Real-world impact of non–breast cancer–specific death on overall survival in resectable breast cancer

Abstract

BACKGROUND: The real-world occurrence rate of non–breast cancer–specific death (non-BCSD) and its impact on patients with breast cancer are poorly recognized. METHODS: Women with resectable breast cancer from 1990 to 2007 in the Surveillance, Epidemiology, and End Results database (n = 199,963) were analyzed. The outcome events of breast cancer were classified as breast cancer–specific death (BCSD), non-BCSD, or survival. Binary logistics was used to estimate the occurrence rates of non-BCSD and BCSD with different clinicopathological factors. The Gray method was used to measure the cumulative incidence of non-BCSD and BCSD. The ratio of non-BCSDs to all causes of death and stacked cumulative incidence function plots were used to present the impact of non-BCSD on overall survival (OS). Models of Cox proportional hazards regression and competing risk regression were compared to highlight the suitable model. RESULTS: There were 12,879 non-BCSDs (6.44%) and 28,784 BCSDs (14.39%). The oldest age group (>62 years), black race, and a single or divorced marital status were associated with more non-BCSDs. With adjustments for age, a hormone receptor–positive (HoR+) status was no longer related to increased non-BCSDs. In patients with grade 1, stage I disease and an HoR+ status as well as the oldest subgroup, a great dilution of non-BCSD on all causes of death could be observed, and this led to incorrect interpretations. The inaccuracy, caused by the commonly used Cox proportional hazards model, could be corrected by a competing risk model. CONCLUSIONS: OS was largely impaired by non-BCSD during early breast cancer. For some future clinical trial planning, especially for the oldest patients and those with HoR+ breast cancer, non-BCSD should be considered a competing risk event. Cancer 2017;123:2432–43. © 2017 American Cancer Society.