Keywords: subgroup, surrogate marker, treatment effect
The utilization of surrogate markers offers the opportunity to reduce the length of required follow-up time and/or costs of a randomized trial examining the effectiveness of an intervention or treatment. There are many available methods for evaluating the utility of surrogate marker information including both parametric and nonparametric approaches. However, in some settings, the utility of a surrogate marker may be heterogeneous i.e. vary by patient characteristics. In this paper, we propose an approach to identify surrogate markers that may be useful for certain subgroups of the population, with emphasis on how the identification and validation of such markers can improve efficiency and follow-up in the design of a future study. We focus on a time-to-event outcome setting where both the primary outcome and surrogate marker measurement are subject to censoring. We illustrate the approach using a simulation study and demonstrate how this method can correctly identify the specific subgroups of interest. We also illustrate the estimation and inference procedures using data from the Diabetes Prevention Program to examine multiple potential surrogate markers for diabetes.