Commentary

TL;DRAbstract

Doing a meta-analysis is easy. Doing it well – well… Ingram Olkin, at a leisurely breakfast on a sunny Sunday of this century When it comes to the effectiveness and safety of treatments, does nature prefer multiplicative or additive effects—or something else? Empirical surveys of meta-analyses comparing treatments for dichotomous outcomes suggest that the choice of effect measure is associated with how (in)consistent, or heterogeneous, study effects appear to be between them.1–3 A typical observation is that statistical tests for between-study heterogeneity are more often significant with a risk difference measure compared with a multiplicative measure, such as the odds ratio.1,2 In this issue of Epidemiology, Poole et al.4 submit that we should not rush to interpret the aforementioned observations as indications that nature works on a multiplicative rather than additive scale, or as an implication of “a certain ontological property about the universe: about human persons and the expos

Chat with Paper

AI Agents for this Paper