Relationship between quality of reports of antidepressant randomized controlled trials and treatment estimates: systematic review, meta-analysis, and meta-regression analysis.
Cipriani A., Malvini L., Furukawa TA., Barbui C.
In the field of meta-analyses of antidepressant randomized controlled trials (RCTs), it is unclear whether a relationship exists between study quality, treatment estimates, and different quality measures. In this systematic review, we investigated whether quality of RCTs allocating patients with major depression to fluoxetine versus any other antidepressant agent influenced treatment estimates. We designed a systematic review and meta-regression analysis of RCTs, obtaining data from the Cochrane Collaboration Depressive Anxiety and Neurosis Controlled Trials Register, Cochrane Controlled Trials Register, MEDLINE, EMBASE, reference lists, and specialist textbooks, with no language restriction. Two reviewers independently extracted data. Efficacy and tolerability outcomes were calculated for the overall sample and for the subgroup of high-quality trials. Quality was assessed using the Jadad scale, the Cochrane Collaboration Depressive Anxiety and Neurosis quality assessment instrument and 3 items of the CONSORT statement. A meta-regression analysis was also carried out. Thirty-nine RCTs contributed to efficacy and 74 to tolerability analyses. We found no correlation between the methodological quality of reports of RCTs and treatment estimates of efficacy and tolerability. Subgroup analyses of high-quality trials provided treatment estimates that did not materially change from overall estimates. This finding was further confirmed by the meta-regression analysis. Quality is an important issue and meta-analyses should take quality into consideration to provide less biased treatment estimates. However, current quality measures are not related with treatment estimates in antidepressant trials and may not be useful weighting tools when meta-analyses of data extracted from antidepressant RCTs are carried out.