Is road traffic pollution really a cause of autism?

Wednesday June 19 2013

“Severe air pollution ‘can double risk of having autistic child’,” reports The Times.

If you’re a parent-to-be you shouldn’t be unduly worried by this story, as the science it’s based on cannot and does not show a definite link. And other than moving house or changing jobs, avoiding environmental pollutants is likely to prove difficult.

This frightening headline was based on research into the exposure to environmental pollutants of mothers with children diagnosed with autism spectrum disorder (ASD).

The study compared women in the highest 20% pollution exposure category with those in the lowest 20% of pollution levels. It found exposure to diesel, lead, manganese, mercury, methylene chloride (an industrial solvent), and an overall measure of metals were all significantly associated with a higher risk of ASD. The risks ranged from 50% higher (for overall metals) to 100% higher (for diesel and mercury). For example, children with ASD were more than twice as likely to be born to mothers with the highest 20% of diesel and mercury exposure, than the lowest 20%.

However, due to its study design this research cannot, and does not, prove that higher air pollution around the time of childbirth causes or increases the risk of a child developing ASD. The causes of ASD are not firmly established and it is likely this research has not accounted for other factors that could be influencing risk. However, it does suggest a potential link that warrants further investigation.

Where did the story come from?

The study was led by researchers from the Harvard School of Public Health in the US and was funded by the US Department of Defence, Army Medical Research and Materiel Command, and National Institutes for Health.

The study was published in the peer-reviewed medical journal, Environmental Health Perspectives.

Neither the Daily Mail nor the Daily Mirror’s coverage discussed the limitations of the research and generally took the findings at face value. However, The Times’ coverage included points from ‘other scientists’ outlining some of the study’s limitations.

What kind of research was this?

This was a case-control study exploring whether exposure to pollution around the time of a child’s birth was linked to the risk of the child developing autism spectrum disorder.

Autistic spectrum disorders (ASDs) cover a range of related developmental disorders, including autism and Asperger’s syndrome. They have key characteristics, including problems with:

  • social interaction with others (such as not being able to respond to others’ emotions)
  • communication (such as difficulties in having a conversation)
  • having a restricted, repetitive collection of interests and activities, rigid routine or rituals

Children described with autism usually have some degree of intellectual impairment and learning difficulties, while children with Asperger’s usually have normal intelligence.

The researchers describe how air pollution contains many toxic chemicals known to affect neurological function and foetal development. Recent studies have reported associations between exposure to air pollutants around the time of childbirth and ASD in children. This study sought to explore this link further.

What did the research involve?

This study involved asking a group of mothers if their children had ASD and then assigning historical pollution data to their address around the time the child was born.

This study used data from the Nurses’ Health Study II, a cohort of 116,430 female nurses from 14 US states. The Nurses’ Health Study II cohort was established in 1989 and has been followed over time with biennial questionnaires.

In 2007-08, researchers sent a questionnaire to the 756 women who had previously reported having a child with ASD, asking about the affected child’s sex, birth date, and whether they were adopted. These represented the “cases” in this study. They were also asked what specific diagnosis the child had been given with autism, Asperger syndrome, and ‘pervasive developmental disorder not otherwise specified’ (PDD-NOS) as possible answers. Cases were excluded if:

  • they had missing data for ASD diagnosis
  • they were adopted
  • the mother did not want to participate
  • the child’s year of birth was missing

This yielded 325 cases that were included in the final analysis.

The ASD diagnoses were validated by telephone using a questionnaire called the Autism Diagnostic Interview Revised. The researchers used 50 randomly selected “case” mothers who indicated willingness to complete the interview.

A group of 22,098 “controls” were used as a comparison group. These were children born from 1987 to 2002 (the years when air pollution data were available) to mothers who indicated they had never had a child with ASD.

Hazardous air pollutant concentrations were assessed by the US Environmental Protection Agency (EPA) National Air Toxics Assessments in 1990, 1996, 1999, and 2002. These used an inventory of outdoor sources of air pollution, including both stationary sources (such as waste incinerators and small businesses) and mobile sources (such as traffic) to estimate average concentrations of pollutants for different communities based on pollution dispersion models.

As pollution levels were not measured every year, the children were assigned pollution concentrations from EPA assessments closest to their year of birth (births 1987 to 1993 used 1990 concentrations; births 1994 to 1997 used 1996 concentrations; births 1998 to 2000 used 1999 concentrations and births 2001 to 2002 used 2002 concentrations).

The researchers recorded family (including grandparents’ education level) and community socioeconomic factors (average community income and education level) that could potentially influence the risk of ASD. They attempted to compensate for these influences in the statistical analysis.

What were the basic results?

The final analysis compared information from 325 cases, with 22,101 controls.

The researchers categorised the children’s level of pollution exposure into fifths (20% of the study group in each pollution level category). They found that those children exposed to the highest versus lowest fifth of diesel, lead, manganese, mercury, methylene chloride, and an overall measure of metals were significantly more likely to have an ASD. The odds ratios for these exposures ranged from 1.5 (for overall metals measure) to 2.0 (for diesel and mercury). This means those in the highest fifth (the top 20%) of diesel and mercury exposure were found to be twice as likely to develop ASD compared with those in the lowest fifth (the bottom 20%).

When the researchers looked at the linear trends, these were also positive and statistically significant for these exposures. This means the risk went up directly as pollution levels went up.

For most pollutants, associations were stronger for boys (279 cases) than girls (46 cases) and were significantly different according to gender.

How did the researchers interpret the results?

The authors concluded that exposure to air pollutants around the time of child birth, “may increase risk of ASD”, and that future studies should investigate the gender differences observed.


The limitations in this study’s design mean it cannot prove that air pollution causes or increases the risk of developing autism spectrum disorder (ASD). However, it does tentatively suggest higher pollution levels may increase the risk, which may prompt further, more reliable investigation.

It is important to consider the limitations of this research before concluding a direct causal link between environmental pollutants and autistic spectrum disorders.

Problems with assessment of pollution levels

The air pollution data was not accurately assigned to the child’s date of birth. Some children were assigned pollution levels three years before they were born, and others three years after. This means we cannot be sure at which point (before or after child birth) pollution may be influencing the risk of ASD, or whether the timing of pollution exposure was important in any way.

This occurred because the researchers used existing pollution data and fitted this as best they could to the dates the children were born. While this was clearly a practical approach, as the dates didn’t match exactly, it will have introduced some inaccuracy. The researchers implied that exposure may be more important before birth as some pollutants can affect the developing baby.

Difficulty specifying causes of autistic spectrum disorders

The possible causes of ASD are not firmly established. While some effort was made to adjust for additional factors that could affect ASD risk outside of pollution, this may not have been complete. Hence, differences in factors such as socioeconomic circumstances (as well as others) may account for some or all of the differences in ASD risk observed.

Problems with comparing risks for boys versus girls

There were very few girls in the study, most likely because ASD is more common in boys than girls. The small number of girls make reliable comparisons between boys and girls difficult. As the researchers point out, for this reason, the conclusions around differences in risk versus exposure profiles between boys and girls are not reliable.

Small number of cases included

The sample size of children with ASD was quite small (325) in this study and represented less than half of the original 756 eligible for the study. Many participants were excluded because they had important information missing such as the year of birth. This small sample may not be representative of the wider group of children with ASD.

The bottom line is that this study does not prove that air pollution increases the risk of a child developing ASD. However, it does highlight a potential link that warrants further investigation.

Analysis by Bazian
Edited by NHS Choices