News The latest developments in autism research.
Profiles Portraits of scientists who are making a mark on autism research.
Toolbox Emerging tools and techniques that may advance autism research.
Spotted A roundup of autism papers and media mentions you may have missed.
Opinion Conversations on the science of autism research.
Viewpoint Expert opinions on trends and controversies in autism research.
Columnists Dispatches from experts on various facets of autism.
Crosstalk Debates and conversations about timely topics in autism.
Reviews Exploring the intersection of autism and the arts.
Q&A Conversations with experts about noteworthy topics in autism.
Deep Dive In-depth analysis of important topics in autism.
Special Reports Curated collections of articles on special topics in autism.
Webinars Presentations by leading experts on their latest research.
Julia Yellow

Environment, genetics may contribute equally to autism risk

by  /  22 May 2014

A large study of autism in Swedish families suggests that about half of the risk for autism comes from genetics and the other half from environmental factors, researchers reported 7 May in The Journal of the American Medical Association1. However, the mathematical model the study used may not be appropriate for evaluating autism risk, experts say.

The study found that the risk of autism rises with increasing genetic relatedness to a family member with the disorder. For instance, a sibling of a child with autism has a tenfold greater risk of autism than the general population, whereas a cousin has double the risk.

Unlike most projects of this nature, the study looks at both the risk of autism’s recurrence in a given family and the role of genetics using the same population, says Diana Schendel, professor of psychiatric epidemiology at Aarhus University in Denmark, who was not involved in the study. “It’s confirming very strongly what we believed based on previous studies of smaller samples.”

Previous reports have pegged the autism risk for high-risk ‘baby sibs’ at 20-fold higher than for controls2,3. And twin studies have shown that identical twins are far more likely to share an autism diagnosis than are fraternal twins.

Together, these studies suggested that up to 90 percent of autism risk is genetic rather than environmental.

However, a 2011 study in California of 192 twin pairs found that more than half the risk could be chalked up to environmental factors such as the age of the parents. The mathematical model used in that study attributed 58 percent of the risk to environmental factors and 38 percent to genetics.

At the time, some researchers noted that the California study was not large enough to draw conclusions about autism’s causes. It also didn’t take into account spontaneous mutations, which play a big role in autism risk, says Michael Wigler, professor of genetics at Cold Spring Harbor Laboratory in New York, who was not involved in either study.

Family ties:

The new study uses the same mathematical model, but relies on a larger sample size of more than 2 million people.

“We brought in the biggest sample yet,” says Sven Sandin, researcher of epidemiology at Karolinska Institutet in Stockholm, who led the statistical analysis.  “That’s why we think that our results are more generalizable.”

Sandin and his colleagues looked at people born in Sweden between 1982 and 2006, counting people multiple times based on their number of familial relationships. The cohort includes 2,642,064 full-sibling pairs; 432,281 half-siblings who share a mother; 445,531 half-siblings who share a father; 5,799,875 pairs of cousins and 37,570 pairs of twins. In all, 14,516 children have a diagnosis of an autism spectrum disorder.

“The nice thing about this is it’s not only based on twins, but also other family members,” says Joachim Hallmayer, associate professor of psychiatry and behavioral science at Stanford University in California, who led the 2011 California study. He was not involved in the new research.

The identical twin of an individual with the disorder has an autism risk 153-fold higher than that of typically developing twins. The risk is about eightfold higher in fraternal twins and about threefold higher in half-siblings compared with the general population.

“Families who may be concerned about the recurrence of autism in their families can take the results here and have at least some information about the reality of risk,” says Avraham Reichenberg, lead investigator and professor of psychiatry and preventive medicine at Mount Sinai Hospital in New York.

The findings are in line with those from a 2013 Danish study of families with a history of the disorder. Although autism is about three times more common in males than in females in the new study, the recurrence risk for family members is slightly higher for siblings of girls with the disorder than for siblings of affected boys.

The recurrence doesn’t vary over the five-year chunks in which the researchers assessed the risk. This suggests that the causes of autism stayed the same despite its increasing prevalence, Reichenberg says.

