A nomogram for prediction of ERCP success in patients with bile duct leaks: a multicenter study

Bile duct leaks (BDLs) are serious complications that occurs after hepatobiliary surgery and trauma, leading to rapid clinical deterioration. Endoscopic retrograde cholangiopancreatography (ERCP) is the first-line treatment for BDLs, but it is not clear which patients will respond to this therapy and which patients will require additional surgical intervention. The aim of our study was to explore the predictors of successful ERCP for BDLs.

A retrospective analysis was conducted using data from six centers' databases. All consecutive patients who were clinically confirmed as BDLs were included in the study. Collected data were demographics, disease severity, and ERCP procedure characteristics. Univariate and multivariate analysis were used to select independent predictive factors that affect the outcome of ERCP for BDLs, and a nomogram was established. Calibration and ROC curves were used to evaluate the models.

Four hundred and forty-eight consecutive patients were clinically confirmed as BDLs and 347 were excluded. In the 101 patients included patients, clinical success was achieved in 78 patients (77.2%). In logistic multivariable regression, two independent factors were negatively associated with the success of ERCP: SIRS (OR, 0.183; 95% CI 0.039–0.864; P = 0.032) and high-grade leak (OR 0.073; 95% CI 0.010–0.539; P = 0.010). Two independent factors were positively associated with the success of ERCP: leak-bridging drainage (OR 4.792; 95% CI 1.08–21.21; P = 0.039) and cystic duct leak (OR 6.193; 95% CI 1.03–37.17; P = 0.046). The prediction model with these four factors was evaluated using a receiver-operating characteristic (ROC) curve, which demonstrated an area under the curve of 0.9351. The calibration curve showed that the model had good predictive accuracy.

Leak-bridging drainage and cystic duct leak are positive predictors for the success of ERCP, while SIRS and high-grade leak are negative predictors. This prediction model with nomogram has good predictive ability and practical clinical value, and may be helpful in clinical decision-making and prognostication.