Is Routine Preoperative Esophagogastroduodenoscopy Prior to Bariatric Surgery Mandatory? Systematic Review and Meta-analysis of 10,685 Patients

There is a debate about the utility of routine preoperative esophagogastroduodenoscopy (p-EGD) screening of patients undergoing bariatric surgery (BS) [1, 2]. The European and Italian national recommendations advocate the use of presurgery upper gastrointestinal endoscopy together with multiple biopsies in the work-up of patients; conversely, the American Society for Metabolic & Bariatric Surgery only recommends it in selected cases with symptomatic gastric disease [3‐5]. Generally, the question of routine p-EGD has many clinical implications and significant financial repercussions [1].

Some evidence supports routine p-EGD among patients undergoing BS. The reasons include the weak correlation between the patients’ symptoms and p-EGD findings, that p-EGD is convenient, safe, applied easily [6‐8], and p-EGD findings may alter the management and hence eliminate the future development of gastric pathology [9], or detect asymptomatic benign or pre/malignant lesions. Missing asymptomatic lesions in some BS where the distal stomach and/or duodenum is rendered unreachable by esophagogastroduodenoscopy could lead to missing some lesions in the bypassed stomach that p-EGD could have discovered [10‐16]. Some authors endorse that all BS patients have p-EGD, as after surgery, the endoscope may not reach the gastric/duodenal mucosa [17]. In agreement, others recommended that all BS patients should have upper gastrointestinal endoscopy [8]. For some procedures (e.g., laparoscopic adjustable gastric banding and vertical banded gastroplasty), p-EGD could provide information that might influence the operative procedure, particularly due to upper gastrointestinal lesions that often require medical therapy [7, 18].

It remains contested whether routine p-EGD should be undertaken for all patients undergoing e.g., laparoscopic sleeve gastrectomy (LSG) [19]. Some authors support routine p-EGD in patients with upper gastrointestinal symptoms (symptomatic cases only) [3, 20, 21]. Others suggest a selective approach for asymptomatic cases, because of the weak clinical relevance of most lesions discovered on routine p-EGD, its cost, and invasiveness [22, 23]. Still, other research found that routine p-EGD in LSG might require further justification for asymptomatic patients due to its low utility in managing such patients in regions with low prevalence of upper gastrointestinal cancers [2]. Only 2% of asymptomatic patients had any abnormality detected at p-EGD, none of which affected their treatment plan, and hence a focus on symptomatic patients only can safely reduce p-EGD rate by 80% [24].

Others reported that most of the pathology identified at p-EGD among patients scheduled for gastric banding did not significantly influence their management; however, two early cancers were detected [25]. In addition, although obesity is a risk factor for gastroesophageal reflux and esophageal adenocarcinoma, research could not confirm a high prevalence of Barrett’s esophagus among 233 patients selected for laparoscopic gastric banding [26]. Likewise, the association between obesity and reflux remains controversial [27], and it is unclear whether BS impacts the advancement of gastro-esophageal reflux disease (GERD) [28]. Despite a somewhat inaccessible foregut after bypass surgery, the low gastric cancer incidence among Caucasians [29] may not demand routine p-EGD [30].

Opinions remain divided as to whether p-EGD should be undertaken for all BS patients. One position is that the “intuitive reasons to continue p-EGD screening of BS patients include endoscopic findings that optimize medical management for the healing of their BS in a substantial proportion of patients and/or the endoscopic findings in at least a few patients that alter or delay the surgery itself” (p. 712) [22]. Conversely, others recommended that standard p-EGD is not indicated, as many BS patients are screened in order to discover clinically significant abnormalities [11]. For example, in Turkey, none of the 755 LSG patients had macro/microscopic malignant pathological finding in the preoperative upper gastrointestinal endoscopy [31]. In Brazil, researchers did not perform routine p-EGD on 649 LSG patients and only did when patients complained of abdominal pain or dysphagia; however, even with these symptomatic complaints, most patients had no abnormal findings [32]. Across 93.2% of BS patients, p-EGD findings were negative or had no effect on the preoperative management or choice of surgery; thus, it might not be wise to expose morbidly obese patients to a routine invasive uncomfortable procedure that carries potential (although minimal) risk [21]. Hence, authors have raised the question: “We do not screen the general population for those minor esophagogastroduodenoscopy findings; so why should we do it on people planned for bariatric surgery?” (p. 414) [21]. Likewise, a comment on “Is esophagogastroduodenoscopy before Roux-en-Y gastric bypass or sleeve gastrectomy mandatory?” concluded that p-EGD had no value in prediction or prevention of postoperative complications [33].

