Diagnosing Endometriosis by Endometrial Biopsy

Endometriosis affects 10% of reproductive-age women and 50% with pain and infertility. It is a major public health issue and healthcare cost ($69.4B in 2009 in the US). The gold standard to diagnose endometriosis is surgical exploration and staging of the disease in a hospital setting under general anesthesia. Average time to diagnosis from symptom onset is 11.7 years. To shorten time to diagnosis, reduce invasiveness/risks of the diagnostic method, and optimize therapeutic decision-making, biomarkers in the endometrium (lining of the uterus, obtained by office biopsy) or in blood are being explored. We have IP on the biopsy approach. Using whole genome arrays, we found differential expression of select genes in endometrium of women with vs. without disease and have 2 patents assigned to the UC Regents (“Methods of Diagnosing Endometriosis”, Giudice, Inventor; No.7,871,778 issued 01/18/11; No.8,247,174 issued 08/21/12) based on our findings. We used margin tree classification and resampling analyses to develop a disease/severity diagnostic classifier with a sizeable (n>100) set of meticulously annotated clinical endometrial tissue samples and have 4 main findings: the endometrial transcriptome is altered in the presence of endometriosis and when non-endometrial uterine pathologies (e.g. fibroids) are present, such that classifier analysis can discriminate if these pathologies are entirely absent or one or more of them is present (e.g., in patients with pelvic pain); best performing classifiers diagnose samples from specific menstrual cycle phases; composite classifiers diagnose endometriosis disease and severity (stage) with >90% accuracy; and classifiers determine disease/severity based on small numbers of genes, which are high value candidates to develop diagnostic/biomarker/therapeutic targets. The margin tree classification utilizes sequential binary decisions each based on specific genes that are distinct for different decision nodes and classifiers. Within a classifier family, core genes are present in each decision node and are thus indispensible for diagnosis at the defined accuracy. The next steps and intent of this application are to validate core genes at the decision points in the highest performance classifiers and validate the classifiers in prospectively obtained, existing endometrial tissue from 27 women with chronic pelvic pain who underwent surgery and identification of disease/stage/no disease – outcomes to which we are blinded.