RoBoT: a robust Bayesian hypothesis testing method for basket trials.

Area of Science:

• Clinical Trials Methodology
• Bayesian Statistics in Oncology
• Biostatistics

Background:

• Basket trials in oncology evaluate single treatments across multiple cancer types or subtypes.
• Hierarchical modeling enhances basket trial analysis by borrowing strength across subgroups.
• Existing methods often rely on posterior credible intervals, which may lack interpretability.

Purpose of the Study:

• To introduce RoBoT (Robust Bayesian Hypothesis Testing), a novel Bayesian method for phase II oncology basket trials.
• To provide a formal hypothesis testing framework for robust and interpretable treatment efficacy inference.
• To address limitations of existing methods in decision-making for complex basket trial designs.

Main Methods:

• Proposed a Bayesian hypothesis testing framework, RoBoT, for analyzing phase II basket trial data.
• Assumed latent subgroups within which treatment efficacy probabilities are similar.
• Employed a Dirichlet process mixture model to infer subgroup number and membership from data, avoiding Type I error inflation.

Main Results:

• Computer simulations demonstrated RoBoT's operating characteristics compared to existing methods.
• The method was applied to real-world data from phase II basket trials of imatinib and vemurafenib.
• RoBoT provides interpretable and robust inference, outperforming traditional approaches in specific scenarios.

Conclusions:

• RoBoT offers a statistically rigorous and interpretable approach for decision-making in oncology basket trials.
• The latent subgroup modeling effectively handles heterogeneity across cancer types/subtypes.
• This Bayesian hypothesis testing framework enhances the reliability of treatment efficacy conclusions in complex trial designs.