Background: EMR of Barrett's esophagus (BE)-related superficial neoplasms represents an efficacious staging modality. It also allows for better pathologic grading compared with mucosal biopsy specimens. However, the interobserver variation in the interpretation of EMR specimens has not been tested.
Objective: To evaluate consistency in the diagnosis of BE-related neoplasia on EMR specimens.
Design: Nine pathologists reviewed 25 esophageal EMR specimens and corresponding biopsy specimens independently. Each pathologist classified the cases as either non-neoplastic BE, low-grade dysplasia, high-grade dysplasia, intramucosal adenocarcinoma, or invasive adenocarcinoma. Interobserver concordance for both specimens from EMRs and biopsies was measured by intraclass correlation and Kendall's coefficient of concordance. The proportion of agreement was also calculated for each specimen and compared for EMR and biopsy by using the Wilcoxon signed rank test.
Setting: Teaching hospitals.
Patients: Twenty-five patients who underwent EMR for BE-related neoplasia.
Results: The intraclass correlation and the Kendall's coefficient for the 25 biopsy specimens was 0.938 (95% CI 0.880-0.965) and 0.677, respectively; for the 25 EMRs, these were significantly improved, at 0.977 (95% CI 0.957-0.987) and 0.831, respectively. In addition, the proportion of agreement for EMR specimens was significantly better compared with biopsy specimens (P = .015).
Conclusions: Interobserver agreement of BE-related neoplasia on EMR specimens is significantly higher compared with biopsy specimens. The results may relate to the larger tissue sampling compared with biopsy specimens and the ability to evaluate mucosal landmarks, such as double muscularis mucosae. Thus, we suggest that EMRs, in addition to being a staging and therapeutic procedure, improve diagnostic consistency.