Neoplasia Detection Rate (NDR) in Barrett’s Esophagus and Its Impact on Missed Dysplasia: Results from a Large Population-Based Database

Abstract

Background & Aims:

It is a challenge to detect dysplasia in Barrett’s esophagus (BE) and esophageal adenocarcinomas (EACs) are missed in 25%–33% of cases. The neoplasia detection rate (NDR), defined as the rate of high-grade dysplasia (HGD) or EAC detection during initial surveillance endoscopy, has been proposed as a quality metric for endoscopic evaluation of patients with BE. However, current estimates are from referral center cohorts, which might overestimate NDR. Effects on rates of missed dysplasia are also unknown. We analyzed data from a large cohort of patients with BE to estimate the NDR and factors associated with it, and assess the effects of the NDR on the rate of missed dysplasia.

Methods:

We analyzed data from 1066 patients the Rochester Epidemiology Project-linked medical record system, a population-based cohort of patients with BE (confirmed by review of the endoscopic and histologic reports) from 11 southeastern Minnesota counties from 1991 through 2019. Biopsies reported to contain dysplasia were confirmed by expert gastrointestinal pathologists. The NDR was calculated as the rate of HGD or EAC detected by histologic analyses of biopsies collected during the first surveillance endoscopy. Patients without HGD or EAC at their initial endoscopy (n=391) underwent repeat endoscopy within 12 months; HGD or EAC detected at the repeat endoscopy were considered to be missed on index endoscopy. Factors associated with NDR and missed dysplasia were identified using univariate and multivariate logistic regression models.

Results:

The NDR was 4.9% (95% CI, 3.8–6.4); 3.1% of patients had HGD, 1.8% had EAC, and 10.6% had low-grade dysplasia. Factors associated with higher rates of detection of neoplasia included older age, male sex, smoking, increasing length of BE, and surveillance endoscopies by gastroenterologists. This NDR was associated with a substantially lower rate of missed dysplasia (13%).

Conclusions:

In an analysis of 1066 patients with BE in a population-based cohort, we found a lower NDR and lower rate of missed dysplasia than previously reported. NDR may have value as a quality metric in BE surveillance if validated in other cohorts.

INTRODUCTION

The incidence of esophageal adenocarcinoma (EAC) has risen rapidly over the last 3 decades [1, 2]. Overall outcomes for EAC remain dismal overall, with a less than 20% 5-year survival. [2] Barrett’s esophagus (BE) is the only known precursor to EAC, and dysplastic BE is associated with an increased risk of EAC [3]. Endoscopic eradication therapy can prevent progression to and treat early stage EAC effectively [4, 5]. Hence, societal guidelines recommend screening followed by endoscopic surveillance for BE [6, 7, 8].

However, almost 90% of patients with EAC continue to present outside of a BE surveillance program [9, 10]. While this may represent missed opportunities for screening, we and others have reported that up to 25-33% of patients with HGD/EAC are missed on initial surveillance endoscopy [11, 12]. Dysplasia detection in BE is challenging due to its patchy distribution and often subtle appearance. Lack of compliance with recommended biopsy guidelines is also well documented [13].

Quality benchmarks have been proposed in BE endoscopy, incorporating at least 3 pivotal facets: (i) meticulous examination of the BE segment; (ii) appropriate and effective sampling; and (iii) recommending appropriate surveillance intervals [14]. Analogous to current practice guidelines for colonoscopy, in which the adenoma detection rate is a validated quality metric [15], neoplasia detection rate (NDR) has been proposed as a quality metric in BE surveillance. NDR is defined as the rate of HGD or EAC detected histologically on initial surveillance endoscopy. It has been proposed as a metric for defining adequate surveillance endoscopy in BE, with the potential to reduce rates of missed dysplasia/EAC and improve effectiveness of surveillance endoscopy. [16]. However data linking this metric to outcomes is lacking.

