We conducted an uncontrolled study to evaluate an improved metallic clip (Olympus hemoclip) for the endoscopic treatment of nonvariceal gastrointestinal bleeding. A total of 88 patients (mean age 63 +/- 14, 60 males) with bleeding from a wide range of sources were treated. Seventy-eight patients had active bleeding (spurting in 50, oozing in 28) and 10 patients had a nonbleeding visible vessel. Initial hemostasis was achieved in all patients with active bleeding. A total of 255 clips were placed (average of 2.9 clips per patient, range of 1-10 clips). Spurting arterial bleeders required more clips on average than oozing bleeders (3.2 versus 2.7); active bleeders required more clips than cases with nonbleeding visible vessels (3.0 versus 2.2). Mean follow-up was 397 +/- 148 days. Recurrent bleeding was observed in 5 patients, all of whom had active bleeding on initial presentation. Re-bleeding was successfully treated with hemoclips in 4 patients and one patient underwent surgery. Clips appeared to be retained well; early clip dislodgement resulted in rebleeding in only 1 patient. No complications resulted from this treatment. Clips did not impair healing of peptic ulcers. We conclude that endoscopic hemoclip placement is a highly effective and safe method for treating nonvariceal gastrointestinal bleeding and deserves comparative studies with other methods of endoscopic hemostasis.