RESUMO
OBJECTIVE: To establish diagnostic accuracy and reproducibility of a diagnosis of cervical intraepithelial neoplasia 3 (CIN 3) in menopausal women on routinely stained hematoxylin and eosin (H&E) slides and compare it to slides processed for p16 and Ki-67. MATERIALS AND METHODS: Confirmed cases of CIN 3 and benign atrophic changes were reviewed independently by 4 pathologists. The samples were studied on separate occasions using H&E staining, p16, and Ki-67. Differences in sensitivity and specificity between reviewers or methods were tested for significance using the McNemar test, whereas differences in positive and negative predictive values were tested for significance using a marginal probability generalized linear model for agreement. RESULTS: Sensitivity was high for H&E (93.3%-100%) and Ki-67 (93.3%-100%) and lower for p16 (70.0%-90.0%). Intraobserver variability was also lower for p16 (76.7% vs 90.0%, although this difference was not statistically significant, p = .219). p16 agreement, however, for CIN 3 is significantly lower than that for atrophy (76.7% vs 97.4%, p = .018). CONCLUSIONS: Routine histopathologic diagnosis of CIN 3 in menopausal women is highly accurate and reproducible. Both H&E and Ki-67 are useful immunohistochemical stains in helping differentiate atrophy from high-grade cervical intraepithelial lesions in postmenopausal cervical biopsies. There may be more disagreement among readers using p16.