Introduction
Gastric outlet obstruction (GOO) is a clinical condition caused by a mechanical malignant
blockage of the upper digestive tract at the level of the distal stomach, pylorus
or duodenum. Often encountered in the context of advanced malignancy, it is associated
with debilitating symptoms including intractable nausea and vomiting, inability to
tolerate oral nutrition, abdominal pain and decreased quality of life [1 ]. In addition, these symptoms contribute in large part to malnutrition and poor functional
status in this fragile patient population, which can lead to increased hospitalizations
and delays in proposed chemotherapy treatments. The traditional treatment modality
for this condition is surgical gastrojejunostomy (SGJ) which bypasses the obstruction.
While this method is highly effective, it is invasive and, in turn, can be associated
with high rates of morbidity [2 ]. Endoscopic stenting (ES) provides a less-invasive approach that is associated with
lower risk of adverse events (AEs) and better short-term outcomes including shorter
hospital length of stay [3 ]; however, it is associated with a significant risk of stent obstruction and increased
need for reintervention [4 ].
Endoscopic ultrasound-guided gastroenterostomy (EUS-GE) is a novel modality that aims
to endoscopically bypass the obstruction by connecting stomach to small bowel downstream
from the pathology with a lumen-apposing metal stent (LAMS). Given its endoscopic
approach, it may avoid the substantial morbidity of the surgical alternative while
at the same time providing the durability of a complete enteral bypass. Early data
suggest good efficacy and safety outcomes, yet comparative data contrasting EUS-GE
to traditional modalities have been limited by small sample sizes [5 ].
We, therefore, conducted a systematic review and meta-analysis assessing the efficacy
and safety of EUS-GE compared to both ES and SGJ for the treatment of malignant GOO.
Methods
This study protocol was prospectively registered with the PROSPERO international database
(CRD42021265074). The Preferred Reporting Items for Systematic Reviews and Meta-Analysis
(PRISMA) and Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines
were followed (Supplementary Table 1, Supplementary Table 2) [6 ]
[7 ].
Search strategy
The literature was systematically searched for studies that assessed EUS-GE for the
treatment of GOO due to malignancy. MEDLINE, EMBASE and Web of Science databases were
searched from inception through February 2022 using the following keywords: 1) endoscopic
or EUS; 2) gastrojejunostomy or gastroenterostomy (see Search Strategy in Supplementary
Table 3). Previously published reviews on the topic were hand searched and the references
of included articles were checked for relevant articles. Abstracts from the following
annual, international scientific meetings were searched going back five years: Digestive
Disease Week, American College of Gastroenterology and United European Gastroenterology
Week.
Inclusion and exclusion criteria
Studies were included if they compared EUS-GE to ES or SGJ in patients with malignant
GOO. Randomized controlled trials as well as observational studies of retrospective
or prospective cohorts were included.
Exclusion criteria were: non-English and non-French articles; non-human studies; case
reports and studies with fewer than 10 participants; studies of EUS-GE using magnets;
studies of EUS-GE using Natural Orifice Transluminal Endoscopic Surgery; studies regarding
the treatment of concomitant gastric outlet and biliary obstruction.
Validity assessment, data abstraction and rating of evidence
Studies were independently assessed for inclusion by two authors (CSM and JB) with
discrepancies resolved, as needed, by a third (YC). Study and baseline patient characteristics,
duration of follow-up and outcomes of interest were abstracted. The Cochrane Risk
of bias tool for randomized trials or the Risk of Bias In Non-randomized Studies of
Interventions (ROBINS-I) tool were used to assess the risk of bias when appropriate
[8 ]
[9 ]. Prespecified confounding domains were: cancer type; presence of carcinomatosis;
age. No important co-interventions that have the potential to lead to bias were prespecified.
Outcomes and study definitions
There were two prespecified primary outcomes: clinical success without recurrent GOO
and AEs. The secondary outcomes were technical success and hospital length of stay.
Clinical success was defined as the ability to tolerate at least a liquid diet post-procedure.
