Ovarian recurrence risk assessment using machine learning, clinical information, and serum protein levels to predict survival in high grade ovarian cancer

Abstract

In ovarian cancer, there is no current method to accurately predict recurrence after a complete response to chemotherapy. Here, we develop a machine learning risk score using serum proteomics for the prediction of early recurrence of ovarian cancer after initial treatment. The developed risk score was validated in an independent cohort with serum collected prospectively during the remission period. In the discovery cohort, patients scored as low-risk had a median time to recurrence (TTR) that was not reached at 10 years compared to 10.5 months (HR 4.66, p < 0.001) in high-risk patients. In the validation cohort, low-risk patients had a median TTR which was not reached compared to 4.7 months in high-risk patients (HR 4.67, p = 0.009). In advanced-stage patients with a CA125 < 10, low-risk patients had a median TTR of 68 months compared to 6 months in high-risk patients (HR 2.91, p = 0.02). The developed risk score was capable of distinguishing the duration of remission in ovarian cancer patients. This score may help guide maintenance therapy and develop innovative treatments in patients at risk at high-risk of recurrence.

Serum Proteomic Signatures in Cervical Cancer: Current Status and Future Directions

“…In addition to surgical outcomes, Mysona et al (2019) developed a serous high-grade ovarian cancer (SHOC) score to predict PFS, which implemented the elastic net ML algorithm to create a comprehensive multivariate score based on clinical and serum data [78]. In another study, a ML risk assessment score was able to predict ovarian cancer recurrence in patients following chemotherapy treatment [79]. The score utilized the elastic net ML algorithm to combine multiple clinical parameters and serum proteomic data to predict time-to-recurrence (TTR) and time-to-death (TTD), as well as PFS and OS in both low and high-risk groups [79].…”

“…In another study, a ML risk assessment score was able to predict ovarian cancer recurrence in patients following chemotherapy treatment [79]. The score utilized the elastic net ML algorithm to combine multiple clinical parameters and serum proteomic data to predict time-to-recurrence (TTR) and time-to-death (TTD), as well as PFS and OS in both low and high-risk groups [79]. The use of ML in OC provides some useful building blocks upon which proteomic studies in CC can build.…”