Diagnostic Value Analysis of Serum NSE, CEA, CYFRA21-1 for Lung Cancer Based on ROC Curve and Logistic Regression

Objective To investigate the diagnostic values of tumor markers （NSE、CEA、CYFRA21-1）for lung cancer with receiver operating characteristic （ROC）curve and logistic regression （LR）analysis. Methods The serum concentrations of NSE、 CEA and CYFRA21-1 were measured by chemiluminescent or electrochemica immunoassay. The area under the ROC curve（AUC）, sensitivity, specificity and Youden’s index were calculated and compared by SPSS 16.0 and MedCale software. Results In patients with lung cancer, the ideal threshold point identified from the ROC curve for NSE was >18.48 ng/mL, with a sensitivity of 55.7% （95%CI 53.0%~58.4%） and a specificity of 63.1%（95%CI 56.6%~69.2%）. The ideal threshold point for CEA is >3.36 ng/mL, with a sensitivity of 60.0% （95%CI 57.6%~62.3%） and a Specificity of 79.6%（95%CI 75.7%~83.1%）. The ideal threshold point based on ROC curve for CYFRA21-1 was >4.4 ng/mL, with a Sensitivity of 47.4%（95%CI 45.0%~49.8%） and a Specificity of 83.7%（95%CI 79.3%~87.4%）. Combined two or three indexes can increase AUC significantly. NSE has the best AUC（0.938） in small cell lung cancer, CEA has the best AUC （0.793） in adenocarcinoma, CYFRA21-1 has the best AUC（0.843） in squamous carcinoma. Conclusions Combined analysis three indexes indicated an increase in AUC for lung cancer diagnostis.