Interobserver variability in the diagnosis of canine hepatoid gland tumors

Abstract

The reported prevalence of canine hepatoid gland tumor (HGT) varies widelyowing mainly to the lack of well-defined diagnostic criteria and poor interobserver agreement. The aim of the present study was to improve the level of diagnostic agreement among pathologists in canine HGT. Five pathologists diagnosed and classified morphological features in 57 cases of canine HGT and, based on their reports, diagnostic algorithms were devised using recursive partitioning analysis. The proportion of diagnoses of malignant hepatoid neoplasia among the five pathologists ranged from 26.3 to 50.9%. Interobserver diagnostic agreement was classified as fair (κ=0.54) but improved to good (κ~0.65) following application of two novel diagnostic algorithms based on histomorphological features as sebaceous differentiation, mitotic count, atypical mitosis and cellular atypia. This study has demonstrated that interobserver agreement in the diagnosis of canine HGT could be improved using novel algorithms. Further analyses are warranted to validate the proposed classification systems applying a higher sampling of canine HGTs.