicine, The Johns Hopkins University School of Medicine, the University of Washington School of Medicine, and the Medical College of Pennsylvania/Hahnemann School of Medicine in Philadelphia have begun to implement some of Foreman’s suggestions to strengthen social responsiveness.^3,10-12

Within academic medical centers, departments of family medicine have pioneered placing medical students in community-based settings. Of the 124 medical schools that participated in the annual Liaison Committee on Medical Education survey in 1996, 69% of family practice clerkships had a community-based placement, compared with 40% for internal medicine and 25% for pediatric clerkships.¹³

Family practice residency programs have also striven to respond to the needs of their surrounding communities. In 1999 the Strategic Planning Working Group of the Academic Family Medicine Organization and the Association of Family Practice Residency Directors developed the following list of competencies for family practice residents to acquire during training: (1) family practice residents should understand Community-Oriented Primary Care (COPC) and the practice of population-based medicine; (2) family practice residencies should model COPC or population-based interventions within their practices; and (3) family practice graduates should be capable of recognizing community health needs, developing interventions, and assessing the outcomes.¹⁴ Several family practice residency programs, such as the one at Montefiore Medical Center in New York City have worked to address their communities’ concerns by implementing COPC.¹⁵

Less information is available on the involvement in community activities of individual family medicine professionals, which include faculty medical doctors (MDs), nonfaculty MDs, doctors of philosophy (PhDs), and master’s degree-prepared department members. The objective of our study was to determine the nature and type of community involvement of professionals in departments of family medicine. We also assessed community activities, challenges and incentives to those activities, and desired resources for working in the community. Insights into these topics increase our understanding of how personnel in academic health centers are attempting to meet the challenge of responding to the health care needs of their surrounding communities while they maintain a commitment to the traditional missions of education, research, and clinical service.

Methods

A pilot survey was sent to 25 members of the Society of Teachers of Family Medicine (STFM). Minor revisions were made according to respondents’ feedback, resulting in a 24-item, structured survey that we mailed to a national random sample of physician and nonphysician active members of STFM (N = 770). The first section of the questionnaire asked respondents a series of demographic and descriptive questions, including participants’ age, sex, ethnicity, professional effort (full or part time), length of time in their current department, and the year they completed residency or a doctoral degree. Additional information was collected on a variety of topics, including type of community-based involvement, reasons for that involvement, challenges to community-based involvement, and support or resources desired from their departments. A list of community-based activities was provided on the questionnaire, as was one write-in option Table 1. Although all activities were community-oriented, not all activities were conducted in the community.

Surveys were distributed in 2 mailings over a 6-month period with the second mailing going only to nonrespondents. Descriptive statistics consisting of percentages for categorical variables and medians for continuous, non-normally distributed variables were calculated. Univariate analyses were accomplished with the chi-square test or, in the case of non-normally distributed variables (age, years in the department, percentage of professional time spent on community-based activities), with the nonparametric Wilcoxon rank sum test.

Multiple logistic regression was used to examine the relationship between binary outcome variables and multiple explanatory variables. The logistic regression outcomes we considered were the individual types of community involvement, barriers to community involvement, and support desired. The candidate explanatory variables were chosen a priori: age, sex, degree (4 categories: master’s degree [reference group], MD degree, PhD degree, and both MD and PhD degrees), and years in department. Following the structure of the survey, analyses of barriers and desired support were restricted to those who had some type of community-based involvement in the previous year. A backward selection stepwise technique was used to build the models. Explanatory variable effects are shown as odds ratios (ORs) with 95% confidence intervals (CIs). For all analyses we used the Stata 6.0 statistical software package.¹⁶

Results

A total of 446 usable surveys were returned (58% response rate). Of these, 3 were blank and therefore unusable. Demographic characteristics indicated that respondents were representative of active STFM membership and national family medicine department faculty as reported by the Association of American Medical Colleges¹⁷Table 2.

Ninety-five percent of respondents had participated in a community-based project within the previous 12 months. Projects represented a continuum of involvement with community members. Nevertheless, much of the community-based activity was traditional in nature and included precepting medical students and residents in the community, providing clinical services at community-based sites, and conducting educational presentations in the community Table 1. When we considered only activities actually taking place within the community and excluded education about the community that took place elsewhere (the second, third, and fourth items under the heading “Any Education” in Table 1), 92% of respondents had been involved in a community-based project in the previous 12 months.

Faculty participated in community projects for several reasons, the most prevalent being personal interest or satisfaction (77%). Respondents identified insufficient time as the biggest barrier to involvement in community-based activities and noted sufficient release time as the most important form of support or resources they desired from their departments Table 3. Respondents’ academic institutions were most likely to serve urban communities (60%), followed by suburban (33%), small town (20%), and rural (16%) communities.

The association between types of community involvement and respondents’ sex, age, and professional degree was examined with logistic regression analysis. Even when controlling for degree, more men than women reported educating faculty on topics regarding community-based education and how to precept medical students or residents in community sites (OR = 2.01; 95% CI, 1.20 - 3.37; P = .008) and providing clinical care at community-based sites (OR = 1.73; 95% CI, 1.14 - 2.61; P = .009). The longer a respondent had been a member of a department, the more likely he or she was to report having served as a board, committee, or council member of a community health organization, even after controlling for age (for each 5-year interval spent in their department: OR = 1.23; 95% CI, 1.03 - 1.47; P = .023). Not surprisingly, MDs were 5.27 times more likely to report that they had precepted medical students or residents at community-based sites (95% CI, 1.29 - 21.46; P = .02) and provided medical care at community-based sites (OR = 5.35; 95% CI, 1.08 - 26.47; P = .04) than non-MD respondents. MD and PhD respondents, however, were less likely than those without such degrees to work with community members to develop and implement a research project to meet a community-identified health concern (PhDs: OR = 0.17; 95% CI, 0.04-0.84; P = .03; MDs: OR = 0.28; 95% CI, 0.07-1.09; P = .07).

We also analyzed the type of community served to determine its effect on participation in community activities. Institutions serving rural communities were more likely to have designed a community health curriculum (51% vs 36%; P = .023 by Fisher’s exact test) and to have evaluated a community-based project or program (32% vs 18%; P = .010 by Fisher’s exact test). Those serving a small town were also more likely to have evaluated a community-based project or program (30% versus 18%; P = .026 by Fisher’s exact test). Those serving urban communities were more likely to have taught students to work in a community site (58% vs 48%; P = .052 by Fisher’s exact test), to have designed a community health curriculum (43% vs 31%; P = .010 by Fisher’s exact test), and to have educated faculty on community-based education (27% vs 17%; P = .021 by Fisher’s exact test). Neither community served nor community activity, however, is mutually exclusive.

Some of the barriers to community-based activities and desired support for such work were also associated with respondents’ sex, age, and number of years in the current department. Women were 2.41 times more likely than men to report a lack of technical assistance as a barrier to community-based projects (95% CI, 1.41-4.13; P = .001). However, women were only 1.56 times (95% CI, 0.99-2.46; P = .054) more likely than men to desire technical support from their department. Men were 1.57 times (95% CI, 0.99-2.48; P = .054) more likely than women to desire help in forming relationships with the community. Increased age was associated with a decreased desire for sufficient release or protected time for community-based work. For each decade increase in age, there was a 28% reduction in the perceived need for sufficient release or protected time (OR = 0.72; 95% CI, 0.55-0.95; P = .02). Similarly, respondents who had been in their departments longer were less likely to report a need for faculty development regarding community-based activities (OR = 0.95 for each year [a 5% reduction for each additional year]; 95% CI, 0.91-0.99; P = .009).

Discussion

Advocates of community health have challenged academic institutions to more and better involvement in teaching and researching community health and providing service in the community. However, there are almost no data describing the status quo. Our study of 446 health providers who demographically mirror current STFM members and family medicine department faculty establishes a baseline of current activities. The findings support some of our beliefs, call others into question, and raise a number of specific areas for further study.

First, our results indicate that significant numbers of family medicine personnel are participating in a variety of community-based activities. Ninety-five percent of those responding reported having participated in a community education, service, or research project in the past year; 92% performed those activities in the community itself. The activities included precepting medical students and residents, providing clinical services at community-based sites, and making educational presentations in the community. Although this finding does not obviate the need for more and better services, it does suggest that faculty are fulfilling their responsibilities in this area. Less than half of our respondents participated in research, however, a finding that merits further investigation.

Second, this group of physicians and other family and community medicine personnel reported personal interest and satisfaction as the primary motivation (77%)for participating in community projects. This finding supports attempts to motivate community involvement as a personally rewarding experience. Other motivating factors were health of the community and importance to medical student and resident education.