Model point:

Experts have little complaint with that aspect of the new study. The researchers then used a mathematical model to estimate the contributions to autism risk from genetic and environmental factors.

Specifically, they found that ‘additive genetic effects,’ or the number of genetic factors linked to autism that are inherited from each parent, account for 50 percent of the risk. Non-shared environmental influences account for the other 50 percent. These might include the flu a woman has during a pregnancy, which would not increase the risk for children born from other pregnancies.

“It will not necessarily be the same genes and the exact same environment that lead to autism in two families,” Reichenberg says.

The researchers found that shared environmental factors, such as the milieu in the womb, have negligible effects on autism risk, but didn’t specify which factors might be influential. In contrast, the California study pegged only shared environmental influences, such as low birth weight in twins, in autism risk.

Hallmayer says it’s difficult to directly compare the findings of the twostudies because one looked at twins and the other looked at families.

Interestingly, a Swedish twin study of 16,858 individuals reported in 2010 that genetic effects account for 80 percent of autism risk in that cohort4. The fact that both Swedish studies used the same model and found such variable results highlights its limitations, researchers say.

Because models cannot capture all variables, they are inherently “crude” measures, says Schendel. “With heritability, as with any modeling, you make some assumptions about what is happening in the real world to simplify your model,” she says.

The model assumes, for instance, that the genetic and environmental factors that cause autism do not interact with one another, says Kenneth Ye, associate professor of epidemiology and population health at Albert Einstein College of Medicine in New York. Ye and Wigler frequently collaborate on projects.

Also, the model is best used for disorders caused by small effects from multiple genetic factors. “I think we know from autism studies that there are a good proportion of cases [in which] it comes from one of a few mutations,” Ye says.

This means that the model may be inappropriate for parsing autism risk, he says. A 2007 study by Wigler and his colleagues suggests that in many cases, autism is caused by a few mutations — maybe only one — with strong effects and multiple targets5.

Knowing the severity of the disorder in each individual and whether his or her parents have any symptoms might help to determine whether the disorder is passed on from the mother, the father or both, Wigler says. “Is it a case where one [has] strong [autism symptoms] and the half-sibling is on the spectrum? One would like to know that,” Wigler says.

Not everyone is as critical of the model, but say it does have flaws.

For example, the model assumes that random effects from the environment may increase autism risk, but random mutations also increase risk, says Aravinda Chakravarti, professor of genetics at Johns Hopkins University School of Medicine in Baltimore, who was not involved in the study.

“If you do not specify [what constitutes] the environment, then saying something is environmental as opposed to a random outcome is not a scientific conclusion,” he says.

The researchers say they are aware of these shortcomings, but believe the model still gives a broad view of autism’s causes.

“All we need is a crude measure,” Sandin says. “Is it 90 percent or is it 10 [percent genetic]? I believe that this model can deliver that.”


1: Sandin S. et al. JAMA 311, 1770-1777 (2014) PubMed

2: Bolton P. et al. J. Child Psychol. Psychiatry 35, 877-900 (1994) PubMed

3: Ozonoff S. et al. Pediatrics 128, e488-495 (2011) PubMed

4: Lichenstein P. et al. Am. J. Psychiatry 167, 1357-1363 (2010) PubMed

5: Zhao X. et al. Proc. Natl. Acad. Sci. 104, 12831-12836 (2007) PubMed


  • RAJensen

    Classical twin study design cannot measure ‘heritability’. It can only indicate whether there is a genetic (not heritable)component. Down syndrome (1-750) and Klinefelter syndrome (1-500 male newborns) are the most common genetic syndromes associated with increased autism risk. These syndromes are almost always caused by a reproductive error,a de novo sperm or egg mutation. Simply calculating a ‘heritabilty’ estimate based on the difference in concordance rates seen in identical twins and fraternal twins would give Down syndrome and Klinefelter sydrome the highest ‘heritabily’ estimates, much higher than autism. Classical twin studies also cannot measure the influence of the prenatal environment because they do not segregate MZ twin differences based on chorion status, monochorioic (single placenta, same prenatal environment) or dichorionic (seperate placentas, different prenatal environment). If the twinning event occurs in the first three days after fertilization the pregnancy will be dichorionic (1/3rd of twin pregnancies). If the twinning event occurs more than three days after fertilization the twin pregancy will be monochorionic (2/3rds of twin pregnancies). Autism twin studies need to report their MZ twin data by chorion type. A number of studies have done so and demonstrate the importance of the prenatal environment and chorion status.

