Observer agreement on the grading of gastric atrophy

Assessment of lesser or doubtful degrees of gastric atrophy can be difficult, especially in the antrum, since well established criteria are lacking. At the Houston Working Party on Gastritis in 1994 a visual analogue scale was designed for the grading of histopathological parameters. This was done to promote uniformity in grading by acting as a reference. The purpose of the present study was to measure interobserver variation between pathologists familiar with the Houston visual analogue scale in a specifically selected set of biopsies from patients with lesser or doubtful degrees of atrophy.

Thirty cases with biopsies of the antrum and corpus from a long-term follow-up study on Helicobacter pylori gastritis comprised the current study material. The cases were selected from that study because there had been uncertainty or disagreement on the presence of gastric atrophy. The study set of haematoxylin and eosin (H & E) slides was circulated amongst gastrointestinal pathologists familiar with the visual analogue scale who were unaware of the source of the study set nor had any other clinical information. Interobserver variability was analysed using kappa statistics. The overall agreement for the grade of atrophy in antral biopsies was 0.461; the kappa value was 0.18 (95% confidence limits 0.12-0.24), which is considered poor agreement. The kappa value was nevertheless statistically significant (P < 0. 01). The overall agreement on the grade of atrophy in the corpus biopsies was apparently good (0.833), but the kappa which adjusts for chance agreement was only moderate (0.48; 95% confidence limits 0.42-0.55; P < 0.001).

The studied series comprised a self-selected sample in which there was doubt about the grade of atrophy and such a sample will produce lower kappa values than a random sample of gastric biopsies. The results nevertheless confirm that better guidelines and firm criteria are needed to properly diagnose and grade gastric atrophy. It is suggested that the use of two grades, low- and high-grade atrophy, akin to that in use for grading inflammatory bowel disorder (IBD)- associated dysplasia, could improve agreement. Furthermore optimal biopsy quality with full thickness mucosa and proper orientation appears important for grading gastric atrophy.