ADNEX2 Risk Model: Advancing Ovarian Cancer Diagnosis

How does the ADNEX2 risk model improve adnexal mass ultrasound analysis?

The ADNEX2 model is an ultrasound-based tool designed to estimate the probability that an adnexal mass is malignant. It is an update to the original ADNEX model, which was developed using data from patients managed surgically.

Researchers at the Department of Development and Regeneration and the Leuven Unit for Health Technology Assessment Research at KU Leuven developed this version to incorporate data from both surgically and conservatively managed patients.

As of October 26, 2026, the study describing this model exists as a preprint and has not undergone peer review. It provides a new version of the prediction tool intended for clinical use when benign descriptors do not apply.

How does the two-step diagnostic strategy work?

Benign descriptors were applied to 2,190 patients, representing 32% of the cohort. These descriptors identified 2,190 patients, or 32% of the study group, of whom 0.5% had a malignancy.

For patients where these descriptors do not apply, clinicians use the model to estimate risk. This combined approach achieves an area under the curve of 0.95 for distinguishing between benign and malignant tumors, while the model alone performs at an area under the curve of 0.93 for the remaining 4,657 patients.

What data does the model use to predict risk?

The model requires specific clinical and ultrasound data to calculate risk. These predictors include age, the type of center, lesion diameter, the proportion of solid tissue, the number of papillary projections, the presence of more than 10 cyst locules, ascites, acoustic shadows, and the level of CA125.

While the model includes CA125 as a predictor, it remains functional if this measurement is unavailable. However, researchers recommend measuring this biomarker when you need to estimate the risk of specific malignant subtypes.

How was the model updated and validated?

This study involved 6847 patients with newly detected adnexal masses recruited between 2012 and 2016 across 17 oncology and non-oncology centres in 8 European countries. Of these participants, 4625 had benign tumours, 1710 had malignant tumours, and 512 had uncertain tumour outcomes.

The team managed these patients either surgically or through conservative clinical and ultrasound follow-up. To update the ADNEX2 model, researchers used logistic recalibration followed by Bayesian refitting. They validated these results using leave-one-centre-out cross-validation.

How does ADNEX2 improve upon the original model?

The study evaluated model performance by measuring discrimination using the area under the receiver operating characteristic curve (AUC) and calibration through plots and observed:expected (O:E) ratios. The two-step diagnostic strategy using ADNEX2 shows improved calibration compared to the original ADNEX model. Specifically, the ADNEX2 strategy achieved an O:E ratio of 1.05, whereas the original model had a ratio of 1.14.

Are there any conflicts of interest?

Several authors, including D.T., B.V.C., L.B., L.W., and J.C., are co-inventors of the ADNEX2 model. Gynaia Inc. and Gynaia Belgium NV hold the license for this model. D.T., T.B., B.V.C., and W.F. are authors who possess stock options in Gynaia Inc., which supplies tools such as an app for the ADNEX2 model. You can access the model through the tools provided by this company.