Retrospective evaluation of interventional embolization therapy in the treatment of gastrointestinal hemorrhage over a long-term observation period from 1989 to 1997. Included in the study were 35 patients (age range 18-89 years) with gastrointestinal bleeding (GI) referred for radiological intervention either primarily or following unsuccessful endoscopy or surgery. Sources of GI bleeding included gastric and duodenal ulcers (n = 7), diverticula (n = 3), erosion of the intestinal wall secondary to malignancy (n = 6), vascular malformations (n = 4), and hemorrhoids (n = 2), as well as from postoperative (n = 6), posttraumatic (n = 2), postinflammatory (n = 4) or unknown (n = 1) causes. Ethibloc (12 cases) or metal coils (14 cases) were predominantly used as embolisates. In addition, combinations of tissue adhesive and gelfoam particles and of coils and Ethibloc were used (six cases). Finally, polyvinyl alcohol particles, a coated stent, and an arterial wire dissection were utilized in one case each. Bleeding was stopped completely in 29 of 35 cases (83%). In one case (3%) the source of bleeding was recognized but the corresponding vessel could not be catheterized. In five other cases (14%) there was partial success with reduced, though still persistent, bleeding. The rate of complications was 14%, including four instances of intestinal ischemia with fatal outcome in the first years, and, later, one partial infarction of the spleen without serious consequences. Gastrointestinal hemorrhage can be controlled in a high percentage of patients, including the seriously ill and those who had previously undergone surgery, with the use of minimally invasive interventional techniques. The availability of minicoils instead of fluid embolization agents has reduced the risk of serious complications.