  • Alan Packer

    Thanks for such a thorough treatment of a complex subject. It’s becoming increasingly apparent that the early estimates of the heritability of autism may have been at least somewhat inflated, though the ‘true’ number is probably still substantial. The trouble comes when we try to estimate the environmental component of risk. As Dr. Chakravarti notes, we have to specify what we mean by ‘environmental’. If the conclusion from the new Swedish study is correct—-that the environmental component of risk is ‘non-shared’—-then that is a very big umbrella indeed. It covers, for example, intrinsic developmental variation that probably contains a significant element of randomness. But I suspect when most people think about environmental risk, they think of prenatal and postnatal exposures to extrinsic factors. These things could be involved, but are not necessarily implied by the current study. As such, any claim about the scope of environmental influence on ASD risk really should be attached to a discussion of the full range of factors that could fall under that heading.

    • RA Jensen


      One of the great mysteries and controversies in autism is the problem of the ‘missing heritability’ and the inflation of ‘heritability estimates’ in classical twin studies. Autism is a complex multifactorial condition and there must be a ‘missing heritability’ component that simply hasn’t been recognized as such. I have an invited review paper that explores the problem of the ‘missing heritability’ in autism. The missing heritability in autism in my view is what has come to be labelled the broader autism phenotype (BAP) or autistic-like traits. BAP traits are common, familial and widely distributed throughout the general population. The BAP is a background genetic effect that is always reliant on other genetic, epigenetic and environmental risk factors that cumulatively increase autism risk in any individual case.

      Jensen RA. The background genetic effect of the genes underlying the broad autism phenotype as a unifying feature in gene x gene and gene x environment causal mechanisms in autism. OA Autism 2013 May 01;1(2):11
      The full Open Access article is available here:

  • autism parent

    UC Davis MIND Institute researchers recently confirmed that many kids with autism have an underlying mitochondrial dysfunction. Mitochondria, which supply cellular energy to all the organs in the body, are known to be sensitive to toxins and viruses in the environment. If the mitochondria fail, then the organ will not function properlyand a systemic dysfunction happens – this may explain the behaviors we see in autism (stimming, loss of language, moto delay, sensory issues) So if there is a genetic sensitivity to environmental factores why not look at vulnerability in the mitochondrial DNA that predisposes these children to disease. It would confirm that a sensitive subset of children exists that should be treated more carefully. To identify these kids, look for a family history of a mitochindrial related disease which could provide a valuable clue. Doctors could prescreen these babies early before a dysfunction happens.

  • passionlessDrone

    Hello friends –

    The data in this study that should be opening everyone’s eyes is the stratospheric difference between monozygotic twins and *everything else*. From the paper:

    **The RRR and rate per 100 000 person-years for ASD among monozygotic twins was estimated to be 153.0 (95% CI, 56.7-412.8; rate, 6274 for exposed vs 27 for unexposed ); for dizygotic twins, 8.2 (95% CI, 3.7-18.1; rate, 805 for exposed vs 55 for unexposed); for full siblings, 10.3 (95% CI, 9.4-11.3; rate, 829 for exposed vs 49 for unexposed); for maternal half siblings, 3.3 (95% CI, 2.6-4.2; rate, 492 for exposed vs 94 for unexposed); for paternal half siblings, 2.9 (95% CI, 2.2-3.7; rate, 371 for exposed vs 85 for unexposed); and for cousins, 2.0 (95% CI, 1.8-2.2; rate, 155 for exposed vs 49 for unexposed).**

    Being a monozygotic twin carries a risk more than *fifteen times* that of a dyzogotic twin, and even more than any other kind of relationship. By what possible mechanism could this be driven genetically? As noted above by Robert Jensen, MZ twins frequently inhabit *the same* placenta, something absolutely unique to that population. There have been a few papers recently documenting placenta differences in the autism population:

    and another one at IMFAR2014

    This raises the possibility that the twin studies used to determine heritability, and the *presumed* genetic driven nature of autism, were missing the point. Food for thought.

