Reflux and Barrett's oesophagitis after gastric surgery — long-term follow-up and implications for the roles of gastric acid and bile in oesophagitis


Dr A. S. Taha, Crosshouse Hospital, Kilmarnock KA2 0BE, UK.


Background : The role of gastric acid is difficult to separate from that of bile in oesophageal reflux, and the complications of this can take many years to develop. Gastric surgery patients provide a good model for both significant bile reflux and marked gastric acid inhibition.

Aim : To study the oesophageal abnormalities in gastric surgery patients undergoing long-term follow-up, compared with patients with intact stomachs.

Methods : Two hundred and forty adult patients were endoscoped regardless of their age, sex or type of surgical procedure. Oesophageal damage was graded on a scale of 0–5, and biopsies were taken to exclude neoplasia, to diagnose Barrett's oesophagus and to identify Helicobacter pylori.

Results : Of the 240 patients studied, 140 had undergone gastric surgery 27 years (19–31 years) [median (interquartile range)] prior to endoscopy, and these patients had milder oesophageal scores and fewer cases of Barrett's oesophagitis. Of the 119 patients with post-surgical bile reflux gastritis, 31 (26%) had oesophagitis, two (1.7%) had Barrett's oesophagitis and oesophageal scores of 0 (0–1) were found. These results compared with corresponding values of 37 (37%; P = 0.11), 11 (11%; P = 0.007) and 0 (0–2) (P = 0.046), respectively, in 100 patients with intact stomachs. In addition, of the 83 patients with vagotomy, 19 had oesophagitis (23%; P = 0.05), none had Barrett's oesophagitis and lower oesophageal scores (P = 0.02) were found.

Conclusions : The prevalence and severity of reflux and Barrett's oesophagitis are not increased in patients with a long history of gastric surgery, particularly after vagotomy, and despite being at risk of bile reflux.


To date, and throughout the last century, gastric acid has established itself as the major factor in the pathogenesis of gastro-oesophageal reflux disease and its complications, including Barrett's oesophagitis. Therapeutic success has been directly proportional to the degree of acid inhibition, with proton pump inhibitors being the most effective agents.1 In addition to gastric acid, bile reflux has been implicated in oesophageal damage in general and, more importantly, in the aetiology of the pre-malignant condition of Barrett's oesophagus.2, 3 These possible bile-related effects have been the focus of increasing numbers of basic and clinical investigations, given the steady rise in the incidence of cancer at the gastro-oesophageal junction.4, 5

The potential contribution of the individual gastric constituents to human oesophageal damage and its complications, particularly Barrett's oesophagus, has been controversial and largely speculative for several reasons. These include the difficulty in separating the role of bile from that of gastric acid, the uncertainty about the degrees of acid inhibition and the methodological problems surrounding the measurement of bile and alkaline reflux.2, 3, 6 In addition, the microscopic and macroscopic features of Barrett's oesophagus can take many years to develop, and it is not clear whether the findings in animal models can be generalized to humans.4, 5, 7

Our hypothesis is that the above difficulties could be minimized by studying patients with a long history of gastric surgery. These patients have long-term exposure to significant bile reflux, whilst gastric acid is markedly reduced by truncal vagotomy.8 We studied the oesophageal abnormalities, including Barrett's oesophagus, in subjects with a long history of gastric surgery, compared with patients with intact stomachs.


Adult patients were recruited from the Medical and Surgical Gastroenterology Departments. They were included regardless of their age, gender or type of gastric surgery, and these factors were taken into account in the statistical analyses. They were excluded if they consumed excessive amounts of alcohol (> 15 units/week), had undergone cytotoxic therapy or had an uncontrolled systemic disease or active malignancy. The aim was to recruit 150 patients with a history of previous gastric surgery, and a control group of 100 patients with intact stomachs who had similar clinical presentations or indications for endoscopy as the post-surgical group. The first 150 gastric surgery patients and 100 patients with intact stomachs to satisfy the inclusion and exclusion criteria of the respective groups were enrolled into the study. Informed consent was obtained and the local Ethics and Clinical Effectiveness Committees approved the study.

