Embolization Therapy: Principles and Clinical Applications, 1 Ed.

Upper Gastrointestinal Bleeding

Rakesh C. Navuluri • Brian Funaki

Approximately 100,000 cases of upper gastrointestinal bleeding require inpatient admission annually in the United States. When medical management and endoscopic therapy are inadequate, endovascular intervention can be lifesaving. This chapter begins with a brief review of the preangiographic workup of patients with upper gastrointestinal bleeding. This will be followed by a detailed discussion of the angiographic technique, including a closer look at the use of various embolic agents. The primary focus will be on nonvariceal (arterial) hemorrhage with brief consideration given to variceal (venous) hemorrhage.


Upper gastrointestinal bleeding (UGIB) is defined by a bleeding source proximal to the ligament of Treitz. UGIB accounts for 76% of gastrointestinal (GI) bleeding events.1 The incidence of UGIB in the United States is estimated at 102 per 100,000 per year, with a mortality rate of 5% based on data from 1991.2 UGIB can be divided into arterial (nonvariceal) and venous (variceal) etiologies. This is an important point of differentiation and a major branch in the management decision tree. Causes of arterial UGIB include peptic ulcer disease (up to 40%), Mallory-Weiss tear (15%), hemorrhagic gastritis, pancreatitis-related pseudoaneurysms, neoplasm, aortoduodenal fistula, and trauma. Other rare causes of arterial UGIB include hemobilia from iatrogenic injury and hemosuccus pancreaticus related to chronic pancreatitis. Venous causes include variceal bleeding secondary to portal venous hypertension (e.g., due to cirrhosis or Budd-Chiari syndrome) or splenic vein thrombosis (Table 29.1).


Typical UGIB symptoms include hematemesis or melena. Hematochezia, although commonly attributed to lower GI sources, can also occur depending on the bleeding rate and transit time of blood through the bowel. Patients with nonvariceal bleeding commonly present with coffee-ground emesis and history of nonsteroidal anti-inflammatory drug use. Variceal bleeding is likely to present with clinical signs of cirrhosis, painless hematemesis, and a greater degree of hemodynamic instability.


Current treatment algorithms call for immediate medical stabilization followed by endoscopic diagnosis and intervention. Refractory cases should be referred for either transvenous or transarterial endovascular intervention, depending on the source of bleeding identified at endoscopy. Although surgery is generally considered the last option, endovascular intervention may be reattempted following failed surgery.


Stabilization of blood pressure is the immediate objective in managing patients with UGIB. Fluid resuscitation should be instituted without delay. Patients with hemodynamically significant bleeding should also be immediately typed and crossed to allow for transfusion of blood products as necessary. The goal of transfusion in this setting is to raise the hemoglobin level above 7 g/dL. Any existing coagulopathy (international normalized ratio [INR] >1.5 or platelet count <50,000/µL) should also be corrected.3 It is important to note that units of packed red blood cells (pRBCs) do not contain the factors and platelets needed for thrombogenesis. If these are not replaced during large-volume blood transfusions, a dilutional coagulopathy will result, which may affect the success of subsequent embolization procedures.

Insertion of a nasogastric tube can help to confirm an upper GI source of bleeding. If variceal bleeding is suspected, a Minnesota tube (or Sengstaken-Blakemore tube) can be placed to tamponade esophageal varices. It has been reported to have variable success in achieving hemostasis but is a good option in stabilizing patients in the emergent setting before interventional procedures.

An estimated 10% to 15% of cases require intervention beyond medical management to control bleeding.4 To help sustain the success of subsequent interventional procedures, various medication regimens may be instituted. Administration of proton pump inhibitors (PPIs) have been shown to reduce the rate of rebleeding and surgery in high-risk patients with nonvariceal UGIB.5 This is because hemostasis within the stomach and duodenum is impeded by low intragastric pH. Low pH has been shown to inhibit platelet function and activate pepsin, which disaggregates platelet plugs.6 The use of portal pressure–reducing medications such as somatostatin, octreotide, or vasopressin can improve outcomes in patients with variceal UGIB.7


Endoscopy should be the initial intervention as it allows for localization of bleeding and determination of etiology via visual inspection and biopsy. It can be particularly helpful in differentiating arterial and venous sources of hemorrhage. Endoscopy can also identify upper and midesophageal sources of bleeding that are not amenable to embolization such as esophagitis. Beyond its diagnostic use, endoscopy also offers several therapeutic options. Hemostasis can be achieved endoscopically via thermocoagulation, sclerosant injection, or clips (banding). Endoscopic clip placement, even if ineffectual, can aid subsequent endovascular intervention by directing the interventional radiologist to the area of concern.

Endoscopy performed within 24 hours of presentation is associated with improved patient outcome and decreased hospital stay. However, the source of bleeding may not be seen in up to 24% of patients at the first endoscopic exam.8 Although endoscopic treatment is reported to be effective in 85% to 90% of patients,9 repeat bleeding occurs in 15% to 20%, most commonly in the first 72 hours after treatment.10Endoscopic predictors of rebleeding include visualized active bleeding, nonbleeding visible vessel, adherent clot, ulcer size greater than 2 cm, and ulcer location in the posterior midgastric body or posterior duodenal bulb1113 (Table 29.2). Recurrent bleeding should be followed with a second endoscopic treatment before endovascular intervention is undertaken.14 Further bleeding after a second endoscopic treatment should be referred to interventional radiology. Alternatively, the inability to visualize the bleeding source at endoscopy due to the severity of bleeding warrants urgent endovascular intervention.


Localization of bleeding before angiography is important for several reasons. First, it decreases procedure time by allowing for a more targeted approach. This consequently reduces radiation exposure as well as contrast load used during angiography. Minimizing contrast dose is critical in patients with borderline azotemia. Second, it has also been shown to directly impact the success rate of endovascular therapy.15Last, it permits empiric embolization to be undertaken even if active bleeding is not seen at angiography.

If the source of bleeding cannot be identified, angiography may be deferred. However, in the setting of hemodynamic instability or in patients requiring more than 5 units of pRBCs, angiography should be performed immediately. In these circumstances, angiography is the most practical diagnostic tool because it offers the option of concomitant therapy.16 If the clinical situation permits, localization with computed tomography angiography may be considered before proceeding to conventional angiography.

