Dec 21 2009

Autism Prevalence

The Centers for Disease Control (CDC) has published the results of their latest study on the prevalence of autism. There is no question that in the last 20 years the number of autism spectrum disorder (ASD) diagnoses has increased. What is also clear is that during this time there has been increased surveillance and a broadening of the diagnosis of ASD. Whether or not this accounts for the entire increase in ASD numbers, or if there is a true increase in there as well, is unknown.

Into that context, the CDC adds their most recent numbers, concluding:

In 2006, on average, approximately 1% or one child in every 110 in the 11 ADDM sites was classified as having an ASD (approximate range: 1:80–1:240 children [males: 1:70; females: 1:315]). The average prevalence of ASDs identified among children aged 8 years increased 57% in 10 sites from the 2002 to the 2006 ADDM surveillance year. Although improved ascertainment accounts for some of the prevalence increases documented in the ADDM sites, a true increase in the risk for children to develop ASD symptoms cannot be ruled out. On average, although delays in identification persisted, ASDs were being diagnosed by community professionals at earlier ages in 2006 than in 2002.

That 1 in every 110 children on average now carry an ASD diagnosis is not new news. This CDC data was actually released ahead of publication in October. At the same time a phone survey published in Pediatrics found 110 in 10,000 children carried an ASD diagnosis – or a little more than 1%. This 1% figure seems to be highly replicated – a National Health Services survey released in September also found a prevalence of 1% for ASD in the UK.

Of interest, Autism Speaks chose to release their own press release regarding the CDC numbers, writing:

The CDC report, published in this week’s Morbidity and Mortality Weekly Report (MMWR), states that 1% or 1 in every 110 children has been diagnosed with autism, including 1 in 70 boys. This represents a staggering 57 percent increase from 2002 to 2006, and a 600 percent increase in just the past 20 years. Other significant findings include that a broader definition of ASDs does not account for the increase, and while improved and earlier diagnosis accounts for some of the increase, it does not fully account for the increase. Thus, a true increase in the risk for ASD cannot be ruled out….

“This study provides strong evidence that the prevalence of autism spectrum disorder is, in fact, dramatically increasing,” said Geraldine Dawson, Ph.D., Autism Speaks chief science officer, who noted that recent research indicates that a significant amount of the increase in autism prevalence cannot be explained by better, broader or earlier diagnosis.

This is somewhat misleading. It is true that increased surveillance alone cannot account for the increase, neither can the broadening of the diagnosis by itself. But it is possible that the two together can account for the entire increase. The fact that a real increase “cannot be ruled out” is purely a statement about the statistical power of the epidemiological evidence – it is not the equivalent of evidence for a real increase.

The CDC study argues that broadening of diagnosis does not account for the 2002-2006 increase in ASD prevalence, according to their data. But this is not surprising, as the changes in diagnosis had occurred prior to 2002. The increase from broadening the definition of autism had already occurred.

The epidemiological evidence, in fact, argues against a real increase. That same National Health Service survey I linked to above found that the prevalence of ASD was about 1% in every age group they surveyed. If true prevalence were increasing, then we would expect that autism prevalence would increase with younger age groups. That every age group they looked at had about the same prevalence speaks powerfully to the conclusion that true autism prevalence is static.

But that is not all. Recent evidence suggests that ASD caseloads increase at times when the diagnostic criteria are changed.

Another study published in Pediatrics found that diagnosis rates are probably below the true estimated prevalence rates, leaving room for increase through better diagnosis. They also found what is called “diagnostic substitution” – as the ASD diagnosis increased, other related diagnoses decreased, concluding:

“Prevalence findings from special education data do not support the claim of an autism epidemic because the administrative prevalence figures for most states are well below epidemiological estimates. The growing administrative prevalence of autism from 1994 to 2003 was associated with corresponding declines in the usage of other diagnostic categories.”

Notice the timeframe – 1994-2003 – before the recent CDC data where they conclude there was no diagnostic substitution.


ASD diagnostic prevalence has increased in the last 20 years has increased and is now around 1%. However, the epidemiological evidence strongly suggests that most or all of this increase is due to broadening the diagnosis and increased surveillance – leading to more and earlier diagnosis.

This still means that autism is a huge health issue that deserves funding and attention. It does not imply, however, that there is some causal factor, such as an environmental factor, that is increasing the prevalence of autism.

However, it is also undeniably true that a real increase in prevalence cannot be ruled out, although it would only account for a part of the total increase. The fact that older age groups also have about a 1% prevalence strongly suggests that any real increase in ASD must be small. But it could be there.

