A Breath of Not-So-Fresh Air

A paper published last week in Environmental Health Perspectives had the autism community all atwitter last week, quite literally. It was clear something was up when my Twitter feed suddenly filled up with tweets about a national study, led by Andrea Roberts and colleagues at Harvard University, “confirming the link between autism and air pollution.”

Utilizing Environmental Protection Agency records of hazardous air pollutant concentrations, the researchers examined levels of air pollution in the year and location in which women lived when they gave birth to their children. These mothers were enrolled in the Nurses Health Study II, a large sample of female nurses that has been followed since 1989. To focus their investigation, the researchers studied air pollutants previously found to be associated with autism, such as metals and diesel particulate matter. Overall, they found positive associations between perinatal exposure to several air pollutants and autism, meaning the greater the concentration of these pollutants, the greater chance of a child having autism. They also showed that some of these associations were stronger for boys than for girls.

However, this highly publicized study on air pollution is not exactly a breath of fresh air. Several things reveal weaknesses in the design and interpretation of this study:

1. On Twitter, Forbes writer Emily Willingham, made an interesting point:

Screen Shot 2013-06-26 at 8.20.42 PMAir pollution may be related to autism, just as countless environmental factors are, but it clearly isn’t a major player in its etiology, or causation (as much as I hate to use that word). For more of Emily’s thoughts on the autism/air pollution relationship, see this insightful post about a previous study.

2. While the Nurses Health Study provided a large sample, there is a significant confound with only studying children born to nurses. In an Autism Research paper, Gayle Windham and colleagues demonstrate a relationship between maternal occupational exposure to chemicals, as well as other potential toxins, and autism. Particularly, maternal exposure to disinfectants, as can be seen in nurses and other medical professionals, seemed to be related to autism.  While this finding doesn’t negate the air pollution finding, it certainly complicates the picture.

3. The children in the autism group in this study were noted as having autism, Asperger’s and PDD-NOS, or, as was explained more vaguely, “may have been on the autism spectrum.” Including multiple subtypes of autism in a study sample is common, but, again, it provides another confound, especially considering the subtype-specific findings related to smoking during pregnancy, which I discuss in this post.

4. Regarding the researchers’ interpretation of these findings, there was much to be desired. Yes, an association between air pollutants and autism was demonstrated. But why or how? In other words, what potential genetic or developmental underlying mechanisms are set into motion by air pollution exposure? For now this is just a correlational, not causational, finding. Simply stating the supposed existence of an association doesn’t provide many answers.

While an interesting concept, the autism and air pollution “link” needs further exploration, and, hopefully, a more reserved reception as future studies are published.


Risk factors in autism: A deeper look at the subtype argument

It has been well demonstrated that prenatal (during gestation) and perinatal (after birth) risk factors are associated with autism spectrum disorders. But which risk factors have a significant relationship with the prevalence of autism remains up for debate. Several papers argue that maternal stress during pregnancy is an important risk factor for autism, but even this proposition is muddled, as several prenatal stressors do seem to be related to autism (e.g., the experience of a natural disaster, loss of a loved one), whereas others, albeit surprisingly, do not (e.g., physical abuse).

An answer to these discrepant findings may lie an argument supported by many autism researchers: that autism is a conglomeration of several distinct subtypes, each with their own genetic and behavioral profile. If there exist specific subtypes of autism, is it possible that one risk factor may contribute to one subtype while not influencing the occurrence of another?

A recent paper published in the Journal of Autism and Developmental Disorders asks this very question. Janne Visser and colleagues from Radboud Unversity in The Netherlands argue that defining autism broadly versus narrowly affects which risk factors seem to have a significant impact. In their study, research participants underwent an extensive diagnostic protocol, including the standard diagnostic measures for autism (ADI-R and ADOS) as well as several other diagnostic tools. Following the diagnostic protocol, participants were classified as either having autistic disorder (AD), a narrowly defined diagnosis, or pervasive developmental disorder-not otherwise specified (PDD-NOS), a more broad diagnosis. Parents of these children and those of matched controls completed a survey with questions pertaining to many pre- and perinatal risk factors.

The most robust finding in this study was an expected one. When the AD and PDD-NOS groups were combined into one autism spectrum disorder group, this group demonstrated an increased occurrence of several risk factors, including maternal infections, prenatal stress, and low birth weight, as compared to the control group. The occurrence of these risk factors was also related to the severity of core autism symptoms, such as repetitive behaviors and impairments in communication.

Interestingly, however, when the AD and PDD-NOS groups were compared to each other, one risk factor, smoking during pregnancy, stood out as being significantly related to PDD-NOS, but not AD. Smoking during pregnancy was also related to lower IQ, but only in the PDD-NOS group, a finding that remained even after controlling for birth weight, which is commonly affected by smoking. This distinction between two autism subtypes provides support for the argument that subtypes of autism (narrowly versus broadly defined) may have different environmental and genetic influences.

Smoking during pregnancy is also a risk factor associated with ADHD, suggesting that this disorder may have underlying mechanisms similar to those of PDD-NOS, but not of AD. These mechanisms may be more than environmental, as Visser and colleagues propose. Smoking during pregnancy may be associated with specific genetic traits that can be passed from mother to child. These traits may contribute to the development of disorders, such as PDD-NOS and ADHD.

Why smoking during pregnancy wasn’t related to AD remains to be seen. It’s clear, however, that future studies should examine risk factors as they relate to not just autism in general but to specific subtypes of this disorder. Fortunately, the subtypes argument is being supported increasingly in the field of autism research. Although it begs more questions about each subtype and its associated characteristics, the subtypes argument provides a more focused and realistic approach to studying the true spectrum that is autism.