The current literature provides, at best, inconsistent associations between exposures to conventional air pollutants and adverse birth outcomes. Additionally, very few studies have investigated linkages between toxic air pollutants and acute respiratory-related illness among children. While a variety of data imputation and cleaning techniques are available, the evaluation of such techniques for toxic air pollutants remains limited.
Criteria air pollutant (SO2, CO, NO2 and PM 10) exposures were used in logistic regression models to estimate adjusted odds ratios of low birth weight (LBW), small-forgestational-age (SGA) and preterm (PTB) births among women in Detroit, Michigan (1990-2001). Poisson regression models were used to estimate adjusted risk ratios of emergency department (ED) visits for respiratory illness among children making Medicaid claims in Dearborn, Michigan (2001-2002) and exposures to urban air toxics (UATs) distilled into source classes using receptor models. This study also included the analysis of intra- and inter-laboratory precisions for UATs using 122 pairs of replicate samples, and evaluated the performance of two imputation methods, multiple imputation and optimal linear estimation.
Among 155,000 singleton births examined, SGA was associated with exposure to NO2 (OR=1.10, 95% confidence interval=1.01-1.19), CO (1.14, 1.02-1.27) and PM10 (1.22, 1.04-1.44). SO2 exposure was associated with PTB (1.07, 1.01-1.14) and LBW (1.16, 1.04-1.30). Among the 7,863 children making Medicaid claims, ED visits for respiratory problems increased with exposures to UATs from fuel combustion sources (1.44, 1.03-2.01), photochemical pollutants (1.48, 1.15-1.90) and gasoline exhaust/evaporated gasoline (1.35, 1.05-1.74). For many compounds, the UAT measurements had very poor reproducibility. Data missing at random could be adequately imputed using either imputation method, but imputations for row-wise deletions, the most common type of missingness pattern, were uninformative.
This appears to be the first U.S. study to associate SGA with air pollutant exposures. It highlights the importance of accounting for long-term trends and maternal smoking status. This study also is one of the first to use source-apportioned exposure measures to associate source classes of UATs and respiratory problems in children. This approach is advantageous in that it captures exposures to mixtures, potentially enhances biological plausibility, and accounts for the correlation among pollutants.