Observer variation in assessment of liver biopsies including analysis by kappa statistics

Abstract

In an observer variation study, six histopathologists independently coded 27 features and final diagnoses on the same 60 liver biopsies without knowledge of clinical data. Many features were shown by kappa statistics to have significantly high agreement, and differences were shown to be due in part to differences in definitions of particular features. Intraobserver variation was much lower than interobserver variation. Full agreement with respect to the final histopathologic diagnosis between the six observers occurred in only nine instances--all of these were cases of alcoholic liver disease. Some of the biopsies of acute hepatitis were confused with chronic active hepatitis, while others were mistaken for cholestasis, usually believed to have been related to drugs. Biopsies from patients with extrahepatic biliary obstruction were most commonly misdiagnosed as drug cholestasis. However, agreement was higher when the results were calculated in terms of observer pairs by comparing the final diagnosis of each observed with that obtained from each of the other five observers. The most striking finding here was the very high agreement rate (80%) between the two specialist pathologists compared with that obtained from the remaining fourteen pairs of observers who averaged an (43%) agreement rate.