p16INK4a Immunohistochemistry Improves Interobserver Agreement in the Diagnosis of Cervical Intraepithelial Neoplasia

Top Cited Papers
Publisher Website
Google Scholar
Abstract
It has been repeatedly shown that there is a substantial lack of interobserver reproducibility in the histologic diagnosis of cervical intraepithelial neoplasia (CIN), which might be improved by a more specific diagnostic biomarker. Cervical cancer and CIN, but not other cervical epithelia, express high levels of the cyclin-dependent kinase inhibitor p16INK4a, suggesting that staining for this marker could help to more precisely identify CIN in tissue sections and therefore reduce variation in interpretation of cervical lesions. To test this hypothesis, 194 cervical cone biopsy samples were selected from a routine histopathology laboratory. Two consecutive sections from each biopsy were stained with hematoxylin and eosin and with a p16INK4a-specific monoclonal antibody, respectively. Five experienced cervical pathologists examined the slides. The agreement in the diagnosis between pairs or groups of observers was calculated by kappa statistics. Significant discrepancies were observed in the diagnostic interpretation of hematoxylin and eosin-stained slides, particularly for low-grade lesions (kappa value 0.60 [95% confidence interval 0.58–0.63]). There was significantly better agreement in the interpretation of p16INK4a expression (kappa value 0.91 [95% confidence interval 0.84–0.99]). Expression of p16INK4a was restricted to CIN 2/CIN 3, CIN 1 associated with high-risk human papillomavirus, or cervical cancer. p16INK4a immunostaining allowed precise identification of even small CIN or cervical cancer lesions in biopsy sections and helped to reduce interobserver variation in the histopathologic interpretation of cervical biopsy specimens. Thus, p16INK4a immunohistochemistry can reduce false-negative and false-positive biopsy interpretation and thereby significantly improve cervical (pre)-cancer diagnosis.

This publication has 32 references indexed in Scilit: