Palliative procedures for malignant gastric outlet obstruction: a network meta-analysis

Abstract

Background The optimal treatment for malignant gastric outlet obstruction (GOO) remains uncertain. This systematic review aimed to comprehensively investigate the efficacy and safety of four palliative treatments for malignant GOO: gastrojejunostomy, endoscopic ultrasound-guided gastroenterostomy (EUS-GE), stomach-partitioning gastrojejunostomy (PGJ), and endoscopic stenting.

Methods We searched PubMed, Embase, Cochrane Library, Scopus, and Web of Science databases, ClinicalTrials.gov, and the World Health Organization International Clinical Trials Registry Platform for randomized controlled trials (RCTs) and cohort studies comparing the four treatments for malignant GOO. We included studies that reported at least one of the following clinical outcomes: clinical success, 30-day mortality, reintervention rate, or length of hospital stay. Evidence from RCTs and non-RCTs was naïve combined to perform network meta-analysis through the frequentist approach using an inverse variance model. Treatments were ranked by P score.

Results This network meta-analysis included 3617 patients from 4 RCTs, 4 prospective cohort studies, and 32 retrospective cohort studies. PGJ was the optimal approach in terms of clinical success and reintervention (P scores: 0.95 and 0.90, respectively). EUS-GE had the highest probability of being the optimal treatment in terms of 30-day mortality and complications (P scores: 0.82 and 0.99, respectively). Cluster ranking to combine the P scores for 30-day mortality and reintervention indicated the benefits of PGJ and EUS-GE (cophenetic correlation coefficient: 0.94; PGJ and EUS-GE were in the same cluster).

Conclusion PGJ and EUS-GE are recommended for malignant GOO. PGJ could be the alternative choice in centers with limited resources or in patients who are unsuitable for EUS-GE.

Introduction

Malignant gastric outlet obstruction (GOO) results from various malignancies, such as gastroduodenal, pancreaticobiliary, colorectal, breast, and lung cancers [1] [2]. GOO indicates poor prognosis, and it often presents with nausea, bloating, vomiting, and abdominal pain. It can also cause difficulty in oral intake, which can lead to fluid and electrolyte imbalances and severe malnutrition, from which recovery is difficult [3] [4]. As patients with GOO have a median survival of only 2–10 months, doctors face many challenges when selecting treatment modalities [5] [6]. Palliative procedures for patients with GOO involve re-enabling the passage of food and liquid at the obstructed location or creating a new gastrointestinal pathway that bypasses the obstruction. Therefore, gastrojejunostomy, stomach-partitioning gastrojejunostomy (PGJ), endoscopic stenting, and endoscopic ultrasound-guided gastroenterostomy (EUS-GE) are indicated to restore the normal function of the digestive system ([Fig. 1]).

Fig. 1 Techniques for treating malignant gastric outlet obstruction (GOO). Gastrojejunostomy: reconstructive procedure that creates the anastomosis between the stomach and small intestine. Stomach-partitioning gastrojejunostomy: the stomach is partitioned from the greater curvature to the lesser curvature, leaving only a tunnel 2–3 cm from the lesser curvature; gastrojejunostomy is then performed in the proximal part of the stomach. Endoscopic stenting: placement of a permanent, endoluminal, self-expandable metal stent through the stricture. Endoscopic ultrasound-guided gastroenterostomy: through the ultrasound tip of the endoscope, the position of the duodenum or jejunum next to the stomach wall is determined and an anastomosis is created using a lumen-apposing metal stent.

PGJ is more effective than gastrojejunostomy in terms of reducing delayed gastric emptying [7]. The advantage of endoscopic stenting over EUS-GE is a shorter hospital stay, but at a cost of a higher reintervention rate due to stent blockage or migration [8]. The advantage of EUS-GE over endoscopic stenting is a lower rate of reintervention due to recurrent GOO [9]. Moreover, by establishing the anastomosis far from the tumor location, PGJ and EUS-GE may reduce the risk of recurrent obstruction. However, PGJ and EUS-GE are more invasive and complex procedures, and their risks and benefits remain uncertain.

The lack of pairwise comparison in randomized controlled trials (RCTs) and non-RCTs indicates the difficulty of optimal clinical decisions. Therefore, this network meta-analysis (NMA) comprehensively evaluated these four primary procedures for the treatment of malignant GOO.

Methods

Inclusion and exclusion criteria

Studies were included in this NMA if they were RCTs or prospective or retrospective cohort studies that compared at least two of the treatments (gastrojejunostomy, PGJ, endoscopic stenting, and EUS-GE) for patients with malignant GOO, and reported at least one of the following clinical outcomes: clinical success rate, complication rate, 30-day mortality rate, reintervention rate, or length of hospital stay (LOS). We included RCTs as well as cohort studies that were balanced at baseline, with the rationale that appropriately analyzed non-RCTs could also aid in improving the decision-making process [10]. To minimize bias, only studies with cohorts that were balanced in terms of confounding factors (P > 0.05 for baseline characteristics) or with characteristics that were adjusted at baseline (through propensity-score matching, regression model adjustment, or subgroup analysis) were included [11].

