Endoscopic Ultrasound-Guided Biliary Drainage in Altered Gastrointestinal Anatomy: A Systematic Review and Meta-Analysis

Summary

This study presents a meta-analysis comparing endoscopic ultrasound-guided biliary drainage (EUS-BD) and endoscopic retrograde cholangiopancreatography (ERCP) for biliary obstruction in patients with altered gastrointestinal anatomy.

Abstract

ERCP intubation is often difficult in patients with biliary obstruction and changes in gastrointestinal structure, such as after Roux-en-Y or Billroth II surgery. EUS-BD is an alternative method. This meta-analysis should systematically compare the efficacy and safety of EUS-BD and ERCP in these patients. Studies were retrieved from PubMed, Embase, Web of Science, and the Cochrane Library from 2011 to 2026. Original studies comparing EUS-BD and ERCP in patients with biliary obstruction and a history of gastrointestinal reconstruction surgery were included. Two researchers independently carried out literature screening, data extraction, and quality evaluation. A meta-analysis was conducted using Review Manager. The main outcome measures are technical success rate, clinical success rate, adverse events, and procedure time. A total of 6 studies involving 572 patients (201 in the EUS-BD group and 371 in the ERCP group) were included in the technical success analysis. The meta-analysis demonstrated that the pooled technical success rate was 90.5% (182/201) for the EUS-BD group and 70.4% (261/371) for the ERCP group, yielding an odds ratio (OR) of 3.60 (95% CI: 2.12–6.13; p < 0.00001). For clinical success, the pooled odds ratio was 3.52 (95% CI: 2.07–5.98; p < 0.00001), demonstrating a statistically significant advantage over ERCP. In terms of safety, there was no difference in the overall incidence of adverse events between the two groups (25.9% vs. 20.5%; OR: 1.54, 95% CI: 0.97–2.44). In patients with biliary obstruction and surgically altered gastrointestinal anatomy, EUS-BD was associated with higher pooled technical and clinical success rates in available observational studies, along with shorter procedure time than ERCP. However, current evidence remains insufficient to establish a definitive difference in overall adverse events. EUS-BD represents a viable and highly effective alternative drainage modality for this challenging population, particularly when standard ERCP is anticipated to be technically demanding.

Introduction

Biliary obstruction is an obstructive condition. Pathologically, the biliary system is blocked¹. The etiologies include pancreatic diseases, cholangiocarcinoma, ampullary cancer, gallstones, surgical conditions, etc². The core pathological change is bile stasis and high pressure in the bile duct, which leads to a series of symptoms and complications³. Endoscopic interventional techniques such as ERCP-BD and EUS-BD are clinically used⁴. ERCP-BD is the main palliative treatment method for biliary obstruction, which can relieve symptoms and prolong the survival period⁵. Anatomical problems may affect its success, especially when the ampulla is invaded by the duodenum, which is the main cause of ERCP-BD failure, with a failure rate ranging from 0.5% to 16%. Procedure - related pancreatitis is a major problem in clinical practice^6,7. Despite these limitations, ERCP-BD continues to be a cornerstone intervention, and numerous meta-analyses have been conducted to systematically compare its therapeutic efficacy against alternative drainage methods, such as EUS-BD^8,9.

In recent years, endoscopic ultrasound-guided biliary drainage has rapidly emerged in the treatment of malignant biliary obstruction, with its clinical applications significantly expanding^4,10. EUS is a cutting-edge medical device that combines endoscopic and ultrasonic technologies, offering outstanding advantages: Firstly, it provides a more flexible puncture pathway, allowing direct access to the intrahepatic or extrahepatic bile ducts via the stomach or proximal duodenum, thereby offering a reliable drainage route for patients with duodenal invasion or postsurgical anatomical alterations. Secondly, this technology avoids complications related to ERCP, especially without the need for pancreatic duct intubation to reduce the risk of the pancreas¹¹. In addition, under real-time ultrasonic guidance, EUS-BD can evaluate the tissue around the bile duct and the infiltration of tumors, thereby providing information for supplementary diagnosis and staging. Furthermore, this technique itself serves as a key solution to overcome the clinical impasses of failed ERCP or postoperative reobstruction, providing a strategy for treatment pathways¹².

