Barrett's esophagus (BE) is a precursor to esophageal adenocarcinoma (EAC). Patients with BE are surveyed by endoscopy with biopsies to detect dysplasia or EAC at an early stage when endoscopic eradication therapy (EET) can effectively prevent disease progression (1,2). The effectiveness of surveillance is limited by the random nature of biopsy sampling and observer variation in histologic evaluation (3,4). A diagnosis of low-grade dysplasia (LGD) confirmed by an expert gastrointestinal (GI) pathologist is a significant predictor of progression to high-grade dysplasia (HGD) and EAC, and current guidelines recommend increased surveillance or intervention with EET for patients with confirmed LGD (2,5,6). Most community-based LGD cases are downstaged to nondysplastic (ND) BE when reviewed by expert GI pathologists, whereas patients with confirmed LGD have up to 10% annual risk of progression to HGD/EAC (7–9). However, expert review is poorly defined, prone to interobserver variation, is not widely available, and hence, the reported progression rates for confirmed LGD vary (3,4,10–12). There is a need for objective tests to identify BE patients with a community-based diagnosis of LGD who are at increased risk for progression. These patients could benefit from early intervention with EET. Furthermore, a test that identifies patients at low risk for progression but have been overdiagnosed as LGD could prevent unnecessary interventions and over surveillance.
A risk prediction assay (TissueCypher Barrett's Esophagus Assay) has previously been validated in multi-institutional studies to risk stratify patients with BE (13–15). The assay uses a multiplexed fluorescence imaging platform that automatically extracts quantitative data on multiple tissue biomarkers (16,17) and uses a multivariable classifier that integrates quantitative image analysis data to provide a risk score (0–10) and risk classes of low, intermediate, or high risk for progression to HGD/EAC within 5 years.
The aims of this study were to investigate whether this risk prediction assay can risk stratify BE patients with community-based diagnosis of LGD and to compare the predictive performance of the assay to 3 expert GI pathologists providing histologic diagnoses.
Population and setting
The screening cohort for the SURveillance vs RadioFrequency ablation (SURF) trial consists of Barrett's patients with a community-based diagnosis of LGD who were prospectively identified in 9 Barrett's treatment centers in Europe and their referring institutions (18).
Study cohort and design
All patients with a referral diagnosis of community-based LGD undergoing screening for the SURF trial in whom the natural history of community-based LGD could be followed were included in the current study (patients randomized to the SURF surveillance arm and patients downgraded to NDBE by a single expert pathologist leading to exclusion from the SURF trial).
Follow-up was completed for each patient up to December 2018 by contacting the attending hospital. Endoscopic follow-up was performed according to the SURF study protocol in the patients from the SURF surveillance arm and guided by the histological diagnosis in the patients downstaged to NDBE and excluded from the SURF trial. All endoscopies were performed according to the Seattle protocol (4-quadrant biopsies every 1–2 cm BE). For cases progressing to HGD/EAC, the biopsy specimens were retrieved and confirmed by the 3 pathologists. Patients were excluded if informed consent was declined, if natural outcome was unavailable because they underwent radiofrequency ablation, formalin-fixed paraffin-embedded tissue blocks were unavailable, or less than 3 years follow-up was available after the first reviewed diagnosis of LGD.
Data elements collected were age, sex, segment length, original diagnosis, and review diagnosis of the 3 expert reviewers of every biopsy specimen block, presence/absence of hiatal hernia, worst histologic diagnosis during follow-up, and survival times after diagnoses.
The risk prediction assay was run in a blinded manner on all available specimens from the first LGD endoscopy with expert histology review.
Test results were reported to a consultant statistician who performed the statistical analyses following a prespecified plan. The study was exempted from review by the institutional ethics committee of the Academic Medical Center, Amsterdam. The institutional biobank review committee of the Academic Medical Center approved the ReBus biobank.
