Table 3. Prognostic significance of the common hypoxia metagene versus other hypoxia signatures.
| Data (Table 1) | End point and significant clinical covariates (Cov.)a | In vitro hypoxia signature (Chi et al, 2006) | HN hypoxia metagene (Winter et al, 2007) | Initial seedsb | PCA scorec | CHM 51genes | Reduced CHMd k genes |
|---|---|---|---|---|---|---|---|
| NKI | End point: MFS Cov.: Age, tumour size, nodal status, grade, adj. treatment | 2.94 (1.39, 6.23) P=0.005 | 3.58 (1.53, 8.39) P=0.003 | 2.41 (1.05, 5.53) P=0.038 | 3.22 (1.37, 7.56) P=0.007 | 4.15 (1.73, 9.96) P=0.002 | 5.58 (2.41, 12.90) P<0.001, k=3 |
| GSE2034e | End point: RFS Cov.: NA | 2.20 (1.11, 4.34) P=0.024 | 1.92 (0.97, 3.78) P=0.061 | 2.36 (0.95, 3.77) P=0.014 | 1.98 (1.01, 3.90) P=0.048 | 3.22 (1.63, 6.35) P=0.001 | 4.15 (2.10, 8.18) P<0.001, k=10 |
| GSE3494e | End point: DSS Cov.: ER, PgR, tumour size, nodal status | 1.19 (0.45, 3.13) P=0.732 | 2.07 (0.77, 5.53) P=0.149 | 2.87 (1.25, 4.49) P=0.029 | 3.61 (1.33, 9.82) P=0.012 | 3.16 (1.05, 9.53) P=0.042 | 4.27 (1.53, 11.94) P=0.006, k=2 |
| Chung | End point: RFS Cov.: Intrinsic sign., differentiation, batch (strata) | 3.06 (0.53, 17.6) P=0.210 | 14.83 (1.8, 122.4) P=0.012 | 6.71 (0.93, 48.4) P=0.059 | 1.25 (0.14, 11.4) P=0.840 | 6.25 (0.83, 47.2) P=0.077 | 34.66 (4.26, 281.95) P=0.001, k=2 |
| Beer | End point: OS Cov.: Stage | 2.59 (1.59, 4.2) P=0.829 | 6.90 (1.34, 35.6) P=0.021 | 3.98 (0.72, 22.0) P=0.114 | 3.45 (0.59, 20.0) P=0.168 | 12.84 (1.71, 96.5) P=0.014 | 24.57 (2.83, 213.36) P=0.004, k=23 |
| GSE4573 | End point: OS Cov.: Nodal status | 3.15 (1.32, 7.54) P=0.010 | 1.49 (0.65, 3.43) P=0.350 | 2.31 (0.93, 5.72) P=0.070 | 1.61 (1.14, 2.3) P=0.035 | 2.75 (1.15, 6.56) P=0.023 | 2.90 (1.27, 6.61) P=0.012, k=38 |
Abbreviations: CHM=common hypoxia metagene; DSS=disease-specific survival; ER/PgR=estrogen/progesterone receptor; MFS=metastases-free survival; RFS=recurrence-free survival; OS=overall survival.
a
Reduced models of clinical covariates are derived using backward-stepwise likelihood. Signature scores are entered into the reduced model; hazard ratio, 95% confidence limits and significance (model with and without the signature) are shown.
b
Summary score, E, is calculated for the signature including only the initial seeds.
c
Score obtained using principal components analysis (Supplementary Methods).
d
At convergence in the cumulative forest plots.
e
These two data sets were used to develop the signature but no training on outcome was carried out.