Radiologic Interventions for Gastrointestinal Bleeding

Abstract

Gastrointestinal (GI) bleeding is a frequent and potentially life-threatening emergency for which endoscopy remains the primary diagnostic and therapeutic modality. However, a subset of patients experiences persistent or recurrent bleeding, that requires stabilization before endoscopy or presents with lesions, that are inaccessible or unsuitable for endoscopic treatment. This review outlines the role of radiological intervention in these clinical settings. This article reviews the use of computed tomography angiography as the principal preprocedural imaging modality, focusing on its ability to localize active bleeding, characterize the underlying etiologies, and delineate the vascular anatomy relevant to catheter-based interventions. The indications, techniques, and outcomes of transcatheter arterial embolization for non-variceal GI bleeding are also summarized, including the application of superselective catheterization and commonly used embolic materials. This review describes the hemodynamically targeted interventional approaches for variceal GI bleeding associated with portal hypertension. The technical principles and clinical roles of the transjugular intrahepatic portosystemic shunt (TIPS) for portal decompression and balloon-occluded retrograde transvenous obliteration (BRTO) and its variants for gastric varices are discussed. Percutaneous variceal embolization has also been reviewed as an adjunctive option in selected situations in which endoscopic therapy, TIPS, or BRTO is not feasible or unsuccessful. Overall, this review summarizes the current imaging strategies and interventional techniques for GI bleeding, emphasizing patient selection based on bleeding etiology, vascular and portal venous anatomy, and hemodynamic status.

INTRODUCTION

Gastrointestinal (GI) bleeding is a common and potentially life-threatening emergency that requires timely diagnosis and coordinated multidisciplinary management. Although endoscopy remains the primary diagnostic and therapeutic modality, approximately 10%–20% of patients experience persistent or recurrent hemorrhage, requiring stabilization before endoscopy, or present with lesions that are inaccessible or not amenable to endoscopic treatment [1].

Advances in computed tomography angiography (CTA) have transformed GI bleeding management by enabling rapid non-invasive localization, etiologic characterization, and vascular mapping for interventional planning.

Transcatheter arterial embolization (TAE) is now the definitive therapy for non-variceal GI bleeding (NVGIB) when endoscopy fails, supported by its improved safety and efficacy achieved through superselective techniques and modern embolic materials. In variceal GI bleeding (VGIB) related to portal hypertension, hemodynamically tailored interventions are required, including transjugular intrahepatic portosystemic shunt (TIPS) for portal decompression, balloon-occluded retrograde transvenous obliteration (BRTO) and its variants for gastric varices with large portosystemic shunts, and percutaneous variceal embolization (PVE) for anatomically challenging or refractory cases (Fig. 1 and Table 1).

Fig. 1.

Schematic diagram of gastric varix (A), TIPS (B), BRTO (C), and PVE (D). GRS, gastrorenal shunt; GV, gastric varix; HV, hepatic vein; IVC, inferior vena cava; LGV, left gastric vein; LPV, left portal vein; RPV, right portal vein; RV, renal vein; SMV, superior mesenteric vein; SV, splenic vein; TIPS, transjugular intrahepatic portosystemic shunt; BRTO, balloon-occluded retrograde transvenous obliteration; PVE, percutaneous variceal embolization.

Table 1.

Key characteristics of radiological interventions for gastrointestinal bleeding

Procedure	Main role	Core mechanism	Major strengths	Major considerations
TAE	• Definitive treatment for non-variceal bleeding refractory to endoscopy	• Superselective arterial embolization via catheter-based angiography → reduction of arterial inflow and cessation of hemorrhage	• High hemostasis rate (>85%–95%)	• Bowel ischemia (risk generally low with superselective technique)
			• Minimally invasive	• Non-target embolization
			• Applicable to diverse bleeding etiologies	• Access-site complications
TIPS	• Portal decompression for uncontrolled variceal bleeding or portal hypertensive complications	• Creation of an intrahepatic portosystemic shunt (portal → hepatic vein) → portal decompression	• Most effective for portal hypertension–mediated bleeding	• Hepatic encephalopathy
			• Reduces recurrent variceal bleeding	• Shunt stenosis or occlusion
			• Improves portal hypertensive gastropathy	• Heart failure aggravation
BRTO (PARTO/CARTO)	• Obliteration of gastric varices associated with portosystemic shunts	• Retrograde shunt occlusion with variceal embolization; occlusion of portosystemic shunt with intravariceal sclerosis/embolization → direct obliteration of gastric varix	• Best treatment for gastric varices with a portosystemic shunt	• Portal pressure–related effects (ascites, esophageal variceal progression)
			• Improves hepatic encephalopathy
			• PARTO/CARTO avoid sclerosant-related complications
PVE	• Salvage or alternative therapy for variceal bleeding when other options are not feasible	• Antegrade embolization of varices via portal venous access → direct variceal occlusion	• Useful when TIPS or BRTO cannot be performed	• Systemic embolization via collaterals
			• Effective for ectopic or anatomically complex varices	• Bleeding from puncture tract
				• Portal vein thrombosis

TAE, transcatheter arterial embolization; TIPS, transjugular intrahepatic portosystemic shunt; BRTO, balloon-occluded retrograde transvenous obliteration; PARTO, plug-assisted retrograde transvenous obliteration; CARTO, coil-assisted retrograde transvenous obliteration; PVE, percutaneous variceal embolization.

This review summarizes the current evidence-based interventional strategies for GI bleeding, focusing on imaging evaluation, indications, technical principles, clinical outcomes, complications, and post-procedural follow-up of TAE, TIPS, BRTO, and PVE.