Predictably, the most commonly perceived barrier to community service project participation was a lack of time. More release time was the most desired form of department support for surmounting that barrier. However, we found no data about release time and service. Bland and Schmitz¹⁸ have suggested that dedicating 40% of effort to research is necessary for adequate research productivity. If community service is a mission of a medical school, it seems that protecting time for community service projects would also be necessary. Further research is needed to ascertain whether schools offer faculty protected time for community service and, if so, how much is necessary or optimal.

Participation patterns, perceived barriers, and desired resources varied by age, sex, educational background, and academic rank. These factors are often interrelated and individual effects are difficult to segregate. Greater experience and time with an organization may be associated with higher status (rank), which in turn may lead to greater access to monetary and other resources, more protected time, and greater ability to allocate one’s own time. There are still more male family physicians than female, and more men have higher faculty rank. These factors may affect our findings that men were significantly more often involved in teaching other faculty about community-based education and providing care at a community or school clinic.

Among respondents who desired technical assistance, women were nearly twice as likely as men to report lack of technical assistance as a barrier (61% versus 34%) but desired technical assistance only slightly more often than men. Men reported desiring help in forming relationships with community members more often than women did; the difference approached, but did not achieve, significance. The findings are intriguing, but speculation about their implications would be based on stereotypes. Certainly further investigation is desirable.

Controlling for age, the longer the respondents had been members of their departments, the more often they participated on community health boards, committees, and organizations. Respondents who have been in their departments longer may be better established in their careers and in the community, resulting in more frequent invitations to these activities. Other explanations could include changes accompanying life stages, such as concern for assisting younger generations.

Length of employment in a particular department correlated with less reported need for faculty development around community-based activities. Since we did not attempt to ascertain respondents’ levels of expertise, we cannot interpret this finding. However, it cannot be assumed that long experience and lack of reported need necessarily reflect a high skill level.

Older respondents were less likely to desire release or supported time for their community activities. The perceived need for more time diminished by 28% with each decade of life. It may be that they have already garnered sufficient support and protected time in their institutions.

That MDs were significantly more likely than non-MDs to have precepted students and residents at community sites reflects the requirements of medical education accrediting bodies. The reason for the prevalence of research by respondents with master’s degrees and not those with terminal degrees is not known, although we surmise that at least some may have been hired specifically to conduct research. More study of the role of this small subset of respondents is warranted.

We did not examine differences in practice environments and their effect on community-based activities. University-, military-, and community-based practices have different goals, incentives, and disincentives, as do managed care and fee-for-service organizations. Furthermore, the traditional patterns of these organizations may be changing in response to interest in performance measures.¹⁹ This is another important area for investigation.

Limitations

This is a descriptive, not a definitive, study. The 58% response rate to the survey may limit the generalizability of our findings. Individuals who are involved or interested in community projects may have been more likely to return the survey, resulting in an overestimation of involvement in community-based activities. Although we do not have demographic or community involvement information about nonrespondents, our sample is demographically similar to active STFM membership and national family medicine department faculty. We provided examples of community-based activities; however, individual interpretations of what constitutes such an activity may differ. Using exploratory analyses increased the likelihood that a significant result would occur by chance. Thus, marginally significant results require further study, and those with P values between .01 and .05 should be considered hypothesis generating.

Conclusions

This descriptive study helps establish a baseline for better understanding academic physicians’ current participation in community-based activities. Although the scope of this study is narrow, it suggests that most academic faculty are providing community service and education and are deriving satisfaction from doing so.

Our results also raise a number of questions for further study. Is there enough appropriate research being done within communities to address its health needs? Should women and younger faculty receive additional support in establishing community-based activities, and if so, what kind? If women perceive technical barriers more often, why do they not report a desire for technical assistance more often? Is the difference between men and women in ease of forming community partnerships meaningful? The answers to these questions will provide a richer understanding of the ability of an academic health center to respond to the health care needs of their surrounding communities.

Twenty-four consecutive patients of a single family physician who had achieved NCEP-II goal levels of low-density lipoprotein cholesterol (LDL-C) on a daily atorvastatin dose of 10 mg for at least 6 months were invited to switch to 20 mg weekly. Mean LDL levels for the 22 patients who completed the trial had been reduced by 43% from baseline on 10 mg daily (P < .05) and were reduced by 22% from baseline on the seventh day following the last weekly dose of 20 mg (P < .05). Total cholesterol to high-density lipoprotein cholesterol (TC/HDL-C) ratios were reduced by 31% and 17%, respectively (both P < .05) and triglycerides by 20% and 10% (both P < .05).

Atorvastatin is a potent antihyperlipidemic but is quite expensive, costing up to $700 for a 1-year supply of 10-mg tablets in retail pharmacies. Weekly dosing has the potential to lower costs and increase convenience while maintaining a similar effect on lipids. The active metabolites have a serum half-life of 11 to 57 hours,^1-4 acting on a target that responds slowly to intervention. Furthermore, atorvastatin demonstrates prolonged inhibition of HMG-CoA reductase compared with other statins, presumably because of longer residence of the drug or its metabolites in the liver.⁵ Two abstracts of meeting presentations have reported efficacy for alternate-day dosing on small patient samples,^1,2 but no reported attempts have been made to test longer dosing intervals. The purpose of this pilot study was to investigate the effect of weekly dosing of atorvastatin on patients who were currently well controlled on daily doses of the drug.

Methods

Selection of patients

Twenty-four consecutive patients presenting for routine follow-up in a private clinical practice whose LDL-C levels exceeded National Cholesterol Education Program (NCEP-II) guidelines⁶ on 2 occasions and who had successfully met their goals on 10 mg atorvastatin daily were offered a 12-week trial of 20 mg atorvastatin weekly. All 24 patients gave written informed consent, although one changed his mind and never altered his dosing and another experienced headaches after taking the first 20-mg dose and reverted to a regimen of 10 mg daily.

Study protocol

Study participants were instructed verbally and in writing to make no special efforts to change their lifestyle as a result of their involvement. The potential skewing effect of such efforts on research was explained. If they were already intending to alter their diet or exercise levels, this was acceptable. They were to take atorvastatin at the usual time of day on a day of the week that seemed most convenient. If they forgot a dose, it was to be taken the next day. Compliance with weekly dosing was assessed at each contact by explicit questioning. Fasting chemical and lipid profiles were available on patients’ charts; for purpose of analysis, the last profile before initiation of any statin therapy was used as the pretreatment baseline and the last profile on 10 mg atorvastatin daily as the treatment baseline. Profiles were repeated with the patient fasting on the seventh day after the last 20-mg dose before that dose was repeated.

Statistical analysis

The data were analyzed with repeated measures ANOVA followed by a Student–Newman–Kuels post hoc test to determine differences between specific treatments. Differences with a 2-tailed P value of less than .05 were considered statistically significant.

Results

Baseline characteristics

Table 1 presents the baseline characteristics of participants. The average age was 54 years (range 42 to 72 years). There were 12 men and 10 women. Thirteen subjects had comorbid conditions (9 hypertension, 4 type 2 diabetes, 1 hepatitis C, 2 coronary artery disease, and 2 tobacco use).

TABLE 1
BASELINE CHARACTERISTICS OF STUDY PARTICIPANTS

Cholesterol reduction

Results for LDL-C, HDL-C, triglycerides, TC, TC/HDL-C ratio, and aspartate aminotransferase (AST) are summarized in Table 2. LDL levels fell 43% and 22%, respectively, on daily and weekly dosing; HDL-C levels were essentially unchanged; triglycerides fell 20% and 10%; TC, 33% and 16%; TC/HDL-C, 31% and 17%; and AST, 0% and 21%.

TABLE 2
RESPONSE OF LIPID PARAMETERS TO DAILY AND WEEKLY DOSING WITH ATORVASTATIN

Adverse reactions

The only reported adverse reaction from doubling the dose of atorvastatin was headache in the patient who dropped out for that reason. No attempt was made to repeat the higher dose to see if this reaction was replicable. No subjects reported myalgias. The mean AST actually dropped on weekly dosing. One patient who had hepatitis C and was in clinical remission experienced a fall in pretreatment AST on daily dosing and a further reduction on weekly dosing.

Discussion

With pharmaceutical costs leading medical inflation, a current challenge for clinicians is to alter the cost-benefit ratio of prescriptions to the advantage of patients. Weekly dosing, as has recently been approved for alendronate sodium (Fosamax) and fluoxetine hydrochloride (Prozac), is one approach to this problem. In this preliminary study, weekly dosing of 20 mg atorvastatin resulted in a 22% reduction of LDL-C, measured on the seventh day after dosing. This regimen represents an approximately 80% reduction in yearly cost compared with that of a regimen of 10 mg daily.

Since this study did not investigate the pattern of LDL-C reduction in the interval between doses, further research is needed to delineate the area under the curve and the impact on clinical outcomes before conclusions may be drawn regarding the effectiveness of weekly dosing.

