Construction and validation of a multiparameter diagnostic model based on conventional tumor markers and cytokines for lung cancer

BACKGROUND Early detection of lung cancer is urgently needed in clinical practice. AIM To evaluate the diagnostic value of conventional tumor markers and cytokines for lung cancer and construct a multiparameter diagnostic model lung cancer detection. METHODS A total of 152 healthy controls and 113 lung cancer patients were included in the model. In addition, 21 healthy controls and 36 lung cancer patients were separately included to validate the model. Three conventional tumor markers and 10 cytokines were detected. Four multiparameter joint analysis methods, binary logistic regression analysis, discriminant analysis, a classification tree and a neural network, were used to establish and compare multiparameter joint diagnosis models. RESULTS Six differentially expressed indicators [carcinoembryonic antigen (CEA), cytokeratin 19 fragment (CY211), neuron-specific enolase, interleukin (IL)-8, monocyte chemoattractant protein-1, and tumor necrosis factor-alpha (TNF-α)] were screened out, among which IL-8 [area under the curve (AUC) = 0.957] and TNF-α (AUC = 0.936) had the optimal diagnostic efficacy. The binary logistic regression model was chosen as the optimal multiparameter combined auxiliary diagnostic model. When 152 healthy controls and 113 lung cancer s were differentiated via the model, the AUC was 0.980. After validation, when 21 healthy controls and 36 lung cancer patients were distinguished, the AUC was 0.922, indicating good stability and superior performance to that of CEA alone. CONCLUSION We constructed a multiparameter binary logistic regression diagnostic model that included CEA, CY211, IL-8 and TNF-α for the auxiliary detection of lung cancer. Compared with conventional CEA, it significantly improved diagnostic accuracy.