Autism -- back to blaming the mother?

Two recent reports of the cause of autism reach different conclusions, though they are alike in that neither implicates genetics, at least not directly.  The first, published in the International Journal of Epidemiology ("Maternal lifestyle and environmental risk factors for autism spectrum disorders," Lyall et al.), reviews the evidence for environmental risk factors and finds that diet can influence risk, and that folic acid supplements taken around conception are associated with reduced risk.  Further,

Although many investigations have suggested no impact of maternal smoking and alcohol use on ASD [autism spectrum disorder], more rigorous exposure assessment is needed. A number of studies have demonstrated significant increases in ASD risk with estimated exposure to air pollution during the prenatal period, particularly for heavy metals and particulate matter. Little research has assessed other persistent and non-persistent organic pollutants in association with ASD specifically.

Lyall et al. call for larger epidemiological studies of maternal exposure to vitamins, fats and other nutrients, as well as pesticides and endocrine-disrupting chemicals, even though environmental epidemiological studies of autism have been done for decades.

The second paper, in  Molecular Psychiatry ("Elevated fetal steroidogenic activity in autism," Baron-Cohen et al.), reports the results of looking at hormone levels in amniotic fluid samples collected at between 15 and 16 weeks gestation from a sample taken from a registry of nearly 20,000 male infants in Denmark, born between 1993 and 1999.  The final sample was fairly small, including 128 male infants with autism and 217 controls; the 24 females in the registry who were later diagnosed with autism were excluded from the study because they were atypical for a variety of reasons.  Prevalence of autism is generally higher in males.

We find that amniotic fluid steroid hormones are elevated in those who later received diagnoses on the autism spectrum. Rather than the abnormality being restricted to a specific steroid hormone, a latent steroidogenic factor is elevated, which includes all hormones in the Δ4 pathway, as well as cortisol.

The effect on the developing brain, Baron-Cohen et al. suggest, may be epigenetic.  That is, steroids modify DNA in ways that affect gene expression without changing coding sequence.

Steroids and their receptors act as epigenetic fetal programming influences on early brain development. Through their nuclear hormone receptors, steroids can alter gene expression via direct or indirect influence on multiple epigenetic processes such as histone acetylation, DNA methylation and have transcriptional and post-transcriptional effects on noncoding mRNAs such as microRNAs. Furthermore, during early sensitive periods of brain development, there are sex differences in DNA methylation, methyl-binding proteins, chromatin modifications and microRNA expression, and these effects are mediated in part by early steroid hormone effects.

What is the source of the excess steroid?  "The fetus, the mother, the placenta or other external factors" -- that is to say, it could be anything and this study couldn't answer that question.  Indeed, it is also impossible to know, if the excess hormone really is involved, whether it's the cause of the disorder or the result.  Perhaps maternal stress is the source, the authors suggest, and perhaps, the authors note, steroids such as testosterone and cortisol are also elevated in other disorders with a skewed sex ratio.  In any case, they write, "Each of these sources require further investigation to determine how such influences might affect fetal development in autism."

Cortisol molecule

A story on the BBC website about this work quotes an autism "expert" saying that this is "an important first step" on the path to discovering what causes autism.  First step!? This is a curious way to describe things, since probably billions of dollars have been spent in the last 30 or 40 years on efforts to identify the cause of this disorder, much of it on genetic studies, with no robust results.  Given this track record, what criteria should we use to decide whether this study is worth paying any attention to?

As with many complex diseases and disorders, many genes with small effect have been identified, but none of these explains the high rates of autism now reported around the world.  It is interesting to see these two reports of possible environmental risk factors after a sea of genetic studies, though.  Decades ago autism was believed to be the result of "refrigerator mothering," but then blame swung toward genes and away from environment, and now it seems autism is epigenetic. The gene switch never could have been exactly right given the dramatic, rapid increase in prevalence of autism, and other than because genes are techy and faddish, why would one ever expect genes to be a main cause in the first place, other than as a rationale to do genetics (which we knew how to do) and a paucity of other ideas?  Or, environmental causes being difficult to replicate and confirm.

But many epidemiological studies looking for environmental causes have been done.  A 2010 paper in Current Opinion in Pediatrics reports, with respect to environmental risk factors, e.g.:

...the most powerful proof-of-concept evidence derives from studies specifically linking autism to exposures in early pregnancy – thalidomide, misoprostol, and valproic acid; maternal rubella infection; and the organophosphate insecticide, chlorpyrifos. There is no credible evidence that vaccines cause autism.

Older mothers and fathers have been associated with autism, birth order, toxic chemicals, vaccines and thimerosol, and so forth, though none reliably so.  And of those factors that have been replicated, they can't explain all cases.

