Drug Evaluation Committee Causal Inference for Interpretation - Causal Mediated Analysis
Data Science Subcommittee
June 2024
The deliverables related to causal inference, which were prepared in the hope that they would deepen the understanding of estimand published so far by the Data Science Subcommittee of the Drug Evaluation Committee of the Japan Pharmaceutical Manufacturers Association, presented the concept of causal inference for the purpose of evaluating the causal relationship between a treatment of interest and an outcome. While they are a primary concern of many clinical studies, there may also be interest in the mechanisms and pathways of causality and the influence of variables between treatment and outcome relationships (mediating variables) to further the interpretation of data obtained in clinical studies. In the context of drug development, pharmacodynamic markers and surrogate endpoints can be considered as mediating variables. In order to consider the influence of such mediating variables, a causal inference approach called mediation analysis is required.
We, Task Force 3-1 of the Data Science Committee of the Drug Evaluation Committee of the Japan Pharmaceutical Manufacturers Association (JMPMA) in 2023, prepared the deliverables on causal mediation analysis as one of the useful methods for interpreting results. The first chapter explains the motivation for mediated analysis with specific examples of situations in which it is used, and introduces the roles required of clinical and statistical personnel and what they need to consider when conducting mediated analysis. In Chapters 2 and after, the assumed reader is the statistician, and the basic items of causal mediation analysis are organized, and estimation methods, sensitivity analysis methods, numerical examples, and analysis codes are introduced according to the types of outcomes.
We hope that this publication will assist clinical and statistical staff in planning, analyzing, and interpreting results when implementing causal mediation analysis.
Japan Pharmaceutical Manufacturers Association Drug Evaluation Committee
Data Science Subcommittee Task Force 3-1 for FY2023
Causal Inference for Interpretation - Causal Mediated Analysis