Antibiotic Prescribing for Cough and Symptoms of Respiratory Tract Infection: Title and subTitle BreakDo the Right Thing

Cough is the third most common reason that patients present to physicians for care and was responsible for 3.2% of all outpatient visits in the United States in 2002.¹ Of these visits, more than 30 million were made for a chief complaint of cough, at least 10 million of which were made by otherwise healthy adults diagnosed with acute bronchitis; most of these patients received an antibiotic prescription, often for a broad-spectrum macrolide or quinolone costing US $50 to $100.^{1 - 4}

Although acute cough in healthy adults is common and treatment is a significant expense, the evidence on which treatment decisions are based is relatively limited; until recently, it consisted primarily of 9 studies with 750 patients, many performed a decade or more ago.⁵ A systematic review of these studies suggests a small benefit for antibiotics, with cough resolving about a half day sooner, balanced by a somewhat increased risk of adverse effects due to the antibiotic.⁵ However, that study⁵ and another meta-analysis with similar findings⁶ are limited by the small size and design limitations of the included studies and the possibility of publication bias in favor of clinical trials with positive findings.

In this issue of JAMA, the study by Little and colleagues⁷ more than doubled the size of this evidence base. The investigators enrolled more than 800 adults and children who presented to their primary care physician with cough and at least 1 symptom referrable to the lower respiratory tract (colored sputum, chest pain, dyspnea, or wheezing). Patients with chronic lung disease and those with suspected pneumonia were excluded. Participants were randomly assigned, with allocation concealed, to receive immediate antibiotics (250 mg of amoxicillin administered 3 times daily or erythromycin, if allergic), no antibiotics, or an offer of delayed antibiotics if symptoms had not resolved after 10 days. The delayed prescription could be picked up from the receptionist without having to actually visit the physician. The patients in this study were typical of those diagnosed with acute bronchitis in primary care practice: 2 in 3 patients had a fever and more than 40% reported production of green sputum. Patients monitored their symptoms using diaries for 3 weeks.

The findings of this study will probably surprise many clinicians and most patients. There was no significant difference in any of the primary outcomes (duration of cough until “very little problem,” duration of moderately bad cough, and mean severity of symptoms) between patients receiving antibiotics and those receiving placebo. The duration of a secondary outcome, moderately bad symptoms, was shorter in the immediate antibiotic group, but only by approximately 1 day. Children and patients with colored sputum did not benefit more than other groups and elderly persons were actually less likely to benefit from antibiotics. Not providing antibiotics increased the likelihood that patients would return for reevaluation during the next month (1 in 5 patients vs 1 in 9) and decreased the likelihood that patients would be very satisfied, but only slightly.

Beyond the failure to demonstrate a clinically meaningful benefit for immediate antibiotics, the study provides the most reliable information to date on the natural history of acute bronchitis. Patients had been symptomatic a mean of 9 days before they presented for care and experienced cough for 12 more days after visiting their physician. This is helpful information, as it allows physicians to set more appropriate expectations for patients regarding the duration of acute bronchitis symptoms. Moreover, these data challenge the “dictum” that the patient’s cough will be better in a week if he/she visits a physician or 7 days if the patient does not.

Little et al had previously studied the medicalization of illness in a study of sore throat, demonstrating that prescription of an antibiotic makes it more likely that patients will return for reevaluation and for subsequent similar episodes.⁸ That work is echoed in the current study, again finding that the decision to provide an antibiotic sends a powerful signal to patients. Those patients who received an antibiotic were much more likely to believe in the efficacy of antibiotics for acute bronchitis (75% vs 40% in the delayed group and 47% in the no antibiotic group), while providing an education leaflet had no effect on these beliefs. Apparently actions speak louder than words (or educational leaflets).

Defenders of the antibiotic dispensing status quo might point to the choice of antibiotic or the relatively long duration of symptoms before seeking care in the current study. One question is whether a broader spectrum antibiotic, administered sooner in the course of the illness, would be more effective. In a randomized trial,⁹ 220 adults with a presenting complaint of cough with or without fever and a median symptom duration of only 4 to 5 days (adults with chronic lung disease or evidence of pneumonia were excluded) were randomized in a blinded fashion with allocation concealed to a 5-day course of azithromycin or low-dose vitamin C. After 1 week, 89% in each group had returned to their usual activities and there were no significant differences between groups for any of the prespecified clinical outcomes.⁹ Perhaps it is not a coincidence that manufacturer-sponsored placebo-controlled trials of newer antibiotics for acute bronchitis in healthy adults are absent from the literature.

