(172) Assessing the Use of the Estimand Framework in Recent Pharmacoepidemiologic Post-market Safety Studies

Background: The ICH E9(R1) estimand framework provides a systematic approach to ensure alignment among clinical trial objectives, trial conduct, statistical analyses, and interpretation of results, however, whether it can be readily utilized for pharmacoepidemiologic safety studies has not been previously evaluated.

Objectives: This work aims to assess whether each attribute of the estimand framework was used in recently published pharmacoepidemiologic studies.

Methods: We selected inferential, pharmacoepidemiologic articles on drug safety published in the journal Pharmacoepidemiology and Drug Safety (PDS), during 2020 to investigate whether attributes of the estimand framework were well-defined in the study design and reporting.

Results: Among the twenty-five articles selected, nineteen were cohort studies and six were nested case-control studies. All studies had well-defined exposure, outcome, target population, and population level summary. The term intercurrent event (ICE) was not mentioned in any of the studies; however, many cohort studies discussed drug discontinuation, treatment modification and terminal events, and strategies to handle them. All studies used methods to control for confounding: propensity score methods, covariate adjustment, or both for cohort studies; matching and covariate adjustment for the nested case-control studies.

Conclusions: All attributes in the estimand framework were utilized in the contemporary pharmacoepidemiologic studies examined; however, the approaches to ICE and the appropriate strategy to handle them were either not explicitly stated or consistently considered. It may be challenging to discern the concept of ICE in case-control studies. Pharmacoepidemiology frameworks, such as Target Trial Emulation, STaRT-RWE, HARPER, which include relevant approaches to handling ICE and typically include other and more detailed considerations for bias assessment, could be linked to the estimand framework so as to enhance it to address the additional challenges posed by the use of real-world data.