The risk of oesophageal adenocarcinoma in a prospectively recruited Barrett’s oesophagus cohort

Abstract

Background

Varying rates of oesophageal adenocarcinoma (OAC) complicating Barrett’s oesophagus (BO) have been reported. Recent studies and meta-analyses suggest a lower incidence, questioning the value of endoscopic surveillance.

Aim

We aimed to retrospectively examine the rate of OAC, risk factors and causes of death in a prospectively recruited BO cohort.

Methods

Data from patients with BO from a cohort from 1982–2007 were studied. Patients were subdivided into surveyed, failed to attend surveillance and unfit for surveillance. Standardised mortality ratios (SMR) were calculated for common causes of death. Cox proportional hazards models were used to determine which factors were associated with progression to OAC.

Results

In total, 671 BO patients (61% male) were studied; 37 (76% male) were diagnosed with OAC. OAC incidence was 0.47% per annum and stable across three decades (1982–1991 0.56%, 1992–2001 0.46%, 2002–2012 0.41% (p = 0.8)). All-cause mortality was increased for the whole cohort (SMR 163(95% CI 145–183)). Mortality from OAC appeared higher in patients who failed to attend surveillance (SMR 3216(95% CI 1543–5916)) compared with surveyed (SMR 1753(95% CI 933–2998)) and those unfit for surveillance due to co-morbidity (SMR 440(95% CI 143–1025)). Multivariable analysis identified low-grade dysplasia (HR 4.4(95% CI 1.56–12.43), p = 0.005) and length of BO (HR 1.2(95% (1.1–1.3)), p < 0.001)) as associated with OAC.

Conclusions

Progression to OAC appeared stable over three decades at 0.47% per annum. Patients with BO had a modest increase in all-cause mortality and a large increase in OAC mortality, particularly if fit for surveillance. Low-grade dysplasia and the length of the BO segment were associated with developing OAC.

Introduction

Barrett’s oesophagus (BO) is the replacement, in response to chronic gastro-oesophageal reflux, of the normal squamous lining with metaplastic columnar epithelium.¹ It can be a precursor to the development of oesophageal adenocarcinoma (OAC),² which is associated with a poor overall 5-year survival of less than 15%.³ The increased risk of developing OAC in patients with BO has been reported to be between 10 and 55 times that of the general population.^4-6

Current guidelines recommend endoscopic surveillance of BO in order to diagnose high-grade dysplasia or OAC earlier and maximise the chance of cure.⁷ This has resulted in significant resource allocation towards surveillance programmes for patients diagnosed with BO, although the benefits of this approach are much debated and have not been demonstrated in a randomised controlled trial to date. A key determinant of the cost effectiveness of BO surveillance is the annual incidence of OAC.⁸ The annual risk of OAC in studies of subjects with BO has been reported to be between 0.2 and 2.9%.⁹ Recent studies have reported a lower rate of progression. For example, a population-based study of BO from Northern Ireland reported an annual incidence of OAC of 0.23%,¹⁰ calling into question the value of surveillance. However, during the same period, advances in endoscopic therapy for high-grade dysplasia and early cancer, such as endoscopic mucosal resection and radiofrequency ablation, have potentially increased the benefit of surveillance by avoiding the need for oesophagectomy to treat high-grade dysplasia or early OAC in some patients.^11,12

We have therefore retrospectively examined the risk of developing OAC, associated risk factors for OAC and the causes of death in a cohort of BO subjects prospectively recruited in two adjacent hospitals over three decades.