Recurrent GOO was defined as recurrence of initial symptoms of nausea, vomiting and
inability to tolerate oral intake after clinical success had initially been achieved.
Adverse events were graded in accordance with the American Society of Gastrointestinal
Endoscopy lexicon [10 ] and included infection, perforation, bleeding, leak, pancreatitis, and in-hospital
mortality related to the index procedure. Studies that did not report on recurrent
GOO or that expressed this outcome only as a combined figure with other AEs were excluded.
Technical success was defined for EUS-GE and ES as adequate deployment of the stent
as reported by the endoscopist; for SGJ it was defined as feasibility to perform gastrojejunostomy.
Statistical analysis
For all outcomes, effect sizes were calculated for EUS-GE compared to ES and SGJ combined
with mean differences for continuous variables and odds ratios (ORs) for categorical
variables. The DerSimonian and Laird method for random effect models was applied to
all outcomes to determine corresponding overall effect sizes and their confidence
intervals. Sensitivity analyses were performed using the Mantel-Haenszel method for
fixed effects models when no statistical heterogeneity was noted.
Mean differences were handled as continuous variables using the inverse variance approach.
Presence of heterogeneity across studies was defined using a Chi-square test of homogeneity
with a 0.10 significance level. The Higgins I2 statistic was calculated to quantify the proportion of variation in intervention
effects attributable to between-study heterogeneity. Values of 0% to 40%, 30% to 60%,
50% to 90%, and 75% to 100% represent low, moderate, substantial, and considerable
heterogeneity, respectively. Prediction intervals were calculated and added to the
forest plots. The prediction interval calculates the 95% of where the effect size
will be if a new study is randomly added to the meta-analysis [11 ]. For all comparisons publication bias was evaluated using funnel plots if at least
three citations were identified. All statistical analyses were done using Revman 5.4
and Meta package in R version 2.13.0, (R Foundation for Statistical Computing, Vienna,
Austria, 2008).
Sensitivity and subgroup analyses
Subgroup analyses using ES or SGJ as separate comparators were performed for all outcomes.
Additional subgroup analyses were performed for primary outcome according to full
publication status and continent of publication. Sensitivity analysis was performed
adopting fixed effect models when appropriate. In addition, observational studies
are subject to confounding and other forms of bias [12 ]. We, therefore, performed additional sensitivity analyses based on studies with
low risk, moderate to serious risk, and critical risk of bias due to confounding,
identified from the ROBINS-I risk of bias tool.
Results
Study selection
The search yielded 1,078 citations ([Fig. 1 ]). One study that resulted from the systematic search in abstract form was identified
as a complete manuscript from hand searching [13 ]. After screening based on title and abstract, 73 articles were reviewed in full.
Sixteen articles were included in the qualitative review [1 ]
[13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]
[22 ]
[23 ]
[24 ]
[25 ]
[26 ]
[27 ], all retrospective comparative studies. Eight were fully published articles [1 ]
[13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[22 ]
[27 ] and eight were abstracts [18 ]
[19 ]
[20 ]
[21 ]
[23 ]
[24 ]
[25 ]
[26 ]. One study was excluded from quantitative analysis since data were reported per
stent and not per patient [27 ]. The results of the literature search are summarized in the PRISMA diagram ([Fig. 1 ]).
Fig. 1 PRISMA Diagram. From: Page MJ, McKenzie JE, Bossuyt PM et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.
BMJ 2021; 372: n71.
Risk of publication bias and heterogeneity and risk of bias
Publication bias was noted only for the outcomes of clinical success without recurrent
GOO (Egger’s P = 0.04; Begg’s P = 0.09) and AEs (Egger’s P = 0.08; Begg’s P = 0.04). Moderate to substantial heterogeneity was noted for clinical success without
recurrent GOO (P < 0.01; I2 = 60%). Adverse event rates reporting exhibited low to moderate heterogeneity (P < 0.01; I2 = 54%). No significant heterogeneity was noted for secondary outcomes. The ROBINS-I
showed a low risk of bias due to confounding factors in six studies, moderate to critical
in seven studies, and insufficient information was provided in the last three studies
(Supplementary Fig. 1). Four studies were thought to have serious risk of bias in
the selection of patient participations. All studies demonstrated moderate risk of
bias in the classification of exposure or from intended interventions. Finally, the
risk of bias was low due to missing data or measurement of exposure.
