Sequential linear rank tests for two-sample censored survival data (Q799060)

The author provides a general theoretical framework for two-group clinical trials in survival analysis by using martingale methods and stochastic integrals with respect to counting processes. For general patterns of patient-arrival, allocation to treatments and loss to follow- up, it is shown that, under the null-hypothesis of identically distributed lifetimes, the sequentially computed logrank statistic has uncorrelated increments and satisfies a functional limit theorem. The practical applicability to survival testing in randomized clinical trials is discussed.