Abdominal aortic aneurysms (AAA) and haematemesis are common clinical problems which are expected to increase as the population ages. An AAA is usually detected either incidentally on routine palpation and ultrasound examinations or when complicated with pain or extraperitoneal rupture. Haematemesis is commonly from gastro-oesophageal varicies or ulceration. Nevertheless, rupture of an AAA in to the duodenum, an aortoduodenal fistula (ADF) is a rare complication of an AAA [1] and haematemesis is rarely caused by ADF. This explains diagnostic and therapeutic delays with primary ADF and resultant high mortality.

The need to consider an ADF as the cause of upper gastrointestinal bleeding in a patient with a previous aortic anastomosis is well established [2]. Interestingly, there are recent reports of ADF even after endovascular stent grafts for AAAs [3]. The thinking is not the same as when there is gastrointestinal bleeding in a patient with an incidental AAA. Furthermore, in primary ADF, the combination of a pulsatile abdominal mass and gastrointestinal bleeding is seen only in 23% [4] although this would increase with the use of ultrasound. But even when haematemesis occurs in a patient known to have an AAA initial management is limited to the more common causes such as varices and peptic ulceration. Thus in a hypotensive patient, blood is transfused until pressures are normalized and the bleeding usually stops spontaneously. In the case of an aortic bleed this would cause further bleeding and death. In a normotensive patient, upper gastrointestinal endoscopy would be the initial step in the search for a cause. However, finding lesions without active bleeding does not rule out an ADF and the length of the endoscope does not allow visualization of the distal duodenum where ADF occur [5]. Computed tomography with contrast is probably the most useful as it may show the communication or reveal loss of continuity and air bubbles in the aneurysm wall that are pathognomonic [6]. Percutaneous angiography is rarely of value since the need for it coincides with the need for immediate surgery [7].

The pattern of bleeding from an ADF is of interest. A "herald" haemorrhage, which may be occult or mild, is followed hours, days, or weeks later by catastrophic haemorrhage that is characteristic of an ADF. This initial slow bleed is the result of a small fistula occluded by thrombus. Since 70% survive at least 6 hours [4] and up to 50%, 24 hours [8] after the initial bleed, a "herald" haemorrhage should be viewed as an opportunity for prompt surgical intervention. Despite technological advances in endoscopy and imaging, the cornerstone in the diagnosis of an ADF remains clinical suspicion. The challenge is even greater when the initial bleed is occult and manifest only as anaemia as in our patient. In this instance, the non tender AAA with negative imaging for a leak was not considered the source of bleeding and a primary haematological or gastrointestinal cause was being pursued.

Surgical repair without delay is the only chance for survival [4]. Surgery must disconnect the fistula, debride the retroperitoneum, repair the duodenum and aorta with omental interposition and establish blood flow to the lower extremities. In the absence of gross infection as in our patient, anatomic in situ repair of the aorta with a synthetic graft gives good results [9]. However, when there is gross infection as with mycotic aneurysms, closure of the aorta and extra-anatomic grafting is advisable [10]. More recently, in situ grafting using autogenous vein, either superficial femoral vein [11] or spiraled saphenous vein [12] has been recommended. Finally, successful endovascular stent grafting has been reported in the case of a very high risk patient giving us an additional option for the future [13].

Even though primary ADF is a rare cause of enteric bleeding, what makes it important is the high mortality from lack of awareness and non diagnosis. The key to its successful management is a high degree of clinical suspicion when there is gastrointestinal bleeding in combination with an AAA. We recommend that in such instances bleeding must be assumed to be from an ADF unless another source can be identified with certainty. The search for other causes is advisable only in normotensive patients while hypotensive patients should be referred to vascular surgeons to be operated on without further delay.

Written consent was obtained from the patient for publication of this case report and any accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal.