Endoscopic Classification of Gastric Varices

GVs differ in morphology, pathophysiology, natural history, and response to endoscopic treatment. The vascular anatomy of GVs is classified into 2 types: type 1 (localized type) consists of a single varicose vessel with almost the same diameter as the inflow/outflow vein, and type 2 (diffuse type) consists of multiple varicose vessels with complex vascular connections. Gastric varices may exist as extensions of esophageal varices as 2 types: gastroesophageal varices type 1 (GOV1) are found along the lesser curvature, and gastroesophageal varices type 2 (GOV2) are found at the cardia. Isolated GVs (IGVs) exist as 2 types: IGV1s are located in the fundus, and IGV2s are sporadic. These distinctions are important in predicting the frequency of bleeding and the response to treatment. IGVs have the highest flow rates, are larger in size, and have deeper feeding vessels, resulting in more severe bleeding episodes.

Endoscopic Treatment of Gastric Varices

Endoscopic therapy of variceal bleeding has become established as first-line therapy as recommended by the Baveno IV consensus and American Association for the Study of Liver Disease guidelines. Variceal ligation has performed well in the treatment of esophageal varices; however, results with GVs have not been favorable. Sclerosants have had less success in the treatment of GVs, because they are associated with a high incidence of complications, including gastric ulcerations and perforation, and recurrent bleeding rates of 37% to 89%.

Direct endoscopic cyanoacrylate (CYA) injection of bleeding GVs ( Fig. 1 ), first described by Soehendra and colleagues in 1986, is widely considered first-line endoscopic therapy. N-butyl-2-cyanoacrylate (Histoacryl) has been used in a number of sizable case series with hemostasis rates of greater than 90%, variceal obliteration rates of 70% to 90%, and rebleeding rates less than 30%. As secondary prophylaxis, cyanoacrylate injection has been shown to reduce rebleeding rates as compared with band ligation and propranolol. As primary prophylaxis, cyanoacrylate has been shown to reduce the risk of bleeding and mortality from GOV2 or IGV1 greater than 10 mm diameter as compared with propranolol alone.

Endoscopic injection of gastric fundal varices with CYA glue.

Cyanoacrylate Glues and Injection Technique

A variety of cyanoacrylate glue monomers are commercially available, differing in the length of their alkyl group. N-butyl-2-cyanoacrylate (NB2-CYA), with a 4-carbon alkyl group, has been most widely used for the treatment of GVs. The polymerization time of NB2-CYA is rapid and can result in premature solidification of the glue in the needle or entrapment of the needle within the varix. NB2-CYA is therefore diluted with Lipiodol (ratios from 1:1 to 1:1.6). The injection catheter needs to be primed with distilled water, and then 1 mL of the glue-lipiodol mixture is injected, followed by flushing with water (to clear a catheter dead space of approximately 1 mL) to deliver the entire glue contents from the catheter into the varix. The 2-octyl-cyanoacrylate (2O-CYA) has a longer polymerization time due to a longer carbon alkyl group (8) and therefore can be injected without dilution and more slowly. The injection catheter is primed (and flushed) with saline because there is no risk of premature solidification within the catheter. Rengstorff and Binmoeller reported on the use of 2O-CYA in 25 patients with GVs with similar hemostasis rates and a 4% rebleeding rate over 11 months.

In preparation for endoscopic injection of cyanoacrylate, the endoscope tip is coated with silicone oil, as well as flushed oil through the instrument channel to minimize the risk of glue adherence that can lead to endoscope damage. The injection needle catheter is primed with either sterile water for NB2-CYA injection or saline for 2O-CYA injection. Saline should not be used for NB2-CYA injection because it accelerates the polymerization time, which may lead to premature clogging of the needle. An initial injection with water or saline should be free flowing into the varix and not forming a submucosal injection. Glue is injected into the varix in aliquots of 0.5 to 1.0 mL. The injection time will vary depending on the choice of cyanoacrylate and amount of dilution; undiluted NB2-CYA must be rapidly injected over seconds, whereas undiluted 2O-CYA can be more slowly injected over a minute. After the glue has been injected, the glue in the catheter dead space is flushed out with sterile water or saline. Continuously flushing will keep the needle patent for a possible repeat injection. The varix is palpated with a blunt-tipped instrument to confirm “hardness” from glue obliteration; if the varix is still soft, then additional injections are performed.

