Search

Abstract Background To date, carbohydrate antigen 19‐9 (CA19‐9) and carcinoembryonic antigen (CEA) have been widely used for the screening, diagnosis and prediction of biliary tract cancer (BTC) patients. However, few studies with large sample sizes of carbohydrate antigen 50 (CA50) were reported in BTC patients. Methods A total of 1121 patients from the Liver Cancer Clin‐Bio Databank of Anhui Hepatobiliary Surgery Union between January 2017 and December 2022 were included in this study (673 in the training cohort and 448 in the validation cohort): among them, 458 with BTC, 178 with hepatocellular carcinoma (HCC), 23 with combined hepatocellular‐cholangiocarcinoma, and 462 with nontumor patients. Receiver operating characteristic (ROC) curves and decision curve analysis (DCA) were used to evaluate the diagnostic efficacy and clinical usefulness. Results ROC curves obtained by combining CA50, CA19‐9, and AFP showed that the AUC value of the diagnostic MODEL 1 was 0.885 (95% CI 0.856–0.885, specificity 70.3%, and sensitivity 84.0%) in the training cohort and 0.879 (0.841–0.917, 76.7%, and 84.3%) in the validation cohort. In addition, comparing iCCA and HCC (235 in the training cohort, 157 in the validation cohort), the AUC values of the diagnostic MODEL 2 were 0.893 (95% CI 0.853–0.933, specificity 96%, and sensitivity 68.6%) in the training cohort and 0.872 (95% CI 0.818–0.927, 94.2%, and 64.6%) in the validation cohort. Conclusion The model combining CA50, CA19‐9, and AFP not only has good diagnostic value for BTC but also has good diagnostic value for distinguishing iCCA and HCC.