For over two decades now, so-called “autism experts” have been claiming that autism is more than 90% caused by genes. The influence of these claims on autism policy and research funding is hard to overstate. But few realize that the basis of these claims hangs on a fragile evidence base: two small twin studies--one from Great Britain, the other from Scandinavia--that reported high rates of concordance for autism among identical twins and no concordance at all among fraternal twins. Last week, the largest and most rigorous twin study ever conducted, the California Autism Twin Study (CATS) reported contradictory new evidence that struck a devastating blow to these claims. The CATS identical twins had lower and the fraternal twins higher concordance rates than past studies, a striking finding that suggests that instead of being highly heritable, the vast majority of autism cases stem from environmental causes.
It’s hard to overstate the importance of the CATS findings. They mean that everything leading “autism experts” have been saying for decades is wrong. And everything leading autism parent advocates have been saying for years is right.
The foundations of autism orthodoxy
As the reality of the autism epidemic began settling in over the last decade, an odd drumbeat in the writing of autism geneticists became more insistent. The more obvious it had become that something new and terrible was happening to a generation of children, the more extreme the statements of the genetics researchers became. It’s as if repeating the orthodox statements as frequently as possible would give them more weight. And the more their extreme claims went unchallenged, the more a spurious “scientific consensus” could be claimed. Here’s a small sample: (1) “Autism is clearly among the most heritable of all psychiatric disorders“(2003); (2) “Autism is one of the most heritable complex disorders, with compelling evidence for genetic factors and little or no support for environmental influence” (2004); (3) “Autism spectrum disorders are considered to be among the most heritable of all mental disorders…. recent reviews estimate the heritability of autistic disorder to be more than 90%.” (2010); (4) “ASDs are known to be highly heritable (~90%)” (2010).
Despite the fact that an explosion in autism rates rendered illogical any ongoing belief in the genetic inheritance of autism, the influence of this orthodox position on autism’s research funding remained profound. Hundreds of millions of research dollars were spent over the last decade in a vain hunt for autism genes; a spending binge that has continued unabated, with over $100 million spent at NIH on genetics-only research during the latest two years of the Combating Autism Act alone.
In the meantime, a mini-industry of epidemic denial has emerged among academics willing to posit a newfound popularity among physicians and parents for the autism label, one that has produced increasingly bizarre claims of diagnostic excess in many forms: substitution, oversight, expansion and accretion among them. These claims have been retracted, disavowed and falsified multiple times, yet because of the overriding need to feed “the hungry lie” these claims keep cropping up in novel forms. Meanwhile, the research funding to support them continues, including active support from NIH. And sadly, anyone in the scientific community with the courage to stick their necks out and suggest that autism rates might be going up because, well, there were more cases of autism, found themselves facing reactions ranging from polite ostracism to ruthless censorship.
The basis for this orthodox belief in autism’s heritability rests on a very specific body of autism research: the investigation of concordance rates within twin pairs. These studies take advantage of a seemingly simple test--the presence or absence of similar outcomes in twins--to estimate the relative contribution of nature (i.e., genes) and nurture (i.e., the environment) to a given disease. To the extent that identical twins have the identical outcomes and fraternal twins have different ones, the cause can reasonably be assigned to genes. To the extent that identical twins have different outcomes and fraternal twins have the same outcome, the cause lies in the environment.
In the case of autism, this test has been applied to twins with increasing frequency, but the orthodox belief in heritability hangs on a slender thread of evidence: the first, a British study of twin pairs first recruited in 1977 (5) and then expanded in 1995 (6); the second, a Scandinavian study from 1989 (7). The 1977 British study reported on just 21 twin pairs, 11 identical and 10 fraternal. Of these, only 4 of the identical twin pairs were concordant for autism (a remarkably low rate that is frequently forgotten). The 1989 Scandinavian study with the same small sample size, 11 identical pairs and 10 fraternal pairs, found autism in both identical twins in 10 of 11 pairs and again no concordance in any of the fraternal pairs. Subsequently, nearly two decades after their first study, the British team recruited 28 new pairs in 1995 and pooled them together with some of the previous group. These new pairs showed a similar profile: only one twin in each of new fraternal pairs had autism (i.e., a zero concordance rate); and of the new identical pairs, 11 of 16 were concordant, giving a 69% concordance rate that fell between their original calculation and the Scandinavian group.