In a recently published systematic review and meta-analysis (SRM), NDR at the first surveillance endoscopy was estimated at 7% (95% CI 4%-10%) [17]. However, this is comprised of data from largely referral center cohorts with endoscopy performed by experienced academic gastroenterologists, which may overestimate NDR. Individual patient level data such as patient, provider (gastroenterologist versus others) and endoscopic data (adequacy of biopsies) were not available. Hence, neither the impact of NDR on subsequent rates of missed dysplasia nor predictors of NDR could be assessed, to study the impact of NDR on clinically relevant outcome data.

In view of these knowledge gaps, we aimed to 1. Estimate NDR at initial surveillance endoscopy in a large population-based cohort to reduce the impact of referral bias 2. Identify predictors of NDR using individual patient, provider and endoscopic data and 3. Assess the impact of NDR on the rate of missed dysplasia and EAC.

METHODS

The Institutional Review Boards of Mayo Clinic and Olmsted Medical Center (OMC) approved this study.

The Rochester Epidemiology project

The Rochester Epidemiology Project (REP), formed in 1966, is an initiative of the Mayo Clinic, Olmsted Medical Center and the Rochester Family Medicine Clinic in Olmsted County, Minnesota, containing a linked database of medical information for all residents of Olmsted County, Minnesota. From 2010, this medical record linkage system was serially expanded to counties in southern Minnesota and western Wisconsin. The REP captures virtually all residents of these counties, regardless of age, sex, ethnicity, disease,, socio-economic or insurance status. The availability of a centralized index to access patient medical records makes this database well suited for conducting population-based observational studies [18, 19, 20, 21, 22].

Search strategy and data abstraction

For this study, we included patients from the 11 counties (Olmsted, Dodge, Mower, Goodhue, Fillmore, Wabasha, Winona, Houston, Freeborn, Steele and Rice) of southeastern Minnesota. A high percentage of all medical records of residents in these counties are captured by the REP: all medical data is available for 400,065 patients in this area with a population of 473,064 (84.5%) according to the 2017 US Census (see supplementary figure 1) [19, 23, 24]. BE cases were identified from 1991 to 2018 by performing an electronic search of the REP using the International Classification of Disease, Ninth Edition and Tenth Edition codes for BE (ICD-9-CM 530.85 and ICD-10-CM-Code K22.7). We included records from 1991 onwards as all endoscopy and pathology reports from that year onwards are electronically accessible. BE diagnosis was confirmed by a review of endoscopic and histologic reports. BE was defined as endoscopic evidence of at least 1 cm of visible columnar mucosa in the esophagus and histology demonstrating intestinal metaplasia. All dysplastic histology was confirmed by expert gastrointestinal pathologists at Mayo Clinic Rochester. Patients with endoscopic BE length < 1cm, those with no index endoscopy report available and those presenting with alarm symptoms revealing obstructing esophageal masses were excluded from this analysis. We included patients with dysphagia and GI bleed, who had no endoscopic findings of obstructing esophageal masses or malignant ulcers and carefully reviewed their notes to exclude any esophageal cause of these symptoms. NDR was calculated using histology from the first endoscopy where surveillance biopsies were taken as defined in the endoscopic report from the medical records (this was a combination of the first endoscopy at which BE was diagnosed on pathology or the second endoscopy at which surveillance biopsies were taken after the initial endoscopic evidence of BE).

Electronic medical records of all cases were reviewed for demographics, symptoms at presentation, endoscopic findings and histology. Data were also extracted regarding the presence of visible lesions (nodularity, ulcers and non-obstructive nodules) and the number of biopsy bottles taken. The ratio of the BE segment length to the number of biopsy bottles was calculated to assess adherence to the Seattle protocol. Those where the ratio was ≤ 2 were deemed to be compliant with the Seattle protocol. Endoscopies were performed by gastroenterologists, surgeons, internal medicine and family medicine physicians. Cases meeting the inclusion criteria were divided into two groups based on initial histologic diagnosis: Group 1 (NDBE, indefinite dysplasia, and LGD) and Group 2 (HGD and EAC). Group 1 cases were additionally followed to determine whether HGD/EAC was missed on index endoscopy (defined as the detection of HGD or EAC on endoscopy done within 12 months of initial diagnosis).

