Therapeutic endoscopy (TE) has provided a new means for treating peptic ulcer disease, prompting a reevaluation of surgery's role. The aim of this study was to determine if surgical therapy of bleeding duodenal ulcers has changed since the advent of TE. This retrospective review involved consecutive time periods during which TE was (1985-1990) and was not (1980-1984) widely available. Exclusion criteria were prior gastric surgery, nonpeptic conditions, and untreated ulcers. Inclusion standards were met by 252 patients (180 men, 72 women) whose mean age was 67 years. Patients were grouped by the initial therapeutic intervention. Groups were similar in age, medical condition (mean APACHE II score 16), and morbidity. Seventy-five patients had surgery alone during 1980-1984 and 38 during 1985-1990. TE was initially performed on 134 patients during 1985-1990. Bleeding (n = 30) and perforation (n = 1) prompted emergent operation in 23% of cases following TE. Thus 69 (38 + 31) patients underwent surgery between 1985 and 1990. Preprocedure transfusions averaged 4.1 units in the endoscopic group and 8.2 units in the operative groups (p < 0.0001). Disagreement existed between the endoscopic and surgical descriptions of ulcer location in 53% of cases. Emergent surgery was required in 45% of hemodynamically unstable patients versus 14% of stable patients who initially underwent TE (p < 0.0001). Sixty-one percent of incompletely visualized TE-treated lesions required operation, and 18% of well visualized ulcers underwent operation (p < 0.0001). Hospital mortality was similar (8% versus 16%) in the endoscopic and operated groups (p = 0.7). Mean follow-up was 540 days.(ABSTRACT TRUNCATED AT 250 WORDS)