<scp>OC</scp>14.07: The predictive value of subjective assessment, <scp>IOTA</scp> simple rules (<scp>SR</scp>), <scp>IOTA</scp> logistic regression model 1 (<scp>LR1</scp>), Alcazar's <scp>GI‐RADS</scp> model and a risk of malignancy (<scp>RMI</scp>) model in the preoperative discrimination of adnexal masses

To validate the predictive value of 4 predictive models, and to compare them with the performance of a non-expert ultrasound examiner (“subjective assessment”, SA) containing clinical and ultrasound variables to evaluate the risk of malignancy in adnexal masses. Study group consisted of women who were operated because of adnexal masses and had undergone a standardized transvaginal ultrasound examination. The non-expert gynecologist retrospectively analyzed all collected information on tumors clinical and ultrasound variables, and classified each mass as benign or malignant on the basis of subjective evaluation of ultrasound findings and 4 currently used models. The accuracy of the risk estimates were compared. Final analysis included 271 women with adnexal masses (78 malignant, 28.8%), 95 (35.1%) patients were postmenopausal. Mean age of women was 49.9 (range: 19–85) yrs. Table 1 presents results of calculated predictive values of various methods. The AUC for used methods for all tumors were as follows: subjective assessment AUC=0.965 (SE=0.011, 95% CI:0.945-0.986), for LR1 AUC=0.951 (SE=0.014,95% CI: 0.924-0.978) and for GI-RADS AUC=0.913 (SE=0.019,95% CI; 0.876- 0.950). Subjective assessment provided better than other models discrimination between benign and malignant ovarian tumors. However, all used models could provide an objective and very good to fairly accurate risk of malignancy estimation in hands of a no-expert examiner.