Causal Inference in Randomized Controlled Trials

One key goal of drug development is to assess causal effects of investigational treatments versus control treatments. Randomized controlled trials (RCTs) have traditionally served as the basis for establishing these causal effects. While the ICH E9(R1) pharmaceutical guideline does not use the word 'causal', its ideas around precisely defining treatment effects (via estimands) are closely related to defining causal effects (via potential outcomes). Indeed, oftentimes causal inference methods are utilized to address intercurrent events (post-randomization events that affect the interpretation or existence of outcomes). We believe there is a need to raise awareness of the role of causal inference in RCTs and to equip statisticians working on clinical trials with knowledge of relevant causal inference tools. This half-day course introduces basic concepts of causal inference and specific topics that are most applicable to RCTs. We start with a general introduction to the causal inference framework, including main assumptions and an overview of the role of causal inference in RCTs. Specific topics include estimation of causal effects, mediation analysis, principal stratum estimands, and conditional and marginal treatment effects, with emphasis on applications in RCTs. The course assumes basic familiarity with statistical inference. Prior knowledge of causal inference is not required.