The British team then added an important new element: a heritability calculation. Using a complex approach (with formulae that defy clear explanation), they took all these concordance rates banged them against two different background rates of autism, ran them through a model and declared, “The estimates of broad heritability were 93% and 91% for the base rates of 1.75 and 10 in 10,000.”
And thus was born the belief that autism was more than 90% genetic.
The orthodoxy in quiet crisis: discordant evidence on twin concordance
Oddly enough, the 1995 date of this publication coincided with the first awareness among public health officials of the autism epidemic, with an inflection point generally assigned to the birth years 1989-90 (8-9) and with awareness of this shift emerging with a lag as these new birth cohorts were being diagnosed 3-5 years later. Because of this confluence of events—the convergence of some “autism experts” around a self-proclaimed “scientific consensus” on autism’s cause and an epidemic rise that directly contradicted that consensus—for over 15 years now, autism research has been hijacked by a bizarre cognitive dissonance between a deeply held belief in autism’s heritability and the impossibility of a genetic epidemic.
Interestingly, there was plenty of evidence available, even as early as 1995, to show that the extravagant conclusions of the British team weren’t true. Most notably, other collections of twin reports found much higher concordance rates in fraternal twins. First among these was a 1985 twin study from Utah (one with a larger sample than either of the initial British and Scandinavian studies) that found a 24% concordance rate among fraternal twins (10). Because this rate was too high for the orthodox scientists to explain (the Utah study authors found such a high autism rate among identical twins they suggested that autism might come from a single recessive gene) the Utah findings were effectively written out of history. The reason? The Scandinavian authors argued that the Utah twin registry began recruiting twins by putting an advertisement in an autism newsletter. And since the evidence from Utah didn’t fit the orthodox story, the evidence was rejected in order to keep the orthodoxy intact.
But just as the 90% heritability mantra was spreading in the years following the 1995 British paper (aptly titled to promote the message “Autism as a strongly genetic disorder: evidence from a British twin study”), more discordant evidence began accumulating. A 2003 Missouri twin study that collected (but didn’t report) autism concordance rates contained an intriguing parenthetical comment (11). Referring to the previously reported differences in the ratio of identical to fraternal autism twin concordance, the authors described these previously reported relationships as “concordance ratios greater than 2 (which we did not observe in our sample).” Another discordant note came from a 2008 twin study from Japan (12), which reported results that were nearly identical to the Utah findings, but with an even higher 31% concordance rate in fraternal twins. Reinforcing the dissonance, a 2009 study using an America registry of autism cases reported results that were nearly identical to the Utah and Japan studies (13).
Alongside all this other discordant evidence came the early rumblings from what eventually became the California Autism Twin Study (CATS). The study of California twins had begun as much as a decade before last week’s CATS publication and hints of their eventual findings have leaked out before. One 2002 abstract (14) on California twins raised the idea that “heritability estimates from previous studies may have overestimated the role of genetics and underestimated the role of environmental factors in the etiology of autism.” The formal launch of the CATS project took place a couple of years later, in July 2004; even so the dimensions of this early suggestion took many more years to come out into the open.
“It was better than CATS”
Orthodox scientists will undoubtedly spin the California results, arguing that the study was flawed in some fatal respect and that other studies were better than CATS. But actually, it’s pretty hard for a twin study to be much better than CATS; certainly none of the previous autism twin investigations come close. The California project is without question the most ambitious study of autism twins ever attempted. First of all, CATS is the largest twin study, enlisted from a registry based on the entire California autism population and including a final sample of 192 twin pairs, more than four times larger than any other previous twin study and nearly an order of magnitude larger than the first British and Scandinavian studies. The registry based approach (identifying twins from a complete roster of California autism cases rather than selective outreach) also makes the CATS sample less vulnerable to recruiting bias.
Perhaps most important at all, it’s hard to argue that the CATS author group is biased against inherited genes and in favor of environmental causation. The author group includes a healthy mix of orthodox autism scientists: Lisa Croen and Judy Grether are among the original “epidemic deniers” in autism (although their early claims of “diagnostic substitution” in California proved incorrect and, to their credit, they retracted them); Claire Lajonchere and Janet Miller are leaders of AGRE, Autism Speaks’ “autism genetic resource exchange”; first author Joachim Hallmayer and Linda Lotspeich of Stanford have been members in good standing of the Autism Genome Project Consortium; and senior author Neil Risch has been active in autism genome scanning work for over a decade.
