This paper proposes a Bayesian adaptive basket trial design to optimize the dose–schedule regimes of an experimental agent within disease subtypes, called “baskets”, for phase I–II clinical trials based on late-onset efficacy and toxicity. To characterize the association among the baskets and regimes, a Bayesian hierarchical model is assumed that includes a heterogeneity parameter, adaptively updated during the trial, that quantifies information shared across baskets. To account for late-onset outcomes when doing sequential decision making, unobserved outcomes are treated as missing values and imputed by exploiting early biomarker and low-grade toxicity information. Elicited joint utilities of efficacy and toxicity are used for decision making. Patients are randomized adaptively to regimes while accounting for baskets, with randomization probabilities proportional to the posterior probability of achieving maximum utility. Simulations are presented to assess the design’s robustness and ability to identify optimal dose–schedule regimes within disease subtypes, and to compare it to a simplified design that treats the subtypes independently.
The authors thank the handling Editor, the Associate Editor, the two referees, and the Editor for their many constructive and insightful comments that have led to significant improvements in the article. RL was funded by NIH/NCI grants P30 CA016672 and P50 CA221703, PFT was funded by NIH/NCI grants P30 CA016672 and P01 CA148600, and YY was funded by NIH/NCI grants P50 CA098258, P50 CA217685, and P50 CA221707.
We thank the handling Editor, the Associate Editor, the two referees, and the Editor for their many constructive and insightful comments that have led to significant improvements in the article.
"A Phase I–II Basket Trial Design to Optimize Dose-Schedule Regimes Based on Delayed Outcomes." Bayesian Anal. 16 (1) 179 - 202, March 2021. https://doi.org/10.1214/20-BA1205