Background: Oral epithelial dysplasia (OED) remains the principal histopathological marker for stratifying malignant risk in oral potentially malignant disorders. Current World Health Organisation (WHO) three-tier and binary grading systems underpin diagnosis and clinical management, yet both demonstrate modest reproducibility and restricted prognostic discrimination, particularly within the mild-to-moderate spectrum. A major under-recognised challenge in routine practice is failure to identify dysplasia at all, especially in lesions where architectural disturbance predominates over cytological atypia.
Methods: This position paper from COST Action INTERCEPTOR Working Group 2 synthesises evidence from reproducibility studies, feature-based prognostic models, and computational pathology analyses to propose that explicit morphological pattern recognition can enhance existing grading systems.
Results: Recognisable patterns including basaloid (conventional), keratinising (differentiated), verrucous, papillary, mixed, and HPV-associated, provide intuitive anchors for dysplasia recognition and support grading within pattern rather than replacement of established frameworks. Feature-based prognostic models show architectural traits e.g. bulbous rete ridges and cohesion loss etc. can predict transformation more effectively than grade alone.
Conclusions: We propose a pattern-anchored, grade-retaining framework as a hypothesis-driven refinement of current practice, compatible with WHO grading and requiring formal validation before wider adoption. This approach addresses real-world diagnostic constraints, particularly in non-specialist settings, and provides a pragmatic, testable foundation for potentially improving risk communication.