What levels of agreement can be expected between histopathologists assigning cases to discrete nominal categories? A study of the diagnosis of hyperplastic and adenomatous colorectal polyps

Aims: To assess the levels of agreement between histopathologists for a two -class nominal categorization process-the discrimination between hyperplast ic and adenomatous colorectal polyps. Methods: Fifty hyperplastic and 50 ad enomatous polyps received consecutively in the laboratory were categorized by nine histopathologists, and the level of agreement between all observers and the original diagnosis was assessed using kappa statistics. Results: F or the eight observers with 11 months or more experience in histopathology, there was a high level of agreement with kappa statistics ranging from 0.8 4 to 0.98. This process was performed rapidly with an average of 13 to 22 s econds spent on each case. One observer with only 6-weeks' experience of hi stopathology had a lower overall level of agreement with kappa statistics r anging from 0.46 to 0.54, but the performance on the later cases was much h igher. Conclusions: The level of agreement in the distinction between hyper plastic and adenomatous colorectal polyps is high among histopathologists w ith at least moderate amounts of experience in histopathology, The one virt ually naive observer showed a marked learning response during the study wit hout feedback on case outcome. This suggests that histopathologists are ver y reliable in assigning cases to distinct nominal categories and that learn ing of these processes occurs early in a histopathologist's career.