Preamble

The International Federation for the Surgery of Obesity and Metabolic Disorders (IFSO) has been playing an integral role in educating both the metabolic surgical and the medical community at large about the importance of surgical and/or endoscopic interventions in treating adiposity-based chronic diseases. The occurrence of chronic conditions following bariatric/metabolic surgery (BMS), such as gastro-oesophageal reflux disease (GERD) and columnar (intestinal) epithelial metaplasia of the distal oesophagus (also known as Barrett’s oesophagus (BE)), has long been discussed in the metabolic surgical and medical community. Equally, the risk of neoplastic progression of Barrett’s oesophagus to oesophageal adenocarcinoma (EAC) and the resulting requirement for surgery are the source of some concern for many involved in the care of these patients, as the surgical alteration of the gastrointestinal tract may lead to impaired reconstructive options. As such, there is a requirement for guidance of the community. The IFSO commissioned a task force to elucidate three aspects of the presenting problem: First, to determine what the estimated incidence of Barrett’s oesophagus is in patients presenting for BMS; second, to determine the frequency at which Barrett’s oesophagus may develop following BMS (with a particular focus on the laparoscopic sleeve gastrectomy (LSG)); and third, to determine if regression of Barrett’s oesophagus may occur following BMS given the close relationship of obesity and the development of BE/EAC. Based on these findings, a position statement regarding the management of this pathology in the context of BMS was developed. The following position statement is issued by the IFSO Barrett’s Oesophagus task force and approved by the IFSO Scientific Committee and Executive Board. This statement is based on current clinical knowledge, expert opinion and published peer-reviewed scientific evidence. It will be reviewed regularly.

Background

Bariatric/metabolic surgery (BMS) has gained substantial popularity to treat the obesity epidemic, with hundreds of thousands of procedures being performed worldwide every year [1]. The subsequent alterations in the anatomy of the gastrointestinal tract may result in chronic conditions such as gastro-oesophageal reflux disease (GERD), which may in turn confer a risk of changes in the distal oesophagus such as Barrett’s oesophagus (BE) or oesophageal adenocarcinoma (EAC). Equally, patients with obesity have higher rates of pre-existing GERD [2, 3], and obesity is a recognised risk factor for both BE and EAC [4]. Thus, patients presenting for BMS may already bear changes in their distal oesophagus putting them at increased risk of EAC formation. As the role of systematic preoperative screening as well as postoperative surveillance endoscopy for patients presenting for or undergoing BMS remains to be elucidated, the magnitude of the presenting problem remains poorly understood. Accordingly, treatment decisions are largely guided by some higher level data as well as anecdotal evidence and small case-series, which indicate that performing laparoscopic sleeve gastrectomy (LSG) results in higher rates of postoperative de novo GERD compared to laparoscopic Roux-en-Y gastric bypass (RYGB) procedures [5, 6]. Equally, to what extent other procedures, such as laparoscopic adjustable-gastric banding (LAGB) or one-anastomosis mini-gastric bypass procedures (OAGB), induce chronic alterations of the distal oesophagus that may be deleterious to long-term patient outcomes remains largely unclear.

Therefore, the task force undertook a systematic review to summarise the current evidence on the incidence of Barrett’s oesophagus both before and after BMS with the aim of providing the most up-to-date information to guide practice.

Methods

Literature Search Strategy

The electronic bibliographic databases MEDLINE, EMBASE, PubMed and Cochrane Library were searched to identify eligible studies published between January 1990 and September 2019 using broad Medical Subject Heading (MeSH) terms and text words to encompass all studies relating to Barrett’s oesophagus and any BMS procedure. Procedure-specifying terms were also used (i.e. gastric band, sleeve gastrectomy, gastric bypass). A full list of search terms is provided in Appendix. To minimise the risk of publication bias, conference abstracts and proceedings were searched through Web of Science, EMBASE and Scopus. Furthermore, the following major gastrointestinal and bariatric conferences were manually searched for relevant reports: Digestive Disease Week (DDW/SSAT), Society of American Gastrointestinal and Endoscopic Surgeons (SAGES), International Federation for the Surgery of Obesity and Metabolic Disorders (IFSO) and American Society of Metabolic and Bariatric Surgery (ASMBS). Manual searching of reference lists from reviews, as well as references from selected primary studies, was performed to identify any relevant additional studies. The search was done in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [7].

