Bayesian Adaptive Methods for Clinical Trials of Targeted Agents

This chapter presents general Bayesian concepts and some specific designs for human clinical trials of targeted agents. The designs employ decision rules that use each patient’s protein or gene expression biomarkers, and possibly conventional prognostic variables, to choose an individualized treatment regime that may include one or several targeted agents. The Bayesian rules are sequentially adaptive in that they are refined repeatedly during the trial by using posteriors updated as new patient data are acquired. A design’s final conclusion is not one recommended treatment regime for all patients. Rather, it is a function that maps each patient’s covariates to a treatment regime targeting that patient’s abnormally expressed biomarkers. Illustrations include dose-finding trials, extensions of the randomized discontinuation design, and a variety of randomized comparative group sequential trial designs.