Figure 2. Relations among end-points and nonparametric data in imatinib treated (n=58) and nilotinib treated (n=52) CML patients.

Cox proportional-hazards modeling and formal test for interaction: starting from logistic regression analysis in which Y was the analyzed variable; the variable X₁ and X₂ were the presence or not (1 or 0) of the LOX-1 polymorphism or other non-parametric variables (history of cardiovascular classical risk factors, here named TCRF, and X₃ the combination). The simplified formula for calculation was: Y = β₀+β₁X₁+β₂X₂+β₃X₃ and the null hypothesis was tested as H₀: β₃ = 0. Final validation of data was assessed by a resampling technique (exact tests) and discrimination analysis by the Hosmer–Lemeshow method [G²_HL= Σ¹⁰_J=1 (O_j – E_j)²/E_j (1-E_j/n_j) ∼X²⁸], where nj = number of observations in the j^th group, Oj = Σ1yij = observed number of positive cases in the j^th group, Ej = Σpij = Expected number of positive cases in the j^th group. (Each reported p is that obtained by this technique and significant if < 0.05)