Bayesian Baskets: A Novel Design for Biomarker-Based Clinical Trials.

Area of Science:

• Clinical Trial Design
• Biomarker Development
• Precision Medicine

Background:

• Biomarker-based clinical trials are crucial for therapeutic development and precision medicine.
• Existing trial designs struggle to efficiently integrate varying levels of pretrial biomarker evidence.
• There is a need for flexible trial designs that explicitly account for pretrial biomarker confidence.

Purpose of the Study:

• To develop an efficient and flexible clinical trial design incorporating pretrial biomarker evidence.
• To compare the performance of the proposed Bayesian basket (BB) design against traditional (TB) and biomarker-agnostic (BA) designs.
• To assess the ability of the BB design to generate information on both therapeutic efficacy and biomarker predictive capacity.

Main Methods:

• Developed a novel randomization procedure that explicitly incorporates pretrial estimates of biomarker predictive capacity.
• Simulated hypothetical multiarm clinical trials under various scenarios to compare BB, BA, and TB designs.
• Evaluated designs based on efficiency, power, and information generated regarding biomarker predictive capacity.

Main Results:

• The Bayesian basket (BB) design demonstrated greater efficiency compared to the biomarker-agnostic (BA) design.
• BB generated more information on biomarker predictive capacity than both BA and traditional basket (TB) designs.
• BB increased statistical power over BA when biomarkers were predictive and therapeutics were broadly effective.

Conclusions:

• The Bayesian basket (BB) design provides an efficient method for evaluating experimental therapeutics and predictive biomarkers simultaneously.
• This flexible design accommodates varying levels of pretrial biomarker evidence within a unified platform.
• The BB design enhances the explicitness of clinical trial design decisions regarding biomarkers.