1. ¹Department of Anatomic Pathology, Cleveland Clinic Foundation, Cleveland, OH, USA
2. ²Department of Quantitative Health Sciences, Cleveland Clinic Foundation, Cleveland, OH, USA

Correspondence: Professor MP Bronner, MD, Department of Anatomic Pathology, Cleveland Clinic Foundation, L-25, 9500 Euclid Avenue, Cleveland, OH 44195, USA. E-mail: bronnem@ccf.org

Received 29 June 2009; Revised 31 August 2009; Accepted 31 August 2009; Published online 16 October 2009.

Abstract

Dysplasia is the gold standard biomarker of cancer risk in Barrett's esophagus, but its diagnosis remains difficult. This is due in part to its multitude of histological appearances. One aspect receiving little attention concerns gastric-type Barrett's dysplasia, which is distinctly different from the well-established intestinal variant. Recognition of gastric-type dysplasia and development of separate grading criteria are required. The prevalence, diagnostic criteria, and natural history of gastric-type Barrett's dysplasia were systematically evaluated in 1854 endoscopic biopsies from a cohort of 200 consecutive Barrett's dysplasia patients. Goblet cells were present in all cases, confirming the utility of this defining feature of Barrett's esophagus. The prevalence of Barrett's gastric-type dysplasia was 15% at the patient level (30 of 200 patients) and 20% at the biopsy level (166 of 852 dysplastic biopsies). Gastric-type dysplasia uniformly showed non-stratified, basally oriented nuclei as the major criterion for distinguishing it from intestinal-type Barrett's dysplasia. As such, loss of nuclear polarity, as the most objective criterion to distinguish intestinal-type low- and high-grade dysplasia, cannot be applied to gastric-type dysplasia. Rather, discriminatory features included increased nuclear size with a high-grade dysplasia cutoff by receiver operating characteristic (ROC) analysis approximating 3–4 times the size of a mature lymphocyte, providing an optimal sensitivity, specificity, and area under the curve of 0.78, 0.90, and 0.90 (95% CI: (0.87, 0.93)), respectively. Crowded, irregular glandular architecture (P<0.001) was more common in high-grade lesions (P<0.001), as was eosinophilic and oncocytic cytoplasm relative to the mucinous cytoplasm (P<0.001), prominent nucleoli (P<0.001), mild nuclear pleomorphism (P<0.001), and villiform architecture (P<0.001). During follow-up, 64% (7 of 11) of patients with pure gastric and 26% (5 of 19) with mixed gastric and intestinal dysplasia underwent neoplastic progression. The recognition of Barrett's gastric-type dysplasia and use of the proposed grading criteria should promote better diagnostic classification of the Barrett's neoplastic spectrum.