ABSTRACT
Response-adaptive randomization (RAR) is a procedure that sequentially modifies the randomization probability of treatment allocation based on accumulating information from past treatment assignments and previously treated patients’ responses, in order to achieve the objective of allocating more patients to the potentially better treatment and possibly other goals. The modification of treatment allocation probability leads to random numbers of patients on treatment arms and introduces dependency into the collected trial data. Recently, many methods have been proposed to handle the dependency and statistical inference approaches have been established for RAR.
