Home > News & Opinion > News > Genetic tests for autism debut amid concerns about validity

Genetic tests for autism debut amid concerns about validity

Popularity tracker
Virginia Hughes
1 November 2012


Wide array: Unlike most genetic tests for autism, which detect rare variants, the new tests analyze combinations of common markers.

A genetic panel intended to predict the risk of developing autism debuted for clinical use in April, while another is in commercial development and a third was published in Molecular Psychiatry in September1. But some experts are concerned, saying the tests are based on preliminary scientific evidence.

Several tests for autism are already in widespread clinical use, but those detect rare genetic variants that account for only a small percentage of cases.

The new tests instead analyze combinations of common markers. Two of the new tests detect single-gene variants, and the third looks at expression patterns of several hundred genes in blood.

As the cost of sequencing continues to plummet, experts say families are likely to be offered a dizzying array of similar tests that claim to screen for autism.

The average child with autism is diagnosed at age 4 — several years after symptoms emerge and after behavioral interventions are thought to be most effective.

For this reason, scientists have been searching for an early and reliable biomarker of the disorder. So far, however, studies of several thousand people with autism or other psychiatric disorders have failed to find any predictive genetic combinations2.

"We need to be more solid on this data before you’re giving a flyer to a family and saying, ‘Spend some money and this might give you some information,’” says Richard Anney, assistant professor of neurodevelopmental molecular genetics at Trinity College Dublin, who was not involved in the new tests.

Common risk:

Many specialized clinics already offer genetic tests for rare forms of autism. The tests typically rely on chromosomal microarray analysis, which identifies large DNA deletions or duplications. This type of genetic glitch, however, appears in only about ten percent of people with the disorder.

In April, the French biotech company IntegraGen debuted a genetic test that looks instead at common variants — called single-nucleotide polymorphisms, or SNPs — that are more likely to crop up in individuals with autism than in controls. Each SNP confers only a slightly increased risk of autism, and many people who carry autism-linked SNPs do not have the disorder.

IntegraGen’s test is tailored to families who already have one child with autism and want to gauge the risk for a second child. Among younger siblings of children with autism, 1 out of 5 develop the disorder, compared with 1 in 88 in the general population. The $2,895 test purports to pin down the risk more precisely by screening for combinations — one for boys, another for girls — of 65 SNPs.

The test places younger siblings into one of several risk categories, based on the SNPs they carry: For example, boys in the highest category of risk have a 59 percent chance of developing autism.

The test has a different level of accuracy for each risk category. For the highest risk group, for example, the test has about 95 percent ‘specificity’ for correctly identifying children who will not develop autism. This category has 15 percent ‘sensitivity’ for correctly identifying the children who do develop autism, meaning that it misses 85 percent of children who will develop autism. 

However, the relatively high specificity means that few children are falsely identified as being at risk for the disorder. As a result, the test brings attention to children who are most in need of early intervention, says Bernard Courtieu, chairman and chief executive officer of IntegraGen.

The company first tested the predictive value of its SNPs on 545 families from the Autism Genetic Resource Exchange (AGRE), a biobank of ‘multiplex’ families, which have more than one child with autism. The researchers then validated the findings in a second group of 627 families. They presented the data for 57 of the SNPs in May at the International Meeting for Autism Research in Toronto, but have not yet published the findings.

IntegraGen’s team pinpointed the SNPs from the AGRE database, and then prioritized SNPs that showed relatively high statistical significance in several large association studies. The SNPs are located in genes — such as ASTN2, CADM1, GRIN2A, NTRK3 and PCDH10, to name a few — that are expressed in the brain and involved in autism-linked pathways.

IntegraGen and others are also developing screens for children with no family history of the disorder. For example, SynapDX, a company located outside Boston, is working on an autism test that measures the expression of hundreds of genes in the blood. The test is not yet commercially available but has performed well in clinical studies, according to the company’s president.

Contentious combinations:

Independent experts approached by SFARI.org were reluctant to comment on the validity of IntegraGen’s test, as the methodology is unpublished. But based on other studies, they say it’s doubtful that combinations of SNPs can accurately predict autism.

For instance, an Australian group reported in September that a predictive panel of 237 SNPs in 146 genes can predict autism with 70 percent accuracy1. The researchers identified the SNPs by screening samples from AGRE, and then validated them in two other datasets.

