Screening for uterine cancer.

Abstract

e18578 Background: Endometrial cancer is the most common gynecologic malignancy. Historically, endometrial cancer would present at an early stage without disseminated disease. Presently, in certain populations endometrial cancer presents at a late stage and carries a mortality rate similar or even greater to that of ovarian cancer. This increase is tied to the increasing prevalence of risk factors and worsening disparities. Owing to data and trends from decades long past, routine endometrial cancer screening is not recommended by any gynecologic or oncologic society. It is now time to revisit the discussion on endometrial cancer screening in the modern era of the disease. In our community with advanced stage and 60% 5-year mortality, endometrial biopsy is a reasonable approach for those at highest risk for disease and the worse outcomes. Methods: Using nationally available datasets on endometrial cancer (NHANES, PLCO), we analyzed the most common risk factors associated with endometrial cancer to develop a risk assessment nomogram. The variables were age, body-mass index, race, concomitant medication usage (specifically hormone use), race, and prior obstetrical history. We bootstrapped our risk calculations to exclude data from the most recent year and used those cases as our testing set to test the ability of the nomogram to correctly predict endometrial cancer in those patients with known disease. We also tested our nomogram against unpublished endometrial cancer data from the population of an Atlanta safety net hospital. Results: For each learning data set (ie the bootstrapped patients from the NHANES and PLCO data), the association with the candidate variables was as expected. The magnitude of the associations was also consistent with previously reported univariate analyses. In the population of interest, there were 342994 women with 1076 (67.4/100,000) uterine cancers over 5 yrs (2015-20). The incidence rate ratio of cancer with the presence of known risk factors already contained in the population EMR was approximately 6x and 15:1 (15 high risk women identified with 1 expected cancer). Conclusions: Our nomogram was able to accurately predict the presence of cancer in patients included in both the national databases as well as the local safety net hospital community. Using a 100% sensitivity goal yielded a PPV of 6.7%. With these promising retrospective results, we intend to apply our risk assessment model in the clinical setting with a prospective study design to predict asymptomatic endometrial disease in at-risk patients, selecting those for further clinical investigation including triage to soliciting symptoms, ordering ultrasound or even performing endometrial sampling using shared decision making. Further directions for research include the results and subsequent effects on stage at diagnosis and survival of these additional clinical investigations in asymptomatic women identified as high-risk by this nomogram.