Interobserver variation in the diagnosis of adenoma malignum (minimal deviation adenocarcinoma) of the uterine cervix

To examine the interobserver agreement level of the histological diagnosis of adenoma malignum (ADM), 52 proliferative endocervical glandular lesions were evaluated independently by four observers (A to D), each of whom is in charge of gynecological pathology at a different hospital. The correlation of diagnosis by each observer with patient outcome was also examined for 19 of these lesions. When the diagnoses were categorized into benign lesions including hyperplasias, ADM, and common types of adenocarcinoma, consistent diagnoses among all observers were achieved for only 12 lesions (23%), with a slight level of interobserver agreement (kappa=0.115). The points of disagreement were as follows: (i) whether proliferative endocervical glandular lesions preserving lobular structures were diagnosed as benign or as ADM; and (ii) whether proliferative endocervical glandular lesions with a discrete area of obvious adenocarcinoma were diagnosed as ADM or as common-type adenocarcinoma. The mortality rates of patients with ADM diagnosed by observers A, B, C, and D were 60% (3 of 5), 25% (3 of 12), 14% (1 of 7), and 13% (2 of 15), respectively. Therefore, ADM diagnosed by observers A and B was frequently lethal, whereas ADM diagnosed by observers C and D was mostly non-lethal and might contain benign lesions. The diagnosis of ADM covered various spectra of proliferative endocervical glandular lesions among the observers. Disagreement in the diagnosis was suggested to derive largely from the absence of consensus criteria for differential diagnosis among benign hyperplastic lesions, ADM, and common adenocarcinoma, and from differences in the observers' interpretations about cellular atypia and invasion.