Safety and effectiveness of endoscopic ultrasound-guided coil plus glue vs. direct endoscopic glue injection for the treatment of gastric varices: a systematic review and meta-analysis

Abstract

Background/Aims

Although endoscopic glue injection is the conventional first-line treatment for gastric variceal (GV) bleeding, a large body of evidence supports the safety and effectiveness of endoscopic ultrasound (EUS)-guided coil and glue injections. We compared the safety and efficacy of these two modalities.

Methods

A random-effects model was used for the pooled estimates. The outcomes of interest were the odds of technical success, GV obliteration, adverse events, rebleeding, reintervention, and pulmonary embolization (PE).

Results

Eight studies involving 659 patients (EUS-guided coil+glue, n=200; endoscopic glue, n=459) were included in the meta-analysis. The rebleeding (odds ratio [OR], 0.28, 95% confidence interval [CI], 0.16–0.51; I²=0%) and reintervention (OR, 0.28; 95% CI, 0.15–0.52; I²=0%) rates were significantly lower in the EUS-guided coil+glue. EUS-guided coil+glue group had higher rates (OR, 4.56; 95% CI, 2.04 –10.19; I²=30%) of GV obliteration after the first session. No difference in the odds of technical success (OR, 2.03; 95% CI, 0.50–8.22; I²=0%), adverse events (OR, 1.19; 95% CI, 0.44–3.20; I²=45%), abdominal pain (OR, 1.17; 95% CI, 0.36–3.82; I²=44%) fever (OR, 1.09; 95% CI, 0.41–2.92; I²=0%), PE (OR, 0.52; 95% CI, 0.19–1.47; I²=0%) and all-cause mortality (OR, 0.67; 95% CI, 0.32–1.40; I²=0%) was identified.

Conclusions

EUS-guided coil and glue injections are safe and effective for the treatment of gastric varices.

Graphical abstract

INTRODUCTION

Gastrointestinal bleeding secondary to gastric varices (GVs) is less common than that secondary to esophageal varices, but is more severe and often life-threatening. In most cases, GV develop secondary to portal hypertension in patients with decompensated cirrhosis and account for approximately 20% of all cases of varices.1,2 GV can develop in the absence of portal hypertension in diseases such as pancreatic cancer, essential thrombocytosis, polycythemia vera, and acute and chronic pancreatitis.3

GVs are associated with an increased rate of uncontrolled bleeding in up to 25% of patients, with a rebleeding rate of approximately 40% and a 50% rate of all-cause mortality.4,5 Different modalities have been used to achieve hemostasis in GV bleeding, including cyanoacrylate (CYA) glue injection (direct endoscopic or endoscopic ultrasound [EUS]-guided), EUS-guided coil embolization, band ligation, and interventional radiologic interventions such as transjugular intrahepatic portosystemic shunt and balloon-occluded retrograde transvenous obliteration.6,7 The application of CYA injections is the first-line treatment strategy, with a clinical success rate greater than 90% and has a lower rebleeding rate.8,9 Although endoscopic CYA injection has a high clinical success rate, it is associated with serious adverse events, such as pulmonary embolism, peritonitis, fever, abdominal pain, development of ulcers at the injection site, and death.10-12

EUS-guided coil embolization was first successfully performed in 2008 by Levy et al.13 for the management of refractory variceal bleeding. This approach has the advantage of direct visualization of the feeding vessels in many cases. This confirms the therapeutic effect at the time of intervention, potentially reducing the risk of systemic embolization.14 Although there are previous data on the effectiveness and safety of EUS-guided coil plus glue combination therapy, it has been reported to be more efficient than either glue or coil monotherapy.6 Previous systematic reviews on this topic have mainly reported data from single-arm studies. We investigated the effectiveness and safety of the combination of EUS-guided coil and glue therapy with endoscopic CYA glue injection by including only comparative studies.

