We electively compared the distal splenorenal ("selective") shunt with the end-to-side portacaval shunt in 80 prospectively randomized patients with variceal bleeding. Selective shunts required more operative time (3.9 vs. 2.8 h) and blood replacement (4.6 vs. 2.5 U) and postoperative mortality was slightly higher (5 of 38 selective vs. 0 of 40 portacaval). Postoperative complication rates were similar. After 65-mo mean follow-up, both shunts have protected well against late gastrointestinal bleeding (5 selective, 4 portacaval episodes). However, after selective shunts, spontaneous encephalopathy occurred less often (23% vs. 40% of patients), was severe in fewer patients (12% vs. 33%), and precipitated fewer hospital admissions (6 admissions in 4 selective patients vs. 26 admissions in 13 portacaval patients). Furthermore, selective shunt patients remained longer without functional disability (83% vs. 70% of postoperative patient months). Long-term survival was not significantly different in the two groups (5-yr survival: selective 51%, portacaval 56%).