Evaluating surrogate markers

Abstract

We discuss approaches for efficiently evaluating potential surrogate markers; in particular, we focus on case-cohort designs in which marker evaluation is undertaken only for a random sample of subjects within a randomized trial and for all other subjects who develop a major clinical outcome. These designs will be useful in clinical trials in which a highly significant treatment difference on clinical outcome has been obtained. In addition, we describe a method for using data from all available studies using a meta-analysis to explore the association of treatment effects on the potential marker and on clinical outcome. This may be the most effective approach for marker evaluation because it uses data from both large and small trials and incorporates information from trials in which nonsignificant treatment differences on the major clinical outcome are obtained.