Chronic obstructive pulmonary disease (COPD) is a debilitating progressive multi-component disease characterised by airflow limitation that is not fully reversible 1. COPD is associated with an inflammatory response in the airways, together with airway obstruction, mucociliary dysfunction and structural changes within the lungs 1. These features result in combinations of symptoms and physiological changes that affect the ability of patients to function, and ultimately influence survival. In addition to chronic symptoms, patients with COPD may experience acute exacerbations, which have a major impact on morbidity, mortality and healthcare utilization 234.

The burden of COPD is considerable. Currently, COPD is the fourth leading cause of mortality worldwide, predicted to rise to the third leading cause by 2020 5. As a disabling condition that affects the physical and social functioning of the sufferer, COPD is also associated with considerable impact on health status. Thus by 2020, COPD is projected to be the fifth leading cause of disability 5.

With increasing understanding of the pathophysiology of COPD, a number of pharmacological and surgical approaches to management of the disease have been developed. International management guidelines recommend that the goals of treatment should be to prevent and control symptoms, prevent and reduce the severity of acute exacerbations, improve lung function, and improve health status 1. In the design of clinical trials, endpoints that are considered to be clinically important have also been clarified recently, and include withdrawal rates from studies, exacerbation rates, lung function (pre and post-bronchodilator FEV₁) and health status 6.

More recent therapies for COPD include salmeterol, a long-acting inhaled β₂-agonist, anticholinergic agonists such as tiotropium bromide and others such as Phosphodiesterase 4 inhibitors which have resulted in a large body of published data assessing the efficacy from numerous clinical trials. These trials have varied considerably in terms of inclusion and exclusion criteria, endpoints and study duration and often restrict other therapies in order to demonstrate any advantage of the new therapy which provides an impression of efficacy that may not reflect the "real world". However, there have been several controlled trials of salmeterol therapy added to usual treatment in COPD which more likely reflects usual prescribing practice. The opportunity was therefore taken to review these studies in the light of current knowledge about relevant clinical endpoints.

The aim of the subsequent meta-analysis of patients with COPD was to provide pooled estimates of the effect of salmeterol 50 mcg taken twice daily when compared with placebo/usual therapy on several clinically relevant endpoints.

Evidence from a meta-analysis of randomized, controlled clinical trials is usually considered the most influential in management guidelines because of the large number of patients involved. This meta-analysis included 3580 patients recruited from centres in 34 countries across four continents, and showed consistent and statistically significant superiority of salmeterol 50 mcg bid over placebo/usual therapy on all outcome measures evaluated after 3, 6 and 12 months of treatment. Patients were 27–33% less likely to withdraw early from a study (p < 0.0001) and 22–28% less likely to suffer a moderate or severe exacerbation (p < 0.0001) when salmeterol rather than placebo was added to usual therapy. Lung function improved by 73–86 mL (3.0–3.2% predicted FEV₁) and more patients experienced a clinically meaningful improvement in health status when treated with salmeterol compared with placebo/usual therapy (46% vs 38%, p < 0.0001).

The results of this meta-analysis are likely to be widely applicable to patients with COPD. Each of the nine studies included was sufficiently powered to detect a pre-defined difference in one or more of the efficacy measures of interest. Heterogeneity in the results from different trials was observed only for FEV₁ during months 1–6, though the degree of departure was small and the results derived from the random effects model did not differ from those of the fixed effects model.

Demographic and baseline characteristics were well-matched between treatment groups, with the study population reflecting the profile of typical COPD patients. Over half of the subjects had been randomized into studies with duration of 24 weeks or more, and 38% were recruited into studies of 1-year duration. Consequently, a reasonable assessment of long-term (6–12 months) efficacy of salmeterol can also be derived from the results of this meta-analysis. The ITT population corresponded to subjects meeting the American Thoracic Society definition of COPD 17, while the subset of subjects with bronchodilator reversibility <10% (PR, comprising 84% of the total population) corresponded to the ERS definition of COPD 18. Results in the PR population were similar to the ITT population.