One set of samples is from the Simons Simplex Collection, a biobank of families that have only one child with autism, sponsored by SFARI.org’s parent organization. The second batch of samples is from a Wellcome Trust cohort of healthy British people born in 1958.

Anney notes that the study may suffer from a ‘gene-size effect.’ Because of a quirk of the methods used, SNPs that fall in large genes are weighted more heavily even when they don’t necessarily confer more risk. He also notes that whereas the autism samples are from Americans of mixed heritage, the controls are from a comparatively homogeneous British group.

The data look interesting but really need to be tested independently before I would believe the results were scientifically sound,” notes Stephen Scherer, director of the Centre for Applied Genomics at the Hospital for Sick Children in Toronto, who was not involved in the work.

There is much room for error, both in study design and interpretation, and too much at stake here to get this wrong,” Scherer says. Researchers involved in the study did not respond to multiple requests for interviews.

The results from any genetic test for autism are bound to be complex. IntegraGen does not market the test directly to families, for example.

A doctor must order the test and send a sample to the company’s laboratory. The company’s genetic counselors then discuss the results with the doctor or parents. “This is not a simple glucose test, and the explanation of the results is important,” Courtieu says. He declined to say how many tests the company has sold so far.

IntegraGen is supporting clinical trials to refine its SNP panel and make the test more accurate, he adds. For instance, the company is sponsoring a prospective study at the Cleveland Clinic to test 600 children, about half with autism and half with either no psychiatric diagnosis or with other conditions, such as attention deficit hyperactivity disorder, speech delays or anxiety disorders.

“There’s a lot of work to be done to make sure that what they’ve done replicates into a larger population,” says Thomas Frazier, research director of the Center for Autism at the Cleveland Clinic and lead investigator of the clinical trial.

Any reliable test for autism would be an enormous help to clinicians, especially in cases that score on the borderline of gold standard behavioral tests, Frazier adds.

Still, for a disorder as complex as autism, genetic information is likely to contribute only one piece of the whole, Frazier says. “I don’t foresee us having a cadre of genetic tests that can do the whole trick.”


1: Skafidas E. et al. Mol. Psychiatry Epub ahead of print (2012) PubMed

2: Anney R. et al. Hum. Mol. Genet. 21, 4781-4792 (2012) PubMed


Name: yankeegirl
1 November 2012 - 2:26PM

Just the direction I expected. Genetic counseling for prospective parents. No wonder the Autism Genome project researchers want to collect our kids DNA. The true agenda. Not for treatment but to clean up the gene pool

Name: Tom Frazier
1 November 2012 - 3:59PM

Yankeegirl, I think you are missing the point. The test is not being used in utero and it is a risk-based test so it does not say whether a child has autism, it contributes to identifying risk (as do clinical observation measures and parent interviews). In a sense, this is a biological version of a parent interview. And one of the very nice things is parent interviews are not very reliable till age 2, whereas this test can be used earlier in families at higher risk to further inform their options.

Name: Yankeegirl
1 November 2012 - 6:24PM

Apart from doing a physcial examination of the DNA, I think you could just learn a lot from a child's family history. For instance whether there is a family history of autoimmune or neuropsychiatric disorders that could indicate a high risk for autism. No doubt in my mind that a sensitive subpopulation exists. Doctors should screen those babies early and put them on a much more conservative track with regard to provoking their immune systems- these kids have low tolerance.

Name: Venna
20 November 2012 - 2:36AM

Yankeegirl, what about those diagnosed with autism that have no family history of autoimmune or neuropsychiatric disorders? What about children also that have gone on to develop autism without the full schedule of vaccinations also? If there is a sensitive sub-population, as you suggest, they aren't in the groups you think they are. And I don't believe that autism in general has anything to do with the immune system.

Name: Yankeegirl
23 December 2012 - 3:50PM

Actually research is showing that autism does have a connection to the immune system Venna. Some autism like Retts and Fragiole X can be traced to a specific defective gene. Other forms of autism can be traced to immune pathway differences. We are talking "autisms" plural. So there may be different subtypes of Autism Spectrum Disorder.

