1. Breast Cancer Surveillance Consortium (BCSC): A Research Infrastructure sponsored by the National Cancer Institute Breast Cancer Risk Models William Barlow, PhD Senior Biostatistician, Cancer Research and Biostatistics Research Professor, Dept. of Biostatistics

2.  Gail model (Gail et al., JNCI, 1989)  Claus model (several references)  BRCAPRO (many references) Breast cancer risk models in 1994

3.  Population-based screening mammography population  Prospectively collected risk factor information coded in an uniform manner  Huge longitudinal dataset that is increasing  Can examine the very important role of breast density as a risk factor, available on almost all screening mammograms Advantage of BCSC data for estimating risk

4.  Barlow et al., JNCI, 2006 (77 citations)  Short-term prediction of a diagnosis of breast cancer  Tice, et al., Annals Int Med, 2008 (17 citations) Long-term prediction of a diagnosis of breast cancer Simplified model to be used in clinical care  Janes, Pepe, and Gu, Annals Int Med, 2009 (17 citations) Compares the change in risk prediction from the Gail model to the Tice model Statistical technique for evaluating how well a model improves on past models for identifying high risk individuals BCSC breast cancer risk models

5.  Published in the same JNCI issue as the revised Gail model which also includes breast density  Based on one million women and 2.4 million screening mammograms  Published in the same JNCI issue as the revised Gail model which also includes breast density BCSC risk prediction model

6.  Prospective collection of self-reported risk factors:  demographic information (age, race, ethnicity)  family history of breast cancer  previous breast procedures  menopausal status and use of hormone therapy  body mass index (BMI)  Breast density (BI-RADS scale) and mammographic assessment reported by the radiologist  Outcome is a diagnosis of breast cancer within one year of the screening mammogram  Invasive breast cancer  DCIS BCSC risk factors & outcomes

7.  Available online along with the risk program and documentation:  http://breastscreening.cancer.gov/ http://breastscreening.cancer.gov/  Dataset has been downloaded 228 times  What’s not included?  Mammographic outcomes; location; individual-level data  What’s missing?  genetic markers  in-depth family history assessment  environmental factors  dietary factors (other than BMI)  exercise levels BCSC risk data

8.  Prediction of a diagnosis of breast cancer within a year of a screening mammogram  Extrapolated to 5-years for comparison to the Gail model, but not intended for long-term prediction  Most Gail risk factors are included  Adds breast density to the risk model Purpose of the risk model

9. Observed incidence by age and breast density

10.  Risk factors for pre-menopausal and post-menopausal women are different  Required that a factor be significant to the 0.0001 level before including it  Goal was to keep the model as simple as possible  Develop the model on 75% and validate it on the remaining 25%  Good calibration in the validation sample (correct estimation of cancer rates)  Good prediction of individual outcomes (c-statistic) Risk model factors and assessment

11. Risk Model for Premenopausal Women Risk Factor LevelsOR 95% CI Age in years (versus Age 35-39) (versus Age 35-39) Age 40-441.23 0.96 – 1.56 Age 45-491.891.49 – 2.41 Age 50-54 2.361.84 – 3.02 Prior breast procedure (versus No procedure) (versus No procedure) Yes1.471.31 – 1.65 Unknown0.870.69 – 1.10 First degree family hx (versus No family hx) (versus No family hx) Yes, 1 relative1.541.36 – 1.74 Yes, 2 or more relatives2.111.32 – 3.36 Unknown0.890.75 – 1.07 Breast density (BI-RADS) 1: Almost entirely fat 1.00- 2: Scattered fibroglandular densities2.001.26 – 3.18 3: Heterogeneously dense3.342.12 – 5.27 4: Extremely dense3.932.47 – 6.28 Unknown/different system 3.312.09 – 5.24 c = 0.63

12. Postmenopausal Predictors (c=0.62)  Age  Race  Hispanic ethnicity  Body mass index (BMI)  Age at first birth  Family history of breast cancer  Previous breast surgery  Menopausal status  Type of menopause  Hormone therapy use  Breast density  Previous screening result (false positive/true negative)

13. Five-year risk of invasive breast cancer based on extrapolation of the model: High and low risk women

14. Risk model summary  This particular model did somewhat better than the Gail model  All risk models have difficult getting beyond a c-statistic of 0.65  Still cannot identify a group of women who are risk-free  Risk models may identify high-risk groups, but still have difficulty with precise estimation of risk for a particular woman

15. Clinical implications  Breast density was a known risk factor, but we did not appreciate its importance relative to other risk factors  Can better estimate risk for prevention efforts  Breast density may be an early marker of a reduction in risk

16. Next steps  Augment the screening mammogram assessment with risk factor information to better estimate outcomes from positive or negative screening mammograms  Use changes in risk factors to develop better models of long-term risk prediction  Incorporate extensive family history and and SNPs into risk models with breast density (P01 application by BCSC)