Three systems have been proposed to classify endometrial carcinoma precursors, but it is currently unclear which system best predicts cancer risk and is most reproducible. The optimal surrogate endpoint for endometrial carcinoma is therefore unknown. The pathologic diagnosis of endometrial hyperplasia (EH) includes most suspected immediate precursors and many mild, highly reversible proliferations. We propose an exploratory study to assess the feasibility of investigating EH as a source of an endometrial carcinoma surrogate endpoint.

Within a large, population-based health care plan, we will identify 'cases' (i.e., women who were diagnosed with EH at least two years before being diagnosed with endometrial carcinoma) through a computerized search of plan databases. We will retrieve the slides from the matching biopsy and hysterectomy on which carcinoma was diagnosed. Women ages 40 or older who were plan members and received a biopsy or curettage diagnosis of EH between 1970 and 2002 will be eligible to be a case.

We will perform an initial histologic review of cases' index biopsy slides to assess two types of misclassification known to affect the diagnosis of EH: a) false-negative endometrial carcinoma (i.e., prevalent carcinoma t the time of EH diagnosis) and b) false-positive EH (i.e., a benign, non-hyperplastic lesion). From cases' physical records and linked computer records, we will collect data on histopathologic classification of EH lesions and subsequent carcinomas; descriptive data (e.g., patient weight, parity, and menopausal status); and a summary of relevant treatments and follow-up procedures (e.g., hormone therapy or additional clinical procedures).

This pilot study is necessary before launching a case-control study (nested within the retrospective cohort of eligible plan women who had at least one endometrial biopsy during the study period) to explore the risk of endometrial carcinoma after a diagnosis of EH. That study will compare cases with women with EH matched on important clinical criteria who were followed in a similar fashion but did not develop carcinoma. Criteria for defining the sampling populations, sampling frames, statistical power, and sample size cannot be made until we know a) how many potential cases are 'true EH' (i.e., free of the misclassification described above) and b) the distribution of histopathologic, demographic, and clinical data among the 'true EH' cases. Selection of controls is the most costly and time consuming aspect of the case-control study and these pilot data will permit us to approach this task with a great deal of useful information.