A large cohort study to evaluate associations between PM and PM and ASD
2.5
0.1
Decompose highly correlated PM and PM into shared and unique variance
2.5
0.1
An application of Deming regression assuming measurement errors in PM and PM
2.5
0.1
Shared variance between prenatal PM and PM was associated with ASD
2.5
0.1
While fine particulate matter (PM2.5) has been associated with autism spectrum disorder (ASD), few studies focused on ultrafine particles (PM0.1). Given that fine and ultrafine particles can be highly correlated due to shared emission sources, challenges remain to distinguish their health effects. In a retrospective cohort of 318,371 mother-child pairs (4549 ASD cases before age 5) in Southern California, pregnancy average PM2.5 and PM0.1 were estimated using a California-based chemical transport model and assigned to residential addresses. The correlation between PM2.5 and PM0.1 was 0.87. We applied a two-step variance decomposition approach: first, decomposing PM2.5 and PM0.1 into the shared and unique variances using ordinary least squares linear regression (OLS) and Deming regression considering errors in both exposures; then assessing associations between decomposed PM2.5 and PM0.1 and ASD using Cox proportional hazard models adjusted for covariates. Prenatal PM2.5 and PM0.1 each was associated with increased ASD risk. OLS decomposition showed that associations were driven mainly by their shared variance, not by their unique variance. Results from Deming regression considering assumptions of measurement errors were consistent with those from OLS. This decomposition approach has potential to disentangle health effects of correlated exposures, such as PM2.5 and PM0.1 from common emissions sources.
Unlabelled Image
Autism spectrum disorder (ASD) is a complex set of developmental disorders characterized by deficits in social interactions and communication, as well as the presence of restricted, repetitive, and stereotyped patterns of behaviors (American Psychiatric Association, 2000). A recent review reported a global autism prevalence of approximately 1 in 100 children (Zeidan et al., 2022). The estimated prevalence of ASD in the United States has shown a significant increase from 0.66 % in 2002 to 2.78 % in 2020 (Centers for Disease Control and Prevention, 2007; Maenner et al., 2023). ASD imposes considerable lifelong emotional and social burden on affected children (Kuhlthau et al., 2010), their families (Rao and Beidel, 2009), and economic burdens on society (Buescher et al., 2014).
A growing number of studies have shown associations between prenatal air pollution exposures and ASD in offspring, with the most consistent findings for particulate matter (PM) (Jo et al., 2019; Chun et al., 2020; Lam et al., 2016; Rahman et al., 2022). PM is emitted directly to the atmosphere from a variety of sources, including vehicles, industrial sources, and biomass combustion. PM can also form in the atmosphere as a product of chemical reactions (Moreno-Ríos et al., 2022). Due to the dynamic formation processes which involve nucleation, condensation, and coagulation, PM exhibits varying aerodynamic diameters broadly defined as coarse (with aerodynamic diameter < 10 μm; PM10), fine (with aerodynamic diameter < 2.5 μm; PM2.5) and ultrafine particles (with aerodynamic diameter < 0.1 μm; PM0.1) (Moreno-Ríos et al., 2022). These differences in size give rise to distinct aerosol behaviors and toxicity profiles (Kelly and Fussell, 2012). The shared variance between PM of different aerodynamic sizes may indicate common emission sources such as traffic exhaust and biomass burning (Hu et al., 2014a). The unique variance may indicate distinct atmospheric processes. PM0.1 can form in ambient air through nucleation, followed by coagulation and condensation into larger particles (Seigneur, 2009). Secondary aerosols like sulfate and nitrate, which are typically larger than 0.1 μm (Hazi et al., 2003), may contribute to the unique variance of PM2.5. Therefore, distinguishing the shared and unique variance of PM of different sizes can provide insights into the different mechanisms or components that may impact human health.
Recent toxicological and epidemiological studies reported that ambient PM0.1 may have exerted greater toxic effects on health than coarse particles (Terzano et al., 2010; Oberdörster et al., 2005; Oberdürster, 2000). The small size of PM0.1 enables it to penetrate deep into the lungs and enter the systemic circulation, which may induce inflammation and oxidative stress (Terzano et al., 2010). Moreover, PM0.1 may cross biological barriers (Cory-Slechta et al., 2023), such as placenta (Bongaerts et al., 2020) and blood-brain barrier (Oberdorster et al., 2004), and enter the fetal bloodstream and developing fetal brain (Bongaerts et al., 2020). Animal studies have also provided evidence that PM0.1 can be a source of neurotoxicity that can alter microglia stimulation and impact neuronal activity and synaptic plasticity of developing brain (Klocke et al., 2017; Rodulfo-Cardenas et al., 2023; Morris-Schaffer et al., 2019). Therefore, prenatal PM0.1 exposure may have direct adverse health effects on fetal neurodevelopment. Since PM0.1 is not regulated as a criteria pollutant and the modeling techniques for PM0.1 exposures have only recently been developed, few epidemiological studies have examined the health effects of PM0.1 during pregnancy. A recent study reported that prenatal PM0.1 exposure was associated with asthma development in children (Wright et al., 2021). Specific to neurodevelopmental disorders such as ASD, a study from our group showed that aircraft emitted PM, primarily composed of PM0.1, was associated with ASD in offspring (Carter et al., 2023). A recent case-control study has reported the association between increased PM0.1 exposure in the second year of life and ASD development among children (Goodrich et al., 2023).
However, previous studies often treated PM0.1 or PM2.5 as single exposures and examined their isolated health associations, rather than considering the real world scenario where children are exposed to both PM2.5 and PM0.1, which can be highly correlated because of shared sources such as traffic (Yu et al., 2019). Therefore, it remains a challenge to distinguish their associations with health outcomes. The collinearity issue and inflated standard errors may present in the co-adjusted model with two highly correlated exposures (Mason and Perreault Jr, 1991). Therefore, we proposed a two-step variance decomposition approach using ordinary least squares (OLS) linear regression and Deming regression considering measurement errors in both exposures (Linnet, 1993; Deming, 1943) as the first step to decompose two correlated PM2.5 and PM0.1 exposures into unique and shared variance. In the second step, we differentiated the associations of the health outcome with the shared variance and the unique variance that remained in their residuals. We leveraged a validated chemical transport model (CTM) (Hu et al., 2014a; Hu et al., 2014b; Hu et al., 2015) for PM2.5 and PM0.1 exposures and a large retrospective birth cohort in Southern California to examine the associations between ASD in children with the shared and unique variance average PM2.5 and PM0.1 exposures during pregnancy.
This study utilized a population-based retrospective birth cohort that included mothers with singleton deliveries (n = 370,723) at Kaiser Permanente Southern California (KPSC) hospitals between January 1, 2001 and December 31, 2014. Information related to the mothers and children, including maternal residential addresses history and medical conditions, were extracted from high-quality integrated electronic medical records (EMR) maintained by KPSC. A total of 52,352 births were excluded due to
