Objective Surveillance interval protocols after complete remission of intestinal metaplasia (CRIM) post radiofrequency ablation (RFA) in Barrett’s oesophagus (BE) are currently empiric and not based on substantial evidence. We aimed to assess the timeline, location and patterns of recurrence following CRIM to inform these guidelines.
Design Data on patients undergoing RFA for BE were obtained from prospectively maintained databases of five (three USA and two UK) tertiary referral centres. RFA was performed until CRIM was confirmed on two consecutive endoscopies.
Results 594 patients achieved CRIM as of 1 May 2017. 151 subjects developed recurrent BE over a median (IQR) follow-up of 2.8 (1.4–4.4) years. There was 19% cumulative recurrence risk of any BE within 2 years and an additional 49% risk over the next 8.6 years. There was no evidence of a clinically meaningful change in the recurrence hazard rate of any BE, dysplastic BE or high-grade dysplasia/cancer over the duration of follow-up, with an estimated 2% (95% CI −7% to 12%) change in recurrence rate of any BE in a doubling of follow-up time. 74% of BE recurrences developed at the gastro-oesophageal junction (GOJ) (24.1% were dysplastic) and 26% in the tubular oesophagus. The yield of random biopsies from the tubular oesophagus, in the absence of visible lesions, was 1% (BE) and 0.2% (dysplasia).
Conclusions BE recurrence risk following CRIM remained constant over time, suggesting that lengthening of follow-up intervals, at least in the first 5 years after CRIM, may not be advisable. Sampling the GOJ is critical to detecting recurrence. The requirement for random biopsies of the neosquamous epithelium in the absence of visible lesions may need to be re-evaluated.
- barrett’s oesophagus
- oesophageal cancer
- endoscopic procedures
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Recurrence rates of intestinal metaplasia and dysplasia following successful ablation for Barrett’s oesophagus (BE) are well established.
Recent data suggest that the rates of recurrence are highest in the initial year after remission and may decline thereafter, suggesting widening of surveillance intervals after the initial year.
Data on the timeline and endoscopic patterns of recurrence are scarce.
What are the new findings?
Cumulative recurrence rates of intestinal metaplasia and dysplasia appear to increase progressively with follow-up time, in this multicentre international cohort study.
Most recurrences occur at the gastro-oesophageal junction (GOJ) or in the distal 5 cm of the oesophagus.
Most dysplastic recurrences at the GOJ are not visible, while most recurrences in the tubular oesophagus are visible.
The yield of neosquamous epithelium biopsies from the tubular oesophagus in the absence of visible recurrences is very low (<1.0%).
How might it impact on clinical practice in the foreseeable future?
Widening of surveillance intervals to detect recurrence after BE ablation may be premature in the absence of additional long-term data.
Practice guidelines recommending Seattle protocol biopsies in the entire neosquamous epithelium in the absence of visible recurrences may need to be re-evaluated.
Biopsies of the GOJ should be performed even in the absence of visible lesions.
Barrett’s oesophagus (BE) is a condition that develops when the normal squamous epithelium is replaced by columnar mucosa with specialised intestinal metaplasia (IM). This process confers an increased risk of progression to oesophageal adenocarcinoma, with an estimated incidence rate of approximately 0.33 per patient year.1 This risk increases when either low-grade dysplasia (LGD) or high-grade dysplasia (HGD) develops in the setting of BE.2 Radiofrequency ablation (RFA) after endoscopic resection of visible lesions has been shown to reduce the risk of cancer progression in patients with both LGD and HGD and has therefore become the standard of care in those patients.2 RFA is deemed to be successful once complete remission of intestinal metaplasia (CRIM) is achieved both endoscopically and histologically. However, recurrence of both IM and dysplasia occurs, with an estimated annual incidence rate of 9.5% for any recurrence and 2% for dysplastic recurrences.3 4 For these reasons, guidelines recommend regular surveillance every 3 months for the first year following CRIM, then every 6 months for the second year and yearly thereafter.2
Data from a recent systematic review suggest that recurrence rates may be considerably reduced after the first year, raising the question whether surveillance intervals should be extended to every 2 or 3 years instead of yearly after CRIM.5 However, this reduction in recurrence rate was not noted in the subgroup of three studies defining CRIM more stringently as two negative endoscopies.6–8 The finding of columnar mucosa or IM on histology after one negative endoscopy may represent incompletely treated or missed prevalent disease rather than true recurrence and may therefore overestimate the recurrence rate.5 None of the latter studies evaluated the variation in the incidence of recurrence over time across all grades of dysplasia. Moreover, other studies have also been limited by either a small sample size (of patients with CRIM with a small number of dysplastic recurrences)6 9 10 or the inclusion of large numbers of patients treated with no dysplasia whose recurrence patterns may not be reflective of those with dysplasia.11 Hence, there is currently no conclusive evidence to demonstrate either that the recurrence rate following CRIM remains at the same level over time in order to justify currently recommended yearly surveillance or whether the rate decreases after some time which may justify widening of surveillance intervals.
