Pathophysiological management of gastric varices: From hemodynamics to targeted therapies

Abstract

Gastric varices (GV) are a distinct and clinically challenging manifestation of portal hypertension characterized by complex vascular anatomy and unique hemodynamic behavior. Unlike esophageal varices (EV), GVs arise from interactions between afferent portal inflow, large variceal reservoirs, and spontaneous portosystemic shunts, resulting in a low-pressure, high-flow system that often responds poorly to conventional pressure-reducing therapies. Consequently, management strategies extrapolated from EV are frequently inadequate. Advances in endoscopic ultrasound (EUS), cross-sectional imaging, and interventional radiology have facilitated a paradigm shift toward a pathophysiology-driven approach. Contemporary evaluation integrates endoscopic classification based on anatomical location (Sarin), EUS for assessment of feeding vessels and flow dynamics, and computed tomography/magnetic resonance imaging for mapping afferent and efferent pathways using Kiyosue and Saad-Caldwell classifications. These frameworks enable hemodynamic triage into shunt-dominant, pressure-dominant, and complex patterns, which directly inform therapeutic selection. Endoscopic therapies, including cyanoacrylate injection and EUS-guided coil ± glue embolization, target the variceal reservoir and inflow, providing effective local control. Radiologic interventions address systemic hemodynamics, with retrograde transvenous obliteration (Balloon-occluded retrograde transvenous obliteration/plug-assisted retrograde transvenous obliteration/coil-assisted retrograde transvenous obliteration) preferred for shunt-dominant varices and transjugular intrahepatic portosystemic shunt for pressure-driven disease. Surgical options remain relevant in selected conditions such as left-sided portal hypertension. Clinical modifiers, including hepatic reserve, encephalopathy, ascites, and portal vein patency, further refine treatment decisions. A multidisciplinary, mechanism-based strategy is essential to optimize outcomes. Future research should focus on integrated classification systems and prospective comparative studies to establish standardized, individualized management algorithms.

Key Words: Gastric varices; Portal hypertension; Balloon-occluded retrograde transvenous obliteration; Transjugular intrahepatic portosystemic shunt; Endoscopic ultrasound

Core Tip: Gastric varices require a pathophysiology-driven management strategy distinct from esophageal varices. Therapy selection should be guided by hemodynamic classification, integrating Sarin (location), Kiyosue and Saad-Caldwell (inflow-outflow anatomy), and clinical factors. Shunt-dominant varices are best treated with Balloon-occluded retrograde transvenous obliteration or its variants, while pressure-dominant disease requires a transjugular intrahepatic portosystemic shunt for global decompression. Endoscopic ultrasound-guided coil ± glue therapy offers precise, flow-directed treatment in complex anatomy or high-risk patients. Clinical modifiers such as hepatic encephalopathy, ascites, liver reserve, and portal vein patency are critical in decision-making. A multidisciplinary approach combining endoscopic, radiologic, and surgical modalities is essential for optimal and durable outcomes.

INTRODUCTION

Gastric varices (GV) represent a distinct and formidable challenge in the management of portal hypertension. While less common than esophageal varices (EV), occurring in approximately 20% to 25% of patients with cirrhosis, they are associated with significantly higher morbidity and mortality[1]. Bleeding from GVs is usually severe, characterized by higher transfusion requirements, increased rates of initial treatment failure, and a substantial risk of early re-bleeding[2]. GVs arise from complex collateral networks with spontaneous shunts, forming a low-pressure, high-flow system that responds poorly to conventional pressure-reducing therapies, making management distinctly challenging[2-4]. Consequently, therapies extrapolated from EV often provide suboptimal or temporary control[2-4].

Over the past two decades, advances in endoscopic ultrasound (EUS), cross-sectional imaging, and interventional radiology have transformed the management landscape of GVs[2,4-7]. These developments have shifted therapeutic paradigms away from purely pressure-based approaches toward anatomy- and flow-directed interventions. However, the absence of a unified, pathophysiology-oriented framework continues to result in heterogeneous practice patterns and suboptimal outcomes. This review aims to bridge this gap by providing a comprehensive overview of the pathophysiological mechanisms underlying GVs and aligning them with contemporary management strategies.

PATHOPHYSIOLOGY OF GV

Portal hypertension and collateral formation

Portal hypertension is the primary driver of variceal development and results from increased resistance to portal venous flow. Once the hepatic venous pressure gradient exceeds approximately 10-12 mmHg, portosystemic collaterals begin to form as compensatory pathways to decompress the portal system[2]. However, the development of GVs is not solely dependent on portal pressure. In contrast to EV, which reflects direct transmission of elevated pressure via the left gastric vein, GVs are sustained by large collateral channels and shunts, allowing persistence even at relatively lower portal pressures[8,9]. While elevated portal pressure initiates collateralization, the persistence and expansion of GVs are mediated by shear stress–induced endothelial activation, leading to upregulation of angiogenic factors such as vascular endothelial growth factor VEGF, nitric oxide, and other vasodilatory mediators. These factors promote vascular remodeling, neoangiogenesis, and dilation of pre-existing vascular channels, particularly in the gastric fundus[10].

