The aim of the study was to investigate whether a combination of serum CA-125, CA-72.4; CA-19.9 and VEGF levels along with several currently used gray-scale criteria and pulsed and semiquantitative color Doppler blood flow assessment can be useful in preoperative discrimination of ovarian tumors. Ultrasound examinations were performed between 1994 and 1999 with the use of transvaginal probes 5-6.5-7.0 MHz and transabdominal 3.5 MHz probe (B&K Medical 3535 and 2002 ADI, Denmark). Preoperative serum CA-125 and CA-19.9 (Roche, Switzerland) and CA-72.4 (Centocor, USA) were measured with EIA II immunoassays. VEGF (R&D Systems, USA) serum levels were measured by microplate immunoenzymatic method. Retrospective study included 4,876 women referred for sonography of pelvic structures. For final analysis 451 patients with persistent adnexal tumors were available. In all these women age, menopausal status, tumor size, volume and morphology as well as blood flow indices (PI, RI, PSV) and semiquantitative color assessment were noted. In women who were not operated the cysts were followed for 10-12 weeks until resolution which was confirmed by repeated sonography. Multiple regression analysis was used to determine which of the independent variables had prognostic significance in the constructed predictive model. ROC curves were plotted and areas under ROC for each test were calculated and compared. 286 women were operated and in 118 patients their tumors regressed. The sensitivity for Doppler indices was between 85% (RI), 67% (PSV) and 66% (PI). For the subjective color assessment sensitivity was 87%. 294 women had their CA-125 serum levels available. For CA-19.9, CA-72.4 and VEGF the corresponding numbers were 109, 92 and 116 patients, respectively. Optimal cutoff level for CA-125 was 67.5 U/ml and this produced diagnostic accuracy of 69.4%. For CA-19.9 the best predictive value was 13.4 U/ml (accuracy of 74%) and for the cut-off CA-72.4 level of 4.1 U/ml the accuracy of the test was 77.8%. Multiple regression analysis revealed that only RI measurements and subjective color assessment had significant influence on the constructed predictive model. The best cut-off value of regression index "Z" was obtained following ROC curves construction for sensitivity (true positive rate) and 1-specificity (false positive rate). In all 116 patients who had Doppler indices with both CA-125 and VEGF levels measured the highest accuracy was associated with "Z" = 1.45 in postmenopausal women. The prognostic model proposed in this study can be log transformed and further used in the estimation of the true probability of malignancy of a given mass in the new set of prospectively studied patients with ovarian tumors. Color and pulsed Doppler can improve preoperative diagnosis of adnexal tumors when compared to transvaginal sonography alone or tumor markers assessment. The proposed semiquantitative evaluation of tumor vascularity increases the predictive value in terms of sensitivity and specificity. VEGF serum concentration was not useful in the preoperative discrimination of malignant and benign ovarian tumors.