Patient and study characteristics
The main characteristics of the 16 included studies (n = 1541) [1 ]
[13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]
[22 ]
[23 ]
[24 ]
[25 ]
[26 ]
[27 ] are summarized in [Table 1 ]. These were conducted in Asia (India, Japan, Hong Kong), Europe (Belgium, Netherlands,
Spain), and North America (United States). Studies were published between 2016 to
2022. Overall,15 studies (n = 1,441) were included in the meta-analysis [1 ]
[13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]
[22 ]
[23 ]
[24 ]
[25 ]
[26 ]. Six [1 ]
[13 ]
[18 ]
[20 ]
[21 ]
[26 ] (n = 494) compared EUS-GE to ES, seven [14 ]
[15 ]
[16 ]
[17 ]
[22 ]
[23 ]
[24 ] (n = 466) EUS-GE to SGJ, while two others [19 ]
[25 ] (n = 481) described EUS-GE results versus both ES and SGJ. The average age of patients
included ranged between 62 to 71 years. The percentage of female participants ranged
between 32.3% and 64.3%. Eleven articles exclusively analyzed malignant GOO, whereas
five articles also reported on a minority of patients with benign GOO. The etiology
of malignant GOO was reported in 11 studies, with malignancies described as pancreatic,
gastric, duodenal, biliary, metastases or other. Nine studies reported on the presence
of peritoneal carcinomatosis (ranging from 11% to 100%). Fourteen studies reported
average follow-up times between 52 and 269 days.
Table 1 Characteristics of studies included in systematic review.
Publication
Patients
Malignant GOO%
Cancer type
Carcinomatosis %
Follow-up, days***
Author
Year
Country*
Groups
Age, y**
Female %
ITT
study
control
study
control
study (%)
control (%)
study
control
Study
control
ES, enteral stent; EUS-GE, endoscopic ultrasound-gastroenterostomy; GB, gallbladder;
ITT, intention-to-treat; NR, not reported; SGJ, surgical gastrojejunostomy.
Khashab et al →
2017
Japan
EUS-GE
68.7
47.3
93
30
63
100
100
Pancreatic (56), Gastric (17.6), metastases (13), ampullary (6.7), biliary/GB (6.7)
Pancreatic (84.5) Ampullary (14), Duodenal (1.5)
43
11
115
196
Chen
2016
Japan
EUS-GE
66.2
40.2
82
30
52
100
100
Pancreatic (56.7), Gastric (16.7), Metastases (13.3), Ampullary/duodenal (6.7), Biliary/GB
(6.7)
Pancreatic (53.8), Metastases (19.2), Ampullary/ duodenal (13.5), Biliary/GB (7.7),
Gastric (5.8)
46.7
34.6
103
83
Bronswijk
2021
Belgium
EUS-GE
65.3
44
125
77
48
96.1
85.4
Pancreatic (48.1), Duodenal (14.3), Biliary/GB (11.7), Metastases (10.4), Gastric
(9.1), Other (1.3)
Pancreatic (29.2), Duodenal (20.8), Gastric (10.4), Metastases (8.4), Biliary/GB (4.2),
Ampullary (2.1)
26
33.3
76
122
Perez-Miranda
2017
USA
EUS-GE
70.3
38.9
54
25
29
68
100
NR
NR
NR
NR
56
269
Kouanda
2021
USA
EUS-GE
70.2
42.4
66
40
26
90
54
Pancreatic (72.2), Metastases (13.9), Biliary/GB (8.4), Ampullary (2.8), Duodenal
(2.8)
Gastric (57.1), Pancreatic (21.4), Duodenal (7.1), Biliary/GB (7.1), Metastases (7.1)
11.1
42.9
