成组序贯设计中基于实际信息量的临界值校正方法的比较研究

Comparative Study of Method about Boundary Adjustment for Actual Information Level in Group Sequential Design

目的在大样本条件下,比较成组序贯设计中基于实际信息量的不同临界值校正方法之间的差异。方法基于O′Brien-Fleming(OBF)法和指数损耗函数法,通过设定不同的参数以及匹配条件,比较线性插值校正法、OBF损耗函数校正法、指数损耗函数校正法之间的I类错误α和功效Power的差别。结果当设计方法为OBF法和指数损耗函数法(ρ=2)时,与之相匹配的OBF损耗函数校正法、指数损耗函数校正法(ρ=2)以及线性插值校正法均能将I类错误控制在0.05,功效Power控制在0.9;当设计方法为指数损耗函数法(ρ=1或ρ=3)时,与设计方法不匹配的校正方法仍可将I类错误维持在0.05左右,但功效Power明显偏离0.9。结论在进行成组序贯期中分析时,线性插值校正法作为通用方法能够适用于各类设计,损耗函数校正法只适用于已知损耗函数形式的设计。

Objective To explore the differences among various methods about boundary adjustment for actual information level based on the large sample size in group sequential design.Methods Based on the O′Brien-Fleming(OBF)method and the power error spending function method,the Type I error and the Power were compared between the linear interpolation correction method,the OBF error spending function correction method,and the power error spending function correction method by setting different parameters and conditions.Results When the methods specified in the design were the OBF method and the power error spending function method(ρ=2),the corresponding OBF error spending function correction method,the corresponding power error spending function correction method(ρ=2)and linear interpolation correction method could maintain the Type I error and the Power;when the methods specified in the design were the power error spending function methods(ρ=1 orρ=3),the unsuitable method could still maintain the Type I error around 0.05,but the Power significantly deviated from 0.9.Conclusion The linear interpolation correction method can be suitable to various methods specified in the design at each interim analysis,and the error spending function correction method is available for the design based on explicit forms of the error spending function.