Description:

Jun (Vivien) Yin, PhD
The Division of Biostatistics/QDSC Faculty Position Candidate

Assistant Professor in Biostatistics
Mayo Clinic College of Medicine

Presentation Title:
Bayesian Adaptive Designs for Biomarker-driven Clinical Trials

Brief Abstract:
The paradigm for cancer clinical trials has shifted towards individualized treatment due to the blooming discoveries of biomarkers and targeted agents. Because of the deficiencies of screening agents or testing histologic tumor types one at a time, basket and umbrella trials are emerging. We proposed hierarchical Bayesian (HB) designs for designing and conducting basket and umbrella trials, to evaluate biomarker-treatment pairing and identify patient subpopulations that are most likely to benefit from novel agents. Compared to parallel design for individual cohorts, the HB designs have greatly reduced sample size, and hence improve efficiency and decreases the financial costs. We further extended the HB design to allow the parallel cohorts to be dynamic, so that at interim analysis investigators can 1) eliminate or graduate molecular/histologic subgroups, and 2) cluster subgroups that exhibit similar response to treatment to improve power. The designs were calibrated with respect to specific error rates. We conduct extensive simulations to assess the performance, and illustrate with an Alliance NCTN coordinated trial in brain metastases with alteration in MET, ALK, NTRK, ROS1, CDK pathway or PI3K pathway (A071701).