Segmental (left-sided) portal hypertension

A distinct mechanism for GV formation is segmental or left-sided portal hypertension, typically caused by splenic vein thrombosis secondary to pancreatitis, pancreatic tumors, or pseudocysts. In this condition, splenic venous outflow is diverted through the short gastric veins, leading to the development of isolated fundal varices (IGV1)[7]. Unlike cirrhotic portal hypertension, patients with segmental portal hypertension often have preserved liver function and normal portal pressure gradients. The varices arise from localized venous congestion rather than systemic portal hypertension. Consequently, the management strategy differs significantly; therapies aimed at portal decompression, such as TIPS, may be unnecessary, whereas splenectomy or splenic artery embolization can be curative by eliminating the pathological inflow source.

Hemodynamic characteristics of GVs

The formation and persistence of GVs depend on the balance between afferent (inflow) and efferent (outflow) pathways (Figure 1). The inflow and outflow pathways of GVs vary significantly according to the Sarin classification and are critical for understanding their hemodynamic behavior (Figure 2). GOV1, located along the lesser curvature, represents an extension of EV and is primarily supplied by the left (coronary) gastric vein, with drainage occurring through peri-esophageal collaterals into the azygos system, thereby exhibiting a pressure-dominant pattern similar to EV. In contrast, GOV2 and IGV1 (fundal varices) are predominantly supplied by the short gastric and posterior gastric veins, with outflow commonly via a gastrorenal shunt and, less frequently, via gastrocaval or inferior phrenic pathways; this configuration results in a high-flow, shunt-dominant system. IGV2 varices, which occur in ectopic gastric locations such as the antrum or pylorus, demonstrate heterogeneous inflow and outflow patterns depending on the underlying vascular anatomy and are less consistently associated with large spontaneous shunts[1]. In addition to the classical Sarin classification, a modified framework has been proposed to address cases that do not conform to conventional categories. Singh et al[11] described an additional entity, GOV3, representing varices with simultaneous involvement of the esophagogastric region and distal gastric or duodenal segments, characterized by dual drainage into both the superior and inferior vena cava systems. The submucosal veins in the gastric fundus have thicker walls and larger luminal diameters, enabling them to accommodate substantial blood flow and form large vascular reservoirs[8].

Figure 1 Anatomy of gastric fundal varices based on various afferent (inflow) and efferent (outflow) pathways. GV: Gastric varice; PV: Portal vein; LGV: Left gastric vein; PGV: Posterior gastric vein; SGV: Short gastric veins; GRS: Gastrorenal shunt; GCS: Gastrocaval shunt; IVC: Inferior vena cava; SMV: Superior mesenteric vein; LRV: Left renal vein; SV: Splenic vein.

Figure 2 Sarin classification of gastric varices. A: Gastroesophageal varices type 1; B: Gastroesophageal varices type 2; C: Isolated gastric varices type 1; D: Isolated gastric varices type 2. GOV1: Gastroesophageal varices type 1; GOV2: Gastroesophageal varices type 2; IGV1: Isolated gastric varices type 1; IGV2: Isolated gastric varices type 2; LGV: Left gastric vein; PGV: Posterior gastric vein; SGV: Short gastric veins; GEV: Gastroepiploic vein.

DIAGNOSTIC EVALUATION BASED ON PATHOPHYSIOLOGY

GVs arise from complex collateral networks involving multiple inflow and outflow pathways. Therefore, optimal evaluation requires integration of endoscopy, EUS, and cross-sectional imaging to determine variceal morphology, identify feeding vessels and shunts, and guide the selection of appropriate therapeutic interventions.

Endoscopy

Upper gastrointestinal endoscopy remains the first-line modality for diagnosing GVs and assessing the risk of bleeding. It allows direct visualization of variceal morphology, size, and location, enabling classification according to the Sarin system (GOV1, GOV2, IGV1, and IGV2) (Figure 2). Endoscopic findings such as large varices, nodularity, or red color signs may indicate an increased risk of hemorrhage[1,12]. Endoscopy also plays a crucial role in acute bleeding by facilitating immediate therapeutic interventions, such as endoscopic cyanoacrylate injection (ECI) or band ligation, in selected cases[2,4]. However, it is limited to luminal assessment and cannot define underlying hemodynamics. As a result, endoscopy alone cannot reliably identify feeding veins, perforating vessels, or spontaneous portosystemic shunts that sustain the varices. In addition, large fundal varices may be difficult to distinguish from prominent gastric folds, and the extent of collateral circulation cannot be accurately determined.

EUS

EUS has emerged as an important tool for evaluating the vascular anatomy and flow dynamics of GVs. One of the major advantages of EUS is its ability to identify larger submucosal varices, perforators and collateral channels connecting GVs to the portal and systemic venous systems, which may not be apparent on conventional endoscopy. Doppler EUS can further assess blood flow direction and velocity, thereby facilitating therapeutic planning[13,14]. It helps determine the optimal injection or coil deployment site[15], confirms successful obliteration of blood flow after treatment, and may guide repeat interventions if residual flow is detected[14]. The Arakawa classification, based on EUS assessment of vascular architecture, categorizes GVs according to their submucosal structure into two types: Type I (localized), characterized by a single large vessel arising from a feeding vein that penetrates the muscularis propria and drains into a splenorenal shunt, and Type II (diffuse), comprising a complex network of interconnected vessels within the gastric wall[16]. This information is particularly useful in determining the choice of therapy. For example, identification of a dominant feeding vessel may allow targeted treatment using EUS-guided coil embolization or glue injection, reducing the number of coils or the volume of glue required and potentially lowering the risk of embolic complications.

Cross-sectional imaging (computed tomography/magnetic resonance imaging)

Cross-sectional imaging with contrast-enhanced computed tomography (CT) or magnetic resonance imaging provides a comprehensive assessment of the portal venous system and associated collateral pathways. These modalities are particularly valuable for mapping the afferent and efferent vessels of GVs and identifying spontaneous portosystemic shunts such as gastrorenal or gastrocaval shunts. The Kiyosue classification emphasizes the number and type of feeding veins (afferent/inflow) and the pattern of drainage (single shunt vs multiple collaterals) (efferent/outflow), thereby guiding the feasibility of shunt-occlusive therapies such as balloon-occluded retrograde transvenous obliteration (BRTO) (Figure 3)[17]. In contrast, the Saad-Caldwell classification integrates inflow dominance with outflow patterns to stratify varices into clinically relevant types that inform whether shunt embolization or portal decompression transjugular intrahepatic portosystemic shunt (TIPS) is more appropriate[18]. Detailed vascular mapping is essential when considering radiologic interventions [large gastrorenal shunt: BRTO or its variants; no suitable shunt or diffuse portal hypertension: TIPS]. Cross-sectional imaging also helps detect associated conditions such as portal vein thrombosis (PVT), splenic vein obstruction, or other pathologies that may influence management decisions.

Figure 3 Kiyosue classification of gastric varices. A: Classification based on efferent pathways: Type 1 (single afferent vein), type 2 (multiple afferent veins), and type 3 (indirect inflow through collateral networks); B: Afferent pathways: Type A (single draining shunt, usually gastrorenal), type B (single shunt with collateral veins), type C (multiple draining shunts), and type D (absence of a major shunt with drainage through small collaterals).

Venography and angiography

The Hirota, Fukuda, and Matsumoto classifications describe GVs based on venographic and angiographic findings, focusing on shunt dynamics, collateral circulation, and feeder dominance, which are particularly relevant for interventional planning. However, with the advent of cross-sectional imaging, these are less commonly used as primary diagnostic modalities and are mostly employed during BRTO (Table 1). The Hirota classification is based on balloon-occluded retrograde transvenography and grades varices from Grade 1 to 5 according to the degree of variceal opacification and presence of collateral pathways, with Grade 1-2 suggestive of favorable anatomy for BRTO[19]. The Fukuda classification categorizes GVs based on the dominance of afferent (feeding) veins (left gastric vein, posterior gastric vein, short gastric veins, or complex) observed during angiography. This classification helps in understanding inflow patterns and may guide selective embolization strategies[20]. The Matsumoto classification evaluates GVs in relation to flow dynamics within the gastrorenal shunt and the left gastric vein, with particular focus on the direction of flow (hepatopetal vs hepatofugal). It is clinically useful in predicting worsening of EV after shunt embolization, with certain patterns (Type Ib) associated with increased risk following BRTO[21].

Table 1 Classification of gastric varices based on venographic and angiographic findings.

Classification	Basis	Types/grades	Key features	Clinical relevance
Hirota classification	BRTO findings	Grade 1-5	Grade 1-2: Good opacification, minimal collaterals. Grade 3 Partial opacification with collaterals. Grade 4-5: Poor/no opacification, extensive collaterals or large shunt	Predicts the feasibility and success of BRTO; higher grades = lower success
Fukuda classification	Dominance of afferent (feeding) veins on angiography	Type 1-4	Type 1 Left gastric vein dominant. Type 2 mixed (LGV + posterior/short gastric). Type 3 complex bilateral supply. Type 4 right-sided dominance	Helps identify inflow patterns and guides selective embolization
Matsumoto classification	Flow dynamics in gastrorenal shunt and LGV	Type 1-2	Type 1 portosystemic flow present. Type 2 portosystemic flow absent. Subtypes based on flow in LGV: A (hepatopetal)/B (hepatofugal)	Predicts worsening of esophageal varices after BRTO (higher risk in Type 1B)

BRTO: Balloon-occluded retrograde transvenography; LGV: Left gastric vein.

CHOICE OF TREATMENT FOR GV: A PATHOPHYSIOLOGY-DRIVEN APPROACH

Contemporary management of GVs targets the dominant hemodynamic driver, either portal pressure, shunt flow, or local inflow, rather than simply obliterating the varix. Outflow anatomy is the primary determinant of radiologic strategy, while inflow complexity guides adjunctive embolization.

Hemodynamic triage: The initial framework

The first step in therapeutic selection is the clinical context. In patients presenting with active bleeding, rapid hemostasis is paramount. ECI is the most commonly used technique due to its widespread availability, achieving hemostasis in more than 90% of cases[22-24]. However, in active variceal bleeding with significant blood in the fundus, endoscopic visualization of the GVs may be obscured during ECI. In such cases, EUS may be preferred as rescue therapy for active bleeding control[25]. Retrograde transvenous obliteration (RTO), especially plug-assisted RTO (PARTO), has also been used as salvage therapy for acute gastric variceal bleeding after failed ECI[26]. Thus, both endoscopic and radiological modalities can be used for the management of acute gastric variceal bleeding, and the choice of therapy will depend on the availability and expertise.

In secondary prophylaxis, the emphasis shifts toward durable eradication of varices. Primary prophylaxis is reserved for carefully selected, high-risk fundal varices[27]. Endoscopic evaluation defines the Sarin subtype: GOV1 is managed with endoscopic band ligation with or without non-selective beta-blockers. In contrast, GOV2 and IGV1 (cardiofundal varices) mandate further anatomical and hemodynamic assessment using CT portal venography and/or EUS. A pivotal decision point is whether the patient has generalized portal hypertension or a shunt-dominant physiology (Table 2).

Table 2 Differentiation of efferent/shunt-dominant or afferent/inflow-dominant varices based on various features.

Parameter	Efferent (shunt)-dominant varices	Afferent (inflow)-dominant varices
Primary driver	Low-resistance outflow (drain-driven)	High inflow (supply-driven)
Key imaging feature (CT/MR)	Large single shunt (e.g., gastrorenal/gastrocaval)	Multiple feeders, no dominant shunt
Collateral pattern	Limited collaterals, dominant drainage pathway	Dense collateral network
EUS findings	Prominent draining channel, rapid flow toward the shunt	Multiple perforators/feeding vessels, complex flow
Flow dynamics	High-flow toward systemic circulation	Multidirectional or inflow-heavy flow
Portal hypertension features	Less prominent	Marked (ascites, collaterals)
Clinical clues	HE, preserved liver function	Ascites, decompensation, less HE

EUS: Endoscopic ultrasound; CT: Computed tomography; MR: Magnetic resonance; HE: Hepatic encephalopathy.

Shunt-dominant GVs

Radiological techniques: In patients with Kiyosue Type A or B anatomy, GVs are sustained by a large, dominant outflow shunt, typically a gastrorenal shunt. These patients often have relatively preserved liver function, minimal ascites, and may present with recurrent bleeding or hepatic encephalopathy (HE) due to shunting of portal blood away from the liver. In this setting, RTO represents the treatment of choice. The mechanistic basis of RTO lies in occlusion of the efferent shunt, which eliminates the low-resistance outflow pathway and promotes thrombosis of the variceal complex[28]. Among RTO techniques, BRTO is particularly effective in large, high-flow shunts where prolonged sclerosant dwell time is required. PARTO and coil-assisted retrograde transvenous obliteration (CARTO) techniques have emerged as valuable alternatives, offering improved procedural efficiency and safety, especially in anatomically favorable or complex shunts[29,30]. The selection of the RTO technique is guided by shunt geometry and flow dynamics.

PARTO is advantageous when the shunt is relatively straight and of suitable caliber to allow deployment of a vascular plug, thereby reducing procedural time and eliminating the need for prolonged balloon occlusion[29]. However, PARTO has been reported to have a higher recurrence rate of GVs on long-term follow-up than BRTO[31,32]. Larger shunts (> 20 mm) may necessitate modification of the PARTO procedure (Double plug technique or glue-in plug technique), as the largest available plug is 22 mm. CARTO, on the other hand, is particularly useful in torturous, elongated, or anatomically complex shunts where navigating a balloon catheter or vascular plug into the shunt, or maintaining stable positioning within it, is not feasible[29]. In cases with collateral outflow pathways, adjunctive embolization of collateral veins may be necessary to prevent diversion of sclerosant and ensure complete obliteration. In PARTO and CARTO, venography should be performed in cases where pre-procedural imaging is inconclusive or when complex anatomy is anticipated. The use of adjunctive techniques, such as placement of a parallel 4 or 5 Fr catheter distal to the plug/coil, allows occluded venographic assessment and facilitates coaxial advancement of a microcatheter for selective embolization of residual efferent channels, thereby ensuring complete variceal obliteration. Furthermore, the incorporation of cone-beam CT during venography can significantly enhance anatomical delineation, improve identification of small collateral vessels, and refine procedural strategy[28,33].

Endoscopic techniques: In GVs, EUS-guided therapies have demonstrated high technical and clinical efficacy, with pooled technical success rates of 94%-99% and overall obliteration rates generally exceeding 90% across meta-analyses[34,35]. EUS-guided therapy appears to perform better than ECI, with better obliteration and lower recurrence/long-term rebleeding, while the coil scaffold may reduce glue migration in high-flow shunt-associated varices[36,37]. A study specifically in patients with spontaneous portosystemic shunts found that EUS-guided coil plus glue was more effective and safer than ECI alone, supporting its use in this high-risk hemodynamic setting[38]. Two recent studies have reported comparable rates of variceal obliteration and rebleeding between EUS-guided intervention and RTO[39,40]. When a dominant feeder or two identifiable feeders are present, feeder embolization is preferred, while intravariceal embolization is recommended in patients with complex collateral anatomy.

Clip-assisted cyanoacrylate injection (Clip-ECI) is a modification of conventional glue therapy in which one or more endoscopic clips are deployed at the variceal neck or puncture site before or after injection to reduce flow and mechanically retain the glue cast[41]. Compared with standard ECI, this approach can improve intravariceal retention, reduce required glue volume, and potentially lower systemic embolization, particularly in high-flow fundal varices with shunt[42]. It may also facilitate hemostasis in actively bleeding lesions by providing immediate mechanical tamponade. Another recent study reported that Clip-ECI had an obliteration rate comparable to that of EUS-guided therapy, with shorter procedural duration and lower cost[43]. Consequently, while Clip-ECI is a useful, accessible enhancement of conventional ECI, especially where EUS expertise is limited, EUS-guided therapy is generally preferred for anatomically complex, high-flow, or shunt-dominant GVs, and for achieving more durable hemodynamic control. However, because endoscopic therapies do not eliminate the underlying shunt, they are often best viewed as a local or bridging therapy, or as part of a hybrid strategy with shunt-directed radiologic interventions when durable hemodynamic control is needed.

Pressure-dominant GVs

In contrast, patients with Kiyosue Type D anatomy lack a dominant shunt and exhibit diffuse portal hypertension. For such patients, TIPS is the preferred intervention[4,44]. TIPS creates a low-resistance channel between the portal and hepatic veins, effectively reducing portal pressure and addressing both GVs and their associated complications[44,45]. In selected cases, TIPS may be combined with antegrade transvenous obliteration or RTO to enhance control of fundal varices[46,47]. However, in the presence of large spontaneous shunts, blood flow may preferentially bypass the TIPS, reducing its efficacy. Additionally, TIPS is associated with HE and risk of liver decompensation, particularly in patients with advanced cirrhosis. Therefore, careful evaluation of hepatic reserve and comorbidities is critical before proceeding with TIPS. EUS-guided intravariceal embolization of coil and glue is still a feasible option in patients at risk of adverse events with TIPS. Rarely, RTO may still be attempted via a prominent alternative outflow vein (e.g., the inferior phrenic or pericardiophrenic vein) if technically accessible[48].

Complex hemodynamics

A subset of patients presents with complex or mixed hemodynamic patterns, corresponding to Kiyosue Type C, characterized by multiple afferent and efferent pathways without a single dominant shunt. These cases pose significant technical challenges for radiologic interventions. In such scenarios, EUS-guided therapy provides a targeted alternative to radiological intervention. EUS-guided coil embolization and glue injection allow direct obliteration of the variceal lumen or feeding vessels[14]. By addressing the inflow component, EUS-guided therapy minimizes the need for global hemodynamic alteration and is particularly valuable in patients with poor liver reserve or high procedural risk. EUS-guided interventions also serve as salvage therapy in cases of failed ECI, RTO or TIPS, and their ability to confirm flow obliteration using Doppler imaging enhances procedural precision and outcomes. Patients with complex or dual-drainage varices (e.g., GOV3) often exhibit mixed hemodynamic patterns, necessitating individualized or hybrid therapeutic strategies[11]. Recognition of such variants is important, as single-modality approaches may be insufficient when parallel outflow pathways are present.

Left-sided/sinistral portal hypertension

In left-sided (sinistral) portal hypertension, management is directed at correcting splenic venous outflow obstruction rather than reducing global portal pressure[7]. Splenectomy remains the definitive therapy, eliminating the source of increased inflow. In a previous meta-analysis of therapies in sinistral portal hypertension, splenectomy was found to be superior compared to non-surgical therapies in terms of the reduction of the incidence of variceal bleeding[49]. Partial splenic artery embolization offers a less invasive alternative by reducing splenic inflow and variceal pressure, particularly in patients who are poor surgical candidates[7,50,51]. Percutaneous transluminal angioplasty, with or without stenting of the splenic vein, may be considered in selected cases with focal venous stenosis to restore physiological flow, although its applicability is limited[7]. EUS-guided therapies are effective for acute bleeding control but do not address the underlying pathology and therefore serve as bridging or adjunctive measures[52].

Table 3 summarizes the preferred and alternative/adjunctive therapies for the management of GVs based on the hemodynamics. Table 4 summarizes the meta-analyses comparing the various therapeutic modalities for the management of GVs[6,35,53-58].

Table 3 Hemodynamics of gastric varices based on cross-sectional imaging classification, along with their preferred and alternative interventions.

Type		Venous anatomy	Venographic/hemodynamic findings ⇨ key issue	Preferred intervention	Alternative intervention	Key technical notes during RTO
Efferent venous pattern
A	Single draining shunt (usually gastrorenal; rarely gastrocaval); no collaterals		Complete opacification of the varix on balloon-occluded venography	Standard BRTO	PARTO/CARTO/EUS-guided	Deep microcatheter placement ⇨ inject sclerosant till minimal afferent filling
B1/B2	Single shunt + small/multiple collateral veins		Incomplete opacification due to preferential collateral flow ⇨ Collateral “leak”	Modify the flow to isolate the varix by coil/plug embolization	EUS-guided	Balloon beyond collaterals; deep positioning ⇨ flow-directed embolization ⇨ microcatheter beyond collaterals ⇨ sclerosant
B3	Single shunt + large collateral veins		Poor opacification until collaterals are controlled ⇨ Significant collateral drainage	Pre-embolize collaterals ⇨ BRTO	EUS-guided	Selective catheterization ⇨ coil embolization ⇨ repeat venography
C1	Two shunts (gastrorenal + small gastrocaval)		Partial opacification due to the second shunt ⇨ Additional minor outflow	Eliminate second shunt ⇨ treat as B3	BATO/PTO/trans-TIPS access/EUS-guided	Coil embolization of the gastrocaval shunt via microcatheter
C2	Two large shunts (gastrorenal + gastrocaval)		Incomplete opacification due to dual drainage ⇨ Major dual outflow	EUS-guided/BATO/PTO/trans-TIPS access	Combined BRTO approaches for dual shunt control	Dual balloon occlusion (gastrorenal + IJV gastrocaval) ⇨ then sclerosant
D	No shunt		No large draining shunt ⇨ No retrograde access possible	TIPS (± embolization)	BATO/PTO/EUS-guided	-
Afferent venous pattern
Type 1	Single afferent vein supplying gastric varix		Sclerosant stagnates with minimal reflux into the afferent vein ⇨ Risk of excessive reflux into the portal vein if over-injected	Standard BRTO	EUS-guided	Endpoint = minimal afferent vein filling; avoid forceful injection
Type 2	Two afferent veins (left + posterior gastric veins)		Reflux preferentially into the lower-pressure vein ⇨ Partial obliteration due to persistent higher-pressure inflow	Staged BRTO is often required after coil/glue embolization of feeders	BRTO + TIPS if associated esophageal varices present/EUS-guided	Endpoint = reflux in one vein; repeat the session for the remaining varix
Type 3	Separate afferent vein drains directly into the shunt (no variceal communication)		Sclerosant preferentially flows into this vein ⇨ portal reflux ⇨ Ineffective variceal filling + risk of portal spill	BRTO + Selective embolization of the aberrant afferent vein	EUS-guided	Deep microcatheter positioning; embolize extra afferent (transhepatic/transjugular) if needed

EUS: Endoscopic ultrasound; BRTO: Balloon-occluded retrograde transvenous obliteration; PARTO: Plug-assisted retrograde transvenous obliteration; CARTO: Coil-assisted retrograde transvenous obliteration; RTO: Retrograde transvenous obliteration; BATO: Balloon-occluded antegrade transvenous obliteration; PTO: Percutaneous transhepatic obliteration; TIPS: Transjugular intrahepatic portosystemic shunt.

Table 4 Summary of meta-analyses comparing the various therapeutic modalities in the management of gastric varices.

Ref.	Study type	Comparison	Sample size	Key efficacy outcomes	Safety outcomes	Key conclusion
Yu et al[55]	Meta-analysis	BRTO vs TIPS	435 (5 studies)	Similar success (91% vs 89%)	Rebleeding lower with BRTO (10.6% vs 18.7%); HE lower (0% vs 23%)	BRTO preferred; TIPS increases HE
Paleti et al[53]	Meta-analysis	BRTO vs TIPS	676 (7 studies)	No difference in technical success/hemostasis	Rebleeding ⇩ with BRTO (OR 0.30); HE ⇩ (OR 0.06); mortality ⇩ (OR 0.43)	BRTO superior to TIPS for rebleeding, HE, and survival
Wang et al[54]	Meta-analysis	BRTO vs TIPS	9 studies	Similar immediate hemostasis	Rebleeding higher with TIPS (RR 2.61); HE ⇧ with TIPS (RR 16.1)	BRTO provides better survival and lower rebleeding
Osman et al[56]	Network meta-analysis (RCTs)	BRTO vs TIPS vs ECI vs NSBB	647 (9 studies)	BRTO lowest rebleeding (RR 0.04 vs NSBB)	β-blockers worst outcomes	BRTO most effective for secondary prophylaxis
Giri et al[6]	Network meta-analysis	Multiple (radiological and endoscopic)	2783 (34 studies)	BRTO highest obliteration (SUCRA 95.1); EUS-C+G second (80.9)	Adverse events highest with TIPS; lowest with thrombin	BRTO and EUS superior to ECI; top-ranked therapies
Florencio de Mesquita et al[35]	Meta-analysis	EUS (coil + glue) vs ECI	445 (6 studies)	Rebleeding ⇩ with EUS (OR 0.22); reintervention ⇩ (OR 0.29) with EUS	No difference in embolism/mortality	EUS superior to ECI with similar safety
Mohammadpour et al[57]	Meta-analysis	EUS-coil + glue vs endoscopic modalities	579 (9 studies)	Higher obliteration (RR 1.18); rebleeding ⇩ (RR 0.36)	Adverse events ⇩ (RR 0.55)	EUS combination therapy superior to endoscopic options
Biswas et al[58]	Individual patient meta-analysis	BRTO/TIPS vs ECI	1240 (15 studies)	Rebleeding ⇩ with BRTO (sHR 0.15), TIPS (sHR 0.49)	Ascites ⇧ with BRTO; HE ⇧ with TIPS	Endovascular therapies superior to ECI

EUS: Endoscopic ultrasound; ECI: Endoscopic cyanoacrylate injection; BRTO: Balloon-occluded retrograde transvenous obliteration; TIPS: Transjugular intrahepatic portosystemic shunt; NSBB: Non-selective beta-blockers; RR: Risk ratio; OR: Odds ratio; HE: Hepatic encephalopathy.

Clinical modifiers in the management of GVs

While anatomical and hemodynamic classification systems such as Kiyosue and Saad-Caldwell provide the structural framework for selecting therapy in GVs, clinical modifiers play a decisive role in refining and, at times, overriding these choices. These modifiers reflect the patient’s global disease burden, hepatic reserve, and competing complications of portal hypertension, ensuring that the selected intervention is not only effective but also safe.

Refractory ascites or hepatic hydrothorax: One of the most important modifiers is the presence of refractory ascites or hepatic hydrothorax, which indicates significant portal hypertension and systemic congestion. In such patients, even if a spontaneous shunt is present, TIPS is often preferred because it provides global portal decompression, addressing both varices and fluid-related complications[45,46]. Performing shunt-occlusive procedures, such as BRTO, in this setting may worsen ascites by increasing portal pressure, thereby exacerbating the underlying condition[28].

HE: HE is a key modifier that favors shunt-occlusive strategies such as BRTO, PARTO, or CARTO. In patients with large spontaneous portosystemic shunts, encephalopathy results from diversion of ammonia-rich blood away from hepatic metabolism, and occluding the shunt can significantly improve, or even resolve, HE[30]. A recent case series reported the role of EUS-guided transgastric shunt obliteration using coil and glue embolization[59]. Conversely, TIPS may worsen encephalopathy by further increasing systemic shunting and, therefore, should be avoided or used cautiously in such patients.

Hepatic functional reserve: Hepatic functional reserve, commonly assessed using the Child-Pugh or MELD score, is another critical determinant. Patients with advanced liver disease (Child-Pugh C) are at increased risk of post-procedural liver failure following TIPS due to the sudden alteration in portal hemodynamics[60]. In these patients, EUS-guided therapies or carefully selected RTO procedures are often preferred, as they achieve local or shunt-targeted control without significant systemic hemodynamic shifts[61,62].