While numerous meta-analyses have systematically compared EUS-BD and ERCP for conventional biliary drainage^12,13,14, evidence remains scarce regarding patients with surgically altered gastrointestinal anatomy (SAGA). In this specific population, standard ERCP is technically demanding and often unsuccessful due to extensive anatomical distortion. It is hypothesized that EUS-BD, which inherently bypasses these anatomical barriers via direct transmural access, will demonstrate superior technical and clinical success rates alongside a comparable safety profile when compared to ERCP. To the best of current knowledge, this is the first systematic review and meta-analysis specifically designed to evaluate the comparative efficacy and safety of these two modalities in the SAGA population. By providing a robust synthesis of current evidence, this study aims to offer novel insights and establish evidence-based guidance for clinical decision-making in this challenging demographic.

Protocol

Guidelines and official logging

The reporting complied with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement; the completed PRISMA checklist is provided in the Supplementary Materials. The study protocol was prospectively registered in PROSPERO (CRD420261297962). All the tools or materials used for this study are listed in the Table of Materials.

Search strategy

Three reviewers independently performed the comprehensive electronic search in duplicate, covering records from January 1st, 2011, to January 1st, 2026. This study employed a coverage strategy in its search, systematically retrieving electronic databases including PubMed, Embase, Web of Science, and the Cochrane Central Register of Controlled Trials (CENTRAL). Eligible studies are confined to observational cohort studies. To capture the most recent evidence, a supplementary search was conducted using the same databases and search strategy to identify any newly published observational cohort studies that met the inclusion criteria.

Inclusion criteria

The systematic review employed search terms such as: “biliary obstruction,” “obstructive jaundice,” “ERCP,” “EUS,” and “endoscopic ultrasound.” “surgical altered anatomy”, “common bile duct stones,” and “digestive tract reconstruction”. The detailed Boolean search strings and syntax specific to each database are provided in Supplementary Tables 1–4. There were no language restrictions, but the search was restricted to the last 15 years, from January 1st, 2011, to January 1st, 2026. For patients, the inclusion criteria were as follows: eligible patients were over 20 years of age¹⁵; EUS-BD (inclusive of any transmural or transpapillary ultrasound-guided approaches) or ERCP (including standard and balloon enteroscopy-assisted ERCP) was performed to manage their biliary obstruction; and they had a history of upper gastrointestinal surgery with Roux-en-Y or Billroth II reconstruction¹⁶. Technical or narrative reviews, noncomparative studies, case reports, and studies reporting on other organ obstructions were excluded. The references of the articles identified after the initial search were also manually reviewed.

Data extraction

Two reviewers (Yu, Liu) independently extracted detailed trial characteristics into Table 1, rigorously recording: specific EUS-BD and ERCP techniques, underlying surgical anatomy, benign versus malignant indications, study-specific definitions of technical and clinical success (including bilirubin reduction thresholds and assessment time points), and operator or institutional expertise. The data analyzed includes technical success, clinical success, cholangitis, pancreatitis, adverse events, stent occlusion, survival time, procedure time, length of hospital stay, stent dysfunction, and stent migration. For observational cohort studies with more than two groups or employing factorial designs that allowed multiple comparisons, only the relevant information and data pertaining to the research question as reported in the original publications were selectively extracted. If a meta-analysis indicated the inclusion of unpublished data obtained directly from the original authors, this data was extracted from its forest plots, and the original trial reports were cross-referenced to verify eligibility against the inclusion criteria. Data confirmed as relevant to our outcomes of interest were then integrated with data extracted from the primary trial publications for analysis.

Definitions

In this meta-analysis, EUS-BD was evaluated as a comprehensive interventional category. The specific procedural approaches subsumed under EUS-BD included EUS-guided hepaticogastrostomy (EUS-HGS), EUS-guided choledochoduodenostomy (EUS-CDS), EUS-guided antegrade intervention (EUS-AG), and the EUS-guided rendezvous technique (EUS-RV). For the purpose of the primary statistical analysis, outcome data from all these specific EUS-BD modalities were comprehensively pooled into a single experimental cohort to be compared against the ERCP control group. Technical success was determined based on the following criterion: the planned placement of a metal stent or removal of stones was successfully completed via EUS or ERCP. While clinical success was predominantly defined as a decrease in serum total bilirubin to below half of the baseline level within 1 to 4 weeks^15,17,18,19, the specific definitions utilized by each original study were adopted. These minor study-specific variations in defining clinical success (e.g., differences in the evaluation timeframe or the inclusion of clinical symptom resolution) were meticulously documented during data extraction.

