Abstract 595P
Background
Ovarian cancer is often not detected (∼70% of the cases) until it has spread to regional or distant sites, thus limiting curative options. In contrast, when detected in a localized state, the 5-year survival rate increases dramatically to ∼90% in comparison to the ∼30% survival rate for distant cases. Currently, only 15% of cases are detected at early stages highlighting the need for the development of non-invasive detection strategies such as liquid biopsies.
Methods
We tested the feasibility of using protein biomarkers from extracellular vesicles (EVs) isolated directly from blood plasma via a microelectrode array technology followed by multiplex ELISA. The EV-proteins were evaluated for their ability to differentiate between ovarian cancer cases and controls. The case cohort was comprised of 69 ovarian cancer cases (Stage I = 50, Stage II = 19) with a median age of 54 years. The control cohort (no known cancer diagnosis) was only female donors with a median age of 57 years. A machine learning (ML) algorithm was employed to identify the most important features for diagnostic purposes and to build a classifier for ovarian cancer. The performance is reported from 10 repetitions of 5-fold cross validation.
Results
The ML algorithm produced an AUC of 0.968 (95% CI: 0.923 – 0.985) with an overall sensitivity of 85.5% (CI: 75.3 – 91.9) at a specificity of 95% (CI: 92.3 – 96.8). When further stratified, stage I sensitivity was 82.0% (CI: 69.2 – 90.2) and stage II sensitivity was 94.7% (CI: 75.4 – 99.1) at the 95% specificity threshold. Furthermore, when we examined the earliest possible ovarian cancer cases available in our cohort, stage IA (N = 29), and found a sensitivity of 72.4% (CI: 54.3 – 85.3) at 95% specificity. We also analyzed the endometroid (N = 23) and serous carcinoma cases (N = 35) and found sensitivities of 95.7% (CI: 79.0 – 99.2) and 80.0% (CI: 64.1 – 90.0), respectively, at a specificity of 95%.
Conclusions
Our results suggest the potential of liquid biopsy EV-based approaches for the early detection of ovarian cancer with high sensitivity and specificity. To validate the current observations, future studies will include larger datasets including populations at risk, e.g. known germline mutations, as well as benign ovarian cysts and other potentially confounding malignancies.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Biological Dynamics.
Disclosure
J.M. Lewis: Financial Interests, Institutional, Full or part-time Employment: Biological dynamics. J.P. Hinestrosa: Financial Interests, Institutional, Full or part-time Employment: Biological Dynamics. G. Schroeder: Financial Interests, Institutional, Full or part-time Employment: Biological Dynamics. H. Balcer: Financial Interests, Institutional, Full or part-time Employment: Biological Dynamics. P. Billings: Financial Interests, Institutional, Member of the Board of Directors: Biological Dynamics. All other authors have declared no conflicts of interest.