Introduction
Enteral access through a gastrostomy tube is often required in patients unable to maintain adequate volitional intake of food. The most common techniques for performing a gastrostomy include percutaneous endoscopic gastrostomy (PEG) and percutaneous radiological gastrostomy (PRG) [1]. PEG is often performed by gastroenterologists and surgeons using an endoscope placed within the stomach [2]. PRG, on the other hand, is typically performed by radiologists using fluoroscopy and administration of oral contrast. Both techniques are considered safe and effective [3]. However, in general, gastrostomy tube placement is associated with a significant risk of adverse outcomes, including perforation, infections, peritonitis, and death.
There are conflicting data regarding the comparative safety of PEG and PRG. Most of the data are restricted to single-center retrospective analyses that were underpowered to conclusively determine the comparative outcomes and adverse events [4]. Hence, the preferred technique for safely performing a gastrostomy remains undetermined.
A prior systematic review and meta-analysis by Strijbos et al was limited to 16 studies reporting data on only 934 PEG and 1093 PRG procedures [4]. The authors noted no difference for 30-day mortality or infectious complications, whereas tube-related complications were higher in the PRG group. However, recent studies from large, nationally representative databases encompassing multiple hospitals nationwide have reported a statistically significantly greater risk of 30-day mortality for PRG compared to PEG [1,5]. Given the significant discrepancy in the results and the availability of newer high-quality studies, we performed this systematic review and meta-analysis to compare the risk of adverse outcomes between PEG and PRG. In addition, we performed a Grading of Recommendations, Assessment, and Evaluation (GRADE) analysis to assess the certainty of evidence.
Materials and methods
This systemic review and meta-analysis were performed in accordance with the Preferred Reporting Items for Systemic Reviews and Meta-Analyses (PRISMA) Statement [6].
Data sources and literature search
A comprehensive search strategy was developed to search for appropriate studies for this systemic review and meta-analysis. We queried PubMed, EMBASE and Cochrane from inception through 21st January 2022 to identify citations comparing outcomes of PEG and PRG. Detailed search queries for individual databases are delineated in Supplementary Table 1.
Study selection and data extraction
Randomized controlled trials, cohort studies (retrospective or prospective), and case-control studies comparing PEG with PRG were identified. We excluded duplicate studies, conference abstracts, editorials, short research letters, studies in other languages than English, and studies assessing outcomes of percutaneous endoscopic jejunostomy, PEG with jejunal extension [7] or PRG with jejunal extension. Two reviewers (DR and DK) independently performed an eligibility assessment. Any disagreement was resolved through consultation with a third author (PS).
A standard data extraction excel sheet was created using Microsoft Excel, and data were independently extracted by 2 reviewers (DR and DK). Extracted items included first author, study year, study design, patients in each arm, indication for procedure, number of study centers, country, type of PEG (pull, direct, push), antibiotic prophylaxis, mean age or median age of patients with standard deviation or range, percentages of males in each group, and outcomes of interests.
Quality assessment
The modified Newcastle-Ottawa Scale (NOS) and Jadad quality scale were used for observational studies and randomized clinical trials, respectively [4,8,9], to assess the quality of eligible studies identified after initial extraction. Quality assessment was performed individually for each study. Studies with NOS scores of less than seven and Jadad scores of less than 4 were deemed low-quality studies.
The certainty of the evidence was assessed using the GRADE approach [10]. Two independent researchers (DK and MD) graded the risk of bias, indirectness, inconsistency, imprecision, and publication bias [11]. A ”Summary of Findings” table was created and the certainty of evidence for the systematic review of interventions was graded as “high”, “moderate”, “low” or “very low”, in compliance with the Cochrane Handbook, as well as the GRADE guidelines, using GRADEPro (McMaster University, 2015; developed by Evidence Prime, Inc.).
Primary and secondary outcomes
The primary outcome of interest was the 30-day mortality rate associated with PEG and PRG. Our secondary outcomes included the following: colon perforation, peritonitis, post-procedural bleeding, technical failure of the procedure, infectious complications (peristomal infections or aspiration pneumonia), tube-related complications (such as dislocation, leak, obstruction, or breakdown), readmissions, cost of the procedure, procedure times, and tumor seeding. The all-cause 30-day mortality, colon perforation and peritonitis were considered outcomes of critical importance during the GRADE analysis.
