Fig 1.
Flow diagram of the data selection procedure.
Table 1.
Patient characteristics for all patients, patients who survive and patients who die due to BC or other causes (in %), for analysis comparing molecular subtypes (N = 36 412 patients).
Table 2.
Patient characteristics for all patients, patients who survive and patients who die due to breast cancer or other causes (in %), for analysis comparing BMI grouped patients (N = 18 778 patients with registered height and weight (BMI)).
Table 3.
Number of events, person-years at risk, and incidence rates for everybody and for patients in the molecular subtype groups, for the events BCSD (breast cancer specific death) and OCOD (other cause of death).
Fig 2.
Cumulative incidences for patient-, BMI-, age- and tumor stage groups for patients who died due to breast cancer (blue lines) and due to other causes (yellow), with 95% confidence bands.
Table 4.
Hazard ratio estimates for BCSD, with OCOD as a competing risk. Results are considered significant at p < 0.05.
Fig 3.
Estimated hazard ratios over time for breast cancer specific death for factors entering the model (baseline: molecular subtype HR + /HER2-, age 40−52 years, tumor stage I, no radiation therapy).
The blue line is the hazard ratio fitted by a smoothing spline over time, and the gray area shows a + /- 2-standard-error band around this blue line. The turquoise line is the estimated time invariant hazard rate (HR) in a Cox-model when we ignore the possibility of time varying coefficients. For each level of the factors, we compute the Schoenfeld residuals over time, and the p-value obtained from the Schoenfeld test for each level of the factors is displayed.