Biomarker-driven optimal designs for patient enrollment restriction.

Area of Science:

• Biostatistics
• Clinical Trials
• Personalized Medicine

Background:

• Personalized medicine utilizes patient-specific characteristics (biomarkers) for tailored treatments.
• Continuous biomarkers present challenges in defining patient subgroups compared to binary ones.
• Current methods often categorize continuous biomarkers using data-driven quantiles, potentially leading to suboptimal treatment allocation.

Purpose of the Study:

• To propose a method for determining the optimal cutoff of a continuous predictive biomarker in binary response trials.
• To develop an optimal design for estimating this cutoff, addressing challenges with unattainable constraints.
• To introduce a novel response-adaptive randomization technique for practical implementation.

Main Methods:

• Derivation of an optimal design for estimating the biomarker cutoff based on relative risk.
• Introduction of a covariate-adjusted response-adaptive randomization to minimize allocation distance to the optimum.
• Extensive simulation studies to evaluate the proposed method's performance.

Main Results:

• The proposed method effectively estimates the optimal biomarker cutoff.
• The covariate-adjusted response-adaptive randomization demonstrates good estimation efficiency and variance reduction.
• Simulations confirm the approach's performance in identifying sensitive and insensitive patient subpopulations.

Conclusions:

• The developed method provides an optimal strategy for utilizing continuous predictive biomarkers in clinical trials.
• The novel randomization technique ensures practical and efficient implementation.
• Findings highlight the ethical considerations and potential drawbacks of using data-dependent medians for subpopulation identification.