Information Theory–Based Surrogate Marker Evaluation from Several Randomized Clinical Trials with Binary Endpoints, Using SAS

A. Tilahun,Assam Pryseley,Ariel Alonso,G. Molenberghs

DOI: https://doi.org/10.1080/10543400701697190

2008-03-07

Journal of Biopharmaceutical Statistics

Abstract:One of the paradigms for surrogate marker evaluation in clinical trials is based on employing data from several clinical trials: the meta-analytic approach. It was originally developed for continuous outcomes by means of the linear mixed model, but other situations are of interest. One such situation is when both outcomes are binary. Although joint models have been proposed for this setting, they are cumbersome in the sense of computationally complex and of producing validation measures that are, unlike in the Gaussian case, not of an R 2 type (Burzykowski et al., 2005). A way to put these problems to rest is by employing information theory, already applied in the continuous case (Alonso and Molenberghs, 2007). In this paper, the information-theoretic approach is applied to the case of binary surrogate and true endpoints. Its use is illustrated using a case study in acute migraine and its performance, relative to existing methods, assessed by means of a simulation study. Because the usefulness of a method critically depends, among others, on the availability of software, a SAS implementation accompanies the methodological work.