Diagnostic performance of ultrasonography-based risk models in differentiating between benign and malignant ovarian tumors in a Chinese cohort.

​​backgroundUltrasound-based risk models provide a valuable tool to support non-expert clinicians in evaluating adnexal masses, potentially reducing unnecessary surgeries for benign conditions. However, their implementation in clinical practice across China remains suboptimal, and direct comparisons of their diagnostic accuracy within this specific population are still needed.​objectiveTo evaluate and compare the diagnostic performance of three ultrasonography-based risk models-the International Ovarian Tumor Analysis (IOTA) Simple Rules (with expert adjudication for inconclusive cases), the IOTA Assessment of Different Neoplasias in the Adnexa (ADNEX) model, and the Ovarian-Adnexal Reporting and Data System (O-RADS)-for distinguishing benign from malignant adnexal lesions in a Chinese cohort.MethodsThis retrospective diagnostic study was conducted at a single Chinese academic medical center and included consecutive patients aged 18 to 89 years with adnexal masses that were managed surgically between January 2023 and December 2024. Each adnexal lesion was systematically evaluated using the IOTA Simple Rules, ADNEX model, and O-RADS classification. The diagnostic performance of these models was quantified by the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, positive and negative predictive values, and likelihood ratios, using surgical pathology as the reference standard. A secondary analysis evaluated the models' performance stratified by menopausal status. An additional analysis was to evaluate the supplementary value of tumor markers for the O-RADS model.​resultsA total of 459 female patients with a 15.5% malignant tumor prevalence were included. Mean age was 43.0 (15.2) years overall, with 40.7 (14.4) in benign and 55.3 (13.3) in malignant cases. Overall, 121 (26.4%) patients were postmenopausal. The AUCs for discriminative performance of the ADNEX and O-RADS models were 0.95 (95%CI, 0.93-0.97) and 0.96 (95%CI, 0.94-0.97), respectively. After converting the ADNEX continuous individualized risk into the discrete ordinal categories of O-RADS, the ADNEX performance was reduced to an AUC of 0.92 (95%CI, 0.90-0.95). The AUC of the Simple Rules was 0.95 (95%CI, 0.93-0.97) and yielded a sensitivity of 95.8% (95%CI, 88.1%-99.1%), specificity of 94.3% (95%CI, 91.5%-96.4%). For a binary comparison of all models (benign vs malignant), a uniform 10% risk threshold (O-RADS score of 4-5) was used. The sensitivity and NPV of all models were above 91% and 98%, respectively. The O-RADS model exhibited the highest sensitivity (97.2%; 95%CI, 90.2%-99.7%), whereas the Simple Rules model demonstrated superior specificity 94.3% (95%CI, 91.5%-96.4%) and the highest PPV (75.6%; 95%CI, 65.4%-84.0%). The sensitivities were higher for menopausal groups, while the specificities were higher for premenopausal patients for the ADNEX and O-RADS model. The addition of tumor markers did not improve the diagnostic sensitivity or specificity of the O-RADS model.​conclusionIn this diagnostic study of a Chinese cohort, the Simple Rules, ADNEX, and O-RADS models demonstrated effective performance in distinguishing benign from malignant adnexal lesions, consistent with previous reports from European and American populations. The use of such risk-stratification models can enhance the accuracy and reproducibility of adnexal mass assessment, particularly for non-specialist clinicians, and may support clinical decision-making and potentially reduce​ avoidable surgical interventions.