Estimands and How to Define Them
Craig H. Mallinckrodt, Geert Molenberghs, Ilya A. Lipkovich, Bohdana Ratitch · 2019
Chapter 3 describes how post-randomization events can cloud inferences about treatment effects. A study development process chart is presented as a tool to help deal with and overcome these difficulties. The role of the intention-to-treat principle is discussed in this new framework. Emphasis is placed on beginning the process by defining the objectives, and how these objectives may vary for different stakeholders who make different decisions. With clear objectives, it is then possible to define the treatment regimen of interest and to define which post-randomization events (intercurrent events) are a deviation from the regimen of interest. Specific guidance is given on how to define estimands, along with an overview of strategies for dealing with intercurrent events. Estimands for the special circumstances of safety endpoints, early-phase trials, and quality-of-life endpoints are discussed. Finally, an overview of trial design issues is presented.