Autism is a difficult trait to study.  The trait itself is hard to define, varies enormously, there are no biomarkers with which to make a definitive diagnosis, diagnostic criteria have changed over the years, and so forth.  But many traits are similarly complex -- asthma, schizophrenia, heart disease, etc. -- and similarly resistant to current methods for determining cause.  So it seems fair to assert that many attempts to determine causes of complex traits are fad-following approaches to understanding complexity with reductionist science.

[ ] is a complex disease

The November 1 issue of Nature has a special section on autism.  Or better put, on all the things we don't know about autism, and how that has sobered researchers.  Ten years ago, or even fewer, the focus would have been on the hunt for genes for the disease, but now the recognition has set in, as this section shows, that this is a complex disorder, and genes that explain it aren't going to be found, brain scans don't yield simple answers, there is no cure, even if we knew causative genes there would be no cure, and so on.

GWAS have indeed identified hundreds of genes that are associated with autism, but they explain perhaps a percent or two of the cases.  The long-standing hope that this would be another Mendelian trait shows just how much Mendel influenced and in fact sidetracked the understanding of disease, albeit inadvertently.  The essence of a Mendelian trait, be it normal or disease, is that it is due to one (or, perhaps two or three) genes and mainly to two alleles (functionally variant states) at the gene.  This means that if one of the alleles is 'dominant' its effects are always manifest in an individual, and one has a 50% chance of inheriting the allele--and hence getting the trait--from a parent who has the allele.  Recessive traits behave similarly--one allele is responsible but only if you inherit two copies of it, one from each parent.  Again, each transmission normally occurs 50% of the time.  We're oversimplifying, but only a bit, and the gist of the message is as we describe (and we've described it before; here, e.g.).

A brief primer on 'Mendelism'
Mendelian traits segregate with these 50/50 probabilities. Sometimes the chance of having the trait even if you have the allele is less than 100%, so dominance (or what is known, mysteriously, as 'penetrance') is 'incomplete'.  When the trait really is just a single-gene trait we can understand it even with incomplete penetrance.  Even if non-genetic factors cause some instances, which are traditionally known as 'phenocopies', we can still make correct inferences, though the ability to predict a newborn's trait is compromised.

Traits that segregated in Mendel's pea plants

There are legions of well-known Mendelian diseases, of course, and we've had methods for finding genes to explain them for decades.  That has been the job of the professional clinic-associated people known as genetic counselors (because they advise parents of potential risk to their future children).  The problem is that this success is matched by the general fact that Mendelian disorders are usually quite rare in the population.  But our major health concerns today are common, not rare, and they are not Mendelian.  They aggregate in families, so that if a relative is affected your chance of being affected is raised, but they don't segregate with neatly estimable probabilities, and the reason is that they are due to the effects of many genes, each individually very small, plus complex environmental factors.

Back to autism:
So, it's complex traits like autism that we're left with now, and for these our methods are lacking.  But the problem can be described in much the same way for all complex traits.  Take this paragraph from the commentary on the genetics of autism ("Genetics: Searching for answers"), for example:

The large databases of autism gene candidates that are now available make the quest to explain autism more complicated than researchers had hoped. But the complexity of the condition is stimulating the expansion of approaches taken and enticing scientists to look beyond straightforward genetic explanations for autism. “We've figured out that explaining autism is not simple,” says Geschwind. “But I have a pretty optimistic view. We're going to continue to make progress — and a lot of it is because of great collaboration in the field and an influx of new people tackling autism.”

Substitute any complex disease for autism and it could be equally apt; heart disease, diabetes, schizophrenia, asthma, multiple sclerosis, even so-called 'simple' disorders like familial cancers or Parkinson's disease, and the same will eventually be true of rare diseases like the periodic paralyses.  The cases vary substantially, just as there's variation among people without disease (yet), and this is in part because every genome is unique and everyone is exposed to different intrauterine risk factors at different developmental stages, or during childhood and adulthood.  Science is better at explaining observations that are easily and readily replicable than it is unique events.

Wethinks the geneticist doth protest too much!
The posturing and proclamations of surprise at finding complexity is false on the part of geneticists, or else they have been very unaware of basic biological knowledge that has been around since before they were born. We had every good reason to know, decades ago (and some of us wrote as much back then) that such traits were complex--and why that was so.  The reason for the 'surprise' is that it counters the disingenuousness of the proclamations that 'the' gene(s) 'for' the trait would be found by such means as GWAS. 

The new approaches taken, looking 'beyond straightforward genetic explanations for' [whatever trait] now include all the existing omics coming online -- microbiome, connectome, metabalome, nutriome, etc -- as well as epigenomics, which is the study of alterations to DNA that are not to the gene sequence and may or may not be inherited, may or may not be due to environmental factors such as toxins, and so forth (it's interesting that this is now bringing epidemiology full circle, from environmental factors to genetics and now back to environmental factors again).  Each of these will surely explain some of the susceptibility to complex disease, but just as surely won't explain it all, or in everyone.

