A Multicentric European Study Testing the Reproducibility of the WHO Classification of Endometrial Hyperplasia With a Proposal of a Simplified Working Classification for Biopsy and Curettage Specimens

This study was designed to assess intraobserver and interobserver agreement in the diagnosis of 56 endometrial specimens by five European expert gynecologic pathologists using the WHO classification and to establish which histologic features are significantly associated with each classification category. The seven categories were simple hyperplasia, complex hyperplasia, atypical hyperplasia, well-differentiated adenocarcinoma, proliferative endometria, secretory endometria, and other. Slides were reviewed twice for diagnosis, with accompanying evaluation of a checklist of histologic features. These seven categories were eventually reduced to four and three for the purposes of data analysis. The four modified diagnostic categories consisted of hyperplasia (previously simple hyperplasia and complex hyperplasia), atypical hyperplasia, well-differentiated adenocarcinoma, and cyclical endometrium (previously proliferative, secretory, and other). The three diagnostic categories consisted of hyperplasia, endometrioid neoplasia (previously atypical hyperplasia and well-differentiated adenocarcinoma), and cyclical endometrium. Intraobserver and interobserver agreement was assessed using the percentage agreement and kappa statistics. The associations among the various histologic features and diagnoses was analyzed using multiple logistic regression to identify those features that were useful for distinguishing diagnostic categories. When using seven categories, kappa values ranged from 0.53 to 0.74 (percentage agreement, 61-79%) and from 0.33 to 0.59 (percentage agreement, 43-63%) for intraobserver and interobserver agreement, respectively. When using four categories, kappa values ranged from 0.68 to 0.73 (percentage agreement, 77-80%) and from 0.39 to 0.64 (percentage agreement, 54-73%) for intraobserver and interobserver agreement, respectively. When using three categories, kappa values ranged from 0.70 to 0.83 (percentage agreement, 80-89%) and from 0.55 to 0.73 (percentage agreement, 70-82%) for intraobserver and interobserver agreement, respectively. Data were analyzed in each diagnostic category. When using four or three diagnostic categories, the mean intraobserver and interobserver agreements varied less between categories and achieved higher values, with smaller 95% confidence intervals. The mean percentage agreement was lowest for complex hyperplasia and for atypical hyperplasia. For distinguishing cyclical endometrium versus hyperplasia, the useful histologic feature was glandular crowding. For hyperplasia versus atypical hyperplasia and for hyperplasia versus endometrioid neoplasia, the useful features were nuclear enlargement, nuclear pleomorphism, vesicular chromatin, and nucleoli, but of these, only nuclear pleomorphism achieved substantial mean intraobserver and interobserver agreements. For discriminating atypical hyperplasia from well-differentiated adenocarcinoma, the only useful feature was stromal alterations, which achieved only fair mean intraobserver and interobserver agreements. In summary, in endometrial biopsy or curettage specimens, the lack of agreement in the diagnoses of complex hyperplasia and atypical hyperplasia and the lack of reproducibility in the recognition of the histologic feature of stromal alterations to differentiate atypical hyperplasia from well-differentiated adenocarcinoma suggest that the histologic classification should be simplified by including a combined category for simple and complex hyperplasia, called hyperplasia, and a combined category for atypical hyperplasia and well-differentiated adenocarcinoma, called endometrioid neoplasia. Diagnoses of hyperplasia and endometrioid neoplasia are highly reproducible between observers from different institutions. Glandular crowding is the best histologic feature to differentiate cyclical endometrium from hyperplasia, whereas nuclear pleomorphism is the reproducible cytologic feature to differentiate hyperplasia from endometrioid neoplasia.