I will base this simple non-technical analysis on the California DDS (CDDS) data, readily available online. The reader is encouraged to download the data and verify my claims. The numbers are quite clear and you don't need to be a mathematician to interpret them.
Caveat: The CDDS does not recommend using this data to determine prevalence trends, but I will use it to point out flaws in interpretations that have used this data for such a purpose in the past.
Let's agree on some terminology first:
Number of autistic clients.- This is the number of persons with a diagnosis of autism in the CDDS system in any given quarter. It could be taken as a very rough indication of the autistic population in the state of California below the age of 18. [Errata: The CDDS does keep clients older than 18, so this could be taken as a reflection of the autistic population in the state].
Rate of increase.- This is obtained by subtracting the number of autistic clients in one quarter minus the number of autistic clients in the previous quarter. This is equivalent to the number of autistic clients who enter the CDDS system minus the number of autistic clients who leave the CDDS system in any given quarter. (I also sometimes refer to the annual rate of increase).
New cases.- This is the number of autistic clients who enter the CDDS system in any given quarter. Note that the CDDS does not provide data on new cases as defined here. The term "new cases" has often been used to describe "rate of increase" (defined above) and this is where much of the confusion stems from.
So has there been a drop in cases of autism based on CDDS data? Let's see. There were 28,724 autistic clients in Sep. 2005 and 29,424 autistic clients in Dec. 2005. Does it look like a drop? There were even less a year before (Dec. 2004): 26,576. That is, in the last year, the autistic population in the CDDS has increased by 2,848. This is an increase of 10.72%.
Is 10.72% a big increase in the autistic population? Absolutely. Annual population growth in the state of California is roughly 1%. Any growth in number of clients above that level is indicative of an increase in prevalence.
Don't buy that? Let's see if it works with epilepsy. There were 37,301 clients with epilepsy in Dec. 2005, and 37,076 clients with epilepsy in Dec. 2004. That's a rate of increase of 225 clients, or 0.6%. This is even somewhat lower than annual population growth. (BTW, a true increase in autism prevalence would be expected to be matched by an increase in the prevalence of epilepsy).
Can the rate of increase remain at 10.72% indefinitely? This is impossible. If it did, the autistic population would double every 7 years and would reach 20 million (the total population of the state) in about a century and a half. This rate needs to go down to about 1% eventually, and there are indications that it is already starting to go down. It will be obvious when it reaches this level, as you will hear many reports to the effect that autism prevalence has finally leveled off.
So what do people mean when they say that the number of new cases is dropping? What they mean is precisely this: The rate of increase is dropping, which is totally unremarkable as explained above. The problem is that they say "new cases" instead of "rate of increase" and this is completely misleading. But don't take my word for it. Go and check the data against this common claim: "The number of new cases of professionally diagnosed full syndrome DSM IV autism entering California's developmental services system declined from 734 new cases during the second quarter of 2005 (April through June) to 678 new cases during the just completed third quarter of 2005 (July through September), a 7 1/2% decline in one quarter" [ref].
Predictions of Thimerosal Theory
Some proponents of the Thimerosal Theory of Autism maintain that autism did not exist before the 1930s. If their theory is correct, when thimerosal is completely removed from vaccines the quarterly rate of increase should become negative immediately. And in about 18 years, the number of autistic clients in the CDDS should become zero. [Errata: Only the number of autistic minors in the system should become zero after that time. A quicker prediction is that the number of autistic clients in the 3-5 age range should become zero within 4 years of thimerosal removal. It is clear this is not happening.].
Even those who are not so forceful about this theory do maintain that the "autism epidemic" is largely attributable to Thimerosal. Still, the quarterly rate of increase should become negative, and the number of autistic clients in the CDDS should drop to early 90s or 80s levels in about 18 years.
Predictions of the Broader Criteria Theory
This theory says that our conception of what autism is has been broadening and that parent and pediatrician awareness about autism is increasing. Awareness and criteria will have to stop broadening at some point, since clearly not everyone is autistic. So according to this theory, the quarterly rate of increase should gradually drop until it matches population growth in the state of California. And the number of autistic clients should remain at high levels but only grow slowly.
Based on the trend I can see on graphs now, I would predict the autistic population in California will level off in about 2016.
Evidence in favor of the Broader Criteria Theory
In the movie Rain Man (1988), Raymond Bobbit is referred to as "high functioning" a couple times. Today, it's common for Rain Man to be referred to as a "relatively severe case". It is clear that our conception of what autism is has changed.
If a movie doesn't do it for you, consider twin studies. In twin studies, it's obvious that higher concordance will be found the broader the definition of autism is. (That is, the more identical you expect twins to be, the less concordance you will report). In the first twin study on autism by Folstein-Rutter (1977), concordance for autism was found to be 36% for identical twins (0% for fraternal). But they also found that concordance for "cognitive disorder" was 82% in identical twins (10% for fraternal). Now consider that at least 3 modern studies have found a "classic autism" concordance of 60%-95.7% for identical twins. This is fairly good evidence that researchers' conceptions of what autism is have broadened - never mind parent and pediatrician awareness.
Isn't it also remarkable that as the dosage of thimerosal has increased, autism has become more and more heritable?
Further evidence can be found in the California data itself. As I noted, the rate of increase in epilepsy is 0.6% whereas the current rate of increase in autism is 10.7%. It is known that a certain portion of autistics suffer from epilepsy. (Epilepsy is also a favorite of the Thimerosal camp when they argue that autism is pathological). So if there is a real increase in the prevalence of autism, you would expect to see a matching increase in the prevalence of epilepsy.
The CDDS does provide data on autistics with epilepsy. There were 1,886 in Dec. 2004 and 1,979 in Dec. 2005. This is an increase of 93 clients, or 4.9% in the last year. This is a lot lower than the 10.7% rate of increase for autism. (Why isn't it around 1%? Probably because more and more clients with epilepsy are recognized as also being autistic).
The number of autistic clients with no mental retardation was 16,448 in Dec. 2004, and 18,708 in Dec. 2005. So the annual rate of increase of autistics with no MR is 2260 or 13.74%. This is considerably higher than the 10.7% annual rate of increase for all autistic clients.
So there you have it. More and more autistics are not mentally retarded (presumably based on IQ testing - which incidentally is likely flawed). And less and less autistics suffer from a co-morbidity of epilepsy. Need more evidence?
The regional data is also interesting. We would expect to find that in places of high prevalence, awareness and broadening criteria is leveling off more quickly than in the state as a whole. The Westside regional center appears to be a place of high prevalence (over 30% of all clients are autistic). You will find that the annual rate of increase is about 8.3% there, lower than the 10.7% for the state. In contrast, some places (not all) where prevalence hasn't caught up should have a higher rate of increase. You will find that the Inland regional center, for example, has an annual rate of increase of 15.17%, much higher than the 10.7%. It's as if Inland is catching up to Westside - like going back in time. (Professional's conceptions of autism will probably always differ a little from region to region, however).
In conclusion, it's not correct to say that we just don't know whether thimerosal causes autism. The truth is that the weight of the evidence is overwhelmingly on the side of the broadening criteria theory.
- The "new cases" terminology is misleading, as it assumes that the rate of increase in the number of clients is the same as the number of clients who enter the system.
- Current annual increase in the number of autistic clients in the CDDS is 10.7%, which is much higher than population growth in the state of California.
- The Broadening Criteria theory predicts that the rate of increase should drop gradually to around 1%.
- The Thimerosal theory predicts that the rate of increase should become negative, and the number of autistic clients drop every quarter until it becomes zero (or at least very low).
- The data is overwhelmingly in favor of the Broadening Criteria theory.