Computed Tomography Angiography

Compared with conventional angiography, computed tomography angiography (CTA) has the advantage of greater availability, speed, and noninvasiveness. It also does not suffer from bowel or respiratory motion artifact. In an animal model, CTA has been shown to detect bleeding rates as low as 0.3 mL per minute—better than conventional angiography.17 No studies specific to UGIB have been performed to date, but a study by Yoon et al.18 found CTA to have a sensitivity of 90.9%, a specificity of 99%, and an accuracy of 97.6% in localizing hemorrhage when considering both upper and lower GI sources. CTA, unlike conventional angiography, may also detect lesions, such as ulcers, masses and aneurysms, that are not bleeding at the time of the study. The ability to characterize the inciting lesion can have a significant impact on treatment planning. If, for example, a mass lesion is seen, particle embolization may be performed where it might otherwise be avoided in the treatment of GI hemorrhage to avoid bowel necrosis. CTA also has the benefit of providing information on the vascular anatomy for preinterventional planning. For example, a steep takeoff of the celiac artery may necessitate the use of a Waltman loop or even brachial artery access. Stenotic or tortuous mesenteric arteries indicate that a guiding catheter will be helpful. Similarly, occlusion of the celiac axis may require embolization of the pancreaticoduodenal arcade via the superior mesenteric artery (SMA) (Fig. 29.1).

CTA protocols are similar to those for endoleak evaluation. A noncontrast series is followed by intravenous contrast-enhanced series in arterial and delayed phases. A positive CTA study will show hyperdense contrast material (>90 HU) within the bowel lumen on the arterial phase that increases on the delayed phase.19 This provides up to a 90-second window of time to evaluate for contrast extravasation—depending on the timing of the delayed series—compared with a 10-second window of evaluation for conventional angiography. Oral contrast should not be given during the exam because it will obscure intravenous contrast extravasation into the bowel lumen.

The primary drawback of CTA is the necessity of intravenous iodinated contrast, which may elicit a hypersensitivity reaction or cause renal injury.20 Typically, 1.5 mL/kg (up to 150 mL) is administered. This can be prohibitive in patients with poor renal function, particularly when considering the additional contrast required in a subsequent endovascular procedure. Radiation dose is another drawback to consider.

Nuclear Medicine

Radionuclide scintigraphy may be helpful in identifying sources of chronic GI bleeding, which are otherwise not detectable by endoscopy or conventional angiography. Technetium 99m (Tc-99m)–labeled red blood cell scans can detect bleeding rates as low as 0.2 mL per minute, compared with 0.5 mL per minute for angiography, and can be particularly useful in the setting of intermittent GI bleeding. Unlike angiography and CTA, bleeding scans allows for interrogation over a multiple-hour window, which increases sensitivity for detection of GI bleeding. In patients with hemodynamic instability, studies with an immediate tracer blush are more likely to have a positive angiogram than studies with a delayed blush.21 It is worth noting that positive scintigrams can also occur in cases of variceal bleeding (Fig. 29.2).

In terms of localizing bleeding within the upper GI tract, nuclear medicine is limited by the low resolution of imaging as well as confounding factors such as uptake of free technetium by the gastric mucosa. Consequently, at our institution, we do not routinely recommend radionuclide scintigraphy for localizing the site of bleeding in the upper GI tract.

Once active hemorrhage is documented and localized with a diagnostic radiology study, patients should be transferred immediately to the angiography suite. At our institution, we strive to perform angiography within 1 hour of radiologic diagnosis of active GI bleeding.

Angiography Indications

The primary indication for conventional angiography is the inability to control the source of bleeding endoscopically. Angiography is favored over surgery as the treatment of choice after failed endoscopic therapy, particularly in high-risk surgical patients.16 It is minimally invasive, associated with lower mortality,22 and there has been shown to be no difference in outcome between patients managed with surgery versus arterial embolization.23,24

Angiography is also warranted in instances where the bleeding site cannot be identified by endoscopy or CTA and the patient is hemodynamically unstable as it may be lifesaving. Endovascular embolization is also the primary treatment for hepatobiliary bleeding.


Before angiography, the patient’s renal and coagulation statuses should be assessed. Elevated partial thromboplastin time and prothrombin time/INR as well as thrombocytopenia should be corrected. Embolization is less likely to succeed in the setting of coagulopathy because the most common embolic agent used—coils—causes vessel obstruction by providing a scaffold for thrombus formation rather than by pure mechanical occlusion.16 Furthermore, several studies have demonstrated preinterventional coagulopathy to have a negative effect on clinical success.25,26 Embolotherapy has been reported to be nearly three times more likely to fail in patients with coagulopathy.27 If necessary, blood products may be given intraprocedurally to expedite the procedure.

Even in the setting of uncorrectable coagulopathy, embolization often remains the best option for treatment. In these scenarios, alternative embolic agents including placement of a Gelfoam sandwich or the application of glue or Onyx (Micro Therapeutics Inc., Irvine, California) may prove effective.


Review of imaging, such as CTA, before intervention can profoundly expedite cases by demonstrating vascular occlusions or variant anatomy. One should also consider the angle at which the mesenteric vessels arise from the aorta.

The right common femoral artery is the default access site for mesenteric angiography. We begin with a mapping aortogram to obtain a survey of the vascular anatomy. This is helpful in identifying the ostia of the mesenteric vessels and in guiding subsequent catheter selection. This step takes only a few minutes to accomplish and can save valuable time later on in the procedure. If recent imaging of the vascular anatomy is available, either as a CTA or conventional angiogram, an initial aortogram can be skipped in favor of a selective mesenteric angiogram to decrease contrast load in patients with poor renal function or to expedite the procedure. Contrast extravasation is rarely visible on the mapping aortogram. We use a 5-Fr nonmarking pigtail catheter (Cook Medical Inc., Bloomington, Indiana) placed through a 5-Fr vascular sheath. The pigtail should be positioned just below the level of the diaphragm. Power injection is performed at a rate of 20 mL per second, for a total volume of 30 mL.

Although not required, 1 mg glucagon can be administered before angiography to limit peristalsis motion artifacts on digital subtraction angiography (DSA). For similar reasons, breath hold during DSA is ideal to limit respiratory motion. However, this is not always feasible depending on the patient’s condition and level of sedation.

We favor the use of Rosch celiac (RC-1) or visceral selective (VS1) catheters (Cook Medical Inc., Bloomington, Indiana) to select the celiac artery and SMA. It is important to seat the catheter just beyond the ostium of the mesenteric vessel. If advanced too far, early branching vessels may not be imaged on the angiogram. Power injection of contrast is used for all angiograms to optimize detection of active bleeding. Celiac artery and SMA angiogram injection rates are typically 5 mL per second, for a total volume of 20 to 25 mL. DSA is carried out until the portal venous phase to document patency of the portal vein. This can be important to document in cases that potentially require hepatic artery embolization (Fig. 29.3). Delayed angiograms may also reveal varices that were not readily apparent by endoscopy. If there is hemobilia related to recent percutaneous biliary drain placement, removal of the tube over a guidewire may be necessary before angiography as the relative tamponade effect of the tube may obscure visualization of the bleeding vessel.

If no evidence of bleeding is found at celiac angiography, superselective catheterization of the suspected second-order branch (gastroduodenal artery [GDA] or left gastric) is undertaken. A microcatheter is preferred in these circumstances to avoid inducing vasospasm before the culprit vessel is identified. If these studies are also negative, an SMA arteriogram is performed before terminating the exam.

