A panel of autoantibodies against tumor-associated antigens in the early immunodiagnosis of lung cancer

Objective: Lung cancer (LC) is one of the most common malignant tumors worldwide with low five-year survival rate due to lack of effective diagnosis. This study aims to find an optimal combination of autoantibodies for detecting of early-stage LC. Methods: Nine relatively novel autoantibodies against tumor-associated (TAAs) (PSIP1, TOP2A, ACTR3, RPS6KA5, HMGB3, MMP12, GREM1, ZWINT and NUSAP1) were detected by using ELISA. Diagnostic models were developed by using the training set (n = 644) and further validated in another independent set (n = 248). We also evaluated the diagnostic accuracy of the model to detect benign lung diseases (BLD) from the early-stage lung cancer. Results: The areas under the receiver operating characteristic curve (AUC) for the model with three TAAs panel (GREM1, HMGB3 and PSIP1) was 0.711(95% CI 0.674-0.746) in the training set and 0.858 (95% CI 0.808-0.899) in the validation set, which demonstrated a higher diagnostic capability. The AUC of this three TAAs model was 0.833 (95%CI 0.780-0.878) in discriminating LC from BLD. This model could identify early-stage LC patients from normal control (NC) individuals, with AUC of 0.687(95% CI 0.634-0.736) in training set and AUC of 0.920(95% CI 0.860-0.960) in validation set, and the overall AUC for early-stage LC was 0.779(95% CI 0.739-0.816) when the training set and validation set were combined. Conclusions: The model with three TAAs panel would detect LC with higher effectiveness, and might be potential screening method for the early LC.