We excluded studies that investigated only benign GOO or reported unextractable data on malignant GOO. We also excluded studies without balanced baseline characteristics or suitable adjustments, studies that did not compare palliative treatments, and studies comparing treatment characteristics (e.g. surgical approach, anastomosis type, or stent type) within only one type of intervention.

Search strategy

Two authors independently conducted the search. The designated search terms are presented in the supplementary materials and methods (see Section S1 in the online-only Supplementary material). We searched the PubMed, Embase, Cochrane Library, Scopus, and Web of Science databases (Table 1s). The final search was conducted in March 2023. To evaluate ongoing and related studies, we also searched ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform. All references from original articles and previous systematic reviews were also checked for additional relevant articles. We did not limit our search by language or date of publication. We contacted the corresponding authors of studies when clarification on the outcomes of interest was necessary. The study protocol was registered on PROSPERO (CRD42021265505).

Outcome variables and data extraction

The primary outcome of this NMA was clinical efficacy based on reintervention rate. The primary purpose of treatment for malignant GOO is restoring the movement of food and liquid through the gastrointestinal tract; a lower rate of reintervention due to complication or recurrent obstruction indicates greater clinical efficacy. The secondary outcomes were clinical success rate, complication rate, 30-day mortality rate, and LOS. In addition, the 30-day mortality rate and complication rate were separately combined with reintervention rate, for analysis of the combined safety and efficacy of treatments. Additional details, including the definitions of all procedures and outcomes, are provided in supplementary Section S1.

Two authors independently extracted data. Differences and uncertainties were resolved through group discussion after the authors of the original publications were contacted for clarification. We extracted study characteristics and statistics related to the outcomes of interest. We used WebPlotDigitizer to extract data from figures (Kaplan–Meier curves). We implemented data transformation to means and SDs for continuous data reported by a five-number summary [12].

Network geometries, risk of bias, and level of evidence

The geometries and confidence in the evidence of this NMA were created in accordance with the Confidence in Network Meta-Analysis (CINeMA) approach [13]. The risk of bias within individual studies was evaluated using the Cochrane RoB 2 tool for RCTs and ROBIN-I tool for non-RCTs [14] [15]. To evaluate publication bias, we employed a funnel plot and Egger test, and checked for asymmetry in order from the oldest to the newest treatment comparisons [16].

Statistical analysis and network outcomes

Treatment effects are reported as relative risks (RRs) with 95%CIs for binary outcomes and as mean differences with 95%CIs for continuous outcomes. Our network model followed a frequentist approach in which we used the package netmeta ver.2.1–0 in R ver.4.2.0 [17]. We transformed the effect size data through pairwise comparison to match the network model. We then established a random-effects model. We assessed global network consistency by using the Q statistic under the assumption of full design-by-treatment interaction in the random-effects model. We assessed local network inconsistency by separating indirect and direct evidence through back-calculation [18]. Owing to the limited availability of data from RCTs, we performed a standard NMA with naïve combination of all the evidence from the RCTs and non-RCTs [10]. As the rates of the outcomes of interest were common and very common, we used the inverse variance method with continuity correction for only those studies with zero events in every treatment arm following the default setup for the inverse variance method in the netmeta package.

The network outcomes included forest plots and league tables sorted by treatment ranking for each outcome of interest. We ranked treatment outcomes using P score (0–1), which indicates the probability of a treatment being optimal, with a higher P score indicating better treatment [19]. We combined rankings of primary and secondary outcomes to obtain the optimal treatment. We specified negative effect sizes for complication rate, 30-day mortality rate, reintervention rate, and LOS, meaning that more effective treatments would reduce these values. We performed subgroup analysis to investigate potential effect modifiers in cases of high heterogeneity, and we conducted sensitivity analysis to detect potential inconsistencies in comparisons and publication bias.

Results

Additional data and analyses from this NMA are available in the online supplementary material in Tables 1s–29s and Fig. 1s–6s.

Search overview, study characteristics, and network geometries

The systematic search process identified 10 911 studies. After removing 4605 duplicates through Endnote and manual screening, and excluding 6235 irrelevant studies through title and abstract screening, we retrieved the full texts of 73 studies; 40 studies were deemed eligible for inclusion in our systematic review on the basis of the inclusion and exclusion criteria ([Fig. 2]).

Fig. 2 Flow chart according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses. WHO ICTRR, World Health Organization International Clinical Trials Registry Platform.

Data were derived from 3617 patients from 4 RCTs, 4 prospective studies, and 32 retrospective cohort studies. The included studies were conducted in Asia, Europe, North America, South America, Oceania, or multiple regions, and comprised patients with malignant GOO, pancreatic cancer only, or gastric cancer only (Table 2s, Table 3s). The overall population had a mean age of 67 (SD 13) years and a male:female ratio of 3:2. The overall rates of clinical success, complications, 30-day mortality, and reintervention were 88.9% (95%CI 85.6 to 91.6), 20.7% (95%CI 17.2 to 24.7), 5.4% (95%CI 3.2 to 8.9), and 13.9% (95%CI 10.7 to 17.9), respectively (Table 4s, Fig. 2s).

