Substantial Interobserver Agreement in the Diagnosis of Dysplasia in Barrett Esophagus Upon Review of a Patient's Entire Set of Biopsies

Marcela A. Salomao, Dora Lam-Himlin, Rish K. Pai

Research output: Contribution to journalArticle

2 Scopus citations


The pathologic diagnosis of dysplasia in Barrett esophagus (BE) suffers from interobserver disagreement. Many of the studies demonstrating disagreement in the diagnosis of dysplasia have pathologists review individual biopsy slides in isolation. To more closely mimic daily practice, 3 pathologists reviewed hematoxylin and eosin slides made from 549 individual biopsy jars obtained from 129 unique patients with a diagnosis of BE. Each pathologist reviewed the entirety of a given patient's biopsy material. The grade of dysplasia present in each biopsy jar was given as well as an overall highest grade of dysplasia from the patient's entire set of biopsies. The interobserver agreement in the diagnosis of dysplasia per biopsy jar and per patient's set of biopsies was measured by Fleiss κ statistic for multiple raters. The κ values for each diagnosis was higher in the per patient analysis compared with the per biopsy jar analysis indicating that pathologists are more likely to agree on the overall grade of dysplasia compared with the grade in an individual biopsy jar. In the per patient analysis, the interobserver agreement in the diagnosis of nondysplastic BE and high-grade dysplasia were substantial (κ=0.66; 95% confidence interval [CI], 0.56-0.76 and κ=0.76; 95% CI, 0.66-0.86, respectively). The interobserver agreement in the diagnosis of low-grade dysplasia (LGD) was fair (κ=0.31; 95% CI, 0.21-0.42). When LGD and high-grade dysplasia were collapsed into 1 category of positive for dysplasia, the interobserver agreement in the per patient analysis remained substantial (κ=0.70; 95% CI, 0.60-0.80), suggesting that much of the disagreement in LGD is not due to lack of recognition of dysplastic Barrett's mucosa, but rather the degree of dysplasia. These results indicate that pathologists can reliably distinguish between nondysplastic BE and dysplastic BE when a patient's entire set of biopsies is reviewed as a group. When second opinions are obtained, all available slides from that endoscopic procedure should be sent for review.

Original languageEnglish (US)
Pages (from-to)376-381
Number of pages6
JournalAmerican Journal of Surgical Pathology
Issue number3
StatePublished - Jan 1 2018



  • Barrett esophagus
  • high-grade dysplasia
  • interobserver agreement
  • low-grade dysplasia

ASJC Scopus subject areas

  • Anatomy
  • Surgery
  • Pathology and Forensic Medicine

Cite this