PRE-PROCEDURAL IMAGING

CTA

Pre-procedural radiological evaluation is essential for identifying the cause of GI bleeding and determining the appropriateness of catheter-based interventions. CTA is the primary imaging modality because it is noninvasive, rapidly acquired, widely available, and highly sensitive for the detection of active hemorrhages in diverse clinical settings (Fig. 2) [1,2]. In addition to pinpointing the bleeding site, CTA characterizes the underlying etiologies, such as peptic ulcer disease, pseudoaneurysms, vascular malformations, or malignancy, and provides a vascular roadmap to guide angiography and intervention.

Fig. 2.

A 69-year-old woman with duodenal ulcer bleeding. Two sessions of endoscopic hemostasis using hemoclips and epinephrine spray were performed 1 week earlier. A: CT scan in the arterial phase shows extravasation of contrast media (arrows) in the duodenum. Note the hemoclip (arrowhead). B: Common hepatic angiogram shows extravasation of contrast media (arrow) adjacent to the hemoclip (arrowhead). C: Superselective embolization was performed using a mixture of NBCA and Lipiodol (black arrows). Note the hemoclip (arrowhead) and the microcatheter tip (white arrow) positioned in a small branch of the gastroduodenal artery. NBCA, n-Butyl cyanoacrylate.

CTA is performed without oral contrast, using a standardized triphasic protocol to maximize the detection and characterization of GI bleeding. The non-contrast phase is essential for identifying pre-existing hyperattenuating materials such as intraluminal blood or metallic artifacts, which may mimic contrast extravasation (Fig. 3). The late-arterial phase captures optimal visceral arterial opacification with early portal venous filling, enabling the detection of newly developed hyperattenuating foci consistent with active bleeding and providing detailed visualization of visceral arterial anatomy for interventional planning [2,3]. The portal venous or delayed phase enables confirmation of active contrast extravasation and comprehensive assessment of the associated abdominal pathology, including cirrhosis, portal hypertension, post-operative anatomy, and bowel ischemia. It also allows the evaluation of portal and hepatic venous patency and the detection of hepatic tumors or biliary obstruction. Because this phase covers the liver, spleen, and portosystemic collateral pathways, it is particularly important in portal venous interventions such as TIPS, BRTO, or PVE, as it facilitates the evaluation of hepatic-portal venous relationships, identification of gastrorenal or alternative shunts, and recognition of anatomic variants relevant to procedural planning [4-8].

Fig. 3.

A 27-year-old woman with rectal bleeding. A: CT scan in the arterial phase shows focal extravasation of contrast media (arrow) in the rectum. B: Inferior mesenteric angiogram shows extravasation of contrast media (arrow) in the rectum. The superior rectal artery was selectively embolized with a mixture of NBCA and Lipiodol (not shown). Although embolization achieved initial hemostasis, rebleeding developed 2 days after the procedure. C: Unenhanced CT scan obtained 2 days after embolization shows residual NBCA-Lipiodol mixture (arrows) from the first embolization. D: CT scan in the portal venous phase obtained 2 days after embolization shows recurrent extravasation of contrast media (arrowhead) in the rectum and residual NBCA-Lipiodol mixture (arrows). E: Inferior mesenteric angiogram shows no extravasation of contrast media, with subtraction artifacts from the residual NBCA-Lipiodol mixture (arrows). F: Right internal iliac angiogram demonstrates extravasation of contrast media (arrow). G: Selective angiogram of the inferior rectal artery arising from the internal iliac artery shows extravasation of contrast media (arrow). The inferior rectal artery was selectively embolized with a mixture of NBCA and Lipiodol (not shown). NBCA, n-Butyl cyanoacrylate.

Thin-section images (≤3 mm) are preferred for evaluating bleeding foci, culprit arterial anatomy, and afferent/efferent venous pathways of varices. Historically, CTA detection thresholds were reported at 0.3 mL/min [2,3]; however, modern scanners may detect bleeding rates as low as 0.1 mL/min [9]. Meta-analyses report CTA sensitivity of 82%–94% and specificity of 74%–92% for diagnosis of acute GI bleeding, with higher sensitivity in acute hemorrhage (91%) than in obscure bleeding (50%–81%) [10]. Limitations include radiation exposure and the risks associated with iodinated contrast agents.

Scintigraphy

Scintigraphy using technetium-99m-labeled red blood cells is a valuable adjunct for evaluation of GI bleeding, particularly in cases of slow or intermittent hemorrhage. This modality avoids the use of iodinated contrast, does not require bowel preparation, and offers an extremely low bleeding detection threshold of approximately 0.05–0.1 mL/min. Prolonged image acquisition enables the identification of intermittent bleeding episodes that may be missed using short-time-window imaging modalities [2].

Nonetheless, scintigraphy provides limited information regarding the bleeding etiology, lacks meaningful vascular mapping for interventional planning, and is susceptible to false positives, such as splenosis or extra-enteric tracer uptake. Additional drawbacks include poor anatomical resolution, long examination times, and limited applicability in unstable patients.

Current ACG/SAR guidelines recommend CTA as the first-line imaging modality after endoscopy because of its rapid acquisition (<30 min), high anatomical accuracy, and direct procedural relevance. Scintigraphy is recommended as a second-line intervention when CTA results are negative or when bleeding is too slow for angiographic detection in clinically stable patients [11].

