A Novel predictive probability algorithm to determine risk for malignancy between uterine fibroids versus uterine sarcoma.

Abstract

e18098 Background: The combination of HE4 and CA125 can be used as a predictive probability algorithm to determine the risk of malignancy in women with a uterine mass. Many studies have been done looking at ways to differentiate between benign fibroids and uterine sarcomas with limited success. This study examined the utility of using a logistics regression algorithm containing biomarkers HE4 and CA 125 to predict risk of malignancy of a uterine mass. Methods: This was an IRB retrospective study using de-identified data form 5 pelvic mass studies. Patients were included if they were diagnosed with either uterine fibroids or uterine sarcoma on final pathology. Pre-operative serum levels of HE4 and CA125 were obtained for each patient. A logistics regression analysis was performed in a prior pelvic mass prospective trial and utilized in this analysis. The predictive probability algorithm was used to classify patients into high and low risk groups for sarcoma. Wilson’s score interval was used to determine confidence intervals. Results: There were 71 patients identified with a uterine mass. The mean age of study participants was 54 (range 22-85). There were 10 (14.1%) sarcomas and 61 fibroids (85.9%) identified. Six of the sarcomas were leiomyosarcomas (60.0%). There was 1 adenosarcoma (10%), 1 mixed sarcoma (10%) and 2 sarcomas which were not further characterized. A threshold of 13.1% was used to classify masses as low or high risk. The predictive probability algorithm was found to have a sensitivity of 90.9% (CI 55.5-99.7%), specificity of 60.7% (CI 47.3-72.9%), PPV of 27.3% (13.3-45.5%), and NPV of 97.4% (86.2-99.9%). An elevated risk of malignancy was noted in 9 (90%) sarcomas and 24 (39%) of fibroids. Conclusions: A predictive probability algorithm using HE4 and CA 125 had a high sensitivity for determining high and low risk of malignancy in patients with presumed uterine fibroids with a sensitivity of 90.9% for detecting sarcoma. This algorithm will be validated in a prospective clinical trial.