Angiographic embolization for major haemorrhage after upper gastrointestinal surgery

Aims: Severe postoperative haemorrhage after upper gastrointestinal surgery is a serious complication. This study examined the effectiveness of selective mesenteric angiography (SMA) in localizing a bleeding point and the ability of angiographic haemostatic methods to control bleeding. Methods: The case notes of a consecutive series of nine patients undergoing urgent SMA during 1996–1998 were analysed. Examination of angiography suite records confirmed the accuracy of patient identification. SMA was performed 18 times with 13 embolizations in nine individuals (seven men; median age 54 (27–73) years). Patients underwent the following operations: Whipple pancreaticoduodenectomy (four patients), pancreatic necrosectomy (two), total gastrectomy (one), cholecystectomy (one) and splenectomy (one). The median interval from surgery to haemorrhage was 15 (2–49) days. Six patients presented with haematemesis/melaena and three with bleeding from drains. Seven had evidence of shock (systolic blood pressure less than 100 mmHg, pulse more than 100 per min); the mean preprocedure haemoglobin concentration was 59 g/l. A median of 8 (4–14) units of blood were transfused before embolization and 4 (2–9) units after. Ten initial endoscopies were performed in six patients, seven of which revealed a source of bleeding. Endoscopic haemostasis was attempted in five and achieved temporary control of bleeding in two. Results: Angiography revealed a discrete bleeding point in 13 of 18 procedures in eight patients. Where a bleeding point was identified, angiographic embolization using 3–8-mm stainless steel coils (ten) or a combination of coils and gelatin sponge (three) achieved radiological evidence of haemostasis in all cases. Periprocedural complications occurred in one patient with unintentional partial embolization of the right hepatic artery during embolization of an actively bleeding left hepatic artery pseudoaneurysm. Rebleeding occurred in six patients within 48 h. Three rebleeds were successfully managed with repeat SMA and embolization (one patient required a third embolization); the remaining three required surgery. Definitive radiological haemostasis was achieved in six patients. Five of the nine patients died in hospital, two of whom had been successfully embolized. Conclusions: In this group of patients, endoscopy contributed relatively little to treatment of postoperative haemorrhage. In contrast, SMA identified a bleeding point in eight of nine patients and achieved definitive control of bleeding in six. SMA and embolization appears to have a useful role in patients with this infrequent but potentially lethal complication. © 2000 British Journal of Surgery Society Ltd