NVGIB

Acute NVGIB is a common medical emergency associated with substantial morbidity and mortality. Because NVGIB predominantly arises from the arterial circulation, its pathophysiology and management differ fundamentally from those of portal hypertension-related variceal bleeding. Although endoscopic therapy remains the first-line treatment, a considerable proportion of patients experience persistent or recurrent hemorrhages, have lesions that are not amenable to endoscopic management, or are unsuitable for endoscopy because of clinical instability or technical limitations [12,13].

Since TAE was first introduced by Rösch and Dotter in 1972, it has evolved into a minimally invasive, safe, and effective alternative to surgery for endoscopically refractory NVGIB [12,14-16].

GI bleeding is commonly classified as upper or lower based on its location relative to the ligament of Treitz. Upper GI bleeding (UGIB) is approximately six times more common than lower GI bleeding (LGIB), with peptic ulcer disease being the leading cause, followed by Mallory–Weiss tears, esophagitis, gastritis, neoplasia, and vascular malformations [17]. LGIB, which originates from the colon in approximately 80% of cases, may also arise from the small bowel (5%–10%) or represent refluxed UGIB (10%–15%), most commonly due to diverticular disease [18,19].

Indications of TAE

Although digital subtraction angiography has a higher bleeding detection threshold (≥0.5–1.0 mL/min) compared with CTA (0.1 mL/min), it remains indispensable for therapeutic embolization [2,11,20]. Direct angiographic findings include active extravasation, pseudoaneurysm formation, and arterial cutoff, whereas indirect signs include arterial spasm and luminal irregularity.

TAE is indicated in cases of massive bleeding (≥4 units transfused within 24 hours), hemodynamic instability, failure of medical therapy, or unsuccessful endoscopic hemostasis, and may also be effective for postoperative rebleeding [21,22]. In current practice, CTA findings primarily determine whether angiography should proceed as targeted TAE or is deferred.

Although contrast allergies, renal insufficiency, and uncorrectable coagulopathy are relative contraindications, TAE may be lifesaving in critical situations. Prior upper GI surgery, radiation therapy, or severe visceral atherosclerosis increases the risk of ischemic complications, but does not constitute an absolute contraindication.

Several factors should be considered when choosing between TAE and endoscopic therapy for GI bleeding. Angiography may fail to demonstrate active extravasation in cases of intermittent or minimal bleeding; therefore, endoscopic management should be prioritized. In contrast, adequate endoscopic visualization is often difficult to achieve in massive bleeding, making TAE the preferred therapeutic option. In real-world clinical practice, treatment selection is also influenced by the immediate availability of resources, particularly during nighttime hours or weekends. Accordingly, the establishment of a well-organized emergency care system is essential to ensure the timely and effective management of acute GI bleeding.

Interventional procedure of TAE

Vascular access is typically obtained via the common femoral artery, while radial or brachial access may be considered in patients with marked vascular tortuosity. Access is achieved using a micropuncture technique and upsized to a 4–6 Fr sheath, followed by selective angiography based on bleeding localization from CTA or endoscopy and digital subtraction angiography to identify active extravasation. Therefore, a comprehensive assessment of all potential arterial feeders is essential.

Once the culprit artery is identified, superselective catheterization is performed using a microcatheter (<2 Fr) over a microwire (<0.018 inch) positioned adjacent to the bleeding site (Fig. 2), and embolization is carried out using coils, gelatin sponge, polyvinyl alcohol (PVA) particles, or n-Butyl cyanoacrylate (NBCA), selected according to the vascular anatomy, flow dynamics, and the need for permanent or temporary occlusion.

Each embolic agent has its own distinct advantages. Coils provide precise deployment and preservation of the distal microvasculature but may be associated with higher rebleeding rates in coagulopathic patients and can limit future vascular re-access [23]. The PVA particles and tris-acryl microspheres are suitable for flow-directed embolization, particularly in cases of tumor-related bleeding, whereas gelatin sponges provide temporary occlusion with unpredictable recanalization. The NBCA polymerizes rapidly upon blood contact, enables coagulation- independent distal occlusion beyond the microcatheter tip, and rarely re-canalizes. Its polymerization rate can be adjusted by modifying the NBCA-to-Lipiodol ratio, allowing single-injection occlusion of both feeding and collateral vessels, which is particularly advantageous for UGIB with extensive arterial anastomoses [24]. Because of the risk of non-target embolization or catheter adhesion, NBCA use requires substantial technical expertise.

At our institution, NBCA is used in >90% of TAE procedures for GI bleeding because of its perceived advantage of reducing the risk of rebleeding (Figs. 2 and 3). Coil- or particle-based embolization reduces perfusion pressure, but may be insufficient in coagulopathic patients, leading to higher rebleeding rates. In contrast, NBCA, as a liquid embolic agent, directly seals the bleeding point and distal vascular channels via rapid polymerization, providing immediate and coagulation-independent hemostasis. This mechanism is particularly advantageous in actively bleeding lesions and patients with impaired coagulation, in whom durable occlusion of the bleeding source is critical.

All potential arterial sources supplying the bleeding site should be systematically evaluated during angiography. Thorough angiographic assessment with awareness of vascular anatomical variants is essential for accurate localization and effective treatment of GI bleeding. Esophageal bleeding may originate from the esophageal arteries arising directly from the thoracic aorta (Fig. 4), left gastric artery, or, less commonly, from accessory branches, such as an accessory left gastric artery originating from the left inferior phrenic artery [25]. Gastric bleeding most commonly arises from the left gastric, right gastric, and right gastroepiploic arteries. Less frequent sources include an accessory left gastric artery arising from the left inferior phrenic artery or left hepatic artery, and a posterior gastric artery originating from the splenic artery. In cases of rectal bleeding, angiographic evaluation should include both the inferior mesenteric and internal iliac arteries, as the arterial supply may arise from the superior rectal artery as well as the middle and inferior rectal branches.