Acknowledgment

The author wishes to acknowledge the assistance of William Harris, PhD, in editing and statistical analysis.

Twenty-four consecutive patients of a single family physician who had achieved NCEP-II goal levels of low-density lipoprotein cholesterol (LDL-C) on a daily atorvastatin dose of 10 mg for at least 6 months were invited to switch to 20 mg weekly. Mean LDL levels for the 22 patients who completed the trial had been reduced by 43% from baseline on 10 mg daily (P < .05) and were reduced by 22% from baseline on the seventh day following the last weekly dose of 20 mg (P < .05). Total cholesterol to high-density lipoprotein cholesterol (TC/HDL-C) ratios were reduced by 31% and 17%, respectively (both P < .05) and triglycerides by 20% and 10% (both P < .05).

Atorvastatin is a potent antihyperlipidemic but is quite expensive, costing up to $700 for a 1-year supply of 10-mg tablets in retail pharmacies. Weekly dosing has the potential to lower costs and increase convenience while maintaining a similar effect on lipids. The active metabolites have a serum half-life of 11 to 57 hours,^1-4 acting on a target that responds slowly to intervention. Furthermore, atorvastatin demonstrates prolonged inhibition of HMG-CoA reductase compared with other statins, presumably because of longer residence of the drug or its metabolites in the liver.⁵ Two abstracts of meeting presentations have reported efficacy for alternate-day dosing on small patient samples,^1,2 but no reported attempts have been made to test longer dosing intervals. The purpose of this pilot study was to investigate the effect of weekly dosing of atorvastatin on patients who were currently well controlled on daily doses of the drug.

Methods

Selection of patients

Twenty-four consecutive patients presenting for routine follow-up in a private clinical practice whose LDL-C levels exceeded National Cholesterol Education Program (NCEP-II) guidelines⁶ on 2 occasions and who had successfully met their goals on 10 mg atorvastatin daily were offered a 12-week trial of 20 mg atorvastatin weekly. All 24 patients gave written informed consent, although one changed his mind and never altered his dosing and another experienced headaches after taking the first 20-mg dose and reverted to a regimen of 10 mg daily.

Study protocol

Study participants were instructed verbally and in writing to make no special efforts to change their lifestyle as a result of their involvement. The potential skewing effect of such efforts on research was explained. If they were already intending to alter their diet or exercise levels, this was acceptable. They were to take atorvastatin at the usual time of day on a day of the week that seemed most convenient. If they forgot a dose, it was to be taken the next day. Compliance with weekly dosing was assessed at each contact by explicit questioning. Fasting chemical and lipid profiles were available on patients’ charts; for purpose of analysis, the last profile before initiation of any statin therapy was used as the pretreatment baseline and the last profile on 10 mg atorvastatin daily as the treatment baseline. Profiles were repeated with the patient fasting on the seventh day after the last 20-mg dose before that dose was repeated.

Statistical analysis

The data were analyzed with repeated measures ANOVA followed by a Student–Newman–Kuels post hoc test to determine differences between specific treatments. Differences with a 2-tailed P value of less than .05 were considered statistically significant.

Results

Baseline characteristics

Table 1 presents the baseline characteristics of participants. The average age was 54 years (range 42 to 72 years). There were 12 men and 10 women. Thirteen subjects had comorbid conditions (9 hypertension, 4 type 2 diabetes, 1 hepatitis C, 2 coronary artery disease, and 2 tobacco use).

TABLE 1
BASELINE CHARACTERISTICS OF STUDY PARTICIPANTS

Cholesterol reduction

Results for LDL-C, HDL-C, triglycerides, TC, TC/HDL-C ratio, and aspartate aminotransferase (AST) are summarized in Table 2. LDL levels fell 43% and 22%, respectively, on daily and weekly dosing; HDL-C levels were essentially unchanged; triglycerides fell 20% and 10%; TC, 33% and 16%; TC/HDL-C, 31% and 17%; and AST, 0% and 21%.

TABLE 2
RESPONSE OF LIPID PARAMETERS TO DAILY AND WEEKLY DOSING WITH ATORVASTATIN

Adverse reactions

The only reported adverse reaction from doubling the dose of atorvastatin was headache in the patient who dropped out for that reason. No attempt was made to repeat the higher dose to see if this reaction was replicable. No subjects reported myalgias. The mean AST actually dropped on weekly dosing. One patient who had hepatitis C and was in clinical remission experienced a fall in pretreatment AST on daily dosing and a further reduction on weekly dosing.

Discussion

With pharmaceutical costs leading medical inflation, a current challenge for clinicians is to alter the cost-benefit ratio of prescriptions to the advantage of patients. Weekly dosing, as has recently been approved for alendronate sodium (Fosamax) and fluoxetine hydrochloride (Prozac), is one approach to this problem. In this preliminary study, weekly dosing of 20 mg atorvastatin resulted in a 22% reduction of LDL-C, measured on the seventh day after dosing. This regimen represents an approximately 80% reduction in yearly cost compared with that of a regimen of 10 mg daily.

Since this study did not investigate the pattern of LDL-C reduction in the interval between doses, further research is needed to delineate the area under the curve and the impact on clinical outcomes before conclusions may be drawn regarding the effectiveness of weekly dosing.

Acknowledgment

The author wishes to acknowledge the assistance of William Harris, PhD, in editing and statistical analysis.

Twenty-four consecutive patients of a single family physician who had achieved NCEP-II goal levels of low-density lipoprotein cholesterol (LDL-C) on a daily atorvastatin dose of 10 mg for at least 6 months were invited to switch to 20 mg weekly. Mean LDL levels for the 22 patients who completed the trial had been reduced by 43% from baseline on 10 mg daily (P < .05) and were reduced by 22% from baseline on the seventh day following the last weekly dose of 20 mg (P < .05). Total cholesterol to high-density lipoprotein cholesterol (TC/HDL-C) ratios were reduced by 31% and 17%, respectively (both P < .05) and triglycerides by 20% and 10% (both P < .05).

Atorvastatin is a potent antihyperlipidemic but is quite expensive, costing up to $700 for a 1-year supply of 10-mg tablets in retail pharmacies. Weekly dosing has the potential to lower costs and increase convenience while maintaining a similar effect on lipids. The active metabolites have a serum half-life of 11 to 57 hours,^1-4 acting on a target that responds slowly to intervention. Furthermore, atorvastatin demonstrates prolonged inhibition of HMG-CoA reductase compared with other statins, presumably because of longer residence of the drug or its metabolites in the liver.⁵ Two abstracts of meeting presentations have reported efficacy for alternate-day dosing on small patient samples,^1,2 but no reported attempts have been made to test longer dosing intervals. The purpose of this pilot study was to investigate the effect of weekly dosing of atorvastatin on patients who were currently well controlled on daily doses of the drug.

Methods

Selection of patients

Twenty-four consecutive patients presenting for routine follow-up in a private clinical practice whose LDL-C levels exceeded National Cholesterol Education Program (NCEP-II) guidelines⁶ on 2 occasions and who had successfully met their goals on 10 mg atorvastatin daily were offered a 12-week trial of 20 mg atorvastatin weekly. All 24 patients gave written informed consent, although one changed his mind and never altered his dosing and another experienced headaches after taking the first 20-mg dose and reverted to a regimen of 10 mg daily.

Study protocol

Study participants were instructed verbally and in writing to make no special efforts to change their lifestyle as a result of their involvement. The potential skewing effect of such efforts on research was explained. If they were already intending to alter their diet or exercise levels, this was acceptable. They were to take atorvastatin at the usual time of day on a day of the week that seemed most convenient. If they forgot a dose, it was to be taken the next day. Compliance with weekly dosing was assessed at each contact by explicit questioning. Fasting chemical and lipid profiles were available on patients’ charts; for purpose of analysis, the last profile before initiation of any statin therapy was used as the pretreatment baseline and the last profile on 10 mg atorvastatin daily as the treatment baseline. Profiles were repeated with the patient fasting on the seventh day after the last 20-mg dose before that dose was repeated.

Statistical analysis

The data were analyzed with repeated measures ANOVA followed by a Student–Newman–Kuels post hoc test to determine differences between specific treatments. Differences with a 2-tailed P value of less than .05 were considered statistically significant.

Results

Baseline characteristics

Table 1 presents the baseline characteristics of participants. The average age was 54 years (range 42 to 72 years). There were 12 men and 10 women. Thirteen subjects had comorbid conditions (9 hypertension, 4 type 2 diabetes, 1 hepatitis C, 2 coronary artery disease, and 2 tobacco use).

