Friday, August 12, 2005

Autism Epidemiology

While we're all waiting to see whether David Kirby will concede that the data show continued increases in the California DDS caseload of autistic young children, let's clear up some issues about the use of California DDS data for assessing the impact of autism on vaccines.

Two formidable commenters, Autism Diva and Jim Laidler, have graced our pages with comments on these issues. The Autism Diva reminds us that:

The administration of the DDS is such that the numbers are too unreliable for epidemiology, which is why they have specifically asked that their numbers not be used for epidemiological purposes.
Jim Laidler points out that:

the prevalence within each birth year cohort continues to rise well into the late teens. This is not consistent with the natural history of autism, and so indicates that the California DDS data . . . is . . . not reliable for tracking autism prevalence.
Laidler also points out that we don't need to rely on flawed California DDS data because there are better ways, implemented in peer-reviewed studies, to assess autism incidence. I highly recommend more detailed statements on these issues from Autism Diva and Jim Laidler.

So, if the California DDS aren't useful for autism epidemiology, why is Citizen Cain engaging in a discussion with David Kirby over what the California DDS data mean? Let me explain.

One school of thought holds that autism didn't exist, or that it was comparatively rare, until babies began to be exposed to thimerosal in vaccines starting in the 1930s. Incidence increased further, the argument goes, in the 1990s, because changes in vaccination schedules increased the amount of exposure to thimerosal. According to this theory, autism incidence will decline as children's exposure to thimerosal declines. Since autism is often diagnosed among children aged 3-5, and since significant reductions in thimerosal exposure began in 2000, or maybe 2001, we ought to start seeing reductions in autism incidence about now if this theory were true.

People being naturally impatient, they want to know now whether autism incidence is decreasing among young children. They don't want to wait for an epidemiological study to be published in a peer reviewed journal. So they latch onto such data as exist now, like the California DDS data, which is published every quarter. As Laidler points out, this data is unreliable for assessing past changes in autism incidence, since increased caseload almost certainly reflects changes in diagnostic criteria (even, contra Kirby, within the category of "full-blown autism"), increased awareness, and changes in the level of services available to those who receive a diagnosis of autism.

But I would argue that it might not be so unreasonable to use the California DDS data to get a first read on whether or not autism incidence is now decreasing. Awareness of autism hasn't decreased over the last five years. Diagnostic criteria, as far as I know, aren't becoming more strict. Availability of services, to the best of my knowledge, isn't decreasing. So if, say, over the next year or two, we were to see a sharp drop in the California DDS caseload of young autistic children, we would have a fact that required explanation. It wouldn't prove that autism incidence had decreased, but it would sure make me want to look more closely to see what's going on. People who don't have their minds made up about possible connections between thimerosal and autism find it very persuasive to hear that caseload out in California is starting to decrease, just as the theory predicts! And despite my best efforts, that's what people are hearing.

So rather than just say that the data isn't suitable for epidemiological purposes, I prefer to engage in the discussion and try to establish some ground rules on how to interpret the California data. Reasonable people should be able to agree that what we're looking for is a significant drop in autism caseload among young children, corresponding to the period during which thimerosal exposures decreased. If we see such a drop, we have to look further to see what's going on, because one possibility would be that the drop is the result of reduced thimerosal exposure. Reasonable people ought to agree that the best way to see whether such a decrease has occurred is to track the number of new cases among young children over time. If such data is not available, the next best way is to track the number of total cases among young children (3-5 years old). Reasonable people should also be able to agree that the tracking changes in total caseload is a lousy way to look for such a decrease. Reasonable people should not insist, once the issue has been explained to them, that a decreasing rate of increase in the total caseload indicates a decrease in autism in young children.

Autism Diva and Jim Laidler seem to think that David Kirby is not a reasonable person. Let's see. Kirby promises to sort through the data and respond next week. Let's give him a chance!

I hope that we'll be able to agree that autism caseload is continuing to increase among young children in California, and therefore that the DDS data provide no suport for the idea that autism incidence is decreasing in response to reductions in thimerosal exposure.

Moreover, a knowledgeable correspondent informs me that California DDS recently published an evaluation that illustrates the impact of new cases and drop-outs on net changes in total caseload. See page 7. This evaluation makes it clear that significant number of cases drop out for one reason or another-- the number of drop outs was more than 20 percent of the number of new cases in 2004. Moreover, the number of drop-outs is highly variable, making use of changes in total caseload a poor measure of the trend in new cases. New cases increased from 2,355 in 1999 to 3,524 in 2002. New cases decreased slightly in 2003, before increasing again in 2004, to 3,554. Unfortunately, this analysis does not break out new cases and drop-outs by age.

To summarize-- the best data from California show a recent leveling off in the total number of new cases per year, and continued growth in total cases among young children. In no way can the data be interpreted to show a decrease in new cases among young children.

One final issue to address. After having claimed that new cases seem to be decreasing, Kirby now states that maybe it's too early to start seeing a decrease, because maybe thimerosal exposure didn't really start decreasing until as late as 2003. My knowledgeable correspondent point me towards an article in the December 17, 2003 JAMA which indicates that there were significant shortages in many vaccines, including DTaP, between late 2000 and spring of 2003. As a result, stockpiles of vaccine produced prior to 2000, or 2001, were likely used up pretty quickly. While some may have still been around in 2002, certainly the exposure levels were decreasing. And since the argument is that increased exposures in the 1990s caused increased autism incidence, decreased exposures in 2001 or 2002 should be leading to decreased incidence among 3-4 year olds, if the theory were correct.