Introduction
EUS-guided biliary drainage (EUS-BD) is one of the possible treatments for the relief of jaundice caused by malignant biliary obstruction (MBO) when endoscopic retrograde cholangiopancreatography (ERCP) has failed [1 ]. If the common bile duct cannot be cannulated, or the papilla is infiltrated or unreachable, EUS-BD provides a valuable alternative to percutaneous transhepatic biliary drainage (PTBD) [2 ]. The available evidence shows that EUS-BD is comparable with PTBD in terms of efficacy, although PTBD is burdened by a higher rate of re-intervention and a lower quality of life [3 ].
EUS-guided hepaticogastrostomy (EUS-HGS) was first described in 2003 [4 ]
[5 ]; however, since its introduction, its use in clinical practice has been intermittent, slowly becoming more widespread in the last 10 years. The main reasons for this are: its technical complexity, requiring high technical skills in both ERCP and interventional EUS; the far from negligible rate of AEs that are often severe and require additional modalities (e.g. interventional radiology or surgery); and, for a long time, the lack of dedicated devices [6 ]. In more recent years, different types of stents have been designed in order to make this procedure safer and to overcome the limitations of the previously available devices. These stents are usually self-expandable metal stents (SEMSs), with a longer fully covered portion and an uncovered extremity for the intrahepatic side, and are provided with variable anti-migration systems [7 ]
[8 ]. This step forward in tools and the paradigm change in biliopancreatic endoscopy have led to recent widespread use of the technique.
Despite the growing evidence on EUS-HGS, to date, the available meta-analyses have mainly focused on EUS-BD, including both EUS-HGS and EUS-choledochoduodenostomy (EUS-CDS), with comparative analysis among the two techniques [1 ]
[9 ]. No meta-analysis has previously been performed to consider EUS-HGS alone and, as remarkable heterogeneity exists among studies, the outcomes of this technique may not be completely reliable.
We performed a systematic review with meta-analysis, aiming to evaluate the efficacy and safety of the EUS-HGS procedure. The main outcome was the rate of technical success; secondary outcomes were the rates of clinical success and adverse events (AEs). In addition, we aimed to identify potential modifiers of the efficacy and safety of EUS-HGS through meta-regression analysis [10 ].
Methods
Literature search strategy
A systematic search was made through Pubmed, Scopus, Web of Science and Cochrane databases until December 2022 that were relevant to EUS-HGS, without restriction criterion for the starting date. No search for grey literature was attempted. The meta-analysis was conducted and reported in accordance with the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) [11 ] and Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines [12 ]. The native search for each database syntax is given in Appendix 1s (see online-only Supplementary material).
Literature screening
One author (E.D.) conducted an initial screen to obtain a single list of articles from the queried databases. Once the list of articles had been obtained, the same author identified articles that were irrelevant by the title or abstract. Study selection was then accomplished through three levels of screening. First, reviews, letters, expert opinions, and editorials were excluded. Second, the abstracts of the retained studies were reviewed by two independent reviewers (C.B. and P.G.) and studies that reported the safety and efficacy of patients undergoing EUS-HGS as the only study population or in comparison to other treatments (i.e. percutaneous drainage or other routes for EUS-BD) entered the subsequent screening level. For the final level of screening, the full text was obtained for any relevant articles and for those articles where a decision could not be made on the basis of the abstract. The reference lists of the selected articles were checked for additional studies by another investigator (C.C.).
Inclusion and exclusion criteria
The final inclusion criteria were: (i) a study population that included patients treated with EUS-HGS for the treatment of biliary obstruction whether malignant or benign; (ii) adequate description of the study population; (iii) a complete description of the technical and clinical success rates, together with description of any post-procedural AEs. The following studies were excluded: (i) small cases series with <10 patients; (ii) studies including antegrade stenting; (iii) studies that did not provide sufficient data to evaluate the primary outcome. Where a study was followed by a more complete study or studies that included the original data set, the most recent, the largest, or the most complete report was used for the analysis. Such linked studies were identified on the grounds of authorship, institutions, design, length of follow-up, and study populations. Any divergences were resolved by discussion between the reviewers and a third investigator (A.C.).