TABLE 1
BASELINE CHARACTERISTICS OF STUDY PARTICIPANTS

Cholesterol reduction

Results for LDL-C, HDL-C, triglycerides, TC, TC/HDL-C ratio, and aspartate aminotransferase (AST) are summarized in Table 2. LDL levels fell 43% and 22%, respectively, on daily and weekly dosing; HDL-C levels were essentially unchanged; triglycerides fell 20% and 10%; TC, 33% and 16%; TC/HDL-C, 31% and 17%; and AST, 0% and 21%.

TABLE 2
RESPONSE OF LIPID PARAMETERS TO DAILY AND WEEKLY DOSING WITH ATORVASTATIN

Adverse reactions

The only reported adverse reaction from doubling the dose of atorvastatin was headache in the patient who dropped out for that reason. No attempt was made to repeat the higher dose to see if this reaction was replicable. No subjects reported myalgias. The mean AST actually dropped on weekly dosing. One patient who had hepatitis C and was in clinical remission experienced a fall in pretreatment AST on daily dosing and a further reduction on weekly dosing.

Discussion

With pharmaceutical costs leading medical inflation, a current challenge for clinicians is to alter the cost-benefit ratio of prescriptions to the advantage of patients. Weekly dosing, as has recently been approved for alendronate sodium (Fosamax) and fluoxetine hydrochloride (Prozac), is one approach to this problem. In this preliminary study, weekly dosing of 20 mg atorvastatin resulted in a 22% reduction of LDL-C, measured on the seventh day after dosing. This regimen represents an approximately 80% reduction in yearly cost compared with that of a regimen of 10 mg daily.

Since this study did not investigate the pattern of LDL-C reduction in the interval between doses, further research is needed to delineate the area under the curve and the impact on clinical outcomes before conclusions may be drawn regarding the effectiveness of weekly dosing.

Acknowledgment

The author wishes to acknowledge the assistance of William Harris, PhD, in editing and statistical analysis.

According to a report by the National Arthritis Data Workgroup based on data from the first National Health and Nutrition Examination Survey (NHANES I), the prevalence of symptomatic hip osteoarthritis (OA) is 0.7% in both adult women and men; 0.5% have moderate or severe symptoms.¹ The incidence of symptomatic hip OA was higher in women than in men, and increased with age in both sexes until age 80 years, with a slight decline beyond that age.² In a study of patients in a health maintenance organization, the incidence of symptomatic hip OA in elderly women was 239 per 100,000 person-years at ages 60 to 69 years, 583 per 100,000 person-years at ages 70 to 79 years, and 441 per 100,000 person-years for women older than 80 years. The corresponding rates for elderly men per 100,000 person-years were 158, 445, and 264, respectively.

In contrast, the prevalence of radiographic hip OA is 3.1%; risk factors for the development of radiographic changes appear different in women and men.³ As has been demonstrated in the knee, radiographic hip OA is more prevalent than symptomatic disease. Only 61% of older individuals with confirmed radiologic hip OA reported pain at the hip, while 11% without radiologic changes of OA had significant hip pain, presumably of nonarticular etiology. Factors that affect patients’ reporting of symptoms of hip pain, both from OA and from nonarticular sources, have not yet been fully elucidated.

We are not aware of any current nationally representative reports on the sex-specific and race-specific prevalence of hip pain in older adults. Therefore, the purpose of our investigation was to examine the age-specific prevalence of hip pain using a current nationally representative sample of older US adults. We will also describe the prevalence of self-reported hip pain among older adults from 2 of the largest minority groups in the United States: non-Hispanic blacks and Mexican Americans.

Methods

Sample design

The Third National Health and Nutrition Examination Survey (NHANES III) was conducted by the Centers for Disease Control and Prevention, National Center for Health Statistics. The plan and operation of NHANES III have been described elsewhere.^5,6 Briefly, the survey was designed to produce a nationally representative sample of the civilian noninstitutionalized US population. One of the main goals of this survey was to estimate the national prevalence of selected health conditions and risk factors.

The NHANES III represents a 6-year study that was conducted from 1988 through 1994 consisting of 2 phases lasting 3 years each: phase I, 1988 through 1992, and phase II, 1991 through 1994. The survey was designed so that each phase was a nationally representative sample. Information presented in this report reflect data from phases I and II combined. The NHANES III oversampled Mexican Americans, non-Hispanic blacks, and older adults to ensure weighted, reliable estimates from these groups.

Each interview was conducted in the participant’s home. In addition, a detailed clinical examination was conducted in a mobile examination center. For the purposes of this investigation, we examined the data from the 6596 adults 60 years and older. The interviewing staff consisted of experienced persons, many of Hispanic origin or fluent in both English and Spanish. All staff members attended annual training sessions to ensure maintenance of effective interviewing skills.

Information on self-reported race and ethnicity was used to classify persons as non-Hispanic white, non-Hispanic blacks, or Mexican American (persons of Mexican origin living in the United States). Age was defined as the age, in years, at the time of the household interview, which preceded the medical examination by 2 to 3 weeks.

Hip pain

Participating adults 60 years and older were asked to report “whether they had experienced significant hip pain on most days over the preceding 6 weeks.”

Physical activity assessment

Trained interviewers used a questionnaire to obtain information on leisure time physical activity (LTPA) during the previous month. The questionnaire was adapted from the 1985 National Health Interview Survey (NHIS), which was used to establish baseline estimates for several Healthy People 2000 physical activity objectives. Participants were asked to specify the frequency of participation in LTPA during the previous month for the following activities: jogging or running, riding a bicycle outdoors or an indoor stationary bicycle, swimming, aerobic dancing, other dancing, calisthenics or floor exercises, gardening or yard work, and weight lifting. Four open-ended questions assessed information on physical activities not previously listed. Participants who responded negatively to all LTPA questions, including the 4 open-ended questions, were classified as persons who participate in no LTPA.

Statistical analysis

Statistical analyses were carried out using the SAS⁷ and WesVar PC⁸ software packages. For each survey, sampling weights were calculated that took into account the unequal selection probabilities resulting from the cluster design and from planned oversampling of certain subgroups. All analyses have incorporated sampling weights.

Results

Overall, 14.3% (95% CI, 13.1-15.5) of older US adults reported significant hip pain on most days over the previous 6 weeks. The age-specific and sex-specific prevalence estimates of US adults reporting severe hip pain on most days are shown in Figure 1 Men reported hip pain less frequently (11.9.1%; 95% CI, 10.2-13.7) than women (16.2%; 95% CI, 14.5-17.8). Reports of hip pain are similar in men aged 60 to 70 years, 70 to 80 years, or older than 80 years. The lowest prevalence of hip pain was reported by women aged 60 to 70 years.

The race-specific prevalence of severe hip pain on most days during the previous 6 weeks is shown in Figure 2. Among non-Hispanic white adults aged 60 years and older, 12.4% of men (95% CI, 10.2-14.1) and 16.0% of women (95% CI, 14.0-17.9) reported hip pain. In contrast, 14.8% (95% CI, 11.7-17.9) of non-Hispanic black women reported hip pain, as did 19.3% (95% CI, 14.9-23.7) of Mexican American women. Reports of hip pain among non-Hispanic black men and Mexican American men were similar to those of their non-Hispanic white male counterparts.

Among adults aged 60 years or older, 18.4% (95% CI, 16.5-20.7) of those who participated in no leisure time physical activity reported severe hip pain, while 12.6% (95% CI, 11.7-14.1) of those who did participate in some activity reported such pain.

FIGURE 1
PREVALENCE OF SIGNIFICANT HIP PAIN ON MOST DAYS IN OLDER ADULTS, STRATIFIED BY AGE AND SEX

FIGURE 2
PREVALENCE OF SIGNIFICANT HIP PAIN ON MOST DAYS IN OLDER ADULTS, STRATIFIED BY RACE AND SEX

Discussion

This study offers a current report of the prevalence estimates of significant hip pain among US adults aged 60 years or older. We found that hip pain affects a higher number of older Americans than would be expected from previous studies. For instance, data from NHANES I demonstrated that from 1971 through 1975, 0.7% of patients reported hip pain secondary to OA.¹ NHANES I, however, did not include any subjects older than 74 years. Our study found that 14.3% of US adults aged 60 years or older report hip pain on most days of previous 6 weeks, but we cannot yet determine the etiology of this pain.

There are 2 possible explanations of this apparent increase in hip pain between the time of NHANES I and that of NHANES III. First, the source of most hip pain in the elderly may be nonarticular in nature. Once radiographic data become available for NHANES III, this possibility can be further analyzed. A second explanation may be that the incidence of hip OA is increasing over time. This explanation is plausible, considering both the aging of the population in the United States and the increasing rates of obesity and sedentary lifestyle. Here, too, radiographic data will be useful.