[Fig. 3] depicts the network geometry for the primary outcome, with four interventions and five direct comparisons (EUS-GE–PGJ comparison was not available). The secondary outcomes of clinical success rate, mortality rate, and complication rate had similar geometries to that of the primary outcome, but LOS data were available for only four direct comparisons (EUS-GE vs. PGJ and PGJ vs. endoscopic stenting comparisons were not available) (Fig. 1s).

Fig. 3 CINeMA network geometry for the primary outcome (reintervention rate). There were a total of 24 studies (3 RCTs, 21 non-RCTs) including 2681 patients with data for this outcome. The number of studies for each comparison was as follows: EUS-GE vs. gastrojejunostomy vs. endoscopic stenting (n = 1); gastrojejunostomy vs. PGJ vs. endoscopic stenting (n = 1); EUS-GE vs. endoscopic stenting (n = 1); EUS-GE vs. gastrojejunostomy (n = 2); PGJ vs. endoscopic stenting (n = 1); gastrojejunostomy vs. PGJ (n = 2); gastrojejunostomy vs. endoscopic stenting (n = 16). EUS-GE, endoscopic ultrasound-guided gastroenterostomy; PGJ, stomach-partitioning gastrojejunostomy. Node size = sample size; node color = risk of bias; edge size = number of studies; edge color = major indirectness; red = high; yellow = moderate; green = low.

Primary outcome

Reintervention rate

The risk of reintervention was found to be lower for PGJ (RR 0.14; 95%CI 0.03 to 0.67), EUS-GE (RR 0.27; 95%CI 0.11 to 0.67), and gastrojejunostomy (RR 0.51; 95%CI 0.38 to 0.70) than for endoscopic stenting, with low and very low certainty ([Fig. 4], Table 17s). PGJ and EUS-GE had higher probabilities of being the optimal approach in terms of reintervention rate (P scores of 0.90 and 0.72, respectively). Endoscopic stenting had the lowest probability of preventing reintervention (P score of 0.00). The heterogeneity in this network (τ² = 0.17; I ² = 39.3%) was moderate ([Table 1]). We found no evidence of inconsistency between direct and indirect comparisons, and no publication bias in the network for this outcome (Table 5s, Fig. 4s). The results of standard pairwise meta-analyses were similar to those of the NMA (Table 6s, Fig. 5s).

Fig. 4 Forest plot of network meta-analysis for reintervention rate. EUS-GE, endoscopic ultrasound-guided gastroenterostomy; PGJ, stomach-partitioning gastrojejunostomy.

Table 1 Treatments for malignant gastric outlet obstruction ranked by P score for each outcome of interest.

	Rank by P score
Outcome Optimal treatment1	1	2	3	4
Data are relative risk values (reintervention, clinical success, complications, 30-day mortality) or mean difference values (length of hospital stay) with 95%CIs. Estimates in bold indicate statistical significance at P < 0.05. EUS-GE, endoscopic ultrasound-guided gastroenterostomy; PGJ, stomach-partitioning gastrojejunostomy. 1Order of each league table by ranking of P scores (in parentheses).
Reintervention rate
PGJ	PGJ (0.90)	–	0.33 (0.05 to 2.09)	0.11 (0.01 to 0.87)
EUS-GE	0.53 (0.09 to 3.18)	EUS-GE (0.72)	0.59 (0.19 to 1.80)	0.21 (0.06 to 0.74)
Gastrojejunostomy	0.28 (0.06 to 1.31)	0.52 (0.21 to 1.29)	Gastrojejunostomy (0.38)	0.52 (0.38 to 0.72)
Endoscopic stenting	0.14 (0.03 to 0.67)	0.27 (0.11 to 0.67)	0.51 (0.38 to 0.70)	Endoscopic stenting (0.00)
Heterogeneity: τ 2 = 0.17; Ι 2 = 39.3%
Clinical success rate
PGJ	PGJ (0.95)	–	1.06 (0.90 to 1.27)	1.49 (1.21 to 1.82)
EUS-GE	1.10 (0.92 to 1.32)	EUS-GE (0.69)	1.23 (1.04 to 1.45)	0.99 (0.84 to 1.18)
Endoscopic stenting	1.22 (1.07 to 1.40)	1.11 (0.98 to 1.27)	Endoscopic stenting (0.20)	1.00 (0.94 to 1.06)
Gastrojejunostomy	1.23 (1.07 to 1.41)	1.12 (0.98 to 1.27)	1.00 (0.95 to 1.06)	Gastrojejunostomy (0.17)
Heterogeneity: τ 2 = 0.01; Ι 2 = 69%
Complication rate
EUS-GE	EUS-GE (0.99)	0.63 (0.32 to 1.22)	–	0.43 (0.21 to 0.90)
Endoscopic stenting	0.58 (0.34 to 0.98)	Endoscopic stenting (0.62)	0.77 (0.29 to 2.01)	0.69 (0.54 to 0.89)
PGJ	0.40 (0.18 to 0.93)	0.70 (0.36 to 1.36)	PGJ (0.23)	0.87 (0.42 to 1.83)
Gastrojejunostomy	0.40 (0.23 to 0.68)	0.69 (0.54 to 0.88)	0.98 (0.51 to 1.87)	Gastrojejunostomy (0.16)
Heterogeneity: τ 2 = 0.15; Ι 2 = 39.7%
30-day mortality rate
EUS-GE	EUS-GE (0.82)	–	0.25 (0.03 to 2.06)	0.52 (0.06 to 4.27)
PGJ	0.82 (0.13 to 5.23)	PGJ (0.79)	0.28 (0.08 to 0.96)	0.54 (0.15 to 1.93)
Endoscopic stenting	0.29 (0.06 to 1.46)	0.36 (0.14 to 0.92)	Endoscopic stenting (0.21)	0.97 (0.71 to 1.33)
Gastrojejunostomy	0.29 (0.06 to 1.44)	0.35 (0.14 to 0.90)	0.98 (0.72 to 1.34)	Gastrojejunostomy (0.18)
Heterogeneity: τ 2 = 0.13; Ι 2 = 30%
Length of hospital stay
Endoscopic stenting	Endoscopic stenting (0.92)	−0.72 (−4.47 to 3.02)	−7.14 (−8.69 to −5.59)	–
EUS-GE	−0.99 (−3.94 to 1.96)	EUS-GE (0.75)	−5.15 (−9.00 to −1.30)	–
Gastrojejunostomy	−7.17 (−8.68 to −5.66)	−6.18 (−9.14 to −3.22)	Gastrojejunostomy (0.28)	−2.20 (−6.57 to 2.17)
PGJ	−9.37 (−13.99 to −4.75)	−8.38 (−13.65 to −3.10)	−2.20 (−6.57 to 2.17)	PGJ (0.05)
Heterogeneity: τ 2 = 8.34; Ι 2 = 79.2%

