Results from multivariable regression models investigating the association between serum type I IFN activity and measures of subclinical CVD.

Principal components analysis was first utilized to combine data from five IFN-inducible genes (IFIGs) into three principal components (explaining 97.1% of the total IFIG variation) that could be simultaneously included as independent variables in the multivariable models. Data are presented for separate multivariable models (separate columns), according to outcome.aPrincipal component 1 (≈IFIT1+IFI44+MX-1+IFI44L);bPrincipal component 2 [≈(IFIT1+IFI44)−(MX-1+IFI44L)];cPrincipal component 3 (≈PRKR). Note: IFIGs scaled by factor of 100.*p**p***p†controlled for Framingham score, disease duration, and current medication use of prednisone, antimalarial, and statin; additional modeling controlling for hsCRP, PAI1/TPA, tissue factor, yielded similar results (data not shown).‡SLE only models adjusted for Framingham score, current medication use of prednisone, antimalarial, and statin, plus SLEDAI and PAI-1/TPA; additional modeling including immunosuppressive use (cyclophosphamide, azathioprine, mycophenolate mofetil) yielded similar results (data not shown).