Letter to the Editor on detecting and dealing with heterogeneity in meta-analyses by Cordero and Dans

Abstract

Cordero and Dans give valuable advice on various aspects of detecting and dealing with heterogeneity in meta-analyses. For assessing statistical heterogeneity, they start with a forest plot of the study-level effect estimates and complement it with 2 related numerical measures, Q and I2. Other recommended approaches in the literature include the between-study standard deviation and a prediction interval for the effect in a new study. All these approaches have shortcomings, which investigation of heterogeneity should take into account. Surprisingly, the limitations of Q and I2, the 2 most popular, are not yet widely understood.