Table 3.

Potential contributors to the elevated hazard of breast cancer-related death in Hispanic versus non-Hispanic white women

Variables included in Cox proportional hazards modelHispanic HR (95% CI)Proportion of excess Hispanic HR attributable to:*Proportion of parameter estimate (log HR) attributable to:
Model 1: age at diagnosis1.50 (1.24-1.83)
Model 2: age at diagnosis and method of detection1.45 (1.20-1.75)0.100.08
Model 3: age at diagnosis, method of detection, and census tract income, education and rural residence1.44 (1.11-1.86)0.020.02
Model 4: age at diagnosis, method of detection, tumor size, nodal status, histologic grade1.23 (1.01-1.50)0.540.46
  • *Using the unadjusted HR (HRU) and the HR after adjustment (HRA) for either method of detection or method of detection and tumor characteristics, the proportion of the excess hazard in Hispanic women attributable to the adjustment variable(s) can be calculated as: (HRU- HRA)/(HRU-1).

  • Using the log of the unadjusted HR (log HRU) (the β regression coefficient from a Cox proportional hazards model), and the log of the adjusted HR (log HRA), the proportion of the parameter estimate attributable to the adjustment variable can be calculated as: 1-(log HRA/log HRU).

  • The increased hazard of breast cancer-specific death associated with Hispanic ethnicity was not altered further by adjustment for estrogen and progesterone receptor status or first course of treatment.