All adverse events occurring within 14 days postoperatively were defined as overall adverse events, which included procedure-related events (bleeding, perforation including biliary or duodenal, acute pancreatitis, bile leakage); infectious events (cholangitis, cholecystitis, liver abscess, bacteremia/sepsis); stent-related events (early stent occlusion, migration [intraperitoneal or intraluminal], malposition, late reobstruction due to tumor growth); systemic events (sedation-related cardiopulmonary complications, contrast agent allergy); and other rare events (subcapsular liver hematoma, pneumoperitoneum, portal vein injury, retained sheared guidewire, procedure-related death)^17,19. RBO was defined as a condition characterized by both symptomatic presentation and radiologically confirmed dilation of the biliary tract¹⁵. The procedure time was defined from endoscope insertion to stent placement^17,19.

Quality assessment

Two independent reviewers (Yu and Liu) methodically appraised the methodological quality of the included observational studies utilizing the Newcastle-Ottawa Scale (NOS). This validated tool evaluates studies across three domains: selection of the study groups, comparability of the groups, and ascertainment of the outcome of interest. Studies scoring 7 to 9 points were deemed to be of high methodological quality (low risk of bias), 4 to 6 points as moderate quality, and 0 to 3 points as low quality. Any discrepancies between the reviewers were resolved through consensus or consultation with a third reviewer (Zhang).

Statistical analysis

All statistical analyses were performed using Review Manager (RevMan) software, version 5.4 (The Cochrane Collaboration, London, UK). Dichotomous data were analyzed using odds ratios (OR) and 95% confidence intervals (CIs). A fixed-effects model was primarily utilized for data synthesis. Although a random-effects model is typically recommended when substantial heterogeneity (I² > 50%) is present, the limited number of included studies in our analysis (n = 6) could lead to an imprecise estimation of between-study variance (τ²). Therefore, the fixed-effects approach was primarily maintained to prevent disproportionate weight from being assigned to smaller trials, but importantly, a random-effects model was conducted as a pre-specified sensitivity analysis to evaluate the stability of the pooled estimates amidst clinical heterogeneity. The assessment of publication bias via funnel plot inspection and formal statistical testing was initially considered. However, in accordance with the Cochrane Handbook recommendations, this analysis was not performed because the number of included studies (n = 6) was insufficient (fewer than 10) to yield adequately powered and reliable results.

Study-specific effect estimates were first calculated. For dichotomous outcomes, this was expressed as an odds ratio (OR) with 95% CI, computed via the Mantel-Haenszel method. For continuous outcomes, the mean difference (MD) with 95% CI was derived using the inverse-variance method. Subsequently, pooled outcome proportions for each intervention group were estimated by the inverse-variance and logit transformation method. Heterogeneity was evaluated using the Cochran Q test and I² statistic. To explore potential sources of substantial heterogeneity, subgroup analyses were planned based on different EUS-BD approaches or types of gastrointestinal reconstruction. Additionally, sensitivity analyses were performed by sequentially omitting individual studies to assess the stability and reliability of the pooled results.

Results

Study selection and trial characteristics

A total of 528 records were initially identified through the systematic search^{16,17,18,19,20,21}. Following the removal of 37 duplicates, 491 records underwent title and abstract screening, resulting in the exclusion of 234 studies. A full-text assessment was performed on the remaining 257 articles, of which 251 were excluded, leaving exactly 6 studies that ultimately met the inclusion criteria. A total of 578 patients from the six included studies were initially identified (EUS-BD: 203; ERCP: 375)^{15,16,17,18,19,20}. It is important to note that variations in denominators across specific outcome analyses (e.g., 572 patients evaluated for technical success, 443 for clinical success) reflect missing outcome data within the original primary studies rather than selective patient exclusion from the pooled analysis. All patients in the studies were exclusively from East Asia. The etiologies of biliary obstruction across the included studies encompassed a heterogeneous mix of both malignant and benign conditions. While malignant diseases—primarily pancreatic, gastric, gallbladder, and biliary tract cancers—constituted the majority of cases, benign conditions such as common bile duct stones were also evaluated^{15,16,17,18,19,20}. The common bile duct diameter ranged from 8 mm to 14 mm in both groups^15,16,19. The types of gastrointestinal reconstruction surgery include the following: Billroth II gastrojejunostomy, Roux-en-Y reconstruction, pancreaticoduodenectomy, choledochojejunostomy, and Gastrojejunostomy.

