Observer variability in the Goseki grouping of gastric adenocarcinoma in resection and biopsy specimens.

Abstract

The Goseki grouping of gastric adenocarcinoma has been suggested as a possible prognostic factor. In those centres where it is used, it may be valuable to assess the Goseki grouping of a tumour on the initial diagnostic biopsy as well as on the resection specimen since it may in theory influence management. We examined the robustness of Goseki grouping of gastric adenocarcinoma in representative sections from resection and biopsy specimens in order to assess the consistency of agreement among a group of pathologists. A single representative block from 100 gastric resection specimens was studied using a haematoxylin and eosin and mucin (alcian blue/periodic acid-Schiff) stain. These were circulated in batches to members of a group of 12 pathologists who each completed a simple proforma confirming the presence of carcinoma and assigning a Goseki group. In a second circulation the diagnostic biopsy specimen taken prior to resection was examined in the same way. This allowed comparison of the Goseki group of the biopsy and resection specimens. In both studies kappa statistics showed good agreement on tubular differentiation of the carcinoma, but only moderate agreement for the intracellular mucin production, resulting in moderate agreement for the final Goseki group. Correlation between the Goseki group assigned on the biopsy and resected specimens was seen in 62% of the cases. However, the reproducibility was low (kappa 0.375). The Goseki grouping of resected gastric adenocarcinoma is reproducible and can be used in prognostication. Goseki grouping of biopsy specimens is of limited value in predicting the Goseki group assigned to the resected carcinoma.