Guidelines recommend that during surveillance after CRIM, random biopsies should be obtained from the gastro-oesophageal junction (GOJ) and the neosquamous epithelium at 1–2 cm intervals to cover the extent of the previous BE segment.2 Longitudinal data on the yield of this approach, particularly in the absence of endoscopically visible recurrence, are limited. Such data will be valuable in determining the cost-effectiveness of this practice and in informing future guidelines. Similarly, the location of recurrence in the tubular oesophagus needs to be better defined in order to justify surveillance biopsies over the entire length of the previous BE segment.
Given these knowledge gaps, we aimed to assess the timeline, location and patterns of recurrence following CRIM in a large multicentre and international cohort with the goal of informing future guidelines for endoscopic surveillance after CRIM. We also used a conservative definition of CRIM as two consecutive negative endoscopies with biopsies (from the oesophagus and GOJ) for the reasons detailed above.
Materials and methods
This was a cohort study of patients undergoing RFA for BE in five tertiary referral centres with expertise in the management of this condition. Three centres were located in the USA (Mayo Clinic Rochester, Mayo Clinic Arizona and Mayo Clinic Florida) and two in the UK (Nottingham and Cambridge University Hospitals). Data were obtained from prospectively maintained databases at each of the participating centres. We included patients who were 18 years or older with endoscopically (at least 1 cm of columnar mucosa in the tubular oesophagus) and histologically (presence of IM) confirmed BE with or without dysplasia who underwent RFA. Patients with advanced cancer stage (T2 or higher), are pregnant or with oesophageal varices were excluded from the study. RFA procedures were performed between November 2003 and July 2016. All patients included in this study achieved CRIM between March 2004 and May 2017 and had at least one follow-up endoscopy to check for recurrence. The study was conducted and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.12
Participants and interventions
Patients underwent RFA by expert endoscopists following endoscopic assessment using high-definition white light endoscopy and narrow band imaging with or without endoscopic resection of any visible lesions. Both circumferential and focal RFA were used to treat the BE segments as well as the GOJ. Energy settings followed manufacturer recommendations and RFA was performed every 3 months until CRIM was achieved. CRIM was defined as two consecutive endoscopies at least 3 months apart confirming the absence of IM on biopsies from both the GOJ (defined as the top of the gastric folds within 1 cm of the neosquamocolumnar junction) and tubular oesophagus. In addition to RFA, patients could receive argon plasma coagulation or multipolar coagulation as rescue (adjuvant) techniques for minimal residual BE islands.
Post-CRIM surveillance protocol
Once CRIM was achieved, subsequent surveillance was performed at 3, 6, 9 and 12 months, and every year thereafter. All patients underwent high-definition white light endoscopy and narrow band imaging with random biopsy specimens obtained from the GOJ and every 1–2 cm in four quadrants to cover the area of the previous BE segment as a minimum requirement. Samples from each level were labelled and stored in separate bottles. In addition, targeted biopsies were taken from any visible lesions (both columnar and/or neosquamous). Recurrence was defined as the histological presence of IM with or without dysplasia on biopsy specimens taken from either the tubular oesophagus or the GOJ or both after CRIM was achieved. The location, visibility and dysplasia status of all recurrences were documented.
