Blinded-Into-Unblinded Interim Analyses for Clinical Trials With Time-to-Event Endpoints.

A large proportion of clinical trials do not meet their recruitment targets. Trials using time-to-event endpoints come along with the additional complexity that the amount of available information depends on the number of observed events, which is random for a specified point in time. This means that even if a trial may have recruited the planned number of patients in its recruitment interval, the amount of available information at a fixed calendar time is random. The determination of the sample size underlying a clinical trial is a very important aspect when designing a trial. When adaptations of the pre-specified sample size are a desired design option to address the uncertainty of underlying assumptions, blinded interim analyses are recommended by the authorities. However, blinding cannot be assured when there is a high interest in a potential early trial stop or a high uncertainty about the underlying effect size. In those situations, (adaptive) group-sequential trial designs may be used. In this work, we propose a flexible, hybrid approach. Based on the chances of an early trial stop, either a blinded or an unblinded interim analysis is conducted such that the advantages of the two commonly applied approaches are combined. We focus on trials where the time to an event is the primary outcome. These studies usually come along with a long study duration, where the necessity of design adaptations may be especially appealing.