Expert histologic review
The 3 expert pathologists reviewing the samples were considered as such by their international peers. They have a special interest in the field of BE for over 10 years with a minimum case load of 5–10 mainly dysplastic cases per week and have coauthored more than 10 peer-reviewed publications in this field. The 3 pathologists form the basis for a national review panel for dysplastic Barrett's cases in the Netherlands, and their expertise was used to establish benchmark criteria for participating pathologists. The pathologists independently reviewed hematoxylin and eosin (H&E) slides from all available biopsy levels. P53 immunohistochemical (IHC) slides were available for use adjunctively in rendering their diagnosis. Diagnoses were recorded separately for each biopsy per the Vienna Classification (NDBE, indefinite for dysplasia [IND], LGD, HGD, or EAC) (19).
Risk prediction testing
The risk prediction assay (TissueCypher Barrett's Esophagus Assay) was run on sections from each specimen at Cernostics CLIA-certified laboratory (Pittsburgh, PA) (see Text, Supplementary Digital Content 1, https://links.lww.com/AJG/B757). All assay parameters were prespecified and locked, as previously defined (13,14).
Kaplan-Meier (KM) curves were used to graphically represent the risk of progression to HGD/EAC (as a combined endpoint and as single endpoints) of the risk classes determined by the assay and pathologic diagnoses. Hazard ratios with 95% confidence intervals (CIs) were calculated from Cox proportional hazards regression, and the logrank test was used to evaluate the equality of progression curves of the risk classes and diagnostic groups from the KM analysis. Performance metrics for the risk prediction assay were: sensitivity = proportion of progressors scored intermediate/high risk, specificity = proportion of nonprogressors scored low risk, negative predictive value (NPV) = proportion of patients scored low risk who did not progress and positive predictive value (PPV) = proportion of patients scored intermediate/high risk who progressed to HGD/EAC within 5 or 10 years. The same metrics were calculated for the pathologists' diagnoses: sensitivity = proportion of progressors with IND/LGD, specificity = proportion of nonprogressors with ND, NPV = proportion of patients with ND who did not progress, and PPV = proportion of patients scored IND/LGD who progressed. Area under receiver operating characteristic (AUROC) curves were calculated to compare the predictive performance of the TissueCypher risk score at 5 years vs a quantitative image analysis measurement of p53 immunofluorescence nuclear sum intensity.
In a post hoc analysis, biopsies from additional time points were evaluated in the subset of progressors who scored low risk and matched nonprogressors. The risk prediction assay results were compared between the multiple time points.
All authors had access to the study data and reviewed and approved the final manuscript.
Baseline characteristics of the study population are summarized in Table 1. One hundred fifty-five patients met the inclusion criteria, 34 progressed and 121 did not progress to HGD/EAC during follow-up (Figure 1). Patients had a mean age of 61 ± 10 years and a median maximal Barrett's length of 4 cm (3–6). There was no significant difference regarding age, sex, and maximal Barrett's length between progressors and nonprogressors. Progression to the combined endpoint of HGD/EAC occurred a median of 2.4 years (1.0–5.1) after the baseline sample, whereas median times to progression for HGD (22 progressors) or EAC (12 progressors) as single endpoints were 1.8 years (interquartile range 0.6–3.4) and 4.6 years (2.1–7.0), respectively. Nonprogressors had a median HGD/EAC-free follow-up of 7.9 years (5.9–10.3).
Risk stratification performance of TissueCypher and 3 expert pathologists
The risk prediction assay was evaluated in biopsies from the baseline endoscopy for each patient. Where biopsies from multiple esophageal levels were available, each was tested separately and the risk score was based on the highest scoring level. The rate of progression was similar in the intermediate- and high-risk groups (see Figure, Supplementary Digital Content 2A, https://links.lww.com/AJG/B758), and therefore, the assay was evaluated as a binary classifier (intermediate/high risk combined vs low risk). The same approach was taken to evaluate the diagnostic classes (IND/LGD combined vs ND).
