Abstract:
|
The ICH E9 R1 introduced five strategies for addressing intercurrent events providing a framework for a nuanced consideration of how to account for a confounding impact of post-randomization events on estimation of the treatment effects. Among the five strategies, the treatment policy, composite, and hypothetical are potentially the most widely applicable. In practice, there sometimes is an impression of a foregone conclusion that the treatment policy strategy is the most desirable and robust. Yet, appropriateness of all strategies should be evaluated equally carefully based on clinical relevance, ability to align the chosen strategy with a robust estimator, and feasibility of collecting sufficient and high-quality data for estimation. In this presentation, we will discuss several aspects and examples of intercurrent events that should be considered to evaluate clinical relevance of the estimand strategies. We will highlight the underlying assumptions of the three strategies under which the estimated treatment effects would be expected to be generalizable to future clinical practice. We will also discuss challenges with estimators that may arise under each of the three strategies.
|