METHODS

Data search and screening

We performed an online literature search of various databases, including PubMed, Medline, Google Scholar, Cochrane Library, Ovid Embase, Scopus, and Web of Science, from inception to October 30, 2024. Two reviewers (Z.I.T. and M.G.) conducted the literature search with the help of an experienced librarian team at the University of Missouri. The following keywords were used to perform the literature search: “Endoscopic Ultrasound” AND/OR “EUS,” “Gastric varices,” OR “variceal bleeding,” “Coil,” “Glue” OR “coil plus Glue.” In addition to the above-mentioned free keywords, further searches were conducted using medical subject heading (MeSH) terms and Embase subject heading (EMTREE) terms. We also manually screened the reference lists of the articles included in the final analysis to identify any missing studies. The citations for the shortlisted studies were downloaded from Zotero, which is a free and open-access citation manager, and duplicates were screened.

Inclusion and exclusion criteria

We applied patient, intervention, comparison, and outcome (PICO) frameworks to establish the inclusion criteria. (1) Patients aged >18 years receiving treatment for gastric variceal bleeding; (2) patients in whom EUS-guided coil and glue embolization was used to treat gastric variceal bleeding; (3) Compared to patients of gastric variceal bleeding who received endoscopic glue injection; and (4) Outcomes reported are technical success, adverse events, odds of pulmonary embolism, rebleeding, GV obliteration, reintervention, all-cause mortality, abdominal pain, and fever. All peer-reviewed studies, prospective cohort studies, retrospective case-control or cohort studies, case series with greater than 10 patients, and randomized trials were included in the final analyses. We excluded case reports, case series, studies with fewer than 10 patients, letters to the editor, and review articles. We included only studies with comparative data and excluded single-arm studies. Studies with overlapping or duplicate data were excluded.

Data extraction

Three researchers (Z.I.T., M.G., and U.F.) reviewed the final selection of studies and extracted the data on separate Excel sheets. Any conflicts were discussed and resolved by a senior author (N.T.) and a final version of the data was prepared before conducting the analysis. Information about the study and patient characteristics was extracted, including the study design, year of study, patient age, sex, underlying cause of cirrhosis and varices, type of intervention, volume of CYA glue, grade of varices, follow-up duration, adverse events, and procedure success.

Outcome measures and definitions

The outcomes of interest were technical success, rate of pulmonary embolization, overall adverse events, abdominal pain and fever, rate of reintervention, all-cause mortality, GV obliteration rate in the 1st and 2nd sessions, and volume of glue used. Technical success was defined as the completion of the procedure without interruption. GV obliteration was considered successful if there was no evidence of varices on repeat endoscopy, absence of blood flow on Doppler ultrasound, or hardening of the varix on palpation. Rebleeding was defined as a new onset of hematemesis or melena or a drop in hemoglobin of ≥2 g/dL from baseline after the initial intervention. Reintervention was documented if a repeat procedure was performed because of rebleeding in a previously bled GV or the first episode of bleeding after primary prophylaxis. Ectopic embolism was diagnosed based on the presence of symptoms and confirmed using imaging.

Data synthesis

This systematic review and meta-analysis was performed using Open Meta-analyst.15 Random effects model using the Der Simonian-Laird method was applied to calculate pooled estimates with 95% confidence intervals (CIs). Odds ratios (ORs) with corresponding 95% CI were reported for the binary outcome, and the mean difference (MD) with 95% CI for the continuous variable. Cochrane chi-square and I² statistics were used to assess variance and heterogeneity. The heterogeneity was considered mild if I² <25%, moderate if I² <50%, and substantial if I² >50%. Forest plots were generated to report the pooled odds.

Quality assessment and publication bias

We used the Methodological Index for Non-Randomized Studies (MINORS) criteria16 to assess the quality of the included retrospective studies. We scored the comparative studies on 12 items of the MINORS criteria, and each item was scored from 0–2 (0 if not reported, 1 when reported but inadequate, and 2 when reported and adequate). Thus, a maximum ideal score of 24 was obtained for comparative studies, and 16 for non-comparative studies. Based on MINORS criteria for non-randomized studies, the quality of studies was classified as poor (score ≤5), fair (score 6–10), or high quality (≥11), as described previously. As such, all studies were rated as high quality. We used the Risk of Bias-217 criteria to assess the randomized controlled trial (RCT) quality. The quality assessment of the studies is summarized in Supplementary Tables 1, 2.

