Endoscopic ultrasound-guided therapies in the treatment of gastric varices: An in-depth examination of associated adverse events

Abstract

The Baveno VII consensus, released in 2023, recommends that the endoscopic treatment of choice for managing bleeding gastric varices (GV) is endoscopic ultrasound (EUS)-guided treatment, specifically EUS-guided cyanoacrylate (CYA) glue injection. This approach has been endorsed due to its efficacy in controlling bleeding while reducing rebleeding rates, compared to other endoscopic techniques. Despite its efficacy, CYA injection for GV has been linked to rare but serious adverse events, such as glue embolization leading to pulmonary embolism, infection/bacteremia, splenic infarction, intra-procedural and post-procedural complications. The technique for endoscopic obliteration of GV was subsequently refined using EUS guidance, aiming to more accurately direct the injection into the varix, reduce the volume of injected glue, target the perforating vein when possible, and confirm obliteration of GV via Doppler examination.

Key Words: Gastric varices; Endoscopic ultrasound; Cyanoacrylate injection; Coil; Adverse events

Core Tip: Gastric varices (GV), contribute to 10%-30% of upper gastrointestinal bleeding in patients with cirrhosis. GV account for 20% of variceal-related hemorrhage and are associated with higher risk of uncontrolled bleeding, transfusion needs, and mortality compared to esophageal varices. Recent studies have explored various endoscopic ultrasound (EUS)-guided therapies, such as cyanoacrylate injection and coil embolization. Albeit safe, EUS-therapies may expose patients to serious adverse events such as embolism, infection and bleeding. EUS-guided therapies often demonstrate superior efficacy in variceal obliteration and lower rebleeding rates compared to conventional methods. They come, however, with higher costs and technical challenges, such as precise needle placement and choosing between transgastric and transesophageal approaches.

INTRODUCTION

Gastric varices (GV) are collateral vessels that arise from portal hypertension (PH), which in turn leads to the reversal of blood flow from the splenic vein into the left renal vein[1]. Gastroesophageal variceal bleeding is a significant complication of PH, particularly in patients with cirrhosis, and accounts for 10% to 30% of all upper gastrointestinal (GI) tract bleeding cases[2]. In patients with cirrhosis, gastroesophageal varices (GEV) develop and grow at 7% per year[3]. GV make up roughly 20% of all variceal bleeding episodes. They are linked to a higher risk of uncontrolled hemorrhage, greater transfusion needs, and increased rates of rebleeding and mortality compared to esophageal varices[4,5]. According to the Sarin classification GV are classified into GEV type 1 (GOV1), which extend over the cardia and lesser curvature; GEV type 2 (GOV2), which extend from the esophagus into the gastric fundus; isolated GV type 1 (IGV1), located at the gastric fundus; and isolated GV type 2 (IGV2), found elsewhere in the stomach. GOV1 accounts for 75% of all GV, while GOV2, IGV2, and IGV1 constitute 21%, 4%, and less than 2%, respectively[4-6].

In recent years, various endoscopic ultrasound (EUS)-guided techniques have been developed for both primary and secondary prophylaxis of GV. These methods include EUS-guided N-butyl-cyanoacrylate (CYA) or glue therapy (EUS-glue), EUS-guided coil embolization (EUS-coil), EUS-guided sclerotherapy, EUS-guided thrombin injection, and a combination of EUS-coil and glue injection (EUS-coil/glue). Although several studies have demonstrated the safety and efficacy of these approaches[7-9], a comprehensive understanding of their overall efficacy is still lacking.

EUS ASSESSMENT OF THE PORTAL VENOUS SYSTEM AND ITS COLLATERALS IN PH

Based on a detailed review by Dragomir et al[10], EUS seems to provide a detailed visualization of the vascular system involved in PH, allowing the assessment of vascular diameter, collateral formation, and complications including portal vein thrombosis. In cirrhosis, collateral veins, such as GEV, develop to divert blood from the portal to the systemic circulation. The left gastric vein (LGV) plays a key role in “feeding” these varices and drains into the azygos vein, which connects to the superior vena cava. Approximately 50% of cirrhotic patients develop varices, with 15%-20% experiencing bleeding within 1-3 years. Portal vein measurements typically show a diameter of up to 13 mm, an increase that can be determined by applying EUS visualization from different gastric regions, aiding interventions like pressure measurement, portal liquid biopsy, and thrombus biopsy. Similarly, the splenic vein, normally less than 9 mm in diameter, is best seen at the gastric body and fundus. While splenic vein enlargement can indicate PH, its diameter often remains normal in early-stage cirrhosis. EUS use within the duodenum is also effective in visualizing the superior mesenteric vein, though its role in predicting PH remains unclear.

