Título: Performance metrics in Response Adaptive Randomization designs
Authors: Arkaitz Galbete
Co-authors: Oleksandr Sverdlov
Language: en

Response-adaptive randomization (RAR) allows clinical trials to align patient allocation with specific study goals without compromising statistical validity. However, performing a holistic assessment of various RAR designs is challenging due to the interplay between optimal allocations, procedure choices, design tuning parameters, and the wide parameter space. This work reviews the statistical background of target-driven RAR procedures for binary outcomes and proposes a systematic template to evaluate multiple criteria—including the clinical context, target allocations, asymptotic versus small-sample performance, statistical inference, and design robustness against departures like time trends or selection bias. This framework offers practical guidance to optimize trial design and decision-making.