We recently reported that sex, age, and race have a strong impact on knee pain.⁹ In our current study, we found that these are also important determinants of hip pain, but the findings were less striking than with knee pain. Men reported hip pain less often than women (11.9% versus 16.2%, respectively), but men aged 60 to 69 years were as likely to report hip pain as the men 80 years and older. There was no difference between the women in the 70- to 79-year group and those 80 years and older, but the women in both groups reported pain more often than did those aged 60 to 69 years. These findings are similar to those of Frankel and colleagues,¹⁰ who found that hip pain in a United Kingdom population was higher for women (overall prevalence 173 per 1000) than for men (overall prevalence 107 per 1000) at every age group; for both sexes it increased with age. This result is surprising, given that the prevalence of radiographic OA of the hip is more common in men than women aged 55 to 74 years, and in both sexes the prevalence of radiographic OA increases with age.¹ The effects of sex and age on nonarticular hip pain (eg, bursitis) may partially explain the discrepancy. Additionally, there may be a reporting bias because women are more likely to report pain at any joint than are men.¹¹ In one report, women had a higher rate of elective total hip replacement, but men tended to be younger at the time of surgery.¹²

We found no previous reports of the effect of race on the reporting of hip pain. In our study, Mexican American women were more likely to report hip pain (19.3%) than were non-Hispanic black (14.8%) or non-Hispanic white (15.9%) women. Race did not significantly affect reports of hip pain in men. In addition, people who report no leisure time activity were more likely to report hip pain. A sedentary lifestyle is more common among non-Hispanic blacks and Mexican Americans than among whites^13-15 and may partially explain some of the difference in reporting by racial groups.

We acknowledge the limitations of self-report and the cross-sectional nature of this study. Given the cross-sectional nature of this study, we cannot determine the direction of causality; however, it is of concern that significant hip pain may contribute to a more sedentary lifestyle (and its attendant risks). As such, significant hip pain deserves a thorough investigation.

Conclusions

Our study reports on the most current nationally representative data that provide estimates of the prevalence of hip pain in older adults in the United States. Future studies are needed to further identify those at highest risk for hip pain and resultant debility and to determine optimal treatment of hip pain, particularly in Mexican American women.

Acknowledgments

Dr Andersen’s work is supported by Grant No. 97214-G from The John A. Hartford Foundation and by Grant No. DK 53907 from the National Institute of Digestive, Diabetes, and Kidney Diseases.

According to a report by the National Arthritis Data Workgroup based on data from the first National Health and Nutrition Examination Survey (NHANES I), the prevalence of symptomatic hip osteoarthritis (OA) is 0.7% in both adult women and men; 0.5% have moderate or severe symptoms.¹ The incidence of symptomatic hip OA was higher in women than in men, and increased with age in both sexes until age 80 years, with a slight decline beyond that age.² In a study of patients in a health maintenance organization, the incidence of symptomatic hip OA in elderly women was 239 per 100,000 person-years at ages 60 to 69 years, 583 per 100,000 person-years at ages 70 to 79 years, and 441 per 100,000 person-years for women older than 80 years. The corresponding rates for elderly men per 100,000 person-years were 158, 445, and 264, respectively.

In contrast, the prevalence of radiographic hip OA is 3.1%; risk factors for the development of radiographic changes appear different in women and men.³ As has been demonstrated in the knee, radiographic hip OA is more prevalent than symptomatic disease. Only 61% of older individuals with confirmed radiologic hip OA reported pain at the hip, while 11% without radiologic changes of OA had significant hip pain, presumably of nonarticular etiology. Factors that affect patients’ reporting of symptoms of hip pain, both from OA and from nonarticular sources, have not yet been fully elucidated.

We are not aware of any current nationally representative reports on the sex-specific and race-specific prevalence of hip pain in older adults. Therefore, the purpose of our investigation was to examine the age-specific prevalence of hip pain using a current nationally representative sample of older US adults. We will also describe the prevalence of self-reported hip pain among older adults from 2 of the largest minority groups in the United States: non-Hispanic blacks and Mexican Americans.

Methods

Sample design

The Third National Health and Nutrition Examination Survey (NHANES III) was conducted by the Centers for Disease Control and Prevention, National Center for Health Statistics. The plan and operation of NHANES III have been described elsewhere.^5,6 Briefly, the survey was designed to produce a nationally representative sample of the civilian noninstitutionalized US population. One of the main goals of this survey was to estimate the national prevalence of selected health conditions and risk factors.

The NHANES III represents a 6-year study that was conducted from 1988 through 1994 consisting of 2 phases lasting 3 years each: phase I, 1988 through 1992, and phase II, 1991 through 1994. The survey was designed so that each phase was a nationally representative sample. Information presented in this report reflect data from phases I and II combined. The NHANES III oversampled Mexican Americans, non-Hispanic blacks, and older adults to ensure weighted, reliable estimates from these groups.

Each interview was conducted in the participant’s home. In addition, a detailed clinical examination was conducted in a mobile examination center. For the purposes of this investigation, we examined the data from the 6596 adults 60 years and older. The interviewing staff consisted of experienced persons, many of Hispanic origin or fluent in both English and Spanish. All staff members attended annual training sessions to ensure maintenance of effective interviewing skills.

Information on self-reported race and ethnicity was used to classify persons as non-Hispanic white, non-Hispanic blacks, or Mexican American (persons of Mexican origin living in the United States). Age was defined as the age, in years, at the time of the household interview, which preceded the medical examination by 2 to 3 weeks.

Hip pain

Participating adults 60 years and older were asked to report “whether they had experienced significant hip pain on most days over the preceding 6 weeks.”

Physical activity assessment

Trained interviewers used a questionnaire to obtain information on leisure time physical activity (LTPA) during the previous month. The questionnaire was adapted from the 1985 National Health Interview Survey (NHIS), which was used to establish baseline estimates for several Healthy People 2000 physical activity objectives. Participants were asked to specify the frequency of participation in LTPA during the previous month for the following activities: jogging or running, riding a bicycle outdoors or an indoor stationary bicycle, swimming, aerobic dancing, other dancing, calisthenics or floor exercises, gardening or yard work, and weight lifting. Four open-ended questions assessed information on physical activities not previously listed. Participants who responded negatively to all LTPA questions, including the 4 open-ended questions, were classified as persons who participate in no LTPA.

Statistical analysis

Statistical analyses were carried out using the SAS⁷ and WesVar PC⁸ software packages. For each survey, sampling weights were calculated that took into account the unequal selection probabilities resulting from the cluster design and from planned oversampling of certain subgroups. All analyses have incorporated sampling weights.

Results

Overall, 14.3% (95% CI, 13.1-15.5) of older US adults reported significant hip pain on most days over the previous 6 weeks. The age-specific and sex-specific prevalence estimates of US adults reporting severe hip pain on most days are shown in Figure 1 Men reported hip pain less frequently (11.9.1%; 95% CI, 10.2-13.7) than women (16.2%; 95% CI, 14.5-17.8). Reports of hip pain are similar in men aged 60 to 70 years, 70 to 80 years, or older than 80 years. The lowest prevalence of hip pain was reported by women aged 60 to 70 years.

The race-specific prevalence of severe hip pain on most days during the previous 6 weeks is shown in Figure 2. Among non-Hispanic white adults aged 60 years and older, 12.4% of men (95% CI, 10.2-14.1) and 16.0% of women (95% CI, 14.0-17.9) reported hip pain. In contrast, 14.8% (95% CI, 11.7-17.9) of non-Hispanic black women reported hip pain, as did 19.3% (95% CI, 14.9-23.7) of Mexican American women. Reports of hip pain among non-Hispanic black men and Mexican American men were similar to those of their non-Hispanic white male counterparts.

Among adults aged 60 years or older, 18.4% (95% CI, 16.5-20.7) of those who participated in no leisure time physical activity reported severe hip pain, while 12.6% (95% CI, 11.7-14.1) of those who did participate in some activity reported such pain.

FIGURE 1
PREVALENCE OF SIGNIFICANT HIP PAIN ON MOST DAYS IN OLDER ADULTS, STRATIFIED BY AGE AND SEX

FIGURE 2
PREVALENCE OF SIGNIFICANT HIP PAIN ON MOST DAYS IN OLDER ADULTS, STRATIFIED BY RACE AND SEX

Discussion

This study offers a current report of the prevalence estimates of significant hip pain among US adults aged 60 years or older. We found that hip pain affects a higher number of older Americans than would be expected from previous studies. For instance, data from NHANES I demonstrated that from 1971 through 1975, 0.7% of patients reported hip pain secondary to OA.¹ NHANES I, however, did not include any subjects older than 74 years. Our study found that 14.3% of US adults aged 60 years or older report hip pain on most days of previous 6 weeks, but we cannot yet determine the etiology of this pain.

