Friday, July 15, 2005

Kirby Flubs the Autism Data

David Kirby, writer for the New York Times and author of Evidence of Harm: Mercury in Vaccines and the Autism Epidemic, triumphantly blogs in the Huffington Post that new data from California supports the notion that thimerosal in infant vaccines has been responsible for an increase in the incidence of autism, and that the removal of most of the mercury in childhood vaccines in 1999 and 2000 is correspondingly beginning to cause a decrease in autism incidence. He’s wrong.

First, his argument. Kirby believes that real world rates of autism have increased dramatically in recent decades, and that there is strong evidence implicating exposure to thimerosal as the cause of this increase. Thimerosal is a preservative that contains ethylmercury. The more standard scientific theories hold that actual incidence of autism has not changed appreciably over time, but rather that there has been increased detection of autism as a result of a variety of factors, including loosened diagnostic criteria, increased awareness, and increased availability of social services for autistic children. Reports by the Institutes of Medicine and the American Academy of Pediatrics have rejected the link between autism and thimerosal. But the theory lives on, most notably in Robert Kennedy’s recent Rolling Stone article, which has been persuasively critiqued here and here and here.

One way to resolve the issue, will be to evaluate whether autism rates go down among the cohort of children born after 1999 who are exposed to greatly reduced levels of thimerosal in vaccines. An appreciable decrease in these rates would provide strong evidence for the theory that mercury in vaccines has contributed to autism. Conversely, no change or an increase in autism incidence among these children would support the idea that mercury in vaccines has nothing or little to do with autism incidence.

Now Kirby claims that new data indicates that "fewer children with full-blown autism entered the system" of California social services provided to autistic persons in 2003 than in 2002, and fewer in 2004 than in 2003. This trend continues, he says, through the middle of 2005. Moreover, this trend corresponds perfectly with what you would expect if mercury in vaccines were an important cause of autism, since autism is most often diagnosed among children aged 3 to 5.

Not so fast. The data that Kirby relies on is not data on new cases. Rather, the California Department of Developmental Services reports on the total number of people with autism registered to receive services. Kirby derives an estimate of children who enter the system in 2004, for instance, by subtracting the total number of cases within the system at end of 2003 from the total number of cases in the system at the end of 2004. Kirby calls the change in total cases "new autism diagnoses," which it is not. The Autism Diva discusses some of the problems with using the California caseload data as a substitute for actual data on incidence.

But let’s suppose, for the sake of argument, that we’re going to set aside these concerns, and use the California data to evaluate changes in autism. Has Kirby made the best use of the available California data? Not at all. California reports not only total caseload, but also caseload by age group, including 3-5 year olds and 6-9 year olds. Surely, using this data is better than using the change in total caseload if our objective is to compare autism incidence among children born in 1999 and earlier, who were exposed to high levels of mercury in vaccines, with autism incidence among children born in 2000 and later, who were exposed to much lower levels. So here are the case loads, by quarter, starting in the third quarter of 2002, for 3-5 year olds and 6-9 year olds. At the beginning of this period all children in the 3-5 year old category would have been in 1999 or earlier. By the end, nearly all would have been born after 1999.

Source: Compiled by Citizen Cain from
Note: One incorrect figure in the table was corrected on 7/19/05.

As you can see, caseload in the 3-5 year old group increased during every quarter, at a fairly constant rate, even as exposure to mercury in vaccines was decreasing. Completely the opposite of the picture portrayed by Kirby! Moreover, caseload over this period increased by 38 percent among 3-5 year olds, but by only 34 percent among 6-9 year olds, although even by mid-2005 nearly all of the children in this category were born prior to 1999. If the thimerosal-autism theory were correct, caseloads should have been increasing faster in the 6-9 year old category, in which there has been essentially no change in thimerosal exposure, than among the 3-5 year old category, in which thimerosal exposure has plummeted.

So, if we’re going to trust the California data, it’s pretty clearly telling us that removing thimerosal from infant vaccines isn’t an effective way to reduce autism. Not surprising, because autism rates haven’t decreased in other countries where thimerosal has been removed from vaccines.

I hope Kirby will correct the record.

One final point: even though the evidence of a thimerosal-autism connection is very weak, REMOVING THIMEROSAL FROM VACCINES WAS THE RIGHT THING TO DO. Mercury is a dangerous neurotoxin, and reducing preventable exposures was a prudent public health measure.