Table 5

Choice of the Cox regression model (N = 196)

Stepwise inclusion of variable

2 × increase in partial log-likelihood (χ2 distributed)

df

P


Stage

23.43

3

< 0.0001

Estrogen receptor

7.92

1

0.005

c-erbB2

7.34

1

0.01

Topoisomerase IIα

9.85

1

0.002

Grading

0.01

1

0.92

Topoisomerase IIα × grading

4.64

1

0.03

c-erbB2 × grading

1.74

1

0.19

c-erbB2 × topoisomerase IIα

0.17

1

0.68


Starting from the null model, successive adding of the variables mentioned above improves the Cox proportional hazard model in the first four steps significantly. In this sense the first four variables can be considered as independent prognostic factors. df, degrees of freedom.

Fritz et al. Breast Cancer Research 2005 7:R374-R384   doi:10.1186/bcr1012

Open Data