Observer agreement in histopathological evaluation of oral epithelial dysplasia

grantor: University of Toronto'Objectives'. We studied observer agreement in diagnosing oral epithelial dysplasia, (OED). 'Methods'. Sixty-four slides were each examined twice by three oral pathologists. Information provided for each observation was altered in 50% of second examinations. The modal diagnosis was used as a gold standard to assess conformity. 'Results'. Pairwise interobserver weighted kappa (Kw) ranged from 0.68-0.83 with a group Kw of 0.74. Intraobserver Kw ranged from 0.82-0.96. Using a 3 point ordinal scale decreased Kw levels. Category specific simple kappa for normal, mild, moderate, severe and carcinoma-in-situ (CIS) were 0.51, 0.24, 0.26, 0.30, and 0.57. Information changing observer expectation significantly influenced diagnosis. 'Conclusions'. There was substantial interobserver consistency and almost perfect conformity when diagnosing OED. Agreement was best with CIS and lowest with mild dysplasia, cases. Observer bias was present; consensus and calibration exercises may be appropriate. The full 5 point ordinal scale was best when assessing OED. Referring clinicians should include accurate patient information with specimens.M.Sc