But hungry scientists know enough to keep coining omics categories.  So now we are seeing what,  consciously or not, amounts to using what is either feigned or culpable surprise to market even more scaled-up, larger, longer-term studies, now to include more kinds of 'omics, including very costly 'environomics' (a word we think we've just coined but that is sure to arise soon enough).

Once again, the more we know about all this, the more complex it becomes, not less. Success can eventually come from any direction of course, but we think slowing down the rat-race, calming down and spending more time thinking and less time proclaiming, more musing and less marketing, smaller and more focused rather than all-inclusive approaches would raise the odds.

And, of course, some things may just be irreducibly complicated whether we like it or not.

Changing the diagnosis? Nature does it, too!

Here is a story that discusses the changing diagnosis of autism, a hot topic this week in the science news.  The doctor interviewed, Dr Bryan King, has spent the last 5 years working with a committee charged with revising the diagnosis of autism for DSM-5, the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, the diagnostic standard for psychiatric illnesses.  Reports of a dramatic increase in the prevalence of autism, along with genetic findings revealing autism's complexity (which we've posted about), are in the news.  So Dr King, involved in setting the standard diagnostic criteria for autism or autism spectrum disorder (ASD), is interviewed about the process.

Obviously, neither environmental nor genetic factors cause 'autism' per se, if the very meaning of the term changes. ASD is in some ways a cultural trait, since it's we who define it.  If we change our definition, the risks associated with specific genes or environments necessarily change as well--yet, in physical terms, they have clearly not changed at all!  If a genetic variant conferred a risk of, say, 0.5% of autism 10 years ago, then today on average it would confer nearly 1.0% (twice as much as before).

This is one problem with doing the genetics of 'autism' when the trait you're doing the genetics of is a moveable target.  The politics and other aspects of the diagnostic criteria may or may not be proper, but certainly the behavioral cutoff is cultural both in the sense of its manifestation in a given cultural setting, but also in the way that setting sets diagnostic criteria.

Relevant to MT is that Nature probably works the same way.  Here the key issue is natural selection.  Natural selection is a screen of organisms for traits that are more, or less, compatible with local circumstances.  But those circumstances change, sometimes rapidly.  Thus, like cultural definitions, the criteria that determine the relative fitness--reproductive success--are changing.  This means that here, too, the fitness of particular genetic variants is context-dependent, not fixed or absolute.

This is one of the challenging aspects of evolutionary biology, because it is tempting to view a genotype as inherently good or bad, inherently likely to succeed or not.  That makes theory and modeling of natural selection, evolution, and species formation tractable.

But Nature may not be like that.  If fitness is a shifting phenomenon, which it certainly is to at least some extent, then everything is context-dependent, and relative to circumstances, all the time.  So many of the scenarios proposed to account for what we see today may have a degree of the arbitrariness of the definition of a given trait, like autism.

Novel mutations = novel conclusions?

As reported in the NYT, the results of three new studies, published (here, here, and here) this week in Nature on the genetics of autism has found novel gene mutations that might explain risk as well as evidence that risk increases with the age of the father.  From the Sanders et al. paper:

Here we show, using whole-exome sequencing of 928 individuals, including 200 phenotypically discordant sibling pairs, that highly disruptive (nonsense and splice-site) de novo mutations in brain-expressed genes are associated with autism spectrum disorders and carry large effects. On the basis of mutation rates in unaffected individuals, we demonstrate that multiple independent de novo single nucleotide variants in the same gene among unrelated probands reliably identifies risk alleles, providing a clear path forward for gene discovery. Among a total of 279 identified de novo coding mutations, there is a single instance in probands, and none in siblings, in which two independent nonsense variants disrupt the same gene, SCN2A (sodium channel, voltage-gated, type II, α subunit), a result that is highly unlikely by chance.

It explains risk in only a very small fraction of cases.  Though, Sander et al. suggest that the model may well be useful for explaining many more. As the Times story says:

Experts said the new research gave scientists something they had not had: a clear strategy for building some understanding of the disease’s biological basis.

And,

An intensified search for rare mutations could turn up enough of these to account for 15 percent to 20 percent of all autism cases, some experts say, and allow researchers a chance to see patterns and some possible mechanisms to explain what goes awry. 

This would be great, of course.  Any clues to the bigger picture could be extremely helpful.  However, if autism is like every other complex disorder, a finding that's true at one extreme of the distribution of the phenotype will not necessarily apply to any other part of the distribution.  There are high shared fractions of the genome among relatives, usually many coding changes, too.  So it is going to be difficult to 'prove' that the change observed really is causal. Exome sequence assumes coding changes and in a way is vulnerable to identifying a coding change and assuming it's causal, when regulatory changes are not in the search space; is this the drunk looking for his keys under the lamplight?

Indeed, as O'Roak et al. conclude,

Although there is no one major genetic lesion responsible for ASD, it is still largely unknown whether there are subsets of individuals with a common or strongly related molecular aetiology and how large these subsets are likely to be.