98
166.5
Vazquez-Sequeiros
2020
Spain
EUS-GE
71.2
36.9
92
46
46
100
100
Pancreatic (61), Gastric (15), Biliary/GB (9), Metastases (9), Duodenal (6)
Pancreatic (61), Gastric (28), Biliary/GB (6), Duodenal (5)
NR
NR
76
91
Widmer
2019
USA
EUS-GE
65.9
58.3
24
10
14
100
100
Pancreatic (40), Metastases (30), Gastric (20), Duodenal (10)
Pancreatic (28.6), Duodenal (21.4), Metastases (21.3), Biliary/GB (14), Ampullary
(7), Gastric (7)
NR
NR
90
240
Bondi
2020
USA
EUS-GE
62.3
53.4
52
18
34
100
100
Metastases (33), Biliary/GB (28), Pancreatic (22), Gastric (11), Duodenal (6)
Pancreatic (32), Metastases (29), Gastric (24), Duodenal (12), Biliary/GB (3)
100
100
NR
NR
Marya
2020
USA
EUS-GE
62.4
43.7
364
172
153 (ES)
79.7
88.2 (ES)
NR
NR
11
11.7 (ES)
234
61 (ES)
Iqbal
2019
USA
EUS-GE
68.4
48.3
60
8
52
92
92
NR
NR
NR
NR
NR
NR
Abbas
2022
USA
EUS-GE
61.9
53.8
52
25
27
100
100
Other, pancreatic, billiary/GB, gastric
Pancreatic, gastric, biliary/GB, other
100
93
98
199
v. Wanrooij
2022
Netherlands
EUS-GE
66.5
52.3
214
88
88
100
100
Pancreatic (46.7), other (18.7), biliary tract (14), gastric (11.2), duodenal (9.3)
Pancreatic (65.4), other (11.2), duodenal (9.3), gastric (7.5), biliary tract (6.5)
41.1
25.2
90.5
50
S.M Chan
2021
Hong Kong
EUS-GE
67.1
64.3
117
48
38
100
100
Other, pancreatic gastric
Gastric, pancreatc, other
NR
NR
51.5
163
Rosas
2021
USA
EUS-GE
68.2
40
50
25
25
100
100
NR
NR
NR
NR
190
220
Dhir
2021
India
EUS-GE
64.4
32.3
34
18
16
100
100
NR
NR
NR
NR
120
120
Ge
2019
USA
EUS-GE
65.8
44
100
22
78
100
100
Metastases (40.9), Pancreatic (31.8), Biliary/GB (18.2), Duodenal (4.6), Gastric (4.6)
Pancreatic (51.3), Metastases (24.4), Biliary/GB (10.3), Gastric (10.3), Ampullary
(2.6), Duodenal (1.3)
59.1
47.4
180
NR
Primary outcome
EUS-GE was associated with higher clinical success without recurrent GOO compared
to ES or SGJ combined (OR, 2.60; 95% CI, 1.58–4.28) ([Fig. 2 ]a and [Table 2 ]). Subgroup analysis also showed higher clinical success without recurrent GOO for
EUS-GE compared to ES alone (OR, 5.08; 95% CI, 3.42–7.55), but yielded no statistically
significant difference when compared to SGJ alone (OR, 1.94; 95% CI, 0.97–3.88) ([Fig. 3 ]). Prediction interval remained significant for EUS-GE compared to ES alone, but
not compared to ES or SGJ combined. EUS-GE was associated with significantly fewer
AEs compared to ES or SGJ combined (OR, 0.34; 95% CI, 0.20–0.58) ([Fig. 2 ]b). One study was excluded from this analysis as AEs were not specified and were
presumed to include recurrent GOO [19 ]. On subgroup analysis, EUS-GE was associated with no statistically significant difference
in AEs rates compared to ES alone (OR, 0.57; 95% CI, 0.29–1.14), and fewer AEs compared
to SGJ alone (OR, 0.17; 95% CI, 0.10–0.30) ([Fig. 4 ]). Prediction interval remained significant for EUS-GE compared to SGJ alone, but
not compared to ES or SGJ combined.
