Background Malignant transformation risk in oral epithelial dysplasia (OED) is currently determined by histological assessment. The subjectivity of this approach has led to interest in identifying prognostic biomarkers. Tissue microarrays (TMA) can reduce the utilization of the finite resources of a pathological archive. However, the selectivity involved in TMA construction necessitates the need to ensure that individual cores are representative of the overall features of the donor specimen. We aimed to validate, for the first time, the use of the TMA technique in OED by using a novel virtual array technique.

Methods Sections from 38 cases of OED were stained with H&E and 6 immunohistochemical (IHC) biomarkers. All were then digitally scanned. Virtual cores were generated by image capturing a 0.6mm² area of the IHC slide that corresponded to the same dysplastic area marked on the H&E slide. Two trained blinded observers scored both whole slides and virtual cores independently. The degree of reliability in scores between the individual raters and between virtual TMA cores and slides was assessed using both interclass correlation coefficient (ICCC) and weighted κ statistics.

Results Excellent inter-observer reliability was demonstrated with all the immunohistochemical markers. ICCC ranged from 0.67−1.0 and κ scores >0.8. There was also a high reliability in the scores between whole slides and virtual TMAs, with ICCC of between 0.66 and 0.89 for the 6 markers.

Conclusions This study validates the use of TMAs in OED using a variety of biomarkers. We also report a novel method for achieving this using a novel virtual-array technique.