Estimands, Estimators, and Estimates
摘要
The International Council for Harmonization (ICH) E9(R1) guideline introduces a structured framework to align planning, design, conduct, analysis, and result interpretation for clinical trials. Defining an estimand is one of the key elements in the framework, connecting the initial choice of clinical objectives and the interpretation of the analysis results. It is a precise description of the treatment effect, reflecting the clinical question of interest. An estimand comprises five attributes, which include the population, variable, treatment, intercurrent events, and population-level summary. Intercurrent events are a new concept proposed in the ICH E9(R1) guideline. They occur during the patients’ treatment journey and affect the existence or interpretation of the outcomes. The ICH E9(R1) guideline offers five main strategies to handle intercurrent events. Different strategies reflect different clinical questions of interest and have different impacts on the estimand attributes. For a given estimand, the main estimator provides its main estimate; sensitivity estimator(s) test the robustness of the main estimate to the assumptions for the main estimator. Supplementary analyses are conducted in addition to the main and sensitivity analyses with the intent to provide additional insights into the understanding of the treatment effect. Aligning estimation to specific estimands highlights whether statistical methods assess the same or different clinical questions of interest and links the results of estimation back to whether the trial objectives are met.