Improving the Diagnosis of Endometrial Hyperplasia Using Computerized Analysis and Immunohistochemical Biomarkers

Abstract

Endometrial hyperplasia (EH) is a precursor lesion to endometrial carcinoma (EC). Risks for EC include genetic, hormonal and metabolic factors most notably those associated with obesity: rates are rising and there is concern that cases in pre-menopausal women may remain undetected. Making an accurate distinction between benign and pre-malignant disease is both a challenge for the pathologist and important to the gynecologist who wants to deliver the most appropriate care to meet the needs of the patient. Premalignant change may be recognized by histological changes of endometrial hyperplasia (which may occur with or without atypia) and endometrial intraepithelial neoplasia (EIN). In this study we created a tissue resource of EH samples diagnosed between 2004 and 2009 (n = 125) and used this to address key questions: 1. Are the EIN/WHO2014 diagnostic criteria able to consistently identify premalignant endometrium? 2. Can computer aided image analysis inform identification of EIN? 3. Can we improve diagnosis by incorporating analysis of protein expression using immunohistochemistry. Our findings confirmed the inclusion of EIN in diagnostic criteria resulted in a better agreement between expert pathologists compared with the previous WHO94 criteria used for the original diagnosis of our sample set. A computer model based on assessment of stromal:epithelial ratio appeared most accurate in classification of areas of tissue without EIN. From an extensive panel of putative endometrial protein tissue biomarkers a score based on assessment of HAND2, PTEN, and PAX2 was able to identify four clusters one of which appeared to be more likely to be benign. In summary, our study has highlighted new opportunities to improve diagnosis of pre-malignant disease in endometrium and provide a platform for further research on this important topic.