基于相对度量指标的非劣效界值计算方法的比较

The Comparison between Two Calculation Methods of Non-inferiority Margin for Relative Measures

目的美国FDA在其非劣效临床试验指南中提供了两种基于相对度量指标的非劣效界值计算方法,一种是与绝对度量指标非劣效界值计算方法相对应的对数转换法(简称对数转换法),另外一种是直接基于相对风险的计算方法(简称直接计算法)。本文探讨了这两种方法计算结果的差异,以评估该差异对非劣效试验结果的影响。方法以风险比(hazard ratio,HR)为例,指代阳性对照药与安慰剂的疗效差异的保守估计值,从计算公式上阐述两种方法的关系,并展示两种方法在高优与低优指标中计算得到的非劣效界值的差异。结果当HR在0.80~1.25时,两种方法计算得到的非劣效界值差别在1%以内;当HR越远离1,差别越大。当取相同的效应保留比例f时,直接计算法计算出的非劣效界值总大于对数转换法,此时会导致低优指标使用直接计算法计算出的界值相对更激进,而高优指标则相对保守。当设定相同的非劣效界值δ时,对于低优指标,使用直接计算法所需设定的效应保留比例f高于对数转换法,对于高优指标则相反。结论在以相对度量指标作为主要评价指标的非劣效试验中,研究者应该认识到这两种计算方法对非劣效界值设定和试验结论的影响,综合考虑临床实践和试验药物的获益-风险等,慎重选取非劣效界值。

Objective FDA has proposed two calculation methods of non-inferiority(NI)margin for relative measures in the guidance for NI clinical trials.One method computes the NI margin by logarithmic transformation and the corresponding calculation process of absolute measures,which is noted as logarithmic transformation method.The other method is directly based on the change of relative risk,which is noted as direct calculation method.This paper analyzed the difference of NI margins and its impacts on NI trial results between these two methods.Methods Hazard ratio(HR)represents the conservative estimate of the efficacy of the active comparator relative to placebo.The association between two methods was evaluated in terms of formulas.The difference of NI margins calculated by two methods were estimated in response variables where higher values are to be considered better(HVB)and those lower values are to be considered better(LVB),respectively.Results When HR is within 0.80~1.25,the relative difference of NI margins between two methods limits to 1%.The difference increases when HR becomes larger or smaller.When the proportion of efficacy retention f is fixed,the margin computed by direct calculation method was larger than that computed by logarithmic transformation method.Hence,the NI margin computed by direct calculation method is radical for LVB variables,but conservative for HVB variables.When the NI margin δ is fixed,the f computed by direct calculation method is larger than that computed by logarithmic transformation method for LVB variables,which is reverse for HVB variables.Conclusion In the NI trials where are based on relative measures,researchers should be aware of the impacts on NI margins and conclusions brought by these two calculation methods,which needs to be considered in conjunction with clinical practice,benefit-risk of test drug,etc.to choose the NI margin prudently.