Inclusion and Exclusion Criteria

Studies were selected based on the reporting of the pre- or postoperative occurrence as well as regression of Barrett’s oesophagus in the context of patients presenting for or undergoing BMS. All study designs, study sizes, procedure types and follow-up time frames were accepted. Abstracts were included, but separated from full manuscript publications in subsequent sensitivity analyses. However, case reports on the occurrence of EAC were excluded, as the current review focussed on the management of patients with BE. Equally, studies that did not report adequate information to determine study eligibility or to assess study methods for risk of bias were also excluded. If the same group (identified from author names and institution) published multiple reports with potentially overlapping patient recruitment time periods, BE estimates were extracted from the most recent publication with the largest patient numbers to avoid duplication of data.

Data Extraction

Information extracted from eligible studies included basic study data (year, country, design, study size), demographic data, surgical technique, weight loss and follow-up time. Barrett’s oesophagus specific questions included the following: adjustment of studies for the preoperative presence of BE (i.e. systematic performing of preoperative upper gastrointestinal endoscopy), definitions of BE used (i.e. endoscopic aspect vs. histologically proven as defined by local diagnostic criteria as well as biopsy locations), the length of BE segments (short-segment BE (SSBE) vs. long-segment BE (LSBE)) and the presence/resolution of dysplasia. When assessing BE regression, regression definitions of the corresponding articles were used. These included either a decrease in the length of the BE segment, a regression from dysplastic to non-dysplastic BE or a complete disappearance of BE during follow-up endoscopy.

Risk of Bias Assessment, Subgroup and Sensitivity Analyses

All studies were assessed for their risk of bias based on the Newcastle-Ottawa Scale [8]. Each study was assessed independently by two investigators regarding study selection, comparability, and outcomes. The Newcastle-Ottawa Scale consists of 3 subscales which contribute to a maximum total score of 9. Studies scoring < 3 were regarded as being at high, between 4 and 6 moderate and > 6 at low risk of bias. Equally, predefined subgroup and sensitivity analyses were performed according to study design (abstracts vs. full-text articles and prospective vs. retrospective studies) and aspects relevant to the pathology of interest (procedure type for the postoperative occurrence of BE, adjustment for preoperative presence of BE, length of follow-up).

Statistical Analysis

We performed a meta-analysis of proportions with the goal of obtaining a precise estimate of the overall proportion of patients with BE in the context of BMS (i.e. presenting for, developing after and/or regressing after BMS). Logit transformations were used to make the transformed proportions follow a normal distribution. For final reporting, the transformed summary proportions and corresponding 95% confidence intervals were converted back to regular proportions for ease of interpretation. The inverse-variance method was used to weigh effect sizes according to study size. Because we expected heterogeneity in study estimates across the included studies, we applied a random-effects model for the calculation of the summary prevalence of BE patients [9]. Heterogeneity was tested using Cochran’s Q statistic, with p < 0.1 indicating heterogeneity. The degree of heterogeneity was quantified using the I2 statistic [10]. Sensitivity analysis was performed according to the plan outlined above. Differences between subgroups were assessed with a test for interaction [11]. Publication bias was quantified using the Egger’s regression model and visualised using funnel plot analyses [12]. All data were analysed using the R Programming Software [13] using the metafor and meta packages.

Results

Literature Search

Using the described search strategy, we initially identified 570 records and six further articles during an updated search on 7 April 2020. Following the removal of 111 duplicates, 465 titles and abstracts were screened. Subsequently, 158 full-text articles were assessed for eligibility, with 18 further studies identified by screening relevant reference lists. One hundred and twelve articles were excluded, and consequently 64 studies were included for detailed analysis. Following the removal of studies reporting on duplicate cohorts or being deemed as not statistically exploitable, 56 studies were included in the final quantitative analysis (Fig. 1).

Fig. 1
figure 1

PRISMA flowchart

Full size image

Overall Summary

A total of 56 studies were included, of which 18 were deemed as being at low, 37 intermediate and 4 high risk of bias. Of the included studies, 30 included only data for the calculation of preoperative BE incidence rates, 12 had only data on postoperative BE incidence and the remaining 14 studies provided data which could be exploited for either preoperative or postoperative BE incidence rates and/or the incidence of Barrett’s progression/regression following BMS (n = 8). Basic study characteristics, their corresponding references and cohort demographic data are provided in Table 1.

Table 1 Basic study characteristics, their corresponding references and cohort demographic data
Full size table

Preoperative Incidence of Barrett’s Oesophagus

Thirty-eight studies including 22,270 patients reported on the incidence of BE in patients undergoing systematic preoperative endoscopy. The overall cumulative incidence of BE prior to BMS was 2.1% (95% CI 1.4–3.2%, I2 = 94%, Fig. 2). When sensitivity analysis was performed and one outlier study (Balsiger et al. [18] which reported a preoperative incidence of 29.2% in highly symptomatic VBG-patients) was excluded, the cumulative preoperative incidence was 2.0% (95% CI 1.3–3.0%, I2 = 93%, Supplementary Fig. 1).

