|
Abstract:
|
Surrogate outcomes are variables for which knowing the effect of a treatment of interest on the surrogate allows one to predict the effect of the treatment on a more clinically important outcome. There are two major paradigms for the evaluation of surrogate outcomes: the causal effects paradigm and the causal association paradigm. In the causal effects paradigm, a good surrogate mediates most of the effect of the treatment on the clinical outcome or is closely related to a variable which does. In the causal association paradigm , the effect of the treatment on the surrogate is closely associated with or predicts the effect of the treatment on the clinical outcome. We consider here how marginal structural models, together with their associated estimation methods, can be used for evaluating surrogacy under both paradigms in the presence of time-varying treatments and covariates.
|