Data extraction and quality assessment
The extracted data included study period, design, and number of centers involved, study location, patient demographics, clinical characteristics, route of drainage, and the type of stent used. The number of procedures/year was calculated by dividing the total number of cases treated at each center by the study period in years. Technical success, clinical success, AEs, procedure-related mortality, 30-day all-cause mortality, hospital stay, re-intervention rate, and time to recurrence were all evaluated. Clinical success was evaluated both on an intention-to-treat (ITT) approach (the denominator being all attempted EUS-HGSs) and per successful procedure (the denominator being only patients where technical success was achieved). Prespecified AEs included only the following procedure-related complications: pneumoperitoneum, biliary leakage/peritonitis, stent malfunction/migration, perforation, bleeding, and cholangitis/sepsis. The quality of each selected study was assessed by two investigators (P.G. and C.C.) through the Cochrane tool (RoB-2) for randomized controlled trials (RCTs) [13 ] and the Newcastle–Ottawa scale (NOS) for observational studies [14 ]. Any divergences were resolved by discussion between reviewers and a third investigator (A.C.).
Statistical analysis
The main outcome was the technical success of the procedure. Secondary outcomes were the rates of clinical success and procedure-related AEs. Dichotomous variables were estimated as pooled binomial proportions with 95%CIs applying random-effect models and the Freeman-Tukey double arcsine transformation to retain studies with proportions at 0 or 1 margins and ensuring admissible confidence intervals for the pooled proportions [15 ]. Continuous variables were pooled in weighted means with 95%CI, using random-effect models. When studies reported continuous variables as median and range or interquartile range (IQR), the mean and variance were estimated as proposed by Wan et al. [16 ]. Studies were not weighted for their quality.
Statistical heterogeneity was explored by Cochrane’s Q and inconsistency (I
2 ) statistics [17 ]. Publication bias was evaluated with Egger’s and Begg’s tests [18 ]
[19 ]. The meta-regression analysis included the following prespecified covariates: study characteristics (year, design, center volume); patient demographics (age, sex); proportion of patients with malignancy; indication for HGS (proportion of patients with surgically altered anatomy/hilar stenosis/duodenal invasion); presence of ascites; and procedure details (use of dedicated stents, transgastric route).
We performed the following post-hoc sensitivity analyses: by geographic reason, by study year (before vs. after 2015), and for subgroups of patients using exclusively metal stents, exclusively dedicated stents, or a transgastric approach. Meta-analysis was performed using Stata (StataCorp. 2017. Stata Statistical Software: release 15. College Station, Texas: StataCorp LLC) and the package “metafor” for R-Project 3.2.5 (R Core Team [2016], Vienna, Austria; available from: http://www.R-project.org ).
Results
Literature search
A total of 1460 articles were identified after the removal of duplicate records and, after assessment against the exclusion criteria, 1351 articles were rejected ([Fig. 1 ]). The remaining 109 studies were assessed for eligibility, of which 76 were not included in the final analysis: in particular 14 studies were excluded because they were subsequently enlarged or included in larger multi-institutional analyses, 14 studies because they included <10 patients, and five studies because they combined HGS with antegrade stenting. Ultimately 33 studies, with 1644 patients, were included for the final analysis (Table 1s , see online-only Supplementary material, including Supplementary References). Among these, 13 were prospective studies, including five RCTs, and nine were multicenter studies. The overall quality of the included studies was deemed to be sufficient.
Fig. 1 Preferred reporting items for systematic reviews and meta-analyses (PRISMA) flowchart for study identification, selection, and inclusion [12 ].
