Swiss Personalized Breast Cancer Risk Prediction study
Department of Computational Medicine and Bioinformatics, & Michigan Institute for Data Science, University of Michigan, USA (Ivo D. Dinov) | Unit of Oncogenetics and Cancer Prevention, Geneva University Hospitals, Geneva, Switzerland (Pierre O. Chappuis & Valeria Viassolo) | Swiss Tropical and Public Health Institute, Basel, Switzerland (Nicole Probst-Hensch) | Radio-oncology, University Hospital Basel, Basel, Switzerland (Sophie Dellas)
Ort der Datenerhebung
– Geneva University Hospitals
– Klinik für Radiologie und Nuklearmedizin, Universitätsspital Basel
– Oncology clinic in the Hopital de Delémont
– Michigan CDC
2016 bis 2019
Breast cancer affects about 12% of Swiss women. Predictive models are important in personalized medicine because they contribute to early identification of high-risk individuals, which in turn facilitates stratification of preventive interventions and individualized clinical management. However, existing models have limited discriminatory accuracy (0.6-0.7) and do not include some non-modifiable and modifiable breast cancer risk factors, e.g., mammography density and obesity.
The purpose of the study is to provide clinical decision support for accurate, reproducible, and more reliable individualized forecasting of the absolute risk for breast cancer compared to currently used models e.g., Gail model and Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA).
We employed six different model-free machine-learning methods to predict absolute risk of breast cancer. Using independent training and testing data we quantified and compared the performance of machine-learning methods to the performance of the Gail model and BOADICEA using the following datasets (1) simulated, with no signal; (2) simulated, with artificial signal; (3) a random population-based sample of US breast cancer patients and their cancer-free female relatives (N=1232); and (4) a clinic-based sample of Swiss breast cancer patients and cancer-free women seeking genetic evaluation and/or testing at the Geneva University Hospitals (N=1700). Managing the massive, multi-source, incongruent and heterogeneous data includes data harmonization, model-free predictive analytics, and quantitative comparison of forecasting reliability.
Erwarteter Nutzen / Relevanz
Advanced data-processing protocols are powerful tools to forecast personalized breast cancer risk and can help develop new and updated predictive models specified for Swiss women.