There are 2 possible explanations of this apparent increase in hip pain between the time of NHANES I and that of NHANES III. First, the source of most hip pain in the elderly may be nonarticular in nature. Once radiographic data become available for NHANES III, this possibility can be further analyzed. A second explanation may be that the incidence of hip OA is increasing over time. This explanation is plausible, considering both the aging of the population in the United States and the increasing rates of obesity and sedentary lifestyle. Here, too, radiographic data will be useful.

We recently reported that sex, age, and race have a strong impact on knee pain.⁹ In our current study, we found that these are also important determinants of hip pain, but the findings were less striking than with knee pain. Men reported hip pain less often than women (11.9% versus 16.2%, respectively), but men aged 60 to 69 years were as likely to report hip pain as the men 80 years and older. There was no difference between the women in the 70- to 79-year group and those 80 years and older, but the women in both groups reported pain more often than did those aged 60 to 69 years. These findings are similar to those of Frankel and colleagues,¹⁰ who found that hip pain in a United Kingdom population was higher for women (overall prevalence 173 per 1000) than for men (overall prevalence 107 per 1000) at every age group; for both sexes it increased with age. This result is surprising, given that the prevalence of radiographic OA of the hip is more common in men than women aged 55 to 74 years, and in both sexes the prevalence of radiographic OA increases with age.¹ The effects of sex and age on nonarticular hip pain (eg, bursitis) may partially explain the discrepancy. Additionally, there may be a reporting bias because women are more likely to report pain at any joint than are men.¹¹ In one report, women had a higher rate of elective total hip replacement, but men tended to be younger at the time of surgery.¹²

We found no previous reports of the effect of race on the reporting of hip pain. In our study, Mexican American women were more likely to report hip pain (19.3%) than were non-Hispanic black (14.8%) or non-Hispanic white (15.9%) women. Race did not significantly affect reports of hip pain in men. In addition, people who report no leisure time activity were more likely to report hip pain. A sedentary lifestyle is more common among non-Hispanic blacks and Mexican Americans than among whites^13-15 and may partially explain some of the difference in reporting by racial groups.

We acknowledge the limitations of self-report and the cross-sectional nature of this study. Given the cross-sectional nature of this study, we cannot determine the direction of causality; however, it is of concern that significant hip pain may contribute to a more sedentary lifestyle (and its attendant risks). As such, significant hip pain deserves a thorough investigation.

Conclusions

Our study reports on the most current nationally representative data that provide estimates of the prevalence of hip pain in older adults in the United States. Future studies are needed to further identify those at highest risk for hip pain and resultant debility and to determine optimal treatment of hip pain, particularly in Mexican American women.

Acknowledgments

Dr Andersen’s work is supported by Grant No. 97214-G from The John A. Hartford Foundation and by Grant No. DK 53907 from the National Institute of Digestive, Diabetes, and Kidney Diseases.

According to a report by the National Arthritis Data Workgroup based on data from the first National Health and Nutrition Examination Survey (NHANES I), the prevalence of symptomatic hip osteoarthritis (OA) is 0.7% in both adult women and men; 0.5% have moderate or severe symptoms.¹ The incidence of symptomatic hip OA was higher in women than in men, and increased with age in both sexes until age 80 years, with a slight decline beyond that age.² In a study of patients in a health maintenance organization, the incidence of symptomatic hip OA in elderly women was 239 per 100,000 person-years at ages 60 to 69 years, 583 per 100,000 person-years at ages 70 to 79 years, and 441 per 100,000 person-years for women older than 80 years. The corresponding rates for elderly men per 100,000 person-years were 158, 445, and 264, respectively.

In contrast, the prevalence of radiographic hip OA is 3.1%; risk factors for the development of radiographic changes appear different in women and men.³ As has been demonstrated in the knee, radiographic hip OA is more prevalent than symptomatic disease. Only 61% of older individuals with confirmed radiologic hip OA reported pain at the hip, while 11% without radiologic changes of OA had significant hip pain, presumably of nonarticular etiology. Factors that affect patients’ reporting of symptoms of hip pain, both from OA and from nonarticular sources, have not yet been fully elucidated.

We are not aware of any current nationally representative reports on the sex-specific and race-specific prevalence of hip pain in older adults. Therefore, the purpose of our investigation was to examine the age-specific prevalence of hip pain using a current nationally representative sample of older US adults. We will also describe the prevalence of self-reported hip pain among older adults from 2 of the largest minority groups in the United States: non-Hispanic blacks and Mexican Americans.

Methods

Sample design

The Third National Health and Nutrition Examination Survey (NHANES III) was conducted by the Centers for Disease Control and Prevention, National Center for Health Statistics. The plan and operation of NHANES III have been described elsewhere.^5,6 Briefly, the survey was designed to produce a nationally representative sample of the civilian noninstitutionalized US population. One of the main goals of this survey was to estimate the national prevalence of selected health conditions and risk factors.

The NHANES III represents a 6-year study that was conducted from 1988 through 1994 consisting of 2 phases lasting 3 years each: phase I, 1988 through 1992, and phase II, 1991 through 1994. The survey was designed so that each phase was a nationally representative sample. Information presented in this report reflect data from phases I and II combined. The NHANES III oversampled Mexican Americans, non-Hispanic blacks, and older adults to ensure weighted, reliable estimates from these groups.

Each interview was conducted in the participant’s home. In addition, a detailed clinical examination was conducted in a mobile examination center. For the purposes of this investigation, we examined the data from the 6596 adults 60 years and older. The interviewing staff consisted of experienced persons, many of Hispanic origin or fluent in both English and Spanish. All staff members attended annual training sessions to ensure maintenance of effective interviewing skills.

Information on self-reported race and ethnicity was used to classify persons as non-Hispanic white, non-Hispanic blacks, or Mexican American (persons of Mexican origin living in the United States). Age was defined as the age, in years, at the time of the household interview, which preceded the medical examination by 2 to 3 weeks.

Hip pain

Participating adults 60 years and older were asked to report “whether they had experienced significant hip pain on most days over the preceding 6 weeks.”

Physical activity assessment

Trained interviewers used a questionnaire to obtain information on leisure time physical activity (LTPA) during the previous month. The questionnaire was adapted from the 1985 National Health Interview Survey (NHIS), which was used to establish baseline estimates for several Healthy People 2000 physical activity objectives. Participants were asked to specify the frequency of participation in LTPA during the previous month for the following activities: jogging or running, riding a bicycle outdoors or an indoor stationary bicycle, swimming, aerobic dancing, other dancing, calisthenics or floor exercises, gardening or yard work, and weight lifting. Four open-ended questions assessed information on physical activities not previously listed. Participants who responded negatively to all LTPA questions, including the 4 open-ended questions, were classified as persons who participate in no LTPA.

Statistical analysis

Statistical analyses were carried out using the SAS⁷ and WesVar PC⁸ software packages. For each survey, sampling weights were calculated that took into account the unequal selection probabilities resulting from the cluster design and from planned oversampling of certain subgroups. All analyses have incorporated sampling weights.

Results

Overall, 14.3% (95% CI, 13.1-15.5) of older US adults reported significant hip pain on most days over the previous 6 weeks. The age-specific and sex-specific prevalence estimates of US adults reporting severe hip pain on most days are shown in Figure 1 Men reported hip pain less frequently (11.9.1%; 95% CI, 10.2-13.7) than women (16.2%; 95% CI, 14.5-17.8). Reports of hip pain are similar in men aged 60 to 70 years, 70 to 80 years, or older than 80 years. The lowest prevalence of hip pain was reported by women aged 60 to 70 years.

The race-specific prevalence of severe hip pain on most days during the previous 6 weeks is shown in Figure 2. Among non-Hispanic white adults aged 60 years and older, 12.4% of men (95% CI, 10.2-14.1) and 16.0% of women (95% CI, 14.0-17.9) reported hip pain. In contrast, 14.8% (95% CI, 11.7-17.9) of non-Hispanic black women reported hip pain, as did 19.3% (95% CI, 14.9-23.7) of Mexican American women. Reports of hip pain among non-Hispanic black men and Mexican American men were similar to those of their non-Hispanic white male counterparts.

Among adults aged 60 years or older, 18.4% (95% CI, 16.5-20.7) of those who participated in no leisure time physical activity reported severe hip pain, while 12.6% (95% CI, 11.7-14.1) of those who did participate in some activity reported such pain.

FIGURE 1
PREVALENCE OF SIGNIFICANT HIP PAIN ON MOST DAYS IN OLDER ADULTS, STRATIFIED BY AGE AND SEX

FIGURE 2
PREVALENCE OF SIGNIFICANT HIP PAIN ON MOST DAYS IN OLDER ADULTS, STRATIFIED BY RACE AND SEX

Discussion

This study offers a current report of the prevalence estimates of significant hip pain among US adults aged 60 years or older. We found that hip pain affects a higher number of older Americans than would be expected from previous studies. For instance, data from NHANES I demonstrated that from 1971 through 1975, 0.7% of patients reported hip pain secondary to OA.¹ NHANES I, however, did not include any subjects older than 74 years. Our study found that 14.3% of US adults aged 60 years or older report hip pain on most days of previous 6 weeks, but we cannot yet determine the etiology of this pain.

