|
Abstract:
|
Estimands, especially handling of intercurrent events (ICEs), are critical in clinical trials as they define the target of estimation driven by the trial objective. Different stakeholders may have different preferences in terms of the strategies to handle ICEs. In the practice of clinical trials for Atopic Dermatitis (AD), ICEs have been commonly considered as treatment failures regardless of the cause of ICEs. Such approach may lead to potential bias depending on the proportion and nature of missing data. To account for the different types of ICEs, we classify them into three categories: 1) rescue medication; 2) treatment discontinuation due to efficacy; 3) treatment discontinuation due to other reasons. Considering the cause of ICEs and the missingness mechanisms of the missing data resulting from ICEs, we propose a hybrid estimand combining composite strategy and hypothetical strategy of handling ICEs. We illustrate the estimation of the proposed hybrid estimand through a clinical trial in AD patients and compare the hybrid estimand with other commonly used estimands to evaluate the impact of different choices of analytic approach.
|
