Table 2

Multivariable linear mixed model for prediction of eGFR levels (with baseline eGFR as part of the dependent variable) obtained from AIC-based backward elimination on all candidate predictors (log2 transformed biomarker and clinical)

BaselineSlope
PredictorCoefficientP valueCoefficientP valueR2 decomposition
Constant407.630<0.001−2.1020.396
Cystatin C−11.661<0.001n.s.n.s.9
Endostatin−2.9570.133n.s.n.s.<1
UMOD2.977<0.001n.s.n.s.4.3
CHI3L11.1240.037n.s.n.s.<1
HGF0.0470.9490.4630.044<1
MMP10.7310.187−0.3070.082<1
MMP7−1.2330.064n.s.n.s.<1
TIE24.622<0.001n.s.n.s.3.3
TNFR1−10.888<0.001n.s.n.s.12.9
KIM1−0.0640.922−1.084<0.0013
FGF23−0.9830.2690.4560.086<1
NTproBNP0.0710.793−0.2530.006<1
Age (years)−0.684<0.0010.0480.03927
Current or former smoker2.4600.024n.s.n.s.<1
Mean arterial pressure0.0850.087n.s.n.s.<1
Total cholesterol−0.7620.096n.s.n.s.<1
• The model had an adjusted R2 of 62.5%. Biomarkers had a total contribution of 34.4%, and clinical risk factors had a total contribution of 28.1%. The decomposition of the model R2 combines the contributions of baseline and slope coefficients for each predictor. n.s., not selected.