Greater interobserver agreement by endoscopic mucosal resection than biopsy samples in Barrett's dysplasia

BACKGROUND & AIMS: Endoscopic mucosal resection (EMR) is an important diagnostic, staging, and therapeutic tool for patients with Barrett's esophagus (BE)-associated neoplasia. We analyzed the histopathologic characteristics of specimens collected during EMR compared with biopsy specimens from patients with BE and assessed interobserver variability in pathologists' assessment of EMR and biopsy specimens. METHODS: We evaluated EMR (n = 251) and biopsy (n = 269) specimens collected from patients with BE at 2 tertiary referral centers. A detailed histologic analysis was performed for each EMR and biopsy specimen to determine the grade of dysplasia, depth of the specimen, proportion of specimen with dysplasia, and quality of samples. Interobserver agreement for both biopsy and EMR specimens (among 4 experienced pathologists) was calculated by using kappa statistics. RESULTS: Histologic analysis showed that submucosa was present in the majority of EMRs, compared with biopsy specimens (88% vs 1%, P < .0001). Almost all biopsy specimens (99%) included lamina propria. However, the muscularis mucosa was observed in only 58% of biopsy specimens. For both EMR and biopsy specimens, the highest grade of dysplasia comprised < or =25% of the total area in >50% of the specimens. Interobserver agreement on the diagnosis of dysplasia was significantly greater for EMR specimens than biopsy specimens (low-grade dysplasia, 0.33 vs 0.22, P < .001; high-grade dysplasia, 0.43 vs 0.35, P = .018). CONCLUSIONS: Submucosa can be examined in most samples collected from EMR; the distribution of neoplasia is focal within biopsy and EMR specimens. There is more interobserver agreement among pathologists in the analysis of EMR samples than biopsy specimens for the diagnosis of dysplasia