Methods

Barrett’s oesophagus cohort

All patients diagnosed with BO between 1982 and 2007 at City Hospital, Birmingham, and between 1997 and 2007 at the adjacent Sandwell General Hospital, West Bromwich, were included in a prospective database. To be diagnosed with BO and included in the database, patients had to have columnar-lined mucosa above the proximal margin of the gastric folds at endoscopy and evidence of intestinal metaplasia (IM) on biopsy. The database includes demographic data (age, gender, ethnicity), endoscopic findings (oesophagitis graded by the revised Savary–Miller system),¹³ length of BO segment, Barrett’s ulcer, stricture, hiatus hernia and histology (presence of dysplasia). Patients underwent surveillance if suitable every 2 years until they reached 75 years of age or developed co-morbidity that, in the opinion of the responsible clinician, precluded further surveillance due to the risks of oesophagectomy. For the majority of the time period studied, oesophagectomy was the only available intervention for high-grade dysplasia or OAC. High-grade dysplasia was managed by immediate repeat endoscopy and if also present on the repeat set of biopsies, after histopathology review, oesophagectomy was recommended. Quadrantic biopsies were taken every 2 cm throughout the Barrett’s segment in addition to targeted biopsies of any focal lesions and reported by a gastrointestinal histopathologist.

Patients who were diagnosed with OAC or high-grade dysplasia within 1 year of index endoscopy were excluded from the database to avoid including patients who were incident cases of OAC and not diagnosed in relation to the surveillance programme.

The patients were retrospectively divided into three groups for analysis:

1. Suitable for endoscopic surveillance and complied with surveillance (undergoing at least three endoscopies over at least 5 years of follow-up);

2. Suitable for endoscopic surveillance but failed to attend for non-medical reasons (i.e. failed to attend or declined repeat endoscopy) and were not deceased;

3. Not suitable for endoscopic surveillance due to age or co-morbidity in the opinion of the clinician responsible for the patient.

Cancer outcomes

Following ethical approval from the Black Country Research Ethics Committee (reference number 09/H1202/29), cancer registrations and causes of death were obtained retrospectively from the Office of National Statistics via the NHS Information Centre for Health and Social Care. Data from death certificates were cross-referenced with local pathology data to distinguish squamous cell carcinoma from adenocarcinoma of the oesophagus. OAC incidence was calculated as events per 100 years of follow-up. The end of follow-up was considered to be the date at which OAC was diagnosed, death or 31 December 2012.

Standardised mortality ratios

The common causes of death in this cohort were identified and standardised mortality ratios (SMR) calculated utilising West Midlands age and sex-specific mortality rates (1990 to 2005) (supplied by the West Midlands Cancer Intelligence Unit). The number of person-years at risk was calculated by sex and 5-year age group. The expected number of cases of OAC was calculated by multiplying mortality rates in the West Midlands population by the observed sex- and age-specific number of person-years at risk in the study population. The ratio of observed to expected cases of OAC was reported as a SMR, with a value of 100 indicating the same mortality as the background population and a value over 100 suggesting excess mortality.

Statistical methods

95% confidence intervals for SMRs were calculated from the exact Poisson distribution.¹⁴ Data are quoted as mean (standard deviation) unless otherwise stated. Continuous data was compared using a student’s t-test and categorical data using a chi-squared test. Cox proportional hazards models were used to determine which variables were associated with progression to OAC. SPSS version 20 was used for all statistical analysis (IBM, New York, USA).

Results

Barrett’s oesophagus cohort

In total, 671 patients were diagnosed with BO at City and Sandwell General Hospitals between 1982 and 2007. Within this cohort there were 409 (60.9%) men and 262 women (39.1%). The median age at diagnosis of BO was 64.6 (interquartile range (IQR) 54.4–73.9) years, with men diagnosed at an earlier age than women, (61.2 (IQR 52–71.1)) versus (69.2 (IQR 60.1–76.7) years, p < 0.001). The majority of patients were white Caucasian (628), with 31 South Asian and 12 Afro-Caribbean subjects.

Endoscopy findings at index endoscopy

The length of BO was recorded in 619 (92.2%) patients and the mean length was 5.3 (SD 3.06) cm. There were 206 (31%) patients with reflux oesophagitis (77 mild, 75 moderate, 54 severe). In 355 (53%) patients, a hiatus hernia was described at the index endoscopy. There were 76 (11.3%) patients with a benign oesophageal (Barrett’s) ulcer and 49 (7.3%) with a stricture at index endoscopy.

Histology at index endoscopy

Histology revealed IM at index endoscopy in 626 (93.3%) patients. The remaining 6.7% did not have histology taken at their first endoscopy, but subsequent endoscopies yielded biopsies showing IM with no dysplasia. There were 19 (2.8%) cases of low-grade dysplasia at index endoscopy.