Angiography can detect bleeding rates as low as 0.5 mL per minute. The primary angiographic findings of bleeding are visualization of active contrast extravasation and contrast pooling in the venous phase. A review of studies by Loffroy et al.28 found that angiographic evidence of active extravasation was seen in 54% of cases. Other indirect signs of bleeding on angiography include pseudoaneurysm, vessel spasm or cutoff, early venous filling, and hypervascularity (Table 29.3).

The presence of an abnormal blush may indicate an inflammatory process. This can represent a bleeding source if such an entity was suspected on prior endoscopy.28 In cases of hemorrhagic neoplasm, tumoral blush and neovascularity may be identified. Not uncommonly, trial subselection of vessels is necessary to demonstrate bleeding.

In theory, carbon dioxide (CO2) angiography is more sensitive than conventional angiography with iodinated contrast because the lower viscosity of CO2 should predispose it to extravasating through endothelial injuries (Fig. 29.4). However, in practice, CO2 imaging is degraded by fragmentation of the CO2 bolus and patient motion related to discomfort caused by the CO2 injection. It also has poorer spatial resolution, which may impair subsequent endovascular treatment.29

In one study, a negative bleeding focus was noted in 52% of cases, with a lower incidence in UGIB (46%) compared with lower GI bleeding (66%).30 Failure to localize a bleeding source may be attributed to slow or intermittent nature of the hemorrhage. In such cases, provocative angiography can aid in detection. Several techniques have been reported, including the administration of anticoagulants, vasodilators, and fibrinolytics, to temporarily augment bleeding and increase diagnostic sensitivity. However, this is rarely, if ever, used for UGIB because, unlike lower GI bleeding, angiography for UGIB is nearly always preceded by endoscopy, which commonly elucidates the source of bleeding. Moreover, the feasibility of empiric embolization in the upper GI tract does not justify the risks of provocative angiography.

If no arterial abnormality is seen, empiric embolization of the vessels supplying the area of concern can be performed. Empiric embolization is performed in 46% of endovascular cases of UGIB.28 This technique is low risk due to the rich collateral circulation of the upper GI tract. The two arteries targeted for empiric embolization are the left gastric artery and the GDA. The left gastric artery, which runs along the lesser curve of the stomach, supplies the distal esophagus, cardia, fundus, and incisura. There is collateralization with branches of the short gastric and right gastric arteries, which typically arise from the splenic and hepatic arteries respectively. The GDA supplies the remainder of the stomach and duodenum through the right gastroepiploic artery and branches of the pancreaticoduodenal arcade. There is collateralization with the left gastroepiploic artery, which arises from the distal splenic artery, and branches from the SMA. The SMA provides duodenal supply via the pancreaticoduodenal arcades28 (Table 29.4). There has been shown to be no statistical difference in outcomes between patients treated with empiric embolization versus embolization after angiographically demonstrated contrast extravasation.26,31,32 An alternative to empiric embolization in cases of negative angiography is to target branches supplying the area of endoscopically placed clips.

Although the number of arteries embolized may not impact clinical success of embolotherapy,25 it may affect subsequent surgical therapy. For example, embolization of both the left gastric and gastroduodenal arteries for treatment of a large gastric ulcer may impair attempts at subsequent partial gastrectomy in favor of total gastrectomy (Fig. 29.5).

Superselection of vessels may be necessary to identify bleeding. This typically requires coaxial placement of a 3-Fr microcatheter through a 5-Fr catheter. At our institution, we commonly employ a Renegade (Boston Scientific Corporation, Natick, Massachusetts) or a Progreat (Terumo, Tokyo, Japan) microcatheter. The Renegade microcatheter is available in two types: HI-FLO and STC. The Renegade HI-FLO has a larger diameter (0.027 in) and is best suited for cases where particulate agents are used. The slightly smaller diameter (0.021 in) of the Renegade STC is preferred for the deployment of microcoils as the narrower lumen helps guard against intracatheter coil “formation,” particularly when using detachable coils. Particles can be administered through the Renegade STC with the caveat that the smaller diameter can lead to aggregation of particles and occlusion of the catheter. In such cases, the catheter can be carefully flushed with a 1-mL saline-filled syringe. The Progreat microcatheter is also available in various sizes, including 2.8-Fr (0.027 in) and 2.4-Fr (0.022 in). The former is available to order as a coaxial system that comes preloaded with a hydrophilic microwire.

Selecting the Gastroduodenal Artery

The GDA most commonly arises from the common hepatic artery. Less common variants include branches off the right hepatic artery or directly off the celiac axis.33 In many cases, the GDA is accessible with a 5-Fr catheter. The catheter is advanced from the celiac artery ostium into the proper hepatic artery over a Glidewire (Terumo, Tokyo, Japan). Care should be taken to avoid arterial dissection when using a Glidewire; this is especially true for patients with surgically altered anatomy (e.g., liver transplantation). Rotating the reverse curve catheter counterclockwise as it is being advanced can help when negotiating a tortuous common hepatic artery. Once in the proper hepatic artery, the catheter is carefully withdrawn until the tip engages the GDA.

Selecting the Left Gastric Artery

Normal celiac anatomy, with the left gastric artery being the smallest of the three primary branches of the hepatogastrosplenic trunk, is seen in 89% of patients. Less common variants include direct aortic origin (4.4%) and separate gastrosplenic and hepaticomesenteric trunks (2.6%).34 The left gastric artery courses cranially in a direction counter to the orientation of the celiac trunk. This feature can make it particularly challenging to catheterize. We favor accessing the celiac trunk with a VS1 catheter. The catheter is then carefully withdrawn until the tip is directed cranially and engages the origin of the left gastric artery. A microwire and microcatheter are then advanced into the vessel (Fig. 29.6).


Embolization should be carried out both distal and proximal to the site of injury to prevent continued bleeding through a “back door.” For example, GDA embolization performed for management of a duodenal ulcer calls for coil embolization distally into the right gastroepiploic artery with extension proximally into the GDA. Adequate embolization is confirmed by superior mesenteric arteriogram to exclude back door bleeding through the pancreaticoduodenal arcade. If bleeding persists, the inferior pancreaticoduodenal arcade should be superselected via an SMA approach and embolization should be performed as distally as possible. Likewise, embolization performed via the SMA should be checked with a celiac angiogram before concluding the case (Fig. 29.7).

Distal catheterization can be limited by vasospasm and vessel tortuosity. The latter can be overcome by the use of soft-tipped microwires—the 0.014-in Hi-Torque Balance Middle Weight (BMW) guidewire (Abbott Vascular, Santa Clara, California) is a favorite of the authors. The use of road mapping technique when attempting to subselect vessels can also be helpful. Vasospasm can be managed by infusion of 200 µg nitroglycerin into the affected artery. Although vasospasm may temporarily mask bleeding in the targeted vessel, it can occasionally be a blessing in disguise by revealing bleeding from an adjacent artery.


