Table 2. Multi-gene Cox Proportional Hazards models for all samples and specific subtypes using three different model selection methods: Stepwise regression with age as a covariate (SW with age), stepwise regression without age (SW no age), and Elastic Net.
| HR (LB,UB) | ||||
| Model | Term | SW with age | SW no age | Elastic Net |
| All | Age | 1.03 (1.02,1.04) | NI | 1.02 |
| MX1 | 1.29 (1.06,1.57) | 1.44 (1.19,1.75) | 1.09 | |
| IFIT1 | 0.88 (0.74,1.04) | 0.79 (0.67,0.94) | NI | |
| R2 | 19% | 7% | 31% | |
| Proneural | Age | 1.04 (1.02,1.06) | NI | 1.03 |
| MX1 | 2.12 (1.13,3.97) | 2.44 (1.48,4.03) | 1.88 | |
| IFIT1 | 0.42 (0.23,0.76) | 0.6 (0.4,0.92) | 0.64 | |
| IFI44 | 1.82 (0.97,3.41) | 2.3 (1.28,4.15) | 1.5 | |
| USP18 | 0.46 (0.23,0.95) | 0.36 (0.19,0.68) | 0.6 | |
| ISG15 | 1.93 (0.77,4.81) | NI | 1.27 | |
| OAS1 | NI | NI | 1.02 | |
| R2 | 61% | 47% | 82% | |
| Neural | Age | 1.06 (1.01,1.1) | NI | 1.03 |
| USP18 | 4.06 (1.47,11.19) | 1.65 (1.04,2.62) | 1.3 | |
| MX1 | 0.48 (0.22,1.05) | NI | NI | |
| IFIT1 | NI | NI | 1.06 | |
| IFI44 | NI | NI | 1.05 | |
| R2 | 37% | 14% | 44% | |
| Classical | Age | 1.01 (0.99,1.04) | ||
| R2 | 3% | |||
| Mesenchymal | Age | 1.05 (1.02,1.08) | 1.03 | |
| R2 | 15% | 24% |
Hazard ratios with confidence limits are given for each term added to each model. If the term was not added to a given model, NI is displayed for “not included”. The total explained variance for each model (R2) is also displayed.