O'Roak et al. identified novel mutations in sporadic, non-familial cases, and characterized them as to their severity and type, and they also identified pathways that the affected genes might share.  They conclude that there are likely to be from hundreds to over a thousand genes associated with autism: "Our analysis predicts extreme locus heterogeneity underlying the genetic aetiology of autism."

Of course there must be 'networks', and all of this kind of rhetoric sounds impressive but really is post-facto and in a sense superficial.  Genes interact with other genes, not just in terms of protein-protein interactions but also related to expression level (not assayable by exome sequencing).  So saying that there are hundreds of genes in networks is to some extent big-words to acknowledge that we may find this or that component, but this trait is simply not simple.

Neale et al. report, "Our results support polygenic models in which spontaneous coding mutations in any of a large number of genes increases risk by 5- to 20-fold."  Again, other functional elements in DNA that greatly outnumber the protein-coding parts, are likely to be at least as important.  Indeed, there are findings that 1% or more of autism is due to copy number variation, which may overall swamp the rare variants in importance, if the results hold up.

Each of these studies looked at a subset of the study population -- because autism is such a wide spectrum of disorders, it's important to reduce possible genetic heterogeneity by narrowing the phenotype in any study -- and found novel, or sporadic mutations to be associated with risk.  Because the idea that new mutations, which we all carry a substantial number of, might be causative can't help predict who is at risk, the hope is that if these mutations are indeed associated with risk, they might give some clues as to which developmental pathways are affected in this disorder.  The hope has been for years that genes for autism will be identified. Now that it's looking like this is a polygenic disorder, if indeed genes are a primary cause, and that sporadic mutations might be significant, it's looking more and more likely that those who have long said that major genes that can predict the disorder will not be found have been right.

The Times quotes a well-known population geneticist on this work:

“This is a great beginning, and I’m impressed with the work, but we don’t know the cause of these rare mutations, or even their levels in the general population,” said Dr. Aravinda Chakravarti of the Institute of Genetic Medicine at the Johns Hopkins University Medical School, who was not involved in the studies. “I’m not saying it’s not worth it to follow up these findings, but I am saying it’s going to be a hard slog.”

If these new results can in fact lead to understanding what goes awry in the developing brain to lead to autism, great.  Whether this will ever be clinically significant is another matter.  And one needs to remember that autism is by far mainly environmentally caused!  The report last week that its prevalence has increased by 78% in the past decade alone shows that this is about environments (that increase is unlikely to all be due to changing definitions of the disorder, or changes in diagnostic practices).  Well, a determined geneticist will argue that rapid environmental change could, in principle, have led to higher disease risk by triggering big responses in a few common genetic variants interacting with the environment.  Not so! We've had he environmental change (whatever it is), and ASD is clearly not due to one or two major genes responding to that change.

The same arguments apply to excessively exuberant claims implying simple genetic adaptation due to natural selection, and for the same reasons.

If biomedical research is about doing something about autism, rather than about forcing genetic thinking onto the problem, we're looking under the wrong lamp-post!

News of the Century: Genes Prove Einstein wrong, change faster than neutrinos!

Why bother arguing whether the speed of a neutrino is faster than the supposed upper limit to speed, that of light?  The measurements are so delicate that nanosecond discrepancies can be very difficult to interpret or, more accurately, a theory so well established (deeply entrenched?) as Einstein's theory of relativity will not be overthrown on such scant evidence.  But don't even bother, because we now have the desired falsification, and surprisingly it comes from genetics!

A new report shows (well, says) that the rate of autism is nearly 80% higher than it was a decade ago. That is rather dramatic, perhaps astonishing if you believe the data are disgnostically accurate rather than that it's due to diagnostic changes or predilections to cause behavioral variation 'autism'.  This is of course disturbing, as it shows some serious issues and that we have a major therapeutic, educational, or other kind of burden to bear to help the affected kids and their families.

But that's only part of the story.  Since our research establishment is so committed to the idea that everything, including how you vote and whether you like to be abused as a child, is genetic, and that autism must, simply must, be seriously genetic, the implications of this change in disease risk are widespread.  Indeed, if this is all true, the presumed genetic basis of autism must have evolved faster than evolution makes possible. Faster than parent-offspring transmission and natural selection!  That implies, believe it or not, non-material transmission of genetic variation and its non-material (in the 'ether' that physicists thought they'd long ago shown doesn't exist) changes in the genes and their frequency, while they are literally disembodied before being reinserted into people before they reproduce.

This disembodied evolution changes things faster than neutrinos fly under the Alps from Switzerland to Italy, and shows that even Einstein's laws are wrong!

When is a trait a trait?
It may seem obvious that an organism has various traits, and these must therefore have a genetic basis and, furthermore, must have evolved by natural selection screening on the traits (and hence molding the underlying genetic basis).  But this is not an accurate way to view the living world.

