Introduction

Barrett’s esophagus (BE) is an acquired condition of the esophagus in which the normal distal squamous epithelial lining is replaced by the metaplastic columnar epithelium, which is visible endoscopically (≥ 1 cm) above the gastroesophageal junction and confirmed histopathologically from esophageal biopsies [1]. Long-standing gastroesophageal reflux disease (GERD) leads to chronic inflammatory changes in the lower esophagus due to exposure to gastric contents, which may eventually lead to metaplastic transformation [2, 3]. GERD in patients with BE appears to be more severe and more frequently associated with complications like stricture, ulcer, or dysplasia than in patients without columnar mucosa [3]. In spite of the more severe reflux, symptoms of GERD are often less pronounced or even completely absent in patients with BE due to decreased sensitivity of the columnar epithelium to gastric contents. One of the major concerns about BE is its risk of malignant transformation to dysplasia and esophageal adenocarcinoma (EAC). This risk is dependent on the length of BE and long-segment Barrett’s esophagus (lsBE), defined as maximal length of BE > 3 cm, and carries a higher risk of malignant transformation than short-segment Barrett’s esophagus (ssBE), defined as BE length of < 3 cm [1, 4]. The estimated annual rate of progression of BE to EAC in the western world is 0.07%/year for ssBE and 0.25%/year for lsBE [4].

Another risk factor common to both BE and EAC is obesity. The prevalence of BE is between 0.5 and 2% in an unselected population; however, in those with GERD, the prevalence is higher ranging between 5 and 15% [5, 6]. Studies have confirmed that GERD is approximately twofold more common and BE three times more frequent in the obese population compared with individuals with normal weight [7, 8]. Furthermore, obesity (BMI > 30 kg/m [2]) increases the risk EAC by a relative risk of 2.4 to 2.8 [9, 10]. Current guidelines on BE advocates regular endoscopic screening with esophageal biopsies to detect field changes or a possible transformation to EAC at an early stage [1, 11].

Roux-en-Y gastric bypass (RYGB) is the second commonest bariatric procedure worldwide [12]. Though its role in the management of obesity and GERD is well established, its effect on BE is not yet fully understood. It also appears to be more effective than sleeve gastrectomy, now the most commonly performed bariatric procedure, in the resolution of GERD [13]. In spite of its well-defined role in inducing weight loss and improving GERD in patients with obesity, there is a paucity of literature on its effect on BE. It can be hypothesized that by reducing reflux, it can potentially lead to regression of BE. Existing literature shows a positive effect of RYGB on BE by inducing its regression, but most studies evaluating this effect are of low quality with few numbers of patients [14, 15]. Lack of routine endoscopic evaluation of the esophagus and stomach prior to bariatric surgery in many bariatric units further makes it extremely difficult to identify patients with BE preoperatively.

The aim of this systematic review and meta-analysis was to evaluate the effect of RYGB on BE.

Methods

Eligibility Criteria

The research question was formulated based on Population, Intervention, Comparison, Outcome, and Study design (PICOS) strategy [16, 17]. The population comprised adult patients undergoing RYGB with known BE diagnosed preoperatively on endoscopy and confirmed histologically. Intervention was RYGB done for the treatment of obesity or obesity-related comorbidities. Comparators were preoperative confirmation of BE on endoscopy and biopsy and postoperative response to RYGB with endoscopy and biopsy at least 1 year after surgery. Outcome of interest was regression of BE.

Data Search and Extraction

Published literature was searched for using MEDLINE, PubMed, Scopus (including Embase), Cochrane Central Register of Controlled Trials (CENTRAL), OpenGray, ScienceDirect, and SpringerLink databases without language restrictions till January 31, 2019. Manual searches were carried out for articles in relevant journals and reference lists in key articles. The retrieved studies were managed using the software, Zotero Standalone (version 4.0.29.10) for Windows (Centre for History and New Media, George Mason University, Fairfax, Virginia, USA). The Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines was followed at the time of writing this manuscript [18]. MeSH terms used were “gastric bypass” OR “bariatric surgery” AND “Barrett’s esophagus” OR “Gastresophageal reflux.”

