Central questions for drug development, licensing and reimbursement are to establish the existence, and to estimate the magnitude, of causal treatment effects, both in terms of efficacy and of safety. Estimating such effects requires a solid framework that begins by formalizing the question of interest by specifying: the patient population of clinical interest; the treatment strategies under study; the variable of interest, including the definition of time zero and the timing of key measurements; as well as the summary measure that will provide the basis for treatment comparisons. Strategies for handling intercurrent events may be reflected in these attributes; for example, treatment strategies should be composed by specific treatments, events that can affect such treatments and decision rules to apply when such events happen. Any remaining intercurrent events not reflected in these four attributes should also be detailed. Two frameworks have been developed in the context of clinical studies which do or do not randomize patients at baseline. This talk will introduce them, discuss their similarities, and reflect on how they might be deployed in combination to define estimands for RWE studies.
