Evaluating diagnostic tools for adnexal masses in novice hands

Evaluating diagnostic tools for adnexal masses in novice hands | Image Credit: © Georgii – stock.adobe.com.

A retrospective study presented at the 2025 ACOG Annual Clinical and Scientific Meeting and led by researchers at the Mayo Clinic in Rochester, Minnesota, examined how 4 commonly used ultrasound-based algorithms perform in distinguishing benign from malignant adnexal masses when used by a novice operator. The algorithms—ADNEX (Assessment of Different Neoplasias in the Adnexa), a 2-step strategy (benign descriptors followed by ADNEX), O-RADS 2019, and O-RADS 2022—were applied to patients by a European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) level I operator.¹

Study design and population

The analysis included 556 women with adnexal masses treated in 2019. Eligible patients had undergone surgery within 3 months of diagnosis or had at least 10 months of follow-up. Among them, 452 had benign and 104 had malignant lesions.

Each adnexal mass was evaluated using all 4 diagnostic approaches. The performance of each algorithm was measured using the area under the receiver operating characteristic curve (AUC), a standard metric for diagnostic accuracy.

Diagnostic performance of algorithms

The 2-step strategy yielded the highest AUC at 0.91 (95% CI, 0.88–0.94), followed closely by ADNEX at 0.90 (95% CI, 0.87–0.94). O-RADS 2019 and 2022 both had an AUC of 0.88 (95% CI, 0.84–0.91). Statistically, the 2-step strategy performed better than both O-RADS versions (P = .005 and P = .002). However, the authors noted that these differences may not be clinically meaningful.

All models were able to differentiate between benign and malignant masses effectively, even in the hands of a novice operator. This aligns with previous research on the ADNEX model, which found it maintained high predictive accuracy when used by less experienced practitioners familiar with IOTA terminology and measurement protocols.

Elevated malignancy rates in low-risk categories

Across all models, the observed malignancy rate among lesions categorized as “almost certainly benign” ranged from 1.9% to 2.2%, roughly double the expected rate of less than 1.0%. These misclassified lesions included borderline tumors (n=4) and secondary metastatic tumors (n=3), highlighting an important limitation in current risk stratification frameworks.

Prior studies have shown that while the ADNEX model provides absolute risk estimates for 5 tumor categories—benign, borderline, stage I cancer, stage II-IV cancer, and secondary metastatic cancer—the predicted risks for rare malignancy subtypes may remain low due to their baseline prevalence, making misclassification more likely.²

Context and clinical implications of ADNEX

Developed by the International Ovarian Tumor Analysis (IOTA) group, the ADNEX model includes 3 clinical predictors (age, CA-125, and type of center) and 6 ultrasound predictors (lesion size, solid tissue proportion, papillary projections, cyst locules, acoustic shadows, and ascites). Its ability to distinguish among malignant subtypes makes it unique, although interpretation must be tailored to individual clinical settings.²

Evidence from the IOTA study shows that ADNEX had a validation AUC of 0.943 with CA-125 and 0.932 without, suggesting strong performance even in the absence of tumor marker data. This enables clinicians to initially screen patients using ultrasound alone and follow up with CA-125 testing if malignancy risk is high.

The choice of malignancy cut-off varies by institution and can influence management strategies. For example, a center prioritizing sensitivity may use a low threshold (e.g., 5–10%) to maximize malignancy detection, while others may favor specificity and use higher thresholds to reduce unnecessary referrals.

Conclusion

The authors concluded, “In the hands of a novice operator, all algorithms performed well and were able to distinguish benign from malignant lesions. Although the 2-step strategy performed slightly better than the O-RADSs, the difference did not appear to be clinically meaningful.”

The unexpected malignancy rate in lesions deemed almost certainly benign underscores the importance of combining algorithmic assessments with clinical judgment, especially when managing borderline and metastatic tumors.

References:

1. De Vitis LA, Schivardi G, Grcevich L, et al. Diagnostic Algorithms for Adnexal Masses in the Hands of a Novice Operator. Obstetrics & Gynecology 145(4):p 368-376, April 2025. doi:10.1097/AOG.0000000000005853

2. Van Calster B, Van Hoorde K, Froyman W, et al. Practical guidance for applying the ADNEX model from the IOTA group to discriminate between different subtypes of adnexal tumors. Facts Views Vis Obgyn. 2015;7(1):32-41.