Studies evaluating the effect of RYGB on BE objectively with a preoperative confirmation of BE endoscopically and histologically and assessing the postoperative response to surgery with a repeat endoscopy at > 1 year after surgery were considered for the meta-analysis. Regression of BE was defined as histological regression of intestinal metaplasia to normal squamous mucosa, or from dysplasia to intestinal metaplasia or normal squamous mucosa on postoperative endoscopic biopsy, or disappearance of BE or reduction in the length of BE by at least 3 cm on postoperative endoscopy. Subanalysis based on the length of BE (ssBE [< 3 cm) and lsBE [> 3 cm]) was done to evaluate the effect of RYGB on ssBE and lsBE respectively using the same eligibility criteria. Subanalysis was also done to evaluate the effect of RYGB on associated GERD in the selected cohort of patients with BE. Improvement in GERD after RYGB was defined as improvement in symptoms of reflux after surgery, or improvement in endoscopic stigmata of GERD such as esophagitis, esophageal ulcer, and peptic stricture, or improvement in GERD demonstrated by pH study. Progression of BE was defined as dysplastic or cancerous transformation of intestinal metaplasia on postoperative endoscopic biopsy or an increase in the length of BE by 1 cm.

Case reports and reviews, studies reporting BE at only one time point (either preoperative or postoperative) such that a comparison of the outcome before and after RYGB could not be made, studies evaluating the effect of other bariatric procedures on BE, studies evaluating the effect of RYGB on GERD only without evaluation of BE, or data on intestinal metaplasia of the cardia and not BE, or studies with data not suitable for meta-analysis of the effect of RYGB on BE were excluded.

Selection of studies was done by two independent reviewers utilizing a blinded model. Studies with post-RYGB data at one or more than one time point were also included for comparison with pre-RYGB data. Disagreement in the selection of studies between the two reviewers was resolved with consensus and additional opinion utilizing Cohen’s kappa statistical model [19].

Data was extracted by four reviewers into tables using SPSS (version 24) for Windows (International Business Machines Corporation). All attempts were made to retrieve missing information as per the study protocol by contacting the primary authors of the included studies.

Assessment of Risk of Bias

Assessment of risk of bias in the included studies was done by three reviewers utilizing the methodological index for non-randomized studies (MINORS) criteria [20]. Assessment of risk of bias was blinded, and all disagreements were resolved with consensus and opinion from additional reviewers utilizing Cohen’s kappa statistical model [19].

Statistical Analysis

Outcome of the meta-analysis was measured using forest plots [21]. Funnel plots were used to draw conclusions about publication bias and heterogeneity at different time frames [22]. These were formulated with respect to individual outcome measures rather than to assess the overall quality of the studies included in the meta-analysis. Funnel plots and forest plots were generated using the software, Review Manager (RevMan version 5.3.5) (Cochrane Collaboration, Oxford, UK) [21, 22]. The experimental group of the meta-analysis was BE after RYGB, and the control group was BE before RYGB. The magnitude of the experimental effect was calculated in terms of risk difference (RD) and confidence intervals set at 95% to reflect a significance level of 0.05. The same statistical principles were applied for subgroup analysis done to evaluate the effect of RYGB based on the length of BE (ssBE [< 3 cm] and lsBE [> 3 cm]), and on GERD.

Regression of BE was also compared with progression of BE in the selected group of sample and data presented as odds ratio (OR) with confidence intervals set at 95% to establish significance at < 0.05.

Tests for heterogeneity using tau [2], Cochran’s Q (chi [2]), and I [2] are incorporated in the forest plots [23]. Studies were assumed to be heterogeneous at the outset; hence, a Mantel-Haenszel, random-effects model was used for meta-analysis instead of the fixed-effect model to allow the outcome measures to vary in a normal distribution between studies.

Data Availability

All data generated or analyzed during this study are included in this published article.

Results

Description of the Studies

Figure 1 summarizes the flow of information in different phases of the review. A Cohen’s kappa score of 1 was achieved between the two primary reviewers.

