In this prospective, multicenter trial, 140 cirrhotic patients with no previous upper gastrointestinal bleeding and with esophageal varices endoscopically judged to be at high risk of hemorrhage were randomized to receive either sclerotherapy or conservative treatment for the prevention of first variceal bleeding. The end-points of the study were bleeding and death. Life table curves showed that prophylactic sclerotherapy significantly diminished the incidence of variceal hemorrhage (p less than 0.001) and overall mortality (p less than 0.01). Two-year cumulative bleeding rate was 18% in the sclerosis group (95% confidence interval = 10 and 31) and 57% (95% confidence interval = 40 and 72) in the control group. Two-year cumulative mortality rate was 30% (95% confidence interval = 19 and 45) in the sclerotherapy group and 56% (95% confidence interval = 39 and 72) in the controls. One patient died after hemorrhage from an ulcer secondary to sclerotherapy. Analysis by the Cox model of the factors potentially confounding or interacting with the effect of sclerotherapy suggested that sclerotherapy was more efficient in preventing first bleeding in patients with decompensated disease (Child B and C) than in those in good condition (Child A). However, the 2-year cumulative bleeding rate of untreated Child A patients was only 19%, showing how in this group the endoscopic findings were unreliable in selecting high-risk varices and explaining why after a 2-year follow-up prophylactic sclerosis did not show any benefit in such patients. We conclude that sclerotherapy can decrease the incidence of first variceal bleeding and death for a period of 2 years in cirrhotic patients with high-risk varices.