Such inconsistency highlights a gap as to whether routine p-EGD is sufficiently justified for all BS patients, and inspired the current systematic review and meta-analysis of the significance of routine p-EGD screening in BS. To the best of our knowledge, there exists no systematic review of the English literature on the topic, and no meta-analysis has been undertaken to answer this important question. Globally, many upper gastrointestinal endoscopies are performed for inappropriate indications, and the overuse of healthcare negatively affects healthcare quality and places pressure on endoscopy services [34]. Therefore, the current systematic review and meta-analysis assessed the justifications as to whether p-EGD should be routinely undertaken for all BS patients.

This systematic review and meta-analysis was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement. The study was registered at the International prospective register of systematic reviews (PROSPERO CRD42020157596).

A systematic review was carried out using PubMed, Cochrane Central Register of Controlled Trials (CENTRAL), WHO International Clinical Trials Registry Platform, Cochrane Library, MEDLINE, Scopus, clinicaltrials.​gov, and Google scholar electronic databases. We used the keywords “bariatric surgery” “Esophagogastroduodenoscopy,” “preoperative” [in Title/Abstract]. The medical subject headings (MeSH) terms used were bariatric surgery (All Fields) AND “Esophagogastroduodenoscopy” (MeSH Terms); bariatric surgery (All Fields) AND “preoperative AND Esophagogastroduodenoscopy” (MeSH Terms); bariatric surgery (All Fields) AND “preoperative OR Esophagogastroduodenoscopy” (MeSH Terms). Additional searches were conducted using the reference lists of studies and review articles for a selection of relevant articles. The references of all included articles or relevant reviews were cross-checked.

The inclusion criteria were (1) original studies, (2) English language, (3) published from 01 January 2000 through 30th April 2019, (4) assessed “Esophagogastroduodenoscopy” and “bariatric surgery,” and, (5) patients of any age, gender, and ethnicity. Articles other than original studies such as commentaries, letters to the editor, reviews, case reports, and studies that did not include outcomes or comparisons were also excluded. The consensus on the inclusion/exclusion criteria was premised on the fact that whether a given study provided information on the association between p-EGD and post-operative outcomes among bariatric surgery patients. Therefore, even studies with smaller sample sizes were also included in the initial evaluation. Three authors independently abstracted the data.

To assess the significance of routine p-EGD screening in BS, the specific objectives were to:

In order to gauge the value of routine p-EGD screening in BS, we employed Sharaf et al.’s classification [6] of predetermined criteria to categorize p-EGD findings into four groups:

The titles of the research articles obtained from the initial database searches were screened and relevant papers were selected. Then the abstracts and full texts were reviewed according to the inclusion criteria for final selection. Three authors independently reviewed the studies based on the exclusion and inclusion criteria. Initially, titles of the studies identified from the search were assessed for inclusion. Titles approved by the authors were moved to abstract screening. If three authors rejected a study at this stage, it was excluded from the review. In the third stage, full text articles were screened for eligibility. Only those studies approved by the three authors were included in the review. Agreement between the authors on the quality of the articles ranged between 90 and 100%. All disagreements were resolved by consensus among the authors. Data extracted from the selected articles included authors, the origin of studies, source population, study settings and duration, inclusion/exclusion criteria, data sources and measurement, sample size, and the yield of p-EGD findings in terms of the four groups of Sharaf et al.’s classification [6].