The LGV, which originates from the lesser curvature of the stomach, is important for predicting variceal bleeding. Studies show that LGV flow velocity correlates with bleeding risk, and EUS-guided treatments, such as CYA injection and coil deployment, show high success rates in treating varices. Following such interventions, a reduction in LGV diameter and flow velocity is observed, with faster recurrence in patients exhibiting higher flow velocities. Lastly, the azygos vein, involved in blood flow through gastroesophageal collateral vessels, can be visualized using color Doppler EUS. Its flow, however, does not consistently correlate with hepatic venous pressure gradient, limiting its predictive value in PH[10].

OVERALL ADVERSE EVENTS

A comprehensive analysis of adverse events associated with various endoscopic treatments for GV offers several insights into the safety profile of these approaches. Bick et al[11] found no significant difference in the overall rate of adverse events between the direct endoscopic injection (DEI) (7/40, 17.5%) and EUS-guided injection (13/64, 20.3%) groups (P = 0.723), with similar rates of moderate to severe adverse events (P = 0.361). CYA glue injection, however, can lead to localized e.g. ulcers and systemic manifestations such as cerebral stroke, splenic infarction, pulmonary embolism, and death[12-14]. In the study of Chandan et al[12] the overall pooled rate of adverse events was 11.9%, including 11 patients with post-procedure fever, 24 with chest and/or abdominal pain, and 16 with pulmonary/systemic embolism. Likewise, two separate studies reported low rates of transient adverse events with no difference between DEI and the EUS-guided technique[15,16]. In terms of efficacy, however, DEI was found to be less effective than its EUS counterpart with lower rates of variceal obliteration (63% vs 84%), higher recurrence (18% vs 9%), late rebleeding (16% vs 12%), and complication rates (50% vs 25%)[16]. Lôbo et al[17] observed early epigastric pain in 18.8% of patients undergoing EUS-coil/glue and 6.3% of EUS-glue, with late abdominal pain in 25% and 0% respectively, and no reported febrile episodes. Thiruvengadam and Sedarat[18] found EUS-CYA to have similar adverse effects to DEI-CYA, including abdominal pain (8%-15%), fever (8%-9%), transient bacteremia (2%-6%), and injection site ulcers (3%). Coil placement, with or without CYA, has been associated with adverse effects such as mucosal scarring (9%), minor GI bleeding (6%-50%), and major bleeding (10%)[19]. Khoury et al[20] noted major bleeding from the puncture site in 1 of 10 patients (10%). Rare events associated with endoscopic gastric variceal obliteration are coil migration and bleeding due to extensive mucosal damage[14,21].

