Background: Multiplicity in systematic reviews is an often neglected issue. For every systematic review in The Cochrane Library primary endpoints should be defined. However, primary endpoints of meta-analyses and included studies often do not match. Objectives: To assess the impact and direction of the nature of the study endpoint (primary vs. non-primary) on results of meta-analyses of controlled trials. Methods: We randomly selected 50 systematic reviews of controlled trials with meta-analyses from The Cochrane Library, Issue 4, 2007. To be included, meta-analyses had to incorporate at least five studies and report a dichotomous outcome. An endpoint was considered as primary if it was explicitly stated as primary endpoint, was used for sample size determination, or was the only outcome being reported. The first eligible meta-analysis in each systematic review was considered. The method of statistical analysis was described previously [1]. For each systematic review, we calculated the combined effect estimate separately for studies with the same primary endpoint as the meta-analysis (SAME) and studies with a different or undetermined primary endpoint (non-SAME), using the meta-analytical method used in the systematic review. The ratio non-SAME/SAME below one indicates that non-SAME studies show a more beneficial effect. These ratios were combined by means of random-effects meta-analysis. Calculations were performed with RevMan5 and MS Excel. Preliminary Results: Up to now, nine systematic reviews with 51 studies were evaluated. In comparison with the corresponding meta-analysis, 21 studies had the same primary endpoint, 30 showed a different or undetermined primary endpoint. The rate ratio was 0.77 (95% CI 0.53 to 1.11). Further results will be presented, including comparisons of SAME vs. non-SAME, SAME vs. no primary endpoint and SAME vs. other primary endpoint. Reference: [1] Jüni P, Holenstein F, Sterne J, Bartelett C, Egger M. Direction and impact of language bias in meta-analyses of controlled trials: empirical study. Int J Epidemiol 2002;31:115-23.