Statistical Analysis

Data were summarized as mean (± SD) for quantitative variables and proportions (%) for discrete variables. The neoplasia detection rate (with 95% confidence intervals), was calculated as the rate of HGD or EAC detected histologically on initial surveillance endoscopy. For categorical univariate outcomes of interest, proportions were compared utilizing chi-square significance testing or the Fisher’s exact test (if the “expected” value was less than 5). For continuous univariate outcomes of interest, the t-test was used to compare means. Univariate analysis was performed by logistic regression to assess association of variables of interest with NDR. The primary outcome was detection of HGD or EAC in surveillance histology (dichotomized to yes : HGD/EAC versus no : NDBE/Indefinite dysplasia/LGD).We then created a multivariate logistic regression model including all variables which reached statistical significance (defined as p < 0.05) on univariate analysis, in order to analyze the effect of multiple factors on NDR outcomes. Variables of interest in the univariate models included age (in 1 year increment), gender, smoking status, time period (in 5 year increments) , index EGD indication, presence of hiatal hernia, BE length (in 1 cm increments), presence of visible lesion (ulcers, nodules, or masses distinct from “flat” BE) and BE length/bottle ratio. All analyses were done using JMP^® version 15.0 (SAS Institute, Cary, NC). A p-value of less than .05 was considered statistically significant.”

RESULTS

We identified 1118 patients in the BE population cohort. Of these, 1066 patients met inclusion criteria (Figure 1). Baseline characteristics are provided in Table 1. The cohort was largely male (71.6%) with a mean (SD) age of 63 (13.3) years. The mean (SD) BE length was 3.6 (3.0) cm and a hiatal hernia was present in almost 60% of patients. A substantial proportion of patients had reflux (36.8%) as an indication for endoscopy followed by dyspepsia (16.1%), dysphagia (15.2%) and iron deficiency anemia (14.2%). Seventy-seven percent (824) of surveillance endoscopies were performed by gastroenterologists and the remaining by non-gastroenterologists (including family practitioners, surgeons and internal medicine physicians): due to the community nature of practices some endoscopy continues to be performed by non-gastroenterologists. Approximately 60% of patients were adequately biopsied per Seattle protocol (BE segment length/no. of bottles ≤ 2).

Figure 1:

Flow diagram of BE cases in the identified and their follow-up.

Table 1:

Baseline characteristics of Barrett’s esophagus patients.

Age
Mean (SD)	63.2yrs (13.3)
Sex
Male (N,%)	763, 71.6%
Female (N, %)	303, 28.4%
Smoking Status
Current/Past	53.4%
None	37%
Unknown	9.5%
Index EGD Indication
Reflux	36.8%
Dyspepsia	16.1%
Dysphagia*	15.2%
Iron Deficiency Anemia	14.2%
Abdominal Pain	9.7%
GI Bleed+	9.2%
Chest Pain	6.6%
Other	12%
Hiatal Hernia
Present	59.3%
Mean Length (range)	3.6cm (3.5-3.8)
Barrett’s Length
Mean (SD)	3.7 cm (3)
Median (range)	2 cm (2-5)
Histology at Index EGD
NDBE (N, %)	900, 84.4%
LGD (N, %)	113, 10.6%
HGD (N, %)	33, 3.1%
EAC (N, %)	20, 1.8%

^*:No evidence of malignant obstructive luminal pathology

^+:No evidence of malignant obstructive luminal pathology leading to GI bleeding.