Statistical analysis
Pooled effect sizes were analyzed to examine the impact of PEG vs. PRG for primary and secondary outcomes. We examined heterogeneity among the outcomes via a sensitivity analysis. A subgroup analysis assessing high-quality studies was performed to examine the impact of high-quality evidence. We also performed a “leave-one-out” meta-analysis as a sensitivity analysis to identify outlier studies affecting outcomes. Meta-analyses were only performed if 2 or more studies were available for each outcome. Continuous variables were analyzed using the mean difference, while the odds ratio (OR) was calculated for dichotomous variables. Under the prior assumption that studies from different populations worldwide would be heterogenous, we used a random-effects model with the Mantel-Haenszel method to calculate the OR. Prediction intervals are reported along with pooled effect sizes to reflect the uncertainty expected in the summary effect. Studies with zero in both the intervention and control arms were not included in pooled effect size per standard meta-analysis practice, because such studies do not indicate either the direction or the magnitude of the relative treatment effect. Statistical heterogeneity among different studies was determined using the I2 statistic. The threshold of I2 statistics being more than 50% was used to define substantial heterogeneity among the studies. Funnel plot visualization and the test of funnel plot asymmetry (i.e., Egger’s test) were used to assess the presence of publication bias [12]. We did not assess publication bias if the number of studies for any outcome was less than 10, as the test of funnel-plot asymmetry would lack the statistical power to detect bias in the setting of a small number of studies [13]. “Leave-one-out” meta-analysis and test of publication bias were performed on high-quality studies. If publication bias was detected, Rucker’s limit meta-analysis method was used to assess for significant effects on pooled analysis. This method builds meta-analysis models explicitly accounting for bias due to small-study effects [14]. P-values <0.05 were considered statistically significant for the meta-analysis and Egger’s test. Statistical analysis was performed using “R” software.
Results
Literature search
Four hundred sixteen citations were identified on the initial literature search. After exclusion of studies based on the selection criteria defined above, 46 studies were reviewed in detail and 33 were included in the final analysis (Fig. 1).
Figure 1 PRISMA diagram
PEG, percutaneous endoscopic gastrostomy; PEG-J, percutaneous endoscopic transgastric jejunostomy; PIG, per-oral image-guided gastrostomy
Population and study characteristics
Of 467,808 patients included in the analysis, 275,117 underwent a PEG procedure (59.6% male; mean age 62.6±2.6 years), whereas 192,691 patients underwent a PRG procedure (58.9% male; mean age 63.9±2.1 years). There were no differences in the sex ratio or mean age between groups.
Indications for percutaneous gastrostomy (PEG or PRG) varied among different studies. Seven studies reported outcomes among patients with motor neuron disease, 3 reported data for patients with head and neck malignancies, and the rest reported a mix of indications. Of 33 studies, 11 were from the USA and 8 from the UK (Table 1). Twenty-five studies were single-center, 6 were multicenter, and 2 studies were based on nationwide administrative databases. Twenty-seven studies were retrospective studies, 5 were prospective nonrandomized cohort studies, and only 1 study was a randomized controlled trial, specifically performed in children. Of 33 studies, 26 were deemed high-quality studies based on quality assessment. A study quality assessment using the modified NOS and Jadad score is provided in Supplementary Table 2.
Table 1 Study and population characteristics
Primary outcome: 30-day all-cause mortality
Twenty-two studies reported the all-cause 30-day mortality rate [1,5,15-34]. The pooled incidence of 30-day mortality was 7.8% in PEG (3279/41,663) and 8% in PRG patients (13,163/165,010). Patients undergoing PEG had 27% lower odds of 30-day mortality compared to PRG patients (OR 0.73, 95% confidence interval [CI] 0.58-0.93; P=0.012; I2=50%, Table 2).
Table 2 Odds ratios for outcomes with percutaneous endoscopic gastrostomy compared to percutaneous radiological gastrostomy
Data from 18 high-quality studies showed a similar pooled incidence of 30-day mortality of 7.87% (3262/41,450) in the PEG vs. 7.97% (13,135/164,774) in the PRG group, with 25% lower odds of 30-day mortality in the PEG group (OR 0.75, 95%CI 0.60-0.95; P=0.02; I2=51%; Fig. 2A). “Leave-one-out” meta-analysis did not demonstrate any significant impact on the pooled OR. No publication bias was observed on the funnel plot, and the test of funnel plot asymmetry was non-significant (Egger’s test P=0.223).