SYSTEMIC AND PULMONARY EMBOLISM AFTER EUS-GUIDED CYA INJECTION WITH OR WITHOUT COIL PLACEMENT

Several studies have examined adverse events linked to EUS-guided treatments for GV, particularly CYA injection and coil embolization. In a retrospective cohort study by Bick et al[11] one patient out of 13 developed pulmonary embolism three weeks post-procedure. Robles-Medranda et al[19] found that EUS-guided CYA injection alone had a higher risk of pulmonary embolization than EUS-coil. Chandan et al[12] reported 16 cases of pulmonary/systemic embolism across 18 studies involving 604 patients. Kozieł et al[22] observed pulmonary embolism in 4 out of 16 patients (25%) in a study of EUS-guided CYA glue injection and coils. Bazarbashi et al[15] noted intraprocedural adverse events in 2 out of 106 patients, including one fatal pulmonary embolism. Lôbo et al[17] reported early complications in 8 out of 16 patients (50%) treated with coil and CYA, including pulmonary embolism in 4 (25%), and in 10 out of 16 patients (62.5%) treated with CYA alone, with 8 (50%) experiencing pulmonary embolism (P = 0.144), all asymptomatic. Thiruvengadam and Sedarat[18] indicated that DEI-CYA can cause distal embolization, with pulmonary embolism occurring in up to 5% of cases. Romero-Castro et al[23] found pulmonary embolism in 9 out of 19 patients (47%) in an EUS-CYA study. Bhat et al[24] observed pulmonary embolism in 1 out of 125 patients (0.8%) in an EUS-coil/CYA study. Mukkada et al[25] found that, among 30 patients who re-bled after CYA injection, one developed symptomatic pulmonary embolism, while nine in the CYA group had asymptomatic pulmonary emboli on CT scans, compared to none in the coil group. They concluded that the coil group had fewer adverse events, though rare cases of pulmonary embolization and coil migration to the liver were reported.

Damage to the endoscope and procedure-related complications

The use of CYA has been accompanied by instrument damage due to polymerization of CYA in the working channel, injection needle, and the scope tip on some occasions[13-15]. According to Guo et al[14], blockage of the injection catheter during glue injection may occur in 2.71% of cases. Other CYA instrument-related complications included glue adhesion to the endoscope, making it difficult to withdraw, and the sticking of the needle to the varix, reported in 1.43% of cases[14].

ADVERSE EVENTS RELATED TO INFECTION AND BACTEREMIA

Bick et al[11] reported a severe case where a patient developed splenic infarcts and required an 11-day hospitalization for Escherichia coli bacteremia, diagnosed two days post-injection. Jamwal et al[16] noted that three patients in the DEI group and two in the EUS-coil/glue group died from causes unrelated to the procedure, including pneumonia, spontaneous bacterial peritonitis, acute kidney injury, and hepatic encephalopathy, all occurring more than three months post-procedure. Thiruvengadam and Sedarat[18] found that EUS-CYA has a similar adverse effect profile to DEI-CYA, with abdominal pain in 8%-15% of patients, fever in 8%-9%, transient bacteremia in 2%-6%, and ulcers at the injection site in 3%. Mukkada et al[25] reported three deaths due to septicemia, upper GI bleeding, and liver failure, occurring 9 days, 120 days, and 143 days post-procedure, respectively.

BLEEDING-RELATED ADVERSE EVENTS

Complications related to bleeding remain a significant concern in EUS-guided treatments for GV, impacting both the frequency and severity of adverse outcomes. Mohan et al[7] reported that the pooled rate of early rebleeding was 7.0% (95%CI: 4.6-10.7, I² = 0) with EUS-guided treatment and 5% (95%CI: 3.3-7.4, I² = 72, P = 0.7) with DEI-glue treatment. The pooled rate of late rebleeding was 11.6% (95%CI: 8.8-15.1, I² = 22) with EUS-guided therapy and 17% (95%CI: 12.3-22.9, I² = 92, P = 0.1) with DEI-glue. The pooled rate of mortality due to gastric variceal bleeding was 7.7% (95%CI: 4.9-11.9, I² = 29). EUS-guided therapy showed superior obliteration rates (84% vs 63%, P = 0.02) and nearly reached superiority in reducing recurrence (9% vs 18%, P = 0.06) compared to DEI-glue therapy. Bick et al[11] found that GV rebleeding (8.8% vs 23.7%, P = 0.045) and non-GV-related GI bleeding (10.9% vs 27.5%, P = 0.030) were less frequent in the EUS-needle injection group compared to the DEI group. Rebleeding from GV occurred in 5/57 patients (8.8%) in the EUS-guided fine needle injection (FNI) group within 30 days of therapy, resulting in three deaths. EUS-FNI had significantly lower GV rebleeding rates overall (P = 0.045) and at 30 days (P = 0.041), as well as lower all-cause bleeding rates (P < 0.001). Robles-Medranda et al[19] reported higher reintervention and rebleeding rates in patients treated with EUS-guided coiling alone compared to combined therapy. Chandan et al[12] found that post-therapy bleeding was 4.9% (95%CI: 1.8%-12.4%, I² = 0) for primary prophylaxis, with no early rebleeding, and 18.1% (95%CI: 13.1%-24.3%, I² = 16) for secondary prophylaxis. Kozieł et al[22] reported a duodenal varix rupture during coil implantation managed with CYA injections. Bazarbashi et al[15] found recurrent bleeding in 15 patients (14.1%), with a mean occurrence of 33 days ± 37 days. Lôbo et al[17] noted thrombosis of the treated vessel and subsequent bleeding from adjacent varices. Mukkada et al[25] reported rebleeding within 365 days in six patients after EUS-guided coil application, while 51% of patients re-bled within one year of glue injection. Guo et al[14] described persistent bleeding in one patient managed with additional glue injections.

