Background

Expeditious identification of the source of gastrointestinal (GI) bleeding is key to minimising patient morbidity and mortality. Most posterior wall duodenal bulb ulcers erode into the gastroduodenal artery (GDA), which is the first vessel that is angiographically interrogated in these cases.1 2 Very few cases have been previously reported involving bleeding from hepatic arterial branches.3–11 Here we report a case of GI bleeding caused by duodenal ulcer erosion into the proper hepatic artery, highlighting the potential for non-GDA vessels to drive serious and life-threatening GI bleeding and emphasising the clinical need for carefully interrogating vessels more distal in the coeliac axis in all patients with an upper GI bleed when the source of bleeding is not initially localised on angiography.

Management of GI bleeding is a cross-disciplinary activity that requires coordination between gastroenterologists, interventional radiologists and surgeons. After the patient is resuscitated and the source of the bleed has been localised, lesions are ideally treated with endoscopic or angiographic interventions, reserving surgery for refractory bleeding. Innovation in endoscopic and endovascular techniques has expanded non-operative options for management of GI bleeding, including undersewing,12 epinephrine injection,13 standard endoscopic clips,14 over-the-scope clips,15 haemostatic sprays16 and transcatheter arterial embolisation using a variety of embolic agents (coils, polyvinyl alcohol, gelatin sponge, N-butyl cyanoacrylate).17 Using minimally invasive interventions is important in minimising patient morbidity, since operative management of upper GI bleeding inherently carries a greater risk of adverse events and prolonged hospital stay.1 18

Case presentation

A man in his 60s had a history of smoking, hypertension, infrarenal abdominal aortic aneurysm treated with endovascular placement of Palmaz stent, coronary artery disease treated with a drug-eluting stent, and diverticulitis that was complicated by large bowel stricture and obstruction requiring an emergency sigmoid colectomy and colostomy creation. The patient then developed a large ventral hernia and subsequently underwent a planned colostomy takedown and complex abdominal wall reconstruction with anterior component separation and placement of underlay mesh. He recovered from this operation and was discharged home. He was then readmitted on postoperative day 10 with upper GI bleeding. Over the ensuing weeks, he had persistent intermittent GI bleeding requiring two separate hospital admissions including an intensive care unit (ICU) stay. He underwent five upper endoscopies with attempts to obtain haemostatic control, as detailed below. In addition, four mesenteric angiograms were performed with the goal of durable haemostasis. In addition, at two admissions in the hospital, the patient underwent CT angiogram of the abdomen with contrast. Neither scan showed any evidence of active extravasation of contrast. He also completed an empirical full course of triple therapy for Helicobacter pylori during this time.

Endoscopy consistently identified a posterior ulcer in the duodenal bulb as the source of bleeding. An attempt was made to control the ulcer with an over-the-scope clip; however, even this intervention did not definitively control the bleeding. Multiple angiograms did not reveal any contrast extravasation to identify the exact source of haemorrhage; thus, the patient underwent empirical embolisation of the GDA and associated arcade including both inferior and superior pancreaticoduodenal arteries and the proximal right gastroepiploic artery. Although this intervention carried the risk of subsequent bowel ischaemia, it was deemed an acceptable risk given his refractory, life-threatening bleed. Despite these attempts, he continued to experience GI bleeding. His final endoscopy 3 days prior to emergency surgery showed arterial bleeding from the ulcer base (figure 1A). Bleeding was temporarily controlled with epinephrine injection. We would note that throughout this process, we considered surgical intervention. However, given his complex history of abdominal surgery and recent complex abdominal closure, we were concerned about serious complications from the surgery. As such, exhaustive efforts were undertaken to avoid surgery and treat this bleed through embolisation or endoscopy. Unfortunately, they repeatedly failed.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Representative images from final endoscopy and angiogram prior to surgery of the patient. (A) Endoscopy 3 days prior to surgery showed continued brisk bleeding from the base of the ulcer, adjacent to the previously placed over-the-scope clip. (B) Angiography demonstrating branches of the coeliac artery, including previously coiled GDA and R gastroepiploic arteries (green arrow), closely abutting the ulcer and over-the-scope clip (blue arrow) with no evidence of active extravasation, abrupt vessel cut-off or other signs of active haemorrhage. GDA, gastroduodenal artery.

