rBMA: A robust Bayesian Model Averaging Method for phase II basket trials based on informative mixture priors.

Area of Science:

• Oncology
• Biostatistics
• Clinical Trial Design

Background:

• Targeted therapies have advanced oncology drug research.
• Basket trials test antitumor activity across indications with shared genomic alterations.
• Fragmented patient populations necessitate improved statistical methods for basket trials.

Purpose of the Study:

• To propose a robust Bayesian model averaging (rBMA) technique for phase II oncology basket trials.
• To facilitate cross-indication learning and improve statistical power in basket trial design.
• To develop a flexible method that accommodates mixed binary endpoints across indications.

Main Methods:

• Proposed a robust Bayesian model averaging (rBMA) technique.
• Weighted posterior distributions using three models with varying priors (enthusiastic, pessimistic, non-informative).
• Determined posterior weights based on treatment effects across all indications.

Main Results:

• The rBMA approach demonstrated flexibility in supporting cross-indication learning with mixed endpoints.
• Simulation studies evaluated and compared the performance of rBMA against competing methods.
• The proposed method enhances statistical power in phase II basket trials.

Conclusions:

• The rBMA technique offers a robust statistical framework for designing and analyzing phase II oncology basket trials.
• This method effectively enables cross-indication learning, particularly valuable in biomarker-defined subgroups.
• The rBMA approach provides a flexible solution for trials with heterogeneous endpoints, advancing oncology trial methodology.