Skip to main content
. Author manuscript; available in PMC: 2017 May 5.
Published in final edited form as: J Am Stat Assoc. 2016 May 5;111(513):27–42. doi: 10.1080/01621459.2015.1086354

Table 4.

Survival analysis in METABRIC: classification models trained in each single study and combined meta-framework are applied to METABRIC. P-value of survival differences of identified subgroups were evaluated based on log-rank test. The previously published PAM50 model was also compared. The number in () indicates the actual number of genes used in the prediction model since a few genes were not observed in the METABRIC array platform.

Model Number of Samples number of selected genes p value
Meta(TCGA+Wang+Desmedt) 533+260+164 203(194) 3.79 × 10−25
TCGA 533 239(233) 1.46 × 10−19
Wang 260 220(214) 3.31 × 10−14
Desmedt 164 197(193) 7.81 × 10−14
PAM50 50 1.01 × 10−20