Characteristics of included patients
From the 33 included studies, the mean age of the included patients varied from 57 [20 ] to 76 years [21 ] and the percentage of male patients from 40% [20 ] to 80% [22 ]. The cause of biliary obstruction was exclusively malignancy in 24 studies; in the remaining studies, malignant obstructions ranged from 60% [22 ] to 95% [23 ]. The proportion of patients with duodenal/papillary invasion and surgically altered anatomy ranged from 0% to 91% [20 ]
[24 ] and 0% [20 ]
[25 ]
[26 ]
[27 ] to 70% [28 ], respectively. The rate of patients with hilar stenosis was highly variable (0% [27 ]
[29 ]
[30 ]
[31 ]
[32 ]
[33 ]
[34 ] to 100% [20 ]
[25 ]
[26 ]
[35 ]
[36 ]). The transgastric approach was used exclusively in 24 studies, whereas the use of this approach ranged from 80% [21 ] to 97% [37 ] in the other studies; one study used only the transduodenal approach [38 ]. Pooled analysis of the clinical characteristics from the included studies is reported in [Table 1 ].
Table 1 Pooled analysis of the clinical features of the study populations submitted to endoscopic ultrasound-guided hepaticogastrostomy (EUS-HGS).
Number of studies
Weighted analysis (95%CI)
I2 , %*
Cochrane’s Q; Wald-type/LR
* I
2 describes the amount of heterogeneity resulting from meta-analysis. A rough guide to interpretation is as follows: 0%-40%, might not be important; 30%–60%, may represent moderate heterogeneity; 50%–90%, may represent substantial heterogeneity; 75%–100%, considerable heterogeneity.
Clinical features
Age, years
33
66.4 (65.0–67.9)
89.6
306.3
Sex, male, %
32
57.9 (54.4–61.3)
29.8
46.6; 46.6
Malignant biliary obstruction
33
99.6 (97.5–99.9)
53.7
69.11; 406
Indication for EUS-HGS, %
30
18.4 (11.3–28.6)
79.9
144.2; 302.2
25
16.0 (2.3–60.5)
76.1
100.4; 484.4
15
1.95 (1.93–1.98)
25.4
18.8; 213.1
24
34.8 (24.3–47.0)
80.3
116.7; 195.6
Ascites present, %
10
20.2 (14.3–27.6)
73.5
34.0; 39.6
Procedure features
Type of stent used, %
33
99.9 (96.3–100)
83.3
191.2; 1113.3
33
8.3 (7.1–9.8)
0
31.2; 765.9
31
42.6 (40.2–45.0)
36.6
47.3; 1826.2
Transgastric puncture/drainage, %
32
99.9 (98.2–99.9)
0
6.8; 334.9
Meta-analyses of clinical outcomes
When evaluating the primary outcome, EUS-HGS showed a summary technical success of 97.7% (95%CI 96.1%–99.0%) ([Fig. 2 ]), with no heterogeneity among studies (I
2 = 0%). The summary ITT and per-procedure clinical success rates were 88.1% (95%CI 84.7%–91.2%; I
2 = 33.9%) and 91.0% (95%CI 88.2%–93.3%; I
2 = 16.7%), respectively, with moderate-to-low heterogeneity among studies ([Fig. 3 ]). Procedure-related AEs occurred at a summary rate of 12.0% (95%CI 9.8%–14.5%; I
2 = 20.4%) ([Fig. 4 ]), with cholangitis/sepsis 2.8% (95%CI 1.5%–5.2%), bleeding 2.3% (95%CI 1.6%–3.3%), biliary leakage 1.6% (95%CI 0.9%–3.0%), and pneumoperitoneum 1.6% (95%CI 0.8%–3.0%) the most common ([Table 2 ]). Procedure-related mortality was practically nil. The summary in-hospital stay was 8.2 days (95%CI 5.3–11.0). Recurrence of biliary obstruction during follow-up occurred at a summary rate of 16.2% (95%CI 11.8%–21.8%), after a weighted mean time of 165 days (95%CI 115–195).
Fig. 2 Forrest plots of meta-analysis for the technical success of endoscopic
ultrasound-guided hepaticogastrostomy (EUS-HGS).
Fig. 3 Forrest plots of meta-analysis for the clinical success of EUS-HGS.
Fig. 4 Forrest plots of meta-analysis for adverse events of EUS-HGS.
Table 2 Pooled analysis of clinical outcomes across the study population submitted to endoscopic ultrasound-guided hepaticogastrostomy.
