|
Abstract:
|
The ICH E9(R1) addendum guideline published in May 2021 provides useful information on how to think about various estimands in relation to the corresponding clinical question of interest, in order to facilitate and construct a precise description of a targeted treatment effect. The definition of an estimand involves the specification of five attributes. One key attribute is intercurrent events (ICEs) and the strategies to consider. The choice of strategies results in different estimands and thus different targeted treatment effects. One interesting aspect is the use of the estimand framework in the context of patient-reported outcomes (PROs), which has yet to be extensively considered. In this presentation, we take a PRO endpoint that is continuous and subject to a boundary as an example, focusing on comparing different strategies in handling an ICE and its impact on estimating the targeted treatment effect through simulations. For a given strategy, we also compared different analysis methods in imputing missing data under different assumptions. Two typical ICEs were considered as well as dropout which is assumed missing completely at random (MCAR).
|