Quality assessment of included studies

The methodological quality of the six included retrospective cohort studies was generally high, with NOS scores ranging from 6 to 9 (Table 2). Specifically, all studies demonstrated robust ascertainment of exposure and outcomes, though some variability was noted in the comparability domain due to inherent differences in baseline anatomical reconstructions among the cohorts. No studies were excluded based on the quality assessment, but the identified methodological nuances were strictly factored into the interpretation of the pooled estimates.

Main outcomes

Success-related outcomes

Technical success

Six studies comparing the technical success of EUS-BD (182/201, 90.5%) and ERCP (261/371, 70.4%) were pooled for effect size analysis using a fixed-effects model^{15,16,17,18,19,20}. The odds ratio (OR) was 3.60, with a 95% confidence interval (CI) of (2.12, 6.13). The test for overall effect yielded Z = 4.73, p < 0.00001. These results indicate a significantly higher technical success rate for EUS-BD compared to ERCP, with an absolute risk difference of 20.1% (Figure 1A). Regarding the substantial heterogeneity observed in technical success (I² = 75%), a leave-one-out sensitivity analysis and a random-effects model analysis were performed. The statistical significance, effect magnitude, and confidence intervals remained robust under the random-effects model, indicating that the findings are stable despite the underlying variances in anatomical reconstruction, disease etiology, and specific procedural techniques. A qualitative evaluation suggests that this heterogeneity is predominantly driven by clinical and methodological variances rather than statistical anomalies. Specifically, the included trials encompassed distinct types of gastrointestinal reconstruction (e.g., Roux-en-Y vs. Billroth II) and utilized varying EUS-BD approaches (e.g., hepaticogastrostomy vs. choledochoduodenostomy). Furthermore, the disparate success rates reported in certain studies (e.g., Xu 2022) may reflect the institutional learning curve associated with advanced EUS techniques in extremely complex anatomical variants.

Clinical success

Five studies reported clinical success (443 patients). The odds ratio (OR) was 3.52, with a 95% confidence interval (CI) of (2.07, 5.98). The test for overall effect yielded Z = 4.66, p < 0.00001. The clinical success rate of EUS-BD is higher than that of ERCP (Figure 1B). Compared to the EUS-BD group (25.9%, 52/201), the ERCP-BD group had a lower incidence of adverse events (20.5%, 76/371), as reported across six studies^{15,16,17,18,19,20}. There is no difference in adverse events between EUS-BD and ERCP. (OR, 1.54; 95% CI, 0.97–2.44) (Figure 2A).

Post-procedure pancreatitis

Four studies reported post-procedure pancreatitis (437 patients). Biliary drainage with EUS resulted in no difference in post-procedure pancreatitis compared with ERCP-BD (OR, 0.55; 95% CI, 0.17-1.74; P, 0.73, Z = 1.02) (Figure 2B).

Aspiration pneumonia

Regarding the incidence of aspiration pneumonia, the two studies revealed a difference between the two groups: not a single case occurred among the EUS-BD group, whereas two cases were observed among the ERCP group. The risk for aspiration pneumonia was not significantly different between groups (OR, 0.66; 95% CI, 0.07-6.44; P, 0.72; Z=0.36) (Figure 2C).

Intestinal perforation

Three studies evaluated intestinal perforation (361 patients). The odds ratio (OR) was 0.51, with a 95% confidence interval (CI) of (0.09, 3.10). The test for overall effect yielded Z = 0.73, p =0.47. Based on these analysis results, there is no significant difference between EUS-BD and ERCP (Figure 2D).

Time-related outcomes - procedure time

A pooled analysis of two studies^17,19 demonstrated a significantly shorter total procedure time for EUS-BD compared to the alternative, with a mean reduction of 37.95 min (95% CI, -39.97 to -35.93; Z = 36.82). However, this result warranted extreme caution in interpretation, as the pooled reduction was essentially driven by a single dataset that contributed approximately 98% of the statistical weight. Given the methodological variances in defining procedure time, this finding was considered fragile and could not be broadly generalized to all EUS-BD and ERCP procedures (Figure 3A).

Other outcomes

Bleeding

There was a trend toward a reduced bleeding risk with EUS-BD, although this difference did not reach statistical significance (OR 3.77; 95% CI 0.44-32.13, Z 1.21, p 0.23) (Figure 3B). Four studies reported RBO rates for the EUS-BD and ERCP-BD groups. There was no significant difference between the groups (OR 0.89, 95% CI 0.53-1.49, Z = 0.45, p 0.65) (Figure 3C).