RESULTS

Literature search and study selection

The initial literature search yielded 1,613 manuscripts. After removing the duplicates, 318 articles were selected for review. After screening titles and abstracts, 245 manuscripts were excluded. The reasons for manuscript exclusion included incorrect patient population (n=121), incorrect study design (n=55), case reports and case series of <10 patients (n=58), and review articles (n=11). We shortlisted 73 manuscripts for full-text review, and finally, eight studies10,14,18–23 were included in the qualitative analysis. The reasons for excluding the remaining studies were as follows: glue versus coil only (n=1), coil or glue versus radiological intervention (n=39), use of thrombin injection (n=2), glue plus coil single-arm studies (n=12), band ligation versus glue (n=8), and coil only (n=4). Figure 1 illustrates the inclusion and exclusion criteria.

Fig. 1.

Prisma flow diagram. IR, interventional radiology.

Patient and study characteristics

Eight studies involving 659 patients (EUS-guided coil embolization plus glue, n=200; endoscopic glue injection, n=459) were analyzed. Two studies were randomized trials and the remaining six studies were retrospective cohorts. The mean age was 52.4±23 in the EUS-guided coil plus glue group and 53.1±21 in the endoscopic glue group. The proportions of male patients were 68.3% in the combination group and 51.8% in the endoscopic glue groups. In the combination group, 32% patients had isolated gastric varices (IGV)-1 type and 68% gastroesophageal (GOV)-2 type varices. In the endoscopic glue group, 61% had IGV-1 and 39% had GOV-2 type GVs. The underlying etiology of GVs in the combination vs. the endoscopic glue group was as follows: alcohol (36.5% vs. 35.8%), hepatitis B virus (HBV, 14.5% vs. 13.3%), hepatitis C virus (HCV, 10.2% vs. 9.2%), non-alcoholic steatohepatitis (NASH, 22% vs. 23%), cryptogenic or others, including autoimmune hepatitis (AIH), schistosomiasis, pancreatitis, non-cirrhotic portal hypertension (NCPH) (5.9% vs. 18.9%). Table 1 shows the patient and study characteristics.6,10,14,18-22

Table 1.

