Background Endometrial tumors arise from a hormonally-responsive tissue. Defining subtypes by hormone receptors might better inform etiology and prediction of patient outcomes. We evaluated the potential role of tumor estrogen receptor (ER) and progesterone receptor (PR) expression to define endometrial cancer subtypes. Methods We measured semi-continuous ER and PR protein expression in tissue specimens from 360 endometrial primary tumors from the Nurses’ Health Study. To explore the impact of different definitions of marker positivity, we dichotomized ER and PR expression at different cut points in increments of 5% positive cells. Logistic regression was used to estimate associations between endometrial cancer risk factors, such as body mass index, with dichotomous ER or PR status. Reclassification statistics were used to assess whether adding dichotomous ER or PR status to standard prognostic factors of stage, grade, and histologic type would improve prediction. Results Compared to not being obese, obesity increased the odds of having an ER-positive tumor at cut points of 0–20% (maximum OR=2.92, 95% CI 1.34, 6.33) as well as the odds of having a PR-positive tumor at cut points of 70–90% (maximum OR=2.53, 95% CI 1.36, 4.68). Adding dichotomous tumor ER or PR status to the panel of standard predictors did not improve both model discrimination and calibration. Conclusion Obesity may be associated with greater endometrial tumor expression of ER and PR. Adding either marker does not appear to improve mortality prediction beyond the standard predictors. Impact Body mass index might explain some of the biological variation among endometrial tumors.

Developing a system of molecular subtyping for endometrial tumors might improve insight into disease etiology and clinical prediction of patient outcomes. High body mass index (BMI) has been implicated in development of endometrial cancer through hormonal pathways and might influence tumor expression of biomarkers involved in BMI-sensitive pathways. We evaluated whether endometrial tumor expression of 7 markers from BMI-sensitive pathways of insulin resistance could effectively characterize molecular subtypes: adiponectin receptor 1, adiponectin receptor 2, leptin receptor, insulin receptor (beta subunit), insulin receptor substrate 1, insulin-like growth factor 1 receptor, and insulin-like growth factor 2 receptor. Using endometrial carcinoma tissue specimens from a case-only prospective sample of 360 women from the Nurses' Health Study, we scored categorical immunohistochemical measurements of protein expression for each marker. Logistic regression was used to estimate associations between endometrial cancer risk factors, especially BMI, and tumor marker expression. Proportional hazard modeling was performed to estimate associations between marker expression and time to all-cause mortality as well as time to endometrial cancer-specific mortality. No association was observed between BMI and tumor expression of any marker. No marker was associated with time to either all-cause mortality or endometrial cancer-specific mortality in models with or without standard clinical predictors of patient mortality (tumor stage, grade, and histologic type). It did not appear that any of the markers evaluated here could be used effectively to define molecular subtypes of endometrial cancer.

Benign normal (NL), premalignant (endometrial intraepithelial neoplasia, EIN) and malignant (cancer, EMCA) endometria must be precisely distinguished for optimal management. EIN was objectively defined previously as a regression model incorporating manually traced histologic variables to predict clonal growth and cancer outcomes. Results from this early computational study were used to revise subjective endometrial precancer diagnostic criteria currently in use. We here use automated feature segmentation and updated machine learning algorithms to develop a new classification algorithm. Endometrial tissue from 148 patients was randomly separated into 72-patient training and 76-patient validation cohorts encompassing all 3 diagnostic classes. We applied image analysis software to keratin stained endometrial tissues to automatically segment whole-slide digital images into epithelium, cells, and nuclei and extract corresponding variables. A total of 1413 variables were culled to 75 based on random forest classification performance in a 3-group (NL, EIN, EMCA) model. This algorithm correctly classifies cases with 3-class error rates of 0.04 (training set) and 0.058 (validation set); and 2-class (NL vs. EIN+EMCA) error rate of 0.016 (training set) and 0 (validation set). The 4 most heavily weighted variables are surrogates of those previously identified in manual-segmentation machine learning studies (stromal and epithelial area percentages, and normalized epithelial surface lengths). Lesser weighted predictors include gland and lumen axis lengths and ratios, and individual cell measures. Automated image analysis and random forest classification algorithms can classify normal, premalignant, and malignant endometrial tissues. Highest predictive variables overlap with those discovered independently in early models based on manual segmentation.