DATA AVAILABILITY:

The raw data supporting the conclusions of this systematic review and meta-analysis, including the comprehensive data extraction sheets and statistical analysis files, have been deposited in the Zenodo public repository and are openly accessible at https://zenodo.org/records/21455497.

Figure 1: Forest plots comparing success rates between EUS-BD and ERCP. (A) Technical success. (B) Clinical success. Please click here to view a larger version of this figure.

Figure 2: Forest plots comparing safety-related outcomes between EUS-BD and ERCP. (A) Overall adverse events. (B) Post-procedure pancreatitis. (C) Aspiration pneumonia. (D) Intestinal perforation. Please click here to view a larger version of this figure.

Figure 3: Forest plots comparing time-related and other clinical outcomes between EUS-BD and ERCP. (A) Procedure time. (B) Bleeding. (C) Recurrent biliary obstruction (RBO). Please click here to view a larger version of this figure.

Figure 4: Schematic illustrations of altered gastrointestinal anatomy and technical challenges during ERCP. The figure was conceptualized and drawn de novo by the authors using Adobe Illustrator. The anatomical configurations and procedural scenarios were developed based on the authors’ first-hand clinical experience and the standard surgical and endoscopic procedures routinely performed by our team. The anatomical structures, endoscope positions, instrument pathways, and technical challenges were manually drawn as vector graphics. The panels were then assembled, labeled, and exported as a PDF using Adobe Illustrator.No copyright concerns are involved. Please click here to view a larger version of this figure.

Author, Year	Country / Design	Operator / Center Experience*	Patients (ERCP / EUS)	Indication (Benign / Malignant)	Surgically-Altered Anatomy	Specific ERCP Technique	Specific EUS-BD Technique	Definition of Technical Success*	Definition of Clinical Success*
Itonaga 2025	Japan / Multicenter	High-volume tertiary centers	98 (54 / 44)	Exclusively Malignant	Roux-en-Y, Billroth II, Choledocho jejunostomy, Gastro jejunostomy	BE-ERCP	HGS, HGS+AG, AG, HJS	Successful deployment of stent/drain	>50% decrease in bilirubin within 14 days
Iwashita 2023	Japan / Multicenter	Experienced endoscopists	119 (96 / 23)	Exclusively Benign (CBD stones)	Roux-en-Y, Billroth II, Choledocho jejunostomy	BE-ERCP	AG	Successful stone clearance	Resolution of symptoms & cholangitis
Xu 2022	China / Single-center	High-volume center	43 (17 / 26)	Exclusively Malignant	Roux-en-Y, Billroth II, Hepatico jejunostomy	Standard/BE-ERCP	Mixed EUS-BD (Not fully specified)	Successful stent placement across stricture	>50% decrease in bilirubin within 2 weeks
Khashab 2016	Multinational / Multicenter	International expert centers	98 (49 / 49)	Mixed (Benign 63 / Malignant 35)	Roux-en-Y, Billroth II, Pancreatico duodenectomy (assumed)	BE-ERCP	HGS, AGS, RV, HJS, HDS	Successful access and stent placement	>50% decrease in bilirubin at 2 weeks
Hakuta 2024	Japan / Single-center	Experienced endoscopists	150 (118 / 32)	Exclusively Malignant	Pancreatico duodenectomy, Gastrectomy, EHBD resection	BE-ERCP	HGS, HJS	Successful stent placement	>50% decrease in bilirubin within 14-28 days
Shibuki 2025	Japan / Single-center	Experienced tertiary center	70 (41 / 29)	Exclusively Malignant	Roux-en-Y, Billroth II, Pancreatico duodenectomy	SBE-ERCP	HGS	Successful deployment of metal/plastic stent	>50% decrease in bilirubin within 2-4 weeks

Table 1: Baseline characteristics and procedural details of the included studies. The table summarizes the specific interventions, patient demographics, underlying surgically altered anatomy, and the precise definitions of technical and clinical success applied across the included trials. Abbreviations: ERCP = endoscopic retrograde cholangiopancreatography; EUS-BD = endoscopic ultrasound-guided biliary drainage; BE-ERCP = balloon enteroscopy-assisted ERCP; SBE-ERCP = single-balloon enteroscopy-assisted ERCP; HGS = hepaticogastrostomy; AG = antegrade intervention; HJS = hepaticojejunostomy; RV = rendezvous; HDS = hepaticoduodenostomy.

