Table 3.

Reduced regression models showing coefficient estimates for CRP.

Coefficient	Estimate	Standard error	P-value	N*
Model 1				249
Age (years)	−0.015	0.005	0.002
BMI, kg/m2	0.038	0.014	0.007
Adjusted R2 = 0.085, AIC = 750.4, p = < 0.0001
Model 2				249
Age (years)	−0.016	0.005	0.001
Female x BMI	−0.059	0.03	0.049
Adjusted R2 = 0.092, AIC = 750.4, p = < 0.0001
Model 3				94
Age (years)	−0.019	0.007	0.011
Female	−0.327	0.205	0.115
BMI, kg/m2	0.049	0.022	0.026
Adjusted R2 = 0.157, AIC = 263.1, p = 0.001
Model 4				94
BMI, kg/m2	0.051	0.021	0.017
Age x HbA1c bi	−0.04	0.017	0.018
Female x Urine U bi	1.296	0.589	0.031
Adjusted R2 = 0.246, AIC = 258.7, p = 0.0003

P < 0.05 was considered statistically significant.

Reduced models are shown. All full models included demographic (age, gender) and physiologic variables (BMI, HbA1c), and estimated annual water intakes of U and As as predictors.

^*Models 1 and 2 included data from 249 participants. Models 3 and 4 include a subset of 94 participants with available data for urinary Cu. Models 3 and 4 also included continuous concentrations of urinary copper (Cu), vanadium (V), and nickel (Ni), As (log transformed), and U (binary). Models 2 and 4 included second order interactions and significant interactions are reported.