We conducted a secondary analysis of the primary outcome based on the indication for reintervention: obstruction or complication. PGJ and EUS-GE had higher probabilities of being the optimal approach in terms of reintervention for obstruction (P scores of 0.88 and 0.73, respectively) (Table 7s, Table 8s, Table 11s). Endoscopic stenting resulted in the highest probability of reintervention due to obstruction. PGJ was the optimal approach for preventing reintervention due to complications (P score of 0.81). Endoscopic stenting and gastrojejunostomy had lower probabilities of preventing reintervention due to complications (P scores of 0.33 and 0.31, respectively) (Table 9s, Table 10s, Table 11s).

Secondary outcomes

Clinical success rate

Compared with endoscopic stenting, PGJ had a higher probability of clinical success (RR 1.22; 95%CI 1.07 to 1.40), but EUS-GE (RR 1.11; 95%CI 0.98 to 1.27) and gastrojejunostomy (RR 1.00; 95%CI 0.94 to 1.06) had similar probabilities, with low or very low certainty (Table 18s, Fig. 3s). PGJ was the treatment with the highest probability of success (P score of 0.95), EUS-GE was the treatment with the second-highest probability of success (P score of 0.69), and endoscopic stenting and gastrojejunostomy were the approaches with the lowest probabilities of success (P scores of 0.20 and 0.17, respectively). The heterogeneity in this network (τ² = 0.01; I ² = 69%) was high ([Table 1]). Significant inconsistency with the design-by-treatment interaction model was observed (0.029), mainly due to the gastrojejunostomy vs. PGJ and PGJ vs. endoscopic stenting comparisons (Table 5s). We detected no substantial publication bias in this network (Fig. 4s). The results from standard pairwise meta-analyses were not considerably different from those of the NMA (Table 6s, Fig. 5s).

Subgroup analysis was performed to investigate potential effect modifiers (publication year, median age, type of cancer, and single-center or multicenter study) that might account for the high heterogeneity. In most studies with a median age of <68 years or ≥68 years, the effect estimates were inconsistent; however, PGJ was always the optimal approach. The results of the subgroup analyses for other potential effect modifiers (publication after 2013, type of cancer, and single-center design) were comparable to those of the principal network analysis (Table 12s). Sensitivity analysis of gastrojejunostomy vs. PGJ and PGJ vs. endoscopic stenting comparisons revealed a possible cause of inconsistency. Although the exclusion of several studies reduced the inconsistency, the treatment effects and rankings mostly agreed with those of the main network analysis (Table 14s).

Complication rate

Compared with the risk of complications with endoscopic stenting, the risk was lower with EUS-GE (RR 0.58; 95%CI 0.34 to 0.98) and similar for PGJ (RR 1.43; 95%CI 0.74 to 2.76). The risk of complications with gastrojejunostomy was higher than that with endoscopic stenting (RR 1.45; 95%CI 1.14 to 1.85), with very low certainty ([Table 1], Table 19s, Fig. 3s). EUS-GE was the treatment with the highest probability of being the safest (P score of 0.99), followed by endoscopic stenting (P score of 0.62). PGJ and gastrojejunostomy had the lowest probabilities of being the safest treatment (P scores of 0.23 and 0.16, respectively). The heterogeneity for this outcome (τ² = 0.15; I ² = 39.7%) was moderate ([Table 1]). We detected no evidence of inconsistency between direct and indirect comparisons of complication rates (Table 5s). However, significant publication bias was found (Fig. 4s). The results of pairwise comparisons agreed entirely with the network outcomes (Table 6s, Fig. 5s). Sensitivity analysis revealed greater publication bias relating to the gastrojejunostomy vs. endoscopic stenting comparison (Table 15s). Therefore, we downgraded the certainty of the gastrojejunostomy vs. endoscopic stenting comparisons on the basis of reporting bias (Table 15s).

