A survival trial is a clinical trial with a time-to-event primary outcome, such as time to death or cure. Typically, survival trials involve a limited recruitment period with staggered entry and are complicated by censoring because of losses to follow-up and other reasons, including administrative censoring. The primary outcome of a survival trial is usually analyzed using the logrank test, or a member of the logrank family of tests, including the Gehan test or the Peto-Peto Prentice Wilcoxon test.1

Most survival trials comparing two treatments have traditionally employed equal sample sizes in each treatment arm. However, it has long been known that equal allocation does not always yield the most powerful test in survival analyses.2&32, 3 Formulas that give the optimal allocation to maximize power have been derived.4 Equal allocation also results in half of patients being assigned to the inferior treatment arm, if one treatment is superior. To mitigate this problem, several authors have proposed adaptive randomization be used in the context of survival trials.5&65, 6 In adaptive randomization, allocation probabilities are skewed away from 0.5 to favor the treatment that is performing “better” thus far in the trial. One method that has been proposed is to adapt the probabilities as a function of the current treatment effect in the trial. These probabilities can be adapted before each patient is randomized,5 or as groups of patients are randomized.4, 6 We refer to survival trials employing adaptive randomization as adaptive survival trials. The idea of using accruing data to assign the more favorable treatment in a survival setting can be traced to Ref. 7.