American Statistical Association New York City Metropolitan Area Chapter Memorial Sloan-Kettering Cancer Center Biostatistics Seminar

This talk will explore ways to combine historical and randomized controls with the goal of justifying unbalanced randomization (smaller controls arms) in Phase II trials. The methodology relies on a power parameter that controls the relative weight of historical controls. This parameter is updated as data accumulate, depending on how similar the randomized and historical controls are. Simulation studies with a binary endpoint indicate that the method strikes a balance between single-arm and 1:1 randomized trials. Adaptive randomization can be used to improve efficiency with binary endpoints but its utility will be limited with time-to-event endpoints which are becoming increasingly popular. SAS and R code are available for trial-based simulations.