Background: Ovarian cancer (OC) lacks of specific symptoms and biomarkers for early diagnosis. Advanced stages are associated with poor survival. CA125 and HE4 are overexpressed in OC. Their role in early diagnosis is controversial discussed. Ultrasound criteria (e.g. IOTA algorithms) seem to improve the sensitivity and specificity for risk assessment. Methods: In a prospective, multicentric study performed in 676 pelvic mass patients, transvaginal ultrasound together with blood sampling and biomarker analysis were performed. All patients had an ovarian (including para-ovarian and tubal) mass and underwent a standardized ultrasound examination, according to IOTA criteria, prior to surgery. ROMA algorithm combining CA125, HE4 and menopausal status, was used to distinguish between benign and malignant disease. An online data base was provided for the prospective documentation of clinical parameters, ultrasound, surgical methods and histology. Patients were included at 7 clinics in Berlin. All samples were analyzed centralized at Labor Berlin. These results represent the analysis of the training cohort. Results: From the 676 patients, 57% were premenopausal, 13.3% were diagnosed with OC, 4.3% with borderline tumors, 0.2% Krukenberg tumors and 0.8% rare malignancies of the ovary. The area under the receiver operating characteristic curve (AUC) for the discrimination between benign and malignant tumors was for CA125 0.840 and 0.899, for HE4 0.860 and 0.885 and for ROMA 0.863 and 0.930 in pre-and postmenopausal patients, respectively. The IOTA LR2 model includes one clinical and five ultrasound predictors: age, presence of ascites, presence of papillations with detectable blood flow, maximum diameter of largest solid component, irregular cyst walls and presence of acoustic shadows. IOTA LR2 reached an AUC of 0.971 in premenopausal and of 0.748 in postmenopausal patients. Conclusions: HE4 performed slightly better than CA125 in risk assessment in premenopausal patients, nevertheless ultrasound criteria could best discriminate between benign and malignant disease in premenopausal patients. In postmenopausal patients, ROMA reached the best sensitivity and specificity.