When to stop treatment arms in a clinical trial assessing time...

Two-arm time-to-event (or survival) trials are powered to continue until a required number of events is reached. The authors discuss how the required number of events should be defined for a study with three or more arms with various pairwise comparisons and a common control arm. They advocate stopping one active arm when the required number of events is observed in the applicable pairwise comparison but continuing with the other active arm and the control arm until the required number of events is observed in that pairwise comparison, thereby ensuring that the study continues until enough events are observed in each pairwise comparison. This article is the result of considerations during the design of a three-arm microbicide trial.