KM analysis indicated that patients who scored intermediate/high risk with the assay were 6.7x (95% CI 3.2–13.8) more likely to progress to HGD/EAC than patients who scored low risk (Figure 2a). Patients diagnosed as IND/LGD by the 3 pathologists were 4.3x (95% CI 2.0–9.3), 5.9x (95% CI 2.7–12.9), or 6.6x (95% CI 3.1–13.8), respectively, more likely to progress than patients who were downstaged to ND (Figure 2b–d). The risk prediction assay detected 67.7% of patients who progressed within 5 years, and the 3 expert pathologists detected 75.6%, 75.8%, and 75.8% of progressors, respectively (Table 2). The specificity of the assay was 78.6%, compared with 63.9%, 68.6%, and 77.0%, respectively, for the 3 pathologists. The PPV of the risk prediction assay was 38.4% for prediction of progression within 5 years, and 56.9% for prediction of progression within 10 years. By contrast, the PPVs associated with the 3 pathologists' diagnostic classes of IND/LGD were 29.2%, 32.2%, and 39.2% for progression within 5 years, and 40.7, 50.7%, and 51.9% within 10 years (Table 2). The NPV of the assay was 92.5%, compared with 93.0%, 93.5%, and 94.2% for the ND groups according to the 3 pathologists.
In a subanalysis that evaluated progression to HGD and EAC as separate end points, the risk prediction assay and the pathologic diagnoses provided statistically significant risk stratification in predicting both endpoints (see Figure, Supplementary Digital Content 3, https://links.lww.com/AJG/B759). However, the sample size within each subset of progressors was very limited; 22 progressed to HGD and only 12 progressed to EAC.
Next, the predictive power of the assay and clinicopathologic variables was compared. In multivariate Cox models in which progression was evaluated in relation to age, sex, segment length, expert diagnosis and risk prediction assay results, both the diagnosis and the risk classes remained independent predictors of progression to HGD/EAC (see Table, Supplementary Digital Content 4, https://links.lww.com/AJG/B760). A comparison of the predictive performance of the TissueCypher risk score vs a quantitative image analysis measurement of p53 nuclear sum intensity assay demonstrated that the risk score provides superior risk prediction (AUROC = 0.7907 for prediction of progression within 5 years, P < 0.0001) over p53 alone (AUROC = 0.6689 at 5 years, P = 0.0085). Representative images of specimens from progressor and nonprogressor patients are shown in Supplementary Digital Content 5 (see Figure, https://links.lww.com/AJG/B761).
Variability in expert pathology
The 3 pathologists assigned the same diagnosis to 51.7% of cases (Table 3). There was significant variability in the diagnoses for the other 48.3% of cases; 9.7% were confirmed LGD by 2 pathologists, 21.9% were confirmed LGD by only 1 pathologist, and the remaining 16.8% were ND or IND. TissueCypher provided significant risk stratification in the subset with 100% agreement by the pathologists (n = 80) and in cases with discordant diagnoses (n = 75) (see Figure, Supplementary Digital Content 6, https://links.lww.com/AJG/B762). When the TissueCypher and pathology results were assessed adjunctively, 79.4%–85.3% of progressors were detected (i.e., scored TissueCypher intermediate/high risk and/or IND/LGD).
TissueCypher performance in patients downstaged to NDBE
TissueCypher detected 50.4%–56.1% of the progressors that were downgraded to ND by the expert pathologists (Figure 3), whereas the pathologists detected 33.3%–50.0% of the progressors that scored low risk. The PPV of the assay in patients who were downgraded to ND ranged from 16.5% to 20.3% depending on the pathologist. This indicates that patients who are downstaged to ND on expert review but score intermediate/high risk with TissueCypher progress at a rate of 3.3%–4.1% per year. TissueCypher also risk stratified within cases confirmed to be LGD by at least 1 pathologist (n = 68), and in cases downstaged to ND or IND by all 3 pathologists (n = 87) (Table 3 and see Figure, Supplementary Digital Content 6, https://links.lww.com/AJG/B762).
Post hoc analysis in progressors scoring low risk
Biopsies from additional time points were tested for 9 progressors that scored low risk (mean 5.0 ± 2.9 years between tested sample and detection of progression) and 9 matched nonprogressors. Five of the 9 progressors scored intermediate/high risk, and 4 remained low risk on a biopsy from a time point closer to progression. Two of the 9 nonprogressors were upgraded from low to intermediate/high risk. KM analysis showed that the assay had sensitivity of 77.7%, specificity of 75.9%, NPV of 94.1%, and PPV of 40.8%, demonstrating improved assay performance when additional temporal information from these 18 patients was assessed (see Figure, Supplementary Digital Content 7, https://links.lww.com/AJG/B763).
