Background:
The ability to accurately identify the diagnosis of cholangiocarcinoma(CCA) for patients with obstructive jaundice would facilitate preventative and therapeutic Endoscopic retrograde cholangiopancreatography (ERCP) interventions delivered to proper patients. Hence, we aimed to develop, validate, and evaluate a diagnostic utility online model combining blood laboratory variables and imaging omics data to predict CCA.
Methods
From 2018 to 2022, consecutive jaundice patients who got ERCP from Gansu Provincial Hospital were enrolled retrospectively. Random forest(RF) regression was used to screen variables and logistic regression was used to reconfirm the risk factors and build the online nomogram model. Corrected linear correlation analysis was employed for the linear test. The model predictive performance was assessed by the receiver operating characteristic (ROC) curve as well as calibration, clinical decision analysis(DCA), and impact curves(CIC). An external cohort of 144 patients was validated. The spyglass subgroup application was analyzed in CCA patients.
Results
A total of 506 obstructive jaundice patients were enrolled (93 patients(18.38%) with CCA). The integrated model identified 6 factors including red blood cell count(RBC), total bilirubin (TBIL), uric acid (UA), total cholesterol (TC), bile.duct.diameter, and Spyglass operation. The area under the curve (AUC) of the ROC was 0.929 (95% CI, 0.903–0.950). The calibration and clinical decision or impact curves showed good predictive accuracy. A validation AUC of 0.948 (95% CI, 0.899–0.978) was obtained in the external validation set.
Conclusions
We presented a promising model to endoscopic individualize cholangiocarcinoma diagnosis for Jaundice patients especially those with ERCP and spyglass with a good discriminative ability.