A real increase could be due to many factors. One strong possibility is increasing paternal age. People are having children later in life, and older age is associated with an increased risk of many disorder. A recent study found that increasing paternal age also increases the risk of autism:

In adjusted models that included age of the other parent and demographic covariates, a 10-year increase in maternal age was associated with a 38% increase in the odds ratio for autism (odds ratio = 1.38, 95% confidence interval: 1.32, 1.44), and a 10-year increase in paternal age was associated with a 22% increase (odds ratio = 1.22, 95% confidence interval: 1.18, 1.26).

We would therefore expect a real increase in ASD from this factor alone. There may be other factors as well. Much attention is paid to environmental factors, but so far none have been proven – including vaccines.

If the increase in ASD prevalence is largely due to increased surveillance and a broadened definition, with perhaps a small real increase thrown in, then we would expect the measured prevalence rates to eventually level off. We may be getting close to that point now. Only time, and further epidemiology, will tell.

22 responses so far

22 thoughts on “Autism Prevalence”

  1. cyberdraco says:

    There is something else my biology professor mentioned that seems plausible, although admit-tingly there isn’t much evidence for causation or correlation.

    His nephew has high functioning autism and is amazing with physics and mathematics, but suffers the draw-back of not being very social or having the natural facilities to socialize.

    He hypotheses that a mutation(s) for increased intellect in academia comes with the decreased ability to interact with other people.

    Are there any studies about linking those?

    I have only met a few people with autism and spoke with them or their parents, but in all cases they go together.
    It’s highly speculative, I know, but it does make one wonder.

    If I had a choice, I think I would give up part of my ability to interact with fellow humans face-to-face for an increased capacity to perform higher mathematics and the like.

  2. Skeptico says:

    It’s worth pointing out the obvious, namely that there has been no decrease – although Thimerosal was removed from childhood vaccines in 2001. Remember when David Kirby said this:

    Because autism is usually diagnosed sometime between a child’s third and fourth birthdays and thimerosal was largely removed from childhood vaccines in 2001, the incidence of autism should fall this year.

    I do. It was in 2005.

  3. Skeptico – you are write, and I have been hammering Kirby about that ever since.

    cyber- others have speculated that the rise of engineering (the alpha-nerds) has provided selective pressures for those on the spectrum who are highly functioning and successful in science and engineering fields. Interesting idea – but I have seen no data to back it up.

  4. cyberdraco says:

    The data to back up the idea of ‘alpha nerds’ would take a cunning experiment to determine any type of causation.
    Trying to isolate variables alone would be daunting, then how would you setup up controls? Engineer’s without any form of Autism?

    Indeed it might be too speculative to ever properly test for, but if we do see a genuine increase in both areas, it will be worth remembering and trying to solve.

  5. HHC says:

    With genetic engineering, a definite enviromental factor, there is an increased likelihood of defective zygotes and gestation problems. This is a problem of our new age of science.

  6. maybrick says:

    Is there a higher incidence of “genius” amongst the autistic than in the general population?

    I always assumed the “link” between autism and genius (i’m not sure if thats the correct word) was something of a myth made famous by Rain Man and maybe Forest Gump.

  7. daedalus2u says:

    I do think that position on the autism spectrum is a “feature”. I lay out my hypothesis in my blog post on “theory of mind vs theory of reality”.

    I think the 3 main features of autism, social isolation, communication difficulties and repetitive motions are evolved features that facilitated the production of stone tools.

    One of the most important abilities of an inventor or innovator is the ability to ignore people telling you “it can’t be done”, or “you are wrong”. Many people can’t do that even when they are correct.

    Making stone tools requires significant practice and manual dexterity, dexterity that can only be obtained by long hours of practice while producing nothing of value. Early human and pre-human ancestors had to develop these skills while their peers were telling them they were wrong and to stop banging stones together uselessly.

    I think that the only way the many genes that seem to be involved in autism could have evolved together is if they encode a “feature” with a positive survival benefit. I think that is a contingent feature, a feature that is expressed under conditions of maternal stress in utero, the only time a brain can be fundamentally re-wired, while the neurons that comprise it are undergoing proliferation, differentiation, epigenetic programming and apoptosis, setting up the gross neuroanatomy that will determine function later in life. It is in the first trimester that the number of minicolumns is fixed (one of the most characteristic physical features of ASDs). It is during the first trimester that teratogens can cause autism. Position on the ASD spectrum can be discordant between monozygous twins. There is more to autism than just DNA.

    I think the trade-off happens because the maternal pelvis is limited in size, which limits the size of the infant brain that can be successfully born. A gigantic brain is such a survival feature that cephalopelvic disproportion kills a significant fraction of mothers without access to C-section. Optimizing brain function is obviously extremely important when brain size is limited. I think that is the evolved reason for people to have different positions on the autism spectrum.

  8. daedalus2u says:

    The problem of “genius” is how to define it and how to measure it.

    When different tests give answers that are 70 percentile points different on the same individual, those different tests cannot be measuring the same thing.

