|
Abstract:
|
In randomized trials, surrogates which allow one to predict the effect of the treatment on the outcome of interest from the effect of the treatment on the surrogate are important when it is time-consuming or expensive to measure the primary outcome. Unfortunately, it is possible that the effect of the treatment on the surrogate is positive, the surrogate and outcome are strongly positively correlated, but the effect of the treatment on the outcome is negative. This phenomenon is referred to as the ``surrogate paradox’’ and may lead to disasters in practice. In this talk, I will introduce our new criteria for surrogates to avoid the surrogate paradox based on the principal stratification framework. The evaluation of these criteria requires the identification of principal causal effects. However, we cannot assume the commonly used exclusion restriction assumption for identification. We propose a new identification strategy by utilizing the homogeneity across the multiple trials from a motivating colon cancer clinical dataset. We also develop estimation and inference methods and a model checking procedure. Finally, we apply the method to evaluate whether three-year recurrence of ca
|
