Skip to main content

Table 3 Cox multivariate regression analysis for breast cancer survival according to CpG methylation profile, clinical factors, or intrinsic subtypes

From: DNA methylation profiling in the Carolina Breast Cancer Study defines cancer subclasses differing in clinicopathologic characteristics and survival

Prognostic variable Univariate Multivariatea
Hazard ratio 95% CI Pvalue Hazard ratio 95% CI Pvalue
Methylation cluster (n =517)       
4 (luminal-enriched) (n =187) (reference) 1.00 -   1.00 -  
1 (mixed) (n =123) 1.11 0.66-1.86 0.70 1.09 0.61-1.95 0.78
2 (basal-enriched) (n =108) 1.91 1.19-3.08 0.0075 1.41 0.76-2.64 0.28
3 (luminal-enriched) (n =99) 1.71 1.04-2.79 0.033 1.27 0.75-2.17 0.37
Clinical factor (n =517)       
Age at diagnosis (continuous) 0.97 0.96-0.99 0.004 0.99 0.96-1.01 0.38
Premenopausal (versus post) 1.76 1.20-2.57 0.004 1.27 0.72-2.22 0.41
African-American (versus white/other) 1.65 1.15-2.35 0.006 1.60 1.08-2.39 0.02
HR- (versus HR+) 1.58 1.09-2.30 0.02 1.07 0.65-1.76 0.80
HER2+ (versus HER2-) 1.32 0.89-1.96 0.16 - - -
Stage (1, 2, 3, 4) 2.76 2.22-3.43 <0.0001 1.74 1.22-2.49 0.002
Grade 2/3 (versus 1) 2.51 1.46-4.31 0.0009 1.21 0.66-2.22 0.54
Lymph node-positive (versus negative) 5.25 3.44-8.00 <0.0001 3.40 2.03-5.71 <0.0001
Tumor size 2-5 cm (versus ≤2 cm) 2.32 1.52-3.55 0.0001 1.24 0.77-1.99 0.38
Tumor size >5 cm (versus ≤2 cm) 5.01 2.91-8.63 <0.0001 1.44 0.73-2.83 0.28
Intrinsic subtype (n =393) b       
Luminal A (n =212) (reference) 1.00 -   1.00 -  
Luminal B (n =65) 1.14 0.62-2.09 0.68 0.97 0.51-1.84 0.93
Basal-like (n =86) 1.49 0.88-2.50 0.13 1.10 0.63-1.91 0.74
HER2+/HR- (n =26) 2.41 1.24-4.70 0.01 1.06 0.48-2.34 0.88
Unclassified (n =24) 1.34 0.57-3.17 0.50 1.19 0.49-2.87 0.70
  1. aMultivariate Cox proportional hazards regression models for methylation-based clusters were adjusted for age (continuous), menopausal status (pre/post), race, stage (1, 2, 3, 4), hormone receptor (HR) status, grade (1 versus 2 + 3), lymph node status, and tumor size. Multivariate Cox proportional hazards regression models for intrinsic subtypes were adjusted for age (continuous), menopausal status (pre/post), race, stage (1, 2, 3, 4), grade (1 versus 2 + 3), lymph node status, and tumor size. bThe reduced number of tumors included in models for intrinsic subtypes reflects missing data for subtype or other covariates. CI, confidence interval; HER2, human epidermal growth factor receptor 2.