    Exactly what they are measuring is not known.

    There is a tremendous tendency to label as trivial tasks that nearly everyone can do, and label as difficult tasks that only a few can do.

    Language acquisition by children is considered trivial because everyone can do it. Computers cannot and many person-years of effort have not figured out a way for computers to do it. Mathematical calculations that humans find intractable are trivial for computers.

    Everything is easy once you know how to do it. Figuring out how to do something for the first time is what is difficult.

  9. ausduck says:

    Steve, has there been any breakdown of the age of the parents for the children that are being diagnosed on the spectrum? HHC sort of sparked a bit of an idea re zygotes. We’re having children when we’re older now (esp. here in Australia), and thanks to the progress in IVF as well as careers/lifestyle there are a lot of first time parents that are well into their 30s or 40s. I’d be interested to know if that factors in at all.

  10. provaxmom says:

    I know that there is lots of speculation and hypothesizing about the ‘geek syndrome;’ the notion that smarter/scientific parents are more likely to have autistic children. But I have yet to ever read a concrete study on it. I know that anecdotes hold little weight ’round this place 😉 but I find it to be true with most spectrum individuals that are in our social circles. Then again, I think it might be interesting to note that individuals who are of a scientific nature, also might be more career focused, and therefore more likely to have kids at an older age. Thus only substantiating the older dad=more likely to be autistic connection.

  11. Andrew Bienert says:

    The engineering link has certainly been looked at.

    A Simon Baron-Cohen study from 1997.

  12. daedalus2u says:

    It appears that there are about 15% of autism cases that can be “explained” through genetics, this includes Rett Syndrome (due to lose of MeCP2 gene) and some others. Each of the different mutations, deletions, duplications, etc only amounts to less than 1% of autism cases but collectively they all amount to ~15% (but the statistics on this are variable and poor because they are all rare events).

    There appear to be two broad classes of autism; autism associated with dysmorphic facial features and autism not associated with dysmorphic features. Dysmorphic types of autism tend to be singleton that is there is one individual in the extended family that exhibits the phenotype. Many (but not all) of the autism cases due to deletions and duplications are of the dysmorphic type. These can be inherited, but many are de novo. Non-dysmorphic autism is what tends to clusters in extended families. Asperger’s is a type of non-dysmorphic autism. It is the non-dysmorphic autism that has the most complex genetics with little data connecting phenotype with genotype.

    I think one of the main problems with autism genetics research is trying to combine these two data sets. It is in this sense that many researchers say that “autism” isn’t one disorder, but is many because each individual “cause” (i.e. deletion, duplication, etc.) can be (and in some ways should be) thought of as a different disorder.

    This is the problem that those looking at the genetics of autism are facing; it is extremely complex and heterogeneous, with no single gene amounting to more than a few percent of the “causation” of autism.

    There is an association of increased paternal age with autism, there doesn’t appear to be an association of increased maternal age. There is the hypothesis of increased gene-damage-load due to replication errors in germ cell stem cells in aged fathers, but there isn’t any published data that actually looks at the genetics of autism in the context of aged fathers. It may be that there are epigenetic differences. Different epigenetic programming of gametes in aged fathers compared to younger fathers and that might be the cause. There might be environmental effects, a pregnant woman with an aged partner might exhibit greater stress and exposure to that stress in utero might affect the ASD status of the fetus. The father-age research was done in Israel, a place where many individuals experience significant stress. Exposure to psychosocial stress in utero does increase the incidence of ASDs.

  13. beche-la-mer says:

    I read this morning that the man who inspired the character of Rain Man died recently; and also that later in life his diagnosis had been revised. Apparently he was not autistic, he lacked the nerves that normally link the two hemispheres of the brain.

  14. HHC says:

    Andrew Bienart, Read the British study at the link site. Its nice to know that engineers are interested in joining societies which support providing information to autistic parents. But I really think the coincidences between joining the society and autism in the family is related to interest in learning how things work, how the mind-body connenction gives you a complete and whole human being. The statistics in favor of the engineers being overrepresented in the group may be their scientific curiosity.

  15. Andrew Bienert says:

    I have read the study a number of times myself, as well as quite a few other from S. Baron-Cohen. Being both an engineer and having a son with Aspergers Syndrome I find it interesting, although do not claim this an any kind of evidence. But you said, “The statistics in favor of the engineers being overrepresented in the group may be their scientific curiosity”.

    The study however claims to have controlled for this:

    “Similar information was also collected from 4 control groups: (a) parents of children with Tourette Syndrome (TS), via the Tourette Syndrome Association (TSA) in the UK (n = 40 couples). This served as a control group, to test if patterns of occupations of parents of children with autism or AS were a function of the sorts of people who become members of a national charity focusing on a childhood psychiatric disorder. Given that TS can be associated with autism, only “pure” cases of TS were included. (b) Parents with a child with Downs Syndrome suffering from cardiac disease, via the charity, Downs Heart, in the UK (n = 464 couples).”

