An adaptive enrichment design using Bayesian model averaging for selection and threshold-identification of predictive...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Precision Medicine

Background:

• Precision medicine tailors treatments for improved patient outcomes and reduced costs.
• Biomarker-driven adaptive enrichment designs are increasingly used to identify patient subgroups with enhanced treatment effects.
• Current methods often assume prior biomarker knowledge or simple linear relationships, limiting their applicability.

Purpose of the Study:

• To propose a novel Bayesian adaptive enrichment design for identifying predictive biomarkers in clinical trials.
• To address limitations of current methods by accommodating complex, nonlinear biomarker-treatment interactions.
• To improve the efficiency and accuracy of identifying treatment-sensitive patient subgroups.

Main Methods:

• Developed a Bayesian adaptive enrichment design using free knot B-splines for flexible modeling of continuous biomarkers.
• Employed Bayesian model averaging to estimate parameters across various biomarker combinations.
• Incorporated interim analyses for early stopping (efficacy/futility) and adaptive enrollment based on biomarker-defined subgroups.

Main Results:

• The proposed design effectively identifies predictive variables from a set of candidate biomarkers.
• Simulations demonstrate the operating characteristics and performance compared to existing methods.
• The approach handles both pre-categorized and continuous biomarkers, including those with complex relationships.

Conclusions:

• The novel Bayesian adaptive enrichment design offers a flexible and powerful tool for precision medicine clinical trials.
• This method enhances the ability to detect patient subgroups with differential treatment effects.
• It provides a robust framework for optimizing clinical trial design and patient selection.