|
Abstract:
|
As the Greek philosopher Heraclitus said “the only constant in life is change”, in where most Statisticians will probably say “the only constant thing in a clinical trial is missing data”. The presence of intercurrent events leading to missing data needs to be adequately reflected in formulating the estimand. A commonly used strategy for addressing intercurrent events is the treatment policy strategy. Estimation of a treatment policy estimand is challenging, when data after discontinuation of the study drug is missing. Approaches like LOCF are often discouraged by regulators, because likelihood based analyses allow to better reflect statistical uncertainty than a single-imputation approach. In this presentation we would like to illustrate the retrieved dropout approach (EMA, 2010). The basic idea is that some patients are actively followed after they discontinue the study treatment but still remained in the study to obtain the final measurement. Only this population is then used to impute other patients, who discontinued in the study.
|
