Thirty-two patients with non-cirrhotic portal system obstruction and oesophageal varices of non-malignant etiology were recruited over 13 years. Diagnosis was based on the presence of oesophageal varices at endoscopy, minor alterations in liver function tests and liver histology, a low hepatic venous pressure gradient, and pertinent angiographic patterns. Twenty-three had portal vein thrombosis, nine had splenic vein thrombosis. Twenty-one had idiopathic portal vein obstruction, 11 had secondary obstruction. The outcome was compared with a group of 32 patients with cirrhosis and portal hypertension, matched for age, Child-Pugh class, previous history of gastrointestinal bleeding, and size of oesophageal varices. Patients with non-cirrhotic obstruction of the portal system were followed for up to 171 months (mean 94 months). During follow-up ten patients had gastrointestinal bleeding, and eight died (five of gastrointestinal bleeding). After 6 years of follow-up, the cumulative risk of gastrointestinal bleeding was 24%, the cumulative risk of death was 17%, and the cumulative risk of death from gastrointestinal bleeding was 14%. Cumulative probability of death by any cause and the probability of gastrointestinal bleeding were significantly lower in patients with non-cirrhotic obstruction of the portal system than in patients with cirrhosis comparable for liver function and portal hypertension (p = 0.04 for both). The cumulative probability of death by gastrointestinal bleeding was not significantly different. In conclusion, the prognosis for non-cirrhotic obstruction of the portal system is significantly better than for patients with cirrhosis with comparable levels of liver function impairment and severity of portal hypertension.