Predicting Neoplastic Progression in Barrett's Esophagus

Abstract

Patients with Barrett's esophagus have a significantly increased risk of esophageal adenocarcinoma, 40-125 times higher than the general population. Since only a small fraction of Barrett's esophagus patients will actually progress to esophageal adenocarcinoma, there is a need to develop markers that may accurately predict which patients with Barrett's esophagus are likely to have aggressive disease and progress to cancer versus patients who will remain histologically stable and have a benign course. This would allow for better risk stratification of patients with Barrett's esophagus in order to target aggressive surveillance and intervention towards only those patients at highest risk for neoplastic progression. Predictive biomarkers may thus have significant clinical utility in the management of Barrett's esophagus patients. The detection of dysplasia in esophageal biopsies is currently the only standard method used in clinical practice as a marker for increased risk of cancer. However, dysplasia has not been a accurate or reliable marker for predicting malignant progression and suffers from poor interobserver agreement among pathologists and sampling error. A multitude of potential biomarkers have been studied over the years. It is likely that the best model for predicting progression to esophageal adenocarcinoma in Barrett's esophagus patients will ultimately involve a combination of biomarkers, dysplasia grade and other pathological characteristics, as well as clinical and demographic attributes. In this review, we will discuss the most promising biomarkers that have been studied thus far.

Introduction

The primary risk factor for esophageal adenocarcinoma (EAC) is Barrett's esophagus (BE), which is a premalignant condition involving a change in the lining of the esophagus from normal squamous epithelium to columnar epithelium with specialized intestinal metaplasia.¹ BE is one of the most common premalignant conditions and affects an estimated 1.6% of the adult population.² Patients with BE have a significantly increased risk of EAC, 40-125 times higher than the general population.^{1, 3} However, among all patients with BE, only a small proportion will actually progress to EAC.^4-7 BE progression is thought to proceed in a multi-step fashion in a metaplasia-dysplasia-EAC sequence over a period of many years.⁸

The triggers for neoplastic progression of BE are likely to be multifactorial in nature, involving chromosomal aberrations, genetic and epigenetic events, as well as environmental factors. It is thought that as cells accumulate genetic abnormalities, they are selected to proliferate and spread in the BE segment with a survival advantage over normal cells in a process called clonal expansion, ultimately leading to the development of EAC.⁹ Our increased understanding of carcinogenesis has allowed for the identification of biomarkers for predicting risk of progression from BE to EAC.

The ideal biomarker should have high specificity, be relatively noninvasive, and easily detectable at little cost. Similar to the drug development process, the National Cancer Institute has recommended five phases that a biomarker should pass through in order for it to be confidently used in clinical practice (Table 1).¹⁰ Others have also advocated for an additional sixth phase which would involve validation of the biomarker in multiple centers and in multiple populations.¹¹

Table 1.

Phases of biomarker development¹⁰

Phase	Description	Study Design	Aims
1	Preclinical exploratory studies	Case-control studies comparing tumor vs. nontumor tissues	Identify leads for potentially useful biomarkers by comparing tumor tissue with nontumor tissue to identify characteristics unique to tumor tissue
2	Clinical assay development	Case-control studies comparing cancer patients vs. noncancer patients	Assess the ability of the biomarker to distinguish subjects with cancer from subjects without cancer by determining sensitivity and specificity for the clinical biomarker assay
3	Retrospective longitudinal repository studies	Longitudinal nested case-control studies of patients who develop cancer vs patients who do not	Evaluate capacity of biomarker to detect preclinical disease and to define criteria for a positive screening test
4	Prospective screening studies	Prospective cohort	Determine the operating characteristics of the biomarker-based screening test in a typical population such as the detection rate, the stage of the cancer at the time it can be detected by the biomarker, and the number of subjects falsely screening positive and referred for work-up
5	Cancer control studies	Randomized controlled trials	Estimate reduction in cancer mortality associated with the biomarker screening test

Methods

We used the Pubmed search engine with the search terms “Barrett's esophagus” and “progression” and “biomarkers” to retrieve all English language articles published from November 1999 until November 2009. These articles were then evaluated carefully for inclusion in this review.

