(168) A step-by-step guide for incorporating estimands in observational studies

Background: Precise definitions of study objectives is paramount to align research questions with clinical study goals. The incorporation of estimands in observational studies is gaining prominence to explicitly articulate study targets. While the adoption of estimands in observational studies is not as advanced compared to randomized clinical trials, the inherent lack of control over exposure, treatment, population, and possible patient pathways adds complexity.

Objectives: To develop a comprehensive guide and underscore critical elements of integrating estimands into observational studies.

Methods: A multidisciplinary team, comprising of epidemiologists, statisticians, and analyst, developed a visual step-by-step guide for estimands. The guide addresses challenges particularly related to intercurrent events (ICEs) in observational studies, using real-world data.

Results: We present a 9-step visual guide outlining essential aspects of the estimand framework and ICEs. Steps 1 to 4 cover essential yet commonplace elements in study design, including stakeholders’ needs, research question, study design, patient population, and selection of appropriate effect measures.
Steps 5 and 6 introduce more novel concepts, emphasising the impact of ICEs (i.e., post treatment initiation events) on treatment evaluation. Step 5 involves identifying possible patient pathways, with considerations of patients’ behaviour in a real-world setting, and ICEs that could affect the interpretation of study outcomes. The occurrence of different ICEs may vary depending on the nature of a disease or population studied. Importantly, ICEs have a direct impact on the definition of the estimands (step 6) and subsequent estimation. Step 6 emphasizes the need for detailed description of estimands and strategies applied for each ICE, recognizing their relevance in addressing distinct clinical questions.
Steps 7 to 9 encompass standard stages in observational study conduct, involving the identification of data sources, description of appropriate statistical methodologies, and conducting sensitivity analyses.

Conclusions: This visual guide describes essential steps and highlights key conceptual challenges related to the significance of estimands and ICEs in observational studies. Detailed and transparent reporting of estimands in observational studies is necessary to facilitate informed decision-making by all stakeholders. The guide serves as a valuable resource, aiding researchers in the precise definition and navigation of estimands when utilizing real-world data.