摘要
目的探讨含安慰剂组三臂临床试验基于bootstrap再抽样的非劣效评判的方法。方法用Monte Carlo模拟方法,产生服从正态分布、对数正态分布和Gamma分布的随机样本,进行Welch校正t检验法和bootstrap法的α-模拟和power模拟的验证和比较。结果当数据服从正态分布,在样本量较大时,Welch校正t检验法和bootstrap法均表现出较好的统计性能,但当数据呈偏态分布时,Welch校正t检验法的第一类错误率会偏离预先给定的α-水平,而bootstrap法在样本量较大时,第一类错误率基本保持在预先给定的水平。Welch校正t检验法和bootstrap法的power模拟结果基本相同。结论含安慰剂组的三臂临床试验在数据不服从正态分布时,bootstrap法可作为一种有效的非劣效评判方法。
Objective To introduce a bootstrap re-sampling method to assess non-inferiority in three-arm clinical trials including placebo group.Methods Generated random samples follow normal distribution,lognormal distribution and Gamma distribution by Monte Carlo simulation method,test and compare Welch t test with the bootstrap re-sampling method in α-simulation and power simulations.Results When simulate with a large sample that fellows normal distribution,Welch corrected t test method and bootstrap method achieve a good statistical performance,respectively.When run as non-normal distribution,Welch corrected t test result would deviate from the type I error levels,whereas Bootstrap method with large sample,type I error maintain the prespecified α levels.Conclusion The bootstrap re-sampling would be a reliable method of non-inferiority assessment in three-arm trials including placebo group in non-normal distribution.
出处
《中国卫生统计》
CSCD
北大核心
2012年第1期65-69,共5页
Chinese Journal of Health Statistics
基金
江苏省"六大人才高峰"课题资助(06-C-031)