Observer variation in histopathological diagnosis and grading of cervical intraepithelial neoplasia.

To assess the variability among histopathologists in diagnosing and grading cervical intraepithelial neoplasia eight experienced histopathologists based at different hospitals examined the same set of 100 consecutive colposcopic cervical biopsy specimens and assigned them into one of six diagnostic categories. These were normal squamous epithelium, non-neoplastic squamous proliferations, cervical intraepithelial neoplasia grades I, II, and III, and other. The histopathologists were given currently accepted criteria for diagnosing and grading cervical intraepithelial neoplasia and asked to mark their degree of confidence about their decision on a visual linear analogue scale provided. The degree of agreement between the histopathologists was characterised by kappa statistics, which showed an overall poor agreement (unweighted kappa 0.358). Agreement between observers was excellent for invasive lesions, moderately good for cervical intraepithelial neoplasia grade III, and poor for cervical intraepithelial neoplasia grades I and II (unweighted kappa 0.832, 0.496, 0.172, and 0.175, respectively); the kappa value for all grades of cervical intraepithelial neoplasia taken together was 0.660. The most important source of disagreement lay in the distinction of reactive squamous proliferations from cervical intraepithelial neoplasia grade I. The histopathologists were confident in diagnosing cervical intraepithelial neoplasia grade III and invasive carcinoma (other) but not as confident in diagnosing cervical intraepithelial neoplasia grades I and II and glandular atypia (other). Experienced histopathologists show considerable interobserver variability in grading cervical intraepithelial neoplasia and more importantly in distinguishing between reactive squamous proliferations and cervical intraepithelial neoplasia grade I. It is suggested that the three grade division of cervical intraepithelial neoplasia should be abandoned and a borderline category introduced that entails follow up without treatment.