Approximately 50% of patients with liver cirrhosis have gastroesophageal varices (GOVs).¹ Gastric varices (GVs) are less common than esophageal varices (EVs); however, bleeding from GVs tends to have more severe consequences, with higher mortality and rebleeding rates than those observed for EVs.^2,3 Plug-assisted retrograde transvenous obliteration (PARTO) has been validated as a safe and effective treatment for GVs.^3,4 This technique is characterized by blocking the gastrorenal shunt (GRS) with an Amplatzer Vascular Plug (AVP) followed by embolization of the varices with gelatin sponge particles (GSPs). Technically, the AVP is recommended to be oversized by at least 20% compared to the GRS diameter.³ However, the maximal diameter of the commercially available plug is 22 mm, making the procedure unfeasible for GRSs ≥ 18 mm. Coil-assisted retrograde transvenous obliteration (CARTO) has been proposed for giant GRSs, with high technical and clinical success rates.^5,6 However, CARTO requires the deployment of multiple coils, which is much more expensive and time-consuming than PARTO. We report two challenging cases of giant GRSs, which were successfully occluded using 22-mm AVPs. According to our institutional policy, institutional review board approval is waived for case reports.

Case 1

A-58-year-old man presented with hematemesis and melena. His medical history included heavy alcoholism, decompensated cirrhosis, and multiple episodes of EV bleeding treated by endoscopic variceal ligation (EVL). On admission, he was alert and his vital signs were within normal limits. A complete blood count (CBC) showed a red blood cell (RBC) count of 2.52 × 10¹² cells/L, a hemoglobin (HGB) level of 61 g/L, and hematocrit (HCT) of 21%. He was administered a 2-mg intravenous bolus of terlipressin (Glypressin®; Ferring) followed by 1 mg every 4 hours. Upper gastrointestinal (GI) endoscopy revealed a huge fundic GV (GOV 2). A tortuous GRS with transverse diameters of 24 to 30 mm (Fig. 1) was seen on abdominal contrast-enhanced computed tomography (CECT). The shunt had multiple segments separated by waists. The transverse diameter of the proximal waist was 14 mm. After a multidisciplinary discussion, PARTO was deemed to be the optimal treatment.

Figure 1. Contrast-enhanced computed tomography portal venous phase. (A) On axial view, extensive varices (GOV 2) were seen in the gastric fundus (arrowheads). A giant, tortuous gastrorenal shunt (arrows) was appreciated on coronal (B) and sagittal views (C). The transverse diameter was measured as 24–30 mm. The shunt had multiple aneurysmal segments separated by waists. The diameter of the proximal waist was 14 mm. GOV, gastroesophageal varix.

The right common femoral vein (CFV) was accessed using a micro-puncture set (Cook Medical) under ultrasound guidance. A left renal venogram was obtained to confirm a GRS. The GRS was then cannulated and an Amplatz stiff wire (Boston Scientific) was exchanged. A 10-Fr long sheath was finally secured at the mid-portion of the GRS. Retrograde venography confirmed a giant, tortuous GRS with multiple aneurysmal segments and waists without collaterals, corresponding to grade 1 in the Hirota classification. A 2.7-Fr microcatheter (Progreat) was advanced into the GRS. Under fluoroscopic control, a 22-mm AVP II (St. Jude Medical) was introduced, then the long sheath was slowly withdrawn while the AVP was simultaneously pulled down until it was secured between the 2 dilated segments of the GRS and the middle lobe of the AVP was anchored at the waist. After 10 minutes, a venogram from the microcatheter confirmed total occlusion of the GRS, and the GV was embolized with GSPs (Spongostan; Ferrosan) in standard PARTO. The completion venogram showed no flow to the GRS while the renal venous flow to the inferior vena cava (IVC) was maintained. The plug was eventually detached (Fig. 2).

Figure 2. Plug-assisted retrograde transvenous obliteration. (A) A venogram obtained from a 5-Fr catheter revealed a giant high-flow gastrorenal shunt (GRS) (arrow) with two aneurysmal segments separated by a waist. (B) A 10-Fr long sheath (arrowheads) was secured at the upper dilated segment. (C) A 22-mm Amplatzer Vascular Plug (AVP) was positioned within the GRS so that the middle lobe was anchored at the waist. (D) The final venogram through the sheath confirmed a complete occlusion of the GRS while leaving the renal venous outflow patent.

No post-procedural complications were documented except for a mild epigastric discomfort. Endoscopic ultrasound (EUS) demonstrated complete thrombus formation within the GV. The patient was discharged uneventfully after 7 days. No recurrent GI bleeding was recorded during 24 months.

Case 2

A 47-year-old man presented with hematemesis and melena. He had a history of heavy alcohol consumption, cirrhosis, and several episodes of EVL due to recurrent EV bleeding. On admission, his hemodynamic status was relatively stable. A CBC showed the following findings: RBC count, 2.44 × 10¹² cells/L; HGB level, 80 g/L; and HCT, 23.7%. A 2-mg bolus of terlipressin was administered intravenously. An extensive GOV 2 was seen on upper GI endoscopy. On abdominal CECT, multiple gastro-phrenic collaterals to the IVC and a tortuous aneurysmal GRS measuring 23–32 mm in diameter were also noted. Multiple waists were seen within the shunt and the diameter of the proximal waist was 16 mm. Elective PARTO was indicated.