Baseline histology was recorded for all patients on entry into the study and classified as non-dysplastic BE (NDBE), indefinite for dysplasia (ID), LGD, HGD and cancer. Biopsy specimens at each centre were examined by an expert GI pathologist. The worst grade of dysplasia detected on tissue sampling at baseline (pre-RFA) and at post-CRIM surveillance (post-RFA) was assigned to that patient.
The Kaplan-Meier method was used to estimate the cumulative incidence of recurrence after CRIM. The primary analysis addressed the extent to which the recurrence rate changed over time by fitting a Weibull survival model to the data and estimating the ‘shape’ parameter, γ. If γ=1 this would correspond to a recurrence rate that remains constant over time, whereas if γ<1 the rate would decrease and if γ>1 it would increase. For greater interpretability, a transformed version of this parameter, 2γ-1, referred to herein as the ‘relative recurrence rate’ (RRR), was used as the focus of the analysis as it corresponds to the recurrence rate (hazard) at any given follow-up time, t (eg, at 4 years), relative to the recurrence rate at a follow-up of half that time, t/2 (eg, at 4 years vs 2 years). An RRR of substantially less than 1 would indicate that the recurrence risk decreases over time and justify widening of surveillance intervals to longer follow-up times, while an RRR of 1 or more would not be consistent with widening of surveillance intervals. An estimate and profile likelihood CI were obtained for the RRR. Similar secondary analyses were performed for subgroups of patients and for other incidence outcomes. Cox proportional hazards models were used to assess the associations of baseline variables with recurrence. Multivariable Cox models were used to estimate adjusted effects of each variable on recurrence, with Firth estimation in cases with low recurrence totals. A sensitivity analysis was performed for excluding NDBE recurrences at the GOJ given concern in some studies that this may represent residual IM or IM of the cardia rather than true BE recurrence.13
Five hundred and ninety-four patients achieved CRIM as of 1 May 2017 and were included in the analysis (table 1). Figure 1 shows the patient flow chart. The mean (±SD) age was 67 (±10) years and 86% were male. The median (IQR) BE segment length was 4 (2–6) cm. Of the patients, 90% were treated for dysplasia or carcinoma. One hundred and fifty-one subjects developed recurrent BE (table 2) over a median (IQR) follow-up of 2.8 (1.4–4.4) years.
Recurrence incidence and timeline
The observed overall annual incidence rates of any recurrence, dysplastic recurrence and HGD/cancer recurrence for the entire cohort are shown in table 3. There was no evidence of a decrease in recurrence risk over time; the estimated RRR with a doubling of follow-up time was 1.02 (95% CI 0.93 to 1.12). The RRR of dysplasia was also close to 1 (0.99, 95% CI 0.85 to 1.20) and also for HGD/cancer (0.99, 95% CI 0.82 to 1.29) (figure 2). When NDBE recurrences at the GOJ were excluded, the recurrence rate of any BE, dysplastic BE and HGD/cancer continued to demonstrate no evidence in change over the follow-up duration (online supplementary figure 1). Similarly, when stratified by baseline histology pre-RFA, the RRR of any recurrence remained constant over the follow-up duration for the three outcomes of NDBE/ID, LGD and HGD/cancer (figure 3). The hazard rate of dysplastic recurrence also remained constant for all three outcomes (figure 4).
Supplementary file 1
BE recurred at the GOJ in 74.2% (n=112) of subjects and in the tubular oesophagus in 25.8% (n=39) (table 2). Of the 112 GOJ recurrences, 27 (24.1%) were dysplastic. All 85 non-dysplastic GOJ recurrences were non-visible. Of the 27 dysplastic GOJ recurrences, 59.3% (n=16) were visible endoscopically (cancer n=9, HGD n=5, LGD n=2) and 40.7% (n=11) were non-visible (HGD n=3, LGD n=8) and only detected on random biopsies of the GOJ. Overall, only 14% of the GOJ recurrences were visible.
Of the recurrences in the tubular oesophagus, 82.1% (n=32) were visible endoscopically and 84.4% of those were detected within 5 cm of the GOJ (figure 5A), and 17.9% (n=7) were non-visible and only detected on random biopsies of the neosquamous epithelium (figure 5B) (five were subsquamous and two had no mention of columnar mucosa on the endoscopy report, but histology showed IM with no squamous epithelium in the specimen bottle). Six out of those seven patients had NDBE at 2 cm (n=3), 7 cm (n=2) and 9 cm (n=1) from the GOJ. One patient had LGD at 4 cm from the GOJ. Therefore, the overall yield of random biopsy sampling for NDBE recurrence was only 1.0% (6/594) for any recurrence and 0.2% (1/594) for dysplastic recurrence.
