A Difficult Step in Meta-analysis: Refining the Search

Abstract

To the Editor: I found the meta-analysis performed by Edmonds et al (June 2002)1Edmonds ML Camargo Jr, CA Brenner BE et al.Replacement of oral corticosteroids with inhaled corticosteroids in the treatment of acute asthma following emergency department discharge: a meta-analysis.Chest. 2002; : 1798-1805Abstract Full Text Full Text PDF Scopus (46) Google Scholar to be very instructive, but in some areas it needs to be refined. The inclusion in the meta-analysis of patients who had been treated with a combination of corticosteroids (CS) and long-acting β2-agonists clearly should have been avoided in a meta-analysis in which the primary outcome was the relapse rate. It has been demonstrated2Markham A Jarvis B Inhaled salmeterol/fluticasone propionate combination: a review of its use in persistent asthma.Drugs. 2000; 60: 1207-1233Crossref PubMed Scopus (34) Google Scholar that this rate is lower when using a combination of long-acting β2-agonists and CS, compared with studies using short-acting β2-agonists and CS. Some of the outcomes described are too heterogeneous. In my view, to give the same significance and to pool together the “treatment failures” during the first 3 days and to pool together the relapses at 7 days and after is a mistake. While the first pooling might be due to specific triggers, the latter could have a more complex causality, including noncompliance. In the face of significant heterogeneity, reviewers need to be mindful; they should refrain from the statistical pooling of data and should refine their search, focusing more on the exploration and description of the sources of heterogeneity. A Difficult Step in Meta-analysisCHESTVol. 123Issue 3PreviewTo the Editor: Full-Text PDF