A basket trial design based on constrained hierarchical Bayesian model for latent subgroups.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Oncology

Background:

• Basket trials offer efficient drug evaluation across multiple cancer types.
• Heterogeneity in treatment effects across cancer types can violate assumptions in standard basket trial designs.
• Existing methods may not adequately address the variability of treatment effects within different cancer types.

Purpose of the Study:

• To develop and validate a novel statistical approach for basket trials that accommodates heterogeneous treatment effects.
• To improve the efficiency and statistical power of clinical trials involving multiple cancer types.
• To address the limitations of the exchangeability assumption in existing basket trial models.

Main Methods:

• Simplified constrained hierarchical Bayesian model for latent subgroups (CHBM-LS) using two classifiers.
• Latent subgroup modeling to aggregate distinct cancer type baskets.
• Information borrowing within identified latent subgroups to enhance statistical power.

Main Results:

• The simplified CHBM-LS approach demonstrated superior performance in real-world basket trial data compared to existing methods.
• Simulation studies confirmed the CHBM-LS approach yields higher statistical power.
• The model effectively controlled Type I error rates across various simulated scenarios.

Conclusions:

• The simplified CHBM-LS model provides a robust and efficient framework for analyzing basket trials with heterogeneous treatment effects.
• This approach offers improved statistical power and reliability for oncology clinical trials.
• The latent subgroup modeling effectively handles variability, making it a valuable tool for future trial designs.