The methodological quality of the selected studies was assessed based on five STROBE criteria from the checklist, namely, study design, setting, participants, data sources/measurement, and study size. The STROBE checklist and the five criteria selected from the checklist were most relevant in the assessment of the methodological quality of observational studies in epidemiology (Table 1).

Prevalences were calculated for categorical variables. The decision to employ either a fixed-effect or random effect model depended on the results of statistical tests for heterogeneity. Data heterogeneity was assessed using the Cochrane Q homogeneity test (significance set at p < 0.10). If the studies were statistically homogeneous, a fixed-effect model was selected. A random effect model was used when studies were statistically heterogeneous. The Higgin’s I² test is the ratio of true heterogeneity to the total variation in observed effects. A rough guide to interpretation of I² test is 0–25%: might not be important; 25–50%: may represent moderate heterogeneity; 50–75%: may represent substantial heterogeneity; and > 75%: considerable heterogeneity. Publication bias was visually estimated by assessing funnel plots. Pooled estimates were calculated using the R 3.5.1 software.

The search generated a total of 1256 articles; 1209 articles were either non-relevant to the topic, duplicates, or review articles which were excluded. The relevant titles and/or abstracts and full text of the remaining 47 articles underwent detailed evaluation, after which 22 articles were further eliminated as these were mainly based on protocol development and narrative reviews. Finally, 25 original studies met all the review criteria and were considered for the final meta-analysis (Fig. 1 and Table 1) [2, 6, 8, 9, 11, 14, 18, 21‐25, 28, 30, 35‐45].

Median study duration was 4 years with an inter quartile range of 2–7 years. Overall average age was 40.7 years, and overall average percentage of males (25%) was lower than females (75%). All studies were non-randomized controlled trials, comprising 15 prospective and 10 retrospective studies. These studies had low or unclear risk of bias, unlikely to seriously alter the results. In addition, these studies had no serious risk of bias that can downgrade the quality. There was no inconsistency: the study populations were BS patients, and outcome assessment was consistent, namely the yield of p-EGD findings in terms of the four groups of Sharaf et al.’s classification [6].

The total number of patients pooled was 10,685. Figure 2 depicts the meta-analysis of the 4 groups (groups 0–3) of patients based on their p-EGD findings. The largest group was Group 0 (no abnormal p-EGD findings, 56%, 95% CI: 45–67%) followed by Group 1 (abnormal p-EGD findings that do not necessitate changing the surgical approach or postponing surgery, 26%, 95% CI: 18–35%). These were followed by Group 2 (abnormal p-EGD findings that change the surgical approach or postpone surgery, 16%, 95% CI: 11–21%) and Group 3 (p-EGD findings that signify absolute contraindications to surgery, 0.4%, 95% CI: 0–1%). H. pylori infection was positive among about one-fourth of patients, and hiatal hernia was present in a mean of 17% of patients.

The results for the test of heterogeneity for the meta-analysis among bariatric surgery patients are displayed in the bottom line to the left of each Forest plot. For Group 0 (no abnormal p-EGD findings), Q [χ²] = 1285.41, P = 0.001, I² = 99%, tau² = 0.0159 (Fig. 2a); for Group 1 (abnormal p-EGD findings that do not necessitate changing the surgical approach or postponing surgery), Q [χ²] = 165.03, P = 0.001, I² = 99%, tau² = 0.140 (Fig. 2b); for Group 2 (abnormal p-EGD findings that change the surgical approach or postpone surgery), Q [χ²] = 557.02, P = 0.001, I² = 97% tau² = 0.077 (Fig. 2c); for Group 3 (p-EGD findings that signify absolute contraindications to surgery) Q [χ2] = 557.02, P = 0.001, I² = 72%, tau² = 0.007 (Fig. 2d); for H pylori infection Q [χ²] = 1207.84, P = 0.001, I² = 98%, tau² = 0.007 (Fig. 2e); and, for hiatal hernia Q [χ²] = 556.10, P = 0.001, I² = 96%, tau² = 0.196 (Fig. 2f). However, as I² was > 25%, a random effect model was considered. Tau² reflect the amount of true heterogeneity among the studies.

