Event-free survival analysis
The Lasso penalized Cox regression analysis identified as potential
predictors of EFS: standard deviation (p =0.0194) at medium scale
(SSF = 4), mean gray-level (p =0.0411) at coarse scale (SSF = 6),
mean gray-level (p =0.0040) at medium scale (SFF=0), mean
gray-level (p =0.0180) at medium scale (SFF=3) and age at
diagnosis (p =0.0732). In multivariate Cox regression analysis,
mean gray-level at coarse scale (SSF = 6) (HR = 1.06 [1.02; 1.12],p =0.006), standard deviation at medium scale (SSF = 4) (HR=1.03
[1.003; 1.07], p =0.03) and age at diagnosis were
significantly associated with EFS (HR = 1.11 [1.05; 1.17],p =0.002) (Table 2B).
When dichotomized at the optimal threshold identified using ROC curve,
mean gray-level greater than -7.63 at coarse scale (SSF = 6) (HR = 2.4
[1.0; 5.9], p =0.06, figure 2B) and standard deviation greater
than 32 at medium scale (SSF = 4) (HR = 2.6 [0.9; 7.2], p =
0.07, figure 2C) tended to be associated with lower survival time.