搜索
Vol.24No.7ChinaJournalofModernMedicineMar.20141005-8982（2014）07-0024-04··张华1，钟白云1，王新华2，冯斯斯1应用logistic回归模型结合多指标联合诊断试验ROC分析评价血清t-PSA及f-PSA/t-PSA在前列腺癌诊断及联合诊断中的效果。采用酶联免疫荧光法对56例前列腺癌和80例良性前列腺增生患者进行血清t-PSA和f-PSA检测。分析ROC曲线，确定两项指标最佳诊断值及其灵敏度、特异度。采用Logistic回归建立预测概率模型，获得新的统计量，计算各曲线下面积（AUC），获得最佳诊断点。t-PSA和f-PSA/t-PSA比值对前列腺癌具有最佳诊断价值的切点分别为7.69ng/mL，0.165，AUC分别为0.784和0.817，二者联合检测AUC为0.868。t-PSA和f-PSA/t-PSA比值的联合诊断有效的提高了前列腺癌的诊断率，ROC曲线结合logistic回归模型简单有效，适用于多指标联合诊断试验的分析评价。前列腺特异性抗原；ROC曲线；前列腺癌；logistic回归模型R737.25AEvaluationofthediagnosticvalueofserumt-PSAcombinedwithf-PSA/t-PSAinpatientswithprostatecancerZHANGHua1,ZHONGBai-yun1,WANGXin-hua2,FENGSi-si1(1.DepartmentofClinicalLaboratory,XiangyaHospital,CentralSouthUniversity,Changsha,Hunan410008,P.R.China;2.DepartmentofClinicalLaboratory,YiyangcentralHospital,Yiyang,Hunan413000,P.R.China)Abstract:【Objective】ToexploretheapplicationoflogisticmodelinROCanalysis,andtoevaluatetheclinicalvalueanddiscriminatorypoweroftotalprostate-specificantigen(t-PSA)andfreeprostate-specificantigen(f-PSA)inprostatecancer.【Methods】t-PSAandf-PSAlevelsweremeasuredin56patientswithprostatecancerand80patientswithbenignprostatichyperplasia(BPH)withenzyme-linkedfluorescentassay(ELFA).Sensitivity,specifici－tywerecalculatedforeachtest.Receiver-operatingcharacteristiccurves(ROC)wereanalyzed.Basedonthebinarylogisticregressionmodel,thepredictorsorprobabilitieswereobtainedandappliedtoestablishtheempiricalandbi－normalmodeloftheROCcurvestocomparetheareaunderthecurve(AUC).【Results】Theconfirmedlimitationwiththebestdiagnosticvalueofthet-PSAandf-PSA/t-PSAinprostatecancerwere7.69ng/mLand0.165respec－tively,theAUCofwhichwere0.784and0.817respectively.ThecombinedpredictedROCAUCwas0.868.【Con－clusion】Combinedmeasurementoft-PSAandf-PSA/t-PSAcanimprovethediagnosticaccuracyofprostatecancerefficiently,ROCanalysiscombinedwiththelogisticmodelissimpleanduseful,especiallyforthescreeningtestwithmultiplemarkersforclassification．Keywords:prostate-specificantigen;roccurve;prostatecancer;logisticregressionmodel24··、。。。。±±。、。。。、。。x±、。。。x±覮≥≤。。2.3.1Logistic模型的拟和±覮±覮±±25··。。2.3.2联合预测因子的形成βββ……ββββ。β。。。。2.3.3最佳界点值的确定。。。、、、、、。。。。。。。。“”。。。26··。、、、、、“”27··