For all of the above analyses, sensitivity analysis yielded consistent results. Based on a visual inspection of the funnel plots, there was evidence of publication bias for the included studies (Fig. 3). The funnel plots exhibited presence of studies with large standard error and they were not symmetrical.

The studies included in this meta-analysis did not report the frequency of multiple abdominal conditions. Rather, the studies reported the frequency of each abdominal condition separately. Hence, there might be a probability of multiple abdominal conditions for a single patient which would influence the overall estimation in Groups 1 and 2.

The current systematic review and meta-analysis is the first to assess the yield of p-EGD findings in terms of four groups [6], in order to gauge justifications as to whether p-EGD should be routine for all BS patients. Routine p-EGD can diagnose rare gastric pathologies [19]. The current review showed that 82% of patients had either no abnormal p-EGD findings (Group 0) or abnormal p-EGD findings that do not necessitate changing the surgical approach or postponing surgery (Group 1). Another 16% of patients required changing the surgical approach or postponing surgery based on the p-EGD findings (Group 2). Only 0.4% of patients had p-EGD findings that signified absolute contraindication to surgery (Group 3).

Generally, EGD carries risks to patients, as well as legal risks [46]. Hence, in addition to the p-EGD ‘yield’ in discovering/excluding pathologies, the appropriate gauging of whether routine p-EGD is justified for all BS patients needs to consider several parameters. These include the following: adverse effects of routine p-EGD; missing or over-diagnoses of lesions (false negatives, false positives); skill level of the esophagogastroduodenoscopy personnel; availability and cost of alternative (non-invasive) diagnostic methods to discover upper gastrointestinal pathology; and the costs of routine p-EGD. A related point is the changes that could occur to any missed pathology across time: i.e., initially before and then subsequent to BS (histological patterns of cellular alterations after gastric surgeries).

Adverse effects of esophagogastroduodenoscopy include infections, bleedings or perforations [47, 48], acute pancreatitis (direct trauma/gas insufflation) [49]; cardiopulmonary events [48]; methemoglobinemia (genetic predispositions/use of topical anesthetics) [50]; hypoxic respiratory failure/critical events requiring bronchoscopic intratracheal oxygen insufflation [8, 51]; orbital hematoma [52]; and Takotsubo cardiomyopathy with complete heart block [53]. Other effects include pre-endoscopy anxiety (unsedated esophagogastroduodenoscopy) [54], effects related to comorbidities of e.g., morbidly obese diabetic patients where the overnight fasting challenges the metabolic status, and sleep apnea (needs surveillance during sedation) [8]. Despite these, some authors suggest that the infrequent adverse events should not limit routine p-EGD [55].

As for missing important lesions (false negatives), the quality of the esophagogastroduodenoscopy varies [56]. In Spain, 17 out of 187 gastric cancer patients had prior esophagogastroduodenoscopy (9.1%), and 12 of those 17 missed gastric cancer had prior esophagogastroduodenoscopy with abnormal findings [57]. P-EGD is also frequently inaccurate at diagnosing hiatal hernia (particularly large hernias), where 23 patients undergoing sleeve gastrectomy had paraesophageal hernia intraoperatively; many of these patients were asymptomatic, and p-EGD revealed large hiatal hernia in only 4 patients [58‐60]. Conversely, hiatal hernia repair was performed in 56 (5%) of patients positive for intraoperative findings despite a negative p-EGD for hiatal hernia [55]. A related point here pertains to the probability of changes of a given missed lesion, i.e., the changes of pathology across time and the histological cellular alterations after gastric surgeries [61]. Pre-surgery biopsies of 798 LSG patients showed non-significant findings in 86.2%; among them, 99.7% maintained a pattern without relevance for its follow-up; and some patients who had intestinal metaplasia reversed its histopathology (maybe following H. pylori treatment) [62]. Others found that the pre-operative inflammatory alterations were reduced post-operatively, where the chronic gastritis with inflammatory activity associated with H. pylori was reduced by 16.7%, and foveolar hyperplasia was reduced by 25% [61]. Further research can evaluate whether such improvements are due to treatment of H. pylori [61].