A subsequent repeat angiogram that interrogated the coeliac axis and the superior mesenteric artery (SMA) continued to show no active haemorrhage with no further potential embolisation targets (figure 1B). He then experienced new onset tachycardia, voluminous melaena and hypotension and thus was taken emergently to the operating room.

Treatment

The patient’s ventral bioprosthetic underlay mesh was well incorporated into the abdominal wall and was opened in the midline on entry to the abdomen. Extensive lysis of adhesions was performed since the patient had undergone abdominal surgery 6 weeks prior. The hepatic flexure of the colon was mobilised, and a Kocher manoeuvre was performed to expose the duodenum. During the dissection, bleeding was encountered from the gall bladder and adjacent liver bed, prompting a cholecystectomy.

Given the patient’s dense adhesions and the difficulty of mobilising any small bowel for resection and reconstruction, the decision was first made to attempt to achieve haemostasis by oversewing the bleeding ulcer. Accordingly, a longitudinal duodenogastrotomy was made on the anterior surface of the first portion of the duodenum extending over the pylorus and partially onto the antrum of the stomach. This immediately exposed a 20 mm ulcer which was still bleeding, along with the previously placed Ovesco clip on the posterior wall of the first portion of the duodenum. The ulcer was oversewn with 2-0 silk stitches placed in a figure-of-eight fashion. Despite placing multiple stitches with progressively deeper bites into the base of the ulcer, the ulcer continued to bleed, and each needle pass elicited brisker arterial bleeding. Thus, after holding direct pressure on the ulcer, it was concluded that suture ligation of the ulcer base was not sufficient to obtain haemostasis and an antrectomy with D1 duodenectomy was performed. The duodenum was lifted off the ulcer crater which revealed bleeding from the porta hepatis. A Pringle manoeuvre was performed by applying a tourniquet around the hepatoduodenal ligament. This substantially decreased the haemorrhage. The bleeding resumed after releasing the tourniquet, but once again stopped after a vascular clamp was placed more proximally on the common hepatic artery. After this was done, a clear point of disruption was visualised on the proper hepatic artery, which was therefore ligated to achieve haemostasis. The distal duodenal stump was closed with a single layer of Gambee stitch. The patient had been massively transfused (including roughly 6.5 L of red blood cells, 5.2 L of fresh frozen plasma, 750 mL of platelets and 215 mL of cryoprecipitate) with significant visceral oedema so the abdomen was temporarily closed with a vacuum dressing, and the patient was transferred to the ICU.

Outcome and follow-up

The patient was brought back to the operating room 48 hours later and re-explored. There was no longer any bleeding from the area of the ulcer, and the liver appeared healthy. A retrocolic gastrojejunostomy was performed to reconstruct the patient’s GI tract and his abdominal fascia was closed with a mesh-mediated closure. The patient’s postoperative course was complicated by prolonged ileus, with a high output nasogastric tube for approximately 2 weeks. He was given total parenteral nutrition until he regained bowel function, at which time his nasogastric tube was removed and he was able to tolerate a regular diet. The patient was discharged to inpatient rehabilitation 19 days after his antrectomy and was discharged home 5 days later and has made a complete recovery.

The patient was doing well 6 months later without recurrent bleeding and tolerating a regular diet. Notably, the patient also did not suffer any long-term liver injury after ligation of the proper hepatic artery. The patient underwent repeat CT angiogram at 6-month follow-up which revealed the known ligated proper hepatic artery and hypertrophy of an accessory left hepatic artery with its takeoff proximal to the GDA from the common hepatic artery which was not seen on prior angiography (figure 2). The liver is then perfused by intrahepatic collaterals. The hypertrophy of this vessel likely protected the patient from developing liver injury or biloma despite the ligation of the proper hepatic artery.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Representative image from CTA 6 months following the patient’s surgery. Red arrow points to the area of the ligated proper hepatic. Just proximal, the green arrow indicates an accessory left hepatic artery, which takes off proximal to where GDA takeoff had been prior to intervention. We suspect this hypertrophied accessory hepatic artery is supplying segments 2 and 3. The liver is then perfused via intrahepatic collateral vessels. CTA, CT angiogram; GDA, gastroduodenal artery.