TOPLINE:

In a prospective cohort study, the ultrasound-based International Ovarian Tumour Analysis (IOTA) Assessment of Different Neoplasias in the adnEXa (ADNEX) model at a threshold of 10% demonstrated superior sensitivity than the current standard Risk of Malignancy Index 1 in detecting ovarian cancer in premenopausal women.

METHODOLOGY:

• Researchers conducted a prospective cohort analysis with 1211 premenopausal women (median age, 43.2 years; 84.7% White; 2.2% mixed; 6.0% Asian; 2.5% Black; and 1.2% other) from 23 hospitals in the UK between June 2015 and March 2023, with follow-up until March 2023.
• Patients included were divided into two groups: the pre-protocol change cohort, consisting of presurgical and conservatively managed patients until June 2018; and the post-protocol change cohort, including only presurgical patients recruited after June 2018.
• Patients completed a symptom questionnaire, blood sampling, and transabdominal and transvaginal ultrasound scans, with premenopausal status determined by age (< 50 years) and menstruation in the past 12 months.
• The analysis compared the diagnostic accuracy of Risk of Malignancy Index 1 (thresholds 200 and 250), cancer antigen 125, and three IOTA tests: the ADNEX model, the simple rules risk model (thresholds 3% and 10%), and the simple rules classifier.
• The primary outcome was the diagnostic accuracy of index tests in identifying primary invasive malignant neoplasms vs benign findings through surgical or biopsy histology or 12-month surveillance, and secondary outcomes included a broader spectrum of malignancies, encompassing metastatic and borderline neoplasms.

TAKEAWAY:

• Risk of Malignancy Index 1 demonstrated low sensitivity at thresholds of 200 (48.9%; 95% CI, 34.1%-63.9%) and 250 (42.6%; 95% CI, 28.3%-57.8%) but high specificity at thresholds of 200 (95.4%; 95% CI, 93.4%-96.9%) and 250 (96.5%; 95% CI, 94.7%-97.8%).
• All other models showed significantly higher sensitivity, with the IOTA ADNEX model exhibiting the highest sensitivity at 89.1% (P < .001). Similarly, all the models showed significantly lower specificity than Risk of Malignancy Index 1, with the Risk of Malignancy Algorithm having the least specificity at 73.1% (P < .001).
• All index tests had high negative predictive values, ranging from 95.7% (95% CI, 93.8%-97.2%) for Risk of Malignancy Index 1 at a threshold of 250 to 98.9% (95% CI, 96.7%-99.8%) for IOTA ADNEX at a threshold of 3%. Positive predictive values ranged from 10.4% (95% CI, 7.6%-13.7%) for the Risk of Malignancy Algorithm to 49.0% (95% CI, 34.4%-63.7%) for IOTA simple rules.
• Results remained consistent across both primary and secondary outcome analyses, including the detection of metastatic cancers and borderline tumours.

IN PRACTICE:

"Primary care implementation could potentially improve survival through earlier detection in symptomatic premenopausal women, but this requires further research alongside investment in sonographer training and quality assurance," the authors wrote.

SOURCE:

This study was led by Sudha Sundar, PhD, Pan-Birmingham Gynaecological Cancer Centre, Sandwell and West Birmingham Hospitals NHS Trust, Birmingham, England. It was published online on January 29, 2026, in The BMJ.

LIMITATIONS:

The recruitment aligned with UK census patterns comprising mainly White patients; therefore, the results had potentially limited applicability to people of different ethnicities. The decision to restrict recruitment to presurgical patients following the observation of low ovarian cancer prevalence led to potential systematic differences between the two cohorts. Additionally, not scheduling additional hospital visits for transvaginal IOTA ultrasound during the COVID-19 pandemic resulted in missing data, particularly for ultrasound variables.

DISCLOSURES:

This study was funded by the National Institute for Health and Care Research Health Technology Assessment. One author reported receiving research grants from AOA Diagnostics and honorariums from AstraZeneca, Merck, and GSK. Additional disclosures are noted in the original article.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.