Assessing the quality of reporting in meta-analyses of randomized controlled trials.

Abstract

Evidence-Based Health Care is becoming increasingly popular, and in some settings is influencing the direction of health-care policy. If Meta-analyses (MAs) are to continue to be useful, consideration must be given to how they are reported. This study was carried out to: (a) assess the quality of reporting and general characteristics of paper-based MAs published from 1977 to 1995; (b) assess over time; (c) compare with Cochrane Collaboration (i.e., electronically published) MAs. Paper-based MAs were randomly selected from a database of systematic reviews. Cochrane Collaboration MAs were randomly selected from the Cochrane Database of Systematic Reviews, Issue 3, 1996. Data were analyzed using univariate and multivariate approaches. One hundred and fifty-one MAs were reviewed. The overall mean quality score of MAs published in the 'Early Years' was 3.04 (95% Confidence Interval (CI) 2.53,3.56); 'Later Years' 3.35 (95% CI 2.83,3.87), and the Cochrane Library MAs 3.42 (95% CI 2.92,3.93). This study reveals that there has been a trend, over time, towards improvement in the reporting quality of paper-based published MAs and that the reporting quality of (Cochrane) electronically based MAs are no different from that of paper-based MAs.