An Evaluation of Surrogate Chemical Exposure Measures and Autism Prevalence in Texas

Abstract

There is currently considerable discussion in the scientific community as well as within the general public concerning the role mercury (Hg) exposures may play in the apparent increased incidence of neurodevelopmental disorders (particularly autism) in children. Although the primary focus of this debate has focused on ethylmercury from vaccinations, linkage to other sources of Hg has been proposed. An ecologic association between 2001 Toxic Release Inventory (TRI; www.epa.gov/tri) data for Hg and 2000–2001 school district autism prevalence was previously reported in Texas. Evaluations using industrial release data as surrogate exposure measures may be problematic, particularly for chemicals like Hg that have complex environmental fates. To explore the robustness of TRI-based analyses of the Hg–autism hypothesis in Texas, a detailed analysis was undertaken examining the extent of the ecological relationship during multiple years and examining whether surrogate exposure measures would yield similar conclusions. Using multilevel Poisson regression analysis and data obtained from a number of publicly available databases, it was found that air Hg release data were significantly associated with autism prevalence in Texas school districts when considering data for 2001 and 2002 (2001: RR = 4.45, 95% CI = 1.60–12.36, 2002: RR = 2.70, 95% CI = 1.17–6.15). Significant associations were not found using data from 2003 to 2005. A significant association was not observed when considering air Hg data for 2000 or 2001 and school district autism prevalence data for 2005–2006 or 2006–2007, an analysis allowing for a 5-yr time period between presumed exposure and entry into the public school system (2000: RR = 1.03, 95% CI = 0.59–1.83, 2001: RR = 0.94, 95% CI = 0.59–1.47). Significant associations were not observed for any year nor for the time lagged analyses when censored autism counts were replaced by threes instead of zeros. An evaluation of TRI air emissions data for several other pollutants did not find significant associations except for nickel (RR = 1.71, 1.12–2.60), which has no history of being associated with neurodevelopmental disorders. An evaluation using downwind location from coal-fired power plants as the exposure surrogate variable also did not yield statistically significant results. The analysis suggests Hg emissions are not consistently associated with autism prevalence in Texas school districts. The lack of consistency across time may be the result of the influence of a more significant factor which remains unidentified. Alternatively, it may be that the significant association observed in 2001 and 2002 does not represent a true causal association.

Acknowledgements

This work was supported by a grant from the Electric Power Research Institute. The authors also acknowledge the technical assistance of Sunessa Schettler and Richard Monteiro.