Résumé : Background/Aims: Acute hemorrhage of the upper gastrointestinal tract occurs at a rate of 50 to 100 per 100,000 annually in the Western adult population. With the increased use of therapeutic endoscopy, the role of surgery is decreasing; surgical intervention is now only used in cases of failure of endoscopic hemostasis. The goal of this study is to determine whether there are predictive factors associated with high-risk post-operative mortality. Methodology: This retrospective study included 30 patients treated from March 1996 to September 2008 at Brugmann Hospital. These patients presented with upper gastrointestinal non-variceal hemorrhage that was treated first endoscopically then surgically for recurrent hemorrhage. Multiple risk factors (variable and fixed) and parameters were evaluated to determine their influence on mortality. Results: Of 30 patients, 10 (33%) developed recurrent hemorrhage following surgical treatment. A total of 8 (26.6%) deaths occurred of which 4 were related to hemorrhage. Three deaths occurred after the first intervention and 5 occurred after a second intervention. Logistic regression analysis revealed that the total number of blood units transfused and the presence of at least one surgical reintervention both significantly increased mortality rate (p=0.0426 and p=0.0068). Other parameters were not sig-nificant. However, there is a lack of power due to the small sample size. Conclusion: For recurrent massive upper gastrointestinal hemorrhage following endoscopic treatment and necessitating more than 19 blood transfusions, early surgical intervention is recommended and surgical reintervention should be avoided. If reintervention is necessary, radical surgery is recommended. However, the small number of patients treated over a 12-year period limits the results of this study, and these results may represent simple coincidences. © H.G.E. Update Medical Publishing S.A.