Table 2 Primary, secondary and subgroup analyses.
N studies
N patients
OR or WMD
P value for heterogeneity
I2
CI, confidence interval; ES, enteral stenting; GOO, gastric outlet obstruction; OR,
odds ratio; SGJ, surgical gastrojejunostomy; WMD, weighted mean difference.
Primary outcomes
Clinical success without recurrent GOO
14
1407
2.60 (1.58; 4.28)
< 0.01
60%
ES only
7
864
5.08 (3.42; 7.55)
0.51
0%
SGJ only
9
763
1.94 (0.97; 3.88)
< 0.01
61%
Adverse events
15
1441
0.34 (0.20, 0.58)
0.02
50%
ES only
8
937
0.57 (0.29; 1.14)
0.16
35%
SGJ only
9
763
0.17 (0.10; 0.30)
0.16
33%
Secondary outcomes
Technical success
15
1441
0.32 (0.16; 0.64)
0.44
1%
ES only
8
898
0.44 (0.18; 1.12)
0.29
19%
SGJ only
9
763
0.17 (0.06; 0.49)
0.99
0%
Length of stay
3
227
0.03 (–2.31, 2.36)
0.30
18%
ES only
1
82
1.80 (–1.47, 5.07)
–
–
SGJ only
2
145
–1.06 (–3.75, 1.63)
0.41
0%
Fig. 2
a Clinical success without recurrent GOO. b Adverse events.
Fig. 3 Clinical success without recurrent GOO. a EUS-GE vs. ES. b EUS-GE vs. SGJ.
Fig. 4 Adverse events. a EUS-GE vs. ES. b
EUS-GE vs. SGJ.
Secondary outcomes
EUS-GE was associated with a significant decrease in technical success compared to
ES or SGJ combined (OR, 0.32; 95% CI, 0.16–0.64) and compared to SGJ alone (OR, 0.17;
95% CI, 0.06–0.49); however, there was no statistically significant difference when
compared to ES alone (OR, 0.44; 95% CI, 0.18–1.12) ([Table 2 ]). Hospital length of stay was only reported in three studies with extractable data,
two comparing EUS-GE to SGJ and one comparing EUS-GE to ES. There were no significant
differences in lengths of hospital stay when comparing EUS-GE to ES or SGJ combined
(mean difference (MD), 0.03; 95% CI, -2.31–2.36), or when compared to ES (MD, 1.80;
95% CI, -1.47–5.07) or SGJ (MD, -1.06; 95% CI, -3.75–1.63) alone.
Sensitivity and subgroup analyses
Among fully published studies, EUS-GE was associated with higher clinical success
without recurrent GOO compared to SGJ or ES combined (OR, 3.51; 95% CI, 1.88–6.56)
([Table 3 ]). Results were similar when assessing North American studies (OR, 2.28; 95% CI,
1.23–4.21) and European studies only (OR, 3.77; 95% CI, 1.51–9.42). There was no statistically
significant difference found between the two groups in studies published as abstract
only (OR, 1.91; 95% CI, 0.86–4.27). Results remained robust when including only low
risk of confounding bias and there was no heterogeneity noted (OR, 3.51; 95% CI, 2.33–5.27),
whereas no difference between the two groups was found for moderate to serious and
critical risk of confounding bias (Supplementary Fig. 2).
Table 3 Additional sensitivity and subgroup analyses.
No. studies
No. patients
Odds ratio (95% CI)
P value for heterogeneity
I2
CI, confidence interval; GOO, gastric outlet obstruction.