Fig. 1
figure 1

Flow of information in different phases of the systematic review and meta-analysis. RYGB, roux-en-Y gastric bypass; BE, Barrett’s esophagus

Full size image

Eight studies [24,25,26,27,28,29,30,31] with more than 10,779 patients undergoing RYGB identified a total of 117 patients suitable for meta-analysis of the effect of RYGB on BE. Six out of the eight studies [24, 25, 27, 28, 30, 31] categorized the effect of RYGB based on the length of BE (ssBE or lsBE). While all eight studies evaluated the effect of RYGB on BE, four studies evaluated the effect on ssBE [24, 28, 30, 31], five studies on lsBE [24, 25, 27, 28, 30], and five studies on GORD [24, 25, 27,28,29]. Regression of BE was analyzed in terms of histological regression from intestinal metaplasia to cardiac mucosa in all the included studies [24,25,26,27,28,29,30,31]. In addition, five studies [24, 25, 29,30,31] also evaluated histological regression from low-grade dysplasia to intestinal metaplasia. Regression in terms of reduction in the length of BE was also evaluated in three out of eight studies [25, 26, 30]. Demographics of the patients in studies included for the meta-analysis are summarized in Table 1. The response of BE, ssBE, lsBE, and GERD to RYGB is summarized in Table 2.

Table 1 Characteristics of studies included for meta-analysis of Barrett’s esophagus before and after Roux-en-Y gastric bypass
Full size table
Table 2 Response to treatment with Roux-en-Y gastric bypass on Barrett’s esophagus in the studies included for meta-analysis
Full size table

Five [24, 25, 29,30,31] studies included a total of 10 patients with dysplastic BE (low-grade dysplasia or indeterminate dysplasia) of which six (60%) downgraded to intestinal metaplasia with no evidence of dysplasia or non-metaplastic cardiac mucosa. Seven [24, 25, 27,28,29,30,31] out of the eight studies showed no progression of non-dysplastic BE to dysplasia or dysplastic BE to a higher grade in any patient in while one study [26] showed progression of BE to dysplasia in two out of 14 patients (14.2%) after RYGB.

Two studies [14, 15] initially included in this review were excluded from the meta-analysis due to data not being compatible to conduct the meta-analysis. Both showed regression of BE after RYGB.

Assessment of Risk of Bias

A kappa score of 0.9 was achieved in the assessment of risk of bias between the three reviewers. Risk of bias was found to be relatively high across studies. Six out of eight studies were found to be of “low” quality [25, 26, 28,29,30,31] while 2 studies were found to be of “intermediate” quality [24, 27]. Assessment of risk of bias of the included studies as per MINORS criteria is summarized in Table 3.

Table 3 Assessment of risk of bias in studies included for meta-analysis of Barrett’s esophagus before and after Roux-en-Y gastric bypass
Full size table

Meta-Analysis of the Effect of Roux-en-Y Gastric Bypass on Barrett’s Esophagus

Meta-analysis of the effect of RYGB on histological regression of BE (117 patients; n = 8) is shown in Fig. 2. Subgroup analysis of the effect of RYGB on ssBE (38 patients; n = 4) and lsBE (39 patients; n = 5) are shown in Figs. 3 and 4, respectively. Analysis of the effect of RYGB on GERD in patients with BE (78 patients; n = 5) is shown in Fig. 5.

Fig. 2
figure 2

Forest plot for Barrett’s esophagus before and after Roux-en-Y gastric bypass. A Mantel-Haenszel, random-effects model is used for meta-analysis. Risk difference is shown with 95% confidence intervals. RYGB, Roux-en-Y gastric bypass; BE, Barrett’s esophagus

Full size image
Fig. 3
figure 3

Forest plot for short-segment Barrett’s esophagus before and after Roux-en-Y gastric bypass. A Mantel-Haenszel, random-effects model is used for meta-analysis. Risk difference is shown with 95% confidence intervals. RYGB, Roux-en-Y gastric bypass; ssBE, short-segment Barrett’s esophagus

