Predictive Modeling for Ovarian Cancer Diagnosis: Integration of Positron Emission Tomography/Computed Tomography (PET/CT) Metabolic Parameters, Clinical Risk Factors, and Serum Biomarkers.

BACKGROUND: Ovarian cancer is often diagnosed at an advanced stage, leading to high mortality. This study aimed to develop and validate an integrated predictive model for advanced-stage (International Federation of Gynecology and Obstetrics III/IV) ovarian cancer combining positron emission tomography/computed tomography (PET/CT) metabolic parameters, clinical risk factors, and serum biomarkers in patients with histologically confirmed epithelial ovarian cancer. METHODS: A retrospective cohort study was conducted among 187 patients with histologically confirmed epithelial ovarian cancer at King Hussein Medical Center (from January to December 2023). A multivariable logistic regression model was developed integrating PET/CT metabolic parameters (maximum standardized uptake value (SUVmax) and metabolic tumor volume), clinical risk factors (age, nulliparity, breast cancer gene (BRCA) status, and ascites), and serum biomarkers (cancer antigen 125 (CA-125), human epididymis protein 4). Model performance was assessed using area under the curve (AUC), sensitivity, specificity, net reclassification improvement (NRI), and decision curve analysis. Internal validation was performed using 1,000 bootstrap resamples. RESULTS: The integrated model demonstrated superior discrimination for advanced-stage disease (AUC = 0.91, 95% confidence interval (CI): 0.87-0.95) compared with PET/CT alone (AUC = 0.79, p = 0.002) or clinical evaluation with CA-125 (AUC = 0.74, p < 0.001). Key independent predictors included SUVmax >10 (adjusted odds ratio (aOR) = 2.67, 95% CI: 1.48-4.82), CA-125 ≥350 U/mL (aOR = 2.35, 95% CI: 1.32-4.18), ascites (aOR = 3.78, 95% CI: 2.04-7.01), and nulliparity (aOR = 1.98, 95% CI: 1.12-3.51). High-grade serous carcinomas demonstrated significantly higher metabolic activity (median SUVmax 12.4) compared with other subtypes (p < 0.001). The full model provided a significant NRI of 19.3% over the PET/CT-plus-clinical model, with an optimism-corrected AUC of 0.89. Decision curve analysis confirmed clinical utility across the 20%- 80% threshold probabilities. Model performance was superior in BRCA mutation carriers (AUC = 0.94) vs. noncarriers (AUC = 0.85, interaction p = 0.038). CONCLUSIONS: Integration of PET/CT metabolic parameters with clinical and biomarker variables significantly enhances prediction of advanced-stage ovarian cancer. This multivariate approach provides a foundation for preoperative risk stratification, enabling improved surgical planning and optimized referral to specialized centers.