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.

Experts critique statistics, conclusion of autism twin study

by  /  14 July 2011

Double trouble: Studying autism among identical and fraternal twins can help estimate how much of the disorder can be attributed to genetic factors.

A new study of twins proposes the controversial claim that environmental influences during early development are just as important as genetics, if not more so.

The findings, published 4 July in the Archives of General Psychiatry1, are not substantially different from those of previous twin studies, however, and some experts are critical of the study’s statistics.

Many articles in the popular press have misinterpreted the report as saying that genetics is not important in autism. But the point of the new study is not to point the finger solely at environmental factors, says lead investigator Joachim Hallmayer, associate professor of psychiatry at Stanford University, in Palo Alto, California.

“We should get this nature-nurture distinction behind us and embrace a broad approach to autism,” Hallmayer says. “We have to somehow better integrate environmental studies and genetic studies.”

The study is notable for its scrupulous methods: It’s the largest twin study to date that relies on the field’s gold standard tests to make autism diagnoses. Earlier studies based their findings on medical records and parent reports.

Genetic research on complex diseases often begins with twin studies. In the case of both identical and fraternal twins, the fetuses share a womb and other early environmental exposures. The environment in this case refers to a broad range of influences after conception, ranging from chemical changes in gene expression to the absorption of nutrients in the womb.

Because identical twins share all of their genes and fraternal twins share only half, comparing them allows researchers to tease apart the respective contributions of genetics and environment.

The first twin study in autism research was published in 1977, and seven more investigating the disorder’s prevalence have followed.

The first few studies found that if an identical twin has an autism spectrum disorder, the chance that the other twin will also have it is high, around 80 to 90 percent. By contrast, this so-called concordance rate was reported as much lower for fraternal twins, ranging from zero to ten percent2.

However, most of these older studies analyzed fewer than 50 twin pairs. Because autism crops up in only about one percent of the general population, the numbers are too small to make calculations with high statistical certainty.

Careful collection:

Based on data from 192 pairs of twins, the new study found a concordance rate for autism spectrum disorders of 77 percent among identical twin boys and 31 percent among fraternal twin boys.

The researchers then used mathematical models and the prevalence of autism in the general population to estimate that 38 percent of autism spectrum disorders are caused by genetic factors, and 58 percent by unspecified environmental factors.

Previous research teams generally found twins through clinics or patient groups, rather than sampling them from the general population. They also relied on records of autism diagnoses made using a variety of tests rather than diagnosing the children themselves.

In the new study, Hallmayer and his colleagues mined several California medical databases to identify twins born between 1987 and 2004 of whom at least one has autism. The researchers then performed thorough autism assessments using two standard tests: the Autism Diagnostic Interview-Revised (ADI-R) and the Autism Diagnostic Observation Schedule (ADOS). They diagnosed children as having either ‘strict autism,’ meaning that they met a certain list of criteria on both tests, or ‘autism spectrum disorder,’ indicating that they met a less stringent threshold.

“It’s the best twin study of autism, I think, to date because of the methods they used,” says Angelica Ronald, lecturer in quantitative genetics at Birkbeck, University of London.

“But the concordances are quite similar to what’s been published so far,” says Ronald, who published a review of autism twin studies in April3. Although the earliest twin research reported concordance rates among fraternal twins of less than 10 percent, newer studies have found rates around 30 percent4,5, similar to those in new study, she notes.

Still, Hallmayer contends that this is the first study using rigorous testing procedures to highlight the importance of environmental factors.

Like twins, non-twin siblings of children with autism are also at higher-than-normal risk of the disorder, with studies estimating the rate at about 10 to 15 percent6.

Because this is lower than the newly reported rate for fraternal twins, it’s likely that environmental factors related to prenatal development are increasing the risk for twins, Hallmayer says. “I think we have to look very carefully at intrauterine factors.”

For example, another paper in the same issue of the journal reports that pregnant women who take certain antidepressants increase their risk of having a child with autism7. There is also some evidence that maternal infection during pregnancy ups the risk of autism.