    Perhaps I’ve missed something, so I’d be interested to know if you have another take? Or if you could clarify?

  16. provaxmom says:

    My son has dup15q syndrome. Yes, he does have dysmorphic facial features, and it has been reported that dup15q is found in about 3-5% of kids with autism. Chromosome 15q11 – 13 duplications are the most frequently identified chromosome problem in individuals with autism. Both of these reports are 10+ years old, but the technology to test for this is new. And you’d have to factor in the doctors that don’t recommend genetic testing. My son’s features are very slight, not pronounced like a child with Down’s.

    Schroer RJ, Phelan MC, Michaelis RC, et al.: Autism and maternally derived aberrations of chromosome 15q American Journal of Medical Genetics. 1998;76:327-336.

    Wang CH, Villaca-Norat E, Papendick BD: Molecular analysis of the chromosome 15q11-q13 region in children with autism American Journal of Human Genetics. 1998;63.

  17. HHC says:

    Andrew Bienert, Studies in genetics show that human characteristics are distributed along a bell curve, a normal curve. Autism and mental retardation are at the extreme ends of the intelligence curve. If you take this part of the intelligence curve and then try to overlay a curve involving all professionals with handicapped children in Britain, you will find an enigmatic statistic. The study question involves a post hoc analysis of British engineers who joined a particular society because they believe their child to have autism. A clearer pattern in the social psychology of linkage of British careers with autism would examine the frequency as a whole in Britain in contrast to one particular association. Why do individuals join a Charitable society? They wish to provide monies to this group. This brings in the human questions of whom do you know, do your engineering friends want to provide economic benefits to a particular group? Are there regional or familial genetic concerns pertaining to intermarriage among groups in an area of Britain? Are the diagnostic categories among the children correct, or is the charity reaching out to certain groups that may have a fear involving social interaction and acceptance? These are basic questions which one asks when correlations are found post hoc among groups.

  18. Andrew Bienert says:

    “Autism and mental retardation are at the extreme ends of the intelligence curve.”

    That’s not quite accurate is it? Autism doesn’t necessarily equate to mental retardation or even higher or lower intelligence. Skewed yes, but I wouldn’t say at the extreme either. It also depends on which aspects of intelligence you are looking at as well.

    Also you say:
    “The study question involves a post hoc analysis of British engineers who joined a particular society because they believe their child to have autism.”

    Did it? That’s not how I read how the study was designed. Is that how you interpret it, or is that statement simply a criticism of the study and the design of it? The study nowhere indicates either if the respondents “believe their child to have autism”. That would be clearer if the actual questions were available to us, but I don’t think that can be assumed either way.

    Do you think the study was poorly designed and is invalid? Calling it a post hoc analysis suggests that you think this might be the case?

    To me it seemed a simple enough study, which set out a prediction, i.e. “We predicted that if their parent shared this cognitive phenotype, then they should be over-represented in engineering as an occupation”.

    And the authors also make it clear that: “We wish to stress however that the majority of engineers have no connection with autism, and the majority of parents of autism have no connection with engineering. Nevertheless, this link between the two phenomena merits further research.”

    The “merits further research” statement I think is the important conclusion from the study, which you yourself indicated with performing a wider study of the population. However if the study is completely invalid due to methodology; then fair enough I guess.

  19. HHC says:

    Andrew Bienert, I do not like the design of the study because I do not feel it answers any real issues. It is circular logic to argue that a particular small group of engineers who selected to joined an autistic society may have autistic children. There is a larger research question here, but what aspect of intelligence and career choice would you study? Some social science studies will lead you to greater understanding of social problems and other studies because of flaws in hypotheses or design lead you nowhere. I would assume engineering questions and research solutions are handled in the same way as well thought out basic social research and its conclusions.

    As for broadening the psychiatric definition of autism, I have difficulty with this concept. My specific conclusions would be based on neuropsychological evaluation of the individual. If there is damage in a specific area, or if skills are truncated, I would note this as part of an assessment. But to metaphorically paint everyone I assess with the category of autism is an overbroad stroke. Its like saying that humans have various forms of intelligence but everyone has limitations in social intelligence.

  20. HHC says:

    Third sentence should read group of engineers who selected to join an autistic society. The point revolves around self-selection.

  21. nickshakra says:

    It is nice that there is a lot of research and discussion as to what causes ASD. But lets give praise and attention to those programs that currently offer the best solution possible to this problem. Companies like rethink autism – – offer an online ABA curriculum with over 400 video lessons that allows parents and caregivers to supplement existing treatment – thats worthy of my praise.

Leave a Reply