To evaluate the performance of the two-step strategy compared with ADNEX model for preoperative classification of benign and malignant adnexal masses. This was an ambispective diagnostic accuracy study between 2012 and 2018, from women with persistent adnexal masses that underwent preoperative ultrasound evaluation. The histopathological diagnosis was used as reference standard. Two expert ultrasonographers classified prospectively the adnexal masses into benign, malignant or inconclusive using SR (1st step). The inconclusive masses were classified using ADNEX model (2nd step). Also, all masses were classified using ADNEX model. The sensitivity, specificity, positive (PPV) and negative predictive values (NPV), positive (LR+) and negative (LR-) likelihood ratios, and Receiver Operating Characteristic (ROC) curve were estimated. The study included 548 patients and 606 masses. Patients median age was 41 years with an interquartile range between 32 to 51 years, of whom 400 (73%) were premenopausal and 148 (27%) postmenopausal. In the 1st step 517 (85.3%) masses were classified with SR, 470 (90.9%) as benign and 47 (9.1%) as malignant. For the 89 (14.7%) inconclusive masses, ADNEX model was used to estimate the probability that 34 (38.2%) were benign and 55 (61.8%) were malignant. Furthermore, for the totality of 606 masses, ADNEX model estimated the probability that 480 (79.2%) were benign and 126 (20.8%) were malignant. The two-step strategy had a sensitivity, specificity, PPV, NPV, LR+, LR-, and ROC curve of 86.8%, 91.01%, 51.9%, 98.4%, 9.7, 0.1, and 0.889; compared to IOTA ADNEX model that had values of 91.8%, 87.2%, 44.4%, 98.9%, 7.2, 0.09, and 0.895, respectively. No statistical differences were found when these variables were compared in both models (p>0.05). The two-step strategy for the pre-operative assessment for adnexal masses could reduce false positive cases when compared to IOTA ADNEX model. This strategy showed an adequate ROC curve, but p value was >0.05.