Dysplasia

The grade of dysplasia in esophageal biopsies from patients with BE is currently the standard biomarker used in clinical practice as a predictor for increased risk of progression to EAC.^{12, 13} Dysplasia is the basis for current guidelines on the surveillance and management of patients with BE.^14-16 These guidelines recommend the use of a systematic biopsy protocol which involves 4 quadrant biopsies within the Barrett's segment at intervals of 2 cm in patients without dysplasia or those with low-grade dysplasia (LGD), and intervals of 1 cm in patients with high-grade dysplasia (HGD).

Published prospective studies on the natural history of patients with Barrett's esophagus have indicated a range of estimates for risk of progression to EAC, depending on the initial grade of dysplasia.^{3, 17-31} BE patients with no dysplasia are thought to be at very low risk for progression to EAC with a 0-5% risk. Meanwhile, BE patients with LGD are estimated to have a 3-15% risk of progression to EAC. When at least two pathologists agreed on a diagnosis of LGD, there was an even stronger association with progression.²⁶ BE patients with HGD are recognized to have the highest risk of progression to EAC ranging from 16-61%. A recent meta-analysis on patients with BE and HGD calculated more precise estimates on the incidence rate of EAC to be approximately 6 per 100 patient-years, or 6% per year.³²

Furthermore, the extent of dysplasia may influence risk of progression, although study results have been mixed. Among patients with LGD, progressors to EAC had a higher mean proportion of LGD crypts compared to non-progressors (46.4% vs. 26.0%).³³ One study found that extent of HGD did not predict EAC risk.³⁴ However, in another study, patients with BE and diffuse HGD had a 3.7-fold increase in risk of EAC compared to those with only focal HGD (95% CI, 1.21–11.40). In addition, patients with nodularity in their BE (defined as subtle mucosal elevation of diameter <= 1 cm) had 2.6 times the risk of EAC as those without nodularity (95% CI, 1.24–5.43).³⁵

Dysplasia has not been a reliable biomarker for predicting malignant progression, as seen in the wide range of estimates for risk of progression to EAC in phase 4 studies. Furthermore, the use of dysplasia as a biomarker has been limited by poor inter-observer agreement among pathologists for the diagnosis of dysplasia ^{23, 36, 37} and the inherent problem of incomplete sampling of the Barrett's mucosa by standard mucosal biopsy techniques. In addition, the use of dysplasia as a biomarker subjects the patient to numerous invasive endoscopies. Therefore, there is a great need for improved predictors of progression to EAC so that better risk stratification of patients with BE can be performed in order to develop more cost-effective preventive and follow-up programs.

Genomic Instability

One of the best studied biomarkers for progression in BE has been flow cytometry of DNA content in esophageal biopsies, which has shown promise in phase 4 studies. Initial prospective studies showed that 69% (9/13) of patients who had aneupl³⁸oidy or tetraploidy in their initial biopsies later went on to develop HGD or EAC whereas 0% (0/49) of patients who did not have these DNA content abnormalities progressed.³⁹ Subsequent long-term prospective studies of patients with BE confirmed that flow cytometry findings of either aneuploidy (greater than 2.7N) or increased 4N (greater than 6%) were highly predictive of progression to cancer.^{38, 40} In 247 BE patients with lower grades of dysplasia (negative, indefinite, or low-grade), having aneuploidy or increased 4N was associated with a 28% risk of cancer over 5 years. In contrast, in patients without either of these findings, none progressed to cancer.²² High diploid S-phase has also recently been found to be predictive of EAC.⁴¹ Abnormal DNA content has been found to be a result of inactivation of p53.^41-44

Bacterial Artificial Chromosome (BAC) array comparative genomic hybridization is an alternative way to measure DNA content abnormalities by measuring copy number alterations, which correlated highly with DNA aneuploidy. DNA copy number changes involving > 70 Mbp were associated with a higher risk of progression to esophageal cancer.⁴⁵

