Selective nonoperative management of leaks after gastric bypass: Lessons learned from 2675 consecutive patients

Objective: To compare outcomes of patients with leaks after primary Roux-en-Y gastric bypass (GBP) managed operatively with those managed nonoperatively and subsequently derive indications for selective nonoperative management. Summary of Background Data: There is no consensus on the management of leaks complicating GBP, which remains the commonest cause of death. Methods: We evaluated 2675 consecutive GBP procedures, determining incidence and outcomes of leaks in a program emphasizing early detection, routine drainage, and selective nonoperative management. Results: Leaks occurred in 46 patients (41 women) with mean (±SD) age of 46.9 ± 8.7 years, weight and body mass index (BMI) of 307.8 ± 56.9 lb and 51.2 ± 9.5 kg/m, respectively. Leaks were initially identified by upper gastrointestinal contrast swallow (UGI) on the first postoperative day (22), abnormal drain output (11), delayed UGI (3), or on clinical suspicion (10) with a respective interval to diagnosis of 1.1*, 6.5, 7, and 7.9 days (*P < 0.007 vs. other groups). Leaks were located in the gastrojejunal (GJ) anastomosis (37), gastric pouch (4), gastric remnant (2), jejuno-jejunostomy (1), Roux limb (1), and cervical esophagus (1), and were radiologically contained (40) or diffuse (3) or not demonstrable (3). Contained leaks were treated nonoperatively (31), by operation (7), or required no treatment (2). Patients with diffuse leaks or bilious drain output were operatively managed. They were similar in duration for nil per oral order, drain and antibiotic use and readmission rates, whereas hospital stays were longer in the operative group, P < 0.01. There were no deaths. Conclusions: Many leaks after gastric bypass are radiologically contained GJ and pouch leaks and can be safely managed nonoperatively. Radiologic features and bilious drainage were key determinants of treatment, with operative treatment used for diffuse GJ leaks, bilious drainage, or clinical suspicion with a negative UGI. Outcomes were similar in both groups.