Skip to main content


Archived Comments for: A comparison of temporal trends in United States autism prevalence to trends in suspected environmental factors

Back to article

  1. The air pollution anomaly

    John Cannell, Vitamin D Council

    21 October 2014

    The air pollution anomaly

    As the author points out, a number of studies have shown a positive association between autism and air pollution. If air pollution were the temporal driver of autism, then the autism epidemic should have occurred in the 1950's and 60's, before the Clean Air and Water Act was enforced.

    It is well known that air pollution lowers 25-hydroxy-vitamin D [25(OH)D] levels:

    Also, 25(OH)D levels are inversely associated with autism.

    The air pollution anomaly is explained by the fact that, in spite of cleaner air, 25(OH)D levels have been falling over time, due to sun avoidance:

    Also, surface UVB is inversely associated with autism prevalence:

    There are multiple reasons to think that low vitamin D is one of the important environmental agents in the gene/environment interaction that causes autism.




    John Cannell


    Competing interests

    I am employed by the Vitamin D Council, a non-profit, and recieve remuneration from Purity Products.
  2. Response to Dr. Cannell's comment on air pollution and vitamin D

    Cynthia Nevison, University of Colorado, Boulder

    21 October 2014

    I thank Dr. Cannell for his comment.  The relationship between air pollution, solar radiation and autism is a topic I have investigated in prior (unpublished) research on a specific air pollutant, atmospheric mercury.  A comparison of IDEA autism data and the climatological data described below suggests that autism prevalence tends to be lower in states receiving the most solar radiation (Figure 1a).  Meanwhile, a comparison of IDEA data and modeled total gaseous mercury (TGM) suggests that autism prevalence tends to be higher in states with higher atmospheric mercury (Figure 1b).  However, since atmospheric mercury, like many other air pollutants, is photoreactive, solar radiation and modeled TGM are themselves strongly anticorrelated (Figure 1c), suggesting that these two variables could be confused for one another in epidemiological studies.   A multivariate model of autism prevalence as a function of both TGM and DSWRF shows that the combination of the two variables yields little additional predictive power than either alone (Figure 1d).


    Figure 1:


    Figure 1.  Statewide correlations of a) autism prevalence vs. downward shortwave radiative forcing (DSWRF), a proxy for vitamin D-producing UVB radiation, b) autism prevalence vs. total gaseous mercury (TGM) obtained from a chemical tracer transport model, c) TGM vs. DSWRF, d) autism estimated from a multivariate linear regression model function of TGM and DSWRF vs. true autism prevalence.  In panels a-c, the solid black line represents a linear regression (with correlation coefficient R2 and p values reported), while in panel d, the solid black line is the 1:1 line for model vs. observed.

    While one cannot distinguish with confidence based on the ecological analysis in Figure 1 whether atmospheric mercury (TGM) or vitamin D deficiency or some other unknown factor that correlates to both is a more important contributor to autism, there are reasons for skepticism that TGM can be the main causal factor.  Among other things, there is a relatively narrow range of variability in TGM across the United States (Figure 1a) but a greater than factor of 4 range in the autism prevalence data (which themselves contain significant uncertainty among states).  This same narrow range of variability is evident in other air pollutants in many of the published air pollution/autism correlation papers in the literature.  For example, Volk et al. (2014) find a significant reduction in risk of autism in California among genetically susceptible children at ambient PM2.5 levels of 12.7 compared to 17 microgram/m3 and above.   Meanwhile, reported autism prevalence in China (Sun et al., 2013), where PM2.5 routinely reaches 200 microgram/m3 or more, is lower than in the United States.  A number of the air pollution/autism correlation studies have used modeled air pollution estimates, which likely come with supporting data on solar radiation.  In future studies of air pollution and autism, it may be useful to include estimates of solar radiation and associated risk of vitamin D deficiency to assess whether the latter may be a confounding influence.


    Cynthia Nevison

    Oct. 20, 2014



    Data and methodology used in Figure 1


    Autism Data

    State by state autism prevalence was taken from the Individuals with Disabilities Education Act (IDEA) database (  Mean prevalence for ages 3 to 21 reported in 2009 is shown. 


    Downward Shortwave Radiative Forcing (DWSRF)

    The NCEP North American Regional Reanalysis (NARR) dataset was used to quantify downward shortwave radiative forcing (DWSRF), which is a good proxy for vitamin D-producing UVB radiation.  NARR data are from the Research Data Archive (RDA), which is maintained by the Computational and Information Systems Laboratory (CISL) at the National Center for Atmospheric Research (NCAR).  NARR data are output on a 32 km x 32 km Lambert conformal grid, with latitude and longitude coordinates provided. The two-dimensional latitude/longitude coordinates of the Lambert conformal grid were matched to the closest grid cell of a 0.25 x 0.25 degree gridded U.S. state map and the statewide, population-weighted mean values of DSWRF were calculated.  The population weights, obtained from the GWPv3 world population density 2.5 arc minute grid (, were employed to best estimate the solar radiation received in the regions of each state where the largest number of people reside.  (In practice, the population weighting was important mainly for California and Oregon.)


    Total Gaseous Mercury (TGM)

    Output from the North American Mercury Model Intercomparison Study (NAMMIS) (Bullock et al., 2008) was used to estimate surface level concentrations (in ng/m3) of total gaseous mercury (TGM), which is the sum of gaseous elemental mercury Hg(0) and reactive gaseous mercury Hg(+2).  While NAMMIS compared 9 global and regional atmospheric chemistry and transport model permutations, only the GEOS-Chem (global) and CMAQ (regional) permutation is shown here.  TGM output was available on a 36 km x 36 km Lambert conformal grid and was convolved into statewide, population-weighted annual mean values, as described above for DSWRF.



    Bullock, O. R., Jr., et al. (2008), The North American Mercury Model Intercomparison Study (NAMMIS): Study description and model-to-model comparisons, J. Geophys. Res., 113, D17310, doi:10.1029/2008JD009803.

    Sun, X. et al.  (2013), Prevalence of autism in mainland China, Hong Kong and Taiwan: a systematic review and meta-analysis, Molecular Autism, 4:7.

    Volk, H. et al., (2014), Autism Spectrum Disorder Interaction of Air Pollution with the MET Receptor Tyrosine Kinase Gene, Epidemiology, 25(1).




    Competing interests