Table 4. Multivariable survival analyses: model selection according to p-values and Bayes Information Criterion.
Step | Number of Predictors in Model | Breslow's Thickness | DCCD | Ulceration | Age | Nodal Statusa | Localization | Gender | BIC |
1 | 7 | <0.0001 | 0.0068 | 0.0009 | 0.0051 | 0.0371 | 0.1380 | 0.8052 | 1,657.8 |
2 | 6 | <0.0001 | 0.0069 | 0.0008 | 0.0050 | 0.0340 | 0.1035 | — | 1,650.9 |
3 | 5 | <0.0001 | 0.0011 | 0.0008 | 0.0054 | 0.0475 | — | — | 1,646.6 |
4 | 4 | <0.0001 | <0.0001 | 0.0005 | 0.0127 | — | — | — | 1,643.6 |
5 | 3 | <0.0001 | <0.0001 | 0.0002 | — | — | — | — | 1,642.9 |
6 | 2 | <0.0001 | <0.0001 | — | — | — | — | — | 1,649.8 |
7 | 1 | <0.0001 | — | — | — | — | — | — | 1,679.1 |
p-Values are for testing whether a hazard ratio equals 1; low BIC identifies best model.
As determined by routine histopathology.