In this single-blinded cohort study, we independently validated an automated assay that risk stratifies BE patients with community-based LGD. The assay stratified BE patients with a community-based diagnosis of LGD with overall predictive accuracy comparable with 3 expert pathologists. The assay provided objective risk stratification, whereas there was significant variability between the 3 pathologists who agreed on only 51.7% of cases in this study. Furthermore, the assay identified approximately half of the progressors that the expert pathologists downstaged to NDBE. These patients represent a high-risk group who may be missed by the current standard of care but could be detected early by this risk prediction assay and undergo preventative EET. In a post hoc analysis of biopsies from additional time points, the assay detected 5/9 progressors who were not detected at baseline, improving the overall predictive accuracy of the assay. The assay provides an objective risk score, which may be a practical solution to the lack of standardization of expert pathology review of LGD. Importantly, this is the fifth study to independently validate the ability of the locked assay to risk stratify patients with BE.
An expert-confirmed diagnosis of LGD is valuable to identify patients who will benefit from EET (2,18,20). However, confirmation of LGD can be challenging, and agreement is poor even among expert pathologists (4). This was confirmed in our current study with a disagreement regarding the final histological diagnosis between the 3 pathologists in almost half of all reviewed samples (48.3%). Although patients with confirmed LGD progress at a rate of up to 10% per year, 26%–28% have no detectable LGD on follow-up (9,18,21). Up to 85% of LGD diagnoses rendered by generalist pathologists are downstaged to ND/IND on expert review, and a subset of those downstaged to ND will progress during a 3- to 5-year surveillance interval (9,18,21). The test validated in this study objectively identified 67.7% of the at-risk patients who will benefit from EET, including those who were downstaged to ND by expert pathologists. Most of the missed progressors progressed more than 5 years after the tested sample, which is outside of the 5-year predictive window of the assay. Most patients who initially scored low risk were accurately identified as intermediate/high risk when samples closer to progression were tested. A recent study demonstrated that the sensitivity of the assay can additionally be increased by testing additional biopsy levels vs a single level (22).
In contrast to the confirmation of LGD by expert pathologists, risk stratification by the assay is objective and reproducible. Based on the comparable overall performance compared with 3 renowned expert pathologists, this assay may be a valid alternative to expert review in patients diagnosed with LGD. In clinical practice, patients scoring TissueCypher assay high risk may be selected EET, regardless of whether the diagnosis is confirmed or downstaged by an expert pathologist.
Patients who score low risk with TissueCypher may be effectively managed by surveillance, potentially reducing unnecessary procedures. Overdiagnosis and the resulting unnecessary procedures is costly to the healthcare system. The average excess cost associated with overdiagnosis of LGD is $5,557 (range $3,115–$8,072) per patient in the United States (23). Although EET procedures are relatively safe, the risk of adverse events is 8.8% (6). TissueCypher may be an objective solution to overdiagnosis and overtreatment of a subset of patients with LGD. A previous study demonstrated that TissueCypher can be cost effective when used in this manner to guide clinical decision-making (24). The assay is billed once for each endoscopy encounter independently of the number of specimens submitted for testing.
The risk prediction assay offers advantages over traditional pathology methods. IHC-based assessment of p53 with manual scoring can aid the diagnosis of dysplasia (25,26), and other biomarkers have also shown promise for risk stratification (27–29). However, IHC approaches are limited by the subjective, qualitative assessment of 1 biomarker/slide. The technology validated here can also be used on standard formalin-fixed paraffin-embedded materials and objectively quantifies multiple biomarkers in tissue images of, with automatic integration of extracted data into a predictive risk score and class. Molecular diagnostic approaches have shown promise in initial studies (30,31). However, these approaches result in loss of structure, and none have been independently validated for risk prediction in BE. The technology underlying TissueCypher quantifies tissue biomarkers in situ, providing valuable contextual information.
