Keywords: adaptive randomization, Bayesian adaptive design, hierarchical model, operating characteristics, platform design
Traditional drug development is a discrete, one drug, one trial, one phase at a time process. It is a time consuming, inefficient, and costly. A recent estimate of the cost for developing a novel drug reaches $3 Billions. How can we do better? One solution is to take the master protocol approach to integrate platform design with adaptive randomization to assign more patients into more promising treatments based on available data and early stopping for futility and/or efficacy. The design allows effective treatments to graduate early, ineffective treatment to be dropped early, and new treatments as modules to be added to the platform to be evaluated. Hierarchical model allows information borrowing across subgroups to increase the study efficiency. Bayesian approach provides a natural and consistent framework to incorporate all information for statistical inference. Several user-friendly tools for study design, simulations, and conduct will be presented and are freely available in the following sites. https://biostatistics.mdanderson.org/softwareOnline/ https://biostatistics.mdanderson.org/softwareDownload/ http://trialdesign.org