A new paper in Pediatrics asks:
…whether the increasing prevalence of autism, on the basis of educational data, in Wisconsin between 2002 and 2008 was uniform in all school districts or was greatest in districts with lower baseline (2002) prevalence.
In other words, was there a greater increase in the school districts where the prevalence of autism had previously been lowest? In this ‘catch up’ scenario, there was _still_ no epidemic of autism just a gradual levelling out as the school districts with lower prevalence in 2002 ‘caught up’ with the school districts with the higher prevalence in 2002.
In order to answer this question, the authors
…grouped [the districts] into 8 categories (octiles) according to their baseline prevalence, and prevalence trends were plotted according to octile.
The results were as expected – whilst the _overall_ prevalence increased from 4.9 cases per thousand to 9 cases per thousand, in each different octile, the results were not uniform. If an octile already had near 9 cases per thousand then they increased by a very small amount. If an octile was closer to the 4.9 starting point then they increased over the same amount of time (2002 – 2008) by a much larger amount until both reached the same amount – roughly 1%.
As we know from other studies, a prevalence of 1% seems to be emerging from differing areas of the world as well as different areas of the US. As John Harrington states in a companion commentary, this study shows
There seem to be no “hot spots” or high-risk areas with some ominous environmental toxin that can be postulated; it is more likely that educational services for autism were better coordinated in 1 area versus another.
Autism services are playing catch up in the school system in the US according to this paper. Thats all. Still no epidemic.
Left Brain Right Brain 
Back in 2006, I wrote a similar argument, except for California:
Educational data, as well as data from other sources like the CDDS, have big limitations. There is another paper in Pediatrics by Jim Laidler discussing the limitations on special ed (IDEA) data.
http://pediatrics.aappublications.org/cgi/content/full/116/1/e120
This current paper shows the effect of the limitations. IDEA data are not a count of all autistics in the schools. IDEA data are counts of those with an educational label of autism, who are seeking/receiving services and who have been correctly identified.
It is not surprising that some geographical areas would see increases while others less so.
Still, I wonder if this will be incorporated into a future talk by David Kirby. “Autism rates level off after removal of thimerosal”.
He was never one to let the details (which often went 180 degrees against his hypotheses) stop him.
Sure, there’s big limitations with this paper – but I think its an initial interesting foray into the subject. I’d like to see more studies done on other data sources that look at the data the same way.
Kev,
I should have been more specific–IDEA data have big limitations in measuring the actual number of autistics in the schools. As this paper clearly shows, IDEA “autism” rates are very dependent on social factors.
If we accept the concept that IDEA data show an epidemic, then we have to accept that autism risk is going up with time but mostly in areas which had low autism rates to begin with.
Cumbersome sentence, I know.
One thing that has to be acknowledged is that the autism prevalence will eventually have to level off. If nothing else, there cannot logically be an autism prevalence in excess of 100%.
Right now, the administrative autism prevalence (from IDEA, social service providers, etc.) is close to 1%. In population genetics, this is a sort of “magic number”, where a rare genetic difference begins to become part of the “diversity” of a species.
Setting aside for a moment (but just a moment) issues of functioning and variable severity, the rising autism prevalence is rapidly approaching the point where it cannot be logically considered a “disorder” and crosses over into being a “variation”, like being left-handed (~10% of the population) or being near-sighted.
Of course, most studies that have tried to compare severity of autism over time have shown that the “median severity” has decreased as prevalence has increased. I have little doubt that, if the current trend continues unabated, the entire faculty of the Mathematics department (and much of Physics and Engineering) will be labeled as “autistic”. This doesn’t mean, of course, that there aren’t autistic people who need help and many who will need life-long supervision or residential care, but it is a mistake (made all too often by people in the orbit of AoA etc.) to assume that “autistic = completely unable to care for self”.
Finally, this recent paper – which has already been touted as “showing that thimerosal causes autism” – helps refute much of the “environmental toxins” hypothesis of autism so popular at AoA. If certain areas had higher autism prevalence – relative to neighbouring areas – because of a greater exposure to “toxins”, you would expect that they would continue to have elevated autism prevalence (again, relative to other areas) even as overall autism prevalence increased.
However, if regional autism prevalence was driven by something not restricted to the local environment – such as diagnostic criteria and awareness – then you would expect to see a “leveling” as those ideas penetrated more evenly.
Of course, people not interested in what the data show – such as the writers of Age of Autism – will likely try to “spin” this in some way to support their dogma.
