2.1. Study population

We randomly selected mothers of children born between 2005 and 2010 for a nested case-control study of autism using California birth records and matched controls in a 1:10 ratio by sex and birth year. All mothers eligible to contribute to the present study lived in the California Central Valley region according to their residential address recorded on the birth certificate (n = 1466). The offspring also had to be born at a gestational age between 21 and 46 weeks and have a birth weight between 500 g and 6800 g (n = 1433). California’s Central Valley is one of the regions with the greatest pesticide use. For this reason, we limited our analyses to mothers who according to our geographic information system (GIS) and state-mandated pesticide use report system had not been exposed to suspected neurotoxic pesticides (organophosphates, pyrethroids, glyphosates, fungicides, or neonicotinoids) (Rull and Ritz, 2003). Among the women randomly selected from birth certificates, we estimate percentiles of air pollution exposure and defined as highly exposed during early pregnancy those for whom three modeled traffic-related air pollutants (carbon monoxide (CO), nitrogen oxides (NO_x), and particulate matter with aerodynamic diameter ≤ 2.5 μm (PM_2.5)) fell at or above the 75th percentile of exposures in controls using a modified version of the California LINE Source Dispersion Model Version 4 (CALINE4) model (Yan et al., 2019).

Based on air pollution estimations, we included 116 children with autism and 98 autism-free control births with high levels of traffic- related air pollution in early and mid-pregnancy for our study (Fig. S1, Supplementary Material). We utilized maternal serum samples taken mid-pregnancy (around 16th week of gestation) as part of the California Prenatal Screening Program (CPSP) in which about 74% of all California women participate. We collected the mother-child pair’s demographic and other information such as maternal age, race/ethnicity, education, parity, child sex, and birth year from birth records. Other potentially confounding variables such as maternal smoking, body mass index (BMI), or diet were not available for the span of the study period.