    • Mark Carew

      The outcome of the headline study indicates 50/50 split environment and genetics. This outcome is of course driven by the fundamental and inescapable truth that those genes themselves are an outcome of life evolving genetically in an environment. It seems to me this basic truth does not need a scientific proof or any further controversy. Despite my lack of scientific training I can see that the twinning environment itself must be a risk factor (it seems very big risk factor) which surely must invalidate a large portion of the genetic-autism-twin statistics. Autism is a problem which is driven by our environment, and its subsequent input to our gene pool, our genes we cannot change.

  • kathryn page

    “…that the risk of autism rises with increasing genetic relatedness to a family member with the disorder. For instance, a sibling of a child with autism has a tenfold greater risk of autism than the general population, whereas a cousin has double the risk…”

    We cannot leave out a teratogenic effect, which will hit identical more often than fraternal twins, and which may be higher in cousins where the teratogen is a family tradition–like drinking. Many autistic-like symptoms are caused by prenatal alcohol exposure. To infer genetic effect from higher family prevalence is myopic.

  • ASD Dad

    This work flows into studies surrounding viral and bacterial immune activation in the maternal environment as an important area of autism etiopathology.

    Japanese researchers have investigated viral pathogens (influenza) in pregnancy, trophoblasts and schizophrenia outcome.

  • Jonathan Sebat

    From the standpoint of the genetic models that were used for additive and dominant effects, this study is deeply flawed. It does not adequately account for the contribution form spontaneous de novo mutation. High MZ concordance, low DZ concordance, and VERY LOW recurrence risk in 2nd and 3rd degree relatives is the hallmark of highly heritable dominant neurodevelopmental disorder. Down syndrome for instance is 100% genetic and dominant. The MZ twin concordance for Down Syndrome is 95-100% and DZ twin concordance is virtually nil. It would be interesting to apply the exact same model (and the same dataset) to estimate the liability to down syndrome (DS) that could be explained by the environment. I expect that this estimate will a bit inflated.

    • passionlessDrone

      Is Down Syndrome an appropriate analogy to autism? The manifestation is clear, unambiguous, and the result of a single genetic anomaly that was identified over fifty years ago.

      This is quite a bit different from the way genetics are supposed to be acting in autism, with dozens, or hundreds of candidate genes providing very low penetrance, and a very wide manifestation of effect. Is the biggest similarity between DS and autism that MZ twins seem to have a larger footprint when we use twin studies to evaluate for heritability?

      I’m curious on your thoughts on the gulf between MZ twins and all other relationships in this study. Shouldn’t MZ, DZ and full siblings all share ~ 50% of the same genes with each other? If so, and genetics are the predominant cause, why would we see such a huge drop off from MZ to the rest of the population? Why would the *other* populations seem to show a stepwise progression you might expect; i.e., DZ and full siblings (sharing ~ 50 genes) show some risk, half siblings show ~ 1/2 that risk, and cousins showing a smaller risk again. If the model was wrong, wouldn’t these numbers also look screwy?

      As far as de novo mutations, it has occurred to me that perhaps MZ twins are suffering from spotaneous de novo mutations at a higher rate than others, perhaps as a function of the cleaving process at an early state of embyogenesis. (?)

      But again, we are left in a place where *other* highly dominant neurodevelopmental disorders are characterized by single, high penetrant genetic impacts that can be easily identified, but autism is (?) characterized by spontaneous de novo mutations. This may be true, but I don’t think that the hunt for spontaneous de novo mutations has done a particularly good job at finding anything that can explain more than a fraction of autism cases, though I may be wrong.

      • Jonathan Sebat

        Mr. Drone, I’m not sure where what your conception of the genetics of autism is based on. Most mutations that have been strongly linked to autism are in fact mutations with high relative risks (of between 10 and 50). These include FMR1 (also associated with fragile X syndrome), MECP2 (Rett Syndrome), PTEN (cowden syndrome), and several CNVs including deletion of 16p11.2 and duplication of 15q11-13. The patterns of familial inheritance for these mutations is similar to what is observed for Down Syndrome. Most mutations occur spontaneously (or very recently in an ancestor)… they tend NOT inherited from the (asymptomatic) parent. This is why DZ twin concordance is low… in order to observe the same mutation in both twins, it would need to be carried by the mom or the dad.