30-day mortality

Compared with the risk of 30-day mortality with endoscopic stenting, the risk was lower with PGJ (RR 0.36; 95%CI 0.14 to 0.92) and similar for EUS-GE (RR 0.29; 95%CI 0.06 to 1.46) and gastrojejunostomy (RR 1.02; 95%CI 0.75 to 1.39), with very low certainty ([Table 1], Table 20s, Fig. 3s). EUS-GE and PGJ had similar probabilities of being the safest treatment (P scores of 0.82 and 0.79, respectively), and endoscopic stenting and gastrojejunostomy had the lowest probabilities of being the safest (P scores of 0.21 and 0.18, respectively). The heterogeneity (τ² = 0.13; I ² = 30%) for this outcome was moderate ([Table 1]). We observed no evidence of inconsistency between direct and indirect comparisons, and no publication bias relating to this network outcome (Table 5s, Fig. 4s). The results of pairwise comparisons were mainly in agreement with the network outcomes (Table 6s, Fig. 5s).

LOS

Compared with the LOS with endoscopic stenting, LOS with EUS-GE was similar (mean difference 0.99; 95%CI −1.96 to 3.94 days), whereas those with gastrojejunostomy (mean difference 7.17; 95%CI 5.66 to 8.68) and PGJ (mean difference 9.37; 95%CI 4.75 to 14.00) were longer, with very low certainty (Table 21s, Table 22s, Fig. 3s). Endoscopic stenting had the highest probability of preventing a long LOS (P score of 0.92), followed by EUS-GE, gastrojejunostomy, and PGJ (P scores of 0.75, 0.28, and 0.05, respectively). The heterogeneity (τ² = 8.34; I ² = 79.2%) for this outcome was substantial ([Table 1]). The evidence from direct and indirect comparisons was consistent for this network outcome (Table 5s). No substantial publication bias was detected in this network (Fig. 4s). The results of standard pairwise meta-analyses were in agreement with those of the NMA (Table 6s, Fig. 5s). Subgroup analysis was conducted to investigate potential effect modifiers to account for the substantial heterogeneity. In studies with a median age of <68 years and those conducted in a single center, EUS-GE and endoscopic stenting had high P scores (>0.7). The results of subgroup analyses of other potential effect modifiers (median age ≥68 years, publication after 2012, and type of cancer) were comparable to those of the principal analysis (Table 13s).

Combined safety and efficacy: cluster rankings of probability of optimal treatment

The P scores for the combination of safety (30-day mortality rate, complication rate) and efficacy (reintervention rate) are presented in [Fig. 5] and Fig. 6s. In [Fig. 5], PGJ and EUS-GE had the highest probabilities of being the optimal approach in terms of combined safety and efficacy. The poorest choice for combined safety and efficacy was endoscopic stenting. A high cophenetic correlation coefficient (c = 0.94) indicated confidence in the distances between clusters. Moreover, PGJ and EUS-GE are in the same cluster (red), and gastrojejunostomy (purple) and endoscopic stenting (blue) are dispersed throughout the lower left corner. With the complication rate used as the indicator of safety, the sparse plot was unsuitable for clustering, as indicated by the cophenetic correlation coefficient (c = 0.63; Fig. 6s).

Fig. 5 P score combining safety and efficacy. The x-axis indicates the probability of a treatment being optimal in terms of safety (30-day mortality). The y-axis indicates the probability of a treatment being optimal for efficacy (reintervention rate). Different clusters are represented by different colors. EUS-GE, endoscopic ultrasound-guided gastroenterostomy; PGJ, stomach-partitioning gastrojejunostomy.

Risk of bias, publication bias, contribution of evidence and certainty of evidence

The results indicating the risk of bias within individual studies, within studies for each comparison, risk of publication bias, contribution of evidence, and certainty of evidence according to the CINeMA approach are provided in Table 16s–29s and Fig. 4s.

Discussion

Our systematic review and NMA comprehensively compared the results of treating GOO with gastrojejunostomy, PGJ, endoscopic stenting, and EUS-GE. PGJ and EUS-GE were more advantageous than gastrojejunostomy and endoscopic stenting in terms of combined safety and efficacy. When EUS-GE cannot be used because of technical limitations, PGJ is recommended, especially in hospitals that lack an advanced endoscopy team.

