Logistic回归和ROC曲线综合评价CEA、NSE和CYFRA21-1对肺癌的诊断价值

Evaluation of the diagnostic values of serum CEA,NSE and CYFRA21-1 in lung cancer

目的应用Logistic回归和ROC曲线探讨血清癌胚抗原(carcinoembryonic antigen,CEA)、神经元特异性烯醇化酶(neuron-specificenolase,NSE)和细胞角蛋白19片段(cytokeratin 19 fragment,CYFRA21-1)在肺癌诊断中的应用。结论采用电化学发光免疫分析仪(E170)检测不同病理类型肺癌组、肺良性疾病组以及健康人血清CEA、NSE和CYFRA21-1的水平,通过Logistic回归建立回归模型,用ROC曲线分析三指标对肺癌诊断的意义。结果肺癌组CEA、NSE和CYFRA21-1的水平显著高于肺良性疾病和健康人组(P<0.01)。腺癌组血清CEA水平最高,NSE在小细胞肺癌中灵敏度最高(81.6%),取特异性90%时,CYFRA21-1对肺癌的诊断灵敏度最高(59.5%)。建立回归模型Y=1/[1+EXP(5.830-0.249X1-0.198X2-0.643X3)],新变量Y的灵敏度、特异性和准确性率分别为80.9%,91.3%和84.6%。结论血清CEA、NSE和CYFRA21-1对肺癌的诊断具有较高的价值,综合运用Logistic回归和ROC曲线可以提高其肺癌诊断价值。

Objective To explore the application of serum carcinoembryonic antigen （CEA）, neuron-specificenolase （NSE） and cytokeratin 19 fragment （CYFRA21-1） for the diagnosis of lung cancer. Methods The levels of serum CEA,NSE and CYFRA21-1 in patients with lung cancer,patients with benign lungs diseases were determined by electrochemiluminescence （ECLIA）. Logistic regression and ROC curve were applied to analyze the data and evaluate the diagnostic values. Results The concentrations of serum CEA,NSE and CYFRA21-1 in patients with lung cancer were significantly higher than the other two groups（P0.01）.The serum level of CEA was highest in patients with adenocarcinoma.NSE was the most sensitive tumor marker（81.6%）for small cell lung cancer.When the specificity was 90%,CYFRA21-1 had the highest sensitivity in lung cancer. According to the regression model,Y=1/[1＋EXP（5.830-0.249X1-0.198X2-0.643X3）], the sensitivity, specificity and accuracy of the new variable Y was 80.9%,91.3% and 84.6%,respectively. Conclusion Serum CEA,NSE and CYFRA21-1 in lung cancer has a high diagnostic value.The use of Logistic regression and ROC curve can improve the diagnosis of lung cancer.