Withdrawals from clinical trials of COPD present a considerable problem in analysis and interpretation 24. A significantly higher withdrawal rate on placebo/usual therapy (due to lack of potential benefits, thereby leaving a subset who may have less severe disease) may bias the results against the treatment. This imbalance between the treatment arms, with placebo/usual therapy results being artificially improved, would therefore reduce apparent treatment benefits. The meta-analysis showed a highly significant difference in the rate of early withdrawal from clinical studies between the treatment groups. Patients receiving salmeterol were up to 33% less likely to withdraw from studies than those receiving placebo/usual therapy (p < 0.0001). Furthermore, despite the greater withdrawal rate in the control group that may reduce apparent differences between treatments, highly significant treatment benefits with salmeterol were still detected. This suggests that salmeterol provided treatment benefits recognised and valued by patients that outweighed any potential side effects. Indeed, a recent meta-analysis has shown a good safety profile for salmeterol 25, supporting this hypothesis.

Lung function improved by 3.2%, 3.0% and 3.0% at 3, 6 and 12 months, respectively, suggesting that there was no tolerance, or decrease of efficacy, over time for up to 1 year of therapy. Some studies 2627 have suggested that response may decrease with continued use whereas other studies have failed to find evidence of such tolerance 912. The current meta analysis supports the latter studies although the reality and clinical relevance of any decrease continue to be an important focus of attention in COPD therapy.

Exacerbations are common in COPD and may have serious consequences 22829. Recovery from exacerbations may take more than a month in around a quarter of patients, or may even be incomplete 30. A high frequency of exacerbations is associated with a more rapid decline in lung function 31, increased risk of hospitalisation 32 and reduced survival 33, with nearly half of patients hospitalised for a COPD exacerbation dying within 3 years 33. A regular treatment that reduces the frequency and severity of exacerbations could, therefore, have an impact on both morbidity and survival. The current meta-analysis showed that salmeterol significantly (p < 0.0001) reduced the risk of experiencing a moderate or severe exacerbation by up to 28% compared with usual therapy. In addition to the potential patient benefits, this should also reduce medical resource utilization.

COPD is characterised by progressive decline in lung function of around 60–70 ml per year 3435. During an exacerbation, lung function falls by an average of 24 ml 30, while a short-term increase of 90 ml in patients with emphysema is sufficient to improve dyspnoea and exercise performance 36. Consequently, the highly significant improvement in lung function of 73–86 ml observed with salmeterol compared with placebo/usual therapy is likely to be beneficial and clinically meaningful. Importantly, the measurements were made shortly before the next dose of study medication, representing the lowest value in the 12-hour dosing period, so that lung function at other time points could be expected to show a greater treatment difference. The difference between salmeterol and placebo/usual therapy was apparent from the earliest assessment point (3 months) and maintained at 6 months and 12 months, indicating that there was no decline in efficacy over this period.

There is increasing recognition that patient-centred outcomes, such as health status, are important in assessing the efficacy of medical interventions for COPD 37. There was a consistent, statistically significant and clinically detectable improvement in health status in more patients treated with salmeterol than with placebo/usual therapy. A clinically meaningful change with the SGRQ (change in total score of 4 units) corresponds to patients, for example, 'no longer having to walk more slowly than other people, no longer being breathless on getting washed and dressed or on bending over' 38. These changes are generally more relevant to patients than spirometric changes, although they are likely to reflect changes in the latter. The true extent of this benefit as a clinically relevant difference in health status has also been assessed and found to be greater with salmeterol.

The findings of the two populations analysed (Intent to Treat and Poorly Reversible) were consistent, indicating that the degree of reversibility to albuterol has little impact on response to salmeterol therapy, as reported previously 39. There was some evidence of greater improvements in lung function with salmeterol treatment in those with better lung function at baseline, but all patients showed similar benefits in terms of exacerbation rate and health status. These findings suggest that salmeterol treatment is likely to benefit a wide spectrum of patients with COPD to a similar degree.

