Over a period of 10 months, we conducted a prospective randomized controlled trial of endoscopic injection for the arrest of peptic ulcer hemorrhage in 84 patients. We injected pure alcohol (PA, 99.8%), 50% glucose in water (G/W), 3% NaCl solution, and normal saline solution (N/S, controls) to stop bleeding. The ultimate success rates in the four groups were: 81% (17/21) in the PA group, 80% (16/20) in the 50% G/W group, 65.2% (15/23) in the 3% NaCl group, and 80% (16/20) in the N/S group (p greater than 0.05). The volume of blood transfusion (mean +/- s. e. m.) for each patient was 1,200 ml +/- 433 ml in the PA group, 1,130 +/- 356 ml in the 50% G/W group, 1,704 +/- 330 ml in the 3% NaCl group, 1,400 +/- 548 ml in the N/S group (p greater than 0.05). The duration of hospitalization (mean +/- s. e. m.) was 5.7 +/- 1.5 days in the PA group, 3 +/- 1.3 days in the 50% G/W group, 6.1 +/- 1.1 days in the 3% NaCl group, and 5 +/- 1.1 days in the N/S group (p greater than 0.05). No complication of perforation or aspiration pneumonia was observed during hospitalization. We suggest that endoscopic injection with the above solutions be used as the first line of therapeutic endoscopy for arrest of peptic ulcer hemorrhage.