A human trait like autism illustrates the point.  If we screen more intensely, or set up new diagnostic criteria, or if there are reasons that autistic children qualify for special educational assistance, or we invent new imaging or biochemical measures of brain function that are used to specify a trait, then what is changing is the trait itself!  That's because the trait is to this extent in our minds, not inherently in the individuals.  Autism today is not autism yesterday.

In that case, the genetic basis will change and no neutrinos need be involved.  Because if we categorize or measure something differently, there is every reason to expect it has a different underlying genetic architecture.  Eyes are made by different genes than fingers, but if we re-name 'eyes' in a way that includes what used to be called fingers, we have to expect genetic associations to change.  Likewise if we rename what 'autism' is.

Even natural selection doesn't see 'traits' per se, but only is reflected in reproductive success.  We may care about, say, running speeds of rabbits and foxes, but nature 'cares' only about rabbits that escape and foxes that get dinner, and hence live to reproduce.  Running--something we define and measure--may or may not be part of the selection.  But if we choose to study the genetic basis of escape or pursuit speed, we can ask questions about its evolution.  But that's different from knowing that the us-defined trait is what in some trait-specific sense caused an us-defined adaptation, or vice versa.

An alternative explanation
It is, of course, possible that we've misinterpreted and that all that's really happened is something in the environment, including diagnostic definitions and screening intensity or other reasons people may look for, or look more intensely for, autism.  In that case, autism then isn't really the same as autism now, so the principles of materialistic genetics haven't really been violated.  That's less exciting, but more likely....and raises an important principle.

So why the insistence on intense genetic approaches?
One can ask why we keep pouring money into genetic studies like GWAS and related high-throughput but hypothesis-free 'omics' approaches.  One answer often given by geneticists is that they are, after all, geneticists and that's what they do.  Another favorite of epidemiologists is that once we identify the genes, we can treat them as fixed risk factors and correct for them (statistically) in searching for the important causes, which are environmental.  Whether you find these explanations adequate or think there may be other motivation behind them is your diagnosis.

But one defense of genetics over everything else is that when something changes faster than a speeding neutrino, it must be because there are common underlying variants with large response effects to the environmental changes.  So if we identify those genes we can....remove the environmental trigger so we don't have to identify those genes.  But a lot of experience and intense studies suggest this is just not the case, and is more wishful thinking or scientific momentum-preserving.

Normal traits like stature, obesity, and blood pressure that aren't diseases, as well as many complex diseases like diabetes, and hypertension, and asthma and numerous others have risen in frequency very rapidly during past decades.  They should have these same characteristics of common-variants responding in a big way to rapid, major environmental change, if that explanation is correct. These traits have been GWAS'ed to death, so to speak, and not only do we have the hyped-up issue of 'hidden heritability' (family risk not identified by GWAS studies), but we also clearly find that a few common variants with big effects simply do not explain these diseases.  So applying this it-must-be-major-genes argument to autism is questionable to say the least.

Genetic variation responding to environmental change is always a factor, but there is no reason whatever to think genetics will contribute much to the overall problem.  Genetics is most likely to contribute to those cases of autism--and they exist--for which a specific high-risk genetic variant has been identified.

What to do about a major societal problem, a real problem, like autism is less clear. But just plowing ahead isn't necessarily the most societally responsible approach.

Individualized attention for every kid: personalized disease?

We all want to feel unique, special, deserving of note or attention.  With our huge research establishment, almost anybody can achieve this yearning.  And not just in terms of skills and interests.  We can, apparently we will eventually, all have our own personalized disease(s).

Results from a large study of autism among Korean children are making a big splash in the health news (e.g., here).  Prevalence of this disorder was studied in a "total population sample" and the results published in the May 9 issue of the American Journal of Psychiatry.  The study is making a splash because estimates of prevalence among Korean kids, 2.64% or more than twice as high as among American or British kids, where it's around 1%.  (We are unable to access the paper online, but the Autism Science Foundation Blog cites it generously here, the source of our quotes from the paper.)

This study was conducted between 2005 and 2009 in the Ilsan district of Goyang City, South Korea, a stable, residential community near Seoul (area, 102 km2; population, 488,590) and representative of the general South Korean population (Korean Statistical Information Service, Capital Region Population, 2006). The target population (N=55,266) included all children born from 1993 through 1999 (ages 7–12 years at screening) and attending Ilsan elementary schools, as well as children in the same age group enrolled in the Ilsan Disability Registry between September 2005 and August 2006. Thirty-three of 44 elementary schools agreed to participate; 36,592 children were enrolled in participating schools and 294 in the Disability Registry...

