Seventy five patients with severely bleeding peptic ulcer were included in a controlled comparative trial to assess the efficacy and safety of endoscopic injection therapy in preventing rebleeding from peptic ulcers that presented at endoscopy with a protruding vessel. Twenty five patients were treated with injection of epinephrine followed by polidocanol, 25 were treated with injection of absolute alcohol, and 25 with sham injection. Rebleeding occurred in 44% of patients in the sham group, 40% of those treated with epinephrine and polidocanol, and in 20% of those treated with absolute ethanol. The difference in the haemostasis rate between the control and ethanol treated subjects nearly reached significance (p = 0.07). A second therapy session resulted in haemostasis rates of 68% in the epinephrine-polidocanol group and of 88% in the absolute ethanol group. These rates after two treatments as well as the emergency surgery rates (32% in the epinephrine-polidocanol group and 8% in the absolute ethanol group; p = 0.07) were not significantly different. In eight of the 11 patients with rebleeding in the sham treatment group, definitive haemostasis was achieved by elective injection therapy. Overall transfusion requirements were mean (SD) 6.0 (0.7) units in the sham group, 6.0 (0.9) in the epinephrine-polidocanol group, and 3.9 (0.5) in the absolute ethanol group. Only the difference between ethanol and sham was significant (p = 0.02). This study shows that injection with absolute ethanol reduces rebleeding in these patients and significantly lowers transfusion requirements. Absolute ethanol was superior to epinephrine-polidocanol, which was not significantly better than sham therapy.