Group Sequential Designs Based On Restricted Mean Survival Time In Clinical Trials

In randomized clinical trials with time-to-event endpoints,one of the main objectives is to compare the difference in survival distributions between two treatment groups.When the proportional hazard assumption does not hold,that is,the hazard ratio of the two groups is no longer constant over time,the traditional log-rank test may fail.In this case,the test based on restricted mean survival time(RMST)is a good alternative because of its strong clinical interpretation and independence of model assumptions.In a certain experiment,if the data are accumulated stably over a period of time,we can adopt the idea of group sequential to conduct the research,that is,to monitor the results in the interim.Especially in clinical trials,a study may take several years or even longer,and the interim analysis of data is conducive to the early termination of the trial under reasonable circumstances,saving sample size and cost.The main work of this paper is to combine the advantages of the above two methods,construct the group sequential design based on RMST,and propose the adaptive design for the re-estimation of sample size.Specifically,in this paper,we prove that the asymptotic distribution of the sequence of standardized test statistics is the canonical joint distribution,use the error spending function to obtain the critical values of the sequential test,and propose two different methods for the re-estimation of sample size given the target power.Finally,we verify that the proposed methods can achieve the expected effect through simulation.