CT-Derived Body Composition Enhances Risk Prediction for Early Progression and Platinum Resistance in High-Grade Serous Ovarian Cancer.
Authors
Affiliations (8)
Affiliations (8)
- Department of Radiology, University of Pittsburgh, Pittsburgh, Pennsylvania, USA.
- Department of Research, University of Miami Miller School of Medicine, Miami, Florida, USA.
- Department of Obstetrics, Gynecology, and Reproductive Sciences, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA.
- Department of Pathology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA.
- Women's Cancer Research Center Magee-Womens Research Institute and Foundation and Hillman Cancer Center, Pittsburgh, Pennsylvania, USA.
- Department of Epidemiology, University of Pittsburgh School of Public Health, Pittsburgh, Pennsylvania, USA.
- Department of Ophthalmology, University of Pittsburgh, Pittsburgh, Pennsylvania, USA.
- Department of Bioengineering, University of Pittsburgh, Pittsburgh, Pennsylvania, USA.
Abstract
Patients with high-grade serous ovarian cancer (HGSOC) who develop resistance to platinum-based chemotherapy experience significantly poorer outcomes and shorter progression-free survival (PFS). This study evaluates whether pretreatment CT-based abdominal body composition is associated with platinum resistance and PFS, and whether its incorporation alongside clinical variables improves prognostic model performance. This retrospective study included 321 patients with HGSOC (mean age 65 ± 10 years; 28.0% platinum-resistant) treated with platinum-based chemotherapy between 2011 and 2023. Deep learning algorithms quantified abdominal body composition features from pretreatment CT scans, including skeletal muscle, bone, subcutaneous (SAT), visceral (VAT), and intramuscular (IMAT) adipose tissues. Platinum resistance, defined as disease progression during therapy or within 6 months of the final infusion, was the primary outcome. PFS was analyzed as a secondary outcome using Cox proportional hazards modeling. Analyses were stratified by chemotherapy sequence (neoadjuvant vs. adjuvant). Survival models incorporating clinical and CT-derived features were evaluated using 10-fold cross-validated time-dependent AUC over a 5-year follow-up window to predict PFS. External validation was partially performed in an independent cohort of 58 patients with HGSOC. Increased SAT and IMAT and decreased skeletal muscle density were independently associated with platinum resistance, with high SAT remaining significant in both the internal and external cohorts (p < 0.05). High adiposity and low muscle density were similarly associated with shorter PFS. Survival models combining clinical and CT-derived features consistently outperformed clinical-only models across all evaluated model classes. The Elastic Net Cox regression model incorporating both feature sets achieved the highest AUC of 0.704 (95% CI: 0.660-0.748). Pretreatment CT-derived abdominal body composition is independently associated with platinum resistance and PFS in HGSOC, and its incorporation into predictive clinical models improves discrimination. Body composition assessment may help identify patients at increased risk of platinum resistance and early disease progression.