We report here on our experience of clinically relevant bleeding after endoscopic sphincterotomy (ES). Relevant bleeding was defined by the occurrence of (a) hematemesis or melena and (b) at least a two-point drop in hemoglobin, with no other bleeding source on endoscopy. These two criteria were met in 16 patients between 1983 and 1992. They represented 0.65% of all ES procedures performed during this period. Bleeding occurred immediately after ES in five cases, and was delayed in 11 cases from one to eight days (mean two days). Patients were retrospectively classified into three groups according to the severity of bleeding and subsequent clinical management. In six cases (group 1), bleeding developed slowly without shock and stopped spontaneously. In five cases (group 2), bleeding developed rapidly with melena and a drop in hemoglobin, but without shock. These patients were successfully managed with sclerotherapy without any further complications. The five patients in Group 3 had brisk bleeding with hematemesis and shock. Endoscopic hemostasis could not be performed; emergency arteriography disclosed active bleeding in four patients, and embolization of the gastroduodenal artery was performed. Bleeding stopped in all patients. Billroth II anastomosis appeared to be the only factor associated with an increased risk of clinically relevant bleeding. It was possible to control bleeding following ES using endoscopic or angiographic hemostasis, surgery being avoided in all cases.