IOTA ADNEX Model Outperforms Current Ovarian Cancer Triage Standard

Key Clinical Summary

• The International Ovarian Tumour Analysis Assessment of Different Neoplasias (IOTA ADNEX) model demonstrated substantially higher sensitivity than the Risk of Malignancy Index 1 (RMI1) for detecting ovarian cancer in premenopausal women.
• At a 10% threshold, the IOTA ADNEX model achieved 89.1% sensitivity compared with 42.6% for the RMI1 at the currently used 250 threshold.
• Investigators concluded that the IOTA ADNEX model should be considered the new standard-of-care triage test in secondary care settings.

A large prospective study found that ultrasound-based IOTA models significantly outperformed the RMI1, the current National Health Service standard, for identifying ovarian cancer in premenopausal women presenting with symptoms and abnormal test results.

The findings suggest that implementation of the IOTA ADNEX model could improve early detection and referral of ovarian cancer while supporting more accurate triage decisions in secondary care.

Head-to-Head Evaluation of Ovarian Cancer Prediction Tools

Researchers conducted a prospective cohort study across 23 hospitals in the United Kingdom between 2015 and 2023.

The analysis included 1211 premenopausal women referred to secondary care with symptoms suggestive of ovarian cancer and either elevated cancer antigen 125 (CA 125) levels or abnormal imaging findings. Most patients were referred through the National Health Service urgent suspected cancer pathway.

Investigators directly compared 6 commonly used ovarian cancer risk prediction models and scoring systems, including the RMI1, Risk of Malignancy Algorithm, IOTA ADNEX model, IOTA Simple Rules Risk Model, IOTA Simple Rules, and CA 125 alone.

Among 799 women included in the primary diagnostic analysis, 88 were diagnosed with primary invasive ovarian cancer.

IOTA ADNEX Delivered the Highest Sensitivity

The currently used RMI1 threshold of 250 demonstrated a sensitivity of only 42.6% while maintaining high specificity of 96.5%.

In contrast, the IOTA ADNEX model achieved a sensitivity of 89.1% at a 10% threshold, more than doubling the cancer detection rate observed with RMI1. Specificity was lower at 75.1%, reflecting the tradeoff between identifying more cancers and increasing false-positive results.

Other models also outperformed RMI1 in sensitivity. The Risk of Malignancy Algorithm achieved 79.2% sensitivity, while the IOTA Simple Rules Risk Model reached 83.0%.

CA 125 alone demonstrated a sensitivity of 55.1%, modestly higher than RMI1 but substantially lower than the ultrasound-based IOTA models.

The IOTA Simple Rules approach achieved a sensitivity of 75.0%; however, results were inconclusive in 120 patients, limiting its practical utility.

Implications for Clinical Practice

The findings are particularly relevant because ovarian cancer remains difficult to diagnose early, especially in younger women with nonspecific symptoms.

Although RMI1 has long been used to guide referral decisions in secondary care, its relatively low sensitivity may result in delayed recognition of malignancy.

The authors noted that the IOTA ADNEX model produced the largest improvement in cancer detection while maintaining specificity comparable to other alternative approaches.

Implementation would require structured ultrasound training and quality assurance programs to ensure consistent application of IOTA terminology and scoring methods.

Looking Ahead

The investigators concluded that ultrasound assessment using the IOTA ADNEX model at a 10% threshold should replace RMI1 as the preferred triage strategy for premenopausal women undergoing evaluation for possible ovarian cancer.

If adopted broadly, the approach could improve referral pathways and facilitate earlier diagnosis of ovarian malignancies in secondary care settings.

Reference

Sundar S, Agarwal R, Scandrett K, et al. Diagnostic tests for ovarian cancer in premenopausal women with non-specific symptoms (ROCkeTS): prospective, multicentre, cohort study. BMJ. 2026;392. doi:10.1136/bmj-2024-083912