        DZ twins share 50% of their inherited genome. But they share virtually none of their spontaneous mutations. It makes no sense to estimate heritability based solely on the percent of shared DNA. You must take into account the spontaneous mutations that are not shared… and which are definitely contributing to risk.

        No MZ twins are not subject to more mutations. We sequenced the complete genomes of identical twins with autism, and they roughly the same number of mutations as you or I do.

        The hunt for de novo mutations is an approach that is clearly working. It’s the best going at the moment. It doesn’t explain all of autism of course, but it is not a tiny fraction either. The percent of cases that can be explained based on the current data is optimistically 20%. But that’s based on sequencing only 1% of the genome. There is probably more that we are missing.

        I try to resist the temptation to make radical assumptions about what accounts for the unexplained risk. Phenomenon that are now unexplained, will not stay that way forever. Some of it will be due to yet undiscovered genetic or epigenetic factors (or combinations of the above). And yes some if it will be nurture rather than nature. We should seek out credible evidence for both kinds of factors. And soft evidence (such as the Reichenberg paper) should not be sensationalized the way it is.

        • passionlessDrone

          Mr. Sebat –

          “Mr. Drone, I’m not sure where what your conception of the genetics of autism is based on.”

          I guess the fact that such extremely rare genetic conditions such as the list you provided as the ones usually presented as ‘strongly linked to autism’ helped guide to me to my conclusion. My status as a rank amateur also likely contributes.

          The mutations and associated conditions you point out are very, very rare compared to autism. Cowden’s syndrome, for example, is estimated to have an incidence around one in two hundred thousand . Rett Syndrome is around one in twenty thousand girls. While the manifestation is clearly extreme, and indeed, highly penetrant for these conditions, using whatever autism prevalence de-jour from the past couple of years doesn’t let us approach more than a fraction of autism being caused by those mutations. I’m not trying to be daft, but if a mutation is present in one in twenty thousand girls, how can we assume that it is having a meaningful effect on autism prevalence, at say, one in 100? If those are our mutations with the most robust evidence set behind them, where does that leave the rest in terms of strength of evidence?

          Another way of looking at it is the despite being very rare, the genetic misprint has been identified in these extremely rare cases. You seem to be up to speed on the genetics much more than me, could you explain to me like I’m a fifth grader some possibilities why the genetic mutation for Cowdens with high penetrance has been identified in a tiny population, but similar high penetrant mutations in a much larger population (idiopathic autism) have remained hidden?

          “DZ twins share 50% of their inherited genome. But they share virtually none of their spontaneous mutations. It makes no sense to estimate heritability based solely on the percent of shared DNA. ”

          OK. But, do you find the fact that a (expected?) stepwise progression of risk in associations *other than* MZ twins to be interesting? I guess where I’m going is that if we expect the model to be flawed for failing to take into consideration only shared DNA, wouldn’t we expect to see curious data across the relationship spectrum as it were, as opposed to only when evaluating MZ twins compared to everyone else? I’m especially intrigued on this considering you mentioning that MZ twins share a similar rate of de novo mutations as the rest of us. From the article above;

          “The identical twin of an individual with the disorder has an autism risk 153-fold higher than that of typically developing twins. The risk is about eightfold higher in fraternal twins and about threefold higher in half-siblings compared with the general population.”

          According to the author, those findings aren’t the modelling component that some people are generally having problems with; they don’t assign genetic vs environmental risk ratios, just ratios between relationships. If de novo mutations occur at the same rate in MZ twins and the rest of us, how could they be participating in this finding, which doesn’t invoke a mathematical model to achieve ‘soft’ results? I understand if you are hesitant to make radical assumptions, but at the very least, aren’t these values problematic for the idea that de novo mutations are causing the increased autism risk in the MZ population?

