基于内部预试验方差及组间差值的样本量调整模拟

目的 通过盲态下内部预试验IPS样本量调整的模拟分析,探索协变量存在前提下有效控制Ⅰ型错误、保证检验效能的合理统计方法.方法 利用蒙特-卡罗法模拟存在协变量时的两阶段自适应设计过程,分别采用协方差分析法和方差分析法分析两阶段数据,采用合并P值法确定检验的最终结果,比较两种方法对Ⅰ型错误、检验效能值的影响.结果 采用方差分析的Ⅰ型错误膨胀较协方差分析要大,检验效能也较协方差分析略低,但是Ⅰ型错误的增大更明显.结论 根据第一阶段的方差和组间均值差调整样本量时,如果存在协变量,应采用协方差分析方法分别分析第一、二阶段的数据,然后采用合并P值法做出统计推断.

Simulation of sample size adjustment using the sample variance and the observed treatment difference based on internal pilot study

Objective To determine the appropriate methods to control the inflation of type I error and the lowering of power on the condition of existing covariance through simulations of analysis for internal pilot study in clinical trials. Methods Two-stage design with covariance was simulated by Monte Carlo methods. Both ANCOVA （Analysis of Covariance） and ANOVA （ Analysis of Variance） were used to analyze the two-stage data. Fisher＇s Combination P test was used to make the final reference in order to compare the different effect of two methods on type I error and power. Results Type I error of ANOVA was lager than that of ANCOVA, and the power of ANOVA was lower. The inflation of the Type I error was more obvious and should be paid more attention. Conclusion ANCOVA should be used to make analysis of the data of the first and the second stage respec- tively, and Combination P test should be used to make the final analysis with condition of existing covariable.