Fig. 4.

A 70-year-old man with esophageal bleeding. A: CT scan in the arterial phase shows focal extravasation of contrast media (arrow) in the distal esophagus. B: Selective angiogram of an esophageal branch directly arising from the descending thoracic aorta shows extravasation of contrast media (arrow).

Tolerance to embolization differs substantially across the GI tract. The stomach is relatively resistant to ischemia because of its rich collateral arterial supply, and clinically significant gastric infarction is uncommon, even when embolization is not strictly superselective (Fig. 5). In contrast, the small bowel and colon have a segmental arterial supply with limited collateralization; therefore, embolization in these regions must be strictly superselective. Embolization involving more than three vasa recta is associated with a markedly increased risk of bowel ischemia and perforation.

Fig. 5.

A 63-year-old woman with gastric bleeding. A: CT scan in the arterial phase shows focal extravasation of contrast media (arrow) in the gastric antrum. B: Selective angiogram of the left gastric artery shows extravasation of contrast media (black arrow). Note the microcatheter tip (white arrow) in the left gastric artery and contrast media reflux into the right gastric artery (arrowhead). C: Superselective catheterization of the bleeding branch failed. A communicating channel between the right and left gastric arteries was embolized with coils (white arrow), and a microcatheter was advanced into the right gastric artery (arrowhead). The bleeding focus (open arrow) and right gastric artery branches (black arrows) were embolized with a mixture of NBCA and Lipiodol. Although relatively extensive embolization was performed, no ischemic complication occurred. NBCA, n-Butyl cyanoacrylate.

For duodenal bleeding, embolization strategies vary by region. Some centers outside Korea perform proximal embolization of the entire gastroduodenal artery, whereas the current practice in Korea generally favors superselective embolization of the identified culprit branch. Catheterization can be achieved via either the celiac axis or the superior mesenteric artery. Because the pancreatico-duodenal arcade provides extensive collateral connections between these vascular territories, failure to recognize this anatomy can lead to persistent or recurrent bleeding.

The angiographic detection of active extravasation is not always possible. Even in the presence of ongoing hemorrhage, extravasation is absent in approximately 10%–20% of UGIB cases because of intermittent bleeding or bleeding rates below the angiographic detection threshold [26]. In such instances, selective angiography of the suspected branch vessels may reveal occult bleeding, particularly when guided by prior CTA findings (Fig. 6). CTA-based localization of the bleeding focus facilitates targeted selective angiography, significantly improving the likelihood of identifying active extravasation and enabling effective embolization. In some cases, cone-beam CT may reveal bleeding foci that are obscured on angiography (Fig. 7).

Fig. 6.

A 40-year-old man with jejunal bleeding. A: CT scan in the arterial phase shows focal extravasation of contrast media (arrow) in the jejunum. B: CT scan in the portal venous phase shows an increased amount of extravasation of contrast media (arrow) in the jejunum. C: Superior mesenteric angiogram shows no extravasation. The dotted circle indicates the suspected bleeding focus based on CT findings. D: Selective angiogram of a jejunal branch shows extravasation of contrast media (arrow) in the jejunum.

Fig. 7.

A 48-year-old woman with colonic diverticular bleeding. A: CT scan in the portal venous phase shows extravasation of contrast media (arrow) in the ascending colon. B: Right colic angiogram shows no extravasation. C: Maximum intensity projection image from cone-beam CT obtained via the right colic artery demonstrates focal extravasation (arrow). D: Superselective embolization was performed using a mixture of NBCA and Lipiodol (black arrow). Note the microcatheter tip (white arrow) positioned in the vasa recta of the right colic artery. NBCA, n-Butyl cyanoacrylate.

When conventional angiography remains negative despite a strong clinical suspicion of ongoing bleeding, provocative angiography may be considered for selected patients. This approach involves the intra-arterial administration of vasodilators, anticoagulants, or fibrinolytic agents to unmask occult bleeding. Given the potential risk of exacerbating the hemorrhage, these techniques require careful patient selection and close hemodynamic monitoring.

In the absence of angiographically visible extravasation, empirical embolization may be considered for UGIB to reduce the risk of rebleeding from untreated culprit lesions. A metaanalysis demonstrated that empirical embolization after negative angiography results in rebleeding and mortality outcomes comparable to those of targeted embolization after positive angiography [26]. Empirical embolization can achieve favorable clinical outcomes if guided by CTA-based lesion localization and anatomical probability. Commonly selected target vessels include the gastroduodenal artery for duodenal ulcers, left gastric artery for gastric bleeding, and right gastroepiploic artery for antral lesions [26-29].

Empirical embolization is generally acceptable in UGIB because of the robust collateral arterial networks, particularly in the stomach. For example, gelatin sponge embolization may be safely performed for suspected gastric bleeding even in the absence of angiographic extravasation, especially in cases of gastric cancer-related hemorrhage. On the other hand, empirical embolization is not recommended for LGIB. Fundamental differences in vascular anatomy and collateral supply between the upper and lower GI tracts render the small bowel and colon far less tolerant for nontargeted embolization, with a substantially higher risk of bowel ischemia or infarction, particularly in patients with small bowel or colon malignancies.