Baseline characteristics of study population

Study	Design, published year	Follow-up period	Total (n)	n		Age (yr)		Male		GV types		MELD		Varix size		Indication for treatment		Etiology
				EUS plus glue	Endoscopic glue	EUS coil plus glue	Endoscopic glue	EUS coil plus glue	Endoscopic glue	EUS+glue	Glue	EUS+glue	Glue	EUS+glue	Glue	EUS coil plus glue	Endoscopic glue	EUS plus glue	Glue
Galvis-García et al.20	Retrospective, 2022	221 d	68	15	53	63.1±9.8	54.3 ±9.4	6	21	IGV-1=0, GOV-2=15	IGV-1= 20, GOV-2=33	9.5 (6–13)	9.5 (6–30)	25±15.1	23±12	Primary prophylaxis 0, active bleeding 0, Secondary prophylaxis 15	Primary prophylaxis 4, active bleeding 1, secondary prophylaxis 48	Alcohol 5, NASH 10	Alcohol 23, NASH 30
Lôbo et al.18	RCT, 2019	12 mo	32	16	16	49.3±14.8	57.69±11.5	8	5	IGV-1=3, GOV-2=13	IGV-1 =3, GOV-2 = 13	NA	NA	36 ±11.3	30 ± 8.8	NA	NA	Cryptogenic 4, HCV 4, alcohol 3, schistosomiasis 2, congenital bile duct atresia 1, Budd Chiari 1, drug 1	Cryptogenic 3, HCV 3, alcohol 3, schistosomiasis 3, HBV 2, HBV+HCV 1, HAV+alcohol 1
Chen et al.14	Retrospective, 2024	13 mo	57	21	36	57.5±7.7	58.8±8	14	27	IGV-1 =7, GOV-2 = 13, GOV-1 = 1	IGV-1=15, GOV-2=19, GOV-1=2	10.2±3.5	9.5±3.4	Size: 18.4±5.9	Size 19.5±7.9	Active bleeding 1, secondary prophylaxis 20	Acute bleeding 4, secondary prophylaxis 32	HBV 13, HCV 1, alcohol 2, schistosomiasis 1, AIH 1, cryptogenic 3	HBV 24, HCV2, alcohol 4, schistosomiasis 1, AIH 2, cryptogenic 3
Jamwal et al.22	Retrospective, 2023	12 mo	80	40	40	44.1±8.2	43.2±7.8	30	28			21±1.54	20±1.41	Large exact size not reported	Large exact size not reported	Secondary prophylaxis 40	Secondary prophylaxis 40	Alcohol 20, NASH 12, HBV 4, HCV 2, others 2	Alcohol 18, NASH 10, HBV 6, HCV 2, others 2
Jhajharia et al.21	RCT, 2025	12 mo	50	24	26	40.2 ±4.88	41.1 ±5.23	17	19	IGV1=8, GOV-2=16	IGV-1 =8, GOV-2 =18	19.6 (17–23)	20.2 (16–22)	>10 mm	>10 mm	Primary prophylaxis 7, active bleeding 5, secondary prophylaxis 12	Primary prophylaxis 4, active bleeding 6, secondary prophylaxis 16	Alcohol 11, HBV 2, HCV 1, NASH 6, others 2, pancreatitis 2	Alcohol 14, HBV 1, HCV 1, NASH 4, others 3, pancreatitis 3
Robles-Medranda et al.10	Retrospective, 2021	10 mo	36	17	19	63.29±8.8	62.83±11.5	20	10	IGV-1=5, GOV-2=12	IGV-1=7, GOV-2=12	NA	NA	22.6±6.8	21.1±8.7	Active bleeding 7, secondary prophylaxis	Active bleeding 19, secondary prophylaxis 0	NA	NA
Samanta et al.6	Retrospective 2023	701.5 d	176	58	118	44.3±12.1	46.63±13.6	NA	NA	IGV-1 =27, GOV-2= 31	IGV-1=90 GOV-2=128	NA	NA	21.46±7.9	16.60±4.5	Active bleeding 7, stigmata of recent bleeding 36, primary prophylaxis 15	Active bleeding 47, stigmata of recent bleeding 109, primary prophylaxis 62	Alcohol 25, HBV 6, HCV 9, NASH 9, NCPH 5, AIH 2	Alcohol 77, HBV 19, HCV 27, NASH 46, NCPH 23, AIH 12
Pegatraju et al.19	Retrospective, 2022	NP	60	9	51	48.5 (10–74)	56 (22–78)	6	33	IGV-1 9	IGV-1 51	9 (6–18)	13 (6–40)	NA	NA	Active bleeding 9	Active bleeding 51	NA	NA

EUS, endoscopic ultrasound; GV, gastric varices; MELD, model for end-stage liver disease; IGV, isolated gastric varice; GOV, gastroesophageal; NASH, non-alcoholic steatohepatitis; RCT, randomized controlled trial; NA, not available; HCV, hepatitis C virus; HBV, hepatitis B virus; HAV, hepatitis A virus; AIH, autoimmune hepatitis; NCPH, non-cirrhotic portal hypertension; NP, not provided.

Outcomes

1) Technical success

All patients in the EUS-guided coil-and-glue combination group achieved technical success, whereas 10 patients in the endoscopic glue only group failed to achieve hemostasis. In the combination group, the pooled event rate of technical success was 98.6% (95% CI, 96.7%–100%, I²=0%). The pooled event rate in the glue only group was 97.7% (95% CI, 95.3%–100%, I²=34%) (Table 2). Though the odds (OR, 2.03; 95% CI, 0.50–8.22; I²=0%) of technical success were high in the EUS coil and glue group, it did not reach statistical significance (Fig. 2). This result showed low heterogeneity, as evidenced by the I² of 0.

Table 2.