Endoscopic assessments

Patients were endoscoped after a fast of ≥ 4 h, using midazolam for intravenous sedation or lignocaine (lidocaine) spray as a local throat anaesthetic. The endoscopist was unaware of the details of the patient's demographic characteristics, Helicobacter pylori status or histological findings.

The main gastric endoscopic findings included a completely normal picture, mucosal redness, erosive gastritis, ulcer and neoplastic lesions. Similar descriptions were applied to the duodenal mucosa. The term ‘bile reflux gastritis’ was given to gastric redness in the presence of bile. This endoscopic evidence of bile reflux was used in the light of the known limitations of currently available methods of assessing this condition.2, 6

The oesophageal endoscopic findings were classified according to a modification of the Savary–Miller system:9 grade 0, completely normal mucosa; grade 1, mucosal erythema without erosions; grade 2, erosions involving < 10% of the circumference; grade 3, erosions involving 10–50% of the circumference; grade 4, confluent circumferential erosions; and grade 5, complicated lesions including deep ulcer, stricture formation and Barrett's oesophagus. Routine biopsies were taken particularly from oesophageal or gastric lesions to exclude neoplasia. More specimens were also taken to diagnose Barrett's oesophagus, and to test for H. pylori by both urease activity and histology. Barrett's oesophagus was diagnosed in instances of an irregular squamo-columnar junction, with one or several tongues of columnar epithelium reaching into the tubular oesophagus, in addition to biopsy-proven intestinal metaplasia containing goblet cells. Intestinal metaplasia associated with a completely regular squamo-columnar junction and irregular junction alone without intestinal metaplasia were not diagnosed as Barrett's oesophagus.5, 10 A biopsy specimen was taken from each quadrant of the lower oesophagus in patients whose columnar epithelium extended to ≤ 5 cm, and biopsies were repeated at each 3–5-cm interval in those with longer segments, in order to confirm the diagnosis of Barrett's epithelium and to detect any co-existing dysplasia.

Statistical analyses

The Mann–Whitney test was used to compare values such as the oesophageal scores, age, etc. Fisher's exact test was used for the comparison of categorical variables, such as gender, number of smokers, cases of oesophagitis, Barrett's oesophagus, etc. In the logistic regression analyses, the occurrence of oesophagitis, including Barrett's oesophagus, was chosen as the outcome variable, whereas the social habits, demography, type of gastric surgery and length of time that had elapsed since gastric surgery served as predictor variables. The odds ratio for logistic regression, with forward stepwise selection of variables to remove the insignificant factors, and the 95% confidence intervals served to express the strength of the influence exerted by the relevant predictor variable.10


Two hundred and forty patients were included: 140 had gastric surgery and 100 had intact stomachs. In the surgical group, 13 other patients were considered, but not entered, because of excess alcohol intake (n = 4) and uncontrolled systemic diseases (n = 9). In those with intact stomachs, 18 other patients were not entered because of recent strokes (n = 3), multiple drug therapies (n = 8) and uncontrolled systemic diseases (n = 7).

The demographic characteristics and clinical presentations of the subjects included in the study are shown in Tables 1 and 2. Patients with post-surgical stomachs were more likely to be males, smokers and older than patients with intact stomachs. The majority of patients in both groups presented with dyspeptic symptoms, suggestive of peptic ulcer, gastro-oesophageal reflux, or both.

Table 1.  Demographic details of patients with or without gastric surgery
 Post-surgical stomachIntact stomach
  • Significant rise in the post-surgical group, P = 0.001.

  • † 

    Median (interquartile range).

