Table 1.

Multivariate Cox regression analysis of methylation gene panel, TNM stage, and survival in training, validation, and independent cohorts

VariableHR (95%CI)P valuea
DFS
 Training cohort
  TNM stage2.05 (1.37–3.07)<0.001
  Methylation gene panel2.26 (1.28–4.01)0.005
 Validation cohort
  TNM stage1.99 (1.34–2.98)0.001
  Methylation gene panel2.08 (1.17–3.68)0.013
 Independent cohort
  TNM stage2.73 (1.74–4.27)<0.001
  Methylation gene panel1.94 (1.08–3.47)0.026
OS
 Training cohort
  TNM stage2.66 (1.63–4.34)<0.001
  Methylation gene panel2.47 (1.30–4.71)0.006
 Validation cohort
  TNM stage2.32 (1.52–3.56)<0.001
  Methylation gene panel1.83 (1.01–3.31)0.046
 Independent cohort
  TNM stage2.35 (1.51–3.66)<0.001
  Methylation gene panel2.09 (1.10–3.98)0.022
  • aMultivariate Cox proportional hazard models were adjusted for methylation gene panel (high methylation vs. low methylation), sex, age (<45 years vs. ≥45 years), TNM stage (IV vs. III vs. II vs. I), World Health Organization pathologic type (undifferentiated nonkeratinizing vs. differentiated nonkeratinizing), and concurrent chemotherapy (yes vs. no) as covariates for each analysis. Variables were selected using the backward stepwise approach.