The main strengths of this study include the cohort study design derived from the screening cohort of a randomized controlled trial (18). Although the analysis was performed retrospectively, the cohort was prospectively enrolled and followed. Additional strengths include the blinded risk prediction testing in a CLIA-certified laboratory. Although the independent review of specimens by 3 expert GI pathologists is a significant strength, the resulting predictive performance of the diagnoses does not reflect routine clinical pathology. In addition, the pathologists revised preselected slides based on dysplasia, whereas the slides tested by TissueCypher were sectioned from deeper levels in the tissue blocks. Because any additional sectioning of a tissue block can lead to a regression to the mean, a lower rate of morphological changes may be detected by the assay. A limitation of the current study is the retrospective design, although the included patients are part of a prospectively enrolled study population from a previous randomized controlled study (SURF trial). Based on the retrospective design, both clinical data and biopsy material were not available for all patients, leading to exclusion of a number of patients. Missing biopsy material did not allow for testing all endoscopic levels, which has been shown to significantly increase the accuracy of the assay. In addition, the decision to combine patients scoring intermediate- and high-risk assay results was made post hoc. However, this is the first study evaluating this objective prediction assay in a cohort of patients with a community-based diagnosis of LGD, and little was known regarding the progression rates of patients scoring intermediate/high risk.
The interobserver variation and rates of progression from confirmed LGD vary between studies with European and US pathologists (4). The pathologists in this study reviewed H&E and p53 IHC slides. Although p53 IHC has been shown to improve diagnostic stratification (32), it is not recommended by the US guidelines and is not routinely used in the United States. Additional studies could include expert GI pathologists in the United States reviewing only H&E slides and multiple community-based pathologists in Europe and the United States.
In summary, the risk prediction assay stratified BE patients with a community-based diagnosis of LGD with predictive accuracy comparable with 3 expert GI pathologists and identified progressors that the experts downstaged to NDBE. The assay provides objective risk stratification for patients with LGD, which may be an effective solution to the lack of standardization and availability of expert pathology review of LGD.
CONFLICTS OF INTEREST
Guarantor of the article: Jacques J.G.H.M. Bergman, MD, PhD.
Specific author contributions: N.F.F.: provided administrative, technical, and material support; acquired data; analyzed and interpreted the data; conceived and designed the study; reviewed and revised the manuscript; and approved the final draft submitted. A.M.K.: analyzed and interpreted the data, reviewed and revised the manuscript, and approved the final draft submitted. K.K.: provided administrative, technical, and material support; acquired data; analyzed and interpreted the data; reviewed and revised the manuscript; and approved the final draft submitted. E.A.B.: provided administrative, technical, and material support; acquired the data; analyzed and interpreted the data; reviewed and revised the manuscript; and approved the final draft submitted. K.S.: provided administrative, technical, and material support; acquired the data; analyzed and interpreted the data; reviewed and revised the manuscript; and approved the final draft submitted. Y.Z.: analyzed and interpreted the data, reviewed and revised the manuscript, and approved the final draft submitted. R.E.P.: analyzed and interpreted the data, reviewed and revised the manuscript, and approved the final draft submitted. F.J.W.t.K: acquired the data, reviewed and revised the manuscript, and approved the final draft submitted. K.A.S.: acquired the data, reviewed and revised the manuscript, and approved the final draft submitted. S.L.M.: acquired the data, reviewed and revised the manuscript, and approved the final draft submitted. R.J.C.T.: provided administrative, technical, and material support; acquired the data; analyzed and interpreted the data; supervised the study; reviewed and revised the manuscript; and approved the final draft submitted. J.J.G.H.M.B.: provided administrative, technical, and material support; analyzed and interpreted the data; supervised the study; conceived and designed the study; reviewed and revised the manuscript; and approved the final draft submitted.
Financial support: None to report. Cernostics provided research support to the Amsterdam UMC, location Academic Medical Center, Amsterdam, to conduct this investigator-initiated, industry-supported study.