One of the limitations of these prospective studies on abnormal DNA content as a predictor of progression to EAC is that these assays have all been done on fresh frozen tissue. The assay would be more difficult to routinely perform in a clinical setting since immediate processing would be required after biopsy in order to avoid false-positive aneuploidy results.³⁸ However, other groups have performed the technique of flow cytometry on paraffin-embedded tissue with comparable results to fresh frozen tissue.^46-48 In addition, fluorescent in situ hybridization (FISH) to detect chromosomal gains from esophageal brush cytology appears to be promising and may be even more sensitive than conventional flow cytometry.⁴⁹ However, prospective studies using FISH or flow cytometry in paraffin-embedded biopsies to evaluate DNA content abnormalities as a biomarker for progression from BE to EAC have not yet been done.

Proliferation Markers

Increased cell proliferation has been found in BE patients, and a number of proliferation markers have been studied in the carcinogenesis of BE and as biomarkers for predicting neoplastic progression.⁵⁰

PCNA, Ki-67

Proliferating cell nuclear antigen (PCNA) is a cofactor of DNA polymerase delta required for cellular proliferation, and was the first proliferation marker that could be detected by immunohistochemical staining in paraffin-embedded tissue. PCNA has been found to be elevated in Barrett's mucosa in some studies^51-53 but not in others.⁵⁴

Ki-67 protein is another known marker of proliferation, and is a nuclear antigen present during all active phases of the cell cycle, but absent in resting cells.⁵⁵ Ki-67 immunohistochemistry has been found to increase with increasing dysplasia ^56-58 but was not always able to distinguish between reactive epithelia from LGD.⁵⁹ Thus, neither PCNA or Ki-67 has progressed beyond phase 2 studies.

Mcm2, Cyclin A

More recently, minichromosome maintenance protein 2 (Mcm2) has been discovered to be another marker of proliferation which is expressed in all cycling cells throughout the cell cycle. In a longitudinal case-control study, BE patients who progressed from no dysplasia to EAC had significantly higher Mcm2 immunostaining than matched BE controls who did not progress (mean, 28.4 and 3.4% positive cells, respectively).⁶⁰ However, Mcm2 was found to have low specificity (35%).

Subsequently, the same group found a related marker of proliferation, cyclin A, to be more sensitive and specific compared to Mcm2.⁶¹ In a longitudinal case-control study, BE patients expressing cyclin A at the surface were significantly more likely to progress to EAC than those who did not express cyclin A (OR=7.5; 95% CI 1.8-30.7).⁶² In addition to traditional tissue immunostaining of esophageal biopsies, Mcm2 and cyclin A were both also able to be detected with high accuracy in esophageal brushings using a liquid-based cytology technique. The sensitivity and specificity of cyclin A expression in brushings for the detection of high-grade dysplasia and cancer patients were 97.8% and 58.7%, respectively, with a negative predictive value of 97.4%. In summary, Mcm2 and cyclin A have shown promise in phase 3 studies, and also have the advantage to be able to be detected in esophageal brushings.

Cell Cycle Regulatory Genes

Several genes involved in controlling progression of the cell cycle have been found to influence the progression of BE to EAC (Figure 1).

Figure 1.

Proposed algorithm for using biomarkers to predict clinical outcome in BE

Cyclin D1

Cyclin D1 is a proto-oncogene which complexes with and activates the cyclin-dependent kinases (CDK4/6). The activated kinases phosphorylate the retinoblastoma protein Rb, thereby inactivating it. In the hypophosphorylated state, Rb is active and blocks progression of the cell cycle from phase G1 to S (synthesis). However, in the phosphorylated state, Rb becomes inactive and therefore allows progression from G1 to S phase, stimulating proliferation. Increased nuclear staining of cyclin D1 has been seen in patients with BE.⁶³ Cyclin D1 overexpression was seen in 8/12 patients who later progressed to cancer vs. 14/49 patients who did not progress (OR=6.85; 95% CI 1.57-29.91).⁶⁴ However, other subsequent studies have not shown correlation between cyclin D1 expression and dysplasia grade or progression to EAC.^{61, 65} Therefore, cyclin D1 has shown mixed results in phase 3 studies.