Neoplasia Detection Rate

The NDR was 4.9% (95% CI 3.8%-6.4%), of which 3.1% (n=33, 95% CI 2.2-4.3) had HGD and 1.8% (n=20, 95% CI 1.2-2.8) had EAC. Of the EAC cases, 60.0% (n=12) were T1 lesions and the remaining 40.0% (n=8) were T2 lesions. These T2 lesions were non-obstructive in endoscopic appearance, without preceding history of dysphagia on review of medical records. Five (25%) EAC cases had evidence of metastatic lymphadenopathy on endoscopic ultrasound or at surgery. These 20 EAC cases were managed with surgery (n=9), endoscopic resection (n=7), and chemoradiation (n=4).

LGD was found in 10.6% of patients (n=113, 95%CI 8.9-12.6) and NDBE in the remaining 84.4% (n=900, 95% CI 82.1-86.4) of patients. NDR appeared to increase significantly from 1991-2019 on univariate analysis (particularly after 2000), but this was not observed on multivariate analysis (Table 2, Figure 2). This was despite the introduction of high definition monitors and high resolution endoscopes in subsequent years.

Table 2:

Predictors of detection of high grade dysplasia or esophageal adenocarcinoma on initial surveillance endoscopy

	Univariate OR (95% CI)	P-Value	Multivariate OR (95% CI)	P-Value
Gender
Female	REF		REF
Male	4 (1.6-10.1)	0.0011	4.8 (1.58-14.74)	0.0056
Age
1-Year Increments	1.05 (1.03-1.08)	<.0001	1.05 (1.02-1.09)	0.0001
Smoking
None	REF		REF
Current	2.6 (1-6.4)	0.0369	2.68 (0.89-8.02)	0.0780
Past	2.26 (1.1-4.6)	0.0247	1.97 (0.72-5.34)	0.1843
Provider
Non-Gastroenterologists	REF		REF
Gastroenterologists	3.6 (1.3-10)	0.0154	2.49 (0.84-7.31)	0.0971
Barrett’s Length
1 CM Increments	1.18 (1.09-1.26)	<.0001	1.12 (1.02-1.21)	0.0116
Time Period
1991-2000	REF		REF
2001-2010	8.16 (1.09-60.8)	0.0405	8.57 (0.91-66.80)	0.0601
2011-2019	9.48 (1.28-70.3)	0.0277	7.89 (0.75-62.676)	0.0926
Visible Abnormality on Endoscopy
Absent	REF		REF
Present	15.61 (8.11-30.05)	<.0001	8.45 (3.69-19.34)	<.0001
Reflux
Absent	REF
Present	1.03 (0.59-1.8)	0.89
Dysphagia
Absent	REF
Present	1.14 (0.54-2.3)	0.72
Iron Deficiency Anemia
Absent	REF
Present	3.3 (1.8-6.1)	<.0001
Dyspepsia
Absent	REF
Present	0.19 (0.04-0.81)	0.0247
Hiatal Hernia
Absent	REF
Present	1.19 (0.66-2.1)	0.54
BE Length/Bottle ratio*
Inadequate Biopsy	REF
Adequate Biopsy	0.54 (0.42-1.3)	0.1141

^*Patients who had targeted biopsy were excluded from the analysis.

Figure 2:

Temporal trends of LGD, HGD and EAC detected in a Southeastern Minnesota Barrett’s esophagus population based cohort over three decades.

We analyzed dysplasia diagnoses made from targeted biopsies separately.Targeted biopsies were taken in 54 patients from a visible lesion. Of these, 26 had NDBE (2.8% of all NDBE diagnoses), 9 had LGD (7.96% of all LGD diagnoses), 9 had HGD (27.3% of all HGD diagnoses, and 10 had EAC (50.0% of all EAC diagnoses). Gastroenterologists reported significantly more visible lesions than non-gastroenterologists (OR= 3.7, p=0.0120). Twenty one patients subsequently underwent endoscopic resection.

Predictors of neoplasia detection on multivariate analysis included older age, male gender, current smoking, increasing BE length and presence of visible abnormality on endoscopy (Table 2). Gastroenterologists had a substantially higher rate of neoplasia detection than non-gastroenterologists on univariate analysis (5.8% vs.1.7%, p=0.0098).

