Five-year AUC with addition of previous mammogram images improves current visit mammogram AUC
By Elana Gotkine HealthDay Reporter
MONDAY, Dec. 9, 2024 (HealthDay News) — Five-year breast cancer (BC) risk prediction is improved with the addition of previous screening mammogram images beyond static models, according to a study published online Dec. 5 in JCO: Clinical Cancer Informatics.
Shu Jiang, Ph.D., from the Washington University School of Medicine in St. Louis, and colleagues analyzed a prospective cohort of 10,099 cancer-free women at entry (Nov. 3, 2008, through February 2012) from WashU. Participants were followed-up through 2020, and 478 pathology-confirmed BCs were identified. A dynamic model using repeated screening mammograms at WashU was trained to predict five-year risk. To assess discriminative performance, the model was applied to an external validation cohort (Emory), including 18,360 women screened from 2013 and followed through 2020, with 332 pathology-confirmed BCs.
The researchers found that a five-year area under the curve (AUC) of 0.80 was obtained in the external validation using three years of previous mammogram images. This represented a significant improvement over the AUC of 0.74 for the current visit mammogram, identified in the same women. Comparing high (>4 percent) to very low (<0.3 percent) five-year risk, a risk ratio of 21.1 was observed when calibrated. The dynamic model classified 16 percent of the cohort as high-risk, including 61 percent of women diagnosed with BC. Comparable performance was seen for the dynamic model in Black and White women.
“This dynamic model incorporates repeated mammograms for long-term risk prediction, showing that adding previous mammograms significantly improves five-year BC prediction,” the authors write.
Several authors disclosed ties to the biotechnology and medical technology industries.
Copyright © 2024 HealthDay. All rights reserved.
