Article Text


Endoscopic surveillance of patients with Barrett’s oesophagus
  1. R J Playford
  1. Gastroenterology Section, Imperial College School of Medicine, Hammersmith Hospital Campus, Du Cane Rd, London W12 0NN, UK;r.playford{at}

Statistics from

Numerous reviews and guidelines encourage us to undertake surveillance of patients with Barrett’s oesophagus (BO). Despite this, many experienced gastroenterologists consider there is insufficient evidence to support this approach. The definition of what constitutes BO remains an issue and is an important element of critically analysing reports. The previous idea that patients could be selected by the length of the Barret’s segment (>3 cm) appears incorrect as it does not have a major influence on the subsequent risk of carcinoma.1 The presence of intestinal metaplasia (IM) is important in the pathogenesis of cancer development but the absolute requirement to identify an area of IM before making a diagnosis of BO appears irrational. This is because virtually all patients with a columnar lined oesophagus also have some IM if enough biopsies are taken.2

Depending on the definition, 0.25–2% of the general population have BO.3,4 The introduction of surveillance for this huge number would therefore have a major impact on NHS finance and endoscopic resources. Most patients in surveillance programmes are initially identified as a result of an endoscopy for a reason other than reflux.5 The vast rump of BO therefore remains undetected in the community. Surveillance programmes, even if shown to be successful for the occasional individual patient, will therefore have little impact on the overall treatment of oesophageal cancer for the community as a whole.5

Studies following up case notes of patients with BO report the vast majority die of unrelated causes and, even in those who do develop oesophageal cancer, many will die from other pathologies.6 These studies provide no information however as to whether surveillance benefits patients.

Studies comparing the stage of disease and survival of patients identified from a surveillance programme against those presenting from the community have suggested a beneficial effect.7 However, these papers (studying very small numbers) do not analyse on an “intention to treat” basis, do not provide the total cost/benefit of their programmes, and fail to statistically separate dysplasia/cancer identified as a result of a true surveillance endoscopy from those having symptoms which would have resulted in an additional (unplanned) procedure. In addition, survival values for the surveillance group tend to consist of patients diagnosed with high grade dysplasia pooled with those having cancer. As the natural history of progression of such lesions is a matter of debate, with values varying from a cumulative cancer incidence at three years of 56%8 to a five year cumulative cancer incidence of only 9%,9 these comparisons have significant flaws.

The numbers of published works that allow a critical review of actual surveillance programmes are few. It is clear however that the median length of time that patients stay in surveillance programmes is short. This is usually less than five years and computer cost benefit modelling stating that the maximal benefit might be achieved if surveillance occurs every five years, for example, shows the limitations of such approaches. Our own review of patients entered into the Leicester General Hospital surveillance programme makes depressing reading. Of the 145 patients surveyed, five developed cancer but in only one was this detected as a result of a surveillance procedure, the remainder being diagnosed from endoscopy outside of surveillance protocol (due to previous default of the patient or the development of new symptoms).5 Importantly, we also showed that this result would not have been altered by following a more extensive biopsy protocol.5 Similarly, Nilsson et al identified five cancers from 199 patients with a mean surveillance duration of 3.9 years, quoting a value of $38 000 per cancer detected.10 However, of these five, one was unfit for surgery and two died early postoperatively and so received little benefit from such intervention.

The main paper that does suggest a benefit is from Wright et al who surveyed 166 patients with annual endoscopy for an average of 2.8 years. The vast majority died from causes unrelated to BO or became unfit for further surveillance.11 This study detected six cancers, five of which were apparently asymptomatic. The subsequent outcome of these patients is unclear although five were reported as having apparent node negative disease. Caution has to be shown however in basing a worldwide programme on one small publication. In addition, their apparent rate of cancer development was about three times higher than that generally considered to be true (0.5 per 100 patient years).12 It is also of interest that 3/6 of their surveyed patients who developed cancer did not have IM detected at initial endoscopy and should therefore not have been entered into the study according to some guidelines.

Some form of surveillance programme may be beneficial for patients with BO. However, an evidence based approach requires us to provide firm data in support of this expensive time consuming activity that uses up valuable endoscopy resources. Patients with dysplasia should probably be monitored. However, in the vast majority of patients with BO, the relatively high workload for every cancer detected means that we should be attempting to find ways of identifying those subjects who are at higher risk.

Key points

  • Most people with Barrett’s die from unrelated causes.

  • Impact of surveillance programmes on the amount of oesophageal cancer in the community is minimal but ties up expensive resources.

  • Studies to date have failed to adequately show any major benefit in altering patient outcome.

  • Guidelines recommending surveillance are inadequately evidence based.

  • Randomised trials showing differences in survival between regularly surveyed patients and no surveillance (or long surveillance intervals if considered unethical) are needed.

For patients without dysplasia, randomised trials with differences in death rates from oesophageal cancer are required, ensuring a distinction between those cancers diagnosed due to a surveillance procedure from those diagnosed from an additional endoscopy. Guidelines induce doctors to perform surveillance for fear of litigation. Based on current evidence, they should also state that not performing routine surveillance is a reasonable course of action.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.