Statistical Issues in Evaluation of Surrogate Endpoints

Among the most important issues in the design of a clinical trial is the selection of the primary endpoint. Fleming et al. (1998) provided two criteria to govern its selection: the endpoint should be (i) sensitive to treatment effects, and (ii) clinically relevant. Endpoints that directly reflect how a patient feels, functions, or survives have clear clinical relevance; we refer to these as true clinical endpoints. The cost in time and resources associated with the choice of such endpoints has led to consideration of whether measures of biologic activity if of a treatment, or biomarkers, can be appropriate as surrogates for clinical endpoints. We use the term "surrogate endpoint" to describe biomarkers intended to substitute for a clinical endpoint in a clinical trial. While use of such endpoints in early phase trials is well accepted, their use in Phase III clinical trials--trials intended to define the role of a therapy in standard clinical practice is more controversial. Fleming and DeMets (1996) reviewed previous experience with use of surrogate endpoints in a variety of disease settings to underscore the difficulties in developing and using such endpoints.