Comparison of surveyed and non-surveyed patients

The demographics, endoscopic and histological findings of patients in the surveyed, failed to attend and unsuitable for surveillance groups are shown in Table 1.

Table 1.

Demographic, endoscopic and histological findings in the patients with Barrett’s oesophagus who were surveyed, failed to attend for surveillance and who were not suitable for surveillance

	Surveyed (n = 247)	Failed to attend surveillance (n = 184)	Not suitable for surveillance (n = 240)
Male	65.9%	70.7% (p = 0.7)	48.3% (p = 0.03)
Age at diagnosis median(IQR) (years)	55.5 (51.2–66.6)	58 (49.2–63.6), p = 0.02	77.2 (73.5–80.4), p < 0.001
Low-grade dysplasia at index endoscopy	8 (3.2%)	2 (1%), p = 0.2	9 (3.8%), p = 0.8
Length Barrett’s segment median (IQR) (cm)	6 (3–8)	5 (3–6), p = 0.1	5 (3–7), p = 0.6
Oesophagitis	83 (33.6%)	50 (27.2%), p = 0.3	73 (30.4%), p = 0.7
Hiatus Hernia	155 (62.7%)	77 (41.9%), p = 0.02	123 (51.3%), p = 0.2
Stricture	22 (8.9%)	8 (4.3%), p = 0.1	19 (7.9%), p = 0.7
Barrett’s ulcer	27 (10.9%)	16 (8.7%), p = 0.5	34 (14.2%), p = 0.4

^*p-values in comparison with surveyed group and not adjusted for multiple comparisons

Surveyed patients

Some 247 (161(65%) male) patients were surveyed. The median age at diagnosis of BO was 55.5 (IQR 51.2–66.6) years. Patients underwent a median of 5 (range 3–18) surveillance endoscopies, with a median study follow-up of 13.5 (range 5–25) years. The median length of BO was 6 (IQR 3–8) cm, with 22 patients having segments less than 3 cm.

Patients who failed to attend surveillance

Some 184 (124 (67%) male) patients did not attend endoscopic follow-up according to the study criteria; 37 had two endoscopies, the others had just one. The median endoscopic follow-up was 1.3 (range 1–4) years. The median age at diagnosis of BO was 58 (IQR 49.2–63.6) years and the study follow-up period from diagnosis 11 (range 1–24) years in this group. The median length of BO was 5 (IQR 3–6) cm and there were 15 patients with segments less than 3 cm.

Patients unsuitable for surveillance

In total, 240 (116 (48%) male) patients were not suitable for surveillance due to age or co-morbidity. The median age at BO diagnosis was 77.2 (range 73.5–80.4) years, the median study follow-up period from diagnosis was 8 (range 1–24) years and the median age at death was 84 (range 75–101) years.

Cancer outcomes and mortality

BO cohort

The median period of follow-up for the whole cohort was 11 (range 1–25, IQR 6.9–15) years and the total number of person-years of follow-up was 7784. There were 37 (28 (76%) male) patients diagnosed with OAC or with oesophageal cancer as a registered cause of death. The 5-year survival from diagnosis of OAC was 22% (8/37), with an additional three patients alive but not yet 5 years from diagnosis by the end of the study period. In total, 20 of the 37 OAC patients underwent resection (19 oesophagectomy and one endoscopic mucosal resection). Of these 20, 17 (85%) were diagnosed within the surveillance programme and their 5-year survival was 35% (6/17), with an additional two patients alive but with less than 5 years follow-up after diagnosis. The overall annual progression rate from BO to OAC was 0.47%. The annual rate of OAC progression from 1982 to 1991 was 0.56% (4 in 972 person-years of follow-up), from 1992 to 2001 0.46% (24 in 5212 person-years of follow-up) and from 2002 to 2012 0.41% (9 in 1601 person-years of follow-up). There was no statistically significant difference in the progression rate over the three decades (p = 0.8). One patient underwent oesophagectomy for high-grade dysplasia in the surveyed group.