p16

p16 is a tumor suppressor gene located on the short arm of chromosome 9 which complexes with and inhibits CDK4/6. It does so by preventing the formation of cyclin D1/CDK4/6 complexes, which then prevents phosphorylation of Rb. When p16 is inactivated, CDK4/6 is able to be activated and promotes Rb phosphorylation, leading to proliferation. In BE, inactivation of p16 commonly occurs through LOH of 9p21 or p16 methylation, and less commonly via mutations.⁶⁶ In 7/7 patients with BE, allelic loss of 9p21(p16) preceded subsequent development of aneuploidy and cancer.⁶⁷ Among BE patients, it appears that p16 LOH, promoter methylation, and mutations have strong independent effects on BE cells in clonal expansions and precedes other lesions such as p53 abnormalities, aneuploidy, and tetraploidy.⁶⁸ Further studies on p16 methylation are discussed below.

p53

p53 is a tumor suppressor gene located on the short arm of chromosome 17 and is involved in regulating cell cycle control. When cells sustain DNA damage and cannot be repaired, p53 induces apoptosis and thereby prevents the replication of genetic instability which could lead to cancer. Although p53 mutations can be inherited in Li-Fraumeni syndrome ⁶⁹, in BE the abnormal p53 usually occurs when one allele has been deleted (usually via mutation) and the other allele is functionally inactivated often due to loss of heterozygosity (LOH) in a two-hit mechanism.^{70, 71}

p53 overexpression by immunohistochemistry

The utility of p53 protein overexpression as a biomarker for neoplastic progression has been evaluated in a number of studies. Missense mutations of the p53 gene result in amino acid substitutions in the protein which causes it to have a much longer half-life than normal, resulting in accumulation in the nucleus where its overexpression can be detected by immunohistochemical staining.⁷² Therefore, the detection of p53 protein overexpression is felt to be a surrogate marker for p53 gene abnormalities.⁷³

Studies have shown increasing p53 overexpression with increasing dysplasia in BE.^74-76 However, studies evaluating the use of p53 overexpression as a predictor of progression to EAC have had mixed results. In a study of BE patients with LGD, 56% (5/9) of patients with positive p53 overexpression progressed on to high-grade dysplasia or cancer compared to 0% (0/16) of patients with negative p53 overexpression.⁷⁷ This was confirmed in a subsequent prospective study of patients with Barrett's esophagus and low-grade dysplasia, where positive p53 overexpression was seen in 60% (3/5) of patients who progressed to high-grade dysplasia or cancer, compared to 16% (7/43) of patients who persisted with LGD or regressed.¹⁷ Patients who were positive for p53 overexpression and had at least 3 pathologists agree on the diagnosis of their LGD were much more likely to progress to HGD or EAC compared to patients who were negative for p53 overexpression and did not have agreement on their LGD diagnosis.⁷⁸

However, in another prospective study of patients with BE using only cancer as the endpoint, p53 overexpression was not a significant predictor of progression to EAC. In this study, p53 overexpression was seen in 36% (4/11) patients who later progressed to cancer vs. 17% (7/41) patients who did not progress.⁶⁴ More recently, in a longitudinal case-control study of BE patients, those who progressed to EAC were significantly more likely to have p53 overexpression compared to those who did not progress (OR=11.7; 95% CI 1.93-71.4). However, despite this strong association, only 32% of progressors actually had initial biopsies positive for p53 overexpression. Therefore, the sensitivity of p53 overexpression was too low to be useful as an independent predictor.⁶⁵

Other issues have arisen with p53 overexpression, including its nonspecificity for p53 gene mutations, with greater than 25% false negative and false positive results compared to DNA sequencing, making it problematic to use as a clinical marker for predicting the development of cancer in patients with BE.⁷⁹ These are likely due to the fact that some p53 gene mutations will cause truncations which result in a protein that can not be detected by immunohistochemistry.^{73, 80} Overall, p53 overexpression has had mixed results in phase 4 studies.

p53 LOH

LOH of 17p(p53) has been found to be a predictor of progression to cancer. LOH of p53 was detected in 6 patients initially negative for dysplasia who subsequently had dysplasia or EAC on follow-up.⁸¹ In a prospective study of patients with Barrett's esophagus, 37% (20/54) patients who had p53 LOH progressed to cancer whereas only 3% (6/202) patients who did not have p53 LOH progressed to cancer (RR = 16; 95% CI 6.2-39).⁸² Therefore, p53 LOH appears promising in phase 4 studies.

p53 antibodies

Antibodies to p53 were detected in the plasma of esophageal cancer patients in a very small study, and in two patients they were detected before progression to cancer.⁸³ This phase 2 study is promising but this biomarker will need to be evaluated in subsequent longitudinal studies.

