Extended Performance of an Image-Based Short-Term Risk Model for Predicting Breast Cancer

Get Permission

In a Swedish case-cohort study reported in the Journal of Clinical Oncology, Eriksson et al found that an image-based short-term risk model outperformed a traditional lifestyle/familial risk–based risk model in predicting the likelihood of development of breast cancer over a 10-year period.

Study Details

The case-cohort study involved a randomly selected subcohort of 8,604 patients from 66,814 individuals in a mammography screening cohort initiated in 2010 in Sweden for women aged 40 to 74 years and a total of 2,028 incident breast cancers in the entire cohort (206 in the subcohort) identified through May 2022.

The image-based model uses age and the mammographic features of density, microcalcifications, masses, and left-right breast asymmetries from study entry mammograms to estimate risk. The Tyrer-Cuzick v8 risk model uses self-reported lifestyle and familial risk factors and mammographic density to estimate risk. Absolute risks were estimated, and age-adjusted receiver operating characteristic area under the curve (aAUC) values for the two models were compared across a 10-year period.

Key Findings

The discriminatory performance of the image-based risk model ranged from aAUC of 0.74 (95% confidence interval [CI] = 0.70–0.78) after 1 year to aAUC of 0.65 (95% CI = 0.63–0.66) after 10 years. By comparison, the performance of the Tyrer-Cuzick model ranged from aAUC of 0.62 (95% CI = 0.56–0.67) at 1 year to 0.60 (95% CI = 0.58–0.61) at 10 years. The aAUC point estimate differences between the image-based and Tyrer-Cuzick models ranged from 0.12 at 1 year (P < .01) to 0.05 at 10 years (P < .01).

For symptomatic cancers, the aAUC values for the image-based model were ≥ 0.75 during the first 3 years of follow-up, whereas aAUC values for the Tyrer-Cuzick model were ≤ 0.65 over the first 3 years.

Among patients with high mammographic density, aAUC values ranged from 0.69 at 1 year to 0.64 at 10 years on the image-based model, and from 0.58 to 0.56 on the Tyrer-Cuzick model. The corresponding values among those with low mammographic density were 0.75 to 0.64 vs 0.64 to 0.60.

Over the 10-year follow-up, 20% of all patients who developed breast cancer were assessed as high-risk at study entry on the image-based risk model vs 7.1% on the Tyrer-Cuzick model (P < .01).

The investigators concluded, “The image-based risk model outperformed the Tyrer-Cuzick v8 model for both short-term and long-term risk assessment and could be used to identify women who may benefit from supplemental screening and risk reduction strategies.”

Mikael Eriksson, PhD, of the Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, is the corresponding author for the Journal of Clinical Oncology article.

Disclosure: The study was supported by the Kamprad Foundation, Stockholm County council, Swedish Cancer Society, and others. For full disclosures of the study authors, visit

The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.