Abstract
Background: Endoscopic retrograde cholangiopancreatography (ERCP) in patients with Roux-en-Y gastric bypass (RYGB) remains technically challenging. Device-assisted ERCP (DAE-ERCP) is widely used for uncomplicated pancreaticobiliary disease but is associated with prolonged procedure times and high failure rates. Endoscopic ultrasound-directed transgastric ERCP (EDGE) offers high technical success but introduces additional cost and the risk of a persistent fistula. We aimed to prospectively identify intra-procedural predictors of DAE-ERCP failure and define an actionable, real-time threshold for early cross-over to EDGE. Methods: We prospectively evaluated consecutive RYGB patients undergoing DAE-ERCP at a tertiary referral center. Patients with established pre-procedural features associated with complex or low-yield DAE-ERCP were triaged directly to EDGE and excluded. Intra-procedural variables were recorded in real time. Univariate and multivariable logistic regression identified predictors of DAE-ERCP failure. Received operating characteristic (ROC) analysis determined optimal cutoffs for cross-over. Results: A total of 94 patients with RYGB underwent ERCP. Amongst these, 42 patients (11 males, 31 females) were included in the analysis and underwent DAE-ERCP with a success rate of 73.8% (n = 31). Significant risk factors of DAE-ERCP failure included excessive resistance to scope advancement (p < 0.0001), failure to reach the ampulla (p < 0.0001), patient position (p = 0.009), BMI (p = 0.004), and time to reach the jejuno-jejunal (J-J) anastomosis (p < 0.0001). Additionally, time needed to reach the J-J anastomosis of ≥23 min [OR 1.360 (95% CI: 1.079–1.713), p = 0.009], excess resistance to scope advancement [OR 2.223 (95% CI: 2.001–4.167)], and failure to reach the ampulla [OR 9.929 (95% CI: 2.004–4.033)] were statistically significant predictors of DAE-ERCP failure. When ≥2 predictors of BA-ERCP failure were present, the likelihood of DAE-ERCP failure was 2.370 with 95.50% sensitivity and 90% specificity with AUC= 0.929 (p = 0.0001). Conclusions: DAE-ERCP remains an effective first-line strategy in appropriately selecting RYGB patients without pre-procedural high-risk features. Real-time intra-procedural predictors can reliably identify impending failure. A structured algorithm incorporating both pre-procedural triage and intra-procedural checkpoints supports timely transition to EDGE, optimizing efficiency, safety, and resource utilization.