Background
Breast cancer is one of the most common cancers affecting women worldwide, with the number of new cases predicted to increase to over three million and the number of deaths to one million per year, by 2040. To prevent breast cancer, there is now increasing demand for personalization of breast cancer screening based on risk levels, considering the individual characteristics of the women.
Methods
We performed a synthesis of the evidence available for risk assessment models (RAMs) to predict breast cancer.
Results
We identified nine systematic reviews that evaluated risk assessment models for breast cancer. Across the systematic reviews, we identified five original models and 13 modifications that were externally validated. The evidence from primary studies that includes deep learning prediction models is limited. Most of the studies only report internal validation results in samples with small number of events.
Conclusions
Results from ongoing clinical trials and cohort studies will likely provide relevant empirical data. In addition, research on genotyping and, as shown in encouraging preliminary reports, research on deep-learning algorithm-based prediction models based on mammographic screening images should further improve the estimation of individual breast cancer risk.
MORGANO Gian Paolo;
GARCIA ESCRIBANO Marta;
PARMELLI Elena;
RIGGI Emilia;
JANUSCH ROI Annett;
BROEDERS Mireille;
CASTELLS Xavier;
FITZPATRICK Patricia;
FOLLMANN Markus;
GIORDANO Livia;
GRAEWINGHOLT Axel;
HOFVIND Solveig;
NYSTROM Lennarth;
QUINN Cecily;
SCHUNEMANN Holger;
AMATO Laura;
VECCHI Simona;
DE CRESCENZO Franco;
ALONSO COELLO Pablo;
RIGAU COMAS David;
CANELO-AYBAR Carlos;
NIETO-GUTIERREZ Wendy;
2024-09-24
Publications Office of the European Union
JRC138682
978-92-68-20459-7 (online),
978-92-68-20576-1 (print),
1831-9424 (online),
1018-5593 (print),
EUR 40013,
OP KJ-01-24-001-EN-N (online),
OP KJ-01-24-001-EN-C (print),
https://publications.jrc.ec.europa.eu/repository/handle/JRC138682,
10.2760/0359784 (online),
10.2760/2311237 (print),
