Abstract
Recent morphologic and molecular evidence suggests that dysplasia in Barrett esophagus (BE) begins in the bases of the crypts [crypt dysplasia (CD)] and progresses with time to involve the upper portions of the crypts and surface epithelium. The aim of this study was to evaluate the criteria and reproducibility of diagnosing CD among 6 gastrointestinal pathologists, all with research interest in BE. Six gastrointestinal pathologists evaluated 2 clinical study sets, the first consisting of 40 BE cases [BE: 10, CD: 9, low-grade dysplasia (LGD): 10, high-grade dysplasia (HGD); 9, and intramucosal adenocarcinoma; 2] and the second consisting of 63 cases (BE: 16, CD: 15, LGD: 15, HGD: 15, and intramucosal adenocarcinoma: 2), at least 4 months apart. In between evaluations, all of the pathologists met at 1 hospital (consensus conference) to review the areas of disagreement and establish more objective criteria. Overall, the level of agreement for all cases was moderate (kappa = 0.44), and the level of agreement did not change significantly after evaluation of the second study set. The highest levels of agreement were obtained for lesions at the low and high end of the spectrum (BE without dysplasia and HGD). Overall, the degree of agreement for CD was moderate after both the first and second study set review (kappa = 0.44 and 0.46, respectively). However, the degree of agreement for CD was higher than that obtained for LGD in both study sets. In the first study set, 4 or more pathologists agreed with the original CD diagnosis in 78% of cases, and this value did not change significantly after review of the second study set. The observers agreed that characteristic features of CD include the presence of unequivocal dysplastic cells, similar in appearance to traditional LGD, involving any part, or all of the length, of the crypt in the absence of intercrypt surface epithelial involvement. Rare cases of CD may show high-grade cytologic features composed of markedly enlarged nuclei with increased nuclear/cytoplasmic ratio, eosinophilic cytoplasm, irregular nuclear membranes, and loss of polarity. The findings in this study suggest that CD can be diagnosed reliably with a moderate level of interobserver agreement. Long-term and multi-institutional studies should be carried out to further determine the biological and clinical significance and natural history of CD in patients with BE.
