【摘要】 目的 在临床随访研究中，基于时间尺度指标的限制平均生存时间(restricted mean survival time, RMST)越来越受到关注，然而目前基于RMST的统计推断主要用于两组比较，缺少进行两组以上比较的方法。方法 本文提出RMST多组间的假设检验法，包括经典法(naive)、对数转换法(log)、双对数转换法(cloglog)三种检验法，并通过Monte Carlo模拟评价其Ⅰ类错误和检验效能，最后进行实例分析。结果 综合Monte Carlo模拟的Ⅰ类错误及检验效能结果，显示所提出的RMST检验可以处理多组比较的问题，特别是cloglog转换法最为稳健。结论 针对生存数据的多组比较问题，若考虑从时间尺度指标分析，推荐使用cloglog转换法的RMST多组检验。更多还原

【Abstract】 Objective In clinical follow-up studies, it is the most common method to quantify treatment differences between groups using a hazard ratio(HR). Moreover, restricted mean survival time(RMST) has attracted more and more attention. However, the current statistical inference method based on RMST is mainly used for the comparison between two groups. Methods In this paper, three RMST tests between multiple groups are proposed, including naive, logarithmic transformation and complementary logarithmic transformation. Monte Carlo simulations were performed to evaluate the type I error and power, and a case study was performed. Results Based on the type I errors obtained by Monte Carlo simulation and the test performance results, it is shown that the proposed RMST test can deal with the problem of multiple sets of comparisons, especially the complementary logarithmic transformation method is the most robust. Conclusions For the multi-group comparison of survival data, if the time scale index is considered, the RMST multi-group test by the complementary logarithmic transformation method is recommended.更多还原