Speaker
Description
Personalized medicine aims to improve treatment decisions using patient-specific covariates. In diseases with heterogeneous treatment responses, estimating treatment-covariate interactions is essential for identifying effective therapies across patient subgroups. Multi-arm clinical trials provide an efficient framework for evaluating several treatments simultaneously; however, the design problem becomes increasingly challenging as the numbers of treatments and covariates increase. In this work, we propose a statistical criterion for evaluating multi-arm trial designs based on interaction estimation across all potential subject covariates, including both continuous and categorical variables. To address the resulting combinatorial optimization problem, we develop a genetic algorithm that efficiently searches for statistically efficient treatment assignments. The proposed approach generates efficient designs by minimizing the maximum subject-covariate variance across treatment groups, thereby reducing uncertainty in treatment assignment under the individualized treatment rule considered. Extensive numerical experiments, including a real clinical trial application, demonstrate that the proposed algorithm consistently outperforms existing methods, yielding more efficient multi-arm trial designs. The proposed methodology provides a flexible and scalable framework for designing multi-arm clinical trials in personalized medicine.
| Classification | Mainly methodology |
|---|---|
| Keywords | Personalized medicine, Multi-arm clinical trials, Genetic algorithm, Individualized treatment rule |