Keywords: heterogeneity of treatment effect, subgroup analysis, Bayesian, clinical trials
Individuals often respond differently to identical treatments, and characterizing such variability in treatment response is important in personalized medicine. In this talk, we will first provide a brief overview of the recent methodological advances in the estimation of individualized treatment effects (ITE). Then, we will describe a nonparametric Bayesian approach to ITE that we have recently developed. Our non-parametric Bayesian accelerated failure time model can be used to analyze heterogeneous treatment effects, when patient outcomes are in the form of time-to-event data. Our approach offers a flexible model for the regression function while placing few restrictions on the baseline hazard. We illustrate the proposed methodology using the data from a large randomized clinical trial (N=9,361) examining the impact of intensive versus standard blood-pressure control on a composite clinical outcome of myocardial infarction, stroke, heart failure or death from cardiovascular causes. We will demonstrate the remarkable flexibility of this approach for estimating a variety of measures of heterogeneity of treatment effect including individual treatment effects, optimal treatment allocation, proportion benefiting from treatment, qualitative interaction. We have developed an R package for implementing the methodology. The software requires little user input in terms of tuning parameter selection or in subgroup specification.