关键词: 临床试验, 删失率, 生存分析, 统计学性质

Abstract:

AIM: To explore the influence on the results of survival analysis when the censored rates are different between the groups. METHODS: Monte-Carlo method was used to simulate a two-group survival data in clinical trials and to estimate the type I error and the test power of the log-rank test and Cox proportional hazards models in different scenario of different censored rates between groups. RESULTS: If censored rates were equal in both test group and control group, the type I errors were well controlled, the estimates of HR were most accurate, and the power was relatively high. If total censored rates were lower than 30%, the lower censored rate difference between two groups, the higher reliability hypothesis testing results. CONCLUSION: In clinical trials, by ensuring enough sample size, the total censored rate should be controlled within 30%, and the difference of censored rate between the two groups less than 10% was suggested.

Key words: clinical trial, censored ratio, survival analysis, statistical property