Outcome
Number of studies
Weighted analysis (95%CI)
I
2 ,%*
Cochrane’s Q
* I
2 describes the amount of heterogeneity resulting from meta-analysis. A rough guide to interpretation is: 0%–40%, might not be important; 30%–60%, may represent moderate heterogeneity; 50%–90%, may represent substantial heterogeneity; 75%–100%, considerable heterogeneity.
Technical success, %
33
97.7 (96.1–99.0)
0
13.4
Clinical success, %
33
88.1 (84.7–91.2)
33.9
48.4
33
91.0 (88.2–93.3)
16.0
38.1
Adverse events, %
33
17.5 (14.7–20.8)
30.8
46.2
33
12.0 (9.8–14.5)
20.4
40.2
33
1.6 (0.8–3.0)
0
17.2
33
1.6 (0.9–3.0)
0
16.3
33
1.3 (0.5–3.2)
0
23.8
33
0.1 (0.01–1.3)
0
1.54
33
2.3 (1.6–3.3)
0
14.8
33
2.8 (1.5–5.2)
0
23.8
Procedure-related mortality, %
31
0.3 (0.06–1.7)
0
1.6
Hospital stay, days
9
8.2 (5.3–11.0)
97.5
325.9
Any re-intervention, %
28
16.2 (11.8–21.8)
65.4
78.1
Time to re-intervention, days
14
165 (115–195)
98.3
779.5
Potential modifiers of the primary and secondary outcomes
Results from univariable meta-regression of studies and patients’ features are reported in [Table 3 ]. On univariable meta-regression analysis, technical success was improved when centers had more experience (>4 cases/year; OR 2.12, 95%CI 1.23–3.67; P = 0.007), in the presence of duodenal invasion (OR 6.56, 95%CI 1.18–36.4; P = 0.03), and when dedicated HGS stents were used (OR 2.22, 95%CI 1.24–3.99; P = 0.007). There was no evidence of publication bias for the primary outcome (Egger’s test, P = 0.91; Begg’s test, P = 0.82).
Table 3 Results from univariable meta-regression of the main outcomes considered.
Odds ratio (95%CI)
Technical success
Clinical success on ITT analysis
Procedure-related adverse events
ITT, intention to treat; RCT, randomized controlled trial. Results in bold indicate significant results (P < 0.05).
Study characteristics
1.05 (0.97–1.13)
1.06 (0.98–1.14)
1.00 (0.94–1.06)
1.02 (0.6–1.75)
1.06 (0.65–1.72)
0.93 (0.61–1.41)
Center characteristics
2.12 (1.23–3.67)
1.47 (1.01–2.13)
0.74 (0.45–1.24)
Patient characteristics
3.13 (0.16–61.9)
1.27 (0.10–15.8)
0.11 (0.01–1.02)
1.03 (0.96–1.10)
1.03 (0.97–1.09)
1.00 (0.96–1.05)
0.88 (0.14–5.49)
0.88 (0.11–6.84)
1.11 (0.26–4.71)
0.77 (0.20–2.88)
0.69 (0.20–2.44)
0.82 (0.29–2.37)
0.75 (0.27–2.08)
0.71 (0.30–1.69)
1.30 (0.64–2.64)
6.56 (1.18–36.4)
2.62 (0.75–9.09)
0.50 (0.18–1.36)
19.2 (0.06–6654)
0.63 (0.02–25.4)
0.08 (0.01–1.11)
Procedure details
2.22 (1.24–3.99)
1.31 (0.79–2.19)
0.62 (0.41–0.95)
0.58 (0.09–3.86)
1.28 (0.43–3.85)
1.00 (0.94–1.06)
Regarding ITT clinical success, the only potential modifier identified was center experience (>4 cases/year), which was associated with an increased chance of clinical success (OR 1.47, 95%CI 1.01–2.13; P = 0.04); this variable accounted for 11.5% of the heterogeneity found. No evidence of publication bias for this secondary outcome was found (Egger’s test, P = 0.87; Begg’s test, P = 0.13).