Fig. 2
figure 2

Preoperative incidence of BE (all studies)

Full size image

In a subsequent sensitivity analysis, the preoperative incidence of BE was adjusted for whether BE had been diagnosed endoscopically or through histopathologic analyses of biopsy specimens. Twenty studies including 8618 patients in which the presence of BE was confirmed by histology provided a cumulative preoperative incidence of BE of 3.0% (95%CI 1.8–4.9%, I2 = 94%, Supplementary Fig. 2). Equally, if the analysis was adjusted for study design (i.e. retrospective vs. prospective studies), the eight prospective studies including 1555 patients reported a cumulative 3.8% (95%CI 1.7–8.3%, I2 = 66%, Supplementary Fig. 3) preoperative incidence of BE. Finally, when only those studies deemed as being at low risk of bias (n = 6, 3510 patients) were included, the cumulative preoperative incidence of BE was 5.9% (95% CI 2.6–12.9%, I2 = 96%, Supplementary Fig. 4).

Postoperative Incidence of Barrett’s Oesophagus

Eighteen studies including 19,775 patients provided postoperative BE incidence estimates. The overall postoperative incidence of BE was 1.9% (95% CI 0.8–4.1%, I2 = 96%, Fig. 3) irrespective of the type of bariatric procedure performed. When only studies were included, in which all patients analysed had also undergone preoperative endoscopy (and thus were deemed true de novo BE patients, n = 15; 19,751 patients), the rate of postoperative BE was 2.6% (95% CI 0.1–5.6%, I2 = 94%, Supplementary Fig. 5). When studies were stratified according to the duration of follow-up, studies with longer follow-up (≥ 2 years, n = 10 of which one study provided separate estimates for LSG vs. RYGB, n total patients = 1827) showed a postoperative BE incidence rate of 4.2% (95% CI 1.9–9.2%, I2 = 89%, Supplementary Fig. 6). When the study by Balsiger et al. was again excluded due to its particular patient cohort, then the cumulative postoperative incidence of BE in studies with a follow-up ≥ 2 years was 3.4% (95% CI 1.5–7.4%, I2 = 88%, Supplementary Fig. 7).

Fig. 3
figure 3

Postoperative incidence of BE (all studies) stratified by bariatric procedure type

Full size image

Postoperative Incidence of Barrett’s Oesophagus Following Laparoscopic Sleeve Gastrectomy

Eight studies including 14,274 patients provided postoperative BE incidence estimates for patients undergoing LSG. The overall postoperative BE incidence was 2.0% (95% CI 0.4–10.2%, I2 = 98%, Fig. 4a). However, when only those studies were included in which patients had undergone systematic preoperative endoscopy and therefore true de novo patients were captured (n studies = 5, n patients = 761), the postoperative BE incidence rate increased to 6% (95% CI 1.8–17.8%, I2 = 89%, Fig. 4b). Equally, when the data that was reported in abstract form only was excluded, the postoperative BE incidence was estimated at 4.6% (95% CI 1.5–13.1%, I2 = 90%, Fig. 4c). Finally, when length of follow-up was taken into account, those studies with a patient follow-up ≥ 2 years (n = 6, n patients = 1022) provided a combined postoperative BE incidence estimate of 4.6% (95% CI 1.5–13.1%, I2 = 90%, Fig. 4d).

Fig. 4
figure 4

Laparoscopic sleeve gastrectomy and BE incidence rates

Full size image

Barrett’s Oesophagus Regression Following BMS

Ten studies including 118 patients reported on the incidence of BE regression following BMS, all of which only included patients having undergone laparoscopic Roux-en-Y gastric bypass (LRYGB). Only studies in which patients had been preoperatively diagnosed by endoscopy were included, and the median length of follow-up of these patients was 28.5 months (IQR 18–37.5 months). In total 62.9% (95% CI 53.4–71.6%) showed signs of BE regression during follow-up endoscopy (Fig. 5). No study reporting on the incidence of BE regression following LSG could be identified during our literature review.

Fig. 5
figure 5

BE regression following LRYGB

Full size image

Publication and Small Study Bias Assessment

Publication bias was assessed by creating funnel plots and performing Egger’s regression upon which we found significant evidence for publication and small study bias in the studies on preoperative BE estimates, whereas there was no significant publication or small study bias in those studies reporting postoperative BE incidence rates (Supplementary Fig. 8a and b, Egger’s p for asymmetry < 0.001 and 0.52 respectively).