Telomeres

Telomeres are repetitive DNA sequences at chromosome ends, which protect them from degradation. With each cell division, normally a fragment of telomeres are lost during the replication process. After a number of cell divisions, the telomeres become too short to protect the chromosomes, causing growth arrest. This process is accelerated by oxidative stress, and shorter leukocyte telomeres have been associated with various cancers, supporting the hypothesis that mean leukocyte telomere length represents the level of cellular injury.⁸⁴ In some cancers, telomerase is activated, which synthesizes and maintains telomeres, causing immortalization of the cancer cells.

Several studies have shown an increase in telomerase with increasing dysplasia.^{85, 86} In a prospective study, shorter leukocyte telomere length in blood samples at baseline predicted risk of EAC in BE patients, with an age-adjusted hazard ratio of 3.45 between the top and bottom quartiles of telomere length (95% CI 1.35-8.78).⁸⁷ Therefore, leukocyte telomere length appears to be a promising biomarker in phase 4 studies.

Methylation

DNA promoter hypermethylation is an epigenetic phenomenon that holds promise as a molecular marker for early detection of malignant transformation, since this alteration is found in pre-invasive neoplastic tissue and involves genes which regulate key pathways in cancer.^88-98 Methylation of the promoter region of a gene results in suppression of gene transcription.⁹⁹ Methylation occurs at cytosine (C) bases that are located 5′ to a guanosine (G) base. Regions rich in C-G dinucleotides (known as CpG islands) are found in the promoters of many genes and are normally unmethylated. In cancers, however, hypermethylation often occurs in CpG islands and inhibits DNA transcription, resulting in the silencing of gene expression. For example, hypermethylation has been found to play a role in the inactivation of several tumor suppressor genes, leading to neoplastic progression.¹⁰⁰ Hypermethylation of various genes have been found in BE and EAC, with p16 methylation being the most well-studied.^{66, 101-109} Other genes of note include runt-related transcription factor 3 (RUNX3, a tumor suppressor gene), HPP1 (a transmembrane-encoding gene frequently methylated in hyperplastic polyposis), adenomatous polyposis coli (APC, a tumor suppressor gene), tissue inhibitor of metalloproteinase 3 (TIMP3, inhibitor of the matrix metalloproteinases which are involved in degradation of the extracellular matrix), and telomerase reverse transcriptase (TERT, ribonucleoprotein polymerase which maintains telomere ends).

In a retrospective longitudinal case control study of BE patients with negative or low-grade dysplasia, methylation of p16 (OR=1.74, 95% CI 1.33–2.20), RUNX3 (OR=1.80, 95% CI 1.08–2.81), and HPP1 (OR=1.77, 95% CI 1.06–2.81) were associated with an increased risk of progression to high-grade dysplasia or EAC.¹¹⁰ Another study found that BE mucosa adjacent to EAC was frequently methylated in APC (100%), TIMP3 (91%), and TERT (92%) compared to BE mucosa from patients negative for dysplasia (36%, 23%, and 17%, respectively).¹¹¹ A subsequent longitudinal case-control study of BE patients showed that methylation of both p16 and APC was a strong predictor of subsequent progression to cancer (OR 14.97; 95% CI 1.73-infinity). Among patients who were negative for both p16 and APC methylation, none progressed from baseline pathology to high-grade dysplasia or cancer.¹¹² In summary, methylation of various genes has shown promise as a biomarker for progression in BE patients in phase 3 studies. Further studies will need to be undertaken in prospective trials.