Study	Selection (0–4)	Comparability (0–2)	Outcome (0–3)	Total Score
Hakuta 2024	3	2	3	8
Itonaga 2025	4	1	3	8
Iwashita 2023	3	2	2	7
Khashab 2016	4	2	3	9
Shibuki 2025	3	1	3	7
Xu 2022	3	1	2	6

Table 2: Quality assessment of the included observational studies using the Newcastle-Ottawa Scale (NOS). Studies scoring 7 to 9 points were considered to have high methodological quality, indicating a low risk of bias, whereas scores of 4 to 6 indicated moderate quality.

Supplementary Table 1: PubMed database. Detailed search strategy and terms used for the PubMed database, including combinations of Medical Subject Headings (MeSH) and free-text keywords.Please click here to download this file.

Supplementary Table 2: Web of Science. Detailed search strategy and terms used for the Web of Science database, outlining the specific Boolean operators and search strings.Please click here to download this file.

Supplementary Table 3: Embase. Detailed search strategy and terms used for the Embase database, outlining the specific Boolean operators and search strings.Please click here to download this file.

Supplementary Table 4: Cochrane Library. Detailed search strategy and terms used for the Cochrane Library database, outlining the specific Boolean operators and search strings.Please click here to download this file.

Supplementary Table 5: Newcastle Ottawa scale. Detailed risk of bias assessment using the Newcastle-Ottawa scale (NOS) for cohort studies.Please click here to download this file.

Supplementary File 1: Checklist. The updated PRISMA checklist for the current study. Please click here to download this file.

Discussion

In this systematic review and meta-analysis of 6 observational cohort studies involving over 578 patients, success rates, safety profiles, and time-related outcomes between EUS-BD and ERCP-BD were compared for biliary tract obstruction in patients with altered gastrointestinal anatomy. The pooled analyses yielded the following key findings: 1) EUS-BD demonstrated significantly higher technical and clinical success rates, alongside a shorter procedure time, when compared to ERCP; 2) Although EUS-BD showed a numerically higher trend in the incidence of adverse events (25.9%) compared to ERCP (20.5%), this difference did not reach statistical significance (OR 1.54; 95% CI: 0.97–2.44; p = 0.07).

EUS-BD was associated with a significantly higher success rate compared to ERCP. Potential explanations for this observation include the following: First, the successful execution of ERCP is highly dependent on reaching and successfully cannulating the duodenal papilla or bilioenteric anastomosis through a retrograde endoscopic approach^21,22. After the gastrointestinal anatomical structure is changed, such as after Roux-en-Y or Billroth II reconstruction, long loops, acute angles, etc., make it difficult for the endoscope to reach the target site. And EUS-BD uses the "transmural puncture" technique to enter the intrahepatic or extrahepatic bile ducts from the stomach or duodenal bulb^23,24. Skip the navigation difficulties brought about by anatomical variables. This "avoid obstacles" strategy may partially explain the higher technical success rate of EUS-BD in this meta-analysis; moreover, ERCP treatment for such cases often requires the help of endoscopic retrograde cholangiopancreatography²². Even if the target position is reached, intubation may fail due to a poor angle or a lack of support. On the contrary, EUS-BD uses special needles, catheters, and endoscopic ultrasound to select new approaches, making the technical goal clearer and more concentrated.

Technical success—specifically, successful tube placement—is intrinsically linked to clinical success. EUS-BD is associated with a higher likelihood of technical success, which serves as a hypothesized mechanism for its corresponding clinical efficacy. Additionally, the drainage pathway created by EUS-BD—particularly in hepatogastrostomy—may be shorter and more direct. This can make bile drainage more physiological or efficient, thereby lowering bilirubin levels more rapidly^23,24,25. Furthermore, in some cases, the metal stent placed under EUS guidance may conform better to the duct wall, providing more stable anchoring and reducing the risk of early drainage dysfunction²⁶.

During EUS procedures, the need for prolonged and uncertain navigation through complex intestinal anatomy is avoided. This navigation process is typically the most time-consuming step in enteroscopy-assisted ERCP. Once the target bile duct is located via ultrasound, the puncture and stent placement steps are relatively standardized and quick²⁵. It should be noted that most of the included studies come from experienced centers^{11,12,13,14,15,16}. In these settings, EUS-BD is already a well-practiced procedure. On the other hand, difficult ERCP cases may require repeated attempts at cannulation or instrument changes, which can significantly increase procedure time²². Analysis showed that ERCP was correlated with a lower overall adverse event rate and a reduced risk of bleeding compared to ERCP-BD, although this difference was not statistically significant. This may be due to statistical chance: First, the overall sample size was small. Second, the studies varied widely, making them hard to compare directly. Setting aside statistical randomness, this result may stem from the fundamentally different risk profiles of the two techniques due to their core procedural differences in achieving biliary drainage.

