To the Editor Dr Dechartres and colleagues1 investigated the association between analytic strategy and estimates of treatment outcomes in meta-analysis. In view of the emphasis that Dechartres and colleagues1 placed on the risk of bias in selecting trials for meta-analysis, it is unfortunate that their meta-analyses relied on the method of DerSimonian and Laird.2 Evidence of its shortcomings has been accumulating for many years, and Cornell et al3 provide an accurate summary: “the most widely used method for pooling heterogeneous studies—the DerSimonian-Laird (DL) estimator—can produce biased estimates with falsely high precision.” This reliance on the DerSimonian-Laird estimator weakens the study.
Hoaglin DC. Assessment of Heterogeneity in Meta-analyses. JAMA. 2014;312(21):2286-2287. doi:10.1001/jama.2014.14346