Obstructive jaundice with rapid onset and complex etiology is a common gastroenterology disease. The proper decision of the clinical treatment methods is fatal, but critical surgical treatment is often not timely and effective[4, 14]. Currently, ERCP is an effective minimally invasive method for diagnosing and treating liver, gallbladder, and pancreatic diseases[15]. The Spyglass technology is a further complement and improvement in real-time diagnosing and treating benign, which can effectively solve some of the hard-to-solve limitations in the operation such as the anatomy and the angle of operation disadvantages[6, 16]. For the diagnosis and treatment of malignant jaundice, there are currently some clear risk factors, such as cirrhosis and viral hepatitis C and B for intrahepatic disease, primary sclerosing cholangitis, and bile duct cystic disorders like Caroli’s disease.
However, there is no specific diagnostic model for malignant CCA combined with endoscopic and clinical laboratory data[2, 17]. Therefore, we aimed to construct a predictive model employing RF algorithms for CCA and develop a public online nomogram platform to improve endoscopic prognosis[3, 8, 9].
This study analyzed the baseline characters and found that there were statistically significant differences in baseline including age and smoking history before ERCP. Then we used 1:1 PSM before spyglass operation to reduce confounding bias effectively. The nomogram included TBIL, UA, TC, Bile.duct.diameter and RBC can strongly discriminate CCA[13, 15]. The model has an AUC of 0.929, a sensitivity of 81.72%, and a specificity of 88.14%. In addition, the p-value of the Hosmer Lemeshow test was 0.749, respectively. The parameters of DCA, CIC, and calibration curves revealed that our integrated model has excellent predictive accuracy. External validation also proved that the integrated model has a better predictive ability[18]. Then, we established a convenient and accurate prediction nomogram tool that could be utilized in clinical decision-making.
As for risk factors, when UA is higher than 385 (sensitivity 93.00%, specificity 25.33%), the probability of CCA will increase and the OR has been applied. Previous studies showed that UA could inhibit the proliferation of CCA cells, and its regulatory mechanism may be by downregulating ATM and BRCA1 signaling molecules to affect homologous chromosome recombination in tumor cells[19].
Bile duct dilation is the most preliminary and intuitive imaging indicator of CCA, so the diameter > 12mm a with an AUC of 0.794,( sensitivity 65.59%, specificity 74.33%). Most patients diagnosed with extrahepatic CCA (ECCA) exhibit cholestasis caused by obstruction of the bile duct. CCA patients may have lipid disorders and exhibit transiently elevated TC and sdLDL-C levels and preliminary falsely low LDL-C results[16]. TC, sdLDL-C, and LDL-C levels could be restored to normal levels after biliary obstruction removal and cholestasis relief. Therefore, the level of TC would typically be higher than 4.6 with an AUC of 0.781 (sensitivity 69.89%, specificity 74.82%) in malignant tumor obstruction patients[10].
Previous studies illustrated preoperative bilirubin level may effectively reflect the severity of perihilar CC and as an important prognostic factor after surgery for perihilar CC patients to some extent [8, 20]. A multicenter European study also showed that a high preoperative total bilirubin level (≥ 3mg/dL) was significantly associated with increased complications after major hepatectomy for perihilar CC. So our study suggested that Total bilirubin > 147.1 with an AUC of 0.789 (sensitivity 76.34%, specificity 73.85%), which may be caused by biliary obstruction and liver dysfunction, is consistent with former studies[20, 21].
Red blood count (RBC) is a routine laboratory blood test that indicates the heterogeneity in the number of circulating erythrocytes especially for inflammatory diseases. Beyazit Y et al reported that RDW can be taken as a prospective diagnostic factor for malignant obstructive jaundice[22]. The possible mechanisms might involve systemic influence in inflammatory response and malnutrition including inflammatory reactions and inflammatory factors affecting iron metabolism, which shortened the life of red blood cells. So when RBC < 4.33(AUC,0.640, sensitivity 62.37%, specificity 60.29%), it would improve the diagnostic rate of CCA even in rare tumors[2, 23].
To our knowledge, this is the first endoscopic diagnostic model matched with spyglass data for malignant CCA. The validated nomogram showed a high predictive value through the calibration and accuracy test. By the nomogram, the AUC for 5 variables for CCA prediction was 0.907 (95% CI, 0.877–0.931), and the p-value of the De Long test is 0.001, which is superior to any single prediction model. The subgroup further analyzed and validated the online nomogram owed a high predictive value combined spyglass parameter. By the integrated model, the AUC for 6 variables was 0.907 (95% CI, 0.881–0.929), and the p-value of the De Long test was 0.001, which demonstrated spyglass had a superior ability than the single model.
Moreover, our analysis incorporated restricted cubic spline methodology to evaluate linear correlation factors for CCA diagnosis. This approach revealed a nonlinear distribution for TBIL, which suggested that the elevation of TBIL before ERCP should avoid empirical bias in clinical prediction practice.
There are still some limitations in our study. In the chosen patients, only ERCP group patients from obstructive jaundice were enrolled instead of all kinds patients. In addition, this single-center retrospective study has an inherent bias in patient selection and data collection. For future modeling validation and development, multicenter and large-scale prospective studies will be conducted.
In conclusion, TBIL, UA, TC, Bile.duct.diameter, and RBC are potential independent factors for the diagnosis of CCA. The integrated nomogram combined spyglass, as a less invasive and convenient alternative, can accurately predict CCA from acute jaundice patients, especially for the ERCP operation to select surgical treatment.