Inhaled Corticosteroids in Patients With Chronic Obstructive Pulmonary Disease

Abstract

To the Editor: We are concerned about possible data discrepancies in the systematic review and meta-analysis of inhaled corticosteroids in patients with stable chronic obstructive pulmonary disease (COPD) by Dr Drummond and colleagues. In the analysis of pneumonia events from the study by Burge et al, Drummond et al reported 87 of 372 events in the inhaled corticosteroid group vs 101 of 370 events in the control group, reflecting a control event rate that is considerably higher than any other trial in the metaanalysis. However, the trial publication actually reports on “lower respiratory serious adverse events” rather than pneumonia events. The company trial report from the GlaxoSmithKline Clinical Trial Register (FLTB3054, the identification number for the study by Burge et al) shows that there were only 18 and 8 pneumonia events in the fluticasone and control groups, respectively. We believe that a substantial proportion of the lower respiratory serious adverse events data in the trial publication were actually COPD exacerbations. We question the inclusion of the trial by Wedzicha et al in this meta-analysis. For an unconfounded comparison to be possible, Drummond et al should have selected trials in which the sole difference between the intervention arms was the use of an inhaled corticosteroid. In the context of a metaanalysis focusing on inhaled corticosteroids, the study by Wedzicha et al represents a confounded comparison of fluticasone/salmeterol combination vs tiotropium. The corticosteroid-treated group differed in 2 important respects from the control tiotropium arm: patients received an inhaled corticosteroid, and patients received salmeterol. Any differences in relative treatment effect could have arisen from the administration of salmeterol rather than from the inhaled corticosteroid. Finally, for the Calverley et al 2007 trial (reference 12 in the study by Drummond et al), we refer to the updated pneumonia data in the UK Medicines and Healthcare Regulatory Authority safety newsletter. Drummond et al appear to have estimated the number of pneumonia events using the percentage probability of pneumonia (based on KaplanMeier analysis) for this trial. However, the percentage probability of pneumonia does not provide an exact estimate of the absolute event rate for pneumonia, as is apparent from the different figures presented within the regulatory agency document. These issues argue for meta-analysis on adverse effects to include data from clinical trial registries as well as those submitted to regulatory agencies to obtain accurate estimates of risk. Moreover, consistent and clear reporting of adverse events is essential. Resolving these discrepancies in the data are likely to remove the substantial heterogeneity that exists in this meta-analysis of pneumonia.