Reliability of endometrial cancer diagnoses in a Swedish Cancer Registry--with special reference to classification bias related to exogenous estrogens.

In this study, various qualitative aspects of endometrial cancer diagnoses were critically evaluated regarding cases included in the Swedish Cancer Registry from the Uppsala Health Care Region. By comparing the number of such cases in the registry with cases reported directly from the departments of pathology and of gynecological oncology in the same region, it was found that approximately 5% of all incident cases had not been notified to the registry. An independent histopathological review of the original specimens showed that almost 9% of the cases did not represent an endometrial neoplastic lesion, the majority of these being reclassified as uterine sarcoma. The review also revealed the problem of a diagnostic bias that might arise in connection with estrogen exposure, in that a significantly higher proportion of the cases observed in a cohort of women who had received estrogen prescriptions were reclassified as a premalignant endometrial lesion than of the cases from the background population without estrogen exposure (33% versus 10%). Additional independent reviews of cases showing discordant diagnoses in the primary review indicated variability in diagnostic criteria among pathologists. It is concluded that when cancer registry data are employed in epidemiological studies of endometrial cancer, the use of additional sources of case recruitment is desirable in order to obtain a complete material; also that an independent histopathological review is necessary to standardize diagnostic criteria and thereby to avoid a classification bias.