Males106 (76%)*49 (49%)
Females34 (24%)51 (51%)
Age (years)69 (61–77)*58 (47–65)
Smokers69/128 (54%)*30/94 (32%)
Alcohol drinkers74/125 (59%)46 (46%)
Hiatus hernia31 (22%)24 (24%)
Table 2.  Clinical presentation in patients with post-surgical (n = 137) and intact (n = 100) stomachs [number (%)]
 Post-surgical stomachIntact stomach
Ulcer symptoms50 (36%)29 (29%)
Reflux symptoms23 (17%)37 (37%)
Ulcer + reflux symptoms12 (9%)18 (18%)
Weight loss22 (16%) 2 (2%)
Blood loss 6 (4%) 6 (6%)
Anaemia24 (17%) 8 (8%)

The details of drug therapy at the time of endoscopy are presented in Table 3. The two groups were not identical (χ2 = 9.99; d.f. = 4; P = 0.04), but there were no significant differences in the numbers of patients taking the individual drugs in either group. The possible influence of drug intake on the prevalence of oesophagitis is described under the logistic regression analyses (see below). Patients were comparable with respect to the prevalence of H. pylori: 24% in post-surgical stomachs and 34% in intact stomachs.

Table 3.  Drug therapy in patients with post-surgical (n = 132) and intact (n = 95) stomachs [number (%)]
 Post-surgical stomachsIntact stomachs
None35 (27%)33 (35%)
Proton pump inhibitors38 (29%)37 (39%)
H2-receptor antagonists16 (12%) 8 (8%)
Simple analgesics16 (12%) 3 (3%)
Low-dose aspirin27 (21%)14 (15%)

In subjects with post-surgical stomachs, the nature of the surgical procedure and when it was performed prior to endoscopy are shown in Table 4. The most common procedure involved vagotomy (n = 83): this was combined with gastro-enterostomy (n = 62) or pyloroplasty (n = 21). In the post-surgical group as a whole, the operation was carried out 27 years (19–31 years) [median (interquartile range)] prior to endoscopy.

Table 4.  Type of gastric surgery and when it was performed prior to endoscopy (years) [median (interquartile range)]
Surgical procedureTime
Vagotomy + gastro-enterostomy (n = 62)27 (20–31)
Vagotomy + pyloroplasty (n = 21)29 (20–30)
Gastro-enterostomy (n = 15)27 (21–35)
Bilroth I (n = 14)32 (10–42)
Bilroth II (n = 28)24 (15–30)
All patients (n = 140)27 (19–31)

Reflux and Barrett's oesophagitis

The prevalence of all cases of reflux, including Barrett's oesophagus, as well as the details of the individual grades of oesophagitis and the overall oesophageal scores, are given in Tables 5 and 6. Patients with post-surgical stomachs tended to have a lower prevalence of reflux oesophagitis (P = 0.06); there were also fewer patients with Barrett's oesophagitis (P = 0.002) and milder oesophageal scores (P = 0.03) in the post-surgical group.

Table 5.  Reflux and Barrett's oesophagitis [number (%)] in subjects with intact stomachs (controls), compared with the index group and subgroups (all gastric surgery patients, vagotomy patients and those with post-surgical bile reflux gastritis)
 OesophagitisBarrett's oesophagus
Intact stomachs (n = 100)37 (37%)11 (11%)
All surgery patients (n = 140)36 (26%)2 (1.4%)
P values0.060.002
Vagotomy (n = 83)19 (23%)
P values0.05
Reflux gastritis (n = 119)31 (26%)2 (1.7%)
P values0.110.007
Table 6.  Grades of oesophagitis [number (%)] and median (interquartile range) of all scores in the gastric surgery group and subgroups, compared with patients with intact stomachs (controls)
(n = 100)
All surgery
(n = 140)
(n = 83)
Reflux gastritis
(n = 119)
  • Comparing all oesophageal scores in the post-surgery group and subgroups with those of patients with intact stomachs (controls).

