Inter- and intra-observer variability in three grading systems for oral epithelial dysplasia

Management of oral premalignant lesions depends on clinical assessment and grading of oral epithelial dysplasia (OED), which remains one of the most important predictors of malignant potential. Numerous grading systems for OED exists with varying sets of assessing criteria and are largely considered subjective. The present study attempted to assess the inter- and intra-observer variability in three grading systems: Binary system, WHO (2005) and Ljubljana systems.

Histopathological grading of 63 cases of leukoplakia was performed by two oral pathologists and one general pathologist, who were blinded, using all the three grading systems at different time intervals and was repeated twice. Inter- and intra-observer variability was then evaluated by multivariate kappa analysis.

Inter-observer agreement in the two set of observations was found to be "slight" in WHO (k = 0.001 and 0.039), "slight" and "poor" in binary (k = 0.108 and -0.007), "poor" and "slight" in Ljubljana's (k = -0.027 and 0.106) grading systems. Intra-observer agreement ranged from "slight" to "fair" (k = 0.128 and 0.295) in WHO, "fair" to "moderate" (k = 0.224 and 0.420) in binary and "slight" to "fair" (k = 0.161 and -0.353) in Ljubljana's grading systems.

The binary system of classification proved to have an overall better inter- and intra-observer agreement. This study also showed better intra-observer agreement in all the grading systems as well as in individual histopathological parameters. Defining the individual parameters more objectively with the reproducible structuring of the grading systems and training of the pathologists would help reduce the variability in diagnosing dysplasia.