Predictors of recurrence
Baseline HGD/cancer but not LGD predicted a higher risk of any recurrence (HR 1.95, 95% CI 1.07 to 3.56; p=0.029) (online supplementary table 1) and dysplastic recurrence (HR 4.81, 95% CI 1.21 to 19.18; p=0.026) (online supplementary table 2). There was weak evidence that the risk of dysplastic recurrence, but not overall recurrence, may be higher in patients for whom other adjuvant ablation techniques were used. There was no evidence of association of recurrence with any of the other eight baseline variables studied.
Supplementary file 2
In this large multicentre and international cohort study, cumulative recurrence rates of NDBE, any dysplasia and HGD/cancer following initial CRIM after RFA did not appear to plateau over the first 5–6 years of follow-up, suggesting that continued yearly surveillance remains important and arguing against extending surveillance intervals at present, particularly in those with HGD/cancer at baseline. The majority of recurrences (74.2%) developed at the GOJ and approximately a quarter of those were dysplastic, out of which a significant proportion (40.7%) were non-visible endoscopically. Most (84.4%) visible recurrences in the tubular oesophagus were located within 5 cm of the GOJ. Finally, the yield of tubular oesophageal biopsies in the absence of visible recurrence was very low for NDBE (1.0%) and dysplastic BE (0.2%), which suggests that the requirement for random biopsies of the neosquamous epithelium in the absence of visible recurrence may need to be re-evaluated.
A recent modelling study based on data from the US and UK RFA registries suggested surveillance endoscopies at 1 and 3 years after CRIM for patients with baseline LGD and endoscopies at 3 months, 6 months and 1 year, and then annually thereafter for those with baseline HGD or cancer.11 The surveillance intervals were estimated only to a limit of 5 years to avoid extrapolation beyond available data. Therefore, there is a lack of clarity from current literature with regard to both the need for and the yield of surveillance beyond this time. Moreover, the latter model’s estimates were based on dysplastic recurrences only and did not account for the non-dysplastic (NDBE) ones. NDBE recurrences following CRIM require therapy as those could represent an incompletely treated or missed prevalent disease and may therefore still have neoplastic potential if left untreated.5 When NDBE recurrences at the GOJ were excluded (given concern that this may represent IM of the cardia rather than true BE recurrence13), the recurrence hazard rate across all grades of dysplasia remained constant (table 3 and online supplementary figure 1). Hence, our data suggest that continued surveillance beyond 5 years remains necessary. The recurrence rate in patients with LGD at baseline also remained constant over time, suggesting that long-term surveillance is warranted in that group as well.
The number of endoscopies required to define CRIM remains a subject of debate. IM is known to be patchy and may be missed on random biopsies.14 Therefore, two endoscopies with biopsies may be required to confidently rule out the presence of IM.14 To our knowledge, there are no data comparing two versus three or more negative endoscopies to define, and this may be an area for further research. None of the three studies that defined CRIM on two negative endoscopies6–8 evaluated the variation in the incidence of recurrence over time across all grades of dysplasia. In one single-centre study, 20% of patients had NDBE at baseline and no recurrences were reported after 3 years, implying little benefit from surveillance beyond 3 years post-CRIM.6 This is in contrast to data from our study in which we estimated the recurrence rate to remain constant, or at least not decrease by more than 7% for every doubling of follow-up time throughout the study period that extended beyond 3 years.
Previously, two smaller, single-centre studies reported that the majority of dysplastic recurrences developed in the gastric cardia and the majority of those were non-visible.10 15 Eighty per cent of recurrences in the tubular oesophagus were visible endoscopically,10 and random biopsies >1 cm proximal to the GOJ had no yield for any recurrence.15 Data from the current larger multicentre cohort show that 82% of tubular oesophageal recurrences were visible, and the yield of random biopsies >1 cm proximal to the GOJ was extremely low (1.2% yield for any recurrence and 0.2% for dysplastic recurrence). The true rate of non-visible, subsquamous recurrence remains hard to measure, but is likely to be rare based on current data.16 Volumetric laser endomicroscopy has been used to image post-RFA subsquamous glandular structures, but correlation with buried BE glands was poor in one study.17 More cost-effective and standardised imaging and sampling techniques are required in order to evaluate this outcome in a more systematic and precise manner. Moreover, data on the natural history of these recurrences are required.
