Diagnostic Accuracy of O‐ <scp>RADS US</scp> , <scp>IOTA</scp> Simple Rules, Subjective Assessment, and the Sequential <scp>IOTA</scp> Simple Rules–Subjective Assessment ( <scp>SR</scp> – <scp>SA</scp> ) Strategy in Adnexal Masses: High Sensitivity V

ABSTRACT Objectives To compare four ultrasound models: O‐RADS US, IOTA Simple Rules (SR), Subjective Assessment (SA), and the Sequential SR–Subjective Assessment strategy (SR–SA) for classifying adnexal masses and to assess whether additional discrimination is possible within intermediate‐risk O‐RADS 3 lesions in a middle‐resource setting. Methods Prospective cohort of 122 women. Two IOTA‐certified radiologists performed ultrasound with on‐scan consensus. Each lesion received all four classifications. The reference standard was histopathology or stability at 12‐month imaging follow‐up. Diagnostic metrics were calculated; false negatives were reviewed; O‐RADS 3 and 4–5 subgroups were analyzed; and decision‐tree models were explored. Results Among 121 adnexal/ovarian masses, 40 were malignant (33.1%). With O‐RADS US (≥ 4 positive), sensitivity and negative predictive value (NPV) were highest (97.5%, 98.0%) with specificity 59.3%. SR sensitivity/specificity were 85.0%/82.7%. SA and SR–SA achieved higher specificities (87.7%, 91.4%) with intermediate sensitivities (85.0%, 82.5%). AUCs were 0.784 (O‐RADS US), 0.840 (SR), 0.853 (SA), and 0.859 (SR–SA). SR–SA and SA exceeded O‐RADS US (ΔAUC 0.086, p = 0.027; ΔAUC 0.080, p = 0.034). False negatives occurred in one O‐RADS US case, six SR cases, seven SR–SA cases, and six SA cases. Among O‐RADS 3 lesions ( n = 17), 1 (5.9%) was malignant; decision‐tree analyses found no additional splits. Conclusions O‐RADS US maximized sensitivity and NPV but had lower specificity. SA and SR–SA improved specificity and discrimination but require experienced operators. Model selection should reflect resource constraints: O‐RADS US for safety‐first triage; SA or SR–SA where expertise is available; and targeted MRI when feasible.