Developing and testing high-efficacy patient subgroups within a clinical trial using risk scores.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Genomic Medicine

Background:

• High-dimensional patient data (genomic, imaging, wearables) can identify subgroups benefiting from treatments.
• The adaptive signature design (ASD) method uses genetic data to test treatment efficacy in sensitive patient groups.
• ASD requires computationally expensive tuning parameter selection.

Purpose of the Study:

• Propose a novel variation to the ASD method, the cross-validated risk scores (CVRS) design.
• Eliminate the need for tuning parameter selection in ASD.
• Assess the computational efficiency and performance of CVRS compared to the original ASD.

Main Methods:

• The CVRS method computes patient risk scores and uses nonparametric clustering.
• Evaluated CVRS against cross-validated ASD using simulation data and a real psychiatry trial.
• Assessed performance across various sample sizes and response rates.

Main Results:

• CVRS significantly reduces computational time compared to ASD.
• CVRS demonstrates improved identification of the sensitive patient group in simulations.
• CVRS shows enhanced power to detect treatment effects in the sensitive group across various scenarios.

Conclusions:

• CVRS offers a computationally efficient and effective alternative to ASD for identifying treatment-sensitive patient subgroups.
• The method shows promise for optimizing clinical trial design and treatment efficacy analysis.
• CVRS was successfully applied to a real-world psychiatry clinical trial.