Regression models to estimate total polychlorinated biphenyls in the general US population: 2001-2002 and 2003-2004.

Certain polychlorinated biphenyls (PCB) have long half-lives and, despite the regulatory bans on the industrial pollutants that expose humans to PCB, are detectable in human serum. However, many of them are not detectable because of the small quantities that may be present in body fluids. For this reason, attempts have been made to estimate the total concentration of PCB (SigmaPCB) using the relationship between SigmaPCB and the concentrations of a few of the PCB congeners which can be reliably measured at detectable levels. PCB 153 or a combination of PCB 153,138, and 180 have previously been used for this purpose. However, because of the unique populations investigated in these studies, the results are not necessarily applicable to the racially/ethnically heterogeneous US population. We defined SigmaPCB as the sum of the concentrations of 12 PCB congeners, and sum of 33 PCB congeners for NHANES 2001-2002 and 2003-2004 respectively. We built regression models in a step-wise fashion using SigmaPCB as the dependent variable and age, race/ethnicity, and gender as the covariates for both whole-weight and lipid-adjusted data. In addition, concentration of PCB 153 was used as the continuous independent variable for 2001-2002 models, and PCB 153 and PCB 180 for 2003-2004 models respectively. R(2) for both models for NHANES 2001-2002 was >86%. The R(2) for both NHANES 2003-2004 models was >81%. Thus, the estimate of SigmaPCB for the general US population can be improved by considering common demographic variables, such as race/ethnicity, and selected congeners.