Grade 063 (63%)104 (74%)64 (77%)88 (74%)
Grade 1 8 (8%)  9 (6.4%) 5 (6%) 8 (7%)
Grade 211 (11%)15 (11%)10 (12%)12 (10%)
Grade 3 2 (2%)  6 (4%) 3 (3.6%) 6 (5%)
Grade 4 1 (1%)  0 0 0
Grade 515 (15%)  6 (4%) 1 (1.2%) 5 (4.2%)
All scores 0 (0–2)  0 (0–1) 0 (0–0) 0 (0–1)
P values*   0.03 0.02 0.046

Subgroup analyses showed that both cases of post-surgical Barrett's oesophagus were found in subjects with bile reflux gastritis. Patients with reflux gastritis still had milder oesophageal endoscopic scores than subjects with intact stomachs, although the prevalence of oesophagitis was not significantly lower. On the other hand, vagotomy patients showed fewer cases of oesophagitis, lower oesophageal scores and no cases of Barrett's oesophagus.

The Barrett's epithelium segment in subjects with or without gastric surgery had a length of 5 cm (3–8 cm) [median (interquartile range)]. The longest segment, 20 cm, was found in a patient with bile reflux gastritis who also had gastro-enterostomy without vagotomy.

Logistic regression analyses

These were carried out for erosive and all cases of post-gastric surgery oesophagitis, and similar findings were noted. The results of the logistic regression analyses for the development of erosive oesophagitis are presented in Table 7. With the exception of time since gastric surgery, various factors tested in the univariate analysis were found to have either no effect or a marginal influence. This was highlighted in the multivariate analysis, which showed that oesophagitis was less likely to develop in older patients and those with a longer time since gastric surgery.

Table 7.  Logistic regression analyses for the development of erosive oesophagitis after gastric surgery
 Odds ratio95% confidence intervalP
Univariate analysis
 Male gender2.080.66–6.500.21
 H. pylori0.930.23–3.770.92
 Reflux gastritis1.360.37–5.020.65
 Acid suppressants1.630.65–4.140.30
 Time since surgery0.950.91–0.990.02
Multivariate analysis
 Time since surgery0.880.80–0.96< 0.001

Neoplastic change

Six patients with intact stomachs (6%) were found to have gastric adenocarcinoma: one affected the distal stomach, and the rest were diagnosed in the proximal body or the gastro-oesophageal junction.

Five other cases of gastric malignancy (3.6%) developed in subjects with previous gastric surgery, 4–31 years after their operations. Four patients had adenocarcinoma: three after vagotomy and gastro-enterostomy, and one after Bilroth II operation. The fifth patient had intra-mucosal carcinoma following gastro-enterostomy.


This study shows that the prevalence of oesophageal damage, including reflux and Barrett's oesophagitis, is low and the oesophageal endoscopic scores are relatively mild in gastric surgery patients, particularly after vagotomy and despite bile reflux gastritis. In addition, the likelihood of developing oesophageal damage seems to decline with time after surgery.

Oesophagitis can develop following distal partial gastrectomy, Bilroth I or Bilroth II, due to the acute exposure of the oesophageal mucosa to excessive alkaline biliary as well as acid refluxate.11 Oesophagitis is almost invariable after proximal gastrectomy carried out for carcinoma of the gastric cardia, and is usually related to increased acid reflux.12 In our study, we included patients with a wide range of surgical procedures, particularly those involving acid inhibition, i.e. vagotomy, and those that might facilitate bile reflux, i.e. pyloroplasty or gastro-enterostomy. The possible relationship between Barrett's oesophagus and other factors, such as H. pylori, was not clear in previous studies.8, 11, 12 Our patients were assessed after a long period, a median of 27 years after surgery, and were more likely to demonstrate Barrett's transformation, dysplastic or neoplastic changes.