Figure 2 Forest plots and funnel plots for 30-day all-cause mortality, colon perforation, and peritonitis
Secondary outcomes
Colon perforation
Ten studies reported the incidence of colon perforation with PEG and PRG [1,5,17,26,28,34-38]. Pooled incidence rates were 0.11% (310/272,866) in the PEG group and 0.20% (390/190,851) in the PRG group. Patients undergoing PEG had 37% lower odds of colon perforation compared to PRG patients (OR 0.63, 95%CI 0.47-0.86; P=0.008; I2=17%, Fig. 3).
Figure 3 Forest plot for all outcomes
PEG, percutaneous endoscopic gastrostomy; PRG percutaneous radiological gastrostomy
Data from 8 high quality studies showed similar results, with 39% lower odds of colon perforation with PEG vs. PRG (OR 0.61, 95%CI 0.41-0.75; P=0.009; I2=0%, Fig. 2B). “Leave-one-out” meta-analysis did not demonstrate any significant impact on the pooled OR.
Peritonitis
Twelve studies reported the incidence of peritonitis with PEG and PRG [1,17,20-22,26,29,31,32,36,39,40]. Pooled incidence of peritonitis was 1.9% (462/24,462) in the PEG group and 2.7% (277/10351) in the PRG group. Pooled OR showed 29% lower odds of peritonitis with the PEG procedure compared to PRG (OR 0.71, 95%CI 0.62-0.81; P<0.001; I2=0%, Fig. 3).
Data from 9 high-quality studies showed a similar 29% lower incidence of peritonitis with PEG compared to PRG (OR 0.71, 95%CI 0.63-0.81; P<0.001; I2=0%; Fig. 2C). “Leave-one-out” meta-analysis showed that excluding a study by Kohli et al led to a statistically non-significant OR. However, Kohli et al was the largest study in the analysis [1].
Post-procedural bleeding
Sixteen studies reported bleeding as an outcome after procedures [1,5,17,19,21,22,25-27,31,35-37,39-41]. Pooled incidence rates of bleeding were 0.3% (955/273,645) and 0.4% (782/191,118) in the PEG and PRG groups, respectively. Pooled OR did not show any difference in bleeding rate between the PEG and PRG groups (OR 0.77, 95%CI 0.48-1.23; P=0.241; I2=80%).
Technical failure
Thirteen studies reported technical success or failure rates [15,16,18,25,27,31,32,34,38-42]. The pooled incidence of procedure failure rates from all studies were 6.1% (73/1196) and 2.5% (18/705) in the PEG and PRG groups, respectively. Pooled OR did not show any difference in the technical failure rate of PEG and PRG (OR 2.52, 95%CI 0.92-6.89; P=0.068; I2=61%).
Peristomal infection
Twenty-three studies reported the incidence of infections at the gastrostomy site [1,5,19-22,25-32,35-37,39-41,43-45]. Pooled incidence of peristomal infection was 1.2% (3310/274399) and 1.2% (2281/192,135) in the PEG and PRG groups, respectively. Pooled OR was 0.87 (95%CI 0.67-1.13; P=0.268; I2=47%).
Data from 19 high-quality studies demonstrated an OR of 0.93 (95%CI 0.68-1.27; P=0.625; I2=54%). Publication bias was observed on the funnel plot, and the test of funnel plot asymmetry was significant (P=0.03). Rucker’s limit meta-analysis showed non-significant changes in adjusted (OR 0.73, 95%CI 0.50-1.08; P=0.114) and unadjusted (OR 0.93, 95%CI 0.68-1.27; P=0.625) OR for small-study effects. “Leave-one-out” meta-analysis did not demonstrate any significant impact on the pooled OR.
Aspiration pneumonia
Eight studies reported aspiration pneumonia as an outcome [21,22,25,27,31,32,34,36]. Pooled incidence rates of aspiration pneumonia were 1.2% (12/976) and 3.4% (17/495) in the PEG and PRG groups, respectively. Pooled OR did not show any difference in aspiration pneumonia rate between the 2 procedures (OR 0.58, 95%CI 0.28-1.21; P=0.114; I2=0%).
Tube-related complications
Twenty-two studies reported tube-related complication as an outcome [1,5,17-22,24,29-32,35-37,39-41,43-45]. Pooled incidence rates were 2.8% (7705/274,214) and 3.5% (6791/191,991) in the PEG and PRG groups, respectively. Pooled OR was 0.56 (95%CI 0.33-0.95; P=0.032; I2=95%).
Data from 19 high-quality studies demonstrated an OR of 0.60 (95%CI 0.35-1.09, P=0.07; I2=96%). No publication bias was observed on the funnel plot, and the test of funnel plot asymmetry was non-significant (P=0.758). “Leave-one-out” meta-analysis demonstrated that omission of studies by Cherian et al [17] or Desport et al [19] led to a statistically lower OR of complications with PEG. Both studies reported a statistically significantly greater incidence of tube-related complications with PEG on univariate analysis. However, on a subsequent multivariate analysis, Cherian et al [24] reported an OR of 5.4 for tube dislodgement with PRG compared to PEG, and noted that the method of gastrostomy insertion was the only predictor for dislodgement. Multivariate analysis was not performed in a study by Desport et al [26]. Exclusion of both studies from the meta-analysis resulted in statistically significantly fewer tube-related adverse events with PEG (OR 0.46, 95%CI 0.29-0.74; P=0.003; I2=95%).
30-day all-cause readmission
Two high-quality studies reported 30-day readmission rates after the procedures [5,34]. Pooled 30-day readmission rates were 17% (2817/16,431) and 19% (29,649/154,052) in the PEG and PRG groups, respectively. Pooled OR did not show any difference in 30-day readmission rate between the 2 arms (OR 14.8, 95%CI 0.006-76.82; P=0.480; I2=63.7%).
Cost
Three studies reported the cost of procedures [17,21,46]. Because of differences in currencies, pooled analyses could not be performed. One low-quality study reported $1861 as the cost for PEG and $1985 for PRG [46]. The 2 high-quality studies reported the following mean cost: Australian dollars (AUD) 1200 for PEG vs. 1366 for PRG, Canadian dollar (CAD) 591 CAD for PEG vs. 407 for PRG [17,21].
Procedure times
Three studies, including 2 high-quality studies, reported the procedure time [32,34,43]. Only 2 studies reported procedure times with mean and standard deviation. Pooled analysis did not demonstrate any difference in procedure time (mean difference: -0.07 min, 95%CI -10.5-9.1; P=0.882; I2=99.7%).
Tumor seeding
Only one study reported tumor seeding events as an outcome. This low-quality study reported 0 events in both groups [40].
Assessment of certainty of evidence by GRADE
The certainty of the evidence was assessed for each outcome using the GRADE methodology (Table 2). Most of the studies were retrospective cohort studies. More than 75% of the studies were high-quality based on the NOS, without any definite selection or attrition bias. Outcomes such as all-cause 30-day mortality, colon perforation, peritonitis, and technical failure of procedure were considered to have uniform definitions across different studies. However, mechanical complications of the feeding tube were not defined uniformly across the various studies.
The certainty of the evidence was downgraded due to serious imprecision noted in the outcome of peritonitis, since one study [1] demonstrated significantly higher odds of peritonitis with PRG, while the other studies did not (Table 3). The certainty of the evidence was upgraded in colon perforation and peritonitis because of a large effect from individual studies. Since peritonitis and colon perforation are relatively uncommon serious adverse events, small single-center studies are likely to be underpowered to detect a significant difference in the incidence. In the outcome of peritonitis, 97% of the weightage can be ascribed to a single, large multicenter study [1] with adequate power to detect differences in outcomes. Similarly, 3 large studies [1,5,37], which together provide 97% of the total weightage, report a higher OR of perforation with PRG. Overall, the certainty of the evidence was considered “low” for all outcomes except colon perforation, for which the certainty was considered “moderate”.
Table 3 GRADE analysis for certainty
Discussion
A percutaneous gastrostomy for a feeding tube is a frequently performed procedure [1,5]. There are endoscopic, radiological, and surgical techniques for performing a gastrostomy, but the surgical technique is used less frequently [5]. However, endoscopic and radiological techniques are widely used, and data regarding comparative outcomes have recently been published. While PEG and PRG are complementary techniques [1], significant differences in outcomes need to be considered when a particular technique is chosen. This systematic review and meta-analysis of 33 studies, including more than 450,000 patients, demonstrated that PEG was associated with a lower risk of 30-day all-cause mortality, colon perforation and peritonitis, compared to PRG. Notably, the 2 techniques have comparable rates of technical failure.
Different studies have variously demonstrated a higher technical success rate for either PEG [27] or PRG [15], or no significant difference [34]. However, the significant heterogeneity in these primarily retrospective studies limited the quality of the evidence. Further, in some studies, it is unclear if PEG or PRG was selected after the failure of the other technique [5]. This meta-analysis demonstrates that PEG and PRG have comparable technical failure/success rates. Notably, the most widely used technique for performing a PEG is the peroral “pull” technique which necessitates the passage of the feeding tube through the mouth and esophagus [47]. This technique may not be feasible in patients with stenotic malignancies and those at risk of implantation metastasis [48], thus contributing to the failure of the PEG technique. Additionally, PEG requires significantly higher doses of sedatives and analgesics than PRG [34]. These limitations may possibly be overcome using a direct PEG technique [44] and propofol-based sedation.
This study demonstrates that PEG is associated with lower 30-day all-cause mortality compared to PRG. This may possibly be explained, in part, by a higher incidence of colon perforation and peritonitis with PRG [1,37]. However, there is also a possible selection bias, whereby sicker patients with probably shorter life expectancy are preferentially referred for a PRG. Indeed, recent studies have demonstrated a statistically significant preference for PRG in patients with malignant causes of dysphagia, such as malignancies of the head and neck. In contrast, PEG is preferentially used in benign causes of dysphagia such as stroke [1]. However, 2 recent, large multicenter studies, with over 160,000 patients, have demonstrated higher odds of 30-day all-cause mortality with PEG after incorporating the presence of malignancy and the Charlson comorbidity index scores [1,5]. Despite the impact of presumed selection bias, high 30-day mortality highlights the dilemma in selecting patients likely to benefit from a feeding tube.
This meta-analysis demonstrated a lower risk for colon perforation and peritonitis with PEG than PRG. The lower risk of colon perforation and peritonitis with PEG has been ascribed to safe endoscopic techniques, such as transillumination, finger indentation, and the “safe-tract” method of using the introducer needle [5]. PRG also uses techniques such as contrast administration to delineate the gastric rugae vs. the colon, but further interventions may be needed to decrease the risk of colon perforation and peritonitis.
It has previously been suggested that patients undergoing a PEG may be at a higher risk for bleeding, since PEG is preferentially undertaken in patients with a stroke typically on dual antiplatelet medications [1]. This meta-analysis, however, demonstrated a relatively low risk of bleeding with a gastrostomy, without a statistically significant difference between PRG and PEG. Notably, it is unclear how many patients who bled were on antiplatelet and/or anticoagulant medications.
Peristomal infection is a common adverse outcome of a gastrostomy. Therefore, the American Society for Gastrointestinal Endoscopy recommends administering periprocedural antibiotics during the procedure [49]. However, the Society for Interventional Radiology until recently recommended prophylactic antibiotics only for the “pull” gastrostomy, but did not have a recommendation for antibiotics for the transabdominal “push” gastrostomy [3]. This may have led some physicians to not administer antibiotics during a PRG [34]. The recent 2018 guidelines, however, recommend prophylactic antibiotics for both types of gastrostomy techniques [50]. Our study did not demonstrate a significant difference in the odds of infection. The source studies did not explicitly state whether antibiotics were administered or not.
Notably, stratification of data based on the administration of antibiotics, type of technique, and extent of follow up was not feasible. Similarly, this meta-analysis did not demonstrate a statistically significant difference in the incidence of mechanical complications of a PEG and PRG. However, there was significant variation in the types of feeding tubes and definitions of complications used by individual investigators.
The strengths of this study include a large dataset from multiple sites, use of a validated quality assessment tool, utilization of “leave-one-out” meta-analysis to assess for outliers, and application of the GRADE process to evaluate the strength of our analysis. Indeed, a recent meta-analysis on this subject reported data on 2027 patients and demonstrated no difference in mortality [4]. However, this current meta-analysis reports the data of 467,808 patients from 33 studies. The study was limited by the retrospective study design of most individual studies, lack of uniform definitions across different studies, and an over-representation of 2 large nationwide studies recently published by our group [1,5]. The analysis was also limited by a lack of ancillary data relevant for specific outcomes, such as the use of antibiotics in relation to the incidence of peristomal infections. Finally, “pull” and direct/”push” gastrostomy techniques were analyzed as a single group, since very few studies reported data on irect/”push” gastrostomies. Notably, despite the low certainty of the evidence for most outcomes, these data represent the best available evidence comparing the risks of PEG and PRG.
In conclusion, this systematic review and meta-analysis demonstrated that PEG was associated with a significantly lower risk of 30-day all-cause mortality, colon perforation and peritonitis, compared to PRG, while having a comparable technical failure rate. Therefore, PEG should be considered the first-line technique for obtaining enteral access in appropriate patients.
What is already known:
-
Percutaneous endoscopic (PEG) and radiological (PRG) gastrostomy are complimentary procedures performed for enteral access
What the new findings is:
-
In our systematic review and meta-analysis of 33 studies, PEG was associated with significantly lowerodds of colon perforation, peritonitis, and all-causemortality at 30 days compared to PRG