Development and validation of a synthetic risk model for stratified disease prevention for breast cancer

Montserrat GARCIA-CLOSAS, National Cancer Institute, United States
BROOK M. Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK , MAAS P. Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, USA , ORR N. The Breakthrough Breast Cancer Research Centre, The Institute of Cancer Research, London, UK , COULSON P. Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK , SCHOEMAKER M. Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK , JONES M. Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK , SWERDLOW A. Division of Genetics and Epidemiology, The Institute of Cancer Research, London, UK , CHATTERJEE N. Biostatistics Department, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD

1 Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, USA

The risk-benefit balance of disease prevention strategies largely depends on the underlying risk of developing the disease. Therefore, risk prediction tools that can accurately stratify a population into categories with sufficiently distinct risks are critical to identify those most likely to benefit.  With the increasing relevance of genetic testing for assessment of disease it is important for these tools to integrate information on genetic and environmental risk factors for disease. We developed a flexible risk model-building tool (iCARE) and used it to build a synthetic risk model for prediction of breast cancer risk in the general population. The model included information on age, reproductive and hormone use history, benign breast disease, lifestyle, family history and a polygenic risk score (PRS) based on 77 single nucleotide polymorphisms (SNPs). Predictions were based on estimates of risk factor relative risks reported in the literature, and risk factor frequencies in the UK population. UK age-specific incidence and mortality rates were used to obtain estimates of absolute risk. Model calibration for relative and absolute risk predictions was assessed in a UK cohort of 106,637 women with a median age of 48 years who were included in the Breakthrough Generations Study. During an average follow-up period of 6.1 years, 1,420 women developed invasive breast cancer.  The synthetic model showed good calibration for 5-year risk and had an AUC of 68% (95%CI 63-71%) for women 50 years of age or younger and 66% (95%CI 63%-69%) for women older than 50 years. This level of risk stratification could have utility to better inform decisions on risk reduction strategies for breast cancer. The flexible modelling approach we developed can be easily extended to include additional risk factors or to build models for other diseases