This study evaluated the relative efficacy of long-acting bronchodilators, long-acting muscarinic antagonist, phosphodiesterase-4 inhibitors and inhaled glucocorticosteroids and found that, depending on the type of drug, various treatment combinations offer benefits for chronic obstructive pulmonary disease exacerbations. Conclusions are supported by data presented but possible publication bias and lack of studies for some comparisons should be considered.

To evaluate the relative efficacy of long-acting bronchodilators, long-acting muscarinic antagonist, phosphodiesterase-4 inhibitors and inhaled glucocorticosteroids, in various combinations, for chronic obstructive pulmonary disease (chronic obstructive pulmonary disease) exacerbations.

PubMed, EMBASE and Cochrane Central Register of Controlled Trials (CENTRAL) were searched from inception to September 2010 for studies published in English. Search terms were reported. The bibliographies of relevant publications were searched for further studies.

Randomised controlled trials (RCTs) of at least six months (24 weeks) duration that evaluated long-acting bronchodilators (formoterol or salmeterol), long-acting muscarinic antagonist (tiotropium), inhaled glucocorticosteroids (fluticasone or budesonide), phosphodiesterase-4 inhibitors (roflumilast) alone or in combination, compared with these interventions or placebo were eligible for inclusion. Patients had to demonstrate moderate to severe chronic obstructive pulmonary disease according to the Global Initiative on Obstructive Lung Disease definition. The primary outcome of interest was exacerbation rates (as defined by the authors).

Most participants in the included studies had a forced expiratory volume (in one second) of less than 80% than the predicted value. Studies were conducted in a variety of countries. Treatment duration ranged from 24 weeks to four years and background and/or rescue therapy was permitted in all of the included trials. A variety of doses were used. Studies reported the total number of exacerbations, the mean annual rate of exacerbations and the proportion of patients with at least one exacerbation.

The number of reviewers that performed study selection and how studies were selected was not reported.

Methodological quality was assessed in terms of randomisation, allocation concealment, blinding, use of intention-to-treat analysis and studies with more than 20% drop-outs.

The number of reviewers that assessed study quality was not reported.

Exacerbation rates reported by the original studies were extracted and classified as moderate or severe where data were available. For the primary analysis based on rates, the number of exacerbations per study arm was extracted and data were extracted to calculate total patient years at risk. For the binomial analysis, data were extracted for the number of patients that had at least one exacerbation during the trial.

Data were extracted independently by two reviewers.

The data were synthesised with a mixed treatment comparison meta-analysis framework and a standard random-effects Poisson regression model. The primary analysis derived the relative effects of the interventions directly investigated in the included studies, whereas the secondary additive main effect model allowed investigation of combinations of drugs not assessed in any of the included studies. Both the primary and secondary analyses produced estimated rate ratios (RR) and corresponding 95% confidence intervals (CI) for rate of exacerbation. Estimated absolute effect of each treatment (as mean exacerbations per patient-years) and estimated probabilities that each treatment was best (associated with the lowest rate of exacerbations in chronic obstructive pulmonary disease per patient-years) were calculated. For the primary model, sensitivity analyses were undertaken for a fixed-effect Poisson regression model, a random-effects logistic regression model (for binomial events), covariates (such as severity of chronic obstructive pulmonary disease exacerbations and publication year), different Bayesian priors for standard deviation. The primary analysis was re-run with the same data as for the secondary analysis. For the secondary analysis, published data were replaced by data not published at the time of manuscript acceptance.

Twenty-nine RCTs were included in the review (36,657 patients; range 79 to 6112). Eighteen RCTs reported randomisation, 16 reported concealment of allocation, 14 reported blinding, 20 were intention-to-treat time periods and 20 RCTs had more than 20% drop-outs. Ten treatment combinations were included in the primary mixed treatment comparison and 15 were included in the secondary mixed treatment comparison analysis.

The primary analysis favoured inhaled glucocorticosteroids plus long-acting bronchodilators plus long-acting muscarinic antagonist (absolute treatment effect 0.82, 95% CI 0.57 to 1.15), and also the addition of roflumilast to long-acting muscarinic antagonist (absolute treatment effect 0.75, 95% CI 0.53 to 1.02) and to long-acting bronchodilators (absolute treatment effect 0.81, 95% CI 0.58 to 1.10). The combination of roflumilast plus a long-acting muscarinic antagonist showed the largest treatment effects compared with the other treatments and had the highest probability of being the best first-line treatment (45%).

The secondary analyses suggested that the highest probability of reductions in exacerbations was found with roflumilast plus inhaled glucocorticosteroids plus long-acting bronchodilators plus long-acting muscarinic antagonist. The secondary analysis again favoured inhaled glucocorticosteroids plus long-acting muscarinic antagonist plus long-acting bronchodilators (absolute treatment effect 0.63, 95% CI 0.54 to 0.73) and indicated that the addition of roflumilast may further reduce exacerbations (absolute treatment effect 0.53, 95% CI 0.43 to 0.64). These findings were robust to the sensitivity analyses.

Depending on the type of drug, combined treatments offer a therapeutic advantage.

The research question was supported by defined inclusion criteria. Several relevant databases were searched, but only published English language studies were included, so the review may have been prone to language and publication bias. Data were extracted by two reviewers reducing the risk of error and bias, but this was not reported for study selection and quality assessment. Quality was assessed with appropriate criteria. Standard mixed treatment comparison methods were used and the results were robust to sensitivity analyses. The authors acknowledged that the power of the study may have been insufficient to detect significant effects as some comparisons only involved one trial.

The authors' conclusions were supported by the data presented but should be interpreted in light of possible publication bias and lack of studies for some comparisons.

Practice: The authors stated that any recommendations for treatment should be alongside individual patient advice, tailored to the patients' situation.

Research: The authors did not state any implications for research.

Merck Sharp and Dohme Ltd; Nycomed Ltd.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.