Clinical outcomes of TAE

Technical success rates of TAE range from 93% to 100% in published studies [12,28,30]. A meta-analysis of 15 studies reported a clinical success rate of 67% [28], whereas more recent series demonstrated higher hemostasis rates of approximately 80%–90% in acute GI bleeding [31,32]. Rebleeding occurs in approximately 15%–25% of patients, particularly those with angiodysplasia, arteriovenous malformations, inflammatory lesions, coagulopathy, or high transfusion requirements, and can often be controlled with repeat TAE, with reported hemostasis rates of 60%–70% [12,27,28].

In UGIB, post-2000 studies reported technical success rates of 75%–100% and clinical success rates of 44%–94% [12,33-35], while in LGIB, technical success ranges from 73%–100% and clinical success from 63%–96% [32,36-42]. Collectively, these data support TAE as a durable and minimally invasive therapeutic option for a broad spectrum of NVGIB etiologies.

Safety, complications, and follow-up of TAE

Access-site complications, including hematoma, pseudoaneurysm, and arteriovenous fistula, occur in up to 6% of procedures [43]. Major complications requiring treatment, such as clinically significant bowel ischemia, hepatic infarction, abscess formation, and access-site vascular injury requiring intervention, occur in <5% of patients [28,33], with higher risk in those with prior surgery, radiation therapy, or advanced underlying vascular disease. Minor ischemic changes, including small post-embolization ulcers, are relatively common but are usually self-limiting after superselective embolization [44]. Upper GI embolization is generally well tolerated because of rich collateral circulation, whereas lower GI embolization carries a higher ischemic risk. The risk of clinically significant ischemia is very low when embolization is limited to fewer than three vasa recta. Advances in microcatheter technology and embolic delivery have reduced major ischemic complication rates by approximately 3% (range, 0%–11%) [22,28], and contemporary ≤2-Fr catheter systems have further decreased ischemic risk to <5% [45].

After successful hemostasis, definitive management of the underlying bleeding etiology should be pursued. If the initial endoscopic assessment was incomplete, repeat endoscopy is recommended to confirm the source of bleeding and to assess post-embolization ischemic injury. In the absence of recurrent bleeding or related symptoms, routine follow-up CT is generally unnecessary. Postprocedural monitoring should include serial hemoglobin measurements, hemodynamic assessment during the first 24–48 hours, and initiation of proton pump inhibitor therapy. According to the 2024 ACG/SAR consensus, follow-up should include puncture site evaluation, coagulation assessment, and selective imaging within 7–14 days, if clinically indicated [11].

VGIB

The VGIB results from portal hypertension, defined as an absolute portal venous pressure >10 mm Hg or a portosystemic pressure gradient >5 mm Hg, with a gradient >12 mm Hg being strongly associated with an increased risk of hemorrhage [46]. Progressive portal venous congestion leads to venous remodeling, hepatofugal flow, and the formation of extrahepatic portosystemic collaterals [46]. Gastroesophageal variceal bleeding is a major and severe complication of portal hypertension, accounting for 60%–90% of GI bleeding in cirrhotic patients, with annual bleeding rates of 5%–15%, 6-week mortality rates approaching 20%, and untreated rebleeding rates of up to 70% [46-48]. Esophageal varices are the most common source and are present in up to 50% of patients with cirrhosis, whereas gastric varices are less prevalent (10%–36%) but are associated with higher transfusion requirements and mortality rates of 14%–45% [46,49]. Ectopic varices account for 2%–5% of variceal hemorrhages and carry a substantially higher bleeding risk depending on the shunt anatomy and location [50,51]. Interventional radiological management includes portal decompression with TIPS, retrograde obliteration techniques such as BRTO and its variants, and antegrade PVE, with treatment selection guided by portal hemodynamics, variceal anatomy, and hepatic functional reserve.

TIPS

TIPS can be performed to decompress the portal venous system, typically when medical or endoscopic therapy fails to control variceal bleeding or portal hypertensive gastropathy, and liver transplantation is not immediately available [52]. By diverting portal inflow directly into the hepatic venous outflow, TIPS bypasses the liver and effectively reduces the portal pressure. The conceptual basis of TIPS was established in 1969 by Rösch et al. [53] using an animal model, followed by the first successful stent-based human TIPS reported by Richter et al. [54] in 1989. This approach parallels surgical portocaval shunts, with a therapeutic goal of reducing the portosystemic pressure gradient to ≤12 mm Hg, thereby decreasing variceal inflow, facilitating hemostasis, and ultimately improving portal hypertension-related complications.

Indications of TIPS

TIPS is indicated for refractory variceal hemorrhage, secondary prevention after gastric or ectopic variceal bleeding, refractory ascites, Budd–Chiari syndrome unresponsive to anticoagulative therapy, and recurrent portal hypertensive gastropathy or hepatic hydrothorax despite medical therapy [52,55,56].

Recent guidelines recommend preemptive TIPS in carefully selected high-risk patients. According to the EASL (European Association for the Study of the Liver) and AASLD (American Association for the Study of Liver Diseases) guidelines, patients who derive a survival benefit are defined as those with Child–Pugh class B >7 with active bleeding despite vasoactive therapy, Child–Pugh class C <14, or hepatic venous pressure gradient >20 mm Hg at the time of bleeding. The GAVAPROSEC trial demonstrated that preemptive TIPS significantly improved 1-year rebleeding-free survival in patients with gastric variceal bleeding (77% vs. 37%; hazard ratio, 0.25) without increasing hepatic encephalopathy incidence (35% vs. 32%) [57-59].