Potential competing interests: E.A. Bossart has ownership interest (stock options) in Cernostics, R.J. Critchley-Thorne has ownership interest (stock, stock options and patents) in Cernostics, and Y. Zhang is a consultant to Cernostics. J.J.G.H.M. Bergman has received financial support for clinical trials from Medtronic, Pentax Medical, C2 Therapeutics, Aqua Medical, Boston Scientific, Erbe Medical, Cernostics, Ninepoint Medical, Fujifilm, and Olympus; he is a recipient of speaker's fees from Fujifilm and is a consultant for Olympus and Fractyl. The other authors declare no conflicts of interest.
WHAT IS KNOWN
- ✓ Expert confirmed that low-grade dysplasia (LGD) is the best predictor of neoplastic progression in Barrett's esophagus.
- ✓ Although recommended by guidelines, expert review is prone to interobserver variation and not widely available.
- ✓ The TissueCypher Barrett's Esophagus Assay has been shown to accurately risk stratify NDBE but was not yet evaluated in LGD.
WHAT IS NEW HERE
- ✓ The TissueCypher Barrett's Esophagus Assay provides statistically significant risk stratification in patients with LGD.
- ✓ The automated objective risk prediction assay may be an effective solution to the lack of standardization of expert pathology review of LGD.
We thank Alicia Angelo, Lianna Paul, Danyale Brazos, Stephen Hayward, Aaron DeWard, Erika Fucci, and Adriane Pawluk for technical assistance in running the TissueCypher Barrett's Esophagus Assay.
1. Spechler SJ, Sharma P, Souza RF, et al. American Gastroenterological Association medical position statement on the management of Barrett's esophagus. Gastroenterology 2011;140:1084–91.
2. Shaheen NJ, Falk GW, Iyer PG, et al. ACG clinical guideline: Diagnosis and management of Barrett's esophagus. Am J Gastroenterol 2015;111:30–50.
3. Montgomery E, Bronner MP, Goldblum JR, et al. Reproducibility of the diagnosis of dysplasia in Barrett esophagus: A reaffirmation. Hum Pathol 2001;32:368–78.
4. Vennalaganti P, Kanakadandi V, Goldblum JR, et al. Discordance among pathologists in the United States and Europe in diagnosis of low-grade dysplasia for patients with Barrett's esophagus. Gastroenterology 2017;152:564–70.e4.
5. Weusten B, Bisschops R, Coron E, et al. Endoscopic management of Barrett's esophagus: European Society of Gastrointestinal Endoscopy (ESGE) position statement. Endoscopy 2017;49:191–8.
6. Qumseya BJ, Wani S, Desai M, et al. Adverse events after radiofrequency ablation in patients with Barrett's esophagus: A systematic review and meta-analysis. Clin Gastroenterol Hepatol 2016;14:1086–95.e6.
7. Wani S, Falk GW, Post J, et al. Risk factors for progression of low-grade dysplasia in patients with Barrett's esophagus. Gastroenterology 2011;141:1179–86.e1.
8. Thota PN, Lee HJ, Goldblum JR, et al. Risk stratification of patients with Barrett's esophagus and low-grade dysplasia or indefinite for dysplasia. Clin Gastroenterol Hepatol 2014;13:459–65.e1.
9. Duits LC, Phoa KN, Curvers WL, et al. Barrett's oesophagus patients with low-grade dysplasia can be accurately risk-stratified after histological review by an expert pathology panel. Gut 2015;64:700–6.
10. Goldblum JR. Controversies in the diagnosis of Barrett esophagus and Barrett-related dysplasia: One pathologist's perspective. Arch Pathol Lab Med 2010;134:1479–84.
11. Krishnamoorthi R, Singh S, Ragunathan K, et al. Factors associated with progression of Barrett's esophagus: A systematic review and meta-analysis. Clin Gastroenterol Hepatol 2018;16:1046–55.e8.
12. Davison JM, Shah MB, Deitrick C, et al. Low-grade dysplasia diagnosis ratio and progression metrics identify variable Barrett's esophagus risk stratification proficiency in independent pathology practices. Gastrointest Endosc 2018;88:807–15.e2.
13. Critchley-Thorne RJ, Duits LC, Prichard JW, et al. A tissue systems pathology assay for high-risk Barrett's esophagus. Cancer Epidemiol Biomarkers Prev 2016;25:958–68.
