A0673
Title: SUDO: A Bayesian subgroup-specific utility-based dose testing-optimization design for multi-dose randomized trials
Authors: Fangrong Yan - China Pharmaceutical University (China) [presenting]
Abstract: Phase II basket trials are increasingly adopted, as they enable the concurrent evaluation of multiple tumor types, thereby expediting the drug development process, especially for oncology treatments. With growing evidence supporting the promising efficacy of lower doses for many novel agents, and in line with the FDA's project optimus, a trending practice in oncology phase IIA studies is to combine routine preliminary testing of treatment effects and dose optimization with investigations of multiple doses in a single trial. A novel Bayesian adaptive design is proposed for testing and optimizing subgroup-specific doses in multi-dose randomized basket trials. To address potential heterogeneity between subgroups, the Bayesian model averaging approach is utilized to adaptively cluster predefined patient subgroups. A Bayesian hierarchical dynamic linear model is developed to facilitate efficient information sharing across multiple doses and within specific subgroup clusters. Under the Bayesian inference framework, proof of concept for the treatment effect in each subgroup can be established, and the subgroup-specific optimal dose can be identified based on a utility function that quantifies the trade-off between toxicity and efficacy. Extensive simulation studies are conducted to evaluate the operating characteristics of the proposed design. The results demonstrate its favorable performance across various scenarios.