The absolute contraindications of TIPS include severe hepatic failure, uncontrolled sepsis, congestive heart failure, and pulmonary hypertension [46,52,59]. Because TIPS increases venous return and cardiac preload, pre-procedural cardiac evaluation, typically by transthoracic echocardiography, is essential [46,60].

The Model for End-Stage Liver Disease (MELD) score is the strongest predictor of post-TIPS outcomes, with increased mortality if the score >18 and particularly poor early survival when the score is >24, for whom elective TIPS is generally discouraged [61,62]. Nevertheless, TIPS remains a critical salvage therapy for acute uncontrolled VGIB when endoscopic or pharmacological treatments fail (Fig. 8) [63,64].

Fig. 8.

A 47-year-old man with alcoholic liver cirrhosis and esophageal variceal bleeding in whom endoscopic and medical management failed to achieve hemostasis. A: Main portal venography shows a dilated left gastric vein (white arrow), gastric varix (arrowhead), and esophageal varix (black arrow). A catheter was advanced through a recanalized umbilical vein. B: The left gastric vein was embolized with a mixture of NBCA and Lipiodol. Main portal venography shows subtraction artifacts (white arrow) from the embolic material in the left gastric vein. The splenic vein (arrowhead) and esophageal varix (black arrow) are visualized. C: A TIPS (arrow) was created between the right hepatic vein and the right portal vein. Note that the esophageal varix is no longer visualized. NBCA, n-Butyl cyanoacrylate; TIPS, transjugular intrahepatic portosystemic shunt.

Interventional procedure of TIPS

TIPS creation is typically performed via the right internal jugular venous access and catheterization of the right hepatic vein. Portal venous access is obtained by puncturing the right portal vein branch with a Colapinto needle, followed by portal venography and pressure measurements. The intrahepatic tract is balloon-dilated, and an 8–10-mm bare stent or polytetrafluoroethylene-covered stent-graft is deployed [65]. For variceal bleeding, reduction of the portosystemic pressure gradient to ≤12 mm Hg is recommended, whereas a lower target of ≤8 mm Hg is often pursued in refractory ascites. Adjunctive embolization of persistent variceal inflow is considered in selected cases to enhance hemostasis and reduce rebleeding [66-68].

Clinical outcomes of TIPS

Published case series reported technical success rates of 93%–100%, although real-world outcomes may be lower because of variability in the operator experience and patient complexity [69-71]. Clinical success rates of 97%–100% have been reported for recurrent variceal bleeding, whereas lower rates of 60%–70% are observed when TIPS is performed for refractory ascites [46,47]. Procedure-related mortality ranges from 0%–2%, and 30-day mortality from 3%–15% [69-71]. Predictors of early mortality include emergent TIPS placement, Child–Pugh score ≥12, alanine aminotransferase ≥100 IU/L, total bilirubin ≥3.0 mg/dL, pre-TIPS encephalopathy, APACHE II score ≥18, and MELD score ≥18 [71-73].

Safety, complications, and follow-up of TIPS

Complications of TIPS include procedure-related injuries, stent dysfunction, and physiological effects of portosystemic shunting [74].

Hepatic encephalopathy is the most common physiological complication, occurring in approximately 10%–44% of patients, and is usually managed medically; however, 3%–7% may require shunt reduction or occlusion [73,75-77]. Post-TIPS hepatic failure may present as acute or progressive disease. Acute failure results from the sudden diversion of portal venous inflow, causing a critical reduction in effective hepatic perfusion, particularly in patients with limited hepatic reserve. Progressive failure may develop over weeks to months because of chronic portosystemic shunting and impaired hepatocyte regeneration. In cases of severe liver failure or refractory hepatic encephalopathy, urgent TIPS reduction or complete occlusion should be considered as essential rescue strategies. Routine follow-up includes laboratory monitoring, mental status assessment, and Doppler ultrasonography to evaluate shunt patency; if findings are inconclusive or signs of portal hypertension recurrence are present, TIPS venography is warranted for definitive evaluation and intervention [66,77].

BRTO

BRTO and its technical variants are well-established interventional treatments for gastric varices. Hemostasis is achieved by occluding the portosystemic shunt with a balloon, vascular plug, or coils, followed by retrograde embolization of the varix using a sclerosant or gelatin sponge [78]. Endoscopic therapy for gastric variceal bleeding is often limited by poor visualization, access to the fundus, and rapid variceal inflow; therefore, interventional approaches such as TIPS and BRTO are frequently preferred [79,80]. As approximately 80%–85% of gastric varices are associated with large portosystemic shunts, most commonly a gastrorenal shunt, BRTO is feasible in the majority of patients [81,82]. Unlike TIPS, which indirectly reduces the variceal pressure through portal decompression, BRTO directly obliterates the varix and has shown superior efficacy in controlling gastric variceal bleeding [83]. Additionally, BRTO is less invasive, associated with fewer complications, and may improve hepatic encephalopathy and hepatic function, which is advantageous in patients with limited hepatic reserve [84]. To mitigate sclerosant-related complications and balloon rupture, plug-assisted RTO (PARTO) and coil-assisted RTO (CARTO) were developed using a vascular plug or coils and substituting a gelatin sponge for a sclerosant, thereby improving procedural safety and technical success [85,86].