Finally, in terms of the rate of procedure-related AEs, the use of dedicated stents was the only variable associated with a reduced risk of AEs (OR 0.62, 95%CI 0.41–0.95; P = 0.03) on univariable meta-regression analysis. There was no evidence of publication bias for this secondary outcome (Egger’s test, P = 0.49; Begg’s test, P = 0.84).
Post-hoc sensitivity analysis
The post-hoc sensitivity analysis according to geographical area showed substantially overlapping results between studies performed in Asia, Europe, and North America ([Table 4 ]). Of note, the number of studies conducted in North America was limited (n = 3).
Table 4 Results of the post-hoc sensitivity analyses.
Number of studies
Percentage rate (95%CI)
Technical success
Clinical success
Procedure-related adverse events
Overall
33
97.7 (96.1–99.0)
88.1 (84.7–91.2)
12.0 (9.8–14.5)
Region
19
97.4 (95.0–98.7)
90.5 (86.1–93.4)
10.7 (7.9–14.2)
10
98.4 (94.3–99.6)
85.4 (78.4–90.4)
14.4 (10.6–19.3)
3
94.4 (91.1–96.5)
84.0 (77.0–89.2)
18.6 (14.6–23.4)
Study year
14
94.6 (92.0–96.3)
83.2 (76.9–88.0)
13.1 (10.3–16.6)
19
98.2 (95.7–99.2)
89.7 (86.0–92.5)
12.3 (9.5–15.9)
Procedure details
25
98.2 (96.7–99.3)
87.9 (84.5–90.9)
12.1 (9.2–15.3)
11
99.1 (97.3–100)
88.0 (81.8–93.1)
10.7 (6.6–15.4)
24
97.3 (95.5–98.8)
86.7 (83.0–90.1)
12.3 (9.8–14.9)
More importantly, we performed a post-hoc sensitivity analysis separately evaluating studies where EUS-HGS was performed after 2015 vs. before 2015 and found that the rate of technical success was significantly higher in the former group (98.2% [95%CI 95.7%–99.2%] vs. 94.6% [95%CI 92.0%–96.3%]; P = 0.02), as was the rate of clinical success (89.7% [95%CI 86.0%–92.5%] vs. 83.2% [95%CI 76.9%–88.0%]; P = 0.04).
The post-hoc sensitivity analysis for studies using only metal stents or only a transgastric approach showed efficacy and safety overlapping with the overall estimates. When considering only dedicated stents, the rates of technical and clinical success were 99.1% (95%CI 97.3%–100%) and 88.0% (95%CI 81.8%–93.1%), respectively, while the complication rate was 10.7% (95%CI 6.6%–15.4%).
Discussion
Since its introduction in 2003 [4 ]
[5 ], EUS-HGS has been increasingly used for the relief of jaundice in patients affected by both distal and hilar MBO [6 ]
[26 ]. We performed a meta-analysis to specifically investigate the efficacy and safety of EUS-HGS, followed by a meta-regression analysis in order to overcome the heterogeneity among the published studies and to identify potential modifiers of EUS-HGS outcomes. Our analysis showed that EUS-HGS has a high technical success rate 97.7% (95%CI 96.1%–99.0%; I
2 = 0%).
Although this technique is currently considered very cumbersome, requiring a steep learning curve, the recent development of dedicated devices and improvement in procedural steps have allowed higher technical success rates to be achieved over the years[1 ]
[29 ]. Indeed, when performed in expert hands, some technical tricks have been described to improve the success of the procedure: (i) dilation of the intrahepatic duct >5 mm with an interposed portion of hepatic parenchyma ≤3 cm is advisable; (ii) puncture of liver segment S3 is preferable over S2, because of the greater distance from the gastroesophageal junction, although puncture of S2 may allow an easier stent deployment; (iii) bile aspiration after intrahepatic puncture has been reported to reduce the rate of post-procedural AEs, such fever and abdominal pain[30 ]. Moreover, the choice of guidewire and its manipulation are crucial points, requiring a flexible tip to gain access and advance into the biliary tree, and adequate stiffness in order to easily advance the stent [39 ]. Last, but not least, in order to minimize the risk of stent misdeployment, the intrachannel release technique is recommended [40 ]
[41 ].
