Statistics from Altmetric.com
For several decades, the incidence of oesophageal adenocarcinoma (OAC) has shown the highest proportional rise among all cancers in the Western world. In order to detect neoplasia at a curable stage, endoscopic surveillance of patients with its precursor lesion Barrett's oesophagus (BO) is recommended by various societal guidelines and as such has been widely implemented over the past decades. However, although surveillance endoscopy is intuitively rational, there is currently no proof from randomised trials that this strategy reduces OAC mortality, and its cost-effectiveness has been questioned. The observation that <8% of OAC is preceded by a BO diagnosis further undermines the impact of surveillance. Recent population-based studies have shown a much lower overall cancer risk in BO than previously anticipated, and together with the growing emphasis on healthcare cost containment, the rationale for endoscopic surveillance has come under greater scrutiny.1 ,2 A 2014 cost-utility analysis of current surveillance protocols, incorporating progression estimates from large population-based studies, suggests that surveillance of patients with non-dysplastic BO is unlikely to be cost-effective.3 Alternative strategies such as ablation have raised hope to omit surveillance in patients; however, in reality considerable recurrence rates of both BO and carcinomas in particularly long segment BO (LSBO) still require continued endoscopic surveillance.4 Identification of patients with BO at highest risk of neoplastic progression is therefore a key priority in order to optimise effectiveness of surveillance programmes and potentially spare the majority of patients with BO the burden of regular endoscopies.
Most guidelines base the interval of surveillance endoscopy solely on the presence of dysplasia, which is present in 5–10% of patients at the time of BO diagnosis. Several other independent clinical predictors of neoplastic progression have been identified, such as age, gender, visceral obesity and molecular markers as p53, but have yet not been validated sufficiently in risk algorithms for further individualisation of surveillance intervals. Of these predictors, BO segment length appears one of the most valuable discriminators. Traditionally, BO has been defined as the replacement of oesophageal squamous epithelium by ≥3 cm of circumferential columnar epithelium with specialised intestinal metaplasia, a length that was purely arbitrary defined. The entity of short segment BO (SSBO), defined as a segment with a length of <3 cm, was only identified in 1994 by Spechler and colleagues.5 LSBO and SSBO are found in respectively 5% and 15% of patients undergoing endoscopy for reflux symptoms. Although OAC develops in both SSBO and LSBO, several recent studies have suggested that increasing segment length is a risk factor for neoplastic progression. In a meta-analysis, a lower annual incidence of OAC in patients with SSBO was found compared with all patients with BO in the study (0.19% vs 0.33% per year).6 In a retrospective study on the association between endoscopic features at BO diagnosis and risk of neoplastic progression, the risk of progression to adenocarcinoma or high-grade dysplasia increased by sevenfold in LSBO compared with SSBO (HR 7.1; 95% CI 1.74 to 29.04).7 A prospective cohort study with 713 patients with BO >2 cm in length reported an 11% increase in the risk of developing high-grade dysplasia (HGD) or OAC for each centimetre increase in BO length.8 In a similar study from the USA, the risk for progression increased by 28% for every 1 cm increase in length of segment.9 These results demonstrated that patients with SSBO have lower cancer risk than patients with LSBO, and thus less likely to benefit from surveillance endoscopy. As non-dysplastic SSBO is far more prevalent than LSBO, risk stratification on BO length could mean an important step towards increased cost-effectiveness of BO surveillance programmes. As the length of the BO segment and presence of dysplasia at BO diagnosis generally remains stable during follow-up, surveillance intervals can safely be established at baseline.1 ,10 At present, most BO guidelines do not differentiate surveillance intervals for SSBO and LSBO. The recently published guideline from The British Society of Gastroenterology does include differentiating surveillance intervals according to the length of the BO segment and recommends intervals of 2–3 years for non-dysplastic LSBO, 3–5 years for non-dysplastic SSBO and no surveillance for BO segments <1 cm.11
In this issue, the data presented by Pohl et al12 substantiate the evidence for risk stratification of patients with BO according to BO length. By combining German OAC incidence data, data from a German cohort of newly diagnosed T1 OAC and BO prevalence estimates from previous BO prevalence studies, they estimated annual cancer transition rates in relation to Barrett's segment length. For those with long, short and ultra-short BO, these rates were 0.24%, 0.03% and 0.01%, respectively. To detect one cancer, 3127 with SSBO and 11 239 with ultra-short BO would need to undergo annual surveillance endoscopy, again questioning the value of surveillance in these patients. Although it should be mentioned that several assumptions were made, the presence of dysplasia was unknown and different data sources were combined, potentially introducing bias; the reported findings appear robust under a broad range of assumptions and sensitivity analysis. The data concentrated on T1 OAC and the author’s acknowledge that most OAC are discovered at a later more advanced stage that may have different biology and BO segment length associations.
The substantiation of BO length as a risk factor for OAC development by Pohl et al should be considered an important piece of the puzzle for improving the effectiveness of BO surveillance. Development and validation of a much more wide-ranging prediction model, based on demographic, environmental, endoscopic, histological and molecular markers, is urgently needed. To improve cost-effectiveness, a prediction model should accurately identify both high-risk and low-risk patients with BO in order to allow discontinuation of surveillance in low-risk subgroups, such as possibly female SSBO patients without dysplasia. In the meantime, segment length is currently a useful discriminator to identify patients with BO who should be prioritised for surveillance. This risk factor has now been sufficiently substantiated, and inclusion in differentiated surveillance strategies is warranted.
Contributors Both authors equally contributed to the concept, design and revised and final version of the manuscript.
Competing interests None declared.
Provenance and peer review Commissioned; internally peer reviewed.
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.