Table III.
RSFa | Cox modelb | HR | Log rank testc | P | |
---|---|---|---|---|---|
Test error rate | P | 95% Confidence interval | |||
Okayama test set (n = 226) | |||||
29.9% | 0.0004 | ||||
Low risk | 1 | 0.002 | |||
Medium risk | 2.37 | 1.03–5.46 | |||
High risk | 3.75 | 1.70–8.28 | |||
Raponi test set (n = 129) | |||||
36.4% | 0.007 | ||||
Low risk | 1 | 0.02 | |||
Medium risk | 2.38 | 1.12–5.10 | |||
High risk | 2.79 | 1.32–5.90 | |||
Bild test set (n = 111) | |||||
35.6% | 0.002 | ||||
Low risk | 1 | 0.03 | |||
Medium risk | 1.54 | 0.78–3.02 | |||
High risk | 2.34 | 1.24–4.42 |
aRSF prediction model built from the 442 training set including 20 genes, age, gender and stage.
bMortality risk index (MRI) as continuous value, LRT was used in univariate Cox model.
cMRI separated test patients to three risk groups (low, medium and high-risk, one-third in each group).