SPLENIC INFARCTION

Bick et al[11] found that 13 patients in the EUS-FNI group had adverse events, including two with splenic infarcts. The second patient with a moderate infarct was hospitalized for 5 days and managed conservatively.

MORTALITY RATES AND CAUSES IN EUS-GUIDED THERAPY FOR GV

Mohan et al[7] reported an all-cause mortality rate of 13.1% (95%CI: 8.3-20.2, I² = 68) for EUS-guided therapy, with 115 deaths across the cohorts, including 32 due to gastric variceal bleeding. The mortality rate due to gastric variceal bleeding was approximately 8%. Robles-Medranda et al[19] found similar overall mortality rates: (1) 30% in the combined treatment group; and (2) 26.7% in the coil-only group (hazard ratio = 0.95, 95%CI: 0.361-2.532; P = 0.90). In the combined group, five patients died from uncontrolled hemorrhage, with two having hepatocellular carcinoma, and all being Child-Pugh class C. The coil-only group had five deaths from uncontrolled hemorrhage, including one with hepatocellular carcinoma, and all were Child-Pugh class C. The median survival was 16.4 months (range 0.6-31.2 months) for combined treatment and 14.2 months (range 0.8-28.2 months) for coil embolization alone (P = 0.90). Bazarbashi et al[15] found a 23.6% overall mortality rate at a mean of 218 days, with most deaths due to non-GV-related causes. Jamwal et al[16] reported that out of 80 patients, five died from causes unrelated to bleeding or embolization (e.g., pneumonia and bacterial peritonitis). Lôbo et al[17] noted no deaths in the EUS-guided coil group but two deaths in the other group, one due to upper GI hemorrhage and another to sepsis of unknown origin. Mukkada et al[25] reported two deaths (4%) in 51 patients, one from liver failure and the other from an uncertain cause, with no significant difference in mortality between groups (Z = 1.1, P > 0.05).

COST COMPARISON BETWEEN EUS-GUIDED THERAPIES VS CONVENTIONAL ENDOSCOPIC THERAPY

The financial implications of EUS-guided therapies compared to conventional endoscopic methods must be considered. Romero-Castro et al[23] noted that 1 mL of histoacryl/lipiodol costs approximately United States $72.30, while a single coil costs United States $99.40, regardless of length. Kouanda et al[26] reported that EUS-guided coil and CYA injection total costs $1831, including $1557 for the facility and $274 for the physician, while inpatient hospitalization for gastric variceal bleeding costs about $11078. In India, Mukkada et al[25] highlighted that 1 mL of CYA costs Rs 1100, coils and microcoils each cost Rs 6500, and the Echotip needle costs Rs 22000, making the procedure relatively expensive.

CHALLENGES AND CONSIDERATIONS IN THE SETTING OF APPROPRIATE TRAINING

Despite their rather lucrative efficacy profile, EUS-guided techniques for obliteration of GV are also demanding. The procedures require significant technical expertise as improper techniques can lead to complications. In particular, a transgastric approach can be problematic due to acute angulation at the scope tip, making the deployment of coils and glue difficult. Additionally, the choice between transgastric and transesophageal techniques remains unclear, necessitating further research to determine the preferred method. The identification of the parietal gastric collateral is also challenging due to the presence of numerous collaterals around the splenic hilum and gastric fundus. This complexity increases the risk of complications, such as misidentification of the correct vein and incorrect injection into splenic veins or adjacent structures[16].

