Poor Interobserver Agreement in the Distinction of High-Grade Dysplasia and Adenocarcinoma in Pretreatment Barrett's Esophagus Biopsies

OBJECTIVE: Grading Barrett's dysplasia at the lower end of the metaplasia-dysplasia spectrum (negative, indefinite, and low-grade dysplasia) suffers from poor interobserver agreement, even among gastrointestinal pathologists. Data evaluating interobserver agreement in Barrett's mucosal biopsies with changes at the upper end of the dysplasia spectrum (high-grade dysplasia, intramucosal, and submucosal adenocarcinoma) have not been published. The accurate categorization of pretreatment biopsies drives therapeutic decision making, but if the diagnostic distinction between cancer and high-grade dysplasia in Barrett's biopsies is inconsistent, then the use of these diagnoses to make management decisions is suspect. To this end, our aim was to assess interobserver reproducibility among a group of gastrointestinal pathologists in the interpretation of preresection biopsies. METHODS: All study pathologists agreed upon the histologic criteria distinguishing four diagnostic categories, including high-grade dysplasia; high-grade dysplasia with marked distortion of glandular architecture, cannot exclude intramucosal adenocarcinoma; intramucosal adenocarcinoma; and submucosally invasive adenocarcinoma. The histologic criteria were used to independently review preresection biopsies from 163 consecutive Barrett's esophagus patients with at least high-grade dysplasia who ultimately underwent esophagectomy. Reviewers recorded the specific histologic criteria used to categorize each case and Kappa statistics were calculated to assess interobserver agreement. RESULTS: Using kappa statistics, the overall agreement was only fair (K = 0.30). Agreement for high-grade dysplasia was moderate (K = 0.47), while agreement for high-grade dysplasia with marked architectural distortion, cannot exclude intramucosal adenocarcinoma and intramucosal adenocarcinoma were only fair (K = 0.21 and 0.30, respectively) and agreement for submucosal adenocarcinoma was poor (K = 0.14). CONCLUSIONS: The overall poor interobserver reproducibility among gastrointestinal pathologists who see a high volume of Barrett's cases calls into question treatment regimens based on the assumption that high-grade dysplasia, intramucosal adenocarcinoma, and submucosal adenocarcinoma can reliably be distinguished in biopsy specimens.