Eighty-two consecutive Child-Campbell class A and B cirrhotic patients were included in a prospective controlled trial to assess the efficacy and safety of portacaval anastomosis vs. endoscopic sclerotherapy as elective treatment of variceal hemorrhage. Forty-one patients were randomized to portacaval anastomosis and 41 to sclerotherapy. After excluding dropouts, 34 patients were treated with portacaval anastomosis and 35 with sclerotherapy. The incidence of variceal rebleeding during follow-up (mean +/- SD, 20.6 +/- 14.2 months) was significantly higher in the sclerotherapy than in the portacaval groups, either considering the overall treated group or only patients completing sclerotherapy (40% and 25% vs. 2.9%; P = 0.0002 and P = 0.01, respectively). The 2-year probability of suffering from at least one episode of hepatic encephalopathy was significantly higher in patients submitted to portacaval anastomosis than in those treated with endoscopic sclerotherapy (40% vs. 12%; P = 0.04). However, disabling encephalopathy only appeared in 3 of 34 patients who underwent surgery (8.8%). Early and long-term mortality did not differ between the therapeutic groups; 2-year survival rates were 83% for portacaval anastomosis and 79% for sclerotherapy. It is concluded that portacaval anastomosis is more effective than endoscopic sclerotherapy in preventing variceal rebleeding in spite of the greater incidence of hepatic encephalopathy. The role of portacaval anastomosis in the elective treatment of variceal rebleeding should be reassessed.