Noncyanoacrylate Glues

Noncyanoacrylate sealants, including fibrin glue and thrombin, have been used to arrest variceal bleeding in small uncontrolled case series. Thrombin plus ethanolamine was found to be equivalent to ethanolamine alone in 1 randomized controlled trial. In 2 retrospective studies, thrombin reportedly achieved hemostasis in bleeding GVs in 75% to 94%. There have been no reported adverse events, specifically, no reports of distant embolization of thrombin. Bovine thrombin (with its putative risk of Creutzfeld-Jacob disease) has been replaced by human formulation, but still does not negate the potential risk of unknown infections (pooled from 4000–5000 plasma donors). Rebleeding rates of 12% to 30%, usually within 3 months, are still of concern, and obliteration of the feeding varix is achieved in only 6%.

Cyanoacrylate Glue Embolization

The most serious adverse events of CYA glue injection therapy is systemic embolization. A recent study by Romero-Castro and colleagues documented a high frequency of pulmonary embolization in 58% of patients treated with NB2-CYA diluted 1:1 with lipiodol. Fortunately, most of these patients were asymptomatic. Sepsis has been reported secondary to embolized glue acting as a septic focus. Embolization into the arterial circulation (via a patent foramen ovale or arteriovenous pulmonary shunt) can result in stroke and multiorgan infarction. Factors that may increase the embolization risk before the glue has hardened include overdilution of NB2-CYA with lipiodol, excessively rapid injection, injection of too large a volume of glue in a single injection, and IGV1 that has high blood-flow rates. Other adverse events described were visceral fistulization, which may occur after accidental paravariceal injection. Entrapment of the needle in the varix by glue and damage to the scope also have been reported.

Endoscopic Ultrasound–Guided Cyanoacrylate Injection

Delivery of CYA under endoscopic ultrasound (EUS) guidance has the advantage of enabling precise delivery of glue into the varix lumen. EUS also enables assessment with Doppler to confirm vessel obliteration after treatment. This may have prognostic significance, as rebleeding risk after CYA injection has been linked to residual patency of treated varices. Furthermore, treatment can be performed without dependency on direct varix visualization; even in the presence of retained food or blood that may obstruct the endoscopic view, the varix lumen can be accurately targeted for glue injection.

EUS can identify the main feeding vein system, which derives from the left gastric vein trunk, the posterior gastric vein, short gastric vein, or outflowing venous system with gastrorenal shunts. Romero-Castro and colleagues described a small case series targeting the perforating “feeder vessel,” rather than the varix lumen proper, under EUS guidance. Targeting the perforating vessel conceptually may minimize the amount of CYA needed to achieve obliteration of GV and thereby reduce the risk of embolization. The glue-lipiodol mixture enabled fluoroscopic visualization of the injected vessel and confirmation that the feeder vessel had been accurately targeted. No rebleeding or complications were observed. The limitation of this approach is that identification of the perforating vessel with EUS can be difficult and time consuming. Importantly, as the perforating vessel may be afferent or efferent, contrast medium must be injected before treatment to determine directional flow relative to the varix.

Endoscopic Ultrasound-Guided Coiling

Two small cases series have described deployment of commercially available stainless steel coils. Levy and colleagues used a 22-gauge needle loaded with a “microcoil.” The stylet was used to advance the constrained coil to the tip of the needle. Once the needle was inserted into the largest (1.4 cm) varix, the stylet was further advanced to deliver the coil. Two additional coils were placed into separate varices. Although rebleeding occurred and repeat EUS therapy performed, the previously treated varices were shown to be thrombosed. Two additional coils were placed into untreated varices. Romero-Castro and colleagues used a 19-gauge needle to deliver 0.035-inch coils of 50 to 150-mm length (coil diameter of 8–15 mm after deployment). In one patient with a large gastrorenal shunt, the investigators failed to achieve obliteration of GVs despite deployment of 13 coils. A further 9 coils were deployed at a subsequent treatment session. The investigators did not comment on cost, but this becomes a consideration when large numbers of coils are required to achieve obliteration. Romero-Castro and colleagues recently reported on a retrospective 4-year cohort study, comparing EUS-guided therapy using CYA to coil embolization in 30 patients (23% never bled), with IGV1 in 15, GOV2 in 14, and GOV1 in 1 patient, respectively. Two-thirds had CYA injection, and one-third received coil insertion targeting perforating veins in 29 of 30 patients. The overall obliteration rate was 97%. A higher number of treatment sessions was required to achieve complete obliteration in the CYA versus coil group ( P = NS), but a single session achieved complete obliteration in 82% of the coiled group. Adverse events were significantly higher in the CYA (58%) versus the coil (9%) group ( P <.01), predominantly related to radiological documentation of 9 asymptomatic pulmonary emboli and 2 symptomatic patients with chest pain and fever, respectively.