The median age at diagnosis of BO was 62.6 (range 39–83) years and at diagnosis of OAC 69.9 (range 48–90) years. There was no significant sex difference in the rate of progression to OAC and in the median age of OAC diagnosis (men 69 (48–90) years, women 76 (66–85) years, p = 0.06) (Table 2). Only one of the 43 non-Caucasian patients with BO within the cohort developed OAC compared with 36 out of 664 Caucasian patients (p = 0.6).

Table 2.

The incidence of oesophageal adenocarcinoma in patients with Barrett’s oesophagus by age and sex

Age at Barrett’s oesophagus diagnosis (years)	Oesophageal adenocarcinoma per years of follow-up	Oesophageal adenocarcinoma per years of follow-up in men	Oesophageal adenocarcinoma per years of follow-up in women
<50	6/1678 (0.4%)	5/1196 (0.4%)	1/482 (0.2%), p = 0.43
50–59	9/1727 (0.5%)	9/1248 (0.7%)	0/479 (0%), p = 0.14
60–69	12/2125 (0.6%)	7/1193 (0.6%)	5/932 (0.5%), p = 0.9
70–79	6/1844 (0.3%)	6/864 (0.7%)	1/980 (0.1%), p = 0.06
>80	4/410 (1%)	1/107 (0.9%)	2/303 (0.7%), p = 1

^*p-value comparing males and females

The SMR for OAC in the whole cohort was increased at 1316 (95% CI 867–1914) and the overall SMR in the cohort was increased at 163 (95% CI 145–183). There were 291 deaths during the study period. The main causes of death were cardio-respiratory (113), old age/dementia (29) and OAC (27) (Table 3).

Table 3.

Standardised mortality ratios and causes of death in the patients with Barrett’s oesophagus who were surveyed, failed to attend surveillance and who were not suitable for surveillance

	Surveyed (n = 88)	Failed to attend surveillance (n = 46)	Not suitable for surveillance (n = 157)
SMR OAC (95% CI)	1753 (933–2998)	3216 (1543–5916)	440 (143–1025)
OAC	11 (12.5%)	9 (19.6%), p = 0.3	7 (4.5%), p = 0.04
Cardiovascular	18 (20.5%)	7 (15.2%), p = 0.6	32 (20.4%), p = 1
Respiratory	15 (17.1%)	8 (17.4%), p = 1	33 (21%), p = 0.5
Lung cancer	7 (8%)	5 (10.9%), p = 0.8	9 (5.7%), p = 0.6
Colon cancer	5 (5.7%)	5 (10.9%), p = 0.3	6 (3.9%), p = 0.5
Dementia	5 (5.7%)	1 (2.2%), p = 0.7	12 (7.6%), p = 0.8
Old age	2 (2.2%)	1 (2.2%), p = 1	8 (5.1%), p = 0.5
Cerebrovascular disease	5 (5.7%)	4 (8.7%), p = 0.5	10 (6.3%), p = 1
Other cancer	12 (13.6%)	6 (13%), p = 1	14 (8.9%), p = 0.3
Other	8 (9%)	0 (0%), p = 1	26 (16.6%), p < 0.001

SMR: Standardised Mortality Ratio; OAC: oesophageal adenocarcinoma

^*p-values in comparison to surveyed group and not adjusted for multiple comparisons

Surveyed patients

Among surveyed patients, 21 (8.5%) developed OAC with an annual rate of 0.58% and a SMR of 1753 (95% CI 933–2998). Four (1.6%) developed OAC a median of 5 (range 3–7) years after finishing endoscopic surveillance, none of whom were suitable for surgery. Among patients undergoing surveillance, there was one interval inoperable cancer, five deaths in the early post-operative period and one death from metastatic disease 1 year after surgery. Nine patients had curative surgery but two have subsequently died (one from pneumonia and one from sepsis). There were 99 deaths in this group. The other major causes of death were cardio-respiratory (33), lung cancer (7) colon cancer (5) and dementia (5) (Table 3).

