The role of specific pathological findings in the upper gastrointestinal tract in chronic renal failure remains uncertain. Most of the studies were conducted in the West, and the number of subjects was small. We have tried to look at that problem in Taiwan. Endoscopy to evaluate the source of upper gastrointestinal hemorrhage was performed in 698 patients over a 37-month period; that represents 4.4% of all patients undergoing upper gastrointestinal endoscopy for miscellaneous reasons in that time span. Fifty-eight patients (8.3%) who had been hemodialyzed for chronic renal failure were selected, as were 640 control patients who did not have renal failure. Patients with renal transplant were not included. Endoscopic diagnoses, contributing factors of bleeding, and the course and outcome of the hospitalization were analyzed. chi 2 Test with or without Yates' correction and Student's t test were used as appropriate. Erosive gastritis was the most frequent source of bleeding in patients with chronic renal failure. Erosive gastritis (p < 0.005), erosive esophagitis (p < 0.001), and esophageal ulcer (p < 0.005) were significantly more common causes of bleeding in the renal failure population than in the group without renal failure. The two groups did not differ significantly (p > 0.05) in smoking, heavy alcohol intake, or use of ulcerogenic medications. The age was older (64.1 +/- 11.4 vs. 55.7 +/- 16.2 years) and the mortality rate higher (13% vs. 2%) in patients with renal failure than in those without. The differential diagnoses of upper gastrointestinal bleeding sites differ in patients with and without chronic renal failure; they are diverse. However, erosive gastritis, rather than gastric ulcer or duodenal ulcer, is the most common cause in the patients with renal failure. The mortality rate is significantly higher in these patients than in the general population.