Study strengths and limitations
This study has several strengths. We evaluated a large sample of patients over a long follow-up duration, which is an important factor to achieve more precise estimates of the recurrence rates. Data were collected from multiple centres located both in the UK and USA, which strengthens the validity and generalisability of our results. Furthermore, the vast majority of our patients (90%) received RFA for dysplastic BE, which is representative of current clinical practice guidelines in contrast to several other studies of US populations where a large proportion of patients (up to 46%)11 have NDBE and therefore do not require RFA based on current evidence.2 Databases were prospectively maintained in all participating centres in order to minimise selection and recall biases. While centralised pathology was not used, all centres had dedicated expert GI pathologists reading all BE histology.
The study also has potential limitations. Although follow-up times were relatively long in this study, it is still possible that recurrence risk may decrease at longer follow-up times. In addition, as always, a larger sample size would have had greater power to detect a reduction in recurrence rate over time. Based on these data, however, we estimate that a future study, identical in design, would have 80% or greater power at the 5% significance level to detect an RRR of 0.887 or lower. In other words, we estimate that this study design is estimated to provide sufficient power to detect a reduction in recurrence risk as small as 11.3% over a doubling in follow-up time. Biopsy sampling techniques and forceps size used were not standardised across centres and may therefore result in sampling error and case ascertainment bias. We attempted to minimise the latter by implementing a more conservative definition of CRIM with two negative endoscopies and biopsies for IM from both the GOJ and tubular oesophagus. The multicentre nature of the study with different operators at centres of expertise makes the study susceptible to some variation in practices, but also helps to make our data more representative of real-world practice. One caveat to the latter is that our study sample comes from tertiary referral centres, which may not be representative of results in community practice. However, the overwhelming number of these procedures is currently performed in settings similar to ours based on current society guidelines.2
Conclusions and implications for clinical practice
This study suggests that the recurrence hazard rate of NDBE, dysplastic BE and HGD/cancer remained constant over time during surveillance in patients who achieved CRIM after RFA when strict criteria for the definition of CRIM are applied. This suggests that diligent long-term (at least yearly) endoscopic surveillance remains important in these patients. The majority of all recurrences developed at the GOJ and a significant proportion of dysplastic recurrences were non-visible endoscopically. This re-enforces the need for careful imaging and sampling of the GOJ despite the absence of any visible lesions. On the other hand, the majority of recurrences in the tubular oesophagus are visible endoscopically, and the yield of random biopsy sampling in the absence of visible lesions was very low in expert centres. These findings may need to be replicated in non-expert centres before further conclusions can be made with regard to the cost-effectiveness of this practice.
Lyndsay Busby for assistance in formatting the manuscript.
Contributors SSS: contributed to the design and coordination of the study, acquisition of data, analysis, interpretation of data, and drafted the manuscript. AR, AK, DS, JS, JO-s, WKT, MGH, MLJ, RL: contributed to the acquisition of data and critical revision of the manuscript for important intellectual content. RAD: contributed to the analysis, interpretation of data and critical revision of the manuscript for important intellectual content. KKW, CLL, DAK: contributed to the design of the study, interpretation of data and critical revision of the manuscript for important intellectual content. KR, MdP, HW, FR, DF: contributed to study supervision, interpretation of data and critical revision of the manuscript for important intellectual content. PGI: contributed to the conception, design and supervision of the study, acquisition of data, analysis, interpretation of data, and critical revision of the manuscript for important intellectual content.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests PGI: research funding from Exact Sciences, C2 Therapeutics and Medtronic; consulting: C2 Therapeutics, CSA Medical and Symple Surgical.
Patient consent Not required.
Ethics approval The study was approved by the institutional review boards of the respective centres.
Provenance and peer review Not commissioned; externally 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.