Missed Dysplasia

Three hundred and ninety one BE patients without HGD/EAC on initial endoscopy had a repeat endoscopy within 12 months. HGD/EAC was detected (i.e. missed on index endoscopy) in 8 of these patients. Of these 8 patients, 6 had LGD and 2 had NDBE on index endoscopy. Details of location and grade of missed dysplasia are provided in supplementary table 1. Since 33 HGD and 20 EAC were diagnosed at initial endoscopy (see above), 8/(53+8=61) or 13% of all HGD/EAC detected in this cohort within 12 months of diagnosis, were missed. There was no statistically significant difference between the segment lengths (4.7cm vs. 3.7cm, p=0.4), adherence to the Seattle protocol (62% vs. 58.7%, p=0.8), visible lesions (OR 0.6 [0.11-3.25], p=0.55), age, smoking history or proceduralist specialty of those with and without missed dysplasia.

DISCUSSION

In this large population-based BE cohort, we estimated a NDR of 4.9%, which is lower than previously reported studies [16]. We also identified predictors of NDR and correlated this NDR with a substantially lower rate of missed dysplasia (a clinically relevant outcome). Given that prior reports consisted largely of tertiary care center cohorts [17], our findings may reflect the absence of referral bias and be more generalizable.

NDR has been proposed as a quality outcome measure and is likely associated with factors such as careful examination of the esophagus and obtaining adequate number of biopsies. Though it would seem that the introduction of high resolution endoscopes, high definition monitors and advanced imaging modalities may impact NDR, we found no difference in the NDR over the almost three decade span of this study. This may suggest that in a low dysplasia prevalence setting, basic techniques such as careful white light inspection of the BE mucosa along with targeted and Seattle protocol biopsies may be more important.

Given the impending paradigm shift from a fee-for-service to value-based care, the need for high quality procedures is paramount, and the use of objective measurements may help determine the adequacy of our procedures and establishment of quality benchmarks. Currently, there are no widely accepted criteria for what constitutes a high-quality endoscopy in BE patients. Factors such as adherence to the Seattle biopsy protocol, proper documentation of Prague criteria, time spent inspecting BE mucosa, and utilization of narrow-band imaging are not included in current reimbursement schemes [25]. NDR has the potential to be a quality parameter for at least some of these tasks [26]. However there are no data on its impact on clinically relevant outcomes. Given the relative rarity of incident EAC, demonstration of a correlation between decreased EAC mortality and NDR would be challenging. Hence in this study we focused on missed HGD/EAC rates.

In a recent SRM by Visrodia et al, it was determined that 24% of EAC were detected within 1 year of the initial BE diagnosis and hence likely missed at index endoscopy [11]. We report a substantially lower rate of missed HGD/EAC in this cohort. As the majority of studies included in the previously published meta-analysis were from referral centers, endoscopy per protocol at specialized centers, by practitioners who may be more experienced with the management of BE patients, may yield higher rates of dysplasia or cancer detection. The implications of this finding should not sway community practitioners from managing BE patients, but rather reinforces the need for objective quality criteria that all endoscopists could adopt.

In our cohort, endoscopy performed by gastroenterologists was associated with an almost four-fold higher reporting of visible lesions and detection of HGD/EAC. This finding is novel and may be due to additional training in endoscopy, lesion recognition and familiarity with surveillance guidelines in gastroenterologists. Similar differences in outcomes (interval colon cancer and cecal intubation rates) have been reported in the colon cancer literature as well [27, 28, 29]. This finding, if replicated in other cohorts, may support recommendations for the performance of surveillance by endoscopists trained in gastrointestinal endoscopy and well versed in surveillance guidelines.

Though adherence to the Seattle protocol was found in only 57% of patients its conitnuing relevance is highlighted by the fact that almost 50% of prevalent EAC and 75% of prevalent HGD were still diagnosed on protocolized biopsies. However, the importance of careful inspection and taking targeted biopsies from visible lesions cannot be overstated given the high yield of these biopsies. Indeed at a community level it appears that both methods continue to contribute substantially to dysplasia detection.