Patients who failed to attend surveillance

There were nine (4.9%) patients who failed to attend surveillance who developed OAC, with an annual rate of developing OAC of 0.42% and a SMR of 3216 (95% CI 1543-5916). Six patients returned with symptomatic inoperable OAC. Three patients underwent oesophagectomy with one patient dying in the early post-operative period and two patients returned with metastases more than a year after surgery. All were recorded as dying from oesophageal cancer. The median age at death in this group was significantly younger than those surveyed (66 (IQR 59–73) years versus 74 (IQR 66–80) years (p < 0.001)). The main causes of death in this group were cardio-respiratory (15), lung cancer (5), colon cancer (5) and cerebrovascular disease (4) (Table 3).

Patients unsuitable for surveillance

Seven (2.9%) patients died of OAC (median age 83 (range 78–90) years), which is significantly less than the surveyed group (p = 0.04). None of these seven patients had surgery. The annual rate of OAC was 0.32% and the SMR for OAC in this group was 440 (CI 143–1025). The main causes of death were cardio-respiratory (55), dementia (12), sepsis (12), cerebrovascular disease (10) and colon cancer (6) (Table 3).

Univariable and multivariable analysis

The results of univariable Cox proportional hazards regression analysis are shown in Table 4. Male sex (HR 2.2 (95% CI 1.01–4.5), p = 0.048)), low-grade dysplasia at index endoscopy (4.9 (1.74–13.87), p = 0.003) and length of BO at index endoscopy as a continuous variable (1.21 (1.09–1.31), p < 0.001) were associated with progression to OAC.

Table 4.

Univariable Cox regression analysis of risk factors at index endoscopy for the progression of Barrett’s oesophagus to oesophageal adenocarcinoma

Variable	Hazard ratio (95% CI)
Male sex	2.2 (1.01–4.5), p = 0.048
Age at Barrett’s oesophagus diagnosis	1.02 (0.99–1.05), p = 0.28
Length of Barrett’s oesophagus	1.2 (1.09–1.31), p < 0.001
Hiatus hernia	1.3 (0.66–2.58), p = 0.45
Oesophagitis	0.46 (0.2–2.2), p = 0.5
Barrett’s ulcer	0.3 (0.07–1.3), p = 0.1
Benign oesophageal stricture	1.35 (0.56–3.28), p = 0.5
Low-grade dysplasia	4.91 (1.74–13.87), p = 0.003

The results of multivariable Cox proportional hazards regression analysis are shown in Table 5. Low-grade dysplasia (HR 4.4 (95% CI 1.56–12.43), p = 0.005) and length of BO (1.2 (1.1–1.32), p < 0.001) remained associated with the development of OAC. However, male sex was not found to be associated with OAC development on multivariable analysis.

Table 5.

Multivariable Cox regression analysis of risk factors at index endoscopy for the progression of Barrett’s oesophagus to oesophageal adenocarcinoma

Variable	Hazard Ratio (95% CI)
Low-grade dysplasia	4.4 (1.56–12.43), p = 0.005
Length of Barrett’s oesophagus	1.21 (1.1–1.32), p < 0.001
Male sex	1.91 (0.9–4.1), p = 0.09

