Interobserver Agreement in the Phenotypic Classification of Barrett Dysplasia
Tze S Khor, Kamran Badizadegan, Andrew Bellizi, Ian Brown, Hiroshi Fujita, Hye-Seung Han, Priyanthi Kumarasinghe, Anthony Mattia, Joseph Misdraji, Do-Youn Park, Amitabh Srivastava, Robert D Odze, Gregory Y Lauwers. Massachusetts General Hospital, Boston, MA; Brigham and Women's Hospital, Boston, MA; Envoi Pathology, Herston, QLD, Australia; PathWest Laboratory, Nedlands, WA, Australia; North Shore Medical Centre, Salem, MA; Pusan National University Hospital, Busan, Korea
Background: The two most common types of Barrett dysplasia are intestinal and (gastric) foveolar. Intestinal dysplasia resembles colon adenomas. Foveolar dysplasia is cytologically characterized by round/oval nuclei, pale eosinophilic cytoplasm and apical mucin cap. Mixed dysplasia combines features of both. Another recognized pattern, non-adenomatous dysplasia, is a rare high grade variant characterized by crowded glands, cuboidal cells, high N:C ratio, non-stratified round nuclei, open chromatin and prominent nucleoli. Some early data suggests that these different types of dysplasia have different natural histories and biologic characteristics. Our aim was to evaluate the interobserver agreement in phenotypic classification of dysplasia in BE based on H&E stained slides.
Design: 40 consecutive endoscopic mucosal resections (EMR) were reviewed independently by 12 pathologists. Each pathologist provided a diagnosis of low grade dysplasia, high grade dysplasia, intramucosal carcinoma or submucosal invasion and classified the dysplasia type as intestinal, non-adenomatous, foveolar or mixed. The intraclass correlation (ICC) and Kendall's co-efficient of concordance (KC) were calculated for overall diagnosis and phenotypic classification. Additionally, Kappa (Κ) values were evaluated for each subtype of dysplasia to determine if observers were better at agreeing on any a particular type of dysplasia.
Results: The ICC and KC for overall grade of neoplasia was 0.604 (95% CI 0.500-0.719) and 0.688, respectively. However, for dysplasia type, the ICC and KC fell to 0.124 (95% CI 0.066-0.220) and 0.230. Κ values for intestinal, foveolar, non-adenomatous and mixed were 0.249, 0.176, 0.06 and 0.01, respectively. In 13 cases (32.5%), one or more pathologists (up to 3) returned a response of "Unsure" further highlighting the difficulty in recognizing dysplasia phenotypes on H&E.
Conclusions: Moderate agreement was achieved for dysplasia. There was poor agreement for dysplasia type classification with best agreement achieved for intestinal type dysplasia. This suggests that more objective measures, such as immunohistochemistry, should be evaluated to better the agreement of dysplasia type.
Monday, March 19, 2012 9:30 AM
Poster Session I Stowell-Orbison/Surgical Pathology/Autopsy Awards Poster Session # 112, Monday Morning