Abstract:
|
Stratified randomization based on prognostic factor(s) is commonly used in comparative clinical trials. We will discuss two topics for such settings, with a focus on time-to-event endpoints: (i) estimation and inference for the "overall" (i.e., averaged across all patients) treatment effect, and (ii) assessment of whether the treatment effect is consistent across strata. Traditional methods (e.g., stratified Cox model) and proposed alternative approaches will be contrasted with illustrative examples and simulations, the latter used to quantify the effect on bias, type 1 error and power.
|