Discussion

The current evidence demonstrates that up to 3.8% of patients presenting for and undergoing BMS have Barrett’s oesophagus. Equally, the present study shows that approximately 1.9% of patients will go on to develop BE irrespective of their type of bariatric procedure, but for patients undergoing LSG, the incidence of de novo BE may be as high as 4.6% within 5 years after surgery. These figures are offset by interesting data, albeit from just over 100 patients, which suggest that BE regression may also occur following LRYGB in up to 63% of patients. However, the present analysis limited by the quality of the included studies which is highly variable, many of which are retrospective cohort studies and this is in turn reflected by the very high inter-study heterogeneity as well as significant publication and small study bias that was identified. Equally, readers should be aware that this study incorporated data from all published sources, including conference abstracts as is recommended by current guidelines [70, 71], and therefore, final estimates may be slightly different to other published series due to alternate search strategies and inclusion/exclusion criteria. Accordingly, the estimates provided in this analysis should be interpreted with caution. Equally, the guidance provided in this position statement is subject to further review as the body of evidence on this topic is set to grow.

The main concern that exists within the bariatric surgical community regarding BE relates to the popularity of the LSG. According to the 5th IFSO Global registry report, LSG was the most frequently performed bariatric procedure from 2014 to 2019 (58.5% of all captured procedures) [1]. Whilst similar excess body weight loss and co-morbidity resolution is achieved with a LSG compared to a RYGB [6, 72], rates of particularly de novo reflux are reported to be substantially higher after LSG [5, 6, 72, 73]. Accordingly, valid concerns exist that the short-term up-sides to performing LSG such as shorter operating times, potentially less perioperative morbidity and improved scalability may be offset by the long-term increased risk of GERD, Barrett’s formation and potential subsequent development of oesophageal adenocarcinoma [74, 75]. The issues surrounding this are twofold: First, because of the increased uptake of BMS throughout the world, by performing mainly LSG, we may be creating a novel patient population with an unprecedented incidence rate of BE. The risk of BE in the general populations is only 1–2% [76]; thus, if the estimates from the present study are correct, then by performing LSG, we would be creating a true “at-risk” population, with an up to 6× higher risk of BE development. Secondly, if the LSG patients subsequently progress from BE to EAC, then the surgical reconstructive options for patients who require more than endoscopic therapy are usually limited to colonic interpositioning. Although this presents a viable reconstructive option with some advantages, the procedure tends to be more complex with longer operating times and higher blood loss and with increased morbidity when compared to gastric transposition [77].

However, whilst these concerns are valid, they remain largely hypothetical. To date, there is no data suggesting that any of this is actually set to occur, and particularly oesophageal cancer development following BMS is limited to case reports [78]. Conversely, whilst some studies suggest a reduced incidence of EAC development following BMS [79], other population-based studies indicate that there is no change in the overall incidence of oesophageal cancer following BMS, but also caution that the incidence rates of oesophageal cancer may be too low to perform time-dependent analyses [80]. This latter aspect is important to consider, because in patients with BE, the risk of EAC development is approximately 0.33% per annum for non-dysplastic BE and as low as 0.19% per annum for short-segment BE [81]. Thus, there remains the risk of a substantial delay between the index surgery and cancer development and the risk-mitigating effects of bariatric surgery–induced weight loss on EAC development are incompletely understood.

Equally, the issue of performing BMS in qualifying patients with obesity who have established BE needs to be considered. Whilst the presented data suggest that RYGB may lead to BE regression, they need to be interpreted with caution. Regression definitions used was either a decrease in length or a reduction in dysplasia severity. Whereas Barrett’s length assessment is subject to observer bias, field effects affecting dysplasia assessment are equally well-established [82]. Accordingly, whilst some studies have aimed to elucidate the physiological changes that may occur in the distal oesophagus following BMS [45, 83], the true impact of RYGB on the natural history of BE development and progression remains unclear. Despite these limitations, there also remains the surgical technical advantage of performing a RYGB, as the remnant stomach may be used as a gastric-conduit for reconstructive purposes, should the patients progress to EAC requiring surgical therapy. Thus, the present data and considerations may encourage a general recommendation of RYGB in the presence of BE. However, some patients may not want or qualify for RYGB for a variety of reasons. Accordingly, the potential health benefits of metabolic and weight loss surgery must be weighed against the potential risk of particularly EAC development. This is a valid debate, particularly when the causes for mortality in patients with BE are considered: Numerous population-based studies suggest that only a minority of BE patients actually die of EAC, whereas the more common causes of mortality are ischaemic heart disease, other non-oesophageal cancers, respiratory and other digestive system diseases [84,85,86,87]. Taken together with the potentially reduced risk of EAC development following BMS, and that data exist that chemoprophylaxis with aspirin and high-dose proton-pump inhibitor therapy may also reduce the risk of high-grade dysplasia and invasive cancer development [88], denying obese patients BE an effective weight loss procedure such as LSG seems somewhat scientifically problematic in view of the paucity of high-quality data. Although extreme caution is mandated when evaluating such therapies in what is effectively an “evidence-free zone”, the 2020 task force has identified an important knowledge gap that mandates further research.