Our finding of a low prevalence of Barrett's oesophagus in post-gastric surgery patients is in agreement with the results of a recent study.10 This concluded that gastric surgery was not a risk for Barrett's oesophagus, and confirmed the suitability of gastric surgery patients for the assessment of the role of bile in oesophageal damage.10 In our single centre study, we also corrected for the presence of bile reflux gastritis and vagotomy, and assessed all grades of oesophageal damage, including Barrett's oesophagitis.

In the post-surgery group, patients were more often male, older and smokers than in the intact stomach group. This reflects two points. Firstly, gastric surgery was performed more commonly 20–30 years ago, prior to the advent of potent anti-ulcer therapy. Secondly, our patients were entered at random into the study, and no significant differences were found in their clinical presentations or indications for endoscopy. Numerically, more reflux symptoms were found in patients with intact stomachs and, in theory, this could have increased the prevalence of oesophagitis in this group. On the other hand, as mentioned above, the surgical group had their own risk factors, such as smoking and older age. Age was found to be a significant risk factor for oesophagitis as demonstrated in the multivariate logistic regression analysis. These considerations are unlikely to have affected the main findings of our study, as the post-surgical group had a lower prevalence of Barrett's oesophagus and milder global endoscopic oesophageal scores. In addition, Barrett's oesophagus was found in 1.4% of all patients with a history of gastric surgery and in 1.7% of those with post-surgical bile reflux gastritis, regardless of the patients' inherent risk factors for developing this condition. This is no different from the 1–2% rates of Barrett's oesophagus that can be extrapolated for the general population.13

It is interesting to note that both patients with Barrett's oesophagus after surgery had bile reflux gastritis and an intact vagus nerve: one after Bilroth II operation and the other after gastro-enterostomy. The latter had the longest Barrett's epithelium segment in the whole study at 20 cm. These two patients were likely to have been exposed to the combined effect of acid and bile reflux.3, 10, 11, 14

The low prevalence of Barrett's oesophagus in the study by Avidan et al.10 has been attributed by some workers, at least in part, to the possible protective effect of H. pylori on the oesophageal mucosa.15 Unlike our investigation, H. pylori status was not taken into account in the study by Avidan et al.10 We found no significant differences in the odds ratio for the development of oesophagitis in the presence or absence of H. pylori. In addition, a subset of peptic ulcer patients infected with H. pylori can be hyper-secretors of gastric acid,16 and this might lead to both peptic ulcer and oesophageal diseases. Therefore, the lower prevalence of oesophageal damage after gastric surgery in our patients, particularly after vagotomy, probably reflects acid hypo-secretion and the interaction between bile and gastric acid, although the measurement of these agents was beyond the remit of our study.

The likelihood of developing oesophageal damage seems to decline with time after gastric surgery, as shown by our logistic regression analysis. This implies that a process of oesophageal adaptation to the duodeno-biliary-gastric refluxate might take place after the initial rise in the incidence of oesophagitis that usually follows gastric surgery.8, 11, 12 This is probably different from gastric adaptation. In a recent study, we found that oesophagitis was observed regardless of the severity of histological bile reflux/chemical gastritis.17

A number of malignant cases were diagnosed in our study. Gastric surgery is a recognized risk factor for neoplastic change and, in our patients, has probably been highlighted by the long period of follow-up after surgery. In those with intact stomachs, the number of malignant cases reflects our criteria for endoscopic examination of patients with dyspepsia after the age of 45 years, and the inclusion of those with alarm symptoms, such as anaemia, weight loss, etc.

In conclusion, we have shown that patients with a long history of gastric surgery have mild grades of oesophageal damage and a low prevalence of Barrett's oesophagus, particularly after vagotomy. In our single centre study, we have assessed and corrected for a number of factors. These include bile reflux gastritis, vagotomy and other types of surgery, the time after surgery and H. pylori. Our findings are relevant to the understanding of the roles played by bile and gastric acid in the aetiology of reflux and Barrett's oesophagus.