There are 2 possible explanations of this apparent increase in hip pain between the time of NHANES I and that of NHANES III. First, the source of most hip pain in the elderly may be nonarticular in nature. Once radiographic data become available for NHANES III, this possibility can be further analyzed. A second explanation may be that the incidence of hip OA is increasing over time. This explanation is plausible, considering both the aging of the population in the United States and the increasing rates of obesity and sedentary lifestyle. Here, too, radiographic data will be useful.

We recently reported that sex, age, and race have a strong impact on knee pain.⁹ In our current study, we found that these are also important determinants of hip pain, but the findings were less striking than with knee pain. Men reported hip pain less often than women (11.9% versus 16.2%, respectively), but men aged 60 to 69 years were as likely to report hip pain as the men 80 years and older. There was no difference between the women in the 70- to 79-year group and those 80 years and older, but the women in both groups reported pain more often than did those aged 60 to 69 years. These findings are similar to those of Frankel and colleagues,¹⁰ who found that hip pain in a United Kingdom population was higher for women (overall prevalence 173 per 1000) than for men (overall prevalence 107 per 1000) at every age group; for both sexes it increased with age. This result is surprising, given that the prevalence of radiographic OA of the hip is more common in men than women aged 55 to 74 years, and in both sexes the prevalence of radiographic OA increases with age.¹ The effects of sex and age on nonarticular hip pain (eg, bursitis) may partially explain the discrepancy. Additionally, there may be a reporting bias because women are more likely to report pain at any joint than are men.¹¹ In one report, women had a higher rate of elective total hip replacement, but men tended to be younger at the time of surgery.¹²

We found no previous reports of the effect of race on the reporting of hip pain. In our study, Mexican American women were more likely to report hip pain (19.3%) than were non-Hispanic black (14.8%) or non-Hispanic white (15.9%) women. Race did not significantly affect reports of hip pain in men. In addition, people who report no leisure time activity were more likely to report hip pain. A sedentary lifestyle is more common among non-Hispanic blacks and Mexican Americans than among whites^13-15 and may partially explain some of the difference in reporting by racial groups.

We acknowledge the limitations of self-report and the cross-sectional nature of this study. Given the cross-sectional nature of this study, we cannot determine the direction of causality; however, it is of concern that significant hip pain may contribute to a more sedentary lifestyle (and its attendant risks). As such, significant hip pain deserves a thorough investigation.

Conclusions

Our study reports on the most current nationally representative data that provide estimates of the prevalence of hip pain in older adults in the United States. Future studies are needed to further identify those at highest risk for hip pain and resultant debility and to determine optimal treatment of hip pain, particularly in Mexican American women.

Acknowledgments

Dr Andersen’s work is supported by Grant No. 97214-G from The John A. Hartford Foundation and by Grant No. DK 53907 from the National Institute of Digestive, Diabetes, and Kidney Diseases.

Infections of the nasal passages are very common¹ and among the most frequent reasons for the prescription of antibiotics.^2,3 Such infections comprise diagnoses that include upper respiratory tract infection (URTI), rhinitis, rhinopharyngitis, and rhinosinusitis, which are very difficult to distinguish because of the lack of specific clinical features or simple office-based diagnostic tests.^4-7 These diagnostic difficulties probably explain why it remains unclear whether and when antibiotics should be used for such patients in clinical practice.

Although evidence shows that a small minority of patients benefit from antibiotic therapy, these patients are extremely difficult to recognize or identify. Three meta-analyses^8-10 on the effect of antibiotics in rhinosinusitis and 5 of 6 recent trials investigating the effect of antibiotics in rhinosinusitis,^11-13 rhinitis, ¹⁴ and bacterial rhinopharyngitis¹⁵ almost exclusively studied patients with a diagnosis established by laboratory or imaging investigation. As a result, implementing the findings is difficult in daily practice, where radiologic or laboratory tests are not obtained for most patients with respiratory infections. Only 1 of the 6 trials¹⁶ included patients with a set of clinical symptoms indicating rhinosinusitis. Because inclusion criteria were rather stringent, however, findings are applicable only to a small group of patients.

The purpose of this trial was to investigate the benefits of antibiotic therapy in a larger group of patients with nose or sinus infections, thereby making the results more widely applicable. Accordingly, we conducted a randomized, double-blind, placebo-controlled trial comparing the effect of amoxicillin with that of placebo in family practice patients with an acute upper respiratory tract infection and presenting with purulent rhinorrhea. Purulent rhinorrhea was chosen as the minimal criterion because it is the symptom most consistently associated with rhinosinusitis in diagnostic studies^5,17-21 and because its presence often leads family physicians (FPs) to prescribe antibiotics.^23-26 The trial was designed as a pragmatic effectiveness trial. Patient inclusion and evaluation were defined on a purely clinical basis to maximize relevance for routine daily practice.

Methods

Study population

Between October 1998 and December 1999, 69 FPs in Flanders, Belgium, agreed to enroll patients meeting the following inclusion criteria: age 12 years or older, presenting with a respiratory tract infection, and having purulent rhinorrhea. Exclusion criteria were allergy to penicillin or ampicillin; having received antibiotic therapy within the previous week; complaints lasting for more than 30 days; abnormality on clinical chest examination; complications of sinusitis (facial edema or cellulitis; orbital, visual, meningeal, or cerebral signs)²⁷; pregnancy or lactation; comorbidity that might impair immune competence; and inability to follow the protocol because of language or mental problems. The Ethics Committee of the Ghent University Hospital (GUH) approved the study. All patients (or their guardians, for those younger than 16 years of age) gave written informed consent.

Treatment assignment and masking

In this double-blind trial, patients were assigned via a computer-generated random number list to receive 500 mg amoxicillin 3 times a day or placebo for 10 days. The trial medication was supplied in numbered uniform cardboard boxes, each containing 30 capsules of the same size, color, and shape for active and placebo treatment. The randomization list, kept at the pharmacy of GUH, was accessible to the participating FPs only in case of a serious adverse event.

To assess the effectiveness of masking, patients and their FPs guessed the treatment group at 10-day follow-up. Data were encoded and entered without knowledge of treatment allocation. Compliance was assessed by counting leftover medication. All patients were allowed to use xylometazoline 1% nose drops and paracetamol or ibuprofen to alleviate symptoms; these data were registered.

Assessment of potential recruitment bias caused by exclusion

First, we compared the characteristics of patients enrolled by high-recruiting FPs (at least 14 patients recruited) with those of patients from low recruiters (at most 5 patients recruited). Second, we asked all participating FPs to complete a short questionnaire over a 6-week period on all patients eligible for the trial but not included in it (sex, age, body temperature, severity of nasal discharge and pain, reason for non-recruitment). Third, to estimate the proportion of sinusitis cases among included patients, all patients were invited for an optional radiologic examination of the maxillary sinuses (single Waters view).²⁸ Radiographs were taken in the nearest radiology unit, collected centrally, and evaluated by a radiologist of the GUH who specialized in the ear, nose, and throat.

Baseline measurements

Randomized patients completed an extensive questionnaire and were physically examined by their FP. To evaluate the symptoms, we used the 20 items of the sinonasal outcome test (SNOT-20)^29,30 supplemented by 3 questions about pain. Symptoms were scored on a 6-category (0-5) Likert scale. Patients were also asked to indicate which of their symptoms (no more than 5) were most troublesome.

Follow-up

During 10 days of treatment, all patients recorded their daily drug intake (trial medication and symptomatic medication); their general feeling of illness; the presence of nasal discharge, pain, and cough; body temperature; the occurrence of presumed adverse drug effects; and absence from work or school. On day 10 they underwent a second physical examination and completed the symptom questionnaire again. In case of insufficient recovery, the FP was then at liberty to prescribe an open antibiotic course (we recommended amoxicillin clavulanate) without revealing the previous treatment phase. Patients who had recovered on day 10 did not have to return on day 15. Any patient with poor recovery on day 10 was asked, regardless of open antibiotic treatment, to continue writing in the diary and to come back on day 15 if complaints were still present.

