On April 22, 2020, we released our ninth short video, Handling Delayed Outcomes. This topic is presented by Dr. Ben Saville of Berry Consultants, LLC.
Clinical endpoints are typically delayed in time. Sometimes this delay is days or weeks (14 day mortality), at other times the delay may be months (6 month heart failure) or potentially years. If we are conducting an adaptive design, where we would like to make decisions partway through a trial like early success, early futility, or dropping an arm, then our interim dataset will contain a number of patients without the delayed endpoint. These patients differ from missing patients in that the patients have not dropped out, but are merely incomplete at the time of the interim analysis.Fortunately, we often also have early longitudinal data or auxiliary variables on the incomplete patients. For example a 6 month cardiac outcome may also be observed at 1 month and 3 months. These interim data can be used to predict the final endpoint for each patient, taking into account that all predictions are imprecise. These predictions allow for better decisions at interim analyses taking into account all the available data. In the video Dr. Ben Saville will describe common methods for making these predictions, how these predictions can guide decision making within a trial, and factors like the accrual rate of the trial that will amplify or diminish the benefit of these methods.