Endoscopic injection was not feasible in 14 of 120 (11.6%) patients because of inaccessibility or massive hemorrhage. The remaining 106 patients underwent endoscopic therapy by injection of adrenaline and polidocanol. The efficacy (achievement of definitive hemostasis) of endoscopy therapy was 83% (88 of 106). Univariate analysis showed that failure of endoscopic injection was related to age, presence of shock, ulcer size greater than 2 cm, and hemoglobin level. Multivariate analysis showed that ulcer size greater than 2 cm (p = 0.005) and the presence of shock (p = 0.03) were factors predictive of endoscopic treatment failure. Failure to achieve hemostasis (p < 0.001) and poor physical status measured by American Society of Anesthesiology classification (p = 0.02) were significantly related to mortality. Conclusions: Ulcer size and severity of hemorrhage are predictive of endoscopic injection failure in patients bleeding from high-risk duodenal ulcers. Survival is determined by clinical status and associated diseases. (Gastrointest Endosc 1996;43:111-6.)