Another potential criticism of the study by Little et al⁷ is failure to blind patients. However, the goal of this study was not only to explore the effect of antibiotics but also to explore the behavior of patients who were treated differently. That is only possible in a pragmatic trial in which patients know whether or not they received an antibiotic. Also, failure to conceal allocation properly is a more important source of bias than failure to blind,¹⁰ and the authors concealed allocation properly in the current study. If anything, lack of blinding would be expected to favor the immediate antibiotic group, given the strongly held belief of many patients in the efficacy of antibiotics for bronchitis. Thus, failure to demonstrate an improvement in the primary outcomes with antibiotics despite this inherent bias in favor of treatment is further evidence against the efficacy of antibiotics in acute bronchitis.

Physicians may argue that a liberal approach to the use of antibiotics for acute bronchitis is more efficient, helps maintain patient satisfaction, and prevents rare complications. However, these arguments are not compelling. First, although it may be more efficient to prescribe an antibiotic than to explain why one is not necessary, the time investment is usually short and the dialogue with the patient is important. Moreover, because the 3% of visits for cough translates into about 1 patient per day (perhaps 2 or 3 during certain times of the year), this hardly seems like an insurmountable barrier. Surely, higher-quality care is worth spending this time.

Second, Little et al did find a small difference in patient satisfaction: 86% of patients were very satisfied or extremely satisfied if they received an immediate antibiotic vs 77% who received a delayed antibiotic and 72% who received no antibiotic. Hamm et al¹¹ have shown that although two thirds of patients expect an antibiotic, and physicians believe that two thirds of patients expect an antibiotic, clinicians were not good at guessing which patients wanted antibiotics and which patients did not. In fact, satisfaction depended on the quality of the physician’s explanation and the care taken with the examination rather than whether the patient received an antibiotic (even if one was expected).¹¹ The primary goal should be to improve patient outcomes rather than to improve patient satisfaction.

And third, complications among patients similar to those in the study by Little et al are rare. In the current study, 1 patient in the no antibiotic group was later diagnosed with pneumonia, treated with antibiotics, and did well. It is not clear whether treating this patient earlier with antibiotics would have prevented the pneumonia; perhaps it would have just led to the development of resistant organisms. Overuse of antibiotics also leads to its own rare but life-threatening complications (eg, Clostridium difficile enterocolitis) and is a major contributor to increasing resistance to antibiotics in the community. Reduced use of antibiotics can reverse this increase in resistance.¹² Although there is a trend toward less use of antibiotics in recent years, when antibiotics are used they are more likely to be broad spectrum. Furthermore, large prospective studies have shown that patients with cough who have a temperature of less than 38.4°C, who are not tachypneic, and who do not have rales or decreased breath sounds have a very low risk of pneumonia (<0.5% given a 5% pretest probability).¹³

However, there may be a small subset of patients with cough who do not have pneumonia but who do benefit from antibiotics, an “antibiotic-sensitive” group of patients. Some hints of that possibility are evident in the small reduction in the duration of more severe symptoms in the current study. However, current diagnostic labels are inadequate for identifying these patients, as are symptoms such as fever or sputum production. Little et al have identified predictors of antibiotic responsiveness in patients with otitis media. Further work is needed to see if a similar small subset exists among patients who are now labeled as having acute bronchitis.¹⁴

What can a clinician gain from the study by Little et al? First, antibiotics provide little or no benefit for patients with cough that is accompanied by lower respiratory tract symptoms provided the patient does not have pneumonia. This is true even for patients who are older and who have a low-grade fever or green sputum production. Second, physicians should be sure to inform patients that whether or not they take antibiotics, they can expect that a cough will last about 3 weeks, and that for at least 25% of patients it will last nearly a month. Third, by prescribing antibiotics it is clear that clinicians are training patients to expect these drugs. Physicians who feel compelled to give an antibiotic should at least use the tactic of delayed prescriptions to mitigate the effects of this prescribing error. Fourth, the patient’s agenda for the visit must be addressed. Physicians should be sure to answer their questions, provide symptomatic care, and consider an inhaled β-agonist if there is evidence of bronchospasm or a history of asthma.¹⁵ Verbally itemizing a checklist of potential bacterial infections for the patient during the physical examination (“Well, your ears look good . . . your lungs are clear so you don’t have a pneumonia . . . ”) may be a way to save time, improve communication, and reassure the patient. For patients in whom pneumonia is suspected, appropriate treatment must be promptly initiated. If the clinician does not suspect pneumonia, the patient should be informed of that assessment, but should be advised to return if symptoms progress. However, physicians should not give antibiotics to 100 patients on the chance that 1 patient may develop pneumonia at some point in the future.¹³

In the current market-based health care system, it is tempting to confuse patient satisfaction with better outcomes, and to confuse more care with better quality care. Physicians have a duty to listen carefully to patients’ symptoms, to examine them carefully, and to take the time to explain their illness to them. However, physicians have no duty to fulfill patients’ expectations for inappropriate care, such as prescribing antibiotics when they are not indicated, and must be mindful of the duty to the larger community that suffers financially and medically when antibiotics are overused.