          “The percent of cases that can be explained based on the current data is optimistically 20%. But that’s based on sequencing only 1% of the genome. There is probably more that we are missing.”

          Fair enough. If we’ve scanned such a small percentage of genome, but can explain a fifth of autism, does that tell us anything about the similarity between what we’ve already found and what lies in waiting? The problem I’m having wrapping my head around is based on these numbers, if we sequence 2% of the genome, we might expect to find 40% of autism, and at such a rate our numbers become meaningless pretty quickly. Perhaps this is a function of intelligent searching, as opposed to random searching. (?)

          “I try to resist the temptation to make radical assumptions about what accounts for the unexplained risk”

          It’s pretty much all I’m good at. 🙁

  • Jonathan Sebat

    One minor but important correction to my post. Twin concordance rates for Down Syndrome are estimated at 90% (MZ) and 7% (DZ) respectively. These numbers would be closer to 100% and 0% if not for the fact that occasionally there is somatic mosaicism for the spontaneous mutation. Mosaicism in the parent leads to an occasional recurrence in their offspring. Mosaicism in the offspring contributes to a decrease in MZ concordance. These are randome genetic effects. Thus, taking de novo mutation AND somatic mosaicism into account, a neurodevelopmental disorder that is completely genetic would be predicted to have twin concordance rates in the ballpark of 90% (MZ) and 10% (DZ). Autism is not 100% genetic. Better estimates of the heritability based on more accurate genetic models are needed. The Reichenberg study, however, is not a step forward.

    • Alexandria

      Does anybody know of hereditability and/ or concordance rate studies for trisomy conditions other than Down Syndrome? If so, do these studies provide any guidance as to how to think about ASD hereditability?

      • RAJensen

        Klinefelter syndrome is genetic syndrome with a high prevalence rate of 1 in 500 newborn males. It is always caused by an extra X chromosome producing an XXY genotype in the fertilized egg in contrast to the normal XY genotype. About half the cases are caused by an XY sperm mutation paternally derived and half are caused by an XX egg mutation maternally derived. Klinefelter syndrome is associated with a high risk for co-occurring autism. The very few case reports of twins have found that Klinefelter syndrome in MZ twins are concordant and Klinefelter Syndrome in DZ twins are not concordant. Klinefelter syndrome is not inherited although very small numbers of cases, as is seen in Downs, are mosaic with some normal XY cells and some cells with the XXY mutation. Several studies have examined for the presence of XY sperms in karyotype normal fathers of KS boys, subfertile males and in healthy volunteer controls using the FISH assay. XY sperms were found in all groups suggesting that all males may produce XY sperm throughout their lives and the production of XY sperm increases with advancing paternal age.
        McAuliffe and colleagues discovered that increasing levels of environmental exposure to polychlorinated biphenyls (PCBs) and dichlorodiphenyltrichloroethane (DDT) congeners, as measured in blood, is associated with increased production of XY and YY sperms in volunteer donors recruited from fertility clinics.

        For further discussion see my paper:

  • ASD Dad

    It would be remiss not to remind fellow readers that the general ‘gist’ of this research was identifying a significant role of environment as well as gene interaction was underpinned by Joachim Hallmayer and his team

    Genetic heritability and shared environmental factors among twin pairs with autism.

    • gregboustead

      ASD Dad – Definitely an important reference point when considering the results of similar segregation modeling applied to a large cohort in this new work. Indeed Joachim Hallmayer was interviewed, and coverage of his 2011 California twin study referenced, in the article above.

  • DR01D

    Thank God the human genome project wasn’t completed until 2001. If it had been completed in 1901 scientists would have pinned Polio on heredity and we still wouldn’t have a vaccine.

  • Rachael Sommerfeld

    Where can i find the study to read it please? I want to know the heritability % of any offspring i might produce as a neurotypical baby sib of asd-asperges brother.
    Thanks in advance. I’m pretty desperate to know this!


Log in to your Spectrum Wiki account

Email Address:



Request your Spectrum Wiki account

Spectrum Wiki is a community of researchers affiliated with an academic or research institutions. To be considered for participation, please fill out this form and a member of our team will respond to your request.


Email Address:

Title and Lab:

Area of Expertise: