Leaks after Roux-en-Y gastric bypass(GB)are a major cause of mortality. This study attempts to define the relationship between the leak site, detection time, method of diagnosis, and mortality rate. Methods: Analysis of 3838 patients in the GB database identified 150 patients with leaks. Statistical tests included chi square, Fisher’s and Wilcoxon score. Results: Of the leaks (3.9% overall), 60/2276 (2.57%) occurred after open gastric bypass (OGB), 57/1024 (5.27%) after laparoscopic gastric bypass (LGB), and 33/378 (8.03%) after revisions(RGB). Leaks were more common in older patients (45.3 ±10.4 vs. 40.6 ±10.3 years, p<0.0001) and males (28.7% vs. 18.2%, p=0.003). Overall mortality rate due to leaks in OGB and LGB was similar [0.64% (15/2336) vs.0.46% (5/1081), p=0.5345], and accounted for half (20 deaths) of all mortality. In OGB, there were 39 leaks (1.7%) at the gastrojejunostomy (GJ), and 11 (0.47%) at the jejunojejunostomy (JJ). In the LGB group, there were 44 leaks (4.1%) at the GJ and 5 (0.46%) at the JJ. Eight LGB leaks occurred elsewhere with 3 from the excluded stomach. Mortality from a GJ leak was higher in OGB than LGB (17.9 vs. 0%, p=0.003)although there was one death at an outside hospital in an LGB patient from a leak of undetermined site. Sixty-eight percent of GJ leaks required reoperation (27/37 OGB, 28/43 LGB). There were no deaths in the nonoperative group. Detection time for a GJ leak in the OGB group was longer than in the LGB group (3 vs 1 days, Wilcoxon score p<0.0001, mean 5.0 ± 6.7 days vs. 2.2 ± 2.6 days, p=0.02). Mortality and detection time after JJ leak was similar in both OGB and LGB. JJ leak was associated with a 40% mortality rate (vs. 8.1% GJ, p=0.003). Initial upper G.I. (UGI) missed 9/10 JJ leaks compared to 16/88 GJ leaks (90.0% vs. 18.2%, p=0.00001). Median detection time was longer in the JJ leak group than the GJ leak group (4 vs 2 days, p=0.033 Wilcoxon). Tachycardia (p>120) was seen in 65% of GJ leaks and 83% of JJ leaks (p=NS). Patients with JJ leaks had a higher BMI than those with GJ leaks(BMI 55.25+/-10 vs. 48.81+/-10 p= 0.022). Discussion: Leak mortality and time of detection was higher in OGB than LGB. JJ leaks are more lethal than GJ leaks and take longer to detect. GBP patients with persistent tachycardia may harbor leaks, especially at the JJ or excluded stomach, even with normal UGI studies. Normal UGI findings should not delay therapy if clinical signs suggest leak.