A1023
Title: Evaluating and utilizing surrogate outcomes in covariate-adjusted response-adaptive designs
Authors: Wenxin Zhang - UC Berkeley (United States) [presenting]
Abstract: The intersection of surrogate outcomes and adaptive designs in statistical research is explored. Current surrogate evaluation methods do not directly account for the benefits of using surrogate outcomes to adapt randomization probabilities in trials, which aim to address treatment effect heterogeneity. Surrogate outcomes can minimize participant regret by enabling rapid adaptation of randomization probabilities, especially when early detection of heterogeneous treatment effects is possible. A novel approach is introduced for surrogate evaluation in sequential adaptive designs, and a new covariate-adjusted response-adaptive (CARA) design is proposed using an online superlearner. This approach adaptively chooses surrogate outcomes to update treatment randomization probabilities. A targeted maximum likelihood estimation (TMLE) estimator is presented to address data dependency challenges, achieving asymptotic normality under reasonable assumptions without relying on parametric models. Simulations demonstrate the robust performance of the adaptive design. The framework not only provides a method to quantify the benefits of surrogate outcomes but also offers an easily generalizable tool for evaluating various adaptive designs and making inferences, providing insights into alternative choices of designs.