14. Critchley-Thorne RJ, Davison JM, Prichard JW, et al. A tissue systems pathology test detects abnormalities associated with prevalent high-grade dysplasia and esophageal cancer in Barrett's esophagus. Cancer Epidemiol Biomarkers Prev 2017;26:240–8.
15. Davison JM, Goldblum J, Grewal US, et al. Independent validation of a tissue systems pathology test to predict progression Barrett's esophagus patients. Am J Gastroenterol 2020;115:843–52.
16. Prichard JW, Davison JM, Campbell BB, et al. TissueCypher: A systems biology approach to anatomic pathology. J Pathol Inform 2015;6:48.
17. DeWard A, Critchley-Thorne RJ. Systems biology approaches in cancer pathology. Methods Mol Biol 2018;1711:261–73.
18. Phoa KN, van Vilsteren FG, Weusten BL, et al. Radiofrequency ablation vs endoscopic surveillance for patients with Barrett esophagus and low-grade dysplasia: A randomized clinical trial. JAMA 2014;311:1209–17.
19. Schlemper RJ, Riddell RH, Kato Y, et al. The Vienna classification of gastrointestinal epithelial neoplasia. Gut 2000;47:251–5.
20. Small AJ, Araujo JL, Leggett CL, et al. Radiofrequency ablation is associated with decreased neoplastic progression in patients with Barrett's esophagus and confirmed low-grade dysplasia. Gastroenterology 2015;149:567–76.e3; quiz e13–4.
21. Shaheen NJ, Sharma P, Overholt BF, et al. Radiofrequency ablation in Barrett's esophagus with dysplasia. N Engl J Med 2009;360:2277–88.
22. Frei NF, Konte K, Bossart EA, et al. Independent validation of TissueCypher to predict future progression in non-dysplastic Barrett's esophagus: A spatial-temporal analysis. Clin Transl Gastroenterol 2020;11:e00244.
23. Lash RH, Deas TM Jr, Wians FH Jr. Healthcare cost of over-diagnosis of low-grade dysplasia in Barrett's esophagus. Adv Ther 2016;33:684–97.
24. Hao J, Critchley-Thorne RJ, Diehl DL, et al. A cost-effectiveness analysis of an adenocarcinoma risk prediction multi-biomarker assay for patients with Barrett's esophagus. Clinicoecon Outcomes Res 2019;11:623–35.
25. Fitzgerald RC, di Pietro M, Ragunath K, et al. British Society of Gastroenterology guidelines on the diagnosis and management of Barrett's oesophagus. Gut 2014;63:7–42.
26. Snyder P, Dunbar K, Cipher DJ, et al. Aberrant p53 immunostaining in Barrett's esophagus predicts neoplastic progression: Systematic review and meta-analyses. Dig Dis Sci 2019;64:1089–97.
27. Kastelein F, Biermann K, Steyerberg EW, et al. Value of alpha-methylacyl-CoA racemase immunochemistry for predicting neoplastic progression in Barrett's oesophagus. Histopathology 2013;63:630–9.
28. van Olphen S, Biermann K, Spaander MC, et al. SOX2 as a novel marker to predict neoplastic progression in Barrett's esophagus. Am J Gastroenterol 2015;110:1420–8.
29. Duits LC, Lao-Sirieix P, Wolf WA, et al. A biomarker panel predicts progression of Barrett's esophagus to esophageal adenocarcinoma. Dis Esophagus 2019;32:doy102.
30. Stachler MD, Camarda ND, Deitrick C, et al. Detection of mutations in Barrett's esophagus before progression to high-grade dysplasia or adenocarcinoma. Gastroenterology 2018;155:156–67.
31. Sepulveda JL, Komissarova EV, Kongkarnka S, et al. High-resolution genomic alterations in Barrett's metaplasia of patients who progress to esophageal dysplasia and adenocarcinoma. Int J Cancer 2019;145:2754–66.
32. van der Wel MJ, Duits LC, Pouw RE, et al. Improved diagnostic stratification of digitised Barrett's oesophagus biopsies by p53 immunohistochemical staining. Histopathology 2018;72:1015–23.