Comment: Is Response-Adaptive Randomization a “Good Thing” or Not in Clinical Trials? Why We Cannot Take Sides

Abstract

The paper by Robertson et al. intends to provide “a unified, broad and fresh review of methodological and practical issues to consider” as a contribution to the ongoing debate concerning RAR in clinical trials. Simulations carried out by different authors seem to disprove its usefulness both for statistical inference and as a safeguard for the care of patients in the trial. I argue that the arguments brought forward so far are inconclusive, since the inferential considerations are sometimes incomplete or incorrect, and some simulation studies unconvincing. A Bayesian stand is very common, but often not fully understood.