Diagnostic performance of the ADNEX model compared with ROMA, RMI, and CPH‑I for ovarian cancer detection: a single‑center retrospective study

Background To evaluate the diagnostic performance of ROMA, RMI, CPH-I, and ADNEX models for ovarian cancer (OC). Methods This single‑center retrospective study included 190 patients with pelvic masses (June 2024–January 2026). Serum CA125 and HE4 were measured, and ultrasonographic features were recorded. Patients were stratified by pathology into benign ( n = 71) and malignant ( n = 119) groups, giving a malignancy prevalence of 62.6%. Diagnostic accuracy of all models was compared using ROC curve analysis, optimal cut‑offs were determined via Youden’s index. Results Serum CA125, HE4, ROMA, RMI, CPH-I, and ADNEX values were significantly higher in malignant versus benign, EOC versus non-EOC, and postmenopausal versus premenopausal groups (all P < 0.001). The ADNEX model achieved the highest AUC (0.954), with 94.1% sensitivity and 93.0% specificity at a pre‑specified cut‑off of 10%. Pairwise comparisons showed ADNEX significantly outperformed HE4( P < 0.001), CPHI ( P = 0.008), ROMA ( P = 0.004), and CA125 ( P = 0.006). Conclusion The ADNEX model demonstrates superior diagnostic accuracy in this high‑risk cohort. Integrated assessment combining serum biomarkers with ultrasonography is essential for accurate preoperative evaluation of pelvic masses.