Risk of adenocarcinoma in Barrett's oesophagus: population based studyBMJ 2003; 327 doi: http://dx.doi.org/10.1136/bmj.327.7414.534 (Published 04 September 2003) Cite this as: BMJ 2003;327:534
- Liam Murray (), senior lecturer in epidemiology1,
- Peter Watson, consultant gastroenterologist2,
- Brian Johnston, consultant gastroenterologist2,
- James Sloan, consultant pathologist2,
- Inder Mohan Lal Mainie, specialist registrar in gastroenterology3,
- Anna Gavin, director1
- 1Northern Ireland Cancer Registry, Department of Epidemiology and Public Health, Queen's University of Belfast, Belfast BT12 6BJ
- 2Royal Hospitals Trust, Belfast BT12 6BA
- 3Ulster Hospital, Dundonald, Belfast BT16 1RB
- Correspondence to: L Murray
- Accepted 3 July 2003
Endoscopic surveillance of Barrett's oesophagus is now routine.1 Cost effectiveness depends on the risk of oesophageal adenocarcinoma.2 The magnitude of this risk is unclear because most previously published studies were small and inconclusive.3 Except for one,4 these studies were not population based but investigated patients at one or more centres. Selection bias or the effect of common losses to follow up were not assessed.5 We investigated the risk of oesophageal malignancy in a large cohort of unselected patients with Barrett's oesophagus in Northern Ireland, where all incident cancers are routinely identified.
Participants, methods, and results
We examined the pathology reports relating to all oesophageal biopsies in Northern Ireland between January 1993 and December 1999. We included every adult identified within Northern Ireland (population 1.7 million) as having oesophageal columnar epithelium. We excluded biopsies taken at the oesophagogastric junction.
We defined Barrett's oesophagus as the presence of columnar metaplasia in the oesophagus irrespective of whether Barrett's mucosa was reported (although we used this fact to further classify the biopsies as “macroscopic Barrett's oesophagus” or otherwise). We did not use data on segment length because it was often absent from reports. We subdivided biopsies finding Barrett's oesophagus further if the pathologist specifically stated that specialised intestinal metaplasia or goblet cells were definitely present or absent. We excuded malignant biopsies.
We identified patients in the cohort and followed them up for death and oesophageal malignancy (oesophageal adenocarcinoma and histologically unspecified oesophageal carcinomas or malignancies) until the end of 2000 by matching with death records from the Registrar General's Office and the Northern Ireland Cancer Registry's database of incident cancers. We excluded oesophageal malignancies diagnosed within six months of the initial biopsy. We identified all patients in Northern Ireland with Barrett's oesophagus who had an oesophagectomy or ablative treatment for high grade dysplasia between January 1993 and December 2000. We calculated person years of follow up until diagnosis of the malignancy, death, or 31 December 2000. We estimated confidence intervals from the Poisson distribution.
Between 1993 and 1999, of 15 670 oesophageal biopsies, 4955 (from 2969 patients) met our criteria for Barrett's oesophagus (table). The mean follow up was 3.7 (range 1 to 8) years, with 11 068 person years of follow up. We found 29 oesophageal malignancies in the cohort. Four patients had an oesophagectomy for high grade dysplasia and two had ablative laser treatment. Oesophageal malignancy was 0.26% (0.18% to 0.38%) a year overall and 0.4% (0.26% to 0.59%) a year for patients with specialised intestinal metaplasia. The malignancy rate in men was 2.5 times that in women. Only for men older than 70 with specialised intestinal metaplasia was incidence greater than 1% per year.
Patients with Barrett's oesophagus are at low risk of oesophageal adenocarcinoma, and this risk is almost exclusively in patients with specialised intestinal metaplasia. Surveillance of patients with Barrett's oesophagus at a risk of malignant transformation of 1% per year may be cost effective,2 but only men aged 70 or more are at this risk, and limiting surveillance to them would miss two thirds of cancers. The length of follow up in our study was brief, but up to eight years after diagnosis we found no increased risk of malignancy with time (data available from authors).
Robust methods for stratifying risk and targeting surveillance in Barrett's oesophagus are needed. Although we did not use specific protocols for oesophageal biopsy, because the data are from routine clinical practice, our findings may have more relevance for standard care than previous studies.
We thank Colin Fox, Richard Middleton, Tom Wylie, and the tumour verification officers of the Northern Ireland Cancer Registry for help processing pathology records; the administrative, medical, and pathology staff of local healthcare trusts; the staff of the Directorate of Information Services for help compiling the Northern Ireland Barrett's register; and Deirdre Fitzpatrick and Chris Patterson for statistical advice.
Contributors LM had the idea for the study, constructed the register, analysed the data, and wrote the intial draft. AG, PW, BJ, and JS, designed the study and interpreted the data. IMLM helped construct the register. All authors commented on draft manuscripts. LM is guarantor.
Funding Ulster Cancer Foundation.
Conflict of interest None declared.
Ethical approval Research ethics committee, Queen's University Belfast.