A study by Takubo et al. which examined the mucosa adjacent to EAC treated with endoscopic mucosal resection found that most (>70%) were bordered by cardiac-type mucosa rather than intestinal-type mucosa and that 56% had no intestinal-type mucosa in any areas of the resection specimens. They concluded that there is a relationship between EAC and cardiac-type mucosa and that a background of intestinal metaplasia may not be a necessary pre-requisite to EAC (6).

Two similar studies by Chandrasoma and colleagues had different findings. The first, which examined esophagogastrectomy specimens resected due to adenocarcinoma, showed cardiac mucosa adjacent to all tumors but also showed residual intestinal metaplasia in 65% of cases overall and in 100% of intramucosal tumors as well as those less than 1 cm in diameter (12). The second study reviewed two groups: (I) cases with visible columnar metaplasia of the esophagus which underwent systematic 4-quadrant biopsies every 1 to 2 cm and (II) cases of dysplasia or EAC which did not receive systematic biopsy. They found that when systematic biopsy was performed, intestinal metaplasia was identified in >87% of cases including all cases with dysplasia or EAC. None of the cases with cardiac type epithelium alone had dysplasia or EAC. In the group which did not receive systematic biopsy but did have dysplasia or EAC, many showed only tumor on biopsy. However, slightly more than half (56%) of those with nontumor mucosa had residual intestinal-type metaplasia (13). They hypothesize that the absence of residual intestinal metaplasia immediately adjacent to many cases of EAC is due to tumor overgrowth and inadequate sampling rather than a true absence (12,13). They also propose that when metaplastic columnar epithelium is adequately and systematically biopsied, patients without intestinal metaplasia have a negligible risk of dysplasia and cancer (13.

Recent data shows that columnar cell epithelium may have an intestinal-type immunohistochemical profile even when goblet cells are not identified. Various studies have shown significantly increased positivity for intestinal markers such as DAS-1 (14-16), CDX-2 (14,17,18), and HepPar1 (19) as well as a similar cytokeratin (16,20) and mucin (20) expression profile in both goblet cell and nongoblet cell columnar epithelia, which suggests a similar origin. There have also been studies showing similar molecular alterations in both non-goblet cell and intestinaltype metaplasia including chromosomal instability (21,22), microsatellite instability (22), and similar DNA content abnormalities (3). Despite the similar phenotypic and molecular profiles, the natural history of columnar cells and goblet cells is not always the same (24) suggesting that additional factors are required for progression toward dysplasia and cancer.

While the presence of GERD symptoms was one of the first recognized and strongest risk factors identified for BE, the presence of GERD alone is not sufficient to recommend screening. Up to 40% of US adults experience GERD on a monthly basis (28), yet despite the increasing incidence of EAC there are still fewer than 10,000 new cases of EAC diagnosed per year (29). Up to 40% of patients who have adenocarcinoma of the esophagus report no history of chronic GERD (30). Eliminating patients from screening based on a lack of symptoms could exclude a large portion of those who might have their cancers detected at an early, presymptomatic stage. Additionally, difficulties recognizing mucosal lesions (31), sampling error (32), and disagreement over pathologic interpretation (33) can decrease the effectiveness of endoscopic screening. For these reasons, the decision of who and when to screen should be individualized (1,4).

Due to the significance of identifying dysplasia, much work has gone into clarifying and refining the criteria used to interpret biopsies (33,54-57). The degree of dysplasia is determined by evaluating the cytology (nuclear and cytoplasmic features), architecture (relationship of glands and lamina propria), and degree of surface maturation (comparison of nuclear size within crypts to nuclear size at the mucosal surface) and interpreting these findings in conjunction with the amount of background inflammation. Features of each category of dysplasia are described below and summarized in Table 1.

Negative for dysplasia – These biopsies can have a minimal amount of cytologic atypia but retain normal architecture, abundant lamina propria between glands, and appropriate maturation with a low nuclear:cytoplasmic ratio at the mucosal surface. The nuclei are regular, have smooth membranes, and are basally situated. If mitoses are present they are within the basal compartment. In the presence of inflammation, increased cytologic atypia is allowed (Figure 1A).

Indefinite for dysplasia – This category is applied to biopsies where the changes seen cannot be definitively described as reactive or neoplastic. It is most often used in the presence of pronounced inflammation or the loss of surface epithelium. Cytologic atypia characterized by hyperchromasia, overlapping nuclei, irregular nuclear borders, and nuclear stratification can be seen in the deep glands or the sides of villiform structures while the surface epithelium is free of atypia. The architecture should be largely normal with, at the most, minimal gland crowding. Surface maturation is present (Figure 1B).

Low grade dysplasia – The most important feature of low grade dysplasia is cytologic atypia extending to the mucosal surface and either minimal or absent surface maturation. Severe architectural distortion is not a feature, though mild gland crowding with decreased intervening lamina propria can be seen. Mitoses may be increased but no atypical forms should be seen. Inflammation is usually minimal. One important note: although cytologic atypia is a key finding, nuclear polarity is preserved. Loss of polarity - where the nucleus is tilted, rounded, or horizontal to the basement membrane - is associated with higher grade lesions (Figure 1C).

