Sample size reestimation for clinical trials with censored survival data

A flexible design with updating of sample size in clinical trials is proposed for censored survival data. Patients enter trials serially and are subject to random loss to follow-up. The statistical inference is based on a weighted average of the linear rank statistics, where the weight function at each look depends on prior observed data. A stopping rule is devised to allow early termination and acceptance of the null hypothesis when the experimental treatment offers no advantage or is inferior to the control. The null hypothesis that two survival distributions are equal may be rejected only at the last step when the weight function is used up. The overall type I error rate is preserved. Independent increments for the sequential linear rank statistic is key to deriving asymptotic properties of the test statistic. Some extensive simulations have been carried out to compare the operating characteristics of the method under different scenarios, and for a comparison with the usual log-rank test with fixed sample design. The methodology is also illustrated with a colon cancer clinical trial.