The Role of Post Intercurrent Event Data in the Estimation of Hypothetical Estimands in Clinical Trials

Estimation of hypothetical estimands in clinical trials typically does not make use of data that may be collected after the intercurrent event (ICE). Some recent papers have shown that such data can be used for estimation of hypothetical estimands, and that statistical efficiency and power can be increased compared to using estimators that only use data before the ICE. In this paper we critically examine the efficiency and bias of estimators that do and do not exploit data collected after ICEs, in a simplified setting. We find that certain imputation G-formula and G-estimators are equivalent in important special cases. We moreover find that efficiency can only be improved by assuming certain covariate effects are common between patients who do and do not experience ICEs. Even when such an assumption holds, we find that in a simple setting gains in efficiency will typically be modest. We derive analytical results for the bias of the estimators when some of their modeling assumptions are violated and evaluate these in simulations. We conclude with a discussion of the relative merits of estimators that do and do not make use of post-ICE data.