Abstract P3-09-06: Predicting cardiovascular versus cancer mortality in a cohort of breast cancer survivors

Abstract

Abstract Background: Given improved survival after breast cancer diagnosis for women with non-metastatic disease, many will likely survive their disease and ultimately die from causes other than breast cancer, the most frequent being cardiovascular disease. There are numerous risk prediction models, such as the Framingham risk score, to identify persons who are at high risk for a cardiovascular event or death. However, these models have been developed for use in the general population and have not been validated in any cohorts of cancer survivors, who are at increased risk for competing causes of death. We evaluated commonly used risk models for cardiovascular events on a contemporary cohort of breast cancer survivors, and developed a new risk model to simultaneously predict the likelihood of death from breast cancer or cardiovascular disease (CVD). Methods: We included all women diagnosed with stage I-III breast cancer between January 1, 2000 and December 31, 2010 in Kaiser Permanente Northern California (KPNC) with follow-up through April 30, 2015. Specifically, we extracted from KPNC clinical and other databases: breast cancer characteristics, cardiovascular risk factors (cholesterol, blood pressure (BP), diabetes, BP lowering medication, smoking status), cardiovascular events, and cause of death. We assessed discrimination for the Framingham, CORE and SCOREOP cardiovascular risk models using the area under the receiver operating characteristic curve (AUC), and calibration by comparing the observed to the expected events. We used a multi-state model based on Cox cause specific hazards (CSH) to jointly model the risk of cardiovascular death and breast cancer death, while accounting for all other causes. Results: In this population of 20,462 KPNC breast cancer survivors with a median follow-up of 7.5 years, there were 695 cardiovascular and 842 breast cancer deaths. The existing cardiovascular risk models discriminated adequately (AUCs ranging 0.64 – 0.78), though models predicting cardiovascular mortality tended to over-predict, while those predicting non-fatal events tended to under-predict. Models developed to predict in a shorter time frame (<5 years), performed slightly better (E/O ratios of 1.08 and 1.18 for Framingham predicting events in the next 2 and 4 years, respectively). In our multi-state model, many of the traditional cardiovascular risk factors were no longer statistically significant (diabetes, BP) in predicting cardiovascular mortality, while the breast cancer characteristics (grade, tumor size, nodal involvement), as well as a prior history of CVD, were most useful in predicting cause of death. The model performed well, with AUCs of 0.85 (95% CI 0.83, 0.86) for 5-year risk of cardiovascular death and 0.85 (95% CI 0.84, 0.87) for breast cancer death. Conclusion: If replicated in an independent cohort, our model suggests that breast cancer characteristics can help predict overall mortality as well as cardiovascular death. Given the risk of cardiovascular death in the population of breast cancer survivors, joint modeling of breast and cardiovascular mortality is warranted. Citation Format: Leoce NM, Terry MB, Jin Z, Kushi LH, Roh JM, Laurent CA. Predicting cardiovascular versus cancer mortality in a cohort of breast cancer survivors [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P3-09-06.