Background: To explore the diagnostic value of combination of exfoliative cytology with detection of tumor markers carbohydrate antigen 125 (CA125), carcinoembryonic antigen (CEA), neuron specific enolase (NSE), cytokeratin 19 fragment antigen 21-1 (CYFRA21-1) and CA15-3 for lung cancer. Methods: A total of 256 patients were enrolled, including 164 males and 92 females aged (64.51 +/- 22.68) years old. Among them, 189 patients (100 males and 89 females) were randomly selected as Tumor group, and the remaining 67 patients were used for validation. Another 514 healthy people receiving physical examination in our hospital during the same period were selected, from which 397 cases (266 males and 131 females) were randomly selected as No Tumor group, and the remaining 117 cases were used for validation. The biochemical criteria were detected in all subjects. The diagnostic value of each index for lung cancer was analyzed using receiver operating characteristic (ROC) curves. Results: The results of ROC curve analysis revealed that in Tumor group, the area under curve (AUC) of exfoliative cytology, CA125, CYFRA21-1, CA15-3, CEA and NSE was >= 0.7, while that of CA72-4, CA19-9, TSGF, AFP, CA242, SCCAg and CA50 was <0.7. The indices in each factor were comprehensively assessed, and then exfoliative cytology, CA125, CA15-3, CYFRA21-1, CEA and NSE were screened to establish the lung cancer prediction model. The diagnostic value was comparable between the prediction model and the combined detection of 9 indices (Z=1.682, P=0.079). Conclusions: The lung cancer prediction model balances sensitivity and specificity without reducing the diagnostic efficiency.
