Abstract 3218: Radiomic biomarkers to optimize treatment decision and predict patient outcomes in serous ovarian carcinoma

Abstract

Abstract Background: There are two first-line treatment recommendations for advanced ovarian cancer: i) upfront cytoreductive debulking followed by chemotherapy and ii) neoadjuvant chemotherapy prior to surgical debulking. The choice between these two treatment strategies is controversial as there are no standardized guidelines for clinical decision support. As such, there remains a critical unmet need to identity biomarkers to personalize the most effective treatment strategies. The primary objective of this study is to identify and validate radiomic biomarkers that predict treatment response among patients with high-grade serous ovarian cancer treated with upfront surgical debulking. Methods: Intratumoral radiomic features (n=308) were extracted from pre-treatment contrast-enhanced CT images; analyses were conducted to remove correlated, non-stable, and non-reproducible features. Patients treated with upfront surgery (N=182) were split into training (N=91) and test (N=91) cohorts and a cohort of patients treated with upfront neoadjuvant (N=116) was used to determine if the radiomic features were prognostic or predictive. Overall (OS) and progression-free survival (PFS) were the main endpoints. Classification and Regression Tree analysis was used to identity the most informative radiomic features in the training cohort, which were then analyzed in the test cohort and the upfront neoadjuvant cohort. Results: Decision tree analysis identified a volumetric feature, ROI volume center of mass (CoM) in X direction, as the most informative radiomic feature which stratified upfront surgery patients into high- and low-risk. In the training cohort, high-risk patients were associated with significantly worse OS versus low-risk patients (HR=2.01; 95% CI 1.07-3.77 vs. 1.00 and 5-year OS=39.9% vs. 56.1%, respectively; log-rank p-value=0.03). In the test cohort, the high- vs. low-risk patients were also associated with poor OS (HR=2.23; 95% CI 1.19-4.17 vs. 1.00 and 5-year OS=23.3% vs. 51.3%, respectively; log-rank p-value=0.01). This radiomic feature was not associated with OS among patients with upfront neoadjuvant (HR=1.23; 95% CI 0.73-2.07 vs. 1.00 and 5-year OS=23.9% vs. 36.6%, respectively; log-rank p-value=0.44). Similar findings were observed for PFS. Conclusion: Utilizing standard-of-care imaging, we identified and validated a predictive radiomic feature associated with outcomes among patients with high-grade serous ovarian cancer treated with upfront surgical debulking but not among patients treated with neoadjuvant chemotherapy. This radiomic biomarker, which describes the location of center of mass inside the tumor in pixels in the x-direction, could be potentially utilized as clinical decision support to guide first-line treatment options. This study was generously funded by a Miles for Moffitt pilot grant. Citation Format: Jaileene Perez-Morales, Christelle M. Colin Leitzinger, Sweta K. Sinha, Melissa J. McGettigan, Daniel K. Jeong, Olya Stringfield, Mahmoud Abdala, Natarajan Raghunand, Robert J. Gillies, Jing-Yi Chern, Lauren C. Peres, Matthew B. Schabath. Radiomic biomarkers to optimize treatment decision and predict patient outcomes in serous ovarian carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 3218.