Acute gastrointestinal haemorrhage

Radiology has a role in the diagnosis and therapy of acute gastrointestinal bleeding. That role is more important in haemorrhage arising from below the Ligament of Treitz since, when the source is from the upper GI tract the endoscopist can make the diagnosis in most patients and often institute endoscopic measures to arrest the bleeding. There is virtually no place for barium studies in acute bleeding, since superficial mucosal lesions are not seen, the source of haemorrhage cannot be determine d when multiple lesions are shown and barium will obscure the field for subsequent angiography. The major role for radiology lies with angiography - both diagnostic and interventional. Whether angiography should be preceded by radionuclide studies is a vexed question. Some radiologists prefer this in haemodynamically stable patients in order to confirm active bleeding (and thus increase the yield of angiography) and to localize the approximate site of bleeding to save time and contrast medium at angiography. However, as discussed below there are problems and pitfalls related to isotope studies.

Two techniques of scintigraphic imaging of gastrointestinal bleeding are generally available. Technetium-labelled sulphur colloid given intravenously is quite rapidly taken up in the liver and spleen. This method therefore will only detect bleeding if it is actively occurring during the intravascular phase of the tracer. This is a problem given that, even in patients with massive bleeding, haemorrhage is of ten intermittent. In addition, bleeding sites in the upper abdomen may be masked by hepatosplenic uptake. However, the technique is very sensitive, being able to detect as little as 0.05 ml per minute oxblood loss away from the background of liver and spleen. Technetium-Iabelled red cells - now able to be labelled in vitro - act as a blood pool agent, allowing imaging up to 24 hours or so after administration, and thus detection of intermittent bleeding. (Indium-labelling enables detection up to 5 days but at high radiation dose.) The minimum rate of blood loss detectable is about 0.1 ml per minute. Images are acquired initially at 30 second intervals for at least 60-90 minutes. Scintigraphy with Technetium-labelled red blood cells is a sensitive, non-invasive procedure that can be used forprolonged surveillance of bleeding, but there are many pitfalls leading to false positive results or inaccurate localisation. These may be overcome by high labelling efficiency techniques, digital imaging for at least 90 minutes, use of cine display to enable demonstration of changes of position with time of areas of abnormal uptake, and strict criteria for the localisation of a source of bleeding seen on delayed images only. This latter problems arises from the rapid antegrade or retrograde movement of extravasated red cells within the bowel leading to false localisation. In addition, pooling can occur in the large bowel from a source more proximally. Figures for correct localisation vary in the literature from 40 to 90%.

It is convenient to discuss upper and lower gastrointestinal haemorrhage separately. It is usually possible to distinguish these on clinical grounds and by the passage of a nasogastric tube. Selective angiography is performed, the choice of vessel first selected being determined by the suspected site of bleeding. Although digital subtraction radiography (DSA) is convenient, many radiologists prefer cut films citing the pitfalls of DSA such as artefacts due to bowel movement, lesser field of view and decreased spatial resolution. It is estimated that conventional angiography can detect blood loss of approximately 0.5 ml per minute; DSA can probably detect 1-1.5 ml per minute. Extravasation is seen as puddling of contrast in the early to mid-arterial phase which changes size and shape as the series progresses. Venous bleeding is hardly ever demonstrated.

Upper gastrointestinal haemorrhage

Patients present with haematemesis and/or melena. Occasionally, in massive bleeding, red blood may be passed per rectum. Endoscopy is usually diagnostic and often, if there is active bleeding, endoscopy measures will arrest this. Angiography is required if the result of endoscopic is equivocal or if angiographic therapeutic intervention is planned. Haemodynamically unstable patients may be examined while undergoing resuscitative measures - nuclear medicine studies are not indicated in these patients. In stable patients the question of whether to perform scintigraphy is vexed (see above); these studies are least accurate for upper GI bleeding.
Coeliac angiography is undertaken, followed by selective left gastric artery (LGA) catheterisation. LGA angiography identifies about 70% of active gastric bleeding. If no abnormality is seen, then further injections of gastroduodenal and superior mesenteric arteries are performed to assess the pancreaticoduodenal arcades and other gastric vessels. It should be noted that venous bleeding, for example, from varices, is hardly ever detected. Transcatheter intervention may be performed by intra-arterial vasopressin infusion or embolotherapy. Vasopressin administration is associated with complications such as hypertension, arrhythmias and vasoconstriction that my result in tissue ischaemia and necrosis that may affect bowel, myocardium and the periphery. The reader is referred to specialist texts for details of dosage. The success rate of vasopressin in arresting bleeding from gastric sources is high (75- 80%). It is usually unsuccessful if there have been previous attempts at endoscopic heat or electrocautery to the bleeding site, and embolotherapy is generally indicated. The success rate of vasopressin for arrest of pyloroduodenal haemorrhage is low (about 30%). This may be due to the dual blood supply of this region, penetration of ulcers into the muscular layers of the wall or the propensity for large vessels such as the gastroduodenal artery to be the bleeding source. Embolotherapy for gastric bleeding usually involves the LGA. Gelatin sponge (Gelfoam) is the most frequently used material; this is a temporary occluding agent, allowing the body's haemostatic mechanisms to be effective at the bleeding site. Permanent occluding agents are used for malignant causes of bleeding. Embolotherapy is successful in approximately 80 %. The risk of infarction is significantly increased if there has been a previous gastric resection. Embolisation for duodenal bleeding sources is more challenging due to the dual blood supply, which may require occlusion of gastroduodenal artery and inferior pancreaticoduodenal artery or branches. This is usually well tolerated when there has been no previous surgery. Occasional complications include doudenal ischaemic necrosis, hepatic or gallbladder infarction.Success rates are quoted as 60-100% for ulcer bleeding.