The difficulty in patient recruitment made it challenging to compare the four strategies and determine the optimal palliative treatment for GOO. Only three published RCTs compared gastrojejunostomy and endoscopic stenting, and only one RCT compared PGJ and endoscopic stenting [4] [5] [20] [21]. Our gastrointestinal surgical teams advocated that clinicians should consider PGJ as a different type of treatment versus gastrojejunostomy. Two systematic reviews made the common gastrojejunostomy and endoscopic stenting comparisons; the two that studied PGJ compared it with only gastrojejunostomy [7] [8] [22] [23]. One systematic review used three arms, EUS-GE, gastrojejunostomy, and endoscopic stenting, but provided no results regarding PGJ [24]. Moreover, to our knowledge, no RCT or non-RCT has compared EUS-GE with PGJ, and no study has included four treatments. Thus, we employed a network strategy to overcome the limitations of conventional meta-analyses in comparing multiple treatments. We hope our findings help clinicians and patients to optimize treatment selection.

We determined the sequence of treatments from convenient and simple to complex and challenging to be endoscopic stenting, gastrojejunostomy, PGJ, and EUS-GE. Some advantages (resumption of regular oral intake sooner and shorter LOS, including same-day discharge) make endoscopic stenting the most convenient and preferable strategy [8]. However, in addition to stent migration, tumor ingrowth and outgrowth may lead to stent blockage and require reintervention, which could increase LOS and adversely affect quality of life [25] [26]. The conventional treatment, gastrojejunostomy, remains indicated for palliation in GOO because it can be performed in hospitals without advanced endoscopy teams [23]. The primary concern regarding gastrojejunostomy is that it often results in delayed gastric emptying and a longer duration until the patient can resume regular food intake, which can result in prolonged recovery, poor nutrition, and a potential delay in chemotherapy [8] [23]. The stomach partitioning in PGJ creates the anastomosis far from the tumor site, enabling oral food intake [7] [27]. The remaining 2–3-cm tunnel in the lesser curvature facilitates further endoscopy for re-evaluation or reintervention [27]. In the case of a successful anticancer treatment response, subsequent tumor resection can be shortened due to simple proximal margin resection at the remaining tunnel. Although PGJ is feasible, it remains rarely performed because of a lack of supporting evidence from prospective studies and RCTs [7]. EUS-GE is a promising approach with a high clinical success rate and low reintervention rate [24]. This technique is technically challenging due to jejunal motility during movement of the echoendoscope, which can result in misdeployment of the lumen-apposing metal stent, intestinal leakage, or perforation; the procedure requires expert hands and training in advanced endoscopy [28]. Our findings suggest that PGJ and EUS-GE have higher probabilities of being simultaneously safe and effective than do gastrojejunostomy and endoscopic stenting, possibly because the anastomoses in PGJ and EUS-GE are far from the obstruction site, resulting in the maintenance of the gastrointestinal tract and maximal patency. However, EUS-GE remains limited by its requirement for advanced endoscopic equipment and well-trained experts. Therefore, PGJ should be the alternative choice to EUS-GE in hospitals that lack an advanced endoscopy team; in addition, PGJ could be the first rescue option in cases of treatment failure following EUS-GE or endoscopic stenting.

The National Comprehensive Cancer Network (NCCN), European Society for Medical Oncology (ESMO), the National Institute for Health and Care Excellence (NICE), Japanese, and Korean treatment guidelines for pancreatic and gastric cancer are inconsistent regarding malignant GOO. Three sets of guidelines for treating pancreatic cancer (NCCN, NICE) and gastric cancer (NCCN) advocate for gastrojejunostomy over endoscopic stenting for patients who are fit for surgery and have an overall survival expectation of more than 3 months [29] [30] [31]. However, the ESMO and Korean guidelines for pancreatic cancer favor endoscopic stenting over gastrojejunostomy because endoscopic stenting has a lower complication rate and results in shorter hospitalization than gastrojejunostomy [32] [33]. Only the Japanese Gastric Cancer Association proposes palliative gastrectomy in certain conditions and indicates no preference for gastrojejunostomy [34]. The other guidelines suggest applying a multidisciplinary approach to help patients decide on treatment [29] [30] [35] [36]. The results of this review indicate that PGJ and EUS-GE should be considered for recommendation in such guidelines. In addition to ongoing RCTs comparing gastrojejunostomy vs. PGJ and EUS-GE vs. endoscopic stenting, RCTs or other well-designed studies comparing EUS-GE vs. PGJ (with an open or laparoscopic approach) and focusing on long-term outcomes are necessary.

Our study has several limitations. The inclusion of non-RCTs introduced the possibility of bias due to unmeasured confounding, and increased the overall uncertainty. The recent introduction of EUS-GE and the unpopularity of PGJ resulted in limited evidence for synthesis; in particular, we had no RCT results for EUS-GE. We excluded unbalanced baseline cohort studies to reduce the bias from confounding and the selection process. Publication year also affected overall certainty because the available technology and techniques change over time. The results of our subgroup analysis of studies published after the median publication year largely agree with those of our main analysis. Subgroup analysis still showed high or substantial heterogeneity in included studies in secondary outcomes, and therefore further well-designed studies that could address this problem are warranted. Owing to the lack of available data, we did not perform anastomosis or stent patency, cost comparison, and subgroup analysis based on technical issue within each intervention (laparoscopic and open gastrojejunostomy or PGJ, anastomosis type, stent type, endoscopic approach for EUS-GE). The clinical definitions and grading systems for complication severity are inconsistent between surgery and endoscopy. We combined data from both endoscopic and surgical procedures for investigation and encourage further cooperation among gastroenterological groups.