Prometheus
Prometheus,
I’ve said it before: people promoting vaccine-induced autism should drop the idea of the epidemic. The fact is that multiple social factors are in play which have caused much, if not all, of the rise in autism prevalence.
If they instead took the stance that vaccine-induced autism is a small fraction of the total prevalence, they could then use these social factors to their advantage. They could claim that the social factors make it very difficult to pull out a small sub population from epidemiological data.
It wouldn’t be proof, but it would be a method to keep the idea alive.
Instead, they cling to the “thimerosal caused an autism epidemic” idea, which is clearly false.
Prometheus:
I agree with this, and with the rest of your comment, I just thought I should mention that what we call autism isn’t a rare genetic difference at all, but a fairly large and heterogeneous group of rare genetic differences that all give rise to a similar (if vaguely defined) behavioral phenotype.
I haven’t been keeping 100% up-to-date with the genetic research, but I thought all the autism-associated mutations to date each only accounted for maybe 1% or a fraction of a percent of all cases of autism?
So, if all cases of autism total 1% of the population, that says to me that no single autism-related gene can have hit that magic number yet … though I do see more mixing of “autistic” genes with the larger gene pool due to changing social conditions — we’re not all institutionalized anymore, and (I think) are marrying and having children in greater numbers than in the past.
Unfortunately, there is a big problem with this study, at least judging by what’s in the paper.
I’ve drafted a post on it, haven’t posted it yet, but here’s the meat:
Imagine you had 100 six-sided dice. You roll them all. You get about equal numbers of 1s, 2s, 3s, etc.
You divide the dice into six categories, from the highest scoring to the lowest scoring. The top category contains mostly sixes. The bottom category contains mostly ones, a 6:1 ratio.
Now you roll all the dice again. To your amazement, in the top category, the scores went down and in the bottom category they went up. All the categories have converged on an average score of about 3.5! The ratio between the highest category and the lowest is pretty much 1.
Of course. You selected your categories based on the first scores, but they were random, so when you rolled the dice again, the categories stopped being meaningful – and the scores, therefore, converged…
The authors say that “the gap in prevalence between districts overall has narrowed” but they didn’t look at the actual raw variability of prevalence scores in 2008 compared to 2002, which is the crucial result, they just looked at the 2002 categories which is like the dice example. Or if they did it’s not in the paper.
@Neuroskeptic: Intitively, the problem with your analogy is that administrative prevalence is not a stationary series with random noise. Noise is minimal in prevalence series of high-population regions.
But this is how I would formally address it: Suppose you roll the dice 3 times. You divide the dice in six categories as you descrrbe, based on the second roll. Is the trend between the first and third roll going to be different for the top and bottom categories? No.
So, basically, all the researchers would need to do is show that the trend was trending the way they describe prior to the baseline prevalence measurements (i.e. prior to 2002). As I recall, this is the case in CalDDS regional centers, if you want to pursue it further.
Easier yet: They could just show that there’s an inverse relationship between prevalence in 2005 and trend between 2002 and 2008.
A reason to take the “levelling off” with a grain of salt: I put up an article months ago at examiner.com where I found that in the 1980s, there was a rise in autism diagnoses in California that slowed and leveled off when the state’s economy hit a recession. My conclusion was that less money to fund public services led to fewer diagnoses. In the current national situation, I don’t think it can be doubted that a similar phenomenon is occurring, to SOME extent.
Joseph: “administrative prevalence is not a stationary series with random noise. Noise is minimal in prevalence series of high-population regions.”
True, but most of the data points in the Pediatrics paper were quite small. They were then categorized into large sets, but the raw data were based on elementary school districts which were noisy – as can be seen from one of the author’s Figure 3 showing the raw data. Many of the school districts had a population of just 250 or 500 kids. And their prevalence in 2002 ranged from 0 (lots of zeros) up to over 2.5%.
Your point about the 3 data point trends is quite right, if you did that and found the same result it would show the authors were right, but as it stands the results could be an artefact.
You’re correct (and I’m taking your word for it, because I don’t have the full text of the paper.) It was a mistake to use the 2002 prevalence as the baseline, as opposed to prevalence in 2005 or even 2008.
I’m pretty sure it ends up being an easily-corrected technicality, though. I’d do a more thorough analysis of CalDDS data to confirm, if I weren’t busy with different passtimes these days. CalDDS RC caseloads are not very small.
I do have an analysis on prevalence vs. proportion of autistics with MR. The principle is similar.
