To determine whether ultrasound features can improve the diagnostic performance of tumor markers in distinguishing ovarian tumors, we enrolled 719 patients diagnosed as having ovarian tumors at Nanfang Hospital from September 2014 to November 2016. Age, menopausal status, histopathology, the International Federation of Gynecology and Obstetrics (FIGO) stages, tumor biomarker levels, and detailed ultrasound reports of patients were collected. The area under the curve (AUC), sensitivity, and specificity of the bellow-mentioned predictors were analyzed using the receiver operating characteristic curve. Of the 719 patients, 531 had benign lesions, 119 had epithelial ovarian cancers (EOC), 44 had borderline ovarian tumors (BOT), and 25 had non-EOC. AUCs and the sensitivity of cancer antigen 125 (CA125), human epididymis-specific protein 4 (HE4), Risk of Ovarian Malignancy Algorithm (ROMA), Risk of Malignancy Index (RMI1), HE4 model, and Rajavithi-Ovarian Cancer Predictive Score (R-OPS) in the overall population were 0.792, 0.854, 0.856, 0.872, 0.893, 0.852, and 70.2%, 56.9%, 69.1%, 60.6%, 77.1%, 71.3%, respectively. For distinguishing EOC from benign tumors, the AUCs and sensitivity of the above mentioned predictors were 0.888, 0.946, 0.947, 0.949, 0.967, 0.966, and 84.0%, 79.8%, 87.4%, 84.9%, 90.8%, 89.1%, respectively. Their specificity in predicting benign diseases was 72.9%, 94.4%, 87.6%, 95.9%, 86.3%, 90.8%, respectively. Therefore, we consider biomarkers in combination with ultrasound features may improve the diagnostic performance in distinguishing malignant from benign ovarian tumors.
Keywords: Rajavithi-Ovarian Cancer Predictive Score; Risk of Malignancy Index; Risk of Malignancy Index model; Risk of Ovarian Malignancy Algorithm; cancer antigen 125; human epididymis-specific protein 4; ovarian masses.