|
Abstract:
|
“Principal stratum” (PS) is one of the strategies in defining estimands in the ICH E9(R1) guidance for estimands and sensitivity analyses in clinical trials. The estimand for a PS, e.g., adherers to the assigned treatment, may be clinically meaningful and can be used to help make treatment decisions. Generally, estimation for a PS-based estimand requires stronger assumptions to allow predict the outcome for patients in the PS or/and identify the PS membership. One common assumption is monotonicity, which implies that if a patient belongs to a certain PS that same patient must also belong to another PS. For example, a patient must be an adherer to the placebo (control) treatment if this patient adheres to the experimental treatment. Therefore, monotonicity that assumes a deterministic relationship between membership in two PS may be unrealistic. In this presentation, we evaluate the plausibility of monotonicity assumption using data from a 2x2 cross-over clinical study, where strata membership defined by both treatments is observed. We also discuss the stochastic monotonicity, a potentially more realistic assumption and its use in the estimation for the treatment effect in a PS.
|