Clinical success without recurrent GOO
Publication status
Fully published article
7
759
3.51 (1.88, 6.56)
0.10
44%
Abstract
7
648
1.91 (0.86, 4.27)
< 0.01
69%
Continent
North American
11
1014
2.28 (1.23, 4.21)
< 0.01
61%
Europe
3
393
3.77 (1.51, 9.42)
0.09
59%
Confounding bias
Low risk of bias
6
643
3.51 (2.33; 5.27)
0.11
44%
Moderate to serious risk of bias
3
482
1.82 (0.34; 9.64)
< 0.01
88%
Critical risk of bias
3
166
1.38 (0.54; 3.53)
0.11
54%
Adverse events
Publication status
Fully published article
8
793
0.33 (0.08, 1.34)
0.13
42%
Abstract
7
648
0.26 (0.10, 0.64)
0.84
0%
Continent
North America
11
1014
0.18 (0.07, 0.44)
0.86
0%
Europe
4
427
0.67 (0.19, 2.35)
0.26
25%
Confounding bias
Low risk of bias
5
526
0.46 (0.28; 0.75)
0.11
47%
Moderate to serious risk of bias
3
482
0.16 (0.05; 0.54)
< 0.01
69%
Critical risk of bias
3
166
0.34 (0.06; 1.75)
0.08
61%
Adverse events were significantly lower for EUS-GE compared to SGJ and ES combined
when limiting the analysis to abstracts alone, as was also the case for North American
studies. There were no significant differences identified when limiting the analysis
to fully published articles or European studies. Results remained robust as well when
including only low risk of confounding bias and there was no heterogeneity (OR, 0.46;
95% CI, 0.28–0.75), whereas no difference between the two groups was found for moderate
to serious and critical risk of confounding bias (Supplementary Fig. 3).
Significant heterogeneity precluded the performance of fixed effect models.
Discussion
EUS-GE has emerged as a promising modality for the treatment of malignant GOO. This
approach exhibits two major potential benefits over the traditional modalities: Given
the complete nature of the bypass of the obstruction created with a gastro-enteric
anastomosis, EUS-GE may offer a more durable treatment compared to enteral stenting
that traverses the tumor and is prone to recurrent obstruction. On the other hand,
the endoscopic nature of the EUS-GE procedure, even if extraluminal, may offer significant
safety advantages over a traditional surgical bypass. In the present systematic review
and meta-analysis of 15 studies that included 1,441 patients, EUS-GE was associated
with higher clinical success without recurrent obstruction and fewer AEs compared
to the traditional treatments of GOO. Subgroup analysis showed higher clinical success
without re-obstruction compared to traditional stenting, with no difference compared
to surgical bypass. On the other hand, EUS-GE was associated with fewer AEs compared
to SGJ, with no difference compared to ES. Taken together, these findings support
the theoretical advantages of EUS-GE for the treatment of malignant GOO that results
in a robust bypass while maintaining a safer, less-invasive approach. Indeed, recent
European Society for Gastrointestinal Endoscopy guidelines have recommended EUS-GE
be “performed in an expert setting, for malignant GOO, as an alternative to enteral
stenting or surgery,” as a strong recommendation based on low-quality evidence [28 ].
Despite the promising data presented herein, EUS-GE is not commonly utilized. This
is likely due to its technically challenging nature relative to other interventional
endoscopic procedures coupled with the lack of standardization of the technique, which
has limited training and dissemination [29 ]. Furthermore, despite the reassuring safety data described above, there remains
reasonable concern regarding the potential complication of stent misdeployment. For
these reasons, the use of EUS-GE is presently limited mostly to high-volume, tertiary-care
endoscopy centers.
The most feared complication of the EUS-GE procedure is stent misdeployment since
the tract being created is extraluminal [30 ]. A recent international, multicentered study reported on outcomes and management
of EUS-GE and found that stent misdeployment occurred in close to 10% of the 467 EUS-GE
procedures, with surgery required in 11% of these and the remaining 89% managed by
conservative or endoscopic means [31 ]. Most resulting AEs were graded as mild, although six cases were severe and one
fatality occurred after an attempt at surgical repair.