Hemobilia is a sign of an arteriobiliary fistula and most commonly results from iatrogenic injury or trauma. Hemosuccus pancreaticus, or bleeding via the pancreatic duct, is rare and usually associated with pancreatitis or splenic artery aneurysm. Embolotherapy is the first-line treatment for hemobilia and hemosuccus pancreaticus as endoscopy is limited by its inability to reach the injured vessel.16

Embolization of the hepatic artery for hemobilia is generally well tolerated due to the dual blood supply of the liver (75% via the portal vein and 25% via the hepatic artery) (Fig. 29.8). However, in the absence of a patent portal vein with centripetal flow, hepatic artery embolization risks hepatocyte ischemia. For this reason, it is imperative that delayed arteriography of the celiac axis is performed before embolization of any hepatic branches is undertaken. Nontarget embolization of the cystic artery via reflux of embolic material (e.g., Gelfoam slurry) can be associated with cholecystitis. When possible, superselection of the targeted hepatic artery branches should be done to reduce the risk of cystic artery embolization. Hepatic abscess formation following hepatic artery embolization has also been reported.35

Even with normal portal hepatic perfusion, arterial embolization can be complicated by biliary ischemia in rare instances. Unlike hepatocytes, the intrahepatic bile ducts do not have a dual blood supply. They are perfused via a peribiliary capillary plexus, which arises from hepatic arterial branches.36 Consequently, there is a risk of biliary necrosis, stenosis, or cholangitis when embolizing the hepatic artery. The risk is theoretically greater with smaller embolic agents (e.g., particles), which cause very distal occlusion. Thus, we recommend embolization with coils when possible.

Evaluation of hemosuccus pancreaticus should involve interrogation of both the celiac artery and SMA. Angiographic findings include opacification of the main pancreatic duct or the presence of aneurysm or pseudoaneurysm.37 Pseudoaneurysms are commonly a result of chronic pancreatitis and typically occur in the splenic, gastroduodenal, or pancreaticoduodenal arteries. In one series, embolotherapy was successful in providing immediate hemostasis in 77.8% of cases.37 The risk of splenic infarction should be considered when embolization of a splenic artery aneurysm is attempted. Embolotherapy of an aneurysm or pseudoaneurysm may also be undertaken as a presurgical measure to improve hemodynamic control.


Although very uncommon, embolization for arterial bleeding of the cervical and midesophagus has been reported. Cases typically involve bleeding ulcers.3840 However, studies are limited by lack of long-term data and small patient populations. The risk of ischemia is thought to be low due to the extensive collateral capillary network created by the complex arterial supply of the esophagus41 (Table 29.4).


There is no conclusive evidence to indicate that one embolic agent is superior to the others. In practice, operator familiarity with each agent and institutional availability determine what is used. The goal of embolotherapy is to reduce blood flow to the site of bleeding without causing bowel ischemia. This is an important principle to keep in mind when deciding on an embolic agent (Table 29.5).


Coils are the most commonly used embolic agent as well as the agent of choice at our institution. They can be precisely positioned and are associated with minimal risk of infarction because they do not affect the microvasculature. They are, however, permanent and may prevent reaccessing the target vessel should bleeding recur.

Coils are available in a wide selection of sizes, allowing one to correctly match the targeted vessel diameter. It is advisable to err on the side of slightly oversizing the coils as there is inevitably some degree of vasospasm associated with catheterization. Once the vasospasm resolves, blood flow can resume around an inadvertently undersized coil and GI hemorrhage will recur. Grossly oversized coils will not form properly within the vessel lumen and will provide a less effective scaffold for thrombus.

We favor the use of standard 0.035-in coils when possible. These larger diameter coils expedite vessel occlusion by their inherent size, and the additional steps involved in positioning the microcatheter are avoided. Moreover, 0.035-in coils are less costly than their microcoil counterparts, not to mention the added cost of the microcatheter equipment through which they are deployed. The obvious limitation to using 0.035-in coils is the inability to access the target vessel using a 5-Fr catheter alone.

Coil placement can be alternated with infusion of a slurry of gelatin sponge (Gelfoam; Pfizer, New York, New York) to create a “Gelfoam sandwich.” This technique helps expedite embolization and is especially useful in patients with underlying coagulopathy in whom thrombus is slow to develop on the fibered coil scaffolding. The Gelfoam sandwich technique also helps limit the number of coils needed to embolize larger and longer vessels such as the GDA (Fig. 29.9).

When the diameter of the target vessel is in doubt or there is a dubious landing zone for coil placement, detachable coils are a useful tool. We prefer the Interlock Fibered IDC Occlusion System (Boston Scientific Corporation, Natick, Massachusetts). These allow the operator to retract and reposition coils or even completely retrieve the coil before final placement. They have the additional advantage of not requiring a separate pusher wire. This can be useful when working with limited support staff.

If inappropriately positioned, coils can preclude subsequent endovascular access to the targeted lesion or vessel. In such instances, surgical intervention may be the only remaining treatment option should bleeding recur. Fortunately, if bleeding continues after coil embolization, it is usually much less severe and the patient is more hemodynamically stable, thus allowing for surgical or endoscopic therapy to be undertaken in more optimal conditions.

Gelatin Sponge

Use of Gelfoam alone provides a variable degree of short-term hemostasis as embolized vessels will recanalize over 2 to 6 weeks.27,42 In theory, this should cause a lesser degree of bowel ischemia compared with other embolic agents. However, because of the rich collateralization of upper GI arterial system, there is less concern for bowel ischemia when treating UGIB. Thus, permanent agents, such as coils, are used without hesitation. Patients with prior bowel surgery (e.g., Whipple procedure) are an exception to this rule. This population has altered vascular anatomy and collateral circulation may be diminished or absent. In such cases, coil embolization may be relegated to the second option after Gelfoam pledgets because of the heightened concern for bowel ischemia. Keep in mind, however, that bowel necrosis can occur within 8 to 12 hours in the setting of acute mesenteric ischemia. Thus, any decision to perform embolization in patients with surgically altered anatomy should be made after consultation with surgical colleagues.

As previously noted, gelatin sponge is used most commonly as a component of a Gelfoam sandwich in conjunction with coils. Gelfoam may also be useful when embolizing hepatic arteries in the setting of compromised portal vein patency. In these cases, Gelfoam may be a better option to address acute hemorrhage while allowing for future blood flow to hepatocytes and bile ducts after recanalization of the vessel.

Gelfoam pledgets can be administered using a microcatheter. We recommend using one of a larger caliber (≥0.27 in) to avoid catheter occlusion. A 1-mL syringe of saline can be helpful to clear catheter occlusions when administering Gelfoam. Note that nontarget embolization is a potential negative repercussion to using Gelfoam in this manner.


