运用Logistic回归和ROC曲线综合评价肿瘤标志物对肺癌的诊断价值

摘要

目的 采用Logistic回归和受试者工作特征(ROC)曲线探讨癌胚抗原(CEA)、甲胎蛋白(AFP)、糖类抗原(CA)15-3、CA19-9、CA72-4在肺癌中的诊断价值.方法 采用电化学发光免疫分析仪检测不同病理类型肺癌组、肺部良性疾病组及健康对照组血清肿瘤标志物(CEA、AFP、CA15-3、CA19-9、CA72-4)的水平,通过Logistic回归建立回归模型,用ROC曲线分析这些指标对肺癌诊断的意义.结果 肺癌组的CEA、CA15-3、CA19-9、CA72-4水平显著高于肺部良性疾病和健康对照组(P<0.05),且CEA在肺癌的诊断中敏感性最高(45.16%).建立回归模型Y=1/[1+EXP(1.982 X1+3.383 X2+2.373 X3+2.167 X4-1.679)],其中CA15-3的OR最高,所生成的新变量Y值在ROC曲线中的曲线下面积最大.结论 AFP对肺癌无临床诊断意义,血清CEA、CA15-3、CA19-9、CA72-4联合检测对肺癌的诊断具有一定的价值,综合运用Logistic回归和ROC曲线可以提高其肺癌诊断水平.%Objective To investigate the diagnostic value of serum CEA ,AFP ,CA15-3 ,CA19-9 and CA72-4 in diagnosing lung cancer by using the Logistic regression and ROC curve .Methods The levels of serum CEA ,AFP ,CA15-3 ,CA19-9 and CA72-4 in the different pathological types of lung cancer group ,benign lungs diseases group and healthy control group were determined by a-dopting electrochemiluminescence immunoassay(ECLIA) .The regression model was established by Logistic regression .The signifi-cance of these indicators in the diagnosis of lung cancer was analyzed by using the ROC curve .Results The levels of serum CEA , AFP ,CA15-3 ,CA19-9 and CA72-4 in the lung cancer group were significantly higher than those in the benign lungs diseases group and healthy control group(P<0 .05) ,moreover CEA had the highest sensitivity (45 .16% ) in the diagnosis of lung cancer .The re-gression model was constructed ,Y=1/[1+EXP(1 .982X1 +3 .383X2 +2 .373X3 +2 .167X4 -1 .679)] ,in which OR of CA15-3 was highest ,the area under the ROC curve of generated new variable Y was biggest .Conclusion Serum AFP had no significance in clin-ical diagnosis of lung cancer ,the combined detection of CEA ,CA15-3 ,CA19-9 and CA72-4 has a certain diagnostic value in lung cancer ,and comprehensive use of Logistic regression and ROC curve can improve the diagnosis level of lung cancer .