A Bayesian adaptive marker-stratified design for molecularly targeted agents with customized hierarchical modeling

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacogenomics

Background:

• Treatment effects of molecularly targeted agents (MTAs) vary significantly based on patient biomarker profiles.
• Evaluating average treatment effects in overall populations may obscure subgroup-specific efficacy.
• Marker-stratified designs (MSDs) are crucial for assessing MTA efficacy in distinct patient subgroups.

Purpose of the Study:

• To propose a novel Bayesian hierarchical model for MSDs that incorporates biomarker information.
• To improve the efficiency of clinical trial designs for MTAs by borrowing strength across standard treatment subgroups.
• To develop a Bayesian adaptive design for treatment allocation and subgroup effect testing.

Main Methods:

• Developed a Bayesian hierarchical model utilizing a hierarchical prior to leverage information across standard treatment subgroups.
• Incorporated biomarker predictability into the model, acknowledging that standard treatment response may be consistent across subgroups.
• Designed a Bayesian adaptive approach for guiding treatment allocation and evaluating subgroup and predictive marker effects.

Main Results:

• The proposed Bayesian hierarchical model demonstrated improved efficiency in estimating subgroup treatment effects.
• The Bayesian adaptive design effectively guided treatment allocation and tested subgroup and predictive marker effects.
• Simulation studies and a real-world trial application confirmed the desirable operating characteristics of the proposed design.

Conclusions:

• The novel Bayesian hierarchical model and adaptive design offer a more efficient and accurate approach for evaluating MTAs in clinical trials.
• This method enhances the assessment of subgroup treatment effects and the identification of predictive biomarkers.
• The proposed design outperforms existing methods, providing a valuable tool for personalized medicine research.