In conclusion, the results of this NMA suggest that PGJ and EUS-GE should be indicated for malignant GOO to optimize efficacy and safety. PGJ could be the alternative choice in centers with limited resources or in patients who are unsuitable for EUS-GE. Additional studies should directly compare EUS-GE and PGJ, and the effects of open and laparoscopic approaches.

Conflict of Interest

The authors declare that they have no conflict of interest.

Acknowledgement

This manuscript was edited by Wallace Academic Editing.

• References

• 1 Bronswijk M, Vanella G, van Malenstein H. et al. Laparoscopic versus EUS-guided gastroenterostomy for gastric outlet obstruction: an international multicenter propensity score-matched comparison (with video). Gastrointest Endosc 2021; 94: 526-536.e522
• 2 Jang S, Stevens T, Lopez R. et al. Superiority of gastrojejunostomy over endoscopic stenting for palliation of malignant gastric outlet obstruction. Clin Gastroenterol Hepatol 2019; 17: 1295-1302.e1291
• 3 Del Piano M, Ballarè M, Montino F. et al. Endoscopy or surgery for malignant GI outlet obstruction?. Gastrointest Endosc 2005; 61: 421-426 DOI: 10.1016/s0016-5107(04)02757-9. (PMID: 15758914)
• 4 Fiori E, Lamazza A, Demasi E. et al. Endoscopic stenting for gastric outlet obstruction in patients with unresectable antro pyloric cancer. Systematic review of the literature and final results of a prospective study. The point of view of a surgical group. Am J Surg 2013; 206: 210-217
• 5 Jeurnink SM, Steyerberg EW, van Hooft JE. et al. Surgical gastrojejunostomy or endoscopic stent placement for the palliation of malignant gastric outlet obstruction (SUSTENT study): a multicenter randomized trial. Gastrointest Endosc 2010; 71: 490-499 DOI: 10.1016/j.gie.2009.09.042. (PMID: 20003966)
• 6 Singh A, Ross WA, Bhattacharya A. et al. Gastrojejunostomy versus enteral self-expanding metal stent placement in patients with a malignant gastric outlet obstruction. Gastrointestinal Intervention 2013; 2: 94-98
• 7 Kumagai K, Rouvelas I, Ernberg A. et al. A systematic review and meta-analysis comparing partial stomach partitioning gastrojejunostomy versus conventional gastrojejunostomy for malignant gastroduodenal obstruction. Langenbecks Arch Surg 2016; 401: 777-785 DOI: 10.1007/s00423-016-1470-8. (PMID: 27339200)
• 8 Upchurch E, Ragusa M, Cirocchi R. Stent placement versus surgical palliation for adults with malignant gastric outlet obstruction. Cochrane Database Syst Rev 2018; 5: CD012506 DOI: 10.1002/14651858.CD012506.pub2. (PMID: 29845610)
• 9 Chandan S, Khan SR, Mohan BP. et al. EUS-guided gastroenterostomy versus enteral stenting for gastric outlet obstruction: systematic review and meta-analysis. Endosc Int Open 2021; 9: E496-E504 DOI: 10.1055/a-1341-0788. (PMID: 33655056)
• 10 Efthimiou O, Mavridis D, Debray TP. et al. Combining randomized and non-randomized evidence in network meta-analysis. Stat Med 2017; 36: 1210-1226 DOI: 10.1002/sim.7223. (PMID: 28083901)
• 11 Kim MS, An MH, Kim WJ. et al. Comparative efficacy and safety of pharmacological interventions for the treatment of COVID-19: a systematic review and network meta-analysis. PLoS Med 2020; 17: e1003501 DOI: 10.1371/journal.pmed.1003501. (PMID: 33378357)
• 12 Shi J, Luo D, Weng H. et al. Optimally estimating the sample standard deviation from the five-number summary. Res Synth Methods 2020; 11: 641-654 DOI: 10.1002/jrsm.1429. (PMID: 32562361)
• 13 Papakonstantinou T, Nikolakopoulou A, Higgins JPT. et al. CINeMA: software for semiautomated assessment of the confidence in the results of network meta-analysis. Campbell Syst Rev 2020; 16: e1080 DOI: 10.1002/cl2.1080. (PMID: 37131978)
• 14 Sterne JAC, Savović J, Page MJ. et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ 2019; 366: l4898 DOI: 10.1136/bmj.l4898. (PMID: 31462531)
• 15 Sterne JA, Hernán MA, Reeves BC. et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ 2016; 355: i4919 DOI: 10.1136/bmj.i4919. (PMID: 27733354)
• 16 Egger M, Davey Smith G, Schneider M. et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315: 629-634
• 17 Balduzzi S, Rücker G, Nikolakopoulou A. et al. Netmeta: an R package for network meta-analysis using frequentist methods. J Stat Software 2023; 106: 1-40
• 18 Higgins JPT, Jackson D, Barrett JK. et al. Consistency and inconsistency in network meta-analysis: concepts and models for multi-arm studies. Res Synth Methods 2012; 3: 98-110 DOI: 10.1002/jrsm.1044. (PMID: 26062084)
• 19 Rücker G, Schwarzer G. Ranking treatments in frequentist network meta-analysis works without resampling methods. BMC Med Res Methodol 2015; 15: 58 DOI: 10.1186/s12874-015-0060-8. (PMID: 26227148)
• 20 Fiori E, Crocetti D, Sapienza P. et al. Palliative surgery or metallic stent positioning for advanced gastric cancer: differences in QOL. Medicina (Kaunas) 2021; 57: 428 DOI: 10.3390/medicina57050428. (PMID: 33925171)
• 21 Mehta S, Hindmarsh A, Cheong E. et al. Prospective randomized trial of laparoscopic gastrojejunostomy versus duodenal stenting for malignant gastric outflow obstruction. Surg Endosc 2006; 20: 239-242 DOI: 10.1007/s00464-005-0130-9. (PMID: 16362479)
• 22 Bian S-B, Shen W-S, Xi H-Q. et al. Palliative therapy for gastric outlet obstruction caused by unresectable gastric cancer: a meta-analysis comparison of gastrojejunostomy with endoscopic stenting. Chin Med J 2016; 129: 1113-1121
• 23 Lorusso D, Giliberti A, Bianco M. et al. Stomach-partitioning gastrojejunostomy is better than conventional gastrojejunostomy in palliative care of gastric outlet obstruction for gastric or pancreatic cancer: a meta-analysis. J Gastrointest Oncol 2019; 10: 283-291 DOI: 10.21037/jgo.2018.10.10. (PMID: 31032096)
• 24 Boghossian MB, Funari MP, De Moura DTH. et al. EUS-guided gastroenterostomy versus duodenal stent placement and surgical gastrojejunostomy for the palliation of malignant gastric outlet obstruction: a systematic review and meta-analysis. Langenbecks Arch Surg 2021; 406: 1803-1817 DOI: 10.1007/s00423-021-02215-8. (PMID: 34121130)
• 25 Maetani I, Akatsuka S, Ikeda M. et al. Self-expandable metallic stent placement for palliation in gastric outlet obstructions caused by gastric cancer: a comparison with surgical gastrojejunostomy. J Gastroenterol 2005; 40: 932-937
• 26 Maetani I, Tada T, Ukita T. et al. Comparison of duodenal stent placement with surgical gastrojejunostomy for palliation in patients with duodenal obstructions caused by pancreaticobiliary malignancies. Endoscopy 2004; 36: 73-78
• 27 Kaminishi M, Yamaguchi H, Shimizu N. et al. Stomach-partitioning gastrojejunostomy for unresectable gastric carcinoma. Arch Surg 1997; 132: 184-187 DOI: 10.1001/archsurg.1997.01430260082018. (PMID: 9041924)
• 28 Ghandour B, Bejjani M, Irani SS. et al. Classification, outcomes, and management of misdeployed stents during EUS-guided gastroenterostomy. Gastrointest Endosc 2022; 95: 80-89
• 29 National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) Pancreatic Adenocarcinoma Version 2.2022 – 6 December 2022. Accessed April 24, 2024 at: https://www.nccn.org/guidelines/guidelines-detail?category=1&id=1455
• 30 National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) Gastric Cancer Version 1.2023 – 10 March 2023. Accessed April 24, 2024 at: https://www.nccn.org/guidelines/guidelines-detail?category=1&id=1434
• 31 The National Institute for Health and Care Excellence of the United Kingdom. Pancreatic cancer in adults: diagnosis and management NICE guideline [NG85]. Accessed April 24, 2024 at: https://www.nice.org.uk/guidance/ng85
• 32 Ducreux M, Cuhna AS, Caramella C. et al. Cancer of the pancreas: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 2015; 26 (Suppl. 5) v56-68 DOI: 10.1093/annonc/mdv295. (PMID: 26314780)
• 33 Committee of the Korean clinical practice guideline for pancreatic cancer and National Cancer Center, Korea. Korean clinical practice guideline for pancreatic cancer 2021: a summary of evidence-based, multi-disciplinary diagnostic and therapeutic approaches. Pancreatology 2021; 21: 1326-1341 DOI: 10.1016/j.pan.2021.05.004. (PMID: 34148794)
• 34 Japanese Gastric Cancer Association. Japanese Gastric Cancer Treatment Guidelines 2021 (6th edition). Gastric Cancer 2023; 26: 1-25
• 35 Okusaka T, Nakamura M, Yoshida M. et al. Clinical practice guidelines for pancreatic cancer 2019 from the Japan Pancreas Society: a synopsis. Pancreas 2020; 49: 326-335
• 36 Oesophago-gastric cancer: assessment and management in adults. NICE guideline [NG83]. Updated 4 July 2023. Accessed April 24, 2024 at: https://www.nice.org.uk/guidance/ng83

Publication History

Received: 04 December 2023

Accepted after revision: 19 April 2024

Accepted Manuscript online:
19 April 2024

Article published online:
31 May 2024