Su1642 Predictors of Failure of Endoscopically Managed Post-Operative Bile Duct Leaks: a Large, Multi-Center Retrospective Study

Su1642 Predictors of Failure of Endoscopically Managed Post-Operative Bile Duct Leaks: a Large, Multi-Center Retrospective Study Anoop Prabhu*, Sunil Amin, Brian Rajca, Dennis Yang, Shailendra S. Chauhan, Peter V. Draganov, Farah Monzur, Laila Menon, Jonathan Buscaglia, Satish Nagula, Juan Carlos Bucobo, Lionel S. D, Petros C. Benias, David L. Carr-Locke, Nazia Hasan, ADAM J. Goodman, Amy Tyberg, Shireen a. Pais, Jawad Ahmad, Susana Gonzalez, Christopher J. Dimaio Gastroenterology, Mount Sinai School of Medicine, New York, NY; Gastroenterology and Hepatology, University of Florida, Gainesville, FL; Gastroenterology, Stony Brook University Medicine, Stony Brook, NY; Gastroenterology, Beth Israel Medical Center, New York, NY; Gastroenterology, NYU Langone Medical Center, New York, NY; Gastroenterology and Hepatobiliary DIseases, New York Medical College, Valhalla, NY Background: Post-operative biliary leaks are a common reason for referral for ERCP. Data is lacking regarding the variables that best predict persistence of these leaks after initial endoscopic therapy. Aims: Assess current practices in endoscopic management of bile leaks and identify predictors of failed resolution at follow-up ERCP. Methods: A multi-center retrospective review was performed of post-surgical biliary leaks referred to ERCP. Case-specific data was extracted on demographics, surgical details, cholangiogram findings, and endoscopic interventions, including use of stent and biliary sphincterotomy. Exclusion criteria were age!18, non-native biliary anatomy, and liver transplant. Data were stratified by persistence of leak at follow-up ERCP. Univariate analysis was performed using chi-squared test for categorical and student t-test for continuous variables. Variables whose univariate p-values were! 0.2 were used in a multivariate model. Results: A total of 219 cases of post-operative bile leaks were identified. The mean age was 55 17.6 years with 51.6% being female (Table 1). The mean time to leak identification was 12.8 28.1 days. The most common means of leak identification was bilious surgical drain output (43.4%). At index ERCP, the most frequent leak site was the cystic duct (43.8%). Patients underwent biliary sphincterotomy alone in 11 cases (5.0%), stenting alone in 50 cases (21.9%), both in 156 cases (71.2%), and neither in 2 cases. Of the 167 patients who underwent sphincterotomy, 26 (15.6%) had choledocholithiasis. Post-ERCP pancreatitis was the most common complication (1.83%). Follow-up ERCP data was available in 151 cases. Mean time between ERCPs was 49 28 days. Resolution of leak occurred in 125 (82.8%) cases. On univariate analysis, predictors of persistent leak were: non-cholecystectomy (p!.001), post-surgical drain placement (pZ.01), leak site (pZ.03), and fewer days between ERCPs (p!.001) (Table 2). On multivariate analysis, factors predictive of persistent leak were added days to leak identification (9.3 11.8 days in leak resolution vs 22.4 51.1 days in leak persistence group; OR 1.04, 95% CI 1.00-1.08) and fewer days between ERCPs (52 25.8 days in leak resolution vs 31.5 24.3 days in leak persistence group; OR 0.85, 95% CI 0.730.99). Age, gender, surgery indication/type, drain placement, leak site, sphincterotomy, or stent choice/location were not predictive of leak persistence on multivariate analysis. Conclusions: The majority of post-surgical bile leaks were treated with combination stent placement and biliary sphincterotomy. Though most leaks resolve with endoscopic management, failure of leak resolution is associated with increased time to leak identification and decreasing time between index and followup ERCP. Neither stent choice nor sphincterotomy predict persistence of leak at follow-up ERCP.