中文摘要：

目的：应用倾向指数匹配法均衡组间的协变量，评价单纯应用放疗（ CRT ）与三维适形放疗联合化疗（ CCRT）对局限期小细胞肺癌（ LD-SCLC）的治疗效果。方法：LD-SCLC患者224例，其中CRT 109例，CCRT 115例。基线资料包括性别（1＝男，0＝女）、年龄、吸烟（1＝是，2＝否）、结核（1＝有，2＝无）、家族史（1＝有，2＝无）、TNM分期（1＝Ⅰ期，2＝Ⅱ期，3＝ⅢA期，4＝ⅢB期）、血清神经元特异性烯醇化酶（NSE）水平、KPS评分、肿瘤数量（1＝单个，2＝多个）等。以分组变量为因变量，以协变量为自变量，建立logistic回归模型，计算倾向指数，然后按照倾向指数进行组间卡钳匹配。分别对匹配前后的数据进行生存分析。结果：匹配前CRT组和CCRT组的中位生存期（95％置信区间）分别为2．15（1．96～2．48）和2．37（2．06～2．72） a，1 a生存率分别为82．65％和83．36％，2 a生存率分别为59．02％和64．30％，3 a生存率分别为25．28％和29．34％，两组生存曲线差异无统计学差异（χ2＝2．173， P＝0．186）。两组共94对匹配成功。匹配后CRT组和CCRT组的中位生存期（95％置信区间）分别为2．14（1．75～2．44）和2．75（2．38～2．92） a，1 a生存率分别为82．24％和85．36％，2 a生存率分别为56．39％和66．20％，3 a生存率分别为23．44％和34．37％，两组生存曲线差异有统计学意义（χ2＝11．045，P＝0．008）。结论：采用倾向指数匹配法能有效降低非随机化临床试验数据的混杂偏倚。

英文摘要：

Aim： To compare the effects of single chemotherapy （ CRT ） and conformal radiotherapy combined with chemotherapy（CCRT） on limited disease small cell lung cancer （LD-SCLC） patients after balancing the covariates by pro-pensity score .Methods：A total of 224 LD-SCLC patients were subjected ,among which ,109 accepted CRT , and 115 ac-cepted CCRT .A logistic regression model was established ,and the treatment assignment was taken as the dependent varia-ble and the covariates as the independent variables .For each LD-SCLC patient , the propensity score was calculated for cal-iper matching and a survival analysis of the matched data was carried out .Results： Before matching , the median survival time （95%CI） of CRT and CCRT were 2.15（1.96-2.48） and 2.37（2.06-2.72） years,respectively;one-, two-, and three-year survival rates were 82.65%and 83.36%, 59.02% and 64.30%,25.28% and 29.34%, respectively.There was no significant difference in survival curve between the two groups （χ2 =2.173,P=0.186）.A total of 94 pairs patients were matched by propensity score .After matching, the median survival time（95%CI） of CRT and CCRT were 2.14（1.75-2.44） and 2.75（2.38-2.92） years;One-, two-, and three-year survival rates were 82.24%and 85.36%, 56.39%and 66.20%, 23.44%and 34.37%, respectively.There was a significant difference in survival curve between the two groups （χ2 =11.045,P=0.008）.Conclusion： Propensity score matching can effectively reduce the confounding bias of non-randomized clinical observational data .