A two-armed response adaptive design for ordinal categorical responses using stochastic superiority measure and the effect of delay on it

Abstract

A two-armed ‘single-dose treatment’ response adaptive randomization (RAR) design is developed for ordinal categorical responses using stochastic superiority measure.^[1–3] Doubly adaptive biased coin design of Eisele^[4] and Hu and Zhang^[5] is implemented to carry out the adaptive procedure. The theoretical results like almost sure convergence and asymptotic normality of the observed allocation proportion are provided for the proposed method. The performance of the design is then evaluated through simulation by the operating characteristics like type-I error rate, expected allocation proportion to the superior treatment and power of a suitable statistical test, and compared to complete randomization and RAR (namely the ‘CatGamma’ design of Biswas et al.).^[6] As instantaneous responses of the patients are assumed to develop the design, effects of patient recruitment and response delay are also studied through a delay model. Moreover, a real clinical trial, that is, prophylaxis of post-operative nausea and vomiting trial of Eberhart et al., ^[7] where the nature of response is ordinal categorical, is redesigned to envisage the practical applicability of the proposed design.

AMS Subject Classification: 62P10