Discussion

In this cohort study, the annual rate of development of OAC in patients with BO was 0.47% and appeared stable over the three-decade period studied. Reports in the literature describe widely varying estimates of the annual risk of OAC in BO from 0.2 to 2.9%.¹⁵ A meta-analysis of 57 surveillance studies from different geographical areas reported a pooled annual OAC incidence in patients with BO without dysplasia at baseline of 0.33(95% CI 0.28–0.38)%.⁹ Two recent large population-based studies suggest that the annual OAC risk in BO is even lower.^10,16 In a population-based study of 8522 BO patients in Northern Ireland, the annual risk of OAC was 0.23 (95% CI 0.18–0.29)% per year in those with IM.¹⁰ In a nationwide study based on histopathology records of 11,028 BO patients in Denmark, an even lower annual risk of OAC of 0.12 (95% CI 0.09–0.15)% was reported.¹⁶ It is important to recognise that other risk factors in BO need to be taken into account when assessing the annual risk of OAC, in particular the length of the BO segment. In a Dutch cohort study of 713 BO patients, the risk of progression to OAC increased by 1.11 per extra centimetre of BO segment length,¹⁷ and in the present study increasing BO segment length as a continuous variable was associated with OAC. The length of BO segment was only available in 20% of patients in the Northern Ireland population-based study¹⁰ and in 0% of the Danish population-based study, since it was based on histopathology records.¹⁶ In Northern Irish patients with a short segment of BO (<3 cm), the annual risk of OAC was 0.07 (95% CI 0.02–0.20)% and it was 0.22 (95% CI 0.13–0.37)% in longer segment BO.¹⁰ Both population-based studies are likely to have included a majority of short-segment BO patients, since they are more common than long-segment BO, thereby lowering their estimates of the annual risk of OAC compared with the present study. Finally, when meta-analysis is limited to studies from the UK, there is a suggestion of a higher pooled annual incidence of OAC of 1 (CI 0.7–1.4)% compared with other geographical regions.⁹

Patients with BO who develop OAC within a surveillance programme tend to have earlier stage disease and an associated improved 5-year survival, compared with patients diagnosed with OAC through investigation of symptoms,¹⁸ accepting that lead time bias plays a part in the improved survival figures. This is borne out in the present study, in which nine of the 15 patients diagnosed with OAC during surveillance underwent successful curative surgery, whereas, in contrast, none of the patients who failed to attend for surveillance or who were unsuitable for surveillance underwent successful curative surgery. It should also be recognised that the recent availability of endoscopic mucosal resection for early OAC has the potential to benefit even more BO patients by avoiding the post-operative mortality seen in the present historical series and extending curative treatment to older BO patients than was feasible during much of the time period studied.

Identifying those who would benefit most from endoscopic surveillance continues to be challenging. An increased incidence of OAC in BO with increasing age has been reported by some^10,19 but not all authors.²⁰ In the present study, it is interesting to note that only one case of OAC developed in a woman under the age of 60 compared with 14 cases in men under 60, and men might be a group who would particularly benefit from surveillance. This is in keeping with previous work showing a 17-year delay in the development of OAC in women compared with men, possibly due to a protective effect of oestrogen that is lost during the menopause.²¹ Despite this apparent age-associated gender difference in the development of OAC, male sex was only associated with OAC development on univariable but not on multivariable analysis when the effects of low-grade dysplasia and BO segment length were controlled for. This is in contrast to other studies in which male sex has been reported to be an independent risk factor for the development of OAC.^22,23 For example, a recent large population-based study from the Netherlands reported that male sex was independently associated with the progression to the combined outcome of OAC/high-grade dysplasia with a hazard ratio of 2.12 (CI 1.7–2.65).²³ Analysis of the Surveillance, Epidemiology, and End Results (SEER) programme revealed that the age-adjusted incidence rates for oesophageal cancer are 6 to 8-fold higher in men than in women.²⁴ Given that the lower 95% confidence interval for the hazard ratio on multivariable regression analysis for male sex was 0.9, it is likely that there was insufficient power within the present study to detect a significant association between OAC and male sex.

It is clear from the present study that identifying low-grade dysplasia at baseline endoscopy is a particularly important predictor of the subsequent development of OAC, with a more than 4-fold increased risk compared with those without low-grade dysplasia. This is consistent with other studies that have reported an association between low-grade dysplasia and the development of OAC.^17,20,23 A recent Dutch study reported that the presence of low-grade dysplasia at index endoscopy was the strongest independent predictor for progression to high-grade dysplasia or OAC (risk ratio 9.7 (95% CI 4.4–21.5)).¹⁷ However, reported rates of progression to OAC in patients with low-grade dysplasia vary significantly, and this is likely to relate to the poor inter-observer reproducibility of a diagnosis of low-grade dysplasia among histopathologists.²⁵ A consensus on low-grade dysplasia among expert pathologists and the addition of p53 immunohistochemistry appear promising in stratifying patients into those at high and low risk of progression to OAC, and guiding future surveillance or mucosal ablation.²⁶

