ABSTRACT
Many meaningful clinical endpoints are time-to-event endpoints-for example, overall survival, time to a response, time to a cardiac-related event, and time to progressive disease. When the intention and the outcome is that all subjects are followed until the event is observed (no censoring), time-toevent endpoints can be analyzed as continuous endpoints. The inferences can be based on the mean, median, or some other quantity relevant to continuous endpoints. However, in most practical cases involving a time-to-event endpoint, all subjects are not followed until an event (i.e., some subjects have their times censored). This limits the types of analyses that can be performed. Nonparametric inferences on means and/or medians may not be possible. To base the inference on means or medians may require following subjects for a long and perhaps impractical length of time.
