RESUMO
AIMS: Although effective clinical management of colorectal polyps detected through the National Health Service (NHS) Bowel Cancer Screening Programme (BCSP) is dependent on the quality of pathological diagnosis, there have been few attempts to formally evaluate inter-observer variability in histological assessment. The aim of this study was to examine the impact of inter-observer variability on the reported prevalence of prognostic features in a large series of screen-detected colorectal polyps. METHODS AND RESULTS: A retrospective series of 1329 screen-detected polyps (2008-10) was identified from computerized records at two histopathology departments participating in the NHS BCSP. Slides from a sample of 239 polyps were exchanged between centres for independent review and measurement of inter-observer (kappa) agreement. There were significant between-centre differences in the prevalence of polyps with high-risk histological features. Diagnostic review demonstrated good reliability with respect to the assessment of adenomatous change (κ = 0.83), excision margin status (κ = 0.74), high-grade dysplasia (0.61) and invasive malignancy (κ = 0.84). By contrast, there were significant inter-observer differences in the classification of villous lesions (0.18) despite recent efforts to standardize reporting practice. CONCLUSIONS: Inter-observer variability in the assessment of screen-detected colorectal polyps limits the prognostic value of histological subtyping and highlights the need for clarification of existing diagnostic criteria.