Biomarker Panels

Given the complexity involved in the carcinogenesis of EAC from BE, it is likely that the best predictor of progression to EAC will be a combination of biomarkers. Several recent studies have found that a combination panel of biomarkers is a better predictor than individual biomarkers alone. Recently, a prospective study of BE patients found that a combination of DNA content abnormalities (tetraploidy and aneuploidy), 17p LOH, and 9pLOH provided the best prediction of risk for EAC (RR = 38.7; 95% CI 10.8-138.5). Patients with all of these findings had at least a 79% EAC risk over 10 years, whereas patients with none of these findings had only a 12% risk of EAC over 10 years.¹¹³

In another prospective study of BE patients, the combination of genetic instability and clonal expansion predicted progression to EAC. Taking into account the sizes of clones with p53 LOH (RR=1.27x for an x cm clone; 95% CI 1.07–1.50) or aneuploidy/tetraploidy (RR=1.31x for an x cm clone; 95% CI 1.07–1.60) predicted progression to EAC compared to the mere presence of such clones. A 5 cm clone containing either p53 LOH or aneuploidy/tetraploidy was associated with a RR of 4.16 (95% CI 2.01-8.95) for progression to EAC.⁹

Meanwhile, environmental exposures such as obesity ¹¹⁴, diet ¹¹⁵, and nonsteroidal anti-inflammatory drugs ^{113, 116} have also been postulated to influence the development or progression of BE. These environmental exposures could also be a potential source of genetic instability. One way to assess genetic instability in a unified manner (whether the source is environmental, genetic, or epigenetic) is to measure clonal and genetic diversity. One prospective study of BE patients found that increased clonal diversity predicted progression to EAC. This suggests that the accumulation of clonal genetic variants is important for neoplastic progression. In this study, the combination of factors that best predicted progression included the number of clones, genetic divergence, p53 LOH, and ploidy abnormalities.¹¹⁷

The use of biomarker panels can be cumbersome, requiring multiple platforms to perform the various assays. Whole genome single nucleotide polymorphism (SNP) arrays provide an alternate way to assess both DNA content abnormalities and LOH in a single platform. A recent study using SNP arrays found that changes in DNA copy number were common in esophageal cancers, with an average of 97 copy number changes per cancer.¹¹⁸ The total number of SNP alterations was highly correlated with DNA content aneuploidy, and both SNP abnormalities and LOH increased in frequency and size from the early to advanced stages of BE.¹¹⁹

Novel Molecular Markers

Gene microarray analysis (Affymetrix cDNA microarray chips) of epithelial cells microdissected from archival BE tissue specimens have identified differentially expressed genes in nondysplastic BE from those in matched biopsies of BE with HGD¹²⁰. Among the overexpressed genes are several previously shown to be increased in the neoplastic progression of BE, as well as novel genes such as lipocalin-2, S100A9, matrix metallopeptidase 12, secernin 1, and topoisomerase IIalpha. Genes decreased in dysplastic epithelium include MUC5AC, trefoil factor 1 (TFF1), meprin A, and CD13. Real-time PCR validated the changes in expression in 24 of 28 selected genes. Immunohistochemistry confirmed increased protein expression for topoisomerase IIalpha, S100A9, and lipocalin-2 and decreased expression of TFF1 across the spectrum of BE-associated dysplasia from nondysplastic BE through adenocarcinoma. These novel molecular markers warrant further investigation.

CONCLUSIONS

In recent years, we have gained an increased understanding in the mechanisms underlying the progression of BE metaplasia to EAC. Based on findings from the studies discussed above, it appears that p16 inactivation via LOH or methylation is one of the earliest events in the carcinogenesis sequence. This is followed by p53 inactivation via mutations and LOH, which then leads to DNA content abnormalities such as aneuploidy and tetraploidy. These studies have led to the identification of promising biomarkers (Table 2).

Table 2.

