倾向指数匹配法在肺癌化疗效果评价中的应用

The Application of Propensity Score Matching in Evaluating Chemotherapeutic Effect of Lung Carcinoma

目的收集临床肺癌病人观测资料,按照临床化疗方法不同将肺癌病人分为NP方案(长春瑞滨+顺铂)和TP方案(紫杉醇+顺铂)两组,用倾向指数匹配法均衡两组间的混杂因素,比较匹配前后差异,评价两种治疗方案的效果。方法采用logistic回归计算肺癌病人的倾向指数,按照倾向指数进行组间卡钳匹配,采用Kaplan-Meier法对匹配后的数据进行分析,用log-rank检验比较两组生存差异。结果匹配前协变量在两组中不均衡,NP方案组和TP方案组的中位生存期分别为2.360年和2.100年,两组生存率无统计学差异(P=0.0516),匹配后协变量在两组中得到很好均衡,NP方案组和TP方案组的中位生存时间分别变为2.560年和2.180年,生存率间有统计学差异(P=0.0134)。结论倾向指数法能够弥补临床观测数据的不足,通过匹配可有效降低组间混杂偏倚,有利于对肺癌患者不同治疗方案疗效做出更准确评价。

Objective The clinical observation data of lung Carcinoma patient were divided into two groups：NP group and TP group according to clinical chemotherapy method. Compared and evaluate treatment effect of two chemotherapy after using propensity score matching method to balance the confounding factors between groups. Methods Calculated each patient＇s propensity score by logistic re- gression model, and made caliper matching according to the propensity score. The Kaplan-Meier method was applied to make the survival analysis. Results The survival rate was not statistically significant between groups before matching（ P = 0. 0516 ）, and the median survival time of two groups were 2. 360Y and 2. 100Y. The survival rate was statistically significant after matched by the Propensity Score （P = 0. 0134）, and the median survival time of two groups were 2. 560 Y and 2. 180 Y. Conclusion Propensity score matching method can ef- fectively reduce the confounding bias of non-randomized clinical observational data, and help us evaluate the therapeutic effect of Lung Carcinoma patients correctly.