Table 3: Predictors of proteinuria in DN by multiple linear regression^{a,b}.

Unstandardized coefficients

Std error

Constant

−0.357

0.083

−4.304

<0.001

Age

−0.115

0.088

−1.306

0.199

BMI

−0.004

0.081

−0.055

0.956

Course

0.080

0.084

0.948

0.349

HbA1c

0.196

0.085

2.288

0.027

Principal component for SBP and DBP^{c}

0.112

0.087

1.298

0.202

Principal component 1 for TG, TC, HDL-C, and LDL-C^{d}

0.052

0.077

0.679

0.501

Principal component 2 for TG, TC, HDL-C, and LDL-C^{d}

−0.070

0.078

−0.894

0.376

Principal component for hs-CRP, TNF-α, uMCP-1, and SAA^{e}

1.184

0.009

13.103

<0.001

Dependent variable: ln (UAE/Ucr). ^{
b}Each variable was standardized by using scores before being entered into the regression model. ^{
c}Since the values of SBP and DBP were correlated, their unique principal component was substituted for them in the model and the principal component = 0.928 * SBP + 0.928 * DBP. In the formula, each variable was no longer the original variable, but standardized variable and the coefficients before the standardized variables represented the correlation coefficients of principal component and the corresponding original variables. So this formula showed that SBP and DBP were highly correlated and the extracted component could nearly represent the variables of SBP and DBP. ^{
d}Since the values of TC, TG, HDL-C, and LDL-C were correlated, their two principal components were substituted for them in the model and the principal component 1 = 0.289 * TG + 0.223 * HDL-C + 0.892 * LDL-C + 0.919 * TC, the principal component 2 = 0.770 * TG − 0.793 * HDL-C − 0.090 * LDL-C + 0.128 * TC. In the formulas, each variable was no longer the original variable but standardized variable and the coefficients before the standardized variables represented the correlation coefficients of principal component and the corresponding original variables. So formula 1 showed that LDL-C and TC were highly correlated and component 1 could represent the variables of LDL-C and TC, while formula 2 showed that TG and HDL-C were highly correlated and component 2 could represent the variables of TG and HDL-C. ^{
e}Since the values of hs-CRP, TNF-α, uMCP-1, and SAA were correlated, their unique principal component was substituted for them in the model and the principal component = 0.841 * hs-CRP + 0.928 * TNF-α + 0.883 * uMCP-1 + 0.944 * SAA. In the formula, each variable was no longer the original variable, but standardized variable and the coefficients before the standardized variables represented the correlation coefficients of principal component and the corresponding original variables. Since only one principal component was extracted among the four inflammatory factors and the correlation coefficients were all close to 1, it showed that the four inflammatory factors were highly correlated and the component could almost contain all the information of the four variables.