What did CATS find that turned previous assumptions about nature and nurture in autism upside down? In brief, they found concordance rates in identical twins that were substantially lower than the Utah and Japan studies and rates in fraternal twins that were far higher than those found in Great Britain and Scandinavia. Identical twin concordance rates in CATS were significantly lower than rates of ~95% reported (and criticized) in Utah and Japan: 43% for full syndrome autism and 59% for ASD. At the same time, fraternal twin concordance was higher than the 0% rate reported in the British and Scandinavian studies, at 8% and 13% for autism and ASD respectively.
Using the concordance data, the CATS team went on to perform a number of elaborate calculations (again, using a complex approach with formulae that defy clear explanation) to assign theoretical causality for autism to three sources: inherited genes (A), common or shared twin environment (C) and non-shared (they also called it “random”) twin environment (E).
The calculations resulted in a dramatically lower rate of heritability (A) than anyone expected, 37% and 38% for autism and ASD, respectively. By extension, the shared twin environment estimates of 55-58% were far higher than expected. And of course, the mere concession that the environment took precedence over genetic causation in autism (even if it was only a 40/60 effect), made news. After all, going from 91-93% heritability to 37-38% was nearly a two thirds blow to the autism gene hunt in a single stroke.
But the talking points that made the news don’t do justice to the real weight of the evidence. Crucially, the low 37-38% number for “A” dramatically overstated the role of heritability, a point the authors conceded in the paper’s fine print, where they acknowledged two major biases.
1. The study design explicitly excluded the possibility of an environmental effect being mediated by genetic vulnerability in a subset of children. In other words, their ACE model excluded the possibility of gene-environment interaction. According to the authors, “The ACE model we used has several inherent assumptions. First, it assumes no gene environment interaction. If such a gene environment effect does exist, it would be confounded with the A parameter in our analysis, implying that as an estimate of pure genetic effect, A may actually be an overestimate.”
2. A more subtle bias of the study involved the way their model treated the twins’ environment in the womb. Twins are known to have variable gestational environments but the authors’ model assumed that both identical and fraternal twins faced the same “shared twin environment.” In fact, identical twins share a fetal environment (via a shared placenta or amniotic sac) far more often than fraternal twins (who almost always share neither). To the extent that the ACE model assigned the higher concordance of identical twins to heritability, the authors’ calculations excluded the fact that identical twins not only have identical genes, but a more nearly identical environment as well. The authors point out how this simplifying assumption raises their heritability estimate. “Similarly, a critical assumption in the model is that the shared twin environmental effect is the same for monozygotic [identical] and dizygotic [fraternal] twins. If, in fact, monozygotic twins share the relevant environment to a greater degree than the dizygotic twins, some of the effect included in the parameter A would actually be environmental rather than genetic; again, A may actually overestimate the true genetic heritability.”
It’s hard to guess how much lower the 37-38% heritability estimate would fall if these two biases were removed. But the effects are more likely large than small and full adjustment for these two biases would almost certainly turn the nature/nurture causation estimates on their head: instead of a world in which >90% of autism causes come from inherited genes, it’s more likely that >90% of its causes come from the environment.
Which is exactly what so many autism parent advocates have been saying for years.
Separated at birth
In the enthusiasm over the move from a 90/10 gene-environment split to one that the CATS team conceded was closer to 40/60, there’s one additional subtlety that has gone overlooked. In a crucial oversight, the authors calculated estimates for both A, C and E, but once they dispensed with genes (A) and turned to the environment, the only variable they reported was the “shared twin environment” (C). In the process, they completely ignored E, or the “non-shared twin environment.” For those interested in environmental causation, this non-shared component E is far more intriguing than C. It includes everything in the twin experience that is unique to an individual twin. Inside the womb, for example, there might be differences between two different amniotic sacs or placentas that put one twin more at risk for autism than the other. More to the point, the twins might experience different environmental exposures outside the womb, exposures that could also account for different autism risks. These non-shared exposures are of special interest because they provide the strongest evidence for exposures that can be avoided so that autism, even in a genetically vulnerable child, can be prevented.
In CATS, the amount of causality assigned to the non-shared twin environment was estimated (although never directly reported) at 4-8%. It’s perhaps not surprising that the numbers were that low, since identical twins are often reared in highly similar environments. There environments are more similar than most siblings and even than fraternal twins. It takes a rare set of circumstances to vary the environment of identical twins enough to provide different autism risks.