Current status of biomarkers for predicting progression from Barrett's metaplasia to esophageal adenocarcinoma based on the National Cancer Institute's phases of biomarker development

Biomarker	Phase of Development	Comments
Dysplasia14	4	Currently used in clinical practice but no clinical trial demonstrating reduced mortality from surveillance has been performed yet
DNA content abnormalities 22, 38-41 (aneuploidy, tetraploidy)	4	Phase 4 studies performed on fresh frozen tissue, need to confirm results in paraffin-embedded tissue
p53 overexpression17,64-65,77-79	4	Mixed results. Low sensitivity of only 32%
p53 loss of heterozygosity (LOH)81-82	4	LOH done on fresh frozen tissue
Leukocyte telomere length87	4	Only one phase 4 study done so far
Combinations:
- p53 LOH, p16 LOH, aneuploidy/tetraploidy113	4
- p53 LOH, aneuploidy/tetraploidy, clone size9	4
- # clones, genetic divergence, p53 LOH, aneuploidy/tetraploidy117	4
Methylation of various genes110-112	3	Prospective studies needed
minichromosome maintenance protein 2 (Mcm2)60	3	Low specificity of only 35%
Cyclin A61-62	3	More sensitive and specific compared to Mcm2, can detect on esophageal brushing
Cyclin D163-65	3	Mixed results
PCNA51-54	1 / 2	Mixed results
Ki-6755-59	1 / 2	Mixed results
p53 plasma antibodies83	1 / 2	Only one small study done so far
Adhesion molecules (α and β-catenin, and E-cadherin)	1 / 2	Further studies needed
Inflammatory markers (COX-2, TNF-α, NF-κB)	1 / 2	Further studies needed

No biomarker currently studied for predicting progression of BE to EAC has completed all five phases. The difficulty in studying biomarkers for predicting progression in BE stems from the fact that even though BE patients are at increased risk for developing EAC, the natural history of BE has shown that over a long period of time only a small proportion will actually progress on to EAC. Therefore, it is difficult to perform prospective studies which will have enough progressors to provide adequate power for statistical analysis, unless multi-center collaborations are pursued. Furthermore, the majority of biomarkers being studied still depend on mucosal tissue sampling. Therefore, like detection of dysplasia, these new biomarkers will still be subject to the inherent problem of sampling error.

However, recent developments in endoscopic imaging may help in targeting biopsies and therefore increase the yield of detecting dysplasia or other biomarkers. High-resolution endoscopy has significantly increased the ability to discriminate among subtle lesions. Narrow-band imaging (NBI) filters out all wavelengths except for a narrow band of blue light, resulting in the ability to highlight the superficial layers of mucosa and enhance vascular visualization. In combination with high-resolution endoscopy, NBI-targeted biopsies detected dysplasia in more patients (57%) compared to standard endoscopy with random biopsies (43%).¹²¹ Autofluorescence differentiates tissue types based on differences in fluorescence emission. Dysplastic areas do not have as intense autofluorescence and appear dark red. Although autofluorescence has been found to have high sensitivity for high-grade dysplasia (100%), it has a 40% false positive rate.¹²² Confocal laser endomicroscopy (CLE) involves real-time, in-vivo microscopic imaging during endoscopy with magnification greater than 1000x. CLE has been found to have high sensitivity and specificity (85-94%) for detection of HGD. Although it has not been proven to increase detection of dysplasia over white light endoscopy, it appears to reduce the number of biopsies needed.^{123, 124}

Thus far, the only biomarkers to be evaluated in phase 4 studies are dysplasia, DNA content abnormalities (aneuploidy, tetraploidy), p53 LOH, p53 overexpression (although with mixed results), and leukocyte telomere length. Biomarkers currently in phase 3 studies include Mcm2, cyclin A, cyclin D1 (although with mixed results), and methylation of various genes. There are numerous other biomarkers which have been investigated in phase 1 and 2 studies including PCNA, Ki-67, and p53 antibodies, chromosomal abnormalities ^{56, 125}, gains of proto-oncogenes ¹²⁶, reduced expression of adhesion molecules (such as α-catenin, β-catenin, and E-cadherin), and inflammation-association markers (such as COX-2, TNF-α, NF-κB) but there have been conflicting results. Ultimately, it is likely that the best model for predicting clinical outcome in BE patients will involve a combination panel of biomarkers as well as demographic attributes and environmental exposures.