From a clinical perspective, the specific adverse event profile of EUS-BD is likely associated with a combination of its technical principles, patient selection criteria, and procedural maturity. EUS-BD is a "route-creating" drainage technique that achieves decompression by establishing a non-physiological fistula through direct puncture²³. Its risk spectrum mainly covers complications such as bile leakage, puncture bleeding, and stent problems, which are different from the risk characteristics of ERCP^24,25. For example, adverse events such as pancreatic and intestinal perforation are frequently associated with the "path tracking" nature of standard ERCP^26,27. EUS-BD is used for "ultra-high-risk" patients who are predicted to have failed ERCP or have extremely complex anatomical structures²⁵. This patient group inherently carries a higher baseline risk of complications^26,27. Furthermore, as an advanced technique still in refinement, the safety of EUS-BD heavily depends on operator experience, and the learning curve may also contribute to a proportion of adverse events²⁸. Consequently, this trend reflects how EUS-BD, while achieving higher technical success rates in addressing "inaccessible" obstructions, does so at the cost of assuming its specific risk profile.

Overall, the "high-risk steps" of the two techniques belong to different categories. However, in the studies included in this analysis, relevant measures were taken—such as covered stents utilization in EUS-BD and prophylactic pancreatic duct stenting in ERCP—to keep their respective main risks at similarly low levels, which may explain why no statistically significant difference was found in overall incidence. Bleeding related to ERCP-BD primarily results from therapeutic sphincterotomy or dilation. This is an active and invasive intervention, and bleeding is one of its main and common complications^26,27. In contrast, the bleeding risk associated with EUS-BD is mainly related to vascular injury along the puncture path. Under real-time ultrasound guidance, the operator can actively select and avoid important vessels in the puncture route, thereby minimizing the risk of bleeding in advance²⁴. Therefore, bleeding in EUS-BD is more of a potential risk that can be prevented through careful technique, rather than an inevitable outcome of the procedure itself.

It is crucial to recognize that the included studies present a confluence of distinct clinical scenarios, notably the admixture of malignant obstructions and benign etiologies (e.g., hepatolithiasis). This distinction is fundamentally important; tumor infiltration often exacerbates anatomical distortion and increases tissue friability, which can intrinsically skew complication profiles and success rates compared to benign stone disease. Crucially, it must be acknowledged that the pooled effect in this analysis represents a comparison of heterogeneous procedural strategies rather than a direct comparison between two uniformly standardized interventions. The experimental and control groups encompass fundamentally different techniques that do not share identical indications, technical endpoints, or adverse-event profiles. Furthermore, the EUS-BD interventions analyzed involved various discrete techniques, including hepaticogastrostomy (HGS), choledochoduodenostomy (CDS), and antegrade approaches. Each modality carries a unique risk-benefit ratio—for instance, HGS might inherently present a different risk profile for stent migration or bile leakage compared to antegrade stone clearance. Because the included studies aggregate malignant biliary obstructions with benign conditions (e.g., hepatolithiasis), which differ fundamentally in treatment objectives, technical endpoints, risk of recurrent obstruction, and expected survival, this etiology admixture represents a major limitation that substantially restricts the clinical applicability of the pooled estimates. Consequently, the findings should not be interpreted as a generalized treatment preference for all patients with surgically altered anatomy; rather, the selection of the drainage strategy must be meticulously individualized based on the underlying etiology.

This study suggests that for biliary obstruction in patients with surgically modified gastrointestinal anatomy, EUS-BD’s observed advantages in rates of technical and clinical success may help improve the likelihood of success on the first treatment attempt²⁹. Clinically, this could mean reducing the need for repeat procedures, multiple anesthesia sessions, and treatment delays that patients might otherwise face after unsuccessful ERCP attempts. For patients with malignant obstruction, this relatively more direct and faster biliary drainage pathway may create more favorable conditions for subsequent anti-tumor therapy³⁰. And the short operation time of EUS-BD can improve the workflow and reduce the intraoperative risk of patients³¹.

Studies have found that, for patients with complex anatomical structures, the treatment strategy may favor EUS-BD over simple ERCP, thereby promoting an individualized, anatomically adapted treatment path. From a practical clinical perspective, EUS-BD is an effective and highly promising alternative drainage strategy for patients with a clear history of gastrointestinal reconstruction. As far as is known, this is the first meta-analysis comparing endoscopic ultrasound biliary drainage (EUS-BD) and endoscopic retrograde cholangiopancreatography biliary drainage (ERCP-BD) for biliary obstruction after surgical alteration of gastrointestinal anatomy. Previous meta-analyses relied solely on observational data and did not specifically evaluate the impact of anatomical changes on the clinical efficacy and safety of EUS-BD. And outcomes such as aspiration pneumonia and intestinal perforation were not evaluated in previous comparisons.

The present meta-analysis has several limitations. First, the supplementary search strategies may have had suboptimal sensitivity; the inadvertent inclusion of restrictive filters in certain databases could have excluded relevant non-randomized comparative studies. Second, while the overall methodological quality of the included studies was acceptable per the NOS evaluation, their inherent retrospective observational design introduces an unavoidable risk of selection bias and unmeasured confounding. Critical residual confounders include selecting EUS-BD following anticipated or prior ERCP difficulties, disparities in institutional expertise, profound differences in baseline anatomy, and variations in deployed stents. Detailed risk-of-bias assessments supporting the individual NOS judgments have been provided in the Supplementary Table 5 to address this concern. This limitation necessitates that the conclusions regarding the superiority of EUS-BD be interpreted with appropriate circumspection, as the lack of randomization may inflate the observed effect sizes. Furthermore, the lack of standardized definitions of clinical success across the primary studies—with slight variations in the required bilirubin reduction and evaluation timeframes—may introduce information bias and contribute to the observed heterogeneity in the pooled estimates. Additionally, the number of included studies is small, with only 572 cases, which may reduce statistical power and fail to detect subtle differences in outcomes, such as post-operative pancreatitis or recurrent biliary obstruction. Therefore, caution is needed when interpreting results that lack statistical significance. Second, the limited number of included studies restricted the ability to conduct robust quantitative subgroup and sensitivity analyses. Consequently, despite substantial heterogeneity in technical success (I² = 75%), it was necessary to report the fixed-effects model estimates to avoid erratic between-study variance weightings. Thus, the pooled results of this study should be considered as an average effect estimate across different clinical scenarios. Their applicability to specific clinical situations requires careful judgment based on individual patient factors. Finally, this meta-analysis did not include any randomized controlled trials (RCTs). Instead, it synthesized the currently available evidence from observational cohort studies, providing important preliminary data to inform treatment decisions.

Given the lack of randomized controlled trials in this field, future research using higher-level study designs will be essential to confirm and refine the current findings. Finally, the limited sample size inherently restricted the ability to perform robust quantitative subgroup analyses. Consequently, it was not possible to statistically stratify clinical outcomes and complication profiles by obstruction etiology (malignant versus benign) or by the specific technical approach (e.g., hepaticogastrostomy versus antegrade interventions). Future multi-center prospective studies with larger cohorts are essential to facilitate these critical subgroup evaluations. In conclusion, while this meta-analysis suggests that EUS-BD yields superior technical and clinical success rates compared to ERCP in patients with surgically altered anatomy, this conclusion must be interpreted with caution due to the limited number of studies, substantial heterogeneity, and the exclusively observational nature of the current evidence. These findings await validation by future well-designed randomized controlled trials.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Cochrane Library	Wiley	URL: https://www.cochranelibrary.com; Database accessed up to Jan 1, 2026	Literature search
Embase	Elsevier	URL: https://www.embase.com; Database accessed up to Jan 1, 2026	Literature search
EndNote	Clarivate	URL: https://endnote.com; Version 21	Reference and bibliography management
PROSPERO	Centre for Reviews and Dissemination, University of York	URL: https://www.crd.york.ac.uk/prospero; Identifier: CRD420261297962	Protocol registration
PubMed	U.S. National Library of Medicine	URL: https://pubmed.ncbi.nlm.nih.gov; Database accessed up to Jan 1, 2026	Literature search
Review Manager (RevMan)	The Cochrane Collaboration	URL: https://revman.cochrane.org; Version 5.4	Meta-analysis and forest plot
Web of Science	Clarivate	URL: https://www.webofscience.com; Database accessed up to Jan 1, 2026	Literature search