Abstract Endogenous hormones have been associated with postmenopausal breast cancer risk and may improve our ability to identify women at high risk of breast cancer. In addition, having high levels of multiple sex hormones or prolactin appears to further increase risk of postmenopausal breast cancer. This suggests that including multiple hormones may improve risk prediction models more than adding one or two hormones individually. Therefore, we evaluated whether the inclusion of multiple hormones (estradiol, estrone, estrone sulfate, testosterone, DHEAS, prolactin, SHBG) improved risk prediction models of postmenopausal invasive breast cancer among 473 cases and 770 controls not using postmenopausal hormones at blood draw. We utilized two previously published breast cancer risk prediction models, the Gail model and the Rosner model. We evaluated the improvement of the AUC for 5-year risk of invasive breast cancer, and secondarily of estrogen receptor (ER) positive disease, using the Gail and Rosner risk scores for each hormone individually. Then we used step-wise regression to identify the subset of hormones that were associated most significantly with risk and assessed improvement in AUC. We used 70% of the data as a training set (n=314 cases/533 control) and 30% as an independent validation set (n=123 cases/240 control). In preliminary analyses, each hormone was associated with risk of invasive breast cancer (OR, doubling in levels=0.82 [SHBG] to 1.45 [testosterone]) in the training set. For invasive cancer, individual hormones improved the AUC by 2.0-7.6 units when included with the Gail score and 0.2-3.9 units for the Rosner score. Estrone sulfate, testosterone, and prolactin were selected by step-wise regression and together increased the AUC by 8.6 points (p<0.001) for the Gail and 3.8 units (p=0.06) for the Rosner score. In the independent set, the corresponding change in the AUC was 5.0 (p=0.09) and 5.1 (p=0.11) units, respectively. Similar results were observed for ER+ disease specifically (223 cases). In the training dataset, the best hormone subset included estrone sulfate, testosterone, prolactin, and SHBG, leading to a change in AUC=7.3 (p=0.004) for the Gail and 5.5 (p=0.01) for the Rosner score. The change in AUC was higher in the independent datasets (12.5 [p=0.006] and 8.3 [p=0.06], respectively). Our results strongly support that endogenous hormones can improve risk prediction for invasive breast cancer and may help identify women who would benefit from chemoprevention. The improvement in the AUC observed when adding circulating hormone levels was similar or better than prior studies that considered adding genetic risk factors or mammographic density, and since hormone levels are not strongly correlated with these other risk factors, future studies should consider simultaneous inclusion of all these factors. Citation Format: Shelley S. Tworoger, Xuehong Zhang, A. Heather Eliassen, Jing Qian, Walter C. Willett, Graham A. Colditz, Bernard A. Rosner, Peter Kraft, Susan E. Hankinson. Inclusion of endogenous hormone levels in risk prediction models of postmenopausal breast cancer. [abstract]. In: Proceedings of the Twelfth Annual AACR International Conference on Frontiers in Cancer Prevention Research; 2013 Oct 27-30; National Harbor, MD. Philadelphia (PA): AACR; Can Prev Res 2013;6(11 Suppl): Abstract nr C21.
Read full abstract