Keywords: Estimand, Tripartite Approach, Causal Inference, Intercurrent Events, Adherence
Whether it is a regulator deliberating about approval, a physician deciding about prescribing, a patient thinking about ingesting or an insurer contemplating about reimbursement for a treatment, the benefits and risks need to be carefully weighed. The central question for all involved with the use of pharmaceutical interventions is this: “What happens when I take this treatment?” An important aspect of any answer to this question needs to acknowledge that all treatments have multiple effects on the patient, most broadly categorized as positive (i.e. efficacy) and negative (i.e. safety). However, these effects can be intertwined as patients and physicians adapt the treatment based on observed efficacy and safety outcomes whether in clinical practice or during the conduct of a clinical trial. In the context of clinical trials, which are a scientific attempt to answer the “what happens” question, intercurrent events [see ICH E9(R1) Addendum] may complicate both WHAT to estimate and HOW to do the estimation. The tripartite approach was proposed by Akacha, Bretz and Ruberg (2017) to more clearly delineate the safety and efficacy benefits of a treatment with particular emphasis on estimating the “what happens” question, that is, when a patient takes the medication as prescribed (the estimand of interest). One element of the tripartite approach suggested the need for causal inference to obtain an appropriate estimator of such a treatment effect. The development of a causal effect estimator will be described, and an example will be given.