an epidemiological smoking gun?

Probably not. But I’d like to see a careful look at these data from a neutral expert.

Back when the possible thimerosal/autism link was a hot topic in blogland, I kept pounding the table demanding epidemiological evidence. Since thimerosal was removed from most infant vaccines in 1999 we ought to be seeing falling rates of autism if there were a real causal link from thimerosal to autism.

Ask, and ye shall receive. A reader sent me a link to a paper that claims to find exactly that: falling rates of autism starting in 2002.

If that’s what the data really show, I withdraw my skepticism. Some caveats, however:

1. The American Journal of Physicians and Surgeons is a fairly obscure publication. A result this clear-cut on an issue this important should have been able to muscle its way into Science or Nature or NEJM or JAMA or the Annals of Internal Medicine or The Lancet or the British Medical Journal, just to name the ones I can think of offhand.

2. The authors are a father-son team. The father, Mark Geier, is an MD/Ph.D. running something called the Genetic Centers of America. He was an assistant professor of OB/GYN at Hopkins, but has no current academic affiliation. He has been a frequent expert witness in vaccine cases, and has at least once attracted serious adverse comment from the bench. The son, David Geier, is a Ph.D. student. at GWU, who runs a “medical-legal consulting” enterprise called MedCon, Inc., on the side. This is not a reassuring profile.

3. The authors note their use of what they call “a novel rapid-sampling epidemiological technique.” Methodological novelty is inversely correlated with reliability.

4. The result might be a reporting artifact. One of the primary data sources is the Vaccine Adverse Event Reporting System (VAERS). If a child who had received thimerosal-bearing vaccines were diagnosed with autism, it would be natural to report that as a possible adverse result of vaccination. If a child who had never been exposed to thimerosal in vaccination were diagnosed with autism, there would be no particular reason to suspect that vaccination had caused the autism. So the rate of autism diagnoses with suspected links to vaccination might have fallen without any drop in the rate of autism diagnoses overall. Other studies suggest that autism diagnosis rates have continued to rise. (I owe this point to Alan B. Cohen. Sc.D., of Boston University.)

So color me unconvinced.

Still, even a blind chipmunk finds an acorn every once in a while, and the data are what they are. In particular, there’s no obvious reason why the California Department of Developmental Services data should be subject to the reporting-artifact problem that renders the VAERS data suspect. So I’d be interested in hearing from any reader who can tell me whether the Geiers have actually found something here.

Update: More here, courtesy of Kevin Drum. One of the commenters makes the obvious analytic point: it’s not raw case counts we want, it’s cohort-specific diagnosis rates by age. Of children born in, say, 2002, how many had been diagnosed by age 3, compared to children born in 1999? That shouldn’t be a hard analysis to do.

Second update A reader points out that the American Association of Physicans and Surgeons, the publisher of the Journal of American Physicians and Surgeons, is a fairly flaky group &#8212 opposing, for example, over-the-counter sales of emergency contraceptives to adults because their availability might increase sexual activity among minors &#8212 and that its journal has published quasi-medical “studies” attacking abortion (using the breast-cancer link), homosexuality, and illegal immigration. So caveat lector.

Third update An expert looks at the study and finds that the statistical analysis “isn’t even third rate.”