Machine learning-enhanced immune signatures optimize cancer antigen 125 performance for epithelial ovarian carcinoma detection

Introduction: Ovarian cancer (OC) ranks as the fifth most common gynecologic malignancy among women worldwide. Objective: The present study evaluates the diagnostic potential of hematological biomarkers for the early detection and differential diagnosis of OC. Methods: A bioinformatic analysis was performed to compare immune cell profiles in blood and tissue samples from patients with OC using data from The Cancer Genome Atlas and Gene Expression Omnibus databases. Subsequently, a retrospective clinical study was conducted at Yichang Central People’s Hospital between January 2015 and January 2021, including three cohorts: (i) Patients with benign ovarian tumors (n = 70), (ii) Patients with OC (n = 70), and (iii) Healthy controls (n = 60). A comprehensive analysis of routine blood parameters and the tumor marker cancer antigen 125 (CA125) was performed. Results: The findings revealed that peripheral blood immune markers exhibited superior diagnostic utility compared with tissue-based indicators. The combination of CA125 with erythrocyte sedimentation rate (ESR) and neutrophil-to-lymphocyte ratio showed high accuracy in differentiating benign ovarian tumors from OC (area under the curve [AUC]: 0.87). Furthermore, a panel combining CA125 and platelet-to-neutrophil ratio showed enhanced diagnostic performance in distinguishing early-stage from advanced epithelial OC (sensitivity: 81.3%; specificity: 96.6%). Notably, the triad of CA125, ESR, and white blood cell count demonstrated strong screening performance for detecting epithelial OC (AUC: 0.941; p<0.001). Conclusion: These results suggest that integrating CA125 with routine hematological parameters significantly enhances the diagnostic accuracy and early detection of epithelial OC compared to CA125 alone, providing a practical and cost-effective screening strategy for clinical implementation.