Furthermore, the present data may be seen as a direct argument to support systematic screening of patients either before BS or following LSG; however, these calls need to be considered carefully. For example, systematic preoperative screening of patients presenting for BMS remains somewhat controversial as some studies report that the proportion of preoperative endoscopies resulting in a change of management is < 10% [89], whereas the recently commissioned 2020 IFSO task force found systematic preoperative endoscopy resulting in a change in practice in 25.3% of cases [90]. Whilst there is no data on the value of systematic postoperative endoscopic surveillance of BMS patients, the general screening and surveillance of non-dysplastic Barrett’s oesophagus patients may not be cost-effective [91]. However, surveillance of high-risk populations may provide a cost benefit and also translate to a survival advantage, in cases of EAC development [92]. Accordingly, offering patients a screening endoscopy at 1 year following a “high-risk” bariatric procedure such as LSG and then every 2–3 years depending on its outcome may be prudent, but warrants further investigation in prospective clinical trials.

Recommendation of the IFSO Barrett’s Oesophagus Task Force

Based on the existing data, the 2020 IFSO task force recommends the following:

  1. 1.

    Patients presenting for BMS need to be carefully assessed for the presence of GERD and complications from GERD such as BE. Particular focus should be placed on the duration of symptoms, any previous upper endoscopies and the use of anti-acid medication. If the patient reviewed represents a potential “at-risk” population according to conventional gastro-enterological guidelines, this patient should undergo preoperative screening endoscopy. However, given that BE patients typically are void of symptom indications for preoperative screening endoscopy should be made generously.

  2. 2.

    If a patient has the presence of “salmon-coloured” mucosa and/or an irregular z-line upon upper endoscopy, then the exact length and circumference according to the Prague Classification needs to be documented as well as the segment of Barrett’s systematically biopsied according to the Seattle Protocol to capture any potential areas of dysplasia.

  3. 3.

    If the patient has any dysplastic BE, then the patient should be considered for evaluation of preoperative BE-therapy.

  4. 4.

    In the presence of long-segment or dysplastic BE, then procedures where the distal oesophagus may subsequently be exposed to higher concentrations of acid or bile (such as LSG or OAGB) should not be performed.

  5. 5.

    If the patient has short-segment BE, then after careful discussion with the patients the benefits of LSG vs. RYGB should be discussed. In general, RYGB is the preferred procedure due to evidence of BE regression; however, a LSG cannot be categorically discouraged due to the potential long-term health benefits of bariatric/metabolic surgery. However, given the lack of high-quality data, the 2020 task force recommends practitioners proceed with extreme caution if considering this option together with their patients, and it is recommended that all such cases be systematically captured and screened in a prospective fashion. This statement cannot be viewed as a blanket approval to perform LSG in patients with BE, but is reflective of the paucity of data regarding the outcomes of patients with BE undergoing potentially refluxogenic bariatric procedures.

  6. 6.

    Given the current evidence suggesting higher incidence rates of BE following LSG compared to the general population, a single screening endoscopy at 1 year postoperatively and then every 2–3 years, depending on its outcome, is recommended.

  7. 7.

    The current analysis mainly includes studies comprising of Caucasian, Middle-Eastern or South-American populations. Accordingly, how the present findings apply to patients of Asian heritage/undergoing BMS in Asian countries is unclear and warrants further research.

  8. 8.

    IFSO supports further high-quality studies in the field, mainly prospective and/or population-based studies to help elucidate the exact magnitude of the issue as well as provide further guidance to the community as necessary. In particular, researchers should pay attention to also identifying potentially confounding factors, such as the presence of hiatus hernia (and how this was addressed intra/postoperatively), pouch sizes and potential pouch pathologies (such as strictures/distal obstructions) when assessing the impact of certain procedures on the development and/or progression of BE.