Endoscopic Ultrasound-Guided Coiling and Cyanoacrylate Injection

In an ex vivo study, we deployed a coil in a container of heparinized blood, followed by injection of 1 mL CYA glue. The glue immediately adhered to the synthetic fibers on the coil and both coil and adherent glue were removed in one piece from the container ( Fig. 2 ). Outside the container, the glue firmly adhered to the coil and no residual glue was identified in the container. The deployment of a coil before CYA injection may serve several functions: (1) the coil itself may contribute to varix obliteration and hemostasis, (2) the coil concentrates the glue at the site of coil deployment, and (3) the coil may prevent glue embolization.

Ex vivo demonstration of CYA glue adherence to the synthetic fibers of a coil, removed in one piece after coil and glue were deposited in a container filled with heparinized blood.

We reported our first use of combined coil and CYA injection as “rescue” treatment after standard endoscopy-guided CYA treatment failed in a patient with massive gastric fundal variceal (GFV) bleeding. We injected 2O-CYA after deployment of a single coil in a series of 30 patients with large gastric fundal varices. The procedure ( Box 1 ) was successful in all patients, with immediate hemostasis achieved for active bleeding ( Fig. 3 ). The average volume of CYA injected was 1.4 mL per patient after coil deployment. Of note, this was 1 mL less than the average amount injected per patient in our previous study using the same CYA injected alone. There was no damage to the echoendoscope related to glue injections and no procedure-related complications. Of 24 patients who underwent follow-up endoscopy, 23 (96%) had complete GFV obliteration after a single treatment session, with no intravariceal flow on EUS color Doppler imaging. Recurrent bleeding from GFV developed in 1 patient at 21 days. This patient underwent a second successful treatment with EUS-guided coil and CYA. No patient required surgical or percutaneous shunt procedures.

Obliteration of gastric fundal varices by using coils and 2O-CYA. ( A ) Large, type 1 isolated gastric varices conglomerate. ( B ) EUS showing deployment of coil into varix. ( C ) EUS showing echogenic glue-filling varix. ( D ) Appearance at 3 months.

Transesophageal Injection

The gastric fundus is well visualized on EUS with the transducer positioned in the distal esophagus ( Fig. 4 ). We therefore elected to treat GFV from the esophagus with the echoendoscope in an orthograde position. Apart from enabling EUS-guided access to GFV, this transesophageal approach is not hindered by gastric contents, such as blood and food, which tend to accumulate in the fundus. There is also no disruption of the gastric mucosa overlying the varix, which is usually thinned and at high risk of “back-bleeding” after varix puncture. The transesophageal approach also allows visualization of the diaphragmatic crus muscle, which is typically “sandwiched” between the esophageal and gastric fundic walls (see Fig. 4 ). When visualized, we intentionally included the crus muscle in our path of access to GFV (“transcrural” puncture), hypothesizing that the crus muscle, a thick fibromuscular bundle approximately 1 cm in thickness, acts as a stabilizing “backboard” to GFV.

Anatomic diagram showing transesophageal-transcrural approach to gastric fundal varices.

Radiologic Therapy of Gastric Varices

Radiological intervention is used as a rescue strategy when endoscopic treatment fails or as primary intervention when endoscopic expertise is not available.

Forward-View Echoendoscope

A limitation of the conventional curved linear array (CLA) echoendoscope is the oblique-viewing optics, which can make it difficult to endoscopically visualize gastric fundal varices. Maneuvering the echoendoscope into retroflexion to visualize the fundus is challenging. The recent availability of the forward-view (FV) CLA echoendoscope addresses this limitation. In addition, the FV-CLA echoendoscope has several conceptual advantages for EUS-guided treatment: (1) more perpendicular needle orientation to the target lesion, (2) uniaxis instrumentation imparting an increased forward transfer of force to the tip of the needle, and (3) accessory water jet channel for water filling and irrigation. The tip of FV-CLA can be angled to provide nearly perpendicular needle access to the fundus from the distal esophagus. These advantages have been previously described in other clinical applications of the FV-CLA echoendoscope.