Antibiotics are commonly prescribed empirically for lower respiratory infections. Infections of the airway mucosa are much more common than pneumonia and the illness they cause is less severe because the infection is superficial, most of the bacteria being found associated with mucus in the lumen. In many cases the infection will resolve spontaneously without antibiotic treatment. Most adult patients are experiencing an exacerbation of chronic lung disease, particularly chronic obstructive pulmonary disease (COPD), when neutrophilic inflammation in response to bacterial infection leads to increased sputum volume and viscosity, and breathlessness due to airflow obstruction. In these circumstances, bacteria are cultured from sputum in about half of the cases which means that, in some of the others, accepting that sputum culture is not a sensitive investigation, antibiotics are given unnecessarily. Antibiotics are essential when a patient with severe COPD presents with purulent sputum and systemic symptoms of infection, but they are often given either to speed up recovery from a bacterial infection that might be expected to resolve spontaneously following a successful host inflammatory response, or in a defensive manner to avoid the risk of airway infection progressing to pneumonia and causing deterioration in a more compromised patient whose host defences are more seriously impaired.

In recent years attention has rightly focused on trying to define which patients benefit from antibiotic treatment, and those in whom antibiotics can be avoided. The size of the likely benefit has to be taken into consideration when making a decision about antibiotic treatment because of the rise in antibiotic resistance among common respiratory pathogens,1 which is directly related to the volume of antibiotic consumption in a community.2 Sputum purulence has proved to be a reliable signal of bacterial infection3 and, together with the symptoms of increased sputum volume and breathlessness, has been used in COPD guidelines for antibiotic use. These cardinal symptoms were used in the study performed by Anthonisen et al.4 Antibiotics or placebo were given in a randomised, double-blind, crossover fashion for COPD exacerbations. Sputum cultures were not performed, so the outcome of the study cannot be related to microbiology. There was a significant benefit from antibiotics that was largely accounted for by patients who had all three symptoms (type 1 exacerbations), whereas there was no significant difference between antibiotic and placebo in patients with only one of the symptoms. However, in the type 1 exacerbations, 43% of patients recovered in the placebo group within 21 days. A recent Cochrane review5 of antibiotics and COPD exacerbations showed that antibiotics reduce the risk of treatment failure (relative risk ratio 0.47) and the number of patients that needed to be treated to avoid a failure was three. Antibiotics influenced resolution of sputum purulence but did not influence recovery of peak flow or gas exchange.

The meta-analysis performed by El Moussaoui et al6 published in this issue of Thorax (see page 415) has addressed an important aspect of antibiotic treatment. Some COPD guidelines have recommended a choice of antibiotic to use during exacerbations, but none has addressed the length of the course of treatment. The authors list the benefits of a shorter course: better patient compliance, fewer side effects and, most importantly, reduced risk of antibiotic resistance development. The result is clear: short-course treatment—which usually means 5 days—is equally efficacious as longer courses (7–10 days). Eradication of bacteria from sputum was also equivalent. This is a very important message for clinicians. The result was the same in trials in which short and longer courses of the same antibiotic were included and when antibiotics were grouped by class. The authors rightly restrict their conclusions to mild to moderate cases. COPD is a very heterogeneous condition and patients enrolled into clinical trials do not usually have life-threatening disease, and protocols exclude sicker patients that are more likely to fail. Trials have usually had a primary end point soon after the end of treatment and so may have missed early relapse due to inadequate treatment.

In the above-mentioned Cochrane review5 a significant benefit for antibiotics versus placebo was found for mortality (relative risk ratio 0.23), but this result was heavily influenced by a single study in patients with severe exacerbations requiring ventilator support.7 Several COPD studies have sought to identify risk factors for poor outcome of an exacerbation.8 Frequent exacerbations, low forced expiratory volume in 1 s, co-morbid diseases (especially cardiovascular and diabetes), low body mass index, current smoking habit, alcohol consumption, duration of COPD and older age have all been identified as risk factors in different studies.8 In these patients, clinical response—particularly sputum colour and, in a hospitalised patient, return of inflammatory markers to baseline—should determine length of treatment. Some patients with COPD who may have regular sputum production and be particularly prone to infective exacerbations have underlying bronchiectasis.9 This is another group in which the course of treatment might need to be longer, although it could be argued that these patients are particularly prone to resistance development because of the larger concentration of bacteria in the airway lumen. Short-course treatment would therefore still be desirable if it was proved to be effective.

A rapid specific biomarker to identify bacterial lower respiratory tract infections would provide a major advance in the antibiotic management of patients with COPD, particularly if it could also be used to judge response to treatment. While procalcitonin has shown some promise in this regard,10 more work is needed to explore its application. For the time being, clinical judgement will determine which patients receive an antibiotic and the length of time for which treatment should be given. Present guidelines are not consistent, but purulent sputum as a marker of bacterial infection, together with increased sputum volume and/or increased breathlessness indicated by the study of Anthonisen et al,4 are recommended to judge the need for antibiotic. The meta-analysis by El Moussaoui et al has shown that, in patients without risk factors for poor outcome, a 5-day course of antibiotic should be used. Further studies are needed in at-risk groups because short courses might still be effective for some antibiotics that penetrate well into the respiratory mucosa, and are active against resistant strains that are more common in at-risk patients who have received antibiotics previously. A weakness of the meta-analysis is that most studies include a new antibiotic seeking registration in the short-term arm and not older agents such as amoxicillin, tetracycline and erythromycin. However, at the present time, few studies have shown superiority of one antibiotic over another in this patient group.8 Future studies should involve follow-up for several weeks after the end of treatment to ensure early relapse does not occur because of bacterial persistence, and should include tools such as patient reported outcome questionnaires to determine speed and extent of recovery rather than the traditional end point of a judgement by the clinician as to whether or not the patient requires more antibiotic.