Full size image
Fig. 4
figure 4

Forest plot for long-segment Barrett’s esophagus before and after Roux-en-Y gastric bypass. A Mantel-Haenszel, random-effects model is used for meta-analysis. Risk difference is shown with 95% confidence intervals. RYGB, Roux-en-Y gastric bypass; lsBE, long-segment Barrett’s esophagus

Full size image
Fig. 5
figure 5

Forest plot for gastroesophageal reflux disease in patients with Barrett’s esophagus before and after Roux-en-Y gastric bypass. A Mantel-Haenszel, random-effects model is used for meta-analysis. Risk difference is shown with 95% confidence intervals. RYGB, Roux-en-Y gastric bypass; GERD, gastroesophageal reflux disease

Full size image

BE showed significant regression on postoperative endoscopy at > 1 year after RYGB (RD − 0.56, 95% c.i. − 0.69 to − 0.43; P < 0.001). On subgroup analysis, significant regression of ssBE (RD − 0.51, 95% c.i. − 0.68 to − 0.33; P < 0.001) and lsBE (RD − 0.46, 95% c.i. − 0.71 to − 0.21; P < 0.001) was observed after RYGB. RYGB also led to a significant improvement in GERD (RD − 0.93, 95% c.i. − 1.04 to − 0.81; P < 0.001) at > 1 year after surgery.

Sixty-four out of 117 (54.7%) patients showed histological regression of BE while 2 patients (1.7%) showed progression. RYGB was strongly associated with regression of BE in the postoperative period compared with progression (OR 31.2, 95% c.i. 11.37, to 85.63; P < 0.001) on meta-analysis (Fig. 6).

Fig. 6
figure 6

Forest plot for regression and progression of Barrett’s esophagus after Roux-en-Y gastric bypass. A Mantel-Haenszel, random-effects model is used for meta-analysis. Odd’s ratio is shown with 95% confidence intervals. BE, Barrett’s esophagus

Full size image

Assessment of Heterogeneity and Publication Bias

The funnel plots for the assessment of heterogeneity and publication bias are shown in Fig. 7. Funnel plots indicate variable asymmetry and are likely due to publication bias or “small study effects” due to inclusion of a large number of studies with small sample size.

Fig. 7
figure 7

Funnel plot of studies included in the meta-analysis of Barrett esophagus before and after Roux-en-Y gastric bypass (a), short-segment Barrett’s esophagus before and after Roux-en-Y gastric bypass (b), long-segment Barrett’s esophagus before and after Roux-en-Y gastric bypass (c), and gastroesophageal reflux disease in patients of Barrett’s esophagus before and after Roux-en-Y gastric bypass (d)

Full size image

Discussion

This meta-analysis of 8 studies (117 patients) [24,25,26,27,28,29,30,31] showed that BE undergoes significant regression at > 1 year after RYGB on endoscopy and biopsy. This effect is observed in both ssBE and lsBE, as shown in a subgroup analysis. RYGB also resulted in significant resolution of GERD in patients with BE at > 1 year after surgery. The odds of BE regressing after RYGB was 31 times higher than progression after surgery.

Existing guidelines on BE endorse regular surveillance and management of BE in an unselected cohort of patients; however, there is no universal recommendation that caters to the management of BE in the obese population [1, 10]. Proton pump inhibitors (PPI), endoscopic therapy with radiofrequency ablation (RFA), endoscopic mucosal resection (EMR), and traditional anti-reflux surgery are the common interventions offered in BE. PPI therapy for symptomatic control, followed by repeat endoscopy, is usually offered to patients of BE without significant risk factors like dysplasia [1, 32]. More aggressive endoscopic therapy like RFA [33, 34] and EMR [1, 35, 36] are reserved for BE with dysplasia. Anti-reflux surgery has shown promising results in the regression of BE in about one-third of patients in retrospective studies [37,38,39]. However, this view has been disputed by a randomized controlled trial [40] which concludes that anti-reflux surgery is no more effective than PPI therapy and does not lead to regression of BE, which can even progress to dysplasia after surgery. One study found the progression rate of BE after traditional anti-reflux surgery to be 0.8%/year [41]. Neither of these modalities address the issue of excess weight which is regarded as one of the primary reasons behind failure of traditional anti-reflux surgery in obese individuals [42,43,44].

