Table 2 Final Cox model fitting on the sample data. 95% CI is calculated by exp(coef ± 1.96 × se).

From: Identification of CT-based non-invasive radiomic biomarkers for overall survival prediction in oral cavity squamous cell carcinoma

Factor

Coef.

H.ratio

Se

95% CI

p value

A. Overall survival

 Radiomics

  HLEs

0.259

1.29

0.103

[1.05, 1.58]

0.014

  GNS

− 0.062

0.94

0.024

[0.90, 0.98]

0.009

  Clinical

 Stage

0.313

1.37

0.121

[1.08, 1.73]

0.009

  ETOH:2

− 0.611

0.54

0.319

[0.29, 1.01]

0.055

  ETOH:3

− 0.280

0.76

0.443

[0.32, 1.80]

0.527

 Goodness of fit

  LRT test

38.58

   

3e−07

  Wald test

33.80

   

4e−04

  Score test

36.94

   

6e−07

B. Absolute survival

 Radiomics

  HLEs

0.274

1.32

0.116

[1.05, 1.65]

0.019

  GNS

− 0.066

0.94

0.031

[0.88, 0.99]

0.032

  Stage

0.343

1.41

0.169

[1.01, 1.96]

0.043

 Clinical

  ETOH:2

− 0.456

0.63

0.357

[0.31, 1.28]

0.202

  ETOH:3

− 1.356

0.26

1.029

[0.03, 1.94]

0.188

 Goodness of fit

  LRT test

32.82

   

4e−06

  Wald test

26.35

   

8e−05

  Score test

30.92

   

1e−05

  1. Both tests yielded significant p-values across all folds, with average p-values of 3e−06 and 6e−06 for the log-likelihood ratio and the score test, respectively, suggesting that our model is highly significant and provides a good fit to the data. The training concordance index (CI) remained stable and high across all iterations, with an average value of 0.727 (SD = 0.011). The testing CI has a mean value of 0.699 and a standard deviation of 0.103. In a simlar vien, the average training CI for absolute survival is 0.751 and testing CI is 0.706.
Back to article page