疑诊消化道恶性肿瘤患者四种常用肿瘤标志物诊断效度评价及Bayes判别式构建

Diagnostic accuracy of four regular biomarkers in patients suspected for malignant tumors of digestive tract and establishing of Bayes discriminant

目的单独使用一种肿瘤标志物或联合多种肿瘤标志物对于消化道恶性肿瘤的诊断具有不确定性。本研究旨在评估单独或联合使用糖类抗原19-9(carbohydrate antigen 19-9,CA19-9)、糖类抗原72-4(carbohydrate antigen 72-4,CA72-4)、甲胎蛋白(α-fetoprotein,AFP)及癌胚抗原(carcino-embryonic antigen,CEA)等肿瘤标志物在疑诊消化道恶性肿瘤患者中的诊断效度,并通过构建Bayes判别的方式探讨该人群中新的肿瘤标志物筛查方法。方法回顾性分析2015-04-30-2017-04-30北京积水潭医院消化内科因疑诊消化道肿瘤住院诊治的患者295例临床资料,入组患者以CA19-9、CA72-4、AFP及CEA等4项肿瘤标志物作为消化道肿瘤筛查指标,同时经胃镜、肠镜检查明确诊断或排除消化道恶性肿瘤。依据是否存在消化道恶性肿瘤将其分为肿瘤组(共87例)和非肿瘤组(共208例)。经二元Logistic回归筛选相关因素后构建Bayes判别方程为Y1=0.000 072 29x1+0.001x2+0.887x3-0.772;Y2=0.006x1+0.004x2+3.253x3-1.854〔x1为血清CEA数值,x2为CA19-9数值,x3为便潜血数值(阳性=1,阴性=0)〕。分别计算单独使用、联合使用上述肿瘤标志物及所得Bayes诊断消化道肿瘤的敏感度、特异度、Youden指数和阳性似然比等指标,绘制受试者工作特征(receiver operating characteristic,ROC)曲线并计算曲线下面积(area under curve,AUC),以进一步评估各肿瘤标志物的诊断效能。结果基线比较结果显示,除AFP以外,肿瘤组患者CEA中位数为3.60ng/mL(四分位数间距10.30ng/mL),高于对照组的1.95ng/mL(四分位数间距1.60ng/mL);肿瘤组患者CA19-9为17.23ng/mL(四分位数间距43.50ng/mL),高于对照组的9.16ng/mL(四分位数间距8.38ng/mL);肿瘤组患者CA72-4为2.80ng/mL(四分位数间距9.30ng/mL),高于对照组的1.50ng/mL(四分位数间距2.30ng/mL),差异均有统计学意义,均P<0.001;肿瘤组患者便潜血阳性率为52.87%(46/87),高于非肿瘤组13.46%(28/208),差异有统计学意义,χ2=49.998,P<0.001。单独使用CEA、AFP、CA19-9及CA72-4诊断消化道恶性肿瘤的敏感度分别为43.68%、6.9%、40.23%和29.89%,特异性分别为91.83%、96.15%、87.50%和88.46%,AUC分别为0.733、0.496、0.686和0.666。肿瘤标志物联合检测结果显示,CEA+CA19-9、CEA+CA72-4、CA19-9+CA72-4和CA19-9+CA72-4+CEA敏感度分别为66.34%、60.51%、58.09%和70.11%,特异度分别为80.35%、81.23%、77.40%和74.04%,AUC分别为0.740、0.754、0.727和0.760。经交叉验证,该判别法诊断消化道肿瘤敏感度62.5%,特异度89.7%,准确度80.0%。结论单独或联合使用2项或3项肿瘤标志物诊断消化道恶性肿瘤的诊断效度不佳,难以满足临床需求。通过构建Bayes判别式诊断消化道恶性肿瘤的特异度、准确度优于单独或联合使用肿瘤标志物,可能更适用于疑诊消化道恶性肿瘤患者的初步诊断。

OBJECTIVE Using one tumor marker alone or combined with multiple tumor markers has great uncertainty in the diagnosis of digestive tract malignant tumors.The purpose of this study was to evaluate the diagnostic validity of common tumor markers such as CA19-9,CA72-4,AFP and CEA alone or in combination in patients suspected for digestive tract malignant tumors.METHODS We analyzed 295 patients retrospectively from Beijing Jishuitan Hospital with suspected diagnosis of digestive tumors from April 30 th 2015 to April 30 th 2017.The included subjects underwent prescreening by CA19-9,CA72-4,AFP and CEA as well as endoscopy.At the same time,the subjects with diagnosis of digestive tumors were enrolled into tumor group while the others with exclusion tumors as non-tumor group.They were divided into tumor group(87 cases)and non-tumor group(208 cases)according to the existence of malignant tumors of digestive tract.ROC curves were drawn for further evaluation of single and combination use of tumor biomarkers.We also established a Bayes discriminant and its diagnostic validity were evaluated in crosswise.The Logistic regression showed that CA19-9,CA72-4 and CEA,but not AFP,were indexes correlated to gastrointestinal tumors.The Bayes discriminant was Y1= 0.000 072 29 x1+0.001 x2+0.887 x3-0.772;Y2= 0.006 x1+ 0.004 x2+3.253 x3-1.854[x1:CEA,x2:CA19-9,x3:occult blood(positive=1,negative=0)].RESULTS CEA[3.60 ng/ml(10.30 ng/ml)vs 1.95 ng/ml(1.60 ng/ml),Z=-6.299],CA19-9[17.23 ng/ml(43.50 ng/ml)vs 9.16 ng/ml(8.38 ng/ml),Z=-5.026]and CA72-4[2.80 ng/ml(9.30 ng/ml)vs 1.50 ng/ml(2.30 ng/ml),Z=-4.489],but not AFP[2.17 ng/ml(1.94 ng/ml)vs 2.19 ng/ml(1.43 ng/ml),Z=-0.116,P=0.581],were higher in tumor group with statistical significance(P<0.001).The positive rate of occult blood was higher in tumor group[46/87(52.87%)vs 28/208(13.46%),χ2=49.998,P<0.001]with statistical significance.Single use of CEA,AFP,CA19-9 and CA72-4 were of low sensitivity(43.68%,6.9%,40.23% and 29.89%,respectively)and relatively high specificity(91.83%,96.15%,87.50% and 88.46%,respectively),with area under curves of 0.733,0.496,0.686 and 0.666,respectively.Combined use of CEA+CA19-9,CEA+CA72-4,CA19-9+CA72-4 and CA19-9+CA72-4+CEA were of relatively high sensitivity(66.34%,60.51%,58.09% and 70.11%,respectively)and relatively low specificity(66.34%,60.51%,58.09% and 70.11%,respectively),with area under curves of 0.740,0.754,0.727 and 0.760,respectively.The sensitivity,specificity and accuracy to gastrointestinal tumors were 62.5%,89.7%and 80.0%respectively.CONCLUSIONS Using two or three tumor markers alone or in combination,the diagnostic validity of two or three tumor markers is not good enough to meet the needs of primary screening of tumors.By constructing the specificity of Bayes discriminant in the diagnosis of malignant tumors of digestive tract,the accuracy is better than that of single or combined use of tumor markers,which may be more suitable for the preliminary diagnosis of patients suspected for digestive tract malignant tumors.