Experimental Designs for Surrogate Evaluation

• Activity Number: 69
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Erin Gabriel*+ and Dean Follmann
• Companies: NIAID and NIAID
• Keywords: Surrogate ; Experimental Design ; Causal analysis

Abstract

There have been many definitions for a surrogate in the statistical literature, as well as in the econometric and social science literature. Prentice[89] was among the first to attempt to give a clear mathematical definition of a surrogate. Since then, many other paradigms for defining surrogates have been proposed which lead to different definitions and criterion for evaluation; Joffe and Greene [09] outline four such paradigms and surrogate definitions within them. Gilbert et al. [08] and Plotkin and Gilbert [12] classified types of surrogates in the setting of randomized vaccine trials by their generalizability and type of surrogate-outcome association, mechanistic or non-mechanistic. We propose a similar classification of a broader class of surrogate definitions based on the generalizability of a validated surrogate and the type of the experiment needed to evaluated the criterion for validation. To determine this classification, we outline experimental designs that allow for the evaluation of a set of given surrogate criterion, while attempting to minimize the number of untestable assumptions.