In terms of false positives, EGD over-diagnosed small hiatal hernias, most did not require repair, and 60% of EGD positive hiatal hernias were found to be negative intraoperatively [55]. Both the presence of symptoms and EGD findings may not always correlate with intraoperative findings [55]. In the current meta-analysis, p-EGD findings suggested hiatal hernia in a mean of 17% of patients (95% CI: 13–21%). However, the data provided by the studies does not enable one to speculate how many hiatal hernias/other lesions were missed or over-diagnosed during these EGDs.

In connection with the skill level, p-EGD has some subjectivity; hence, the endoscopist’s expertise could lead to over/under diagnoses [55, 63]. The endoscopist is vital in missed gastric cancer [57], and training/learning interventions could enhance the quality of endoscopy [63]. About 51.8% of the incomplete endoscopy reports did not have justification for its incompleteness [64]. Patients with no symptoms or no esophagogastroduodenoscopy evidence of hiatal hernias had hernia repairs (4%–6%), suggesting that small hiatal hernias are operator-dependent diagnoses [55]. The studies included in the current meta-analysis did not examine such skills, and we are unable to conclude how this might have affected the p-EGD yield we computed.

In terms of alternative diagnostic methods for gastric cancer pathologies, there are novel noninvasive screening techniques for e.g., Barrett’s esophagus [65] and H. pylori [66‐68]. However, some authors might view that some novel techniques might be inferior to established gold standards, not all institutions might have advanced alternative diagnostic technologies, and esophagogastroduodenoscopy allows both the direct visualization and tissue biopsy [55].

Endoscopy is costly [1]. In the USA, the average hospital cost of an esophagogastroduodenoscopy with and without biopsy was $3732 and $3038 [69]. Endoscopy necessitates time, money, and personnel resources including experienced investigators, anesthesiological support, and special surveillance [8].

The current meta-analysis found that Group 2 patients (abnormal p-EGD findings that change the surgical approach or postpone surgery) amounted to 16%. However, it is not clear what proportion of these patients were postponed solely for H. pylori medical treatment as opposed to a “true” more substantial esophagogastroduodenoscopy-informed change in the surgical approach. This is important, as some might argue that if H. pylori is diagnosed by a non-invasive method (no need for esophagogastroduodenoscopy), and if the surgery waiting list time at a given institution is > 2–4 weeks (sufficient time for H. pylori treatment), then no postponement might have been required. One inquiry [2] examined the postponement, cancelation, or change of surgical approach based on the p-EGD findings across several sleeve gastrectomy studies and found that a considerable number of Group 2 patients were postponed solely for the treatment of H. pylori. This research [2] reported that across three studies, 21.5% [6], 12% [10], and 27% [30] of Group 2 patients had their BS postponed for H. pylori treatment, or waiting for H. pylori test result to assess severity of inflammation after medical treatment. Such findings suggest, that for the present meta-analysis, it might be reasonable to speculate that the proportion of Group 2 patients postponed due to a “true” change in surgical approach could be much less that the current 16%, further questioning the utility of routine p-EGD.

This review searched most of the citation databases and reference lists of the included studies. We also accessed paid articles. Nevertheless, a limitation of the current meta-analysis is that it included only published studies and only the English literature. We could not find “gray” literature, and hence, potential publication bias cannot be excluded. There were no studies from some regions of the world. However, 25 studies were included in this meta-analysis and we had a sizeable sample of 10,685 patients.

The findings of this meta-analysis compel a revisit of current practice, and a re-evaluation of why p-EGD should be routine for all bariatric surgery patients. In 2016, about 634,897 bariatric operations were performed worldwide [70]. It might not be totally judicious to expose very large numbers of morbidly obese patients to a routine invasive uncomfortable procedure that has potential (although minimal) risk and insufficient evidence of effectiveness. Limitations include the lack of studies from some world regions and a small number of studies.