Pooled odds of outcomes in EUS-guided glue plus coil and endoscopic glue group

Outcome	Sample size EUS coil plus glue (n)	Total events coil plus glue (n)	Pooled event rate (%)	95% CI withI2	Sample size endoscopic glue (n)	Total events endoscopic glue (n)	Pooled event rate (%)	95% CI withI2
Technical success	151	151	98.6	96.7–100, I2=0%	268	259	97.7	95.3–100.0, I2=34%
Overall adverse events	200	22	11	3.6–18.4, I2=71%	359	44	9.5	4.1–14.8, I2=72%
Abdominal pain	191	15	6.8	1.2–12.4, I2=64%	311	34	7.8	3.3–12.3, I2=57%
Fever	133	7	4.7	1.2–8.3, I2=0%	190	9	4.4	1.5–7.2,I2=0%
Pulmonary embolism	185	4	1.5	0.02 to 3.1, I2=0%	306	12	2.3	0.05 to 5.1, I2=65%
Rebleeding	200	15	5.3	1.5–9.2, I2=36%	359	92	22.2	13.5–31.0, I2=76%
Reintervention	163	16	9.0	1.3–16.8, I2=72%	253	71	26.0	8.8–43.3, I2=92%
All-cause mortality	142	12	5.7	0.10–10.3,I2=41%	241	33	11.7	4.4–19.0, I2=78%
GV obliteration 1st session	174	156	91.2	84.4–98.1, I2=71%	289	191	68.4	51.5–85.2, I2=92%
GV obliteration 2nd session	133	134	98.6	96.7–100, I2=0%	289	261	89.1	81.2–97.1, I2=85%
Glue volume			1.78	1.59–1.97, I2=89%			2.89	2.29–3.49, I2=96%

EUS, endoscopic ultrasound; CI, confidence interval; GV, gastric varices.

Fig. 2.

Forest plot to compare technical success between endoscopic ultrasound glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

Adverse events

A total of 44 adverse events were observed in the endoscopic glue group compared with 22 adverse events in the combination group. The pooled event rate of overall adverse events was 11% (95% CI, 3.6%–18.4%; I²=71%) and 9.5% (95% CI, 4.1%–14.8%; I²=72%) in the combination therapy and endoscopic glue only group, respectively (Table 2). No difference in pooled odds (OR, 1.19; 95% CI, 0.44–3.20, I²=45%) of adverse events was seen between both groups (Fig. 3A). After excluding one study, the sensitivity analysis results remained consistent (Supplementary Fig. 1A).

Fig. 3.

(A) Forest plot to compare overall adverse events between endoscopic ultrasound (EUS) glue plus coil vs. direct endoscopic glue injection. (B) Forest plot to compare odds of fever between EUS glue plus coil vs. direct endoscopic glue injection. (C) Forest plot to compare odds of abdominal pain between EUS glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

On stratified analysis of adverse events, nine patients in the glue-only group developed fever (pooled event rate, 4.4%; 95% CI, 1.5%–7.2%; I²=0%), and 34 patients had post-procedure abdominal pain (pooled event rate, 7.8%; 95% CI, 3.3%–12.3%; I²=56%) (Table 2). In the combination group, 15 patients had abdominal pain after the procedure (pooled event rate, 6.8%; 95% CI, 1.2%–12.4%; I²=64%), and 7 had fever (event rate, 4.7%; 95% CI, 1.2%–8.3%; I²=0%) (Table 2). On comparative pooled analysis, we did not find any difference in the odds (OR, 1.09; 95% CI, 0.41–2.92; I²=0%) of fever and abdominal pain (OR, 1.17; 95% CI, 0.36–3.82; I²=44%) between groups (Fig. 3B, C).

Variceal obliteration

The pooled event rate of overall GV obliteration was 98.7% (95% CI, 97.0%–100%; I²=0%) in EUS-guided coil plus glue compared to 95.8% (95% CI, 92.2%–99.4%; I²=57%) in the endoscopic glue group. GV obliteration was achieved in a single session in 91.2% (95% CI, 84.4%–98.1%; I²=71%) of the combination group compared to 68.4% (95% CI, 51.5%–85.2%; I²=92%) in the endoscopic glue group. Following the two procedures, the combination group had 98.6% (95% CI, 96.7%–100%; I²=0%) GV obliteration rate compared to 89.1% (95% CI, 81.2%–97.1%; I²=85%) in the endoscopic glue group (Table 2). The pooled odds (4.56; 95% CI, 2.04–10.19; I²=30%) of GV obliteration were significantly higher in the combination group after the first session (Fig. 4A). No difference in the odds of GV obliteration was seen after the 2nd session (2.17; 95% CI, 0.51–8.87; I²= 0%) and in overall GV obliteration odds (2.88; 95% CI, 0.75–10.89; I²=0%) (Fig. 4B, C).

