Breast cancer mortality of older patients with and without recurrence analysed by novel multi-state models.

In older patients with breast cancer, the risk of dying from other causes than breast cancer strongly increases after the age of 70. The aim of this study was to assess contributions of breast cancer mortality versus other-cause mortality after locoregional or distant recurrence in a population-based cohort of older patients analysed by multi-state models. Surgically treated patients ≥70 years diagnosed with stage I-III breast cancer in 2003–2009 were selected from the Netherlands Cancer Registry. A novel multi-state model with locoregional and distant recurrence that incorporates relative survival was fitted. Other-cause and breast cancer mortality were indicated as population and excess mortality. Overall, 18,419 patients were included. Ten-year cumulative incidences of locoregional and distant recurrence were 2.8% (95%CI 2.6–3.1%) and 12.5% (95%CI 11.9–13.1%). Other-cause mortality increased from 23.9% (95%CI 23.7–24.2%) in patients 70–74 years to 73.8% (95%CI 72.2–75.4%) in those ≥80 years. Ten-year probabilities of locoregional or distant recurrence with subsequent breast cancer death were 0.4–1.3% and 10.2–14.6%, respectively. For patients with a distant recurrence in the first two years after diagnosis, breast cancer death probabilities were 95.3% (95%CI 94.2–96.4%), 93.1% (95%CI 91.6–94.6%), and 88.6% (95%CI 86.5–90.8%) in patients 70–74, 75–79, and ≥80 years. In older patients without recurrence, prognosis is driven by other-cause mortality. Although locoregional recurrence is a predictor for worse outcome, given its low incidence it contributes little to breast cancer mortality after diagnosis. For patients who develop a distant recurrence, breast cancer remains the dominant cause of death, even at old age. • Older patients with breast cancer often die from competing causes. • Competing mortality after recurrence has not been extensively studied so far. • Novel multi-state models were used including time after recurrence. • These models can be applied to different datasets as input for clinical decision tools. [ABSTRACT FROM AUTHOR]