The 2 primary endpoints were the therapy success rate on day 10 and the duration of general illness, pain, and purulent rhinorrhea as recorded in the diary. Treatment was considered successful when all symptoms that the patient had included in the list of “most important item affecting my health” scored 0 (absent) or 1 (very mildly present) after 10 days of treatment. Secondary endpoints were the mean change in severity score between day 1 and 10 on the various symptoms, incidence of unfavorable evolution, incidence of side effects, intake of analgesics, and duration of sick leave. The number of patients needed to demonstrate a difference in the therapy success rate of 15% at day 10 (α = 0.05, β = 0.20) was 168 per treatment group.³¹ This determination assumed a success rate of 50% in the placebo group.^11,12

Statistics

Data were analyzed with SPSS-7. Differences in proportions are presented as relative risks with 95% confidence intervals and tested by chi-square test. The duration of symptoms is presented by Kaplan-Meier survival plots. Differences in duration are tested by the log rank test. Other continuous variables are tested by Student’s t test or the nonparametric Mann-Whitney U test.

Results

Participant flow and follow-up

Of 416 patients enrolled in the study, 8 were excluded after randomization. Of the 408 patients remaining, 202 received amoxicillin and 206 placebo; 34 patients (8%) withdrew from the trial. Their personal characteristics and clinical conditions at inclusion were not different from those of patients with follow-up. Figure 1 lists reasons for exclusion or withdrawal. The treatment code was broken once for a suspected allergic reaction and once because of an exacerbation of symptoms. In accordance with the intention-to-treat principle, all enrolled patients were included in the analyses in the groups to which they were originally randomized. Patients who had withdrawn because of side effects were also included in the analysis of side effects.

Complete or partial follow-up data were obtained for 374 patients (90%) after 10 days (mean 10.3 days, standard deviation 1.44): 334 patients completed the questionnaire, 348 returned the diary, and 338 underwent a physical examination. In 265 (71%) patients, data (questionnaire, diary, and physical examination) were complete; in 109 (29%), data at day 10 were partly missing. The two treatment groups were very similar in terms of sex, age, duration of preinclusion complaints, and frequency of various physical signs and symptoms (Table 1).*

TABLE 1
BASELINE CHARACTERISTICS

FIGURE 1
PATIENTS’ PROGRESS THROUGH THE TRIAL

Primary Outcomes

Of the 374 patients with follow-up data on day 10, 334 completed the symptom questionnaire twice. Treatment was successful—defined as a score of 0 (absent) or 1 (very mildly present) for all symptoms that had been included as “the most important item affecting my health”—in 35% of patients in the amoxicillin group (59/170) and 29% in the placebo group (47/164) (Table 2). Relative risk of success was 1.14 (95% CI, 0.92-1.42, P = .24): more patients were cured in the amoxicillin group, but this difference was not statistically significant.

In 82 (19.7%) of the 416 randomized patients (37 amoxicillin, 45 placebo), data on this main outcome are missing. In 40 of these 82 patients, follow-up data are available from the diary (n = 38) or physical examination (n = 2). According to these data, in 13/17 of the amoxicillin group and 11/23 of the placebo group the outcome was favorable: in the diary, the patient reports feeling “well” again at day 10 or sooner, or on physical examination, all signs of respiratory infection have cleared). Eight patients withdrew for clinical exacerbation and 2 patients after full recovery. Adding the 50 patients with a known course of illness to those in the treatment and result groups does not alter the overall result (RR 1.20, 95% CI, 0.98-1.47, P = .08). Furthermore, when considering the 24 nonexcluded patients (13 amoxicillin, 11 placebo) with total lack of follow-up in their allocated treatment group, first as treatment failures (RR 1.18, 95% CI, 0.97-1.44, P = .11) and then as successes (1.20, 95% CI, 0.99-1.46, P = .07), the result also remains the same. Regarding the success rate from the complete diary data (n = 348) and the results of physical examinations (n = 338) (Table 3), we find no significant difference between treatment groups.

Duration of purulent rhinorrhea was significantly shorter in the amoxicillin group than in the placebo group (75% of patients were free of purulent rhinorrhea after 9 days versus after 14 days in the placebo group, log rank P = .007). There is no difference between treatment groups in the duration of general illness or pain (Figure 2).

TABLE 2
MAIN OUTCOME: RATE OF TREATMENT SUCCESS AT 10-DAY FOLLOW-UP

TABLE 3
MEAN SYMPTOM CHANGE BETWEEN BASELINE AND 10-DAY FOLLOW-UP

FIGURE 2
DURATION OF ILLNESS, PAIN, AND PURULENT RHINORRHEA BETWEEN TREATMENT GROUPS

Secondary outcomes

The mean score reduction on the symptom “thick nasal discharge” between day 1 and day 10 is significantly larger in the amoxicillin group than in the placebo group (2.2 vs 1.5, Student’s t test: P <.0001) (Table 3). There is no significant difference in change for any other symptom. Seven patients in the placebo group (3.4%) withdrew before day 10 because of exacerbation of symptoms versus 1 patient (0.5%) in the amoxicillin group (RR 0.25, 95% CI, 0.04-1.56, P = .07). All 8 patients recovered after starting open antibiotic therapy and had no complications or referrals.

The chance of receiving open antibiotic treatment at day 10 follow-up (n = 34: 19 placebo, 15 amoxicillin) or of having to return because of persistent complaints at day 15 (n = 73: 41 placebo, 32 amoxicillin) was not significantly different between the treatment groups (chi-squared test: P = .46 and P = .26, respectively). Diarrhea was more frequent in the amoxicillin group (29% vs. 19%, RR 1.28, CI 1.05-1.57, P = .02). There was no difference in incidence of skin rash, abdominal pain, or vomiting. Absence from work or school was comparable in both treatment groups (RR 0.95, 95% CI, 0.86-1.05, P = .34). Patients in the amoxicillin group took an analgesic an average of 5 times, mainly in the first days of treatment, compared with 4 for the placebo group (Mann-Whitney U test, P = .24).

Other results

The lack of correlation between the estimated and actual treatment demonstrates that masking was maintained. Compliance was good in both groups: 89% of patients in the amoxicillin group and 91% of those in the placebo group took at least 25 of 30 capsules.

Patients from low recruiters were not significantly different from patients enrolled by high recruiters. Included patients had slightly more complaints of pain (58% vs 50%, RR 1.20, CI 1.02-1.42, P = .03) than the 332 eligible but excluded patients registered during the 6-week period. The most frequent reasons for exclusion were the presence of an exclusion criterion (22%), the patient’s refusal to participate (16%), the patient’s request for antibiotic therapy (14%), and lack of time by the FP (10%). Of the 292 patients who agreed to undergo a radiologic examination, about two thirds had abnormalities of the maxillary sinuses.

Discussion

This study produced 3 important findings. First, we found that patients consulting their FP for acute URTI with purulent rhinorrhea do not experience any important benefit from amoxicillin therapy. With treatment, the purulent rhinorrhea disappears more quickly, but this seems to be of little importance in relation to a general recovery. Moreover, amoxicillin therapy increases the risk of diarrhea. We further found that with or without amoxicillin, complaints last long: after 10 days, two thirds of patients still had complaints and about half of the patients still felt ill. The natural course to recovery takes a long time and is not influenced by taking amoxicillin. Finally, we observed that failure to prescribe antibiotics is safe. The placebo group had no complications. A small number of exacerbations occurred, but these responded swiftly to a course of amoxicillin-clavulanate.

To our knowledge, this is the first time that the effect of an antibiotic in adult patients presenting with acute purulent rhinorrhea (but with an otherwise unspecified diagnosis) has been investigated in a randomized, placebo-controlled trial. This trial is in line with a number of other family practice-based pragmatic trials in which patients were included on the basis of respiratory symptoms instead of by diagnosis^16,32-37 and in which the emphasis was on practical relevance rather than on diagnostic accuracy.

Since 1995, 6 randomized clinical trials of high methodologic quality^11-16 have studied the efficacy of antibiotics in general practice patients suffering from various acute infections of the nasal passages and usually presenting with purulent rhinorrhea. In 3 of these trials, no beneficial effect of antibiotics was found. Study populations consisted, respectively, of patients with a set of clinical symptoms (including purulent rhinorrhea) indicating rhinosinusitis¹⁶; patients with clinical suspicion of rhinosinusitis plus sinus abnormalities on conventional radiology¹¹; and patients with clinical suspicion of sinusitis but without the radiologic signs.¹⁴ In the 3 other trials, treatment was (more or less) effective. Included were patients with clinical suspicion of sinusitis and abnormalities on CT scan,¹² patients with unilateral facial pain and elevated C-reactive protein levels or erythrocyte sedimentation rate,¹³ and patients with rhinopharyngitis and positive bacteriologic cultures of nasopharyngeal secretions.¹⁵ These trials show that antibiotics are efficacious in some patients. In our trial, which probably included a mix of all these populations, we also found more patients in the amoxicillin group to be symptom free after 10 days. Despite a fairly large sample size, however, this difference was too small (less than 15%) to be statistically significant.