Researchers Say They've Created a 90% Accurate Blood Test for Autism

An experimental blood test could detect autism in children at younger ages than current diagnostic methods are capable of. Image: Getty
An experimental blood test could detect autism in children at younger ages than current diagnostic methods are capable of. Image: Getty

Autism spectrum disorder (ASD) is a complicated neurological condition caused by a variety of risk factors, including our genes and environment (but no, not vaccines), that interact in ways we still understand little about.


Its symptoms are varied, too—from problems with social interaction to being unable to speak or process sensations normally. This complexity extends all the way to how it’s diagnosed: Children can start to show visible signs of autism by the age of 18 months, but there is no single medical test that can diagnose it, and it often takes years to confirm a suspected case, potentially delaying treatment.

New research, published Sunday in the journal Molecular Autism, might provide one of the first steps needed toward developing an accurate blood test for the condition. And along the way, it might also help us better understand why autism occurs in the first place.

European researchers looked at the blood and urine of 38 Italian children diagnosed with ASD and compared them to a similarly matched control group of 31 children who did not have autism. The children with autism were 7 and a half years old, on average. In those children, they found signs of damage to certain proteins found in blood plasma that were caused by complex processes that involve either oxygen or glucose.

They then created four different predictive algorithms that tried to tell whether a child had ASD or not, based on the presence of these biomarkers. Among other things, the most successful algorithm looked for higher levels of a molecule called dityrosine, which appear when proteins are changed by oxygen-containing free radicals (a process known as oxidation), and advanced glycation end products (AGE), which are proteins or fats that have been altered by contact with glucose. This algorithm predicted if a child had autism with 90 percent accuracy, and predicted if a child didn’t with 87 percent accuracy.

“Our test is expected to improve the accuracy of ASD diagnosis from 60–70 percent currently achieved by experts in neurological disorders to approximately 90 percent accuracy and potentially offered at all well-equipped hospitals with or without high level expertise in neurological disorders,” lead author Naila Rabbani, a biologist at the University of Warwick in the UK, told Gizmodo via email.

In addition to forming the basis of a diagnostic test, the researchers believe their findings suggest that the build-up of AGEs and oxidation may actually help account for the symptoms associated with autism. Children with ASD in their study were also more likely to show signs of their neurons having less amino acids available. That seems to reaffirm a popular theory that some cases of autism can be sparked by having a rare genetic mutation that causes the proteins responsible for transporting amino acids to become dysfunctional.


“Our finding is consistent with low-level inflammation in ASD and also with disturbance in the supply of amino acids to neurons—the latter has also been suggested in previous gene association studies,” Rabbini said.

“We have not studied any relation to vaccines and cannot comment on this,” she added.


Rabbini’s team isn’t the first to try looking for biomarkers of autism in the blood—similar research is ongoing elsewhere, with so far similarly encouraging results. Aside from testing larger samples of people, Rabbini and her colleagues next plan to study the blood of even younger children to see whether their biomarkers can be also be used to predict how severely someone’s autism will progress.

“The test is patented and we are seeking commercial partners for further development towards regulatory approval and clinical availability,” Rabbini said.


The CDC estimates that one in 68 children in the U.S. has autism, though a 2015 survey put the rate at one in 45.

[Molecular Autism]


Born and raised in NYC, Ed covers public health, disease, and weird animal science for Gizmodo. He has previously reported for the Atlantic, Vice, Pacific Standard, and Undark Magazine.


It seems odd to me that it can more accurately predict if they do than if they do not. Is that odd? Any biologists out there who can offer some insight on if that sort of situation is common with other conditions/tests?