Group-Sequential Response-Adaptive Designs

Abstract

Suppose that two treatments are being compared in a clinical trial and that responses are immediate. ?Then a response-adaptive randomisation rule can be used to reduce the number of patients on the inferior treatment. Although most of the existing work deals with response-adaptive randomisation in the context of a fixed trial size, it is often more efficient to conduct a trial group sequentially. ?In this talk, a group-sequential response-adaptive procedure is considered in which an error spending function is used to obtain the appropriate stopping boundaries at the different interim analyses. ?The finite-sample performance of two response-adaptive randomisation rules is compared with complete randomisation by simulation. ?An extension of the work to censored survival responses is briefly discussed. ?(The talk is based on joint work with Hsiao-Yin Liu at Queen Mary, University of London.)