Table 4 Predictive ability of Cox regression models for progression-free and overall survival.

From: Combining radiomic phenotypes of non-small cell lung cancer with liquid biopsy data may improve prediction of response to EGFR inhibitors

Modeling covariates

C-statistic (95% CI)

p versus null1

p versus model without phenotype2

Progression-free survival

Number of mutations

0.50 (0.37–0.63)

0.97

 

Smoking status

0.63 (0.47–0.78)

0.09

 

ECOG performance score

0.69 (0.58–0.8)

< 0.005

 

Radiomic phenotype

0.63 (0.49–0.77)

0.03

 

Number of mutations, smoking status, and ECOG performance score

0.73 (0.59–0.86)

< 0.005

 

Radiomic phenotype, number of mutations, smoking status, and ECOG performance score

0.77 (0.64–0.89)

< 0.005

0.01

Overall survival

Number of mutations

0.55 (0.31–0.8)

0.46

 

Smoking status

0.69 (0.49–0.89)

0.02

 

ECOG performance score

0.71 (0.55–0.88)

0.02

 

Radiomic phenotype

0.62 (0.39–0.85)

0.11

 

Number of mutations, smoking status, and ECOG performance score

0.80 (0.61–0.98)

0.01

 

Radiomic phenotype, number of mutations, smoking status, and ECOG performance score

0.83 (0.67–1)

< 0.005

0.08

  1. 1p value by likelihood ratio test versus the hypothesis that the model is no better than the null model, in which all patients are at equal risk.
  2. 2p value by likelihood ratio test versus the hypothesis that the model is no better than the same model without radiomic phenotype.