Neoplasm-induced hemorrhage is the lone setting in which the use of particles is generally agreed to be safe and possibly advantageous. Although surgery is the only definitive treatment for neoplasm, particle embolization can be used as temporizing measure in cases of emergent GI hemorrhage from primary or metastatic GI tumors.43 Particle embolization has also been successful in shrinking and devascularizing tumors before surgical resection.44

Unlike with benign lesions of the GI tract, ischemia at the arteriole or capillary level is sought when treating bleeding tumors. Use of particle sizes as small as 200 µm has been reported to be technically successful in the embolization of primary GI tumors and is not associated with bowel ischemia or postembolization syndrome.44 Care must be taken to ensure no arteriovenous shunting is present on preembolization selective angiography. This is especially true when using smaller particle sizes. The authors recommend using particles no smaller than 500 µm to avoid the risk of bowel ischemia from nontarget embolization. Reflux into nontarget arteries should also be strictly avoided.

Because of their small size and potential to reach the level of the intramural vasculature, particulate agents, such as polyvinyl chloride or trisacryl gelatin microspheres, are theoretically associated with an increased risk of bowel infarction and organ necrosis. For this reason, we advise against the use of particulate agents in the treatment of nontumoral UGIB.

Liquid Embolics

The primary advantage of N-butyl cyanoacrylate, or glue, (TruFill NBCA; Cordis Neurovascular, Miami Lakes, Florida) and ethylene vinyl alcohol copolymer (Onyx) is that they cause mechanical occlusion of vessels and do not rely on the patient’s ability to form thrombus. They work by polymerizing upon exposure to an ionic environment (blood) and immediately forming a cast of the vessel, causing permanent occlusion. Thus, they are ideal agents when the patient has an uncorrected coagulopathy. Their potential use is affirmed when considering that coagulopathy is associated with a 2.9-fold greater risk of embolization failure and a 9.6-fold greater risk of death from bleeding after embolization.27

Glue and Onyx are also suitable in cases complicated by small and tortuous vasculature where precise delivery of coils is not possible. They can also be delivered through smaller (0.010 in) catheters than microcoils, which may be beneficial in small, tortuous, or spastic vessels.45 These agents are also useful when embolizing pseudoaneurysms with multiple branch vessels that cannot be completely excluded with coils. Embolization of all collateral vessels is necessary in these cases to prevent continued bleeding through retrograde pathways.

The microcatheter should be positioned as close as possible to the site of bleeding. When using NBCA, the degree of dilution with iodized oil should be determined based on the distance of embolization target from the catheter as well as the rate of injection by the operator. A greater ratio of Ethiodol to NBCA increases the polymerization time, allowing it to travel more distally. Polymerization of NBCA occurs within seconds.46 Alternatively, coil embolization before NBCA infusion can be done to help control the flow of glue or to prevent nontarget embolization.47 The end point of infusion is extravasation from the bleeding site or complete filling of the target vessel. Only a small amount of the NBCA–iodized oil mixture, on the order of 1 mL, is typically needed. Once injection is complete, the microcatheter should be removed immediately to prevent adherence to the catheter wall.

The primary concern in using glue or NBCA is that the extent of vascular penetration can be difficult to control, resulting in increased risk of ischemia and nontarget embolization. Conversely, if adequate embolization distal to the target is not achieved, recurrent bleeding may result from collateral flow. Consequently, these agents require more operator experience and diligence. In a study by Lang et al.,42 a high prevalence of duodenal stricture as a late complication of embolization using 6-cyanoacrylate was postulated to be secondary to the embolization technique used by the operators, in which terminal muscular branches were embolized. Yata et al.47 found evidence of ulcers in 3 of 10 patients who underwent postembolization endoscopic evaluation within the first week after treatment. However, all cases showed improvement on follow-up endoscopy and none required surgical intervention. Interestingly, the same study also reported hepatic abscesses in 2 of 2 patients who underwent hepatic artery embolization. Both were successfully treated by percutaneous drainage.

NBCA has been reported to successfully achieve immediate hemostasis in 88% to 94% of cases of UGIB, with recurrent bleeding seen in only 6% to 7%.47,48 Jae et al.49 reported an 83% clinical success rate in patients with underlying coagulopathy and acute UGIB. It is this subset of patients in whom we recommend liquid embolic agents.

Onyx is a much more forgiving liquid embolic agent compared with NBCA. It is less likely to result in catheter adhesion and its lavalike consistency allows for more controlled infusion by the operator.50 The extent of distal diffusion within the vasculature depends on the rate of injection. Onyx may be impractical from a cost perspective as it is itself more expensive and also requires the use of dimethyl sulfoxide–compatible catheters (Fig. 29.10).


Vasopressin causes vasoconstriction of the smooth muscle of the splanchnic blood vessels and the bowel wall which, in turn, decreases perfusion to the site of vascular injury to allow for clot formation. Thus, as with embolotherapy, procedural success depends on a normal coagulation cascade. Historically, vasopressin infusion was considered in cases where embolization was not technically achievable. However, unlike for lower GI bleeding, vasopressin has not been shown to be effective for UGIB. The relatively larger vessels from which UGIB usually arises may not constrict to the same degree as smaller branches associated with lower GI bleeding.51


If a patient is on aspirin, it should be resumed when the cardiovascular risks outweigh the risk of rebleeding. This determination should be made after consultation with all those involved in the care of the patient. A PPI should be considered in patients who developed acute GI bleeding while taking aspirin or clopidogrel (in spite of the potential clopidogrel–PPI interaction).


A review of studies by Loffroy et al.28 found that the overall technical and clinical success of embolization in UGIB were 93% and 67% respectively, with a 33% rebleeding rate. Repeat embolization was successful in approximately half of these patients with rebleeding.28 Technical failure can be attributed to difficult anatomy such as vessel tortuosity or stenosis. Schenker et al.25 demonstrated that patients who underwent clinically successful embolotherapy were 13.3 times more likely to survive than those with unsuccessful procedures. The same study found sixfold greater success rate of embolotherapy when done for trauma or iatrogenic injury.25 This suggests that focal vascular injuries respond better to embolotherapy than inflammatory or neoplastic processes. Schenker et al.25 also found that patients with successful embolization had one-sixth the mortality rate of those with failed embolization. The incidence of surgical intervention for patients with clinically unsuccessful arterial embolization is 9% to 20%.28,52,53

One study found UGIB to be more resistant to hemostasis, with a higher rate of early rebleeding, than lower GI hemorrhage.25 This was hypothesized to be secondary to refilling of injured vessels through collateral circulation distal to the point of embolization. It is also important to remember that embolization does not treat the underlying pathology of UGIB such as peptic ulcer disease. In these patients, gastric acid suppression and treatment of Helicobacter pylori are important adjuncts to prevent recurrence of bleeding. UGIB also tends to be more profuse than lower GI bleeding and is associated with greater risk factors (i.e., sicker patients), leading to treatment failure. Factors associated with clinical failure of arterial embolization include bleeding secondary to trauma or invasive procedures, multiorgan failure, use of anticoagulants, underlying coagulopathy, longer time interval between onset of bleed and embolization, increased number of pRBC transfusions, postinterventional administration of fresh frozen plasma (FFP), hypovolemic shock and/or vasopressor use, corticosteroids, and the use of coils as the lone embolic agent24,25,28,52,54,55 (Table 29.6).

No prospective comparison of embolic agents has been performed to date. There is, however, some evidence that coils, when combined with polyvinyl alcohol particles or Gelfoam, are associated with lower bleeding recurrence compared with the use of coils alone.26 The authors recommend using a Gelfoam sandwich technique, in part, for this reason. These agents are widely available and most interventionalists are well versed in how to deploy them. Moreover, option of retractable coils can make up for inexperience.

The overall postembolization complication rate is 6% to 9%.52,56 Complications include access site hematoma, arterial dissection, contrast nephropathy, and nontarget embolization. Bowel ischemia or infarction can be caused by embolization too far distal in the vascular bed. This is of concern primarily when using particles or liquid embolic agents. Additionally, one must be cognizant that the normally rich collateral blood supply of the upper GI tract that protects against ischemia is compromised in patients who have had prior surgery or radiation therapy. A review of studies by Mirsadraee et al.56 found mortality rate secondary to technical failure or procedural complication ranged from 0% to 33%.


Endovascular embolization has become the de facto second-line treatment after endoscopy because it is minimally invasive and avoids laparotomy in critically ill patients, although it is interesting to note that no significant survival benefit over surgery has been demonstrated in the literature. Although transarterial embolization is associated with fewer complications, it has a higher rebleeding rate when compared with surgery.57,58 The available data is only retrospective in nature and nearly always confounded by older patient populations with more comorbidities in the embolization groups.


Acute variceal bleeding is associated with a high early mortality rate of up to 30%.59,60 Variceal sources of GI bleeding are distinct from arterial bleeding both in etiology and endovascular treatment. For these reasons, it is important to distinguish between nonvariceal and variceal sources of hemorrhage at the outset. Sources of variceal UGIB include gastroesophageal varices from portal venous hypertension and gastric varices from splenic vein thrombosis. Thirty percent of patients with portal hypertension who present with UGIB actually have an arterial source of bleeding.61

Active variceal hemorrhage accounts for about one-third of all deaths related to cirrhosis.62 Variceal bleeding stops spontaneously only in approximately 50% of patients, which is considerably less than in arterial UGIB.6365Following cessation of active hemorrhage, there is a high risk of recurrent hemorrhage. The greatest risk is within the first 48 to 72 hours, and over 50% of all early rebleeding episodes occur within the first 10 days.66 Each episode of bleeding carries a 30% mortality, with mortality rates approaching 70% to 80% in patients with continued bleeding.67,68 The risk of rebleeding is high (60% to 70%) until the gastroesophageal varices are treated.69 Risk factors for early rebleeding include age older than 60 years, renal failure, large varices, and severe initial bleeding as defined by a hemoglobin level below 8 g/dL at admission.66 The goals of management during an active bleeding episode are hemodynamic resuscitation, prevention and treatment of complications, and treatment of bleeding.

Endoscopic therapy is currently the definitive treatment of choice for active variceal hemorrhage and can be performed at the time of diagnostic endoscopy. Two forms of endoscopic treatment are commonly used: sclerotherapy and variceal band ligation. Urgent endoscopic and/or pharmacologic treatments nevertheless fail to control bleeding in approximately 10% to 20% of patients, and more definitive therapy such as portosystemic shunt creation must be immediately instituted.7 Although balloon tamponade is an effective way to achieve short-term hemostasis, due to complications of rebleeding upon balloon deflation, its use is generally reserved for temporary stabilization until more definitive treatment can be instituted.

Transjugular Intrahepatic Portosystemic Shunt

Portal venous hypertension is most commonly attributable to cirrhosis and Budd-Chiari syndromes. Reduction of the portal–venous gradient usually necessitates a transjugular intrahepatic portosystemic shunt (TIPS) creation with or without concomitant variceal embolization. A portosystemic gradient less than 12 mm Hg is associated with a lower risk of bleeding recurrence. Embolization of varices is not routinely performed at the time of TIPS at our institution unless it is in the setting of acute ongoing variceal bleeding. A retrospective study by Tesdal et al.70 demonstrated that the incidence of rebleeding is lower in cases of TIPS with variceal embolization compared with TIPS alone. However, this study did not reveal a statistically significant difference in survival between the two cohorts. We routinely place 10-mm diameter Viatorr stents (W. L. Gore & Associates, Inc., Newark, Delaware) and dilated them as needed to achieve the desired portosystemic gradient. This is typically achieved at 8 mm. If bleeding recurs in the short-term, the stent is fully dilated to 10 mm and additional attempts at variceal embolization are made.

Balloon-Occluded Retrograde Transvenous Obliteration

Gastric varices represent a slightly different pathology and hemodynamic issue than esophageal varices. Most gastric varices are due to portal hypertension, whereas others are secondary to splenic vein thrombosis. Balloon-occluded retrograde transvenous obliteration (BRTO) is a highly effective and minimally invasive treatment for gastric varices, particularly in patients who are not suitable candidates for TIPS due to poor hepatic reserve. BRTO is widely accepted in Japan, with growing use worldwide. This technique uses an occlusion balloon to control the blood flow through prominent draining veins of portosystemic shunts (most commonly a gastrorenal shunt) contributing to the gastric varices. With the shunt outflow occluded, the goal is to sufficiently fill the variceal complex with a sclerosing agent and obliterate the gastric varices without refluxing into the systemic or portal circulation. Successful treatment relies on an understanding of the anatomy and hemodynamic patterns of the gastric varices. For a detailed discussion of BRTO, please refer to Chapter 31.


Gastroduodenal Artery Embolization

• Embolize distally into the GDA at its junction with the right gastroepiploic artery; this is commonly demarcated by a right angle turn of the vessel.

• Gelfoam sandwich—alternating deposition of coils and Gelfoam slurry—can expedite embolization of a lengthy vessel such as the GDA.

• Embolize as close to the GDA–hepatic artery junction as possible to ensure occlusion of all pancreaticoduodenal branches.

• Use retractable coils near GDA–hepatic artery junction to guard against coil placement into the hepatic artery.

• Check the SMA for back door bleeding.

Accessing the Left Gastric Artery

• Select the celiac artery using a reverse curve catheter (e.g., VS1).

• Carefully pull back the catheter so that the tip of the reverse curve catheter is directed cranially and engages the left gastric artery.

• Advance a microwire and microcatheter coaxially through the reverse curve catheter and into the left gastric artery.


 1. Peura DA, Lanza FL, Gostout CJ, et al. The American College of Gastroenterology Bleeding Registry: preliminary findings. Am J Gastroenterol. 1997;92:924–928.

