Ninety-seven patients with recent or active var-iceal bleeding were randomly assigned to oral pro-pranolol, endoscopic sclerotherapy plus oral pro-pranolol, or transhepatic sclerotherapy plus oral propranolol. The effects of treatment on the number of units transfused, rebleeding of any magnitude, major rebleeding, and death were assessed in these patients, 82% of whom were alcoholic and 81% Child's Class C. After a minimum follow-up interval of 2 yr (range, 27-65 mo), major rebleeding rates were 65% for propranolol alone, 45% for endoscopic sclerotherapy plus propranolol, and 60% for transhepatic sclerotherapy plus propranolol. The corresponding death rates were 81% for propranolol alone, 55% for endoscopic sclerotherapy plus propranolol, and 66% for transhepatic sclerotherapy plus propranolol (p = 0.03). Thirty-three patients (34%) never received propranolol; 8 due to medical contraindications and 25 because they died or bled enough to meet the definition of treatment failure within 3 or 4 days of randomi- zations (no significant differences among treatment groups). Patients assigned to propranolol alone bled sooner, bled more units, and had a higher mortality rate than patients treated by endoscopic sclerotherapy plus propranolol. Patients treated with transhepatic sclerotherapy plus propranolol had intermediate results. Propranolol alone is inadequate treatment for esophageal var-iceal bleeding in patients with advanced liver disease.