Using multiple biomarkers for patient enrichment in two-stage clinical designs

Enrichment strategies play a critical role in modern clinical trial design, especially as precision medicine advances the focus on patient-specific efficacy. By targeting biomarker-defined populations most likely to benefit, these strategies improve efficiency, reduce sample sizes and costs, accelerate timelines, and minimize unnecessary exposure to treatment-related risks and side effects. Recent developments in enrichment design have introduced biomarker randomness and accounted for the correlation structure between treatment effect and biomarker, resulting in a two-stage threshold enrichment design. However, existing methods are limited to a single continuous biomarker. While incorporating multiple biomarkers can improve the accuracy of target population identification, no study has examined how to incorporate multiple continuous biomarkers into enrichment designs due to the challenges in determining multiple thresholds. To fully utilize information from all relevant biomarkers, we propose novel two-stage enrichment designs capable of handling two or more continuous biomarkers. Our framework accommodates two popular treatment effect metrics including average treatment effect (ATE) and the standardized ATE. We illustrate our method using a hypothetical clinical trial involving early-stage Alzheimer's patients and assess the impact of stage one sample size on threshold estimation through a simulation study. Overall, our proposed two-stage enrichment designs offer researchers greater flexibility in integrating multiple continuous biomarkers. The findings from our study provide valuable insights for the advancement of enrichment trial methodology.