Identical issues:

Although this is the largest study of its kind, some experts say the new report is still too small to make wide-sweeping conclusions about the relative importance of genes and environment.

“They don’t have the numbers to state definitively that the environment is a greater contributor to autism at this point,” says Paul Law, director of the Interactive Autism Network (IAN) at the Kennedy Krieger Center in Baltimore. “It’s premature to say that, in my opinion.”

In 2009, Law’s team published the largest-ever study of autism twins, based on 277 pairs registered on the network, an internet-based registry of autism families. That study found a concordance rate of 31 percent for fraternal twins and 88 percent for identical twins5.

In the new study, the researchers base their claims largely on the high concordance among fraternal twins: 31 percent for autism spectrum disorders and 21 percent for strict autism, notes Michael Wigler, professor at Cold Spring Harbor Laboratory in New York.

But because of the size of the study, the researchers can only be statistically confident that the rate is between 16 and 46 percent for broader autism, and between 9 and 43 percent for strict autism. “That means any value inside that range is a possible real value,” Wigler says.

Even if the fraternal concordance rates are sound, Wigler adds, it’s likely that the researchers are underestimating the role of genetics in autism.

Even with diseases that are unquestionably genetic, such as congenital heart disease, sometimes one identical twin has the disease and the other does not, he notes. That’s because elements of chance are always involved in developmental processes. For example, spontaneous genetic mutations can crop up in one twin and not the other. Identical twins may also carry different chemical modifications to DNA that change how their genes are expressed.

“Where I think they really go wrong is in ignoring the general concept of penetrance in developmental disorders,” Wigler says. “You can be genetically predisposed and that doesn’t mean with certainty that you’ll have the disorder.”

For example, in March, Shiva Singh‘s group at the University of Western Ontario analyzed the genomes of identical twins among whom only one twin has schizophrenia. Singh’s team found that identical twins carry different spontaneous genetic mutations.

This sort of variation isn’t usually picked up by twin studies, notes Singh, professor of molecular genetics at the University of Western Ontario.

“We have lots and lots of heritability estimates in the literature, but most of these have not really led to the identification of actual genes that may cause the disease,” he says.

Until researchers compare full genetic sequences of identical and fraternal twins, all twin studies come to the same conclusion about nature versus nurture, he says. “We’re still in this limbo of some this, some that.”


  • Anonymous

    The final report of the California Autism Twin Study (CATS) has been published and created a great deal of buzz and excitement in that they interpret their data as suggesting that perinatal environmental factors may be far more important than has been suggested in previous twin studies in autism.

    It therefore may be useful to compare the results in two of the most recent autism twin studies, the CATS twin study (2011) and IAN twin study (2009) to the 1995 British twin study which is the most referenced autism twin study ever published (Bailey et al 1995 ) (Hallmayer et al 2011 ) ( Rosenberg et al 2009).

    The CATS study did state, correctly, that:

    ‘Three studies of twins ascertained from clinical samples with a total of 36 monozygotic pairs (concordance rate of 72%) and 30 dizygotic pairs (concordance rate of 0%) have estimated the heritability of autism, or proportion of liability attributable to genetic factors, at about 90%. The dizygotic concordance is substantially lower than the expected rate based on estimates of sibling recurrence rates’.

    One has to consider the 15 year gap in publication dates between the British twin study and the CATS and IAN twin studies. The 1995 British twin study was published concomitant with the introduction of DSM-IV (1994) with more broadly defined diagnostic criteria and additional sub-categories for autism. In twin pairs discordant for autism the ‘unaffected’ twins in the British twin study usually had a history of childhood language disorders or social impairments persisting into adulthood that would now might meet diagnostic criteria for PDD/NOS or Asperger Syndrome and may also explain the lower concordance rates in DZ twin pairs in the British twin study compared to more recent twin studies.

    Analyzing the underlying data in greater detail in these three important twin studies suggest that they are all in broad agreement with respect to MZ twins.