RYGB seems to be an attractive option in morbidly obese individuals with GERD as it can address reflux and its consequences along with inducing weight loss. According to a study on obese patients with GERD who failed traditional anti-reflux surgery, RYGB when used as a salvage procedure produced symptomatic improvement from GERD in 100% patients with complete resolution in 78% [42]. However, there is a scarcity of literature on the effect of RYGB on BE. The results of this meta-analysis show that there is a 56% regression of BE and 93% improvement in GERD after RYGB at > 1 year after surgery. This further lends support to the view that RYGB should be utilized as the therapy of first choice in suitable morbidly obese patients with GERD.

Another area of debate is the utilization of routine endoscopic assessment of the esophagus and stomach before bariatric surgery. Since obesity is a risk factor for GERD, hiatus hernia, BE, and EAC, routine preoperative screening before bariatric surgery can potentially diagnose abnormalities which may alter the course of management. The European Association of Endoscopic Surgery (EAES) guidelines [45] advocate routine preoperative endoscopy in all patients undergoing bariatric surgery. In contrast, the American Society for Gastrointestinal Endoscopy (ASGE) guidelines [46] recommend preoperative endoscopic evaluation in symptomatic patients only. In our study, BE was diagnosed in 117 out of more than 10,779 patients who underwent RYGB with a selective incidence of less than 1.08%. In a multicenter study [47] on 3219 patients evaluating the role of upper gastrointestinal endoscopy in bariatric surgery, endoscopy was found to be normal in 66% patients who were asymptomatic compared with only 9% in those who had upper gastrointestinal symptoms and the authors concluded that preoperative endoscopy should be considered only in symptomatic patients before bariatric surgery. Similar findings were observed in a study from the USA [48] that found an abnormality detection rate of 61.6% on preoperative endoscopy before bariatric surgery; however, this altered the surgical strategy in only 1.7% patients, and the authors concluded that endoscopic screening before bariatric surgery picks up abnormalities, but the diagnosis of these findings rarely changes the surgical strategy. Furthermore, on cost analysis, the cost of performing routine endoscopy prior to bariatric surgery per clinically important lesion detected that altered surgical management was approximately US$34,800 per lesion [48]. Based on the findings of these studies, and a detection rate of BE of less than 1.08% of all RYGB patients in this meta-analysis, it can be argued that endoscopic screening should be limited to patients with upper gastrointestinal symptoms only.

This review has many strengths. It is the first meta-analysis on the effect of RYGB on BE providing a high-level evidence on this topic. This analysis on 117 patients is the largest among existing literature and suggests that RYGB should be the preferred bariatric procedure in suitable, obese patients with BE as one out of two patients show regression of BE at > 1 year after surgery. The results of this study also showed comparable results for ssBE (regression rate of 51%) and lsBE (regression rate of 46%) at > 1 year after RYGB, a 93% improvement in GERD symptoms after RYGB in obese patients with BE, and 31 times odds of regression of BE after RYGB compared with progression.

The aim of this study was to perform a meta-analysis of the effect of RYGB on BE in obese population based on the existing scientific literature. A limitation of this meta-analysis is the relatively low quality of literature available on this topic. Only 117 patients from 8 studies with BE were identified to have undergone RYGB, but this could be due to a relatively low rate of detection of BE before bariatric surgery and the lack of routine endoscopic screening before bariatric surgery in most units. There is a possibility of an inter-observer variation in the diagnosis of BE between the studies included for this meta-analysis leading to observer bias. Variation in the histopathological diagnosis of BE by the pathologists examining the specimen in the studies included could also confound the results. There was some missing information in the individual studies despite our attempts at contacting the authors. Postoperative use of PPI, although clearly mentioned in only one study, could potentially confound the results of this meta-analysis. Also, this meta-analysis pertains to only obese patients with BE undergoing RYGB and the results of this study cannot be extrapolated to non-obese individuals. Another limitation of the study was that patients who failed to show a regression of BE after RYGB were not followed up in the long term. Well-controlled prospective studies with long-term follow-up are felt necessary to fully understand the effect of RYGB on BE.