 2. Longstreth GF. Epidemiology of hospitalization for acute upper gastrointestinal hemorrhage: a population-based study. Am J Gastroenterol. 1995;90:206–210.

 3. Barnert J, Messmann H. Diagnosis and management of lower gastrointestinal bleeding. Nat Rev Gastroenterol Hepatol. 2009;6(11):637–646.

 4. Billingham RP. The conundrum of lower gastrointestinal bleeding. Surg Clin North Am. 1997;77:241–252.

 5. Leontiadis GI, Sharma VK, Howden CW. Proton pump inhibitor treatment for acute peptic ulcer bleeding. Cochrane Database Syst Rev. 2006;(1):CD002094.

 6. Green FW, Kaplan MM, Curtis LE, et al. Effect of acid and pepsin on blood coagulation and platelet aggregation. A possible contributor prolonged gastroduodenal mucosal hemorrhage. Gastroenterology. 1978;74(1):38–43.

 7. Garcia-Tsao G, Sanyal AJ, Grace ND, et al; Practice Guidelines Committee of American Association for Study of Liver Diseases; Practice Parameters Committee of American College of Gastroenterology. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Am J Gastroenterol. 2007;102(9):2086–2102.

 8. Vreeburg RM, Snel P, de Bruijne JW, et al. Acute upper gastrointestinal bleeding in the Amsterdam area: incidence, diagnosis, and clinical outcome. Am J Gastroenterol. 1997;92:236–243.

 9. Bjorkman DJ, Zaman A, Fennerty MB, et al. Urgent vs. elective endoscopy for acute non-variceal upper-GI bleeding: an effectiveness study. Gastrointest Endosc. 2004;60:1–8.

10. Lim J, Ahmed A. Endoscopic approach to the treatment of gastrointestinal bleeding. Tech Vasc Interv Radiol. 2005;7:123–129.

11. Greenspoon J, Barkun A. A summary of recent recommendations on the management of patients with nonvariceal upper gastrointestinal bleeding. Pol Arch Med Wewn. 2010;120(9):341–346.

12. Lau JY, Sung JJ, Lam YH, et al. Endoscopic treatment compared with surgery in patients with recurrent bleeding after initial endoscopic control of bleeding ulcers. N Engl J Med. 1999;11(340):751–756.

13. Laine L, Peterson WL. Bleeding peptic ulcer. N Engl J Med. 1994;331:717–727.

14. Laine L, Jensen DM. Management of patients with ulcer bleeding. Am J Gastroenterol. 2012;107(3):345–360.

15. Gunderman R, Leef J, Ong K, et al. Scintigraphic screening prior to visceral arteriography in acute lower gastrointestinal bleeding. J Nucl Med. 1998;39:1081–1083.

16. Millward SF. ACR appropriateness criteria on treatment of acute nonvariceal gastrointestinal tract bleeding. J Am Coll Radiol. 2008;5:550–554.

17. Kuhle W, Sheiman R. Detection of active colonic hemorrhage with use of helical CT: findings in a swine model. Radiology. 2003;228:743–752.

18. Yoon W, Jeong YY, Shin SS, et al. Acute massive gastrointestinal bleeding: detection and localization with arterial phase multi-detector row helical CT. Radiology. 2006;239:160–167.

19. Yann G, Rodallec MH, Boulay-Coletta I, et al. Multidetector CT angiography in acute gastrointestinal bleeding: why, when and how. Radiographics. 2011;31:E35–E47.

20. ACR Committee on Drugs and Contrast Media. Manual on Contrast Media: Version 7. Reston, VA: American College of Radiology; 2010.

21. Ng DA, Opelka FG, Beck DE, et al. Predictive value of technetium Tc 99m-labeled red blood cell scintigraphy for positive angiogram in massive lower gastrointestinal haemorrhage. Dis Colon Rectum. 1997;40:471–477.

22. Eriksson L, Ljungdahl M, Sundbom M, et al. Transcatheter arterial embolization versus surgery in the treatment of upper gastrointestinal bleeding after therapeutic endoscopy failure. J Vasc Interv Radiol. 2008;19:1413–1418.

23. Ripoll C, Banares R, Beceiro I, et al. Comparison of transcatheter arterial embolization and surgery for treatment of bleeding peptic ulcer after endoscopic treatment failure. J Vasc Interv Radiol. 2004;15:447–450.

24. Defreyne L, Vanlangenhove P, De Vos M, et al. Embolization as a first approach with endoscopically unmanageable acute nonvariceal gastrointestinal hemorrhage. Radiology. 2001;218(3):739–748.

25. Schenker MP, Duszak R Jr, Soulen MC, et al. Upper gastrointestinal hemorrhage and transcatheter embolotherapy: clinical and technical factors impacting success and survival. J Vasc Interv Radiol. 2001;12:1263–1271.

26. Aina R, Oliva VL, Therasse E, et al. Arterial embolotherapy for upper gastrointestinal hemorrhage: outcome assessment. J Vasc Interv Radiol. 2001;12:195–200.

27. Encarnacion CE, Kadir S, Beam CA, et al. Gastrointestinal bleeding: treatment with gastrointestinal arterial embolization. Radiology. 1992;183:505–508.

28. Loffroy R, Rao P, Ota S, et al. Embolization of acute nonvariceal upper gastrointestinal hemorrhage resistant to endoscopic treatment: results and predictors of recurrent bleeding. Cardiovasc Intervent Radiol. 2010;33:1088–1100.

29. Sandhu C, Buckenham TM, Belli AM. Using CO2-enhanced arteriography to investigate acute gastrointestinal hemorrhage. AJR Am J Roentgenol. 1999;173(5):1399–1401.

30. Kim JH, Shin JH, Yoon H, et al. Angiographically negative acute arterial upper and lower gastrointestinal bleeding: incidence, predictive factors, and clinical outcomes. Korean J Radiol. 2009;10:384–390.

31. Padia SA, Geisinger MA, Newman JS, et al. Effectiveness of coil embolization in angiographically detectable versus non-detectable sources of upper gastrointestinal hemorrhage. J Vasc Interv Radiol. 2009;20:461–466.

32. Dixon S, Chan V, Shrivastava V, et al. Is there a role for empiric gastroduodenal artery embolization in the management of patients with active upper GI hemorrhage? Cardiovasc Intervent Radiol. 2013;36:970–977.

33. Loffroy RF, Abualsaud BA, Lin MD, et al. Recent advances in endovascular techniques for management of acute nonvariceal upper gastrointestinal bleeding. World J Gastrointest Surg. 2011;3(7):89–100.

34. Song SY, Chung JW, Yin YH, et al. Celiac axis and common hepatic artery variations in 5002 patients: systematic analysis with spiral CT and DSA. Radiology. 2010;255:278–288.