Indications of BRTO

BRTO is indicated for the secondary prevention of rebleeding in patients with gastric varices and a catheter-accessible portosystemic shunt, most commonly a gastrorenal shunt. It is also an established therapeutic option for hepatic encephalopathy caused by large spontaneous portosystemic shunts and bleeding ectopic varices when a suitable retrograde venous approach is available. Although primary prophylaxis remains controversial, BRTO may be considered in selected high-risk patients, including those with large fundal varices, prior hepatic encephalopathy, or contraindications for TIPS [87,88]. PARTO and CARTO share the same fundamental indications as BRTO but may offer technical advantages in specific settings [85,86]. In actively bleeding gastric varices, the sclerosants used in BRTO may escape through the bleeding point and limit complete obliteration, whereas the gelatin sponge used in PARTO tamponade the bleeding site and facilitate more uniform filling [86].

Interventional procedure of BRTO

BRTO requires a catheter-accessible portosystemic shunt, most commonly a gastrorenal shunt, although alternative drainage pathways such as the inferior phrenic or pericardiophrenic veins may allow pre-procedural identification of the procedure [89]. The procedure is typically performed via the femoral or internal jugular venous access with catheterization of the left renal vein, followed by venography and balloon occlusion to delineate the shunt-variceal complex [49]. Collateral veins are selectively embolized when necessary.

After satisfactory opacification, the required contrast volume is determined, and contrast stagnation is confirmed prior to sclerosant injection. In Korea, 3% sodium tetradecyl sulfate is the most commonly used [90]. The sclerosant is injected until complete filling of the varix and proximal afferent vein is achieved, followed by a dwell time of 1–24 hours, after which the balloon is gradually deflated with monitoring for leakage into the inferior vena cava. The catheter is removed after confirmation of variceal thrombosis. PARTO and CARTO (Fig. 9) follow similar principles but a vascular plug or coil and gelatin sponge is used instead of sclerosant; by avoiding prolonged balloon occlusion and sclerosant-related complications, PARTO is associated with higher technical success, shorter procedural time, and improved safety, particularly in patients with active bleeding or large-volume shunts [85,86].

Fig. 9.

A 75-year-old woman with liver cirrhosis and gastric varices. A: CT scan in the portal venous phase shows a gastric varix (arrow). B: A balloon catheter (black arrow) was advanced into the gastrorenal shunt, and the gastric varix (white arrow) was opacified with contrast media. C: The gastric varix was filled with a sclerosant (white arrow), and the gastrorenal shunt was embolized with coils (black arrow). D: CT scan in the portal venous phase obtained 1 day after CARTO shows complete thrombosis of the gastric varix (arrow). CARTO, coil-assisted retrograde transvenous obliteration.

BRTO remains a useful option in selected clinical settings, particularly when a sclerosant-based strategy aimed at dense thrombosis and complete vascular fibrosis of the variceal complex is desired or when device availability and cost are important practical considerations. Although PARTO and CARTO offer shorter procedural times and avoid sclerosant-related systemic risks, BRTO may provide a more durable long-term obliteration in certain anatomical variants. Accordingly, the choice between BRTO, PARTO, and CARTO should be individualized based on shunt anatomy, bleeding status, portal hemodynamics, and institutional expertise.

Clinical outcomes of BRTO

A meta-analysis of 24 retrospective studies reported a technical success rate of 96.4% (95% confidence interval [CI], 93.7%–98.3%) and a clinical success rate of 97.3% (95% CI, 95.2%–98.8%) [91]. Technical failure most commonly results from the inability to catheterize the gastrorenal shunt, a shunt diameter exceeding the balloon capacity, complex collateral venous anatomy, or balloon rupture [92]. The reported recurrence rates of gastric varices after BRTO range from 2.7% to 6.3%, whereas hepatic encephalopathy improves in 86%–100% of patients within one week [84,93]. Randomized trials have shown lower rebleeding rates after BRTO for isolated gastric varices (5%–10%) compared with TIPS (15%–20%) [4,58]. Meta-analyses suggest a comparable overall survival between BRTO and TIPS when patient selection is individualized according to the variceal anatomy and hepatic functional reserve [11,22].

Safety, complications, and follow-up of BRTO

Common postprocedural symptoms after BRTO include mild abdominal pain, nausea, vomiting, and low-grade fever, which are generally transient. Elimination of a major portosystemic shunt increases portal venous pressure, ascites, pleural effusion, or portal hypertensive gastropathy, but typically improves within approximately four weeks as hepatic function recovers and portal hemodynamics stabilize [94-98].

Progression of esophageal varices is a recognized consequence of BRTO, with a reported rate of 33.3% (95% CI, 24.6%–42.6%) [49]. Clinically significant esophageal variceal bleeding is uncommon, occurring in approximately 5% of patients and is usually managed with endoscopic therapy [84,97]. Accordingly, early endoscopic evaluation within the first week is recommended to confirm hemostasis, followed by surveillance endoscopy at 1–3 months to document variceal regression and detect newly developed collateral vessels [58].

PVE

PVE is an antegrade approach for variceal bleeding performed via portal venous access obtained through percutaneous transhepatic or transsplenic puncture or by direct puncture of a recanalized paraumbilical vein, followed by catheter-based embolization of the varices [99]. It is described as percutaneous transhepatic or transsplenic embolization/obliteration (PTE/PTO) or balloon-occluded antegrade transvenous obliteration (BATO) based on the access route and technique.

Before the widespread adoption of endoscopic therapy, TIPS, and BRTO, PVE was commonly used for refractory gastroesophageal variceal bleeding from the mid-1970s to late 1980s [100-101]. Its role has since diminished because of the persistent portal hypertension, which may lead to recanalization or variceal transformation of collateral pathways, resulting in recurrent bleeding despite technical success [102]. Nevertheless, PVE continues to serve as a complementary option in selected patients when other treatment modalities are infeasible or unsuccessful, particularly in emergency settings, ectopic varices that are not amenable to endoscopic therapy, or isolated gastric varices associated with splenic vein thrombosis [103].

