Abstract:
|
The ICH E9 addendum suggested a few estimand strategies: treatment policy strategy, composite variable strategy, hypothetical strategy, principle stratum strategy, and while on treatment strategy. While there are different approaches available for other estimands, there are very limited statistical approaches for the principle stratum estimand. The challenge with the principle stratum estimand is that the population of interest in each group is not randomized population at all. Comparing two treatment groups among a principle stratum is like comparing two groups among completers. Commonly used statistical analysis approaches will therefore lead to biased analysis results. This presentation will present a valid statistical approach when the principle stratum estimand is used. Detailed statistical model, model assumption and inferential approach will be presented.
|