Logistic回归和ROC曲线综合评价四种指标对原发性肝癌的诊断价值

Comprehensive Evaluation of the Diagnostic Value of Four Markers for Primary Hepatic Carcinoma by ROC Curve and Logistic Regression

目的运用logistic回归与受试者工作特征曲线(ROC曲线)评价甲胎蛋白(AFP)、甲胎蛋白异质体(AFP-L3)、胸苷激酶(TK1)、分泌型蛋白Dickkopf-1(DKK1)对原发性肝癌(PHC)的诊断价值。方法选择76例原发性肝癌患者作为PHC组、42例良性肝脏疾病患者作为良性肝病组、32健康体检人员作为对照组,采用电化学发光法检测血清AFP水平;采用亲和吸附层析法检测AFP-L3水平,计算AFP-L3在AFP中比例;采用免疫印迹增强化学发光法检测血清TK1水平:采用ELISA试剂盒检测血清标本中DKKI水平。通过ROC曲线分析和Logistic回归结果的ROC曲线下面积(AUC),对4个指标单独及联合检测的临床性能行比较分析。结果 PHC组4个血清指标的含量明显高于良性肝病组及健康对照组,差异有统计学意义(P<0.05),AFP、AFP-L3、TK1和DKK1诊断PHC的敏感性分别为68.4%、76.3%、77.6%和80.3%,特异性分别为90.5%、91.9%、95.9%和94.6%。四项联合检测的敏感性和特异性为93.4%和89.2%;AFP、AFP-L3、TK1和DKK1的ROC曲线下面积(AUC)分别为0.755、0.808、0.913、0.814,联合检测的AUC为0.941,与AFP、AFP-L3、TK1和DKK1比较,差异均有统计学意义(P<0.05)。结论 AFP-L3、TK1和DKK1可以作为诊断PHC血清标志物,与AFP联合检测可提高诊断效能;Logistic回归结合ROC曲线可用于PHC多指标联合诊断的分析评价。

Objective By using logistic regression and ROC curve to analyze the diagnosis values of alpha- fetoprotein （AFP） 、 alpha- fetoprotein heterogeneity （ AFP- L3 ）, serum thymidine kinase 1 （ TK1 ） 、 DKK1 in primary hepatic carcinoma （PHC）. Methods There were 76 patients with primary hepatic carcinoma （PHC group）, 42 patients with benign hepatic benign diseases （benign hepatic diseases group）, and 32 healthy subjects （control group） enrolled in the study. AFP was detected by ECLIA; AFP-L3 was determined by adsorption centrifuging, and the ratio of AFP-L3 to AFP was calculated; and TK1 was determined by Western blot-enhanced chemiluminecence; DKK1 was detected byELISA method. The area under the ROC curve （AUC） of four factors and Logistic results were compared by clinic efficiency. Results All four factors in PHC group were significantly higher than that in benign hepatic diseases group and control group （P 〈 0.05）. The sensibilities and specificities of AFP,AFP-L3~TK1 and DKK1 were 68.4% ,76.3% ~77.6% 80.3% and 90. 5% ,91. 9% ,95. 9% ,94. 6% respectively. The sensibility and specificity of combined examination was 93.4% and 89.2% ; the areas under the ROC curve（AUC） of four factors and the combined examination were 0. 755,0.808、0.913 、0. 814 and 0. 941 respectively ; allhad statistical difference with AFP, AFP-L3 TK1 and DKK1 （P 〈 0.05 ）. Conclusion AFP-L3 、 TK1 and DKK1 can be used as serum markers for the diagnosis of PHC, and the combined examination of AFP 、 AFP-L3, TK1 and DKK1 could improve diagnostic performance of PHC. Logistic regression combined ROC curve is a simple and effective statistical method, and it can be used in the analysis and evaluation of multi-index combined diagnosis of PHC.