Fig. 4.

(A) Forest plot to compare gastric variceal (GV) obliteration after 1st session between endoscopic ultrasound (EUS) glue plus coil vs. direct endoscopic glue injection. (B) Forest plot to compare GV obliteration after 2nd session between EUS glue plus coil vs. direct endoscopic glue injection. (C) Forest plot to compare overall GV obliteration between EUS glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

Pulmonary embolization rate

A total of 12 pulmonary embolization events were observed in the endoscopic glue group compared to 4 events in the combination group. The pooled event rate of embolization was 2.3% (95% CI, 0.05%–5.1%; I²=65%) in the glue group and 1.5% (95% CI, 0.02%–3.1%; I²=0%) in the combination group (Table 2). There was no difference in the pooled odds (OR, 0.52; 95% CI, 0.19–1.47; I²=0%) of embolization when comparing both groups (Fig. 5). The I² of zero signified low-to-no observed heterogeneity. These results were consistent with those of the leave-one-out sensitivity analysis (Supplementary Fig. 1B).

Fig. 5.

Forest plot to compare pulmonary embolization odds between endoscopic ultrasound glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

Rebleeding rate

Fifteen cases of rebleeding occurred in the EUS-guided coil plus glue group, whereas 92 patients experienced rebleeding in the endoscopic glue intervention study group. The pooled event rate of rebleeding in the combination group was 5.3% (95% CI, 1.5%–9.2%; I²=22%) and 22.2% (95% CI, 13.5%–31%; I²=76%) in the glue group (Table 2). The pooled odds (OR, 0.28; 95% CI, 0.16–0.51; I²=0%) of rebleeding were significantly lower in the combination group (Fig. 6). The I² of 0% shows the homogeneity of these results. Sensitivity analysis performed by excluding one study individually did not change the results (Supplementary Fig. 1C).

Fig. 6.

Forest plot to compare rebleeding odds between endoscopic ultrasound glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

Reintervention rate

Sixteen patients in the combination group underwent repeat interventions, whereas 71 patients in the endoscopic glue group required retreatment. The pooled event rate of reintervention was 26.0% (95% CI, 8.8%–43.3%; I²=90%) in the endoscopic glue group, while it was 9.0% (95% CI, 1.3%–16.8%; I²=65%) in the combination group (Table 2). The pooled odds (OR, 0.28; 95% CI, 0.15–0.52; I²=0%) of reintervention were significantly lower in the combination group (Fig. 7). Leave-one-out sensitivity analysis confirmed these results (Supplementary Fig. 1D).

Fig. 7.

Forest plot to compare reintervention odds between endoscopic ultrasound glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

All-cause mortality

Twelve patients in the EUS-guided glue plus coil group and 33 patients in the endoscopic glue group died during follow-up. No procedure-related deaths occurred. The most common causes of death were hepatic encephalopathy, followed by bleeding, hepatocellular carcinoma, and sepsis. The pooled event rate of mortality in the EUS-guided coil plus glue group was 5.7% (95% CI, 0.10%–10.3%; I²=41%) and 11.7% (95% CI, 4.4%–19.0%; I²=77%) in the endoscopic glue group (Table 2). No difference in the odds (OR, 0.67, 95% CI, 0.32–1.40; I²=0%) of all-cause mortality was noted between both groups (Fig. 8). Leaving out the study sensitivity analysis did not change the results (Supplementary Fig. 1E).

Fig. 8.