As noted above, longer segments of BO are associated with a higher risk of OAC. The present study found an association between risk of OAC and increasing BO length (HR 1.2, CI 1.1–1.32). A recent prospective cohort study reported that the risk ratio for the development of high-grade dysplasia or cancer was 1.1 for every centimetre increase in length.¹⁷ Progression to OAC has been reported to be 1.5 times higher in patients with a segment of BO that is 3 cm or longer compared with segments shorter than 3 cm.²⁷ The consistent data on the influence of segment length on risk of OAC strongly suggest that risk stratification for BO surveillance should include this factor, with shorter BO segments being surveyed less frequently than longer segments.⁷

In the present study, patients with BO had an increase in all-cause mortality (SMR 163). Several previous studies have reported increased mortality in BO patients.^28,29 A study from Leicester in the UK reported a striking overall SMR in BO patients adjusted for age of 552 (95% CI 466–638) for men and 455 (95% CI 357–552) for women.²⁸ A larger population-based study from the UK reported a 21% relative increased risk of all causes of death in BO patients.²⁹ In common with the present study, the predominant causes of death among BO patients in these studies were cardiovascular and respiratory. This may relate to an increased prevalence of unreported cardiovascular and respiratory risk factors, such as smoking and increasing body mass index in the BO patients studied, as several studies have reported that BO patients are more likely to smoke and have more central obesity.^30,31 However, two population-based studies have shown no increase in all-cause mortality.^32,33 The overall mortality rate in a cohort study from Northern Ireland of 2373 BO patients was very similar to that of the local population.³² Individuals with BO in the Northern Ireland cohort even appeared to have a lower risk of cerebrovascular disease when compared with the local population (SMR 0.65 (95% CI 0.37–0.93)). The increased mortality rate from OAC among the BO patients studied was expected, but to our knowledge this is the first time it has been studied among BO patients being surveyed, those that defaulted from surveillance and those unsuitable for surveillance. The SMR for OAC among those who failed to attend for surveillance was highest at 3216, although the 95% confidence intervals crossed those of the surveyed group and there was therefore no significant difference between the two groups. However, it was striking that the median age of death was lower among the group that failed to attend for surveillance compared with the surveyed group, and it is possible that failing to attend for surveillance is associated with other high-risk behaviours such as smoking or obesity.

It should be recognised that the present study is prone to a number of biases. The mean BO segment length of the cohort was 5 cm and it is likely that the requirement for IM on biopsy may have excluded a number of patients with shorter segments of BO. Patients with short BO segments will have had fewer biopsies taken at endoscopy to establish the presence of IM that is known to be patchy. The risk of OAC appears significantly lower in patients without IM on biopsy,¹⁰ and this selection bias is likely to explain the higher OAC incidence in the cohort compared with population-based studies.¹⁰ Mortality from OAC was examined rather than survival in an attempt to minimise lead time bias, in which detecting cancers in their pre-clinical phase results in an apparent increase in survival as the diagnosis is made earlier in the course of the disease. This study is also dependent on the accuracy of the diagnosis on death certificates, and it has been reported that gastrointestinal disorders are under-diagnosed as a cause of death.³⁴ The elevated SMR also needs to be interpreted with some caution. The study cohort consists of patients diagnosed with BO at City and Sandwell Hospitals in a deprived area on the Western edge of Birmingham. The mortality rates utilised for comparison were from the whole West Midlands region, covering a larger and potentially more demographically diverse population. Finally, a number of factors that are known to be important in the aetiology of OAC, including increasing body mass index, smoking status and medication usage, could not be accounted for in the present study.

In conclusion, in a prospectively recruited cohort of BO patients over a 30-year period the rate of developing OAC appeared stable at 0.47% per annum. BO patients with low-grade dysplasia at index endoscopy and with longer BO segment lengths were at increased risk of developing OAC, and stratification of the frequency of endoscopic surveillance by these risk factors would be appropriate. BO was associated with a modest increase in all-cause mortality and a large increase in OAC mortality, in patients fit for endoscopic surveillance.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.