Subtyping intestinal metaplasia in patients with chronic atrophic gastritis: an interobserver variability study.

Incomplete gastric intestinal metaplasia (GIM) is associated with an increased risk of gastric cancer. We aimed to examine the interobserver variability of GIM subtyping (incomplete vs complete) in histological diagnosis of patients with chronic atrophic gastritis and to identify factors with potential impact on agreement. Nine international gastrointestinal expert pathologists assessed 46 cases with complete, incomplete or mixed-type GIM on scanned haematoxylin and eosin (H&E)-stained slides. Results were compared with the consensus diagnosis driven by two experts. Interobserver variability was evaluated by kappa statistics. Focusing on the predominant pattern, the agreement between each observer and the consensus diagnosis ranged from 78% to 98%. The level of agreement was moderate to almost perfect (weighted kappa=0.464-0.984). The participating pathologists reached substantial overall agreement (Fleiss' kappa=0.716, 95% confidence interval 0.677-0.755). Misclassification with potential impact on clinical decision making occurred in 5.7% of case ratings. The pattern of GIM (pure GIM versus mixed-type GIM) differed significantly between cases with high and low agreement (p=0.010), while the number of biopsy pieces per sample and the portion of mucosal surface involved by GIM did not. Pathologists who apply subtyping in daily routine performed better than those who do not (p=0.040). In conclusion, subtyping GIM on H&E-stained slides can be achieved satisfactorily with high interobserver agreement. The implementation of GIM subtyping as a risk stratifying tool in current practice guidelines by the European Society of Gastrointestinal Endoscopy (ESGE) and the American Gastroenterological Association (AGA) carries a low rate of misclassification, at least among gastrointestinal expert pathologists.