Development and Multiple Validation of the Protein Multi-marker Panel for Diagnosis of Pancreatic Cancer

Abstract

Purpose: To develop and validate a protein-based, multi-marker panel that provides superior pancreatic ductal adenocarcinoma (PDAC) detection abilities with sufficient diagnostic performance. Experimental Design: A total of 959 plasma samples from patients at multiple medical centers were used. To construct an optimal, diagnostic, multi-marker panel, we applied data preprocessing procedure to biomarker candidates. The multi-marker panel was developed using a training set comprised of 261 PDAC cases and 290 controls. Subsequent evaluations were performed in a validation set comprised of 65 PDAC cases and 72 controls. Further validation was performed in an independent set comprised of 75 PDAC cases and 47 controls. Results: A multi-marker panel containing 14 proteins was developed. The multi-marker panel achieved areas under the curve (AUCs) of 0.977 and 0.953 for the training set and validation set, respectively. In an independent validation set, the multi-marker panel yielded an AUC of 0.928. The diagnostic performance of the multi-marker panel showed significant improvements compared with carbohydrate antigen (CA) 19-9 alone (training set AUC = 0.977 vs. 0.872, P < 0.001; validation set AUC = 0.953 vs. 0.832, P < 0.01; independent validation AUC = 0.928 vs. 0.771, P < 0.001). When the multi-marker panel and CA 19-9 were combined, the diagnostic performance of the combined panel was improved for all sets. Conclusions: This multi-marker panel and the combined panel showed statistically significant improvements in diagnostic performance compared with CA 19-9 alone and has the potential to complement CA19-9 as a diagnostic marker in clinical practice.