Regarding factors that may contribute to stent misdeployment, the endoscopist’s experience
seems to play an important role. In the study by Ghandour et al [31 ], 73% of misdeployments occurred within the endoscopist’s first 13 EUS-GE procedures.
In a European multicentered cohort study of 45 EUS-GE procedures, most of the five
misdeployments leading to technical failure happened at a single center early after
introducing EUS-GE [32 ]. Indeed, a study of a single expert endoscopist’s EUS-GE learning curve using cumulative
sum curve analysis with mean procedure time as the target value suggested that 25
cases are necessary to become proficient, and 40 to achieve mastery [33 ]. Interestingly, in the study by Ghandour et al [31 ], 83% and 90% of stent misdeployments occurred prior to these suggested proficiency
and mastery reference points, respectively. In terms of procedure-related factors,
there has been a shift away from over-the-wire placement of the LAMS to form the gastroenterostomy,
as the wire has been noted to push the small bowel away, potentially resulting in
misdeployment [32 ]
[34 ]. The risk of stent misdeployment highlights the need for further refinement and
standardization of EUS-GE technique. It is also imperative that proper informed consent
be obtained and that plans for salvage maneuvers are considered by the endoscopy team
in advance. At the same time, our findings support the overall safety of EUS-GE compared
to the traditional treatments of malignant GOO, with fewer total AEs compared to SGJ
and no difference compared to ES.
Designated accessory devices will be essential for improvement of the EUS-GE procedure
and its dissemination beyond the walls of the most expert centers. It should be noted
that there are currently no approved dedicated devices for this complex procedure
in North America. Even the LAMS, the stent universally used to form the gastroenterostomy,
is off-label for this indication.
Although our results demonstrate better durability associated with EUS-GE compared
to traditional ES, it is not clear whether EUS-GE allows for as robust a bypass as
a surgical anastomosis. While EUS-GE has almost exclusively been studied using LAMS
with a 15-mm diameter, 20-mm LAMS was recently developed. A retrospective study comparing
EUS-GE using a 20-mm vs 15-mm LAMS demonstrated similar results with regards to technical
success, clinical success and AEs; however, a higher proportion of patients in the
larger LAMS group tolerated soft or complete diets [35 ]. Interestingly, the functional diameter of a SGJ anastomosis may not be more than
20 mm [36 ]. Current recommendations for patients with malignant GOO who have a life expectancy
greater than 2 months favor SGJ over ES given evidence of better long-term patency
[37 ]. While our data suggest no difference in clinical success without recurrent GOO
between EUS-GE and SGJ, high-quality prospective head-to-head studies with sufficient
duration of follow-up are required to address this important comparison.
The current study is limited by mostly retrospective data and relatively small sample
sizes of the included studies. There is clinical heterogeneity in the differences
in EUS-GE technique used, as the procedure lacks standardization. Further heterogeneity
is introduced by inclusion of both malignant and some benign GOO in patient selection.
Observational studies are prone to confounding and other forms of bias. We assessed
each study according to the ROBINS-I tool and performed sensitivity analyses for the
primary outcomes by risk of bias due to confounding. Results demonstrated that the
findings for both primary outcomes remain robust when including only studies at low
risk for bias due to confounding. Furthermore, statistical heterogeneity was no longer
present for these analyses, indicating that an important contribution to the heterogeneity
found in the primary outcomes comes from studies that are higher risk for confounding.
Lastly, the studies included were mainly performed in high-volume tertiary-care centers,
which can impact the generalizability of these results to outside the most expert
centers. Strengths of this meta-analysis include its a priori design with protocol
registration and selection of primary endpoints that are both clinically most relevant
as well as homogenous throughout individual studies.
Conclusions
In conclusion, our systematic review and meta-analysis show that EUS-GE is associated
with higher clinical success without recurrent obstruction and fewer AEs compared
to the traditional standard of care treatments of GOO. Further development of designated
accessory devices and standardization of the technique are required to mitigate the
technical challenges of this promising modality. High-quality randomized trials will
also be needed to better characterize the role of EUS-GE in the treatment of malignant
GOO.