Lower gastrointestinal haemorrhage

Radiologists more frequently have a role here. Endoscopy is rarely diagnostic if the patient is bleeding significantly. However, proctoscopy or rigid sigmoidoscopy should be performed to exclude an anorectal origin. In addition, the upper GI tract is the source in about 10% of patients with severe rectal bleeding; this should be excluded by nasogastric intubation or, preferably, endoscopy. Scintigraphy is at its most accurate in the colon and may be undertaken in the haemodynamically stable patient. Small bowel haemorrhage is most commonly due to anastomotic ulcers, neoplasms, Meckel's diverticulum, arteriovenous malformations or angiodysplasia, and varices. Variceal bleeding is not seen

Figure 80.Superior mesenteric artery angiogram demonstrating extravasation into a bleeding right-sided colonic diverticulum (arrowed).

angiographically, although the varices are demonstrable. Colonic sources of massive haemorrhage are most commonly diverticula (Fig. 80) (80% of diverticula are left-sided but 50% of diverticular bleeding is right-sided), and angiodysplasia. Colonic haemorrhage arrests spontaneously in 8090 % of patients, but recurs in about 25 %. Angiodysplasia is common in elderly asymptomatic individuals and lesions are not uncommonly multiple. Therefore, unless extravasation is seen, the angiographic demonstration of angiodysplasia must be interpreted with caution - it may be an incidental finding in a patient with another cause of bleeding. The angiographic features of angiodysplasia are of a vascular tuft with an early filling prominent draining vein which persists. Lesions occur anywhere but are usually seen in the right colon. Possible transcatheter interventions include vasopressin infusion, embolisation or (for small bowel Iesions) aiding the surgeon in localising the appropriate segment for resection (see below). Vasopressin controls bleeding in approximately 50% of small bowel and 75% of colonic lesions. In the latter, 15% will recur following cessation of infusion. Embolisation carries the risk of bowel infarction; proximal occlusions are safer but less effective, most workers aiming at occluding proximal second order branches or more distally. In one series embolisation controlled haemorrhage in over 90%. The incidence of infarction is difficult to assess, but is probably about 10%. Not all of these will require urgent resection; some are relatively minor and will lead to no sequelae or to later colonic strictures. There is no consensus as to whether vasopressin or embolotherapy is the initial preferred method in colonic bleeding. Both techniques have their advocates. There is some evidence that vasopressin is best in diverticular bleeding and embolisation in others.

Angiography in recurrent obscure bleeding

Recurrent bleeding of obscure origin is of ten a difficult problem. Bleeding has frequently ceased by the time that angiography has been arranged. Attempts to increase the yield of angiography include selecting the timing of the study by scintigraphy and pharmacoangiography. Agents used for the latter have included heparin or thrombolytic drugs to prolong bleeding and vasodilators, such as tolazoline, to enable angiographic detection of extravasation. There are, of course, risks associated with these techniques. However, angiography may be diagnostic even if active bleeding is not occurring at the time. Lesion detection increases with age over 50 years. Neoplasms, arteriovenous mal formations, angiodysplasia and varices may be demonstrated. Results may be interpreted with caution since the only conclusive evidence that a lesion has bled is the demonstration of extravasation.

The angiographer may aid the surgeon in selecting the appropriate small bowel segment for resection in difficult small bowel lesions. The catheter may be left in situ in the vessel supplying that segment and methylene blue instilled through the catheter to enable visual identification of the loop.

Richard M. Mendelson