Cytokeratin 19 Immunostain Reduces Variability in Grading Epithelial Dysplasia of the Non-Keratinized Upper Aerodigestive Tract Mucosa

Epithelial dysplasia (ED) grading systems are criticized for low reliability. The effects of diagnostic tests or stains on reducing examiner variability in grading ED of the upper aerodigestive tract have not been investigated. The study aim was to examine the effectiveness of cytokeratin 19 (K19) immunostain on enhancing inter and intraexaminer reliability of ED grading and to reiterate the relation of K19 positivity with epithelial keratinization. The study sample consisted of 122 paraffin blocks that fulfilled the inclusion criteria. Each paraffin block had three sections cut: one immunostained for K19 and two for hematoxylin and eosin stain (H&E). Each examiner graded the study sample in six rounds; three using H&E stain only and three using paired K19–H&E stains. The study examiners were three American-Board certified practicing oral pathologists. The results were analyzed using Krippendorff’s alpha, ROC curve, Chi square test and binary logistic regression. Upon the use of paired K19–H&E stains the results showed that the intraexaminer reliability coefficients of grading were improved from 0.70, 0.69, 0.78 to 0.73, 0.88, 0.91 for examiners 1, 2, and 3 respectively. Reliability coefficients for inter-examiners improved from 0.55 to 0.73 (Krippendorff alpha). The accuracy of identifying the diseased cases (high-grade dysplasia) increased from 0.82 to 0.94 (ROC curve). Binary logistic regression revealed that K19 positivity is negatively associated with hyperkeratinization of surface epithelium (P = 0.001). To conclude, for grading non-keratinized epithelial dysplastic lesions of the upper aerodigestive tract, paired K19–H&E stains proved to reduce inter and intra-examiner variability by highlighting the extension of dysplastic epithelial cells within epithelial thickness, thus identifying the involved epithelial third and assigning a more reliable and better reproducible grade.