A Better Way Forward: Bayesian Inference, Adaptive Designs, and Model-Assisted Designs for Oncology Trials

Jack Lee, PhD
John G. & Marie Stella Kenedy Foundation Chair in Cancer Research
Professor, Department of Biostatistics, The University of Texas MD Anderson Cancer Center

Join Zoom Meeting Here

Abstract: Simply put, Bayesian 1-2-3 is prior plus data to form posterior. The posterior becomes a new prior. The new prior plus new data generates a new posterior.  Bayesian methods take the “learn as we go” approach and are uniquely suitable for clinical trials. I will first illustrate the concept of Bayesian update and Bayesian inference, then, give an overview of Bayesian adaptive designs.  I will introduce a new class of novel designs, known as model-assisted designs. Model-assisted designs combine the transparency and simplicity of conventional algorithm-based designs with the superiority and rigorousness of model-based designs. They enjoy superior performance yet are simple to implement. A few model-assisted designs will be discussed including the Bayesian optimal interval (BOIN) design, the time-to-event BOIN (TITE-BOIN), the BOIN combination design, and the Bayesian Optimal Phase 2 (BOP2) design. Model-assisted designs establish a new KISS principle: Keep It Simple and Smart! Freely available Shiny applications are provided to facilitate the adoption of model-assisted designs https://trialdesign.org