Two thirds of the cases they found were among kids in the mainstream school population, previously "undiagnosed and untreated".  The authors suggest there could be cultural reasons for this (apparently autism can be a stigma that affects an entire family in Korean society).  They also suggest that kids on the autism spectrum may do better in Korea's highly structured school system than they do in the US, so can be fairly successful even without diagnosis and specialized support, and thus have flown under the radar.

The authors further conclude that prevalence estimates are likely to go up in the US and the UK if their methods are followed.

But, the "autism spectrum disorder" (ASD) category is a fluid one, generally having become broader over the years as more is learned about apparently similar behavioral disorders.  And, as a cynic might say, increasing the topic-pool for researchers, special ed programs, and other interests. The broadening of the definition, as well as the increased awareness of the spectrum have correlated with a rapid increase in prevalence of autism in the US over the past several decades.  Thus, it has been difficult to sort out whether the increase has been largely due to changing definitions and awareness or in fact to more cases.

Therefore, the suggestion that because prevalence is higher in Korea than in the US means that prevalence has been underestimated in the US -- and that thus many children are being denied treatment and schooling that would improve their chances of success -- assumes that prevalence is real and there to be uncovered rather than at least somewhat dependent on definitions and awareness.  We aren't arguing at all that the disorder doesn't exist, simply that if you throw more symptoms into the pot, you'll get more kids with what is called ASD. If behavior has a distribution, as it seems to--such as a 'normal' distribution--then it's a continuum and one can choose whatever cutoff points one wants to define a person's trait.

Not surprisingly, the intensive and very expensive search for genes 'for' ASD has yielded no genes with large effects.  This is because autism is a complex trait, and like all complex traits, is more likely to be polygenic than due to single genes.  The heterogeneous definition of the trait doesn't help, either. The more continuous the distribution of what's being measured, the more likely it is to be complex--that's the nature of nature.

But the suggestion in this paper that it's incumbent upon researchers to estimate the true prevalence of ASD so that affected children can get the individualized attention they deserve (which they certainly do) is troublesome.  What is 'true' depends on where you draw the line, even if measurement and interpretation were perfect.  Wouldn't every child benefit from individualized attention, whether or not they have been diagnosed with a disorder?  The brightest kids get it, and thrive.  Athletic kids get it, and musical kids, kids into mechanics or art or theatre, as well as kids with labeled disorders. But these labeled kids are the only ones with IEPs (individualized education protocols).

There are several issues here.  One, the solution isn't to label more kids but to give more attention to every kid.  Yes, that's asking for a fundamental remake of the public school system but rather than expand our definitions so that, say, shyness becomes social anxiety disorder, and thus druggable, and every child with any symptom along the autism spectrum becomes treatable, it would behoove us to recognize that all behaviors fall along a spectrum.  And every child deserves an IEP.  And two, we should not give in to the pressure to dope everyone up on maintenance meds.

The value of education?

Here is another instance of the relationship -- or not -- between education and an educated public.

Just a slim majority of Americans -- 52 percent -- think vaccines don't cause autism, a new Harris Interactive/HealthDay poll found. 

Conversely, 18 percent are convinced that vaccines, like the measles-mumps-rubella (MMR) vaccine, can cause the disorder, and another 30 percent aren't sure. 

The poll was conducted last week, following news reports that said the lead researcher of a controversial 1998 study linking autism to the MMR vaccine had used fraudulent research to come to his conclusion.

The poll also found that parents who have lingering doubts about the vaccine were less likely to say that their children were fully vaccinated (86 percent), compared to 98 percent of parents who believe in the safety of vaccines.

For many years many people did not have their children vaccinated against childhood diseases like measles, mumps, or pertussis (whooping cough) because there were allegations that the vaccines were toxic in some ways and could cause autism.  Mercury is in thimerosol, a vaccine preservative, and the MMR vaccine was said to have some other brain-damaging properties.

Figure from BBCnews.com

The absolute risk was low but the consequences when the problem arises are lifelong and can be devastating.  So children were going unvaccinated, but this itself had consequences.  Directly, the unprotected child could suffer harm from disease itself, and some children have apparently died as a result.  Indirectly, a large unvaccinated subpopulation could provide a transmission reservoir for the pathogens.

Many expensive studies were carried out to see if the evidence for the connection between vaccination and autism was real -- which meant a lot of research energy thus diverted from other problems.  Lawsuits were instigated by angry and devastated parents against the producers of the vaccine, who naturally denied any liability.

Well, as we've posted earlier and you've seen in the news, the original Lancet paper was apparently very flawed, if not indeed a fraud, with faked or at least incompetently analyzed data.  The Lancet withdrew the paper, the author was disbarred from medical practice in Britain, and so on.  It seems now that there is consensus about the false nature of the first result.  This has been widely publicized, with great sighs of relief by the established infectious disease community: now, finally, we can get with the modern program and have every child vaccinated and remove the burden of these various diseases!