Forest plot to compare all-cause mortality between endoscopic ultrasound glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

Glue volume used

The mean volume of glue applied in the endoscopic glue group was 2.89 mL (95% CI, 2.29–3.49; I²=96%), while 1.78 mL (95% CI, 1.59–1.97; I²=89%) of glue was used in the EUS-guided coil plus glue intervention group (Table 2). The mean volume of glue used in coil plus glue group was less (MD, –1.14; 95% CI, –1.92 to 0.37) compared to glue only group (Supplementary Fig. 2).

DISCUSSION

This systematic review and meta-analysis of 8 studies, 6 retrospective studies, and 2 RCTs showed no difference in the odds of technical success between EUS-guided coil embolization plus glue and direct endoscopic glue injection. This review further reports that the GV obliteration rate was 91.2% in the combination group and 70.3% in the endoscopic glue group after the first session, and 98.6% vs. 89.1% after the second session, respectively. We found no difference in the odds of overall GV obliteration between the two groups, patients who underwent EUS-guided coil and glue injection had significantly higher odds of obliteration after the first session. These findings are similar to data in the published literature.23,24

We found a technical success rate of 100% with the combination of EUS-guided coils and glue, which is in accordance with the previously reported success rate of 100% by Baig et al.23 The difference between the systematic review and meta-analysis by Baig et al.23 and our study is that we further analyzed the pooled event rate of technical success in the endoscopic glue group and compared the success rates between the groups. In our analysis, we found that 9% of patients in the combination group underwent reintervention, while 26% required reintervention in the glue-only group. These values were lower than those reported in a recent meta-analysis25 conducted on the same topic, which included only six studies. We included eight studies with a larger number of patients, which may explain this difference.

Previously, Mohan et al.26 conducted a meta-analysis on the role of EUS-guided treatment of GVs and reported that EUS-guided modalities including glue, coil, or a combination of both have comparable efficacy compared to endoscopic glue injection; however, in this analysis, the majority of the studies were single arm and direct comparison was not performed between EUS-guided coil plus glue and endoscopic glue injection. They only reported comparative data on treatment efficiency, GV obliteration, and rebleeding, whereas we provided a detailed comparative analysis of the odds of PE, overall adverse events, fever, abdominal pain, and reintervention, in addition to the outcomes mentioned above.

Another systematic review on the same topic was published by McCarty et al.24 in which 11 studies were included. This meta-analysis included both single-arm and comparative studies. These findings support the results of the present study. The advantage of our analysis in comparison with the studies mentioned above is that we excluded single-arm studies and included studies that provided a direct comparison between EUS-guided coil-plus-glue and endoscopic glue injections, which improved the robustness of our results.

Since the first successful application of a combination of EUS-guided coil and glue embolization in 2011,27 more studies have documented the safety and efficiency of this combination therapy for both bleeding and non-bleeding GVs.20,28,29 A significant concern regarding the endoscopic treatment of GV is systemic embolization, as reported in previous studies.26,30 In our meta-analysis, we found that patients who underwent EUS-guided coil embolization had an embolic event in 1.5% of cases compared to 2.3% cases in the endoscopic glue group, with no difference in the odds of embolization between both treatment groups, which is the same as reported by the recent analysis by Florencio de Mesquita et al.25

A critical finding of our study was that patients who underwent the combination of EUS-guided coil and glue therapy had lower odds of rebleeding, leading to decreased odds of reintervention, which results in the cost-effectiveness of this dual therapy, as reported by Robles-Medranda et al.10 There were only 15 cases of rebleeding after the intervention in the combination group compared to 92 cases in the endoscopic glue group. The pooled odds of bleeding were 5.3% in EUS compared to 22.2% in the endoscopic glue groups. In the combination group, 11% of patients developed adverse events, whereas 9.5% of patients who underwent endoscopic glue application experienced adverse events. We did not find any statistically significant differences in the odds of overall adverse events between the two groups, which is consistent with previous reports.14,25 In the subgroup analysis of adverse events, no differences in the odds of abdominal pain or fever were noted between the two groups. The all-cause mortality event rate was 5.7% in the combination group compared to 11.7% in the endoscopic glue group, with no difference in the pooled odds of mortality between both groups.

