BACKGROUND: Sclerotherapy is widely used for acute variceal bleeding, although in emergencies bleeding makes it difficult to obtain the clear view required for safe and effective treatment. We investigated whether early administration of somatostatin would improve the efficacy of sclerotherapy. METHODS: In this double-blind, prospective trial, patients who had cirrhosis with upper-gastrointestinal bleeding were randomly assigned natural somatostatin (6 mg per 24 h) or placebo for 120 h. In addition, intravenous bolus doses of somatostatin (250 micrograms) or placebo were injected after the start of the infusion, before emergency endoscopy or sclerotherapy, and when active bleeding was observed. The primary endpoint was treatment failure, defined as the occurrence during the infusion period of at least one of: excess transfusion of blood products, haematemesis, haemodynamic instability, need for rescue therapy, or death. FINDINGS: 205 patients were enrolled: 101 received somatostatin and 104 received placebo. Treatment failed in 35 somatostatin and 57 placebo recipients (p = 0.004); death or use of rescue therapy occurred in nine and 19 patients, respectively (p = 0.05). The mean quantity of blood products transfused over 120 h (adjusted for baseline haemoglobin) was 2.64 (SD 0.35) units in the somatostatin group versus 3.62 (0.35) units in the placebo group (p = 0.05). At endoscopy, active bleeding from oesophageal varices was less frequent (27 vs 42 patients, p = 0.012) and the sclerotherapy procedure was easier (2.8 vs 4.7 cm, p = 0.0027) in the somatostatin than in the placebo group. INTERPRETATION: Early administration of natural somatostatin continued for 120 h, combined with additional bolus injections, is more effective than placebo in the overall control of acute variceal haemorrhage in patients with cirrhosis undergoing sclerotherapy.