906Can mammographic density add value to the Gail model in risk-stratifying women in BreastScreen Australia?

Abstract Background There is significant interest in personalised, risk-based breast cancer screening. This requires high quality risk assessment. The ‘Gail model’ risk assessment tool has been validated on over 40,000 BreastScreen Australia participants. We assess whether adding mammographic density (MD) information improves risk stratification on that cohort. Methods We used questionnaire data, baseline MD readings (using AutoDensity) and linked screening and cancer registry records from 40,158 BreastScreen Australia participants aged 50–69 years (via the lifepool cohort). We investigated incident invasive breast cancer rates by quintiles of Gail model scores, MD, and combinations of Gail and MD. Results Gail scores and MD values were weakly correlated (r≤0.02). Gail and MD were each strong predictors of incident breast cancer, but stronger predictors when used in combination. For example, the odds ratio for incident invasive breast cancer was 3.6 (95%CI 2.5-6.3) for the 17% of women in the upper two quintiles of both Gail and MD scores compared to the 17% of women in the lower two quintiles of both scores. In comparison, the odds ratio for breast cancer between same-size (each 17%) upper and lower groups for Gail score alone was 2.5 (95%CI 1.8-3.4), and for MD 1.9 (95%CI 1.2-2.9). Conclusions Combining Gail and MD categories improves risk stratification on BreastScreen Australia participants, compared to using Gail or MD alone. Key messages While questionnaire data and MD measures are each strong predictors of future invasive breast cancer among BreastScreen Australia participants, risk prediction is stronger when questionnaire and MD measures are combined.