Why most published meta-analysis findings are false

Excerpt

Almost a decade after the controversial paper ‘why most published research findings are false’ was written [1], we re-visit this concern with regard to published meta-analyses. Although reading the title of this article may make some pause for thought (or vehemently disagree), if you simply subscribe to the findings of Ioannidis’s paper then the logical conclusion is that the meta-analyses of primary studies are equally susceptible to being false. Indeed, it was asserted in the paper that meta-analyses of small, inconclusive studies (very common) are probably false [1]. But how can this be when meta-analyses sit unchallenged at the top of the hierarchy of evidence (despite being retrospective and observational in nature)? We will first examine the evidence for our assertion, and then provide reasons why meta-analyses are poor predictors of results from large trials gained from the authors’ experience in perioperative meta-analyses. …