Online Program

Surrogacy Assessment Using Principal Stratification When Surrogate and Outcome Measures are Multivariate Normal

*Anna Sadie Chernin Conlon, University of Michigan 
Michael Elliott, University of Michigan 
Jeremy Michael George Taylor, University of Michigan 

Keywords: Bayesian estimation, Principal stratification, Surrogate endpoints

In clinical trials, a surrogate outcome variable (S) can be measured before the outcome of interest (T) and may provide early information regarding the treatment (Z) effect on T. Most previous methods for surrogate validation rely on models for the conditional distribution of T given Z and S. However, S is a post-randomization variable, and unobserved, simultaneous predictors of S and T may exist. When such confounders exist, these methods do not have a causal interpretation. Using the principal surrogacy framework introduced by Frangakis and Rubin (2002), we propose a Bayesian estimation strategy for surrogate validation when the joint distribution of potential surrogate and outcome measures is multivariate normal. We model the joint conditional distribution of the potential outcomes of T, given the potential outcomes of S and propose surrogacy validation measures from this model. By conditioning on principal strata of S, the resulting estimates are causal. As the model is not fully identifiable from the data, we propose some reasonable prior distributions and assumptions that can be placed on weakly identified parameters to aid in estimation. We explore the relationship between our surrogacy measures and the traditional surrogacy measures proposed by Prentice (1989). The method is applied to data from a macular degeneration study and data from an ovarian cancer study, both previously analyzed by Buyse, et al. (2000).