But no!  This new survey shows that a substantial fraction of the population still believes that the vaccines are dangerous.  Some may feel there's an Establishment conspiracy to silence the risk to feed the greed of Big Pharma.  Some may just not accept the recent outcome because they're so fearful for their children in regard to autism.  And there is the contributing problem that the actual causes of autism remain largely unknown.

So what is the role of education in this kind of situation?  Risks are low (even if they exist) and that means it is difficult to estimate, assess, or dismiss them.  And peoples' sociopolitical views, including suspicion of government, are deeply entrenched.  Conspiracy theories, and advocacy group solidarity, conspire to reinforce tribal feelings of us vs. them.  Since everyone, even scientists (believe it or not) have their own vested interests, it is not easy to define what is 'education' and what is 'propaganda'.  The emotional side of this keeps people in their camps.  Unclear causation is partly responsible.  Low risks or slow risks are very problematic and play into the hands of all of us, in the way we view the world: climate change, evolution, environmental pollutants, and so many other similar issues reflect this.

When we give a lecture in class or write papers, we believe that by educating our students or readers, we are helping in societal progress and enlightenment.  But we're flattering ourselves, at least to some extent.  Knowledge, and action based on knowledge, are vague areas in society and the tide of action is the result of many factors, of which our teaching and writing are but a part, and sometimes perhaps even just a small part.

Art forgeries are better than science forgery

A big challenge for modern biomedical science (and for others, like climate change, for example) is that the evidence is often that effects may be real  but they are slow, slight, or statistical.  What is barely observable year by year can turn into an ice age--or an epidemic after much time or many exposures have occurred.

False ideas under such circumstances can be very costly.  That's why what now appears to have been outright fraud in associating autism with somehow-toxic childhood vaccines has been so serious.  Countless parents shunned vaccine protection because they believed the vaccines were dangerous.  Huge amounts have been spent on the diseases that would have been prevented, on morbidity and serious ill effects, and of research dollars and resources.

The problem became public and political, with parents advocacy groups, lawyers, and politicians getting into the act.  Other problems might have been addressed with the same funds.

The physician who has been accused, Andrew Wakefield, naturally denies the allegations.  Accusations fly back and forth, and those who believe (or hope) in the conspiracy theory that the attack on  him is from Big Pharma trying to protect their profits or from lawsuits, include advocacy groups that have been attacking the vaccines for some time.  Every side, as usual, sticks to its story. We're not in a position to judge, of course, but there's a deeper point.

There is no easy answer for these situations in which causation is statistical and causal effects small.  We have to work out various criteria for making inference about what causes the ill outcome, or what its mechanism or cause-effect pattern is.  We have to work out societal risk-benefit assessments.  So it is very serious indeed when we can't trust the good faith of the evidence as well as its epistemological shakiness.

No matter how much we criticize science today, and no matter how much hyperbole, dissembling, shading and stretching of truths goes on, it is assumed to stay within the lines at least of technical accuracy.  So when there is fraud as has been claimed in this case, or data cover-ups as has been repeatedly claimed for drug trials, the crime is a crime against society and is very serious.

Fortunately real fraud of the data-faking kind seems to be very rare.  When it happens, sanctions have to be strong, to make as sure as we can that it stays very rare.  In situations where the truth is easy to find, fakery has little chance of getting the faker anywhere.  But in the common situation of complex causation, cheating can pay off.  So we have to be rigorous in preventing it.  Generally, we seem to be doing well, and part of the evidence is the degree of publicity that the occasions of real dishonest science receive.

Consequences of mistaken science

The leading medical journal The Lancet has decided to withdraw a 1998 paper that linked autism to a vaccine commonly given to young children to prevent measles, mumps, and rubella. This had stirred up many interests, including of course in the drug industry that stood to be hit with many lawsuits, the medical profession for being casual about exposure to dangerous compounds, and parents devastated by serious autism in their children.

Autism had also been connected to heavy metal--ingestion, not rock concerts--years before when the latter was found at higher concentrations in the hair of autistic children than in non-autistic children. There seem to be many studies that connect autism, or autism severity, to heavy metal toxicity. And concerns included some vaccines that contain a mercury-containing preservative called Thimerosal. However the literature on the causes of autism is complex and extensive, with many attempts to show a genetic causal link as well, some of which seem convincing.

In any case, many studies, at great expense, have been done since the Lancet report, to confirm its serious conclusions, but few if any provided evidence for the link.

But society doesn't trust vested interests these days, for well-earned reasons, so parent advocacy groups sprung up to pursue the issue for some sort of redress. The government was suspected of covering up for vested interests and not coming clean with the public. The advocacy groups are not prone to give up on what they thought could explain the tragedies that have happened to their children.

A lot of cost and anguish!

But it now turns out that the original study had various unethical aspects including some outright misrepresentation of the data so The Lancet decided to withdraw the study, a very rare act of self-policing indeed. (Even if, according to the New York Times, this was done under threat of a soon-to-be-published demand by the British Medical Journal for The Lancet to retract the study.)