We also identified older age, smoking and BE segment length as predictors of neoplasia detection. While these are well-known risk factors for EAC, providers should also consider these risk factors while performing surveillance.

In this study we included only patients who had BE >1 cm on endoscopy and intestinal metaplasia consistent with current diagnostic criteria (to avoid misclassifying intestinal metaplasia of the GE junction as BE). While earlier definitions of BE did not have a length requirement, we believe it is unlikely that there would be a significant number of missed BE in the earlier time periods as guidelines would have recommended biopsies of any length of BE [30]. It has since become clear that IM of the GE junction has outcomes that are distinct from those with bonafide BE. Therefore it is unlikely that exclusion of these patients with columnar segments < 1cm would affect the overall results of our study [22].

The strengths of this study lie in the large, population based cohort with comprehensive linkage of medical records allowing for the abstraction of data over almost 30 years. Furthermore, referral bias is unlikely to affect our results as all included patients reside within the catchment area of the database (as defined by the predetermined inclusion criteria). We acknowledge some limitations of this study. While there was no central pathology read, cases with dysplasia were confirmed by expert GI pathologists. This reduces the possibility of overestimating the NDR. However, we do acknowledge the known interobserver variation among pathologists in diagnosing LGD. The retrospective nature of the study precluded adoption of a consistent biopsy protocol over the entire time period of the study (the Seattle protocol was published in 2000) [31]. However, our results did not change with a subset analysis of procedures performed since 2000 [OR = 0.68 (0.38-1.21) p=0.19].

The long duration of the study also makes results susceptible to changing guidelines, endoscopes and imaging modalities. However the NDR remained fairly stable over time, though a type 2 error cannot be completely excluded. Similar findings were reported in a prior study wherein NDR did not substantially change after the introduction of NBI and chromoendoscopy with surveillance performed by physicians trained in BE surveillance [32, 33]. Prior REP studies have shown low rates of migration out of the catchment area for REP. However this could have contributed in a small measure to low rates of missed dysplasia. Lastly we could not assess the impact of BE inspection time and number of biopsy pieces on NDR given the lack of availability of this data.

In conclusion, we analyzed data from over three decades in a large, population-based cohort and report a lower NDR than previously published, while highlighting that a high NDR, found particularly in gastroenterologist performed surveillance, results in lower missed dysplasia rates . The lower overall NDR may be due to the lack of referral bias in our cohort. Given the lack of objective quality measures in the endoscopic surveillance of BE, using NDR as a quality metric coupled with outcome measures such as missed dysplasia rates could improve adherence to established biopsy protocols and improve the quality of care to patients. Ultimately, this can be an opportunity to develop a high-value, evidence-based quality metric in BE surveillance.

Supplementary Material

s2

Supplementary figure 1:

Geographical map of the 11-county region of the Expanded-REP (E-REP) showing the percentage capture for each county (black numbers or white numbers). The color shading of the counties is proportional to the percentage capture of the E-REP as compared with the US Census estimates.

What You Need to Know.

Background: It is a challenge to detect dysplasia in Barrett’s esophagus (BE) and esophageal adenocarcinomas (EACs) are missed up to 33% of cases. The neoplasia detection rate (NDR), defined as the rate of high-grade dysplasia (HGD) or EAC detection during initial surveillance endoscopy, has been proposed as a quality metric for endoscopic evaluation of patients with BE.

Findings: In an analysis of 1066 patients with BE in a population-based cohort, we found a lower NDR (4.9%) and lower rate of missed dysplasia (33%) than previously reported.

Implications for patient care: NDR may have value as a quality metric in BE surveillance if validated in other cohorts.

Abbreviations:

BE

Barrett’s Esophagus

EAC

esophageal adenocarcinoma

HGD

high grade dysplasia

LGD

low grade dysplasia

NBI

narrow band imaging

NDR

Neoplasia detection rate

REP

Rochester Epidemiology Report

SRM

systematic review and meta-analysis