    The 1995 British twin study was published by Rutter’s group at King’s College ‘Autism as a strongly genetic disorder: evidence from a British twin study’. That study is perhaps the most important and most referenced study published since 1995. The 1995 British twin study replicates the CATS study more closely than what was suggested by the CATS study with respect to MZ twin pairs. The British twin study using a narrow autism definition did report a pairwise 60% concordance rate for MZ twins. When discordant twin pairs where the unaffected twin had a developmental problem not meeting strict diagnostic criteria for autism were included as concordant, the pairwise concordance rate rose to 92%, not all that dissimilar to what the CATS study reported, an MZ probandwise concordance rate of 58% for narrowly diagnosed twins and the combined group consisting of both narrowly defined (genotype concordance and phenotype concordance) autism and more broadly defined autism (genotype concordance but phenotype discordance) reported a probandwise 77% concordance rate for the combined groups.

    The IAN twin study published in 2009 did replicate the 1995 British twin study. At the IMFAR conference in London in 2008 the IAN group reported that concordance rates for narrowly diagnosed pairwise MZ twin pairs (genotype concordance and phenotype concordance) was 60% in both MM and FF MZ twin pairs, exactly the same 60% concordance rate reported in the British twin study. When MZ twins who were discordant for narrowly diagnosed autism (genotype concordance and phenotype discordance) were included, the pairwise concordance rate for MZ twins rose to 88% compared to the 1995 British twin study of 92%. The final IAN report did not include the data presented at the 2008 London IMFAR Conference by the IAN group ( Rosenberg et al 2009).

    The broad agreement between these three twin studies is even more remarkable since the methodology used in recruiting the twins was different in all three studies. The British twin study included twins that participated in a previous twin study and added additional twin pairs recruited from clinical referrals, national parent groups and special schools. The twins in the IAN twin study were all volunteer families recruited by Interactive Autism Network and the CATS twin study was a population based study that identified the twin pairs from records contained in the California Dept. of Developmental Services. The CATS study used a probandwise method of calculation heritability estimates while the British and IAN twin study used a pairwise method in calculating heritability estimates. The probandwise method usually calculates somewhat lower concordance rates than the pairwise method.

    All three of these twin studies are in broad agreement with respect to MZ twins, there is a 2:1 ratio between narrowly defined (or diagnosed) MZ twins and more broadly diagnosed MZ twins. Why is that important? Because the same 2:1 ratio of MZ twins by chorion type (monochorionic and dichorionic) (MC=66%,DC=34%) has been reported by the East Flanders Prospective Twin Survey. The EFPTS twin survey is one of the oldest and largest twin registries in existence and it is the only twin registry that has from the beginning systematically examined and recorded perinatal chorion data and the 2:1 ratio by chorion type was based on 2,000 pairs of MZ twins entered into the registry between 1964 and 2000 ( Blickstein & Keith 2006).

    The CATS study authors have concluded that “Future studies that seek to elucidate such factors and their role in enhancing or suppressing genetic susceptibility are likely to enhance our understanding of autism’. Are MZ/MC (single placenta) twins who may share the same prenatal environment more alike than MZ/DC twins who do not share the same prenatal environment (separate placenta). Segregating concordance rates in MZ twin pairs for narrowly diagnosed (genotype concordance and phenotype concordance) and broadly diagnosed (genotype concordance and phenotype discordance) by placentation status in MZ twin pairs present a rare and unique opportunity of elucidating the strength or weakness of the different interpretations made by the British twin study, the CATS study and the IAN study authors.

