Abstract
The role of biomarkers has increased in cancer clinical trials such that novel designs are needed to efficiently answer questions of both drug effects and biomarker performance. We advocate Bayesian hierarchical models for response-adaptive randomized phase II studies integrating single or multiple biomarkers. Prior selection allows one to control a gradual and seamless transition from randomized-blocks to marker-enrichment during the trial. Adaptive randomization is an efficient design for evaluating treatment efficacy within biomarker subgroups, with less variable final sample sizes when compared to nested staged designs. Inference based on the Bayesian hierarchical model also has improved performance in identifying the sub-population where therapeutics are effective over independent analyses done within each biomarker subgroup.