increase in blood lead was associated with a 19% higher levels of total cholesterol ([exp(1.71×log(2)) –1]×100% = 19%). Since we used these weights to construct ERS rather than interpret the associations of individual pollutants, we presented the direct weights rather than more interpretable estimates (percent changes). Also note that less significant associations in ERS2 compared with ERS1 are mainly due to lower power due to fitting of a larger model with larger number of parameters and with multiple pollutants that are potentially correlated. Two pollutants (1,2,3,4,6,7,8-HpCDD and PCB 177) for total cholesterol and 4 pollutants (PCB 118, PCB 138, PCB 153 and 3,3,4,4,5,5-PnCB) for triglycerides were additionally identified in conditional analyses in which the pollutants selected in the previous two-stage analyses were included as covariates. No further pollutants were identified in relation to HDL and LDL in the conditional analyses. Therefore, a total of 13 pollutants for total cholesterol, 9 for HDL, 5 for LDL and 27 for triglycerides were identified and used to construct ERS for each outcome. Various persistent organic pollutants (POPs) were positively associated with total cholesterol and triglycerides and inversely associated with HDL in single-pollutant models but the association directions for some POPs (2,3,4,7,8-PnCDF, 3,3,4,4,5-HxCB, PCB