Status of the portal venous system: Assessment of portal vein patency and overall portal hemodynamics is critical in selecting and safely performing therapies such as RTO and TIPS. In cases of complete PVT, spontaneous shunts (e.g., gastrorenal shunt) may serve as the primary outflow pathway for splanchnic circulation. Their occlusion can markedly increase mesenteric venous pressure, thereby increasing the risk of bowel ischemia and extensive thrombosis, and thus requires extreme caution and individualized decision-making[4,44]. In contrast, patients with chronic PVT and cavernous transformation often develop extensive collateral networks that may accommodate altered flow dynamics, allowing RTO to be performed more safely after detailed evaluation[63]. In partial thrombosis or a small but patent portal vein, shunt occlusion may enhance hepatopetal flow.

Cardiopulmonary comorbidities: Other important modifiers include cardiopulmonary comorbidities, which may contraindicate TIPS due to increased cardiac preload[64] and necessitate hybrid approaches combining endoscopic, EUS-guided, and radiologic therapies.

Figure 4 summarizes the approach to the management of GVs based on the anatomy and pathophysiology/hemodynamics, along with clinical modifiers.

Figure 4 Approach to the management of gastric fundal varices based on the hemodynamics, along with clinical modifiers. ¹Trans-transjugular intrahepatic portosystemic shunt (TIPS)/balloon-occluded antegrade transvenous obliteration/percutaneous transhepatic obliteration preferred over TIPS for Kiyosue C. GOV1: Gastroesophageal varices type 1; GOV2: Gastroesophageal varices type 2; IGV1: Isolated gastric varices type 1; IGV2: Isolated gastric varices type 2; EVL: Endoscopic variceal ligation; EUS: Endoscopic ultrasound; ECI: Endoscopic cyanoacrylate injection; BRTO: Balloon-occluded retrograde transvenous obliteration; PARTO: Plug-assisted retrograde transvenous obliteration; CARTO: Coil-assisted retrograde transvenous obliteration; RTO: Retrograde transvenous obliteration; BATO: Balloon-occluded antegrade transvenous obliteration; PTO: Percutaneous transhepatic obliteration; SAE: Splenic artery embolization; PTA: Percutaneous transluminal angioplasty; CT: Computed tomography; MRI: Magnetic resonance imaging; TIPS: Transjugular intrahepatic portosystemic shunt.

LIMITATIONS AND FUTURE DIRECTIONS

Despite substantial advances in understanding and managing GVs, several limitations remain that hinder the development of standardized treatment strategies. One of the major challenges is the limited availability of high-quality randomized controlled trials comparing different therapeutic modalities. Much of the current evidence is derived from retrospective studies, single-center cohorts, or observational analyses, which may introduce selection bias and limit generalizability[6]. In addition, heterogeneity in study design, patient populations, and outcome definitions makes direct comparison between endoscopic, EUS-guided, and radiologic therapies difficult. Another limitation is the lack of a universally accepted classification system that integrates anatomical, endoscopic, and hemodynamic characteristics of GVs. While systems such as the Sarin, Kiyosue, and Saad-Caldwell classifications provide valuable insights individually, none alone captures the full spectrum of pathophysiological features relevant to clinical decision-making. Consequently, treatment selection often relies on institutional expertise and multidisciplinary interpretation rather than standardized algorithms. Technical challenges also exist. Endoscopic glue injection carries the risk of systemic embolization, particularly in the presence of large spontaneous shunts. Similarly, procedures such as BRTO or TIPS require specialized expertise and may not be widely available in all centers. Furthermore, the optimal sequencing or combination of therapies, including EUS-guided interventions and radiologic procedures, remains unclear.

Future research should focus on prospective multicenter trials comparing mechanism-based treatment strategies for GVs. Advances in EUS-guided techniques, including targeted coil and glue embolization of feeding vessels, hold promise for improving safety and treatment durability. In addition, incorporating advanced imaging, Doppler flow quantification, and artificial intelligence–assisted analysis may enhance the identification of dominant inflow and outflow pathways. Ultimately, a precision medicine approach, tailored to the underlying vascular anatomy and portal hemodynamics, is likely to shape the future management of GVs.

CONCLUSION

The anatomy and hemodynamics of GVs differ substantially from those of EV, making a one-size-fits-all approach consistently inadequate. The evidence supports a mechanism-based framework in which variceal subtype, dominant hemodynamic driver, and individual clinical context together determine optimal therapy. Shunt-occlusive techniques such as BRTO and its variants remain the cornerstone for shunt-dominant physiology, while TIPS is reserved for diffuse portal hypertension and associated complications such as refractory ascites. EUS-guided coil and glue embolization has matured into a first-line option for complex anatomy and high-risk patients, and serves as a reliable salvage modality when other approaches fail. Clinical modifiers, including hepatic reserve, encephalopathy, portal vein patency, and cardiopulmonary status, are active determinants that can redirect the entire therapeutic pathway, not secondary considerations. Prospective multicenter trials comparing mechanism-stratified treatment arms, alongside standardized classification systems integrating endoscopic, hemodynamic, and anatomical data, remain the most pressing unmet need toward truly individualized management of GVs.