Logistic and linear regression models using uCrn and eGFR to predict %uAs metabolites.

aWe examined confounding or mediation of associations between uCrn and %As metabolites by using nested models, with and without control for eGFR; Model 1 parameters are log(age), sex, current smoking, log(total uAs), log(uCrn), and recruitment year (in total sample only); Model 2 parameters are log(age), sex, current smoking, log(total uAs), log(uCrn), eGFR, and recruitment year (in total sample only). bGeneralized R2. cProbability modeled is %uInAs >12.2 (total sample: %uInAs ≤12.2 N = 168, %uInAs >12.2 N = 310; 2001 sample: %uInAs ≤12.2 N = 123, %uInAs >12.2 N = 245; 2003 sample: %uInAs ≤12.2 N = 45, %uInAs >12.2 N = 65). Logistic and linear regression models using uCrn and eGFR to predict %uAs metabolites.