ABSTRACT
Aim of the current study was to evaluate the inter-observer agreement between pathologists in the diagnosis of celiac disease (CD), in the qualified context of a multicenter study. Biopsies from the "PreventCD" study, a multinational- prospective- randomized study in children with at least one-first-degree relative with CD and positive for HLA-DQ2/HLA-DQ8. Ninety-eight biopsies were evaluated. Considering diagnostic samples with villous atrophy (VA), the agreement was satisfactory (κ = 0.84), but much less when assessing the severity of these lesions. The use of the recently proposed Corazza-Villanacci classification showed a moderately higher level of agreement (κ = 0.39) than using the Marsh-Oberhuber system (κ = 0.31). 57.1% of cases were considered correctly oriented. A number of >4 samples per patient was statistically associated to a better agreement; orientation did not impact on κ values. Agreement results in this study appear more satisfactory than in previous papers and this is justified by the involvement of centers with experience in CD diagnosis and by the well-controlled setting. Despite this, the reproducibility was far from optimal with a poor agreement in grading the severity of VA. Our results stress the need of a minimum of four samples to be assessed by the pathologist.