To determine the prevalence and natural history of gastric varices, we prospectively studied 568 patients (393 bleeders and 175 nonbleeders) with portal hypertension (cirrhosis in 301 patients, noncirrhotic portal fibrosis in 115 patients, extrahepatic portal vein obstruction in 117 patients and hepatic venous outflow obstruction in 35 patients). Primary (present at initial examination) gastric varices were seen in 114 (20%) patients; more were present in bleeders than in non-bleeders (27% vs. 4%, respectively; p < 0.001). Secondary (occurring after obliteration of esophageal varices) gastric varices developed in 33 (9%) patients during follow-up of 24.6 +/- 5.3 mo. Gastric varices (compared with esophageal varices) bled in significantly fewer patients (25% vs. 64%, respectively). Gastric varices had a lower bleeding risk factor than did esophageal varices (2.0 +/- 0.5 vs. 4.3 +/- 0.4, respectively) but bled more severely (4.8 +/- 0.6 vs. 2.9 +/- 0.3 transfusion units per patient, respectively). Once a varix bled, mortality was more likely (45%) in gastric varix patients. Gastric varices were classified as gastroesophageal or isolated gastric varices. Type 1 gastroesophageal varices (lesser curve varices) were the most common (75%). After obliteration of esophageal varices, type 1 gastroesophageal varices disappeared in 59% of patients and persisted in the remainder; bleeding from persistent gastroesophageal varices was more common than it was from gastroesophageal varices that were obliterated (28% vs. 2%, respectively; p < 0.001). Type 2 gastroesophageal varices, which extend to greater curvature, bled often (55%) and were associated with high mortality. Type 1 isolated gastric varices patients had only fundal varices, with a high (78%) incidence of bleeding.(ABSTRACT TRUNCATED AT 250 WORDS)