非随机化医学研究中风险比的一种估计方法

目的提出一种适用于非随机化医学研究的,结合倾向指数与非参数生存分析估计风险比的方法.方法首先对倾向指数进行估计,然后对倾向指数分布分层以消除比较两组间协变量分布的不均衡.其次对分层样本用非参数生存分析的方法估计两组间发病或死亡的风险比.最后比较本法与常用的Cox模型方法并探讨其适用性.结果将本法应用于一项评价某降血脂新药效果的4期临床试验数据后显示:(1)对倾向指数分布分层后基本上消除了由于随机分组方案失败导致的新药组与传统药物组之间协变量分布的不均衡性,使得非参数生存分析方法得以应用;(2)由本法得到的新药效果的估计-风险比与由Cox模型得到的结果基本一致.结论对于非随机化医学研究,结合倾向指数进行非参数生存分析是一种新的可选择的统计方法.

A Method to Estimate the Hazard Ratio in Non-randomized Medical Researehes

Objective Propose a method that combines the propensity score approach and non-parametric survival analysis for hazard ratio estimation in non-randomized medical researches.Methods The non-parametric survival analysis with the stratified sample was used to estimate the hazard ratio after balancing the covariates distribution across 2 treatment,groups by stratifying the estimated propensity score distribution.Results Applying the method proposed with a post marketing trial data of a cholesterol-reduce drug,(1) The imbalance of the covariates distribution across 2 treatment groups was essentially reduced.(2) The hazard ratios estimated from the proposed method and Cox regression model were essentially consistent.Conclusion Combining the propensity score approach and the non-parametric survival analysis for hazard ratio estimation is a new alternative method for non-randomized medical researches.