Interobserver Variability in the Diagnosis of Crypt Dysplasia in Barrett's Esophagus
D Coco, J Goldblum, J Hornick, GY Lauwers, E Montgomery, A Srivastava, H Wang, RD Odze. Brigham & Women's Hospital, Boston, MA; Cleveland Clinic, Cleveland, OH; Massachusetts General Hospital, Boston, MA; John Hopkins Hospital, Baltimore, MD; Dartmouth Medical Center, Lebanon, NH; Beth Israel Deaconess Medical Center, Boston, MA
Background: Recent studies suggest that dysplasia in Barrett's esophagus (BE) begins in the crypt bases [("crypt dysplasia") (CD)] and then extends to the surface epithelium with progression. The aim of this study was to evaluate the interobserver reproducibility of CD among gastrointestinal pathologists with research interest in BE. Diagnostic reproducibility is important if future biology and natural history studies are to be performed accurately.
Design: Glass slides of 40 routinely processed H&E stained mucosal biopsies of BE and related neoplasms [(10 BE without dysplasia, 9 CD, 10 low-grade dysplasia (LGD), 9 high-grade dysplasia (HGD), 2 intramucosal adenocarcinoma (IMCA)] diagnosed by the index pathologist) were sent to 5 other GI pathologists for a blinded evaluation of the grade of neoplasia using pre-determined criteria. A second review was performed on the original 40, plus an additional 23, cases after a meeting in which the criteria were modified. Analysis was performed by Kappa (K) statistics.
Results: The K value for interobserver agreement of all cases was moderate (K=0.44). The degree of agreement was highest for IMCa (K=0.65) and BE without dysplasia (0.57), and lowest for LGD (K=0.31). Notably, no significant differences were observed in the degree of reproducibility in the diagnosis of CD (K=0.44) compared to LGD (0.31) or HGD (0.46). Regarding CD cases, all 5 observers agreed in 50% of cases, and 4 in 69% of cases. In general, when there was a disagreement with the index pathologist regarding a diagnosis of CD (N= 26/80 readings), the majority diagnosed either LGD or HGD (62%) rather than BE without dysplasia (23%). After a second review, K values for all grades, except Barrett's without dysplasia and IMCa, increased.
Conclusions: The level of agreement for a diagnosis of CD is similar to that observed for other grades of dysplasia, both in this and in previous interobserver reproducibility studies in BE. Thus, similar to LGD and HGD, further biology and natural history studies may be performed to elucidate the role of CD in the pathogenesis of cancer in BE.
Wednesday, March 24, 2010 9:30 AM
Poster Session V # 68, Wednesday Morning