Skip to main content
. 2017 Jun 27;2017:1678584. doi: 10.1155/2017/1678584

Table 3.

Results of simple and multiple Cox regression models for overall survival. Hazard ratios (HR) with 95% confidence intervals (CI) and p values (p) and proportions of explained variation (PEV). Full model PEV = 21.8%.

Variables Simple
HR (95%, CI)
p value PEV Multiple
HR (95%, CI)
p value PEV
Alb < 35 1.01 (0.61–1.68) 0.97 0.0 0.92 (0.54–1.58) 0.763 0.0
CRP > 1 1.63 (1.12–2.38) 0.011 1.6 1.47 (0.97–2.24) 0.070 0.3
N1 3.42 (2.01–5.82) <0.001 12.4 1.93 (0.98–3.77) 0.007 3.1
N2∗∗ 3.13 (2.21–4.43) <0.001 12.4 2.37 (1.50–3.74) <0.001 3.1
Nres∗∗∗ 0.90 (0.77–1.05) 0.18 0.3 0.70 (0.59–0.83) <0.001 3.2
nEAC∗∗∗∗ 1.85 (1.21–2.84) 0.005 4.2 1.03 (0.62–1.72) 0.903 1.2
nESCC∗∗∗∗ 0.82 (0.78–1.42) 0.482 4.2 0.85 (0.47–1.53) 0.589 1.2
One-stage surgery 0.56 (0.33–0.94) 0.027 1.5 0.80 (0.46–1.41) 0.447 0.1
UICC <0.001 13.5 0.013 3.0
I versus 0 0.59 (0.26–1.30) 0.186 0.77 (0.33–1.82) 0.555
II versus 0 2.03 (1.00–4.13) 0.052 1.94 (0.89–4.21) 0.093
III + IV versus 0 2.90 (1.45–5.84) 0.003 2.58 (1.08–6.14) 0.033

Effect of at least one pos. LK has time-dependent effect (p = 0.001 in simple model, p = 0.002 in multiple model). HR and 95% CI given are evaluated at 48 months; p value given refers to whole (time-dependent) effect, including neoadjuvant-treated and untreated cases. ∗∗N1 and N2 together also in “simple” model, including neoadjuvant-treated and untreated cases. ∗∗∗HR for log2-transformed variables quantify the effect of a doubling of the respective variable. ∗∗∗∗Neoadjuvant therapy and tumor biology (EAC and ESCC) exhibit a significant interaction (p = 0.021) in the model containing only these two variables but not in the multivariable model (p = 0.604). Thus, the neoadjuvant therapy effect is given separately for nEAC and nESCC.