35. Tzeng WS, Wu RH, Chang JM, et al. Transcatheter arterial embolization for hemorrhage caused by injury of the hepatic artery. J Gastroenterol Hepatol. 2005;20:1062–1068.

36. Sakamoto I, Iwanaga S, Nagaoki K, et al. Intrahepatic biloma formation (bile duct necrosis) after transcatheter arterial chemoembolization. Am J Roentgenol. 2003;181:79–87.

37. Lermite E, Regenet N, Tuech J, et al. Diagnosis and treatment of hemosuccus pancreaticus: development of endovascular management. Pancreas. 2007;34(2):229–232.

38. Park JH, Kim HC, Chung JW, et al. Transcatheter arterial embolization of arterial esophageal bleeding with the use of N-butyl cyanoacrylate. Korean J Radiol. 2009;10:361–365.

39. Kos X, Trotteur G, Dondelinger RF. Delayed esophageal hemorrhage caused by a metal stent: treatment with embolization. Cardiovasc Intervent Radiol. 1998;21(5):428–430.

40. Michal JA, Brody WR, Walter J, et al. Transcatheter embolization of an esophageal artery for treatment of a bleeding esophageal ulcer. Radiology. 1980;134:246.

41. Vogten JM, Overtoom TT, Lely RJ, et al. Superselective coil embolization of arterial esophageal hemorrhage. J Vasc Interv Radiol. 2007;18(6):771–773.

42. Lang EK. Transcatheter embolization in management of hemorrhage from duodenal ulcer: long-term results and complications. Radiology. 1992;182:703–707.

43. Fidelman N, Freed RC, Nakakura EK, et al. Arterial embolization for the management of gastrointestinal hemorrhage from metastatic renal cell carcinoma. J Vasc Interv Radiol. 2010;21:741–744.

44. Kurihara N, Kikuchi K, Tanabe M, et al. Partial resection of the second portion of the duodenum for gastrointestinal stromal tumor after effective transarterial embolization. Int J Clin Oncol. 2005;10:433–437.

45. Frodsham A, Berkmen T, Ananian C, et al. Initial experience using N-butyl cyanoacrylate for embolization of lower gastrointestinal hemorrhage. J Vasc Interv Radiol. 2009;20:1312–1319.

46. Pollack JS, White RI. The use of cyanoacrylate adhesives in peripheral embolization. J Vasc Interv Radiol. 2001;12(8):907–913.

47. Yata S, Ihaya T, Kaminou T, et al. Transcatheter arterial embolization of acute arterial bleeding in the upper and lower gastrointestinal tract with N-butyl-2-cyanoacrylate. J Vasc Interv Radiol. 2013;24:422–431.

48. Lee CW, Liu KL, Wang HP, et al. Transcatheter arterial embolization of acute upper gastrointestinal tract bleeding with N-butyl-2-cyanoacrylate. J Vasc Interv Radiol. 2007;18:209–216.

49. Jae HJ, Chung JW, Jung AY, et al. Transcatheter arterial embolization of nonvariceal upper gastrointestinal bleeding with N-butyl cyanoacrylate. Korean J Radiol. 2007; 8:48–56.

50. Lenhart M, Paetzel C, Sackmann M, et al. Superselective arterial embolization with a liquid polyvinyl alcohol copolymer in patients with acute gastrointestinal hemorrhage. Eur Radiol. 2010;20(8):1994–1999.

51. Darcy M. Treatment of lower gastrointestinal bleeding: vasopressin infusion versus embolization. J Vasc Interv Radiol. 2003;14:535–543.

52. Lundgren JA, Matsushima K, Lynch FC, et al. Angiographic embolization of nonvariceal upper gastrointestinal bleeding: predictors of clinical failure. J Trauma. 2011;70:1208–1212.

53. Jairath V, Kahan BC, Logan RFA, et al. National audit of the use of surgery and radiological embolization after failed endoscopic haemostasis for non-variceal upper gastrointestinal bleeding. Br J Surg. 2012;99(12):1672–1680.

54. Poultsides GA, Kim CJ, Orlando R III, et al. Angiographic embolization for gastroduodenal hemorrhage: safety, efficacy, and predictors of outcome. Arch Surg. 2008;143:457–461.

55. Mensel B, Kuhn J, Kraft M, et al. Selective microcoil embolization of arterial gastrointestinal bleeding in the acute situation: outcome, complications, and factors affecting treatment success. Eur J Gastroenterol Hepatol. 2012;24(2):155–163.

56. Mirsadraee S, Tirukonda P, Nicholson A, et al. Embolization for non-variceal upper gastrointestinal tract haemorrhage: a systematic review. Clin Radiol. 2011;66(6):500–509.

57. Ang D, Teo EK, Tan A, et al. A comparison of surgery versus transcatheter angiographic embolization in the treatment of nonvariceal upper gastrointestinal bleeding uncontrolled by endoscopy. Gastroenterol Hepatol. 2012;24:929–938.

58. Wong TC, Wong KT, Chiu PW, et al. A comparison of angiographic embolization with surgery after failed endoscopic hemostasis to bleeding peptic ulcers. Gastrointest Endosc. 2011;73:900–908.

59. de Franchis R. Evolving consensus in portal hypertension. Report of the Baveno IV consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol. 2005;43:167–176.

60. Garcia-Tsao G, Sanyal AJ, Grace ND, et al. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology. 2007;46:922–938.

61. Lee E, Laberge J. Differential diagnosis of gastrointestinal bleeding. Tech Vasc Interv Radiol. 2004;7(3):112–122.

62. Avgerinos A, Armonis A. Balloon tamponade technique and efficiency in variceal hemorrhage. Scand J Gastroenterol Suppl. 1994;207:11.

63. Brett BT, Hayes PC, Jalan R. Primary prophylaxis of variceal bleeding in cirrhosis. Eur J Gastroenterol Hepatol. 2001;13:349.

64. Gines P, Cardenas A, Arroyo V, et al. Management of cirrhosis and ascites. N Engl J Med. 2004;350:1646.

65. Infante-Rivard C, Esucola S, Villeneuve JP. Role of endoscopic variceal sclerotherapy in the long term management of variceal bleeding: a metaanalysis. Gastroenterology. 1989;96:1087.

66. de Franchis R, Primignani M. Why do varices bleed? Gastroenterol Clin North Am. 1992;21(1):85.

67. Smith JL, Graham DY. Variceal hemorrhage: a critical evaluation of survival analysis. Gastroenterology. 1982;82:968.

68. DeDombal FT, Clarke JR, Clamp SE, et al. Prognostic factors in upper GI bleeding. Endoscopy. 1986;18:6s.

69. Graham DY, Smith JL. The course of patients after variceal hemorrhage. Gastroenterology. 1981;80:800.

70. Tesdal IK, Filser T, Weiss C, et al. Transjugular intrahepatic portosystemic shunts: adjunctive embolization of collateral vessels in the prevention of variceal rebleeding. Radiology. 2005;236(1):360–367.