Read this article in the Boston Children Hospital Vector blog

"A better early blood test for autism- Genetic signatures point to disrupted neuro-immune pathways"

"December 5, 2012

An Excerpt:

Boston, Mass.–"Researchers at Boston Children’s Hospital have developed a blood test for autism spectrum disorders (ASDs) that outperforms existing genetic tests, while presenting evidence that abnormal immunologic activity affecting brain development may help explain some of autism’s origins. The findings also suggest a new direction for genetic research on autism and the search for treatments.

The blood test, described December 5 in the online open access journal PLOS ONE and based on the largest gene-chip investigation ever done in autism, could enable early diagnosis of autism in about two thirds of patients before clear symptoms start to appear (the average age of diagnosis in the U.S. is 5 years)...

"The 55 genes whose expression was altered also suggest more than one path to what we know as autism. Based on their genetic signatures, subjects with ASDs clustered into four subgroups marked by changes in different biological pathways (see image 2):

· Synaptic pathways, specifically long-term potentiation pathways (essential for memory and learning) and neurotrophic pathways (signaling neurons to survive, develop and grow)

· Immune/inflammatory pathways

“In our sample, about half of the autism cases had some sort of alteration on immune pathways, synaptic pathways, or both,” says Kong."

Name: Jon Brock
6 November 2012 - 8:17PM

In addition to the problems noted in the article, there is a huge problem with base rates. That is, the reported specificity of a test assumes that there's a 50/50 chance of the person being autistic. In the general population, it's closer to 1 in 100.

For example, in the case of the Skafidas study, 70% specificity actually translates into 2% of people who test positive actually being autistic (because 30% of the 99% of people who aren't autistic also test positive).

The authors suggested that the test would be most useful for testing younger siblings of autistic kids, where the base rate is much higher. But it turned out that the test was actually very poor at discriminating between autistic kids and their non-autistic siblings, so it wouldn't be much use there either.


Name: Anonymous
12 November 2012 - 8:09AM

I don't know if I agree with your assessment on the utility of genetic testing.

I wonder how does one reconcile your statement on the utility of genetic testing with the fact that the sensitivity of ADOS or ADIR is approximately 85%. I assume that you are familiar with these tests and their limitations. Clinicians need to provide parents like yourself with more information so that you can appreciate the problems with the current testing methodology.

Name: Jon Brock
20 November 2012 - 1:13PM

Hi Anonymous

Specificity more than sensitivity is critical in relation to base rates. Poor specificity means that the overwhelming majority of people with a positive diagnosis are not actually autistic.

While the ADOS and ADI-R certainly have their limitations, I find it hard to believe that they would misdiagnose 30% of typically developing children or adults as being autistic, which is the relevant comparison.

Name: Sharon Luciano
18 January 2013 - 3:33PM

Dies the research indicate any link between in-vitro fertilization and autism? How about adult diagnoses?

Name: Greg Boustead
24 January 2013 - 11:36PM

Good questions, Sharon. Thanks for posting. We'll try to you an informed response soon!

Name: Virginia Hughes
26 January 2013 - 1:24PM

Thanks for your questions, Sharon. I'd like to emphasize here that these genetic tests are not prenatal tests. So in vitro fertilization would not be relevant to what the test is picking up.

As far as the overall link between in vitro fertilization and autism, the evidence is very weak. For example, two studies in July showed that there is no increased risk of autism after IVF or other infertility treatments, bolstering older studies:


Thanks for reading SFARI.org!

Name: Miss Lena
4 April 2013 - 6:21PM

A letter detailing concerns about the experimental design and validity of this study:

Name: Diane
4 September 2014 - 5:10PM

I would just like to add that there has been a study that pretty much concludes that autism could be caused by older fathers. It has to do with genomes. My husband and I have a 3 1/2 yr old with autism who is healthy otherwise and no family history on either side. My husband was 46 when we conceived. The study found that men as young as 30 carry a higher percent increase in fathering an autistic child and the rate goes up with age. I feel that we could very well fall into this category, but we haven't had any genetic testing yet.

Name: James
19 August 2015 - 10:26AM

My wife and I have found a clinic in Barcelona that does genetic tests to detect such risks and we are planning to go there to do genetic tests as we have history in our family. I let you the link to the website for thos who might be interested: http://www.genomicgenetics.com/

Add a Comment

You can add a comment by filling out the form below. Plain text formatting.

Question: What is 10 + 4 ?
Your answer: