Constrained hierarchical Bayesian model for latent subgroups in basket trials with two classifiers

Area of Science:

• Oncology
• Biostatistics
• Clinical Trial Design

Background:

• Basket trials assess one treatment across multiple cancer types.
• Heterogeneity in treatment effects arises from multiple classifiers (e.g., cancer type, biomarkers).
• The assumption of exchangeability is often violated, complicating analysis.

Purpose of the Study:

• To propose a novel constrained hierarchical Bayesian model for latent subgroups (CHBM-LS).
• To address potential heterogeneity of treatment effects in basket trials with multiple classifiers.
• To enable adaptive information borrowing across baskets by identifying latent subgroups.

Main Methods:

• Developed a constrained hierarchical Bayesian model for latent subgroups (CHBM-LS).
• Aggregated baskets into latent subgroups using a latent subgroup modeling approach.
• Evaluated treatment effects within each latent subgroup, assuming approximate exchangeability.

Main Results:

• The CHBM-LS approach effectively handles heterogeneity in treatment effects across baskets.
• Simulation studies demonstrated superior performance compared to other methods.
• CHBM-LS achieved higher statistical power and better control of type I error rates.

Conclusions:

• CHBM-LS provides a robust statistical framework for analyzing complex oncology basket trials.
• The model facilitates adaptive information borrowing, enhancing analytical power.
• This approach is valuable for trials with heterogeneous treatment effects influenced by multiple patient or disease characteristics.