Similarly, the right CFV was accessed under ultrasound guidance. Retrograde venography revealed a giant, tortuous, aneurysmal high-flow GRS with multiple segments and waists (Fig. 3). Collaterals from gastrophrenic veins, subphrenic veins, and hemiazygos veins (Hirota grade 2) were super-selected with a 2.7-Fr microcatheter (Progreat) and subsequently embolized with microcoils. Next, a 22-mm AVP II was deployed using the same manner, followed by GSP embolization of the GV. Additional coils were used to occlude the proximal recess between the plug and the GRS wall and to stabilize the plug. The completion venogram showed total occlusion of the GRS and preservation of the renal venous outflow. The plug was finally released (Fig. 4).

Figure 3. Preprocedural contrast-enhanced computed tomography. (A) Gastric varices (GOV 2) were seen at the fundus. (B, C) Coronal and sagittal reformatted images showed a giant, aneurysmal gastrorenal shunt (GRS) (arrow) with transverse diameter of 23–32 mm. The GRS was separated by multiple waists and the diameter of the proximal waist was 16 mm. GOV, gastroesophageal varix.

Figure 4. Plug-assisted retrograde transvenous obliteration. (A) A 10-Fr long sheath was advanced into the gastrorenal shunt (GRS). A venogram demonstrated a giant GRS. (B, C) Collaterals from the inferior phrenic vein were super-selected and embolized with microcoils. (D) A 22-mm Amplatzer Vascular Plug (AVP) was deployed distally within the GRS. The AVP was pulled down and positioned so that the middle lobe was sealed at the waist. (E) A follow-up venogram showed residual flow through a recess between the AVP and the GRS wall. After thrombus formation, the varices were embolized with large gelatin sponge particles. (F) The completion venogram confirmed occlusion of the GRS and patency of the left renal venous return. Noted that additional coils were used to occlude the most proximal segment and to stabilize the AVP.

No intra- or post-procedural complications were observed except for fever, abdominal discomfort, and an aggravation of a pre-existing EV from grade I to grade II; however, no further treatment was required. EUS on day 5 showed complete thrombosis of the GV. The patient was discharged after 14 days. No recurrent bleeding was documented during a 23-month follow-up.

The informed consent was waived.

Variceal bleeding is a major complication of portal hypertension and a leading cause of death in patients with cirrhosis.⁴ Compared to EVs, GVs have lower incidence but higher rates of mortality and rebleeding.^7,8 GVs are classified according to Sarin’s classification based on their location and relation to EVs.⁸ Approximately 80%–85% of GVs drain into the left renal vein (LRV) via a GRS.⁹

PARTO is a novel technique modified from balloon-occluded retrograde transvenous obliteration (BRTO), which uses an AVP to occlude the GRS instead of a balloon catheter, therefore overcoming the classic disadvantages of BRTO.^3,10 A large multicenter prospective study found that 98.6% of patients achieved complete GV thrombosis after 1 week. During follow-up, no recurrent variceal bleeding or hepatic encephalopathy developed. The Child-Pugh score improved in 40% of patients.¹¹ The consistent drawback of PARTO is aggravation of portal hypertension due to occlusion of the portosystemic shunt, which consequently increases the risk of EV bleeding.

Technically, an AVP must be at least 20% oversized compared to the maximal diameter of the GRS to prevent plug migration and GSP reflux.^11,12 However, the largest size of a commercially available AVP is 22 mm, making the procedure technically infeasible for GRS ≥ 18 mm in diameter. CARTO and CARTO-II have been recognized as effective alternatives for giant GRS up to 25–30 mm in diameter.^2,5,13 However, CARTO requires the use of multiple coils, making it more time-consuming and costly than PARTO, which uses a single AVP.² Additionally, AVP has a lower risk of migration than a coil, even in high-flow situations or short landing zones.¹⁴

Morphologically, a giant, tortuous GRS usually appears as multiple aneurysmal segments separated by waists. Mukund et al¹⁵ found several anatomical factors that influence PARTO outcomes. Specifically, factors of the GRS itself, such as the diameter of the shunt, the entry point, and the waist, were not associated with technical failure of PARTO. Conversely, factors related to the LRV, such as aneurysmal dilatation, extreme acute/obtuse angulation, and an extreme antero-posterior orientation of the shunt in relation to the LRV significantly affected outcomes.¹⁵

Within the AVP family, AVP II is currently the largest (22 mm) and can be delivered through a 7-Fr sheath. It is made out of a densely braided multilayer nitinol mesh with three components, generating six barrier planes.¹⁴ Based on this aspect of its structure, we believe that the waist between the two dilated segments of the GRS itself can serve as an ideal occlusion point because it can firmly seal the middle lobe of an AVP II. In the present cases, the diameters of the proximal waists, where we intended to deploy the AVP’s middle lobe, were 14 mm and 16 mm, corresponding to 57% and 37.5% oversizing of a 22-mm AVP. We believe these diameters are suitable to safely anchor the 22-mm plugs, to completely occlude the GRS, and to prevent the plugs from migration.

In conclusion, Off-label PARTO can be considered in selected patients with giant GRSs ≥18 mm in diameter. A detailed preprocedural evaluation of the GRS morphology on CECT is crucial to ensure a safe and successful procedure.

This study was supported by Hue University under the Core Research Program, Grant No. NCM.DHH.2020.09.