Recently, Salpeter et al. published the results of another meta-analysis of efficacy of various treatments in COPD 40. Some of their conclusions appear initially to differ from ours, namely, that β2-agonists were associated with more respiratory deaths, and led to no difference in severe exacerbations, compared with placebo. The apparent differences are likely to reflect differences in aim, methodology and timing of the analyses. We analysed data only from studies of salmeterol, whereas Salpeter's methodology allowed consideration of any β2-agonist. In that analysis, most respiratory deaths – 60% of the weighting – were from a single trial of formoterol, its erratum, and accompanying unpublished data 40. Similarly, 98% of the weighting for severe exacerbations was from a single study not included in our analysis. Thus, these results and ours presented here are not mutually contradictory.

It should be remembered that the underlying pathologies and concomitant medications in COPD and asthma are quite different. Nevertheless, recent major studies in asthma add to the body of knowledge about β2-agonists. In the SMART study, conducted in asthma patients, there was no significant difference between salmeterol and placebo in the primary combined endpoint of respiratory-related deaths or life-threatening experiences 41. Respiratory-related deaths, and asthma-related deaths, were slightly but significantly more frequent with salmeterol, but these differences were not apparent after a 6-month follow-up period. Similarly, a meta-analysis conducted in asthma patients (in which SMART data contributed 80% of the weighted data) identified a greater risk of life-threatening asthma attacks with β2-agonists than placebo (6-month risk difference 0.12%; 95% CI 0.01–0.3%) 42. Comparisons with the current data however cannot be drawn.

COPD is a multi-component disease, associated with inflammation, airway obstruction, mucociliary dysfunction and structural changes in the lung. Consequently, it is logical to assume that interventions addressing these different components will be more effective than treatments having more limited scope. In addition to effects on bronchodilation, salmeterol may have other effects, including promotion of mucociliary clearance, protection against bacterial-mediated epithelial damage and anti-neutrophil effects 4344. Whether these additional effects play a role in the overall benefit of salmeterol therapy has yet to be determined. The analysis showed consistently greater efficacy of salmeterol than with placebo when added to usual therapy, which included inhaled and oral corticosteroids, anticholinergic agents, methylxanthines and mucolytics. Importantly, there were no apparent trends in relative efficacy for studies with and without these medications, despite the expectation that a "usual therapy" comparator group may reduce treatment differences. This suggests that these other interventions target different components to salmeterol in the underlying pathophysiology of the disease. However, insufficient detail was collected in the studies to allow this to be examined specifically.

Since the meta-analysis was performed, other completed, published studies have been identified which could have been included in the analysis. In a small population of patients with moderate COPD, salmeterol (n = 6) over 52 weeks reduced exacerbations and lung function when compared to placebo but was less beneficial than a combination of salmeterol and fluticasone propionate 45. In addition, results from two studies comparing salmeterol with the long acting anticholinergic tiotropium over 26 weeks have been published 46. It was felt that inclusion of these studies would not alter the conclusion of the study due to the small patient numbers and in addition, we did not have the individual patient data for the analysis.

In conclusion, this meta-analysis of nine large randomized clinical trials involving over 3500 patients in 34 countries has shown a consistent and highly statistically significant reduction in withdrawal rate from studies, reduction in exacerbation rate, improvement in lung function and improvement in health status with salmeterol compared with placebo/usual therapy, with no evidence of tachyphylaxis to bronchodilation over one year. The impact on a broad range of outcome measures suggests benefits from interventions for COPD that can modify more than one aspect of this multi-component disease.

RS has no competing interests. PW and MW are employees of GSK who sponsored this meta analysis.

RS was involved in the concept and planning of the meta analysis and in writing and editing the manuscript; PW and MW were involved in the writing and editing of the manuscript and responsible for the searches of the databases and the statistical analysis. All authors read and approved the final manuscript.