    DC-MZ twins result from separation of blastomeres within 72 h of ovulation; MC-MZ twins arise from later duplication of the inner cell mass suggesting different prenatal environments in these MZ sub-types. Different levels of exposure to environmental pathogens (alcohol) in multiple placenta pregnancies has been found in both humans and animals to vary significantly ( Gareri et al 2009).
    There is good evidence that developmental outcomes vary substantially in MZ/DC twin pairs compared to MZ/DC twin pairs in all sorts of ways, including studies that are directly related to many lines of research in autism. Studies that have confirmed zygosity and chorion type have reported significant differences between MZ/MC and MZ/DC subtypes. Skewed X-inactivation patterns have been suggested in Rett Syndrome as contributing to Rett Syndrome etiology. Trevo et al (1994) reported an unexpected finding, in newborn female MZ twin pairs, skewed X-inactivation patterns varied significantly in MZ/DC female twin pairs in contrast to female MZ/MC twin pairs who had closely correlated patterns of X-inactivation ( Trevo et al 1994 ). Jacobs et al (2001) found substantial variances in IQ scores (WISC-R) in MZ/DC twin pairs compared to MZ/MC twin pairs ( Jacobs et al 2001 ). Sokol et al (1995) found significant variances in scores derived from the McCarthy Scales of Cognitive Ability and Personality Inventory for Children tests in MZ/DC twin pairs compared to MZ/MC twin pairs ( Sokol et al 1995 ).

    Two studies that have not received much interest because they did not examine perinatal chorion data but instead used a controversial methodology. Finger, palm, and footprint data (dermatoglyphics) were scored for an index to retrospectively assess chorion type in MZ twin-pairs. Davis et al (1995) using dermatoglyphics found that concordances schizophrenia in MZ twins with one or more MC markers averaged 60 percent while concordance rates for schizophrenia MZ twins without MC markers averaged 10.7 percent ( Davis et al 1995). Reed at al (2002) using dermatoglyphics in elderly twins found greater structural variation in corpus collosum within pairs with dichorionic placentas ( Reed et al 2002).

  • Anonymous

    The California twin study reported a moderate genetic heritability estimate of 37%. This has been interpreted as suggesting that the genetic heritability for autism has been overestimated by previous twin studies. In fact, the moderate genetic heritability of 37% may itself be an overestimate. The rapidly advancing technology and rapidly lowering costs of genome scans has advanced so rapidly that is now routine that gene scans of probands and first degree relatives (parents and siblings) is becoming routine. All the excitement is now about the discovery that copy number variations (CNV) are associated with autism risk and that a high rate of CNV’s are de novo (not inherited). The definition of a de novo mutation is a reproductive error that is present for the first time in one family member as a result of a mutation in a germ cell (egg or sperm) derived from one of the parents or an error of cell division occurring during early fetal development and that the mutation is not present in the parent(s). De novo mutations are rare, random events that can strike within any family at any time.

    There is a profound flaw in classical twin study design that simplistically calculates a ‘heritability’ estimate based on the difference between concordant rates in MZ and DZ twin pairs. The classical twin study has been in effect, relatively unchanged for over one hundred years. The logical flaw in calculating ‘heritability’ estimates is that it assumes that ‘genetic’ = ‘inheritance’. The heritability estimates cannot control for de novo mutations that are not inherited. Completely overlooked in the excitement over CNV’s and de novo CNV’s is that about 10% or more of autism is associated with the single-gene disorders with high rates of co-occurring autism (Fragile X, Rhett Syndrome, Downs Syndrome etc.). The strong association between the single gene disorders with high rates of co-occurring autism (and the twin studies) are the twin associations that has led to the oft made claim that autism is the most heritable of the developmental disorders.

    What is the ratio of inherited gene mutations and de novo mutations in the single-gene disorders associated with high rates of co-occurring autism? The answer is surprising. Among the single-gene disorders Fragile X mental retardation syndrome stands virtually alone in following a Mendelian pattern of transgenerational inheritance, but in most of the single-gene disorders associated with high rates of co-occurring autism the mutations are primarily de novo mutations and are not inherited.

    Arthur Beaudet (2008) a leading authority on Prader-Willi and Angelmans Syndrome has observed ‘In terms of potential lessons for other disorders, it should be noted that almost all of these genetic and epigenetic cases of PWS and AS are de novo as contrasted to being inherited events’.

    The NIH has established a web accessible genetic reference table that is maintained and updated usually by the leading experts in the field who research the specific genetic disorder. The articles are written in clear English.

