Effect Of Difference Censored Rates Between Groups In Clinical Trails

ObjectiveSurvival time is a widely-used endpoint in clinical follow-up study,and censoring is an important property of survival data.Previous studies have shown that high censored rate would cause bias of efficacy estimation and decreas oftest power.Actually,censored rates between treatment groups are different,and the difference sometimes would be very considerable.But the impact of this different censored rates is unclear.On the other hand,the accruate survival time is unable to be observed in many clinical trials,,such as the actual time of disease progression.There are two topics in this paper.One is to explore the influence of the survial analysis when the censored rates are different between groups.The other is to explore the effects of estimation under three kinds of inexact survival time definitionMethodsMonte-Carlo simulation method was used in our study.Based on exponential distribution and Weibull distribution,two-group survival data with different censored rates were generated.Log-rank test and Cox proportional hazard regression model were used to do analysis.Firstly,the type I error and power of test were calculated to evaluate the impact of different censored rates on the traditional survival analysis.The bias between the true hazard ratio(HR)and the estimated HR was compared to evaluate the impact of different censored rated on the efficacy estimation.Secondly,the median survival time and HR were estimated to verify the difference of the three kinds of inexact survival time.ResultsIn the first part,when the censored rate were equal in two groups,the type I error islowest,and the estimated HR is accurate,and the test power was relatively high.When the total censored rate of two treatment groups was lower than 30%,the lower censored rate difference between groups,the higher reliability ofsurvival analysis.In the second part,the definition of survival time would affect the median survival time estimates.And the median survival time estimates was affect by the duration of visit.Conclusion1.In clinical trial,the total censored rate should be controlled lower than 30%,and the difference between groups should be controlled lower than 10%.2.If the survival time is not exact,the follow up visits duration should be short enough.For the choose of primary endpointin the trial,the difference of median survival time between groups and HR are better than the single group median survival time.