High grade dysplasia – The cytologic changes are severe with markedly enlarged nuclei at the surface, pronounced pleomorphism, and at least focal loss of nuclear polarity. Surface maturation is lost. Mild to marked architectural distortion is a frequent finding, with crowded glands, loss of lamina propria, focal budding, and/or cribriform glands. There should be no evidence of invasion into the lamina propria. Mitoses are increased and atypical mitoses may be seen. Ideally inflammation is minimal or absent. If either the cytologic or architectural changes are severe and extensive, the diagnosis of high grade dysplasia can be made even if other features are only low grade in severity (Figure 1D).

Ideally surveillance endoscopy is performed in patients whose reflux symptoms are controlled, reducing the chance of inflammatory or reactive changes interfering with pathologic interpretation (67). Four quadrant biopsies should be obtained at a minimum of every 2 cm and submitted to pathology in separate containers. The surveillance intervals suggested by the 2008 ACG Guidelines (4) are dependent on the pathology results (Table 1).

Chemoendoscopy involves the application of chemicals that selectively react with and highlight various mucosal features, theoretically improving the detection of abnormalities (70-76). Methylene blue is absorbed by non-dysplastic intestinal-type epithelial cells theoretically helping to detect BE and target biopsies. However, metaanalysis found no significant difference in the detection rates of BE or dysplasia between methylene blue directed biopsy and a standard 4-quadrant approach (74). Additonally, there is some evidence that methylene blue induces DNA damage in BE (77,78). Lugol’s solution is absorbed by glycogencontaining squamous epithelial cells and helps identify the squamocolumnar junction after eradication therapy, which is helpful in the recognition of residual columnarcell islands (75). Indigo carmine dye is combined with magnification endoscopy to distinguish mucosal pit patterns - round, reticular, villous, and ridged (68). While there is good association of certain patterns with intestinal metaplasia (76), it has not been shown to increase the detection of dysplasia beyond that of high-resolution endoscopy (9).

Electronic chromoendoscopy includes optimal band imaging which involves postprocessing to accentuate the contrast between columnar and squamous epithelia (79) and narrow band imaging (NBI) which uses optical filters to highlight vascular patterns on the mucosal surface (80). While studies show good correlation of vascular patterns identified by magnified NBI with BE and high grade dysplasia (80,81), prospective studies comparing the actual diagnostic yield of NBI to standard endoscopy have had mixed results (82- 84). A comparison of NBI to high resolution white light endoscopy showed no significant difference in the detection of BE or dysplasia (84).

Autofluorescence imaging utilizes differences in the endogenous fluorophores found in normal and neoplastic epithelia (68). While the technique has good sensitivity for the detection of high grade dysplasia, studies have shown poor specificity with false positive rates up to 81% (85-87). An analogous process recently described by Bird-Lieberman et al. utilizes a fluorescently labeled wheat germ derived lectin that binds to surface glycans of normal esophageal epithelial cells. Expression of these glycans is decreased or lost during neoplastic progression, so potentially premalignant or malignant regions are highlighted by a negative staining pattern (88). The potential applications are intriguing, but it has yet to be applied in vivo or prospective clinical trial.

Magnifications exceeding 1,000× can be achieved in real time using confocal laser endomicroscopy, allowing for analysis of the crypt architecture and capillaries during endoscopic examination. A few initial studies have shown accuracy rates above 85% in detecting high-grade dysplasia (89,90), fused glands indicating neoplasia with a sensitivity of 80%, and good interobserver agreement (91).

Light scattering spectroscopy and diffuse reflectance spectroscopy use algorithms to analyze light scattered back to the sensing device by the tissue. This spectroscopic information has been able to distinguish neoplastic from non-neoplastic tissue with both good sensitivity and specificity (92,93) in a few small trials. Optical coherence tomography uses variations in the reflectance of nearinfrared light from different tissues to create a highresolution cross-sectional image of the mucosa (94). One study has shown excellent sensitivity (97%) and specificity (92%) in the recognition of BE without dysplasia (95) while another showed good sensitivity (83%) and specificity (75%) in identifying high grade dysplasia (96).

Other biomarkers which test for DNA abnormalities have been evaluated in cross-sectional or retrospective studies. The detection of aneuploidy, increased tetraploidy, and loss of heterozygosity (LOH) for chromosome 17p in patients with no dysplasia or low grade dysplasia on biopsy has been shown to have good predictive value for neoplastic progression, but added little information when high grade dysplasia was detected (100-104). These studies utilized flow cytometry to detect DNA content abnormalities in fresh frozen tissue, which may not be practical in clinical practice. Fluorescence in situ hybridization can theoretically be used to detect these same abnormalities in fixed tissue and most initial studies show promising results (104-108).

Biomarker panels - including detection of chromosomal abnormalities (aneuploidy/tetraploidy, 17p LOH, 9p LOH) or tumor-suppressor gene-methylation patterns - have also been good indicators of progression risk in initial studies (109-111). One methylation-based panel, applied retrospectively, even identified patients who progressed to high grade dysplasia two years before histologic changes were seen (111).