    Fragile X:

    Smith-Lemli-Opitz syndrome:

    Rhett Syndrome:

    Downs Syndrome:

    Klinefelter Syndrome:

    Angleman Syndrome:

    Prader-Willi syndrome:

    Tuberous Sclerosis:

    Williams Syndrome:

    Turner Syndrome:

    16P11.2 deletion syndrome:

    Phelan-McDermid syndrome:

    CHARGE Syndrome:

  • Anonymous

    This is a good example of a non argument. Genetics adapt to interact with environment. Only a fool would try to separate them.

    See Hemingway’s ‘Old Man and The Sea’ for an understanding of the genetically adapted fool. The male generally as in bird of paradise or silver plumed dove, is the part which tests the water for a viable gene, without risk to the gene itself or the gene pool generally.

    Mark (another fool) Life is Love is Faith XX

    There are many secrets locked away in this data, best wishes you who seek the code.

  • Anonymous

    One can certainly argue about statistics and models used to calculate % heritability. However, it seems extremely important that now 2 very large twin studies have come up with the same number for dizygotics – 31% concordance. The fact that this is clearly much higher than any estimate for siblings shows that the fetal environment is very important.

    The comment by Jensen on chorion status is also quite relevant and rarely mentioned.

    Wigler’s comment on penetrance also opens the door for environmental influences.

  • Anonymous

    How complex genetic and environmental interactions operate in contributing to human disease has been elegantly demonstrated by AIDS researchers. Multiple groups have identified lower copy number variations in a single gene, CCL3L1, a common genetic variant located in the region near chromosome 17q12 as associated with risk for infection after exposure to the HIV-1 virus. A meta analysis demonstrated that lower copy numbers than population norms in this gene increases risk for infection. Higher copy number variations than population norms in the same gene confers resistance for infection after exposure. No one is claiming that lower copy number variations in in the CCL3L1 gene is an ‘AIDS’ gene.…..t=abstract

    The claim by behavioral geneticists that copy number variations ‘cause’ autism is simplistic and ignores the observation that correlation does not imply causation

  • Anonymous

    Apologies. Broken link to the HIV study. Here’s the abstract:

    PLoS One. 2010 Dec 30;5(12):e15778.
    CCL3L1 copy number variation and susceptibility to HIV-1 infection: a meta-analysis.
    Liu S, Yao L, Ding D, Zhu H.
    SourceThe State Key Laboratory of Genetic Engineering, Institute of Genetics, School of Life Sciences, Fudan University, Shanghai, China.

    BACKGROUND: Although several studies have investigated whether CCL3L1 copy number variation (CNV) influences the risk of HIV-1 infection, there are still no clear conclusions. Therefore, we performed a meta-analysis using two models to generate a more robust estimate of the association between CCL3L1 CNV and susceptibility to HIV-1 infection.

    METHODS: We divided the cases and controls into two parts as individuals with CCL3L1 gene copy number (GCN) above the population specific median copy number (PMN) and individuals with CCL3L1 GCN below PMN, respectively. Odds ratios (ORs) with 95% confidence intervals (95% CIs) were given for the main analysis. We also conducted stratified analyses by ethnicity, age group and sample size. Relevant literatures were searched through PubMed and ISI Web of Knowledge up to March 2010.

    RESULTS: In total, 9 studies with 2434 cases and 4029 controls were included. ORs for the main analysis were 1.35 (95% CI, 1.02-1.78, model: GCN ≤ PMN Vs. GCN > PMN) and 1.70 (95% CI, 1.30-2.23, model: GCN < PMN Vs. GCN ≥ PMN), respectively. Either in stratified analysis, statistically significant results can be detected in some subgroups. CONCLUSIONS: Our analyses indicate that CCL3L1 CNV is associated with susceptibility to HIV-1 infection. A lower copy number is associated with an increased risk of HIV-1 infection, while a higher copy number is associated with reduced risk for acquiring HIV-1.

  • Anonymous

    There is a good review of some of the strengths and weaknesses of this study on the Autism Jabberwocky blog.



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: