两均数比较的优效性/等效性和非劣性验证的样本量估计  

Sample size determination for comparison between two means in superiority/equivalence/non-inferiority trails

作　　者：袁小红[1] 庄严[1] 陈平雁[1] 

机构地区：[1]南方医科大学公共卫生与热带医学学院生物统计学系,广州510515

出　　处：《中国新药杂志》2009年第23期2205-2209,共5页Chinese Journal of New Drugs

摘　　要：目的:论述两类方法及其两总体均数之差(μA-μB)对样本量估计的影响和适用性。方法:用基于正态分布和非中心t分布的方法估计样本量并经nQueryAdvisor7.0验证。结果:多数情况下,基于非中心t分布估计的样本量较基于正态分布多1～3例。等效性验证中,μA-μB越趋于等效界值两端,样本量越大。非劣性验证中,μA-μB越大所需样本量越小。结论:等效性/非劣性验证中一概假定2总体均数之差为0会导致检验效能不足,应予纠正。样本量估计应基于非中心t分布方法。Objective：To illustrate the effect of two true means difference （μA -μB） and two types of methods on sample size estimation in equivalence/superiority/non-inferiority trails. Methods： The normal distribution and non-central t distribution separately based methods were applied for sample size estimation, followed by conformation using the nQuery Adivisor 7.0. Results ： Usually the sample size estimated by the non-central t distribution - based method is 1 -3 cases larger than that by the normal distribution-based one. The closer to equivalence limits two true ineans difference is, the larger the sample size is in equivalence trials. The sample size increases with the lager μA -μB = 0 in non-inferiority trials. Conclusion：The assumption of P-μA-μB = 0 for any sample size estimation in equivalence/non-inferiority trials should not be encouraged because it may lower testing power. The non-central t distribution-based method is recommended to sample size estimation.

关 键 词：优效性验证 等效性验证 非劣性验证 样本量 

分 类 号：R969.4[医药卫生—药理学]