Abstract PR07: Comparison of risk model recommendations for women at high-risk of breast cancer based on clinical thresholds using the Prospective Family Study Cohort (ProF-SC)

Background: Clinical guidelines for classifying women as high-risk for breast cancer when considering chemoprevention and/or MRI screening options include thresholds of remaining lifetime risk (RLR) of 20% or more and/or a fixed time interval (e.g., 5-year risk of 1.67 or higher, 10-year risk of 3.34 or higher). Although clinicians have noted differences in risk estimates from the existing risk models, there have been few prospective validations using large cohorts to describe the magnitude of the discordancies between these models. Methods: We prospectively followed 16,285 women without breast cancer at baseline for an average of 10.2 years to compare the RLR and 10-year risk assigned by two commonly used risk estimation models for high risk women: 1) The International Breast Cancer Intervention Study tool (IBIS); and 2) the Breast Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA). We compared the model-assigned 10-year risks with subsequent incidence of breast cancer in the cohort. We used chi-square statistics to assess calibration and the area under the receiver operating characteristic curve (AUC) to assess discrimination. Results: We observed differences between risk models in terms of the proportion of women classified as high-risk based on 20% or more RLR (IBIS=56% vs BOADICEA=23%). Only 21% of women were classified as high risk by both models, 35% of women were classified as high risk by IBIS only and 2% of women were classified as high risk by BOADICEA only. The difference was not evident (IBIS=52% vs BOADICEA=51%) when using a 10-year risk threshold of 3.34%. Using this 10-year threshold, 43% of women were classified as high risk by both models, 9% of women were classified as high risk by IBIS only and 8% of women were classified as high risk by BOADICEA only. IBIS risk predictions (mean=4.9%) were better calibrated to observed breast cancer incidence (5.8%, 95% confidence interval (CI)=5.4% to 6.2%) than were those based on BOADICEA (mean=4.2%). When we compared the magnitude of the discordancy between IBIS and BOADICEA by age, race/ethnicity, and number of relatives affected, we observed the extent of discordancy (e.g. one model resulted in a woman being above the clinical threshold when the other did not) depended on age. Specifically, for women under the age of 40 years, only 3.1% of women were high risk with IBIS but not BOADICEA compared with 7.5% classified as high risk by BOADICEA but not IBIS. Both models