Another interesting finding in our analysis was the decreased amount of glue utilization in the combination group compared to endoscopic glue only group. The mean volume of glue applied in the endoscopic glue group was 2.89 mL compared to 1.78 mL used in combination EUS-guided coil and glue group. A previous study by Jamwal et al.22 reported that 6 mL of glue was used in the endoscopic glue group, compared with only 2 mL in the combination group.

The use of EUS-guided injection therapy for the treatment of GVs is gaining popularity because studies have demonstrated its safety and several advantages over conventional endoscopic therapy. EUS enhances treatment options, provides better visualization of bleeding vessels, and confirms treatment success in real time.31 Endoscopic cyanoacrylate injection is still considered the preferred first-line treatment for the management of bleeding GVs and is a widely used method with a good safety profile. It is recommended as initial therapy for bleeding GVs by European Society of Gastrointestinal Endoscopy.32 Significant data are available on the safety and efficacy of EUS-based therapies, which advocate their use as the preferred treatment modality for acute GV bleeding, as outlined by the American Gastroenterological Association practice committee.31 The findings of our meta-analysis reinforce the use of EUS-guided CYA and glue as the preferred initial treatment method for actively bleeding GVs. Although the use of EUS-guided modalities has increased recently, it is still a complex procedure, requires a high level of expertise, and should preferably be performed in higher volume centers.31 RCTs comparing EUS-guided therapies with our treatment options, including interventional radiology-guided therapies, are required to further strengthen the safety and success of this method.

To date, our analysis is the most recent systematic review and meta-analysis. We compared our review with the most recently published meta-analysis by Florencio de Mesquita et al.,25 consisting of six studies. We applied more stringent inclusion and exclusion criteria and found two more studies, which led to a greater number of patients being analyzed. Second, we conducted sensitivity analysis by removing each study individually. Third, we analyzed the variceal obliteration rate in both groups and measured the amount of glue used, which was not reported in a recent meta-analysis. Fourth, we included only studies that reported a direct comparison between the two groups. Fifth, the study outcome definitions were uniform, which further enhanced the strength of our results.

Although we reported the most updated data on the comparison of combination therapy versus endoscopic glue injection, it still has certain limitations. Most of the included studies were retrospective, which carried the risk of measured and unmeasured bias. Secondly, publication bias was not determined because of the small number of studies included. Third, many studies did not provide data on stratified outcomes based on patient characteristics and the underlying etiology of the GVs. In addition, data on technical and clinical success based on the indications for the procedures (active bleeding vs. primary or secondary prophylaxis) were not provided. Fourth, some outcomes, including adverse events, reintervention rate, and GV obliteration, showed substantial heterogeneity, which is likely secondary to different study designs, a low number of patients in both groups, and a significant difference in the basic characteristics of the study population between the groups. We then conducted a sensitivity analysis by removing one study at a time to overcome this limitation, and found no difference between the results. There was also a significant difference in the baseline characteristics of the study population, such as the proportions of IGV1 and GOV2, between the two groups, which was a potential cause of high heterogeneity in some outcomes.

We conclude that EUS-guided coil and glue are safe and effective methods for treating both bleeding and non-bleeding GVs. When compared to direct endoscopic glue injection, it has significantly lower odds of reintervention and rebleeding, which also decreases procedural and hospitalization costs. Further randomized trials with longer follow-up durations will help validate these findings.

Supplementary Material

Supplementary Table 1. Minors criteria for quality assessment of retrospective studies.

Supplementary Table 2. Risk of Bias-2: randomized controlled trials.

Supplementary Fig. 1. (A) Leave out one study sensitivity analysis of overall adverse events. (B) Leave out one study sensitivity analysis of pulmonary embolization. (C) Leave out one study sensitivity analysis of rebleeding rate. (D) Leave out one study sensitivity analysis of reintervention. (E) leave out one study sensitivity analysis of all-cause mortality. CI, confidence interval.

Supplementary Fig. 2. Forest plot to compare gel volume use between endoscopic ultrasound glue plus coil vs. direct endoscopic glue injection. CI, confidence interval.

Supplementary materials related to this article can be found online at https://doi.org/ce.2025.251.