In an extreme case, however, twins will maximize their non-shared environment if they are separated at birth.
In our book, The Age of Autism: Mercury, Medicine and a Manmade Epidemic, Dan Olmsted and I report on the most extreme cases of non-shared post-natal environment in autism twins ever described. One of these was a pair of non-concordant, identical twin boys, given up for adoption at birth by a woman named Kim Stewart who retained contact with the adoptive families. Both families provided a loving environment for their sons, both children lived in similar parts of the country and developed normally, as “healthy, happy boys,” for the first fifteen months of life. The first joined a family of Christian Scientists, received no vaccinations of any kind and continued to develop normally. The second joined a family that followed more conventional infant vaccination schedules and regressed into autism after receiving his fifteen month shots. The only non-shared twin environmental risk the birth mother could pinpoint was that separating the boys at birth led one to be fully vaccinated and autistic and the other to be unvaccinated and healthy. “I became convinced,” she told us, that “vaccinations played a significant role in his regression.”
* * *
The body of twin evidence in autism as most recently revealed by the CATS study demonstrates the power of scientific orthodoxy to sustain itself in the face of overwhelming evidence when its believers have strong research interests and the power to perpetuate a hungry lie. In situations like this, putative claims of “scientific consensus” can become a pernicious influence when that consensus favors a convenient lie.
My friend JB Handley has made this point before and pointed the readers of Age of Autism to a compelling quote by the late Michael Crichton MD, author of The Andromeda Strain and Jurassic Park, who had this to say about scientific consensus.
"I want to pause here and talk about this notion of consensus, and the rise of what has been called consensus science. I regard consensus science as an extremely pernicious development that ought to be stopped cold in its tracks. Historically, the claim of consensus has been the first refuge of scoundrels; it is a way to avoid debate by claiming that the matter is already settled. Whenever you hear the consensus of scientists agrees on something or other, reach for your wallet, because you're being had."
Mark Blaxill is Editor at Large of Age of Autism.
1) Yonan AL et al. A genomewide screen of 345 families for autism-susceptibility loci. Am J Hum Genet. 2003;73(4):886-97.
2) Veenstra-Vanderweele J, Christian SL, Cook EH. Autism as a paradigmatic complex genetic disorder. Annu Rev Genomics Hum Genet. 2004;5:379-405.
3) Lichtenstein P et al. The genetics of autism spectrum disorders and related neuropsychiatric disorders in childhood. Am J Psychiatry. 2010;167(11):1357-1363.
4) Pinto D, Pagnamenta AT, Klei L, et al. (Autism Genome Project Consortium) Functional impact of global rare copy number variation in autism spectrum disorders. Nature. 2010;466(7304):368-72.
5) Folstein S and Rutter M. Infantile Autism: A Genetic Study of 21 Twin Pairs. J Child Psychol Psychiatry. 1977;18(4):297– 321.
6) Bailey A, et al. Autism as a strongly genetic disorder: evidence from a British twin study. Psychol Med. 1995;25(1):63– 77.
7) Steffenburg S, et al. A Twin Study of Autism in Denmark, Finland, Iceland, Norway, and Sweden. J Child Psychol Psychiatry. 1989;30(3):405– 16.
8) McDonald ME, Paul JF. Timing of increased autistic disorder cumulative incidence. Environ Sci Technol. 2010;44(6):2112-8.
9) Blaxill MF. What's going on? The question of time trends in autism. Public Health Rep. 2004;119(6):536-51.
10) Ritvo ER et al. Concordance for the syndrome of autism in 40 pairs of afflicted twins. Am J Psychiatry. 1985;142(1):74– 7.
11) Constantino JN, Todd RD. Autistic traits in the general population: a twin study. Arch Gen Psychiatry. 2003;60(5):524-30.
12) Taniai H et al. Influences on the broad spectrum of autism: study of proband-ascertained twins. Am J Med Genet B Neuropsychiatr Genet. 2008; 147B:844– 849.
13) Rosenberg RE et al. Characteristics and concordance of autism spectrum disorders among 277 twin pairs. Arch Pediatr Adolesc Med 2009;163:907–914.
14) Croen LA, Grether JK, Hallmayer J. A population based study of autism among twins in California. IMFAR, Orlando, FL, November 2002:69.