Although high, the rate of clinical success was suboptimal, being 88.1% (95%CI 84.7%–91.2%; I
2 = 33.9%) and 91.0% (95%CI 88.2%–93.3%; I
2 = 16.7%) in the ITT and per-procedure analyses, respectively. When compared with other EUS-BD modalities, no significant differences in clinical success have been reported [42 ]. Regardless of the technique used, the achievement of clinical success may be affected by several patient and disease features, such as the presence of liver metastases and a complex hilar biliary stricture, although to date no clear correlations have been defined. As for other EUS-BD modalities [43 ], the future challenge will be to better define those patients who will substantially benefit from this type of treatment.
In our study, the overall rate of procedure-related AEs was 12.0% (95%CI 9.8%–14.5%; I
2 = 20.4%), mainly represented by cholangitis or sepsis, bleeding, biliary leakage, and pneumoperitoneum. Although not negligible, the overall rate of AEs is lower than for PTBD [3 ], while it is similar when compared with other EUS-BD techniques [42 ]. Indeed, despite high technical and clinical success rates, previous studies and meta-analyses have reported that PTBD is associated with higher rates of AEs and need for re-intervention than EUS-BD, mainly owing to an increased risk of stent dysfunction and stent dislodgement [3 ]. Moreover, the presence of an external catheter has been associated with abdominal pain at the insertion site and with lower quality of life [44 ].
When comparing EUS-CDS and EUS-HGS, there is actually no evidence of the superiority of one over the other in terms of technical success, clinical success, or AEs [45 ]; however, although the overall rate of complications is comparable, differences have been reported in the types of AEs. In a meta-analysis by Uemura and colleagues, higher rates of cholangitis were shown for EUS-CDS than for EUS-HGS (31% vs. 10% of AEs, respectively) [46 ]; however, other studies have not confirmed the same results [42 ]. In our analysis, the rate of cholangitis was relatively low at 2.8% (95%CI 1.5%–5.2%), comparable with previous studies and to the rate reported with ERCP [45 ].
Minimizing the rate of cholangitis is particularly important as it strongly impacts on oncologic outcomes and patient survival [47 ]. Stent patency is therefore crucial, generally defined as the time from stent deployment to stent dysfunction, and usually indicated by the time to re-intervention. Because the stent is generally far away from the MBO, EUS-HGSs have been assumed to have a lower risk of obstruction from tumor ingrowth/overgrowth; however, obstruction by clogging, by food, or by reactive tissue has been described [48 ]. In our analysis, the pooled incidence of recurrence of biliary obstruction was 16.2% (95%CI 11.8%–21.8%), after a weighted mean time of 165 days (95%CI 115–195). Our data are slightly lower than those reported in a recent meta-analysis, in which the rate of re-intervention for EUS-HGS was 20.9% (95%CI 16.3%–25.6%), being higher than that reported for EUS-CDS (15.8%, 95%CI 12.2%–19.5%) [42 ]. A subgroup meta-analysis from Mao et al. considering only fully covered SEMSs for EUS-HGS showed no significant difference in stent obstruction and re-intervention compared with EUS-CDS (OR 0.25, 95%CI 0.25–1.34; P = 0.11; I
2 = 0%) [49 ]. According to recent data, an intragastric portion of the stent with a length of >3 cm and chemotherapy seem to be associated with a longer median stent patency [39 ].
The procedure-related mortality rate is our study was close to zero. This result is in line with the meta-analysis by Giri et al., which reported a pooled mortality incidence related to EUS-HGS of 0.2% (95%CI 0.0–0.5%) and of 0.1% (95%CI 0.0–0.4%) related to EUS-BD overall [42 ].
Interestingly, on meta-regression analysis, center experience (>4 cases/year) was found to be an independent modifier of both technical and clinical success. This threshold could be considered low for such a challenging technique; however, published data come mainly from referral centers, with highly skilled endoscopists. Moreover, over the years, the advancement of technical skills and the introduction of dedicated devices may have positively affected both the learning curve and patient selection. According to Oh et al., the procedure time shortened and the AE rate decreased after at least 24 overall procedures had been performed, reaching a plateau after 33 procedures [50 ]. In contrast, a later study reported a number of approximately 40 procedures to gain adequate experience in EUS-HGS [51 ]. All these data highlight how the learning curve to achieve proper expertise may be long and needs to be maintained over the years. To achieve this, centralization of such cases to expert hands in referral centers is mandatory, in order to optimize outcomes.
Duodenal invasion was also associated with higher technical success. This may be because duodenal obstruction is usually an expression of distal MBO. Although to date there have been no studies comparing the outcomes of EUS-HGS in patients with distal versus hilar MBO, it is reasonable to believe that the latter may be technically more challenging. In patients with distal MBO, EUS-CDS may be a valuable alternative for the relief of jaundice; however, emerging evidence supports the use of EUS-HGS in patients with combined distal MBO and gastric outlet obstruction, owing to the lower stent dysfunction rate of EUS-HGS compared with EUS-CDS; indeed duodenal invasion may be predictive of EUS-CDS dysfunction [52 ].
Finally, on meta-regression analysis, the use of dedicated stents was found to positively affect technical success and was the only variable associated with a reduced risk of procedure-related AEs (OR 0.62, 95%CI 0.41–0.95; P = 0.03). Dedicated stents are partially covered metal stents, with an uncovered portion of variable length for intraductal drainage, and a fully covered part for the extrahepatic and intragastric portion; in addition, importantly, these devices are fitted with antimigration systems (i.e. anchoring flaps, asymmetric flared shape) [22 ]. Such hybrid stents aim to overcome the main limitations of uncovered and fully covered SEMSs, such as the risk of bile leakage, intrahepatic bile duct obstruction, and migration. On a post-hoc sensitivity analysis, the results when considering only dedicated stents were extremely satisfying, being 99.1% and 88.0% for technical and clinical success, respectively, while the complication rate was 10.7%.
Newly designed plastic stents specifically designed for EUS-HGS have also now been developed, which showing promising results, representing a valuable alternative to metal stents, especially when EUS-HGS is performed for benign indications [53 ]
[54 ]. Our meta-regression did not show any superiority of metal stents over plastic ones. This may be due to the fact that the mean follow-up of the included studies was short (<6 months), affecting the reliability of this result on long-term outcomes.
Our post-hoc sensitivity analysis confirmed that the outcomes of EUS-HGS were superior for studies published after 2015, a landmark year for standardization of the technique and the availability of dedicated devices. Therefore, in the near future, the use of these dedicated devices should be recommended, as they increase the safety of this challenging procedure. As happened in recent years with lumen-apposing metal stents (LAMSs), the development of systems that may reduce the number of steps in the procedure (e.g. the integrated cystotome) is desirable.
The present study has some limitations. First, most of the studies were retrospective or non-randomized, which could lead to a lack of information regarding the stent design and technical issues, such as the length of the intragastric portion, and to selection bias. Moreover, the short follow-up period does not allow conclusions to be drawn on long-term outcomes, such as stent patency, this being beyond the aim of the current analysis. Second, some of the main results, such as the use of metal stents during EUS-HGS, the re-intervention rate, and the time to re-intervention, showed significant heterogeneity. Clinical success on ITT analysis and total AEs also showed moderate heterogeneity. The definitions of outcomes were variable among the studies and some issues, such as the use of chemotherapy and overall patient survival, were not included, thereby limiting the interpretation of the results as to the real efficacy of EUS-HGS. Finally, our meta-regression analysis was limited to those variables that have a plausibility to influence the clinical outcomes and had been commonly reported among most of the studies; however, it is likely that other known and unknown factors can influence the efficacy and safety of HGS and these aspects should be addressed by future prospective studies.
In conclusion, where ERCP and EUS-CDS are not feasible, EUS-HGS is a safe and effective procedure when performed in an expert setting. The development and spread of dedicated devices will positively affect the safety of the procedure going forward.