Mammographic breast density refines Tyer-Cuzick estimates of breast cancer risk: findings from the placebo arm of the International Breast Cancer Intervention Study I

Introduction Mammographic density is well-established as a risk factor for breast cancer, however, adjustment for age and body mass index (BMI) is vital to its clinical interpretation when assessing individual risk. In this paper we develop a model to adjust mammographic density for age and BMI and show how this adjusted mammographic density measure might be used with existing risk prediction models to identify high-risk women more precisely. Methods We explored the association between age, BMI, visually assessed percent dense area and breast cancer risk in a nested case-control study of women from the placebo arm of the International Breast Cancer Intervention Study I (72 cases, 486 controls). Linear regression was used to adjust mammographic density for age and BMI. This adjusted measure was evaluated in a multivariable logistic regression model that included the Tyrer-Cuzick (TC) risk score, which is based on classical breast cancer risk factors. Results Percent dense area adjusted for age and BMI (the density residual) was a stronger measure of breast cancer risk than unadjusted percent dense area (odds ratio per standard deviation 1.55 versus 1.38; area under the curve (AUC) 0.62 versus 0.59). Furthermore, in this population at increased risk of breast cancer, the density residual added information beyond that obtained from the TC model alone, with the AUC for the model containing both TC risk and density residual being 0.62 compared to 0.51 for the model containing TC risk alone (P =0.002). Approximately 16% of controls and 19% of cases moved into the highest risk group (8% or more absolute risk of developing breast cancer within 10 years) when the density residual was taken into account. The net reclassification index was +15.7%. Conclusions In women at high risk of breast cancer, adjusting percent mammographic density for age and BMI provides additional predictive information to the TC risk score, which already incorporates BMI, age, family history and other classic breast cancer risk factors. Furthermore, simple selection criteria can be developed using mammographic density, age and BMI to identify women at increased risk in a clinical setting. Clinical trial registration number http://www.controlled-trials.com/ISRCTN91879928 webcite (Registered: 1 June 2006).