This may give a sigh of relief to infectious disease communities if it will lead to less suspicion of vaccines, and thus increased vaccination rates. It will be perhaps again devastating to parents of autistic children, who may feel they now have to start afresh to look for cause, and hence cure or prevention. But if the study was flawed, The Lancet did the right thing.

An important point here is credibility and honesty in science. It's related to issues like 'climategate' and to many issues of self-promotion, exaggeration, and vested interest that we've blogged about in regard to human genetics and evolutionary reconstructions. Science usually gets through these things, but at what cost? How many children, for example, have suffered serious disease because their parents were afraid to have them vaccinated? Apparently many, which in the UK included a number of hospitalizations and some deaths. Vaccination rates declined substantially in the UK after the 1998 paper appeared, allowing measles to make a come-back.

This is why we think even needless self-promotion in papers or the media, or in medical advertising is wrong and costly, if not downright dangerous. Excess claims are seized on by people desperate to understand the cause of serious disease, who yearn for explanations that can led to prevention or treatment. High integrity is hard to maintain in a society, including in science, that is founded on a belief system that says that competition is the be all and end all of personal success. Dissembling, failure to publish negative results, and the like all contribute to raise the risk of unfortunate abuses such as appear to have characterized the original Thimerosal study.

So, kids, roll up your sleeves. Here comes the needle!

GWAS on my mind

There's a classic blues song Georgia on My Mind that is ironically appropriate. The current Nature has two reports of mapping studies of autism traits (and there is reference to a third study published elsewhere). The study gained a lot of attention (or was given a lot of hype, depending on your view of these things). It isn't the first such story to get attention. These studies are treated as if they are major findings, and tend to reinforce the idea that GWAS are a powerful approach to understanding the genetics of complex disease. Despite the evidence, we can't seem to get GWAS off our mind.

Of course 'major' is a subjective judgment but after filtering the rhetoric in the media, and looking at what the actual papers and authors say, it turns out that, as before, these new findings account for less than 1% of autism cases (and such estimates are often upwardly biased for various statistical reasons). And the genes were known before as potentially relevant. And the different studies did not find the same genes.

This is par for the course, unfortunately, because autism is a sadly damaging disease both to the persons affected and those who care about them. One doesn't want to dismiss any findings that might materially help. But as even the authors of one of the studies pointed out in comments to the media (e.g., here), this is further evidence that GWAS are a fading tool, though that is not how the news media portrayed the result.




The sad fact is that in autism, as in many other diseases, there is plenty of generic evidence for genetic risk factors playing some role (because the disorder seems to cluster somewhat in families), yet the prevalence has grown rapidly only in a recent few decades. Yes, there is the lingering question of whether prevalence has actually risen, or just the probability that a child will be diagnosed with one of a broadened spectrum of disorders, but prevalence has continued to rise rapidly even in the last few years, while the definition of the spectrum has been the same, which makes it seem less likely that rates are simply reflecting increased or altered diagnosis. This shows that even if there are gene-environment interactions, the trait is preponderantly due to environmental factors--unfortunately, despite all sorts of guesses and wild guess, the factors are not yet known.

GWAS and other mapping approaches held out hope to many, and even to some extent to skeptics, that diseases like autism for which there was no good physiological understanding could reveal genes, and hence mechanisms that would lead both to understanding and eventually to treatment. But this hope really hasn't been borne out.

The current typical response (and that of the authors of these studies) is that we need larger studies of various genetic sorts. As we've said a few times in this blog, and as even some GWAS proponents believe, that is most likely to mean much more work to find much less, and unlikely to really crack the problem. Larger studies will in principle find rarer things, but that only works if the increase in study size isn't accompanied by even greater risk heterogeneity (an improved signal to noise ratio). Even complete DNA sequence can't automatically pull the rabbit out of the hat, because the more DNA examined the more variable sites one will find.

Since it is likely that much of what we're looking for is regulatory, and we don't yet know very well how to identify such sequences, we'll be awash in DNA data with no clearer biological picture. There are some subtle points afoot here. They have to do with the statistical nature of these studies. It's possible in principle that even with densely-marked GWAS no marker will have detectably strong association with the true causal site (or, worse, sites) in a chromosome region. Full DNA sequence would in principle include the site(s) directly, but the plethora of data may well obscure it. Generally, if individual signal is strong enough to be breakthrough-generating, we should have seen it by now.

So, each new study like these recent ones presses home the dual points: large association studies simply do not seem to be the way to understand these complex traits, and instead we need some clever person to show us a better way.

We need to get GWAS off our mind (though it doesn't seem likely to happen any time soon). Being negative about GWAS may seem unseemly. But it is an attempt to be constructive. These problems are so worth studying that they are worth studying right rather than just studying again and again in the same unsatisfactory way.