Why The 'Autism Test' Might Not Be What Everyone Is Hoping For

There are some big problems with extrapolating this research to the autistic population as a whole

19/02/2018 10:08 GMT | Updated 19/02/2018 10:08 GMT
FatCamera via Getty Images

A scientifically fascinating paper by Anwar et al published today will claim to have found a biochemical ‘signature’ that distinguishes children with and without autism (ASD), which the team say may act as an ‘indicator’ for the condition. But headlines about a ‘test for autism’, though inevitable, are hugely premature.

Dietary proteins are broken down by the body in a bewilderingly complex process, and a legion of different molecules are produced as a result. Essentially, the research team looked at levels of these intermediate molecules in multiple combinations, until they found one that differed reliably between a population of about 30 autistic children and a similar number of typically developing children. However, there are some big problems with extrapolating this to the autistic population as a whole:

  • It’s a very small population, making a chance finding more likely
  • The fact that multiple approaches were tried, and had highly variable performance, increases the chance that those results presented as significant are instead statistical ‘noise’, albeit that the researchers tried to reduce this problem.
  • The groups did not only differ as regards their autism: the autistic group also ate fewer vegetables. This may seem trivial, but given that the whole analysis is about protein break-down, it is plausible that the difference in diet contributed to the subtle biochemical differences, and the paper certainly does not exclude this.
  • The autistic group of children were skewed towards the severe end of the spectrum, and had an average age of 7. It’s not clear that this analysis would hold in younger children or less affected populations.

But let’s accept that these difficulties can be overcome and a fairly reliable biochemical indicator can be developed in the future- would it be useful in our clinic, where we diagnose and support children with a variety of developmental problems, including autism?

Probably not. Because in order to add to existing diagnostic practice, a test would have to either predict future autism at an early age, or improve upon existing clinical methods of assessment. 

In terms of early, pre-symptomatic, identification, even if these findings did apply to younger children, we would have to test a large population of at-risk infants to make this a useful ‘screening’. Given that the test is only about 84% specific, this would lead to a huge number of false positive identifications, huge worry for parents and, potentially, harmful intervention for children. And we don’t know of any way of reliably preventing autistic difficulties, at this point, so it’s not clear what difference it would make.  

On the other hand, once the child is showing problems with their development, it is not clear what this kind of test will add. Firstly, if a child does not look or act autistic, then no test will persuade any sensible clinician that they are (although it may persuade them to keep an eye on things). Secondly, the question we are asking in clinic is not, ‘is this child autistic, or is she typically developing?’ but, ‘is this child’s atypical developmental profile one that looks like autism?’. Help with that often very tricky question would be welcome, but would require an entirely different research strategy.

Parents of young children are often terrified by the prospect of an autistic child. Those for whom the possibility has been raised need prompt, comprehensive and accurate clinical assessment, and those with the diagnosis need consistent, high-quality care and appropriate intervention. No blood test will help any of these groups, but adequate early intervention, diagnostic and support services certainly would- if we had those available, that truly would be a breakthrough!