Covariate Adjustment in Basket Trials Borrowing Information Across Subgroups

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacology

Background:

• Basket trials efficiently evaluate single therapies across multiple diseases with common molecular targets.
• Bayesian hierarchical models (BHMs) are standard for estimating treatment effects in basket trials, accounting for subgroup heterogeneity.
• The utility of analysis of covariance (ANCOVA) with treatment-by-covariate interactions in small, heterogeneous basket trial samples remains underexplored.

Purpose of the Study:

• To propose and evaluate covariate-adjusted BHMs incorporating ANCOVA for enhanced precision in basket trials.
• To compare the performance of these novel BHMs against unadjusted BHMs and frequentist ANCOVA models.
• To investigate the impact of ANCOVA with and without treatment-by-covariate interactions within the BHM framework for basket trials.

Main Methods:

• Development of two covariate-adjusted BHMs integrating ANCOVA into the data model.
• Simulation studies to assess the performance and advantages of the proposed BHMs.
• Retrospective application of the BHMs to the MAJIC study, a randomized controlled basket trial in blood cancer.

Main Results:

• Covariate-adjusted BHMs demonstrated improved estimation precision compared to unadjusted BHMs.
• The proposed BHMs outperformed traditional frequentist ANCOVA models in simulation scenarios.
• Information borrowing across subgroups was effectively facilitated by the ANCOVA-integrated BHMs.

Conclusions:

• Covariate-adjusted BHMs offer a valuable enhancement for basket trial analysis, particularly in settings with patient heterogeneity and small sample sizes.
• Incorporating ANCOVA within BHMs provides a robust approach to increase estimation precision.
• The proposed methods show promise for improving the efficiency and accuracy of treatment effect estimation in complex clinical trial designs.