Reproducibility of classification in non-squamous cell carcinomas of the uterine cervix.

OBJECTIVE: Reproducibility of histopathologic classification systems is of major importance for their utility in daily practice and in research. The reproducibility of histologic classification of non-squamous carcinomas (non-SCC) of the uterine cervix was evaluated, using population-based material from two 5-year periods in Norway.
METHODS: Histologic slides from 388 tumors were reviewed by three experienced pathologists and analyzed for inter- and intraobserver agreement on histological subtypes and grade.
RESULTS: Kappa values of inter- and intraobserver agreement were moderate to substantial for all major adenocarcinoma subgroups (endocervical, endometrioid, clear cell, or serous carcinoma), and fair to poor for mixed carcinomas and adenocarcinomas not otherwise specified (NOS). Interobserver agreement on villoglandular and adenosquamous carcinomas was poor, and the distinction of adenocarcinoma in situ from well-differentiated carcinoma proved difficult. Reproducibility of the high-risk subgroups of small cell and undifferentiated carcinomas was acceptable from a statistical point of view (kappa values >0.50). However, the authors agreed upon the diagnosis of small cell carcinomas and undifferentiated carcinomas only in 2/3 of these high-risk diagnoses. Patients with high-risk diagnoses showed significantly lower overall survival than patients with non-high-risk diagnoses (P < 0.001). This inferior survival was independent of whether the reviewers had agreed on the high-risk diagnosis or not.
CONCLUSION: Clinicians should be aware of the potential inconsistencies of histopathologic diagnoses. No histopathological classification system will ever be perfectly reproducible. Future histopathologic classification of the uterine cervix should emphasize the distinction between groups of particularly low or high prognostic risks.