Related Experiment Video
Updated: Jan 17, 2026
An R-Based Landscape Validation of a Competing Risk Model
Published on: September 16, 2022
CA125 and age-based models for ovarian cancer detection in primary care: a population-based external validation study
Kirsten D Arendse1, Fiona M Walter1,2, Gary Abel3
1Wolfson Institute of Population Health, Queen Mary University of London Barts and The London School of Medicine and Dentistry, London, United Kingdom.
Background:
Cancer antigen-125 (CA125) is widely used to investigate symptoms of possible ovarian cancer (OC) in primary care. However, OC risk varies with age and CA125 level. We externally validated the Ovatools models, which provide CA125- and age-specific OC risk.
Methods:
The performance of Ovatools in predicting OC diagnosis within 12 months of primary care CA125 was examined using English healthcare data for women <50 and ≥50 years. Discrimination and calibration were examined, accuracy was calculated at varying risk thresholds and compared to CA125 ≥ 35U/ml. We estimated OCs missed/detected by Ovatools in hypothetical diagnostic pathways, including a two-threshold pathway where moderate risk (1-2.9%) triggered primary care ultrasound, and higher risk (≥3%) triggered urgent cancer referral.
Results:
342,278 women were included, 0.63% had OC. The AUC was 0.95 in women ≥50 and 0.89 in women <50. When sensitivity/specificity were matched to CA125 ≥ 35U/ml, Ovatools showed marginally improved performance across other accuracy metrics in women ≥50 years. In a two-threshold pathway (≥50 years), 18.3% identified for urgent referral and 1% identified for ultrasound had OC.
Discussion:
Ovatools performed well on external validation. Ovatools could be used to support informed decision-making and to triage women for further investigation based on cancer risk.
