Table 4. All-subset regression models evaluating combinations of multimodal, noninvasive inflammatory biomarkers for predicting serum C-reactive protein (CRP) levelsa.
| Regression model | Adjusted R2 | |
|---|---|---|
| Model 1 | Serum CRP~ sa-CRPb | 0.518 |
| Model 2 | Serum CRP~u-CRPc | 0.608 |
| Model 3 | Serum CRP~u-CRP/u-Crd | 0.513 |
| Model 4e | Serum CRP~saCRP+u-CRP | 0.761 |
| Model 5 | Serum CRP~saCRP+u-CRP/u-Cr | 0.747 |
| Model 6 | Serum CRP~uCRP+u-CRP/u-Cr | 0.691 |
| Model 7 | Serum CRP~saCRP+uCRP+u-CRP/u-Cr | 0.745 |
a
Each model includes noninvasive markers as predictors: sa-CRP, u-CRP, and u-CRP/u-Cr.
b
sa-CRP: saliva C-reactive protein.
c
u-CRP: urine C-reactive protein.
d
u-CRP/uCr: creatinine-normalized urine C-reactive protein.
e
Model 4, which includes both sa-CRP and u-CRP, demonstrated the highest explanatory power.