生存资料优效性设计的国际多中心临床试验在目标区的决策依据探讨  

The Discussion in the Decision-making about the Multi-Regional Clinical Trial of Superiorty Design of Survival Data in the Target Region

作　　者：戚洋洋 王陵[1] 虞成凯 李婵娟[1] 李晨[1] 蒋志伟[1] 夏结来[1] 

机构地区：[1]第四军医大学军事预防医学院卫生统计学教研室,710032

出　　处：《中国卫生统计》2016年第2期198-200,共3页Chinese Journal of Health Statistics

基　　金：国家自然科学基金资助(项目编号:81273176;81302509);陕西省自然科学基金资助(项目编号:2015JM8440)

摘　　要：目的通过构建生存资料优效性设计包含目标区的国际多中心临床试验数据模型,探讨国际多中心临床试验下目标区生存资料合适的检验标准。方法采用蒙特卡罗模拟方法,利用Log-rank非参数检验方法在总的优效的前提下做目标区优效性检验,根据模拟产生的条件Ⅰ型错误率(CFPR)和条件检验效能(CP),探讨适宜的检验标准,并在此标准上计算所需终点事件数。结果目标区的CFPR和CP均随着检验水准α'、样本比例K的增加而增加,当α'≤0.5时能够保证CFPR控制在0.5以内;当K≥30%时,在f≥0.9的前提下,设置α'=0.3可保证目标区CP能够在72.1%以上;在f≥0.8的前提下,设置α'≥0.4时,可保证目标区CP能够在72.3%以上。结论建议MRCT在方案设计时目标区的样本比例K≥30%,试验数据在目标区的检验水平不应超过0.5。该标准适用于目标区试验药物疗效相当或略低于M RCT的疗效(0.8≤f≤1)的情形,若估计目标区试验药物疗效与M RCT的疗效相等(f=1),检验水平可适当降低,但不建议低于0.3。Objective A study of superiority design was conducted to explore the statistical decision rules for survival datasets in the target region of Multi-regional Clinical Trial（ MRCT）. Methods Double survival datasets were generated by Monte Carlo simulation, the superiority in target region was tested by log-rank method under the established superiority in MRCT to evaluate adjusted statistical significant level α＇ and estimate the cases of endpoint events according to the conditional false positive rate（ CFPR） and conditional power（ CP）. Results The CFPR and CP increase as both of the sample size ratio in the target region and the value of α＇ increase. When α＇≤0. 5, the CFPR can be controlled under 50%. When K≥30% and f≥0. 9,α＇ = 0. 3, then CP≥72. 1%. When K≥30% and f≥0. 8,α＇ = 0. 4, then CP≥72. 3%. Conclusion In the situation of similar or small inferior treatment effects between target region with MRCT（ 0. 8≤f≤1）, sample size of MRCT should be larger than30% and α＇ should be lower than 0. 5 simultaneously. While assumed same treatment effect of target region with MRCT,α＇can be lower appropriately but should not be below 0. 3.

关 键 词：条件Ⅰ型错误率 条件检验效能 国际多中心临床试验 生存资料 MONTE CARLO模拟 

分 类 号：R969.4[医药卫生—药理学]