Odd ratios were calculated per IQR increase in exposure.
Model 1, single-pollutant mode. No adjustment for other pollutants and covariates. The role of each pollutant in the mixing effect was obtained.
In the multi-pollutant models, the VIF value of PM10, PM2.5-10 and PM2.5 was over 5, which was multicollinearity and not included in the model at the same time. PM10 model (Model 2), PM2.5-10 model (Model 3), and PM2.5 model (Model 4) were adjusted for paternal atopy, maternal work during pregnancy, maternal primipara, monthly household income, open windows often, and use an air purifier.
In the multiple logistic regression model of the effect of pollutants during pregnancy on childhood eczema, results from the single-pollutant model showed that O3 exposure was significant for the risk of developing eczema at age 1 year only [1.11(1.04-1.19)] and age 2 years only [1.23(1.06-1.42)]. After adjusting for other pollutants and covariates in the multi-pollutant model, the results showed that PM10 exposure was significant for 1-year old eczema only [1.25(1.11-1.40)] and 2-years old eczema only [1.51(1.16-1.95)]. Dividing PM10 into PM2.5 and PM2.5-10, PM2.5-10 remained positively associated with the risk of 1-year old eczema only [1.32(1.20-1.45)] and 2-year-old eczema only [1.24(1.00-1.54)], while PM2.5 had no significant effect (Table 5).
In the multivariate logistic regression model in the first year of life(Table 5), the results of the single-pollutant model showed that PM10[1.08(1.02-1.15)], CO[1.07(1.00-1.14)], and SO2 [1.07(1.00-1.14)]exposure were all positively associated with only eczema at age 1, and after dividing PM10 into PM2.5 and PM2.5-10, both PM2.5[1.08(1.01-1.15)]and PM2.5-10[1.07(1.00-1.14)] were still positively associated with only eczema at age 1, and in the multi-pollutant model, CO [1.08(1.00-1.18)]was still positively associated with only eczema at age 1. O3 concentration was positively associated with only eczema at age 2, in both the single-pollutant model [1.26(1.09-1.46)] and the multi-pollutant model [1.24(1.03-1.50)] with statistical significance.
Table 5 OR and 95% CI for the effect of air pollutant exposure during pregnancy on childhood eczema in the multiple logistic regression model.