Lindsay,
I am in complete agreement with you – autism is NOT a single disorder but is many different disorders (possibly hundreds of different disorders) with a similar (but not too similar) presentation.
In fact, if you look at a group of autistic people – especially children (because of the changes in diagnostic/assessment criteria), you often find them more different from one another than they each are from “typical” people of their age.
However, even the idea of genetic causes for autism gives certain groups apoplectic fits. After all – as they repeatedly say – “you can’t have a genetic epidemic”. They are so “stuck” on the concept of an “autism epidemic” that they can’t see the data in front of their noses.
Despite their resistance, new candidate genes for autism are discovered every year. I believe that there are a few genes that are thought to be responsible for 1 – 5% (each) of the cases of autism, but I don’t have that information right at hand.
Sullivan,
You’re right on target – if the “vaccines-cause-autism” groups would drop their insistence on an “epidemic” of autism caused by vaccines, they’d be able to evade science for much longer. After all, it would take an incredible amount of research to tease out the few percent of the autism cases that might be caused by vaccines.
Their best hope is to become the “Cause of the Gaps” – they can claim that vaccines are responsible for whatever percentage of autism that has an unknown etiology. Sure, they’d gradually get squeezed out, but they can also forestall the inevitable a bit by claiming that any genes linked to autism that have unknown functions could be “vaccine susceptibility” genes. They could even claim (as some already do today) that some of the autism-associated genes of known function are affected (in some vague, indirect or speculative way) by vaccines.
By insisting that vaccines are a major cause of autism – enough to cause an “epidemic”, they are painting themselves into a corner.
Prometheus
Prometheus,
“By insisting that vaccines are a major cause of autism – enough to cause an “epidemic”, they are painting themselves into a corner.”
They are breaking a cardinal rule of their movement: don’t make falsifiable claims.
Actually you could have a genetic epidemic. If autism were caused by genetic damage, and something came along to cause more genetic damage, you would expect more autism.
for example, increasing parental age has been proposed to do that. but it could be anything, e.g., radiation from the nuclear tests that happened in the 50s and 60s; maybe that damaged the DNA of people who were children at that time, and now it’s been passed onto their children, causing autism…
I’m sure in fact that doesn’t have anything to do with the recent rise in diagnosed autism. But it’s no more crazy than saying it’s vaccines. Actually it’s a lot less crazy.
Neuroskeptic,
I hadn’t intended to open the “genetic epidemic” can of worms, but here it is.
As you mention, increasing parental age could (and, some data suggest, does) cause an rise in autism incidence due to increasing genetic damage in sperm stem cells and ova over time (time: the greatest mutagen of all).
But apart from genetic damage, there are other ways of having a “genetic epidemic”. The simplest of these ways is almost trivial: a relatively fixed genetic susceptibility to a changing (rising, one assumes) “environmental” exposure. I put “environmental” in inverted commas because I use the term in the way population biologists do, to indicate anything that is not part of the genome or its encoded developmental programme.
Thus, we see periodic “genetic epidemics” of type I diabetes, caused by a genetic susceptibility (apparently due to certain MHC proteins) to Coxsackievirus B4. The virus exposure is seasonal (more in the fall and winter), so the “epidemic” of type I diabetes is seasonal as well.
One could also postulate an increased sensitivity to – for example – mercury because of a genetic anomaly such as mutations in the pathways needed to eliminate mercury. These people exist and are – in truth – poor excretors of mercury. The only problem – for the “mercury-causes-autism” groups, anyway – is that they don’t seem to be more prone to autism and their hair mercury levels are significantly higher than “normal” controls.
So far, the “biggest” genetic correlation with autism has been copy number variations – relatively large “chunks” of DNA that are missing (or duplicated) in autistic subjects but not their parents or (non-autistic) siblings. This speaks against the “cumulative genetic damage over generations” hypothesis but fits well with both the “aging parent” and “random event” hypotheses. As it turns out, copy number variations are very common, but there appear to be certain regions of the genome where copy number variations are much more likely to cause trouble (e.g. autism).
By the way (and not that Neuroskeptic said anything like this), the claim that “vaccines damage DNA” is not only unsupported, it still wouldn’t explain how vaccines could cause autism unless you could somehow show that the vaccine “damage” was either highly selective (to just small parts of the brain) or it causes the exact same “damage” to the DNA of every cell in the body.
Either of those hypotheses would require a precision previously unseen in DNA damage.
Prometheus