Grading oral epithelial dysplasia: analysis of individual features

Assessing epithelial dysplasia to predict malignant transformation remains problematic in many tissues because grading systems are poorly structured and individual features poorly defined. Dysplasia grading is criticised for lack of reproducibility and poor predictive value. Grading systems for upper aerodigestive tract dysplasia have evolved over several decades and are not supported by good outcome experimental data.

This study analysed the individual features of dysplasia in 86 oral dysplastic lesions and determined the reproducibility of scoring for each, and correlated them with other features and clinical factors using complex clustering analyses.

A uniform pattern of dysplasia was found in 37 lesions, focal dysplasia in 36 and in 13 lesions dysplasia formed complex discontinuous patterns. There was wide variation in reproducibility of scoring of individual features and many, including thickness, some types of rete morphology, basaloid cell anisonucleosis, basal dyscohesion, and dyskeratosis as deep single cells correlated with sub-sites. Rete morphology, type of keratinisation, hyperchromatism of the basaloid compartment, prickle cell anisonucleosis and extension down salivary ducts correlated with smoking. Conventional grading and oral intraepithelial neoplasia (OIN) grading by 'thirds affected' showed strong correlation overall but scores obtained with the OIN system tended to a higher grade at all sites except soft palate/fauces. There was poor correlation between the systems for moderate dysplasia and also severe dysplasia at some sites. Individual features could not be shown to cluster to form distinct patterns of dysplasia.

These variations may account in part for the lack of reproducibility and poor predictive value of the grading systems in current use and could inform the design of future grading systems.