FUTURE AND EMERGING EUS APPLICATIONS IN PH

EUS is emerging as a valuable tool in the management of PH allowing for complicated and innovative interventions[10]. In the case presented by Zhang et al[27], GV entwined with arteries were successfully treated using EUS-coil/glue therapy. EUS-guided portal vein sampling is such another example whereby blood samples drawn from the portal vein are analyzed for circulating pancreatic tumor cells. This technique has also been used in a small cohort of PH patients, successfully collecting portal blood for metabolomic analysis, in the absence of adverse events. Additionally, EUS-guided selective embolization of the portal vein has been tested in animals, showing promise in promoting compensatory liver hypertrophy before liver resection. Another experimental technique is EUS-guided creation of portosystemic shunts using metal stents, a procedure currently lacking application in humans. Direct injection of glue and coils into spontaneous portosystemic shunts via EUS could also help reduce the risk of encephalopathy, signaling a promising future for EUS in other aspects of PH management[10].

Currently, an effort is at play to couple predictive models and artificial intelligence (AI) with EUS-diagnostic and interventional applications to optimize clinical decisions, patient selection, accuracy, operator-dependent bias, procedure safety, and outcomes[28-30]. Since AI can process vast amounts of data with high accuracy it may offer a more objective, efficient, and rapid assessment during EUS procedures. AI-assisted EUS-guided fine-needle aspiration/biopsy has been effective in precisely localizing lesions and avoiding important blood vessels during puncture, significantly reducing procedure-related risk[30]. Vice versa, AI-assisted EUS could aid in the visualization of vessels while reducing the risk of needle trauma on adjacent structures. A rather intriguing application of AI in this context is the development of systems like ENDOANGEL-GEV, assessing bleeding risk stigmata for GEV which are detected during endoscopic surveillance in patients with cirrhosis. By identifying these risk factors, ENDOANGEL-GEV can pick “candidates” for prophylactic treatment. The system comprises six models that segment GEV, grade them (grades 1-3), and evaluate red color signs (RC) (RC0-RC3). ENDOANGEL-GEV has been trained on 6034 images from 1156 cirrhotic patients across three hospitals (dataset 1) and validated using multicenter datasets comprising 11009 images from 141 videos (dataset 2), as well as a prospective study involving 161 cirrhotic patients from Renmin Hospital of Wuhan University (dataset 3). In dataset 1, ENDOANGEL-GEV achieved intersection over union values of 0.8087 for segmenting esophageal varices and 0.8141 for GV. The system maintained consistent accuracy across images from multiple hospitals in dataset 2. In dataset 3, it outperformed attending endoscopists in detecting RC of GEV and classifying grades (P < 0.001). When risk was combined with the Child-Pugh score, ENDOANGEL-GEV surpassed endoscopists for esophageal varices (P < 0.001) and showed comparable performance for GV (P = 0.152). Notably, ENDOANGEL-GEV may help 12.31% (16/130) more patients receive the appropriate intervention compared to endoscopists[29]. Based on these promising data perhaps the use of this technology before or during EUS-GV treatment could help confirm or refute the need for variceal treatment.

CONCLUSION

EUS-guided treatment of GV offers various advantages, including enhanced visualization and precise targeting of varices, even in complex cases. EUS-guided therapies, may reduce the amount of therapeutic agents needed and minimize associated adverse events. They can be utilized as parts of treatment algorithms where EUS-guided interventions, including EUS-coil/CYA, are first-line options, followed by transjugular intrahepatic portosystemic shunt if initial hemostasis fails. EUS-guided approaches, on the other hand, face limitations, such as the need for significant expertise, variability in techniques and equipment, and the restricted availability of CYA due to regulatory issues. Standardization of EUS-guided methods and obtaining approval from regulatory authorities e.g. Food and Drug Administration are essential for broader adoption. In all, while EUS-guided therapy of GV holds potential, further research preferably incorporating innovative techniques and technology e.g. AI, may help establish its superiority over existing therapies while addressing the aforementioned limitations.