Neurology Center

The media's all too often shabby treatment of neuroscience and psychology research doesn't just propagate bad science - it means that the really interesting and important bits go unreported. This is what's just happened with the controversy surrounding a paper from the Autism Research Center (ARC) at Cambridge University - Bonnie Aeyeung et. al.'s Fetal Testosterone and Autistic Traits. For research published in a journal with an impact factor of 1.538 (i.e. not good), it's certainly attracted plenty of attention - but for all the wrong reasons.


The Autism Research Center is headed by the dashing Simon Baron-Cohen, also one of the authors on the paper. He's probably the world's best-known autism researcher, and the author of some excellent books on the subject including the classic Mindblindness and The Essential Difference. Mindblindness, in particular, probably deserves a lot of the credit for interesting a generation of psychologists in autism. A big cheese, in other words. Surely his greatest achievement, however, is being Borat's cousin.

Baron-Cohen is famous for his theory that the characteristic features of autism are exaggerated versions of the allegedly characteristic features of male, as opposed to female, cognition. Namely, autistic people have difficulties understanding the emotions and behaviour of other people ("empathizing"), but may show excellent rote memory and understanding of abstract, mathematical or mechanical systems ("systematizing"). He and his colleagues have also hypothesised that an excess of the well-known masculinizing hormone testosterone, could be responsible for the hyper-male brains of autistics, just as testosterone is responsible for the development of masculine traits in boys. Amongst other things this would explain why rates of diagnosed autistic spectrum disorders are several times higher in boys than in girls.

Now, this is one of those wide-ranging theories which serves to drive research, rather than strictly following from the evidence. It's a bold idea, but there is, at the moment, not enough data to confirm or reject this idea. The simple view that testosterone = maleness = autism is almost certainly wrong, but it's a neat theory, there's clearly something to it, and, as one of the commentators on the paper puts it

Anyway, the paper reports on an association between testosterone levels in the womb and later "autistic traits" in childhood. 235 healthy children were studied; for all of these kids, the levels of testosterone in the womb during pregnancy were known, because their mothers had had amniocentesis, collecting a sample of fluid from the womb. Amniocentesis is not risk-free and it can't be done for research purposes, but the mothers here got amniocentesis for medical reasons and then agreed to take part in research as well. Testosterone levels in the amniotic fluid were measured; notably, this probably represents testosterone produced by the fetus itself, rather than the mother.

The headline finding was that fetal testosterone (fT) levels were correlated with later "autistic traits", as judged by the mothers, who filled out questionaires about their kid's behaviour at the age of about 8. Here's a nice plot showing the correlation. The vertical axis, "AQ-child total", is the parent's total reported score on the "Autism Quotient" questionaire. Higher scores are meant to indicate autism-like traits (although see below). You'll also notice that fT levels are much higher in the boy fetuses than in the girl fetuses - not surprisingly. That's it - a statistically significant association, but there is still a lot of scatter on the plot. The correlation was still significant if the very high-scoring children were ignored. A similar pattern emerged using a different autism rating scale, but was less significant - probably because many scores were very low.
So, this was a perfectly decent study with an interesting result, but it's only a correlation, and not an especially strong one. How did this get written up? New research brings autism screening closer to reality puffed the Guardian's front page! They suggested that measuring fetal testosterone levels might be a way of testing for autism pre-natally, thus sparking off an entirely formulaic debate about the ethics of selective abortion, the usual denunciations of "eugenics", etc. Long story short - Catholics are against it, the National Autistic Society say it's a dilemma, while a family doctor on Comment is Free is unsure about the "test" because she can't read the article: she doesn't have access to the journal.

Lest it be said that the ethical debate is important in itself, even if the details of the testosterone-based screening test might be inaccurate, bear in mind that "testing for autism" is likely to raise unique issues. Are we talking about a test which could distinguish "low-functioning autism" - which can leave children unable to lead anything like a normal life - from "high-functioning autism", sometimes associated with incredible intellectual achievement? Would the test distinguish classical high-functioning autism from Asperger's? When and if a test is developed, these will be crucial questions. You cannot simply speculate about "a test for autism" in the abstract.

Anyway, after a few days of this nonsense Baron-Cohen rightly protested that the paper had nothing to do with prenatal testing, and that such testing isn't on the horizon yet.

The new research was not about autism screening; the new research has not discovered that a high level of testosterone in prenatal tests is an indicator of autism; autism spectrum disorder has not been linked to high levels of testosterone in the womb; and tests (of autism) in the womb do not allow termination of pregnancies.

Most importantly, there were no autistic kids in the study - all of the children were "normal", although some were rated highly on the autism measures. Moreover, as the plot above shows, any testosterone-based screening test would be very inaccurate. Which is why no experts proposed one.

Just like last time. Back in 2007 the Observer (the Sunday version of the Guardian) ran a front-page article about Simon Baron-Cohen's work on the epidemiology of autism. They said that he'd found that autism rates in Britain were "surging"; they probably aren't, and Baron-Cohen's data didn't show that they were, but despite this the Observer took weeks to clarify the issue (for details of the saga, see Bad Science.) In both cases, some important research about autism from Cambridge ended up on the front page of the newspaper, but the debate which followed completely missed the real point. It would have been better for all concerned if the research had never caught the attention of journalists at all.

The actual study in this case is very interesting, as are the three academic commentaries and a response from the authors published alongside it. I can't cover all of the nuances of the debate, but some of the points of interest include: the question of whether the Autism Quotient (AQ) questionaire actually measures autistic behaviours, or just male behaviours; the point that it may be testosterone present in baby boys shortly after birth, not in the womb, which is most important; and the interesting case of children suffering from Congenital Adrenal Hyperplasia, a genetic disorder leading to excessive testosterone levels; Baron-Cohen et. al. suggest that girls with this disorder show some autism-like traits, but this is controversial. Clearly, this is a crucial point.

Overall, while it's too soon to pass judgement on the extreme male brain theory or the testosterone hypothesis, both must be taken seriously. As for autism prenatal testing, I suspect that this will only come when more of the genetic causes of autism are identified. There is no single "gene for autism"; currently a couple of genes responsible for a small % of autism cases are known: CNTNAP2, for example.

Once we have a good understanding of the many genes which can lead to the many different forms of autistic-spectrum disorders, genetic testing for autism will be possible; I doubt that testosterone levels or anything else will serve as a non-genetic marker, because autism almost certainly has many different causes, and many different associated biochemical abnormalities. Maybe I'm wrong, but even so, if you're worried about hypothetical people aborting hypothetical autistic fetuses, you don't have to worry quite yet. Actual children are dying in Zimbabwe - worry about them.

[BPSDB]

Bonnie Auyeung, Simon Baron-Cohen, Emma Ashwin, Rebecca Knickmeyer, Kevin Taylor, Gerald Hackett (2009). Fetal testosterone and autistic traits British Journal of Psychology, 100 (1), 1-22 DOI: 10.1348/000712608X311731