Indications of PVE

Although most forms of variceal bleeding are technically amenable to PVE, it is generally reserved for selected cases because it requires a detailed evaluation of the variceal anatomy and flow dynamics and is less familiar to many operators than TIPS or BRTO. PVE is most often considered when endoscopic therapy, TIPS, or BRTO is not feasible or has failed, particularly for ectopic varices inaccessible to endoscopy (e.g., duodenal, rectal, or stomal varices); anatomically unsuitable candidates for TIPS or BRTO; isolated gastric varices related to splenic vein thrombosis; or severe bleeding requiring rapid hemostasis [8]. Relative contraindications include severe uncorrectable coagulopathy, ascites, portal vein tumor invasion at the intended puncture site, and tumors or large cysts along the planned access tract, which do not absolutely preclude the procedure if the expected benefit outweighs the procedural risk [8].

Interventional procedure of PVE

PVE is commonly achieved via transhepatic puncture of the portal branches or transsplenic access to the splenic vein.104A recanalized paraumbilical vein, present in approximately 12%–26% of patients with portal hypertension, may be directly punctured under ultrasound guidance when identified on imaging. In stomal varices, direct ultrasonography-guided puncture of the varix is an alternative, and PVE may be performed via an existing or newly created TIPS tract [103,105].

Following portal access, venography is performed to evaluate the afferent and efferent veins, variceal size, and flow dynamics. When active bleeding is absent and hepatopetal flow is preserved, the benefit of PVE is reduced, and the risk of main portal vein thrombosis increases, warranting careful consideration before proceeding. Embolization is typically performed using coils, NBCA, or gelatin particles via a microcatheter (Fig. 10). Proximal coil occlusion alone should be avoided because collateral channels may permit rebleeding; for large or highflow varices, partial coil embolization followed by NBCA or gelatin particle injection is recommended, with the NBCA-to-Lipiodol ratio tailored to the variceal size and flow characteristics [106,107]. Completion venography is required to confirm complete occlusion and exclude nontarget embolization.

Fig. 10.

A 62-year-old man with liver cirrhosis and gastric varices. A: CT scan in the portal venous phase shows a gastric varix (arrow). B: Left gastric venogram shows the gastric varix (black arrow) and multiple draining veins, including the left phrenic vein (white arrowhead), pericardiophrenic vein (black arrowhead), and azygos/hemiazygos veins (open arrow). A catheter (white arrow) was inserted into the left gastric vein via a transhepatic approach. C: Multiple coils (white arrow) were first deployed in the left gastric vein to prevent reflux of liquid embolic materials. The gastric varix (black arrows) was embolized with a mixture of NBCA and Lipiodol. NBCA, n-Butyl cyanoacrylate.

For transhepatic or transsplenic access, the sheath should be withdrawn slowly under fluoroscopic guidance with simultaneous contrast injections to allow tract embolization using coils and NBCA [104,108]. For paraumbilical vein access, manual compression is usually sufficient, and additional embolization is performed when necessary [109].

Clinical outcomes of PVE

The clinical efficacy of PVE varies among studies. In a large 1980s case series by L’Herminé et al. [102], acute bleeding was controlled in 83% of patients, but rebleeding occurred in 55% at 6 months and 81% at 2 years. Subsequent studies reported overall success rates of 44%–100%, reflecting heterogeneity in patient selection and techniques [103]. More recent post-2010 series, although limited by small sample size, reported a high initial hemostasis rate of approximately 95% [110,111]. Rebleeding remains a major limitation, with reported rates of 10%–60% [102], largely because PVE does not correct the underlying portal hypertension and complete embolization of all afferent and efferent channels is often not technically achievable, leading to progressive risk of recurrent bleeding over time [7].

Safety, complications, and post-procedural follow-up of PVE

Serious complications may occur if embolic materials migrate into the systemic circulation through portosystemic collaterals, potentially causing cerebral, coronary, or peripheral arterial embolization [112]. Procedure-related bleeding from the hepatic or splenic puncture tract has been reported to be 0%–6.5% but may exceed 30% when tract embolization is not performed, underscoring the importance of meticulous tract hemostasis [105].

Follow-up imaging should include contrast-enhanced CT with a dedicated portal venous phase to assess variceal thrombosis, confirm response, and detect procedure-related complications. This approach enables reliable visualization of thrombosed varices and adjacent vascular anatomy, facilitates early detection of residual or recurrent flow, and supports the timely planning of repeat interventions when indicated [113].

CONCLUSION

Radiologic interventions are essential for the contemporary management of GI bleeding when endoscopy is unsuccessful, contraindicated, or technically limited. CTA has become central to the diagnostic-therapeutic pathway, providing rapid bleeding localization, etiologic characterization, and procedural vascular mapping, which directly inform interventional strategies.

For NVGIB, TAE is a definitive, minimally invasive therapy with high technical success and clinically durable hemostasis when performed using superselective technique and appropriate embolic selection.

VGIB requires hemodynamically tailored interventions. TIPS provides portal decompression for refractory or high-risk presentations, while BRTO and its variants achieve highly effective and durable control of gastric variceal bleeding in the setting of large portosystemic shunts, with the potential to improve liver function and hepatic encephalopathy. Nonetheless, PVE remains a complementary option for selected patients when endoscopic therapy, TIPS, or BRTO are infeasible or has failed.