Kristjan Wager returns to discuss the Danish autism studies

Excellent.

Having had to work on a talk last night, I didn’t have any time to write anything substantive. Horrified at the thought of this blog going silent on a weekday (going silent on the weekend doesn’t concern me much, given that my traffic almost always falls by around 50% regardless of whether I post on the weekend or not, barring any unexpected links from bigger blogs), I thought about doing what PZ did yesterday and recycling one of my more memorable posts from the old blog. Unfortunately, I couldn’t come up with topic as likely to draw as much interest as PZ’s, and I wasn’t sure if anyone would want to read my old stuff anyway, even though I like PZ’s idea of slowly moving the best of the old blog over to this one. Still, I like fresh material whenever possible, even if much of the old stuff is in essence ancient history, never before read by the majority of my present readership. Consequently, I was happy to remember that my one and only guest blogger from the first incarnation of Respectful Insolence had recently written a new piece that I had been asked to post.

Yes, Kristjan Wager is back. Back in August, Kristjan, a frequent commenter on this and other blogs, guest blogged a post discussing attacks by mercury-autism activists on the Danish Institute that did the widely cited studies showing no decrease in autism incidence after the removal of thimerosal. It was well-written and well reasoned (using English as a second language yet), and, even better, provided a viewpoint bout the Danish health system that Americans might not appreciate. The post, which I published while on vacation, provoked 151 comments, some of them quite heated.

When Kristjan asked me if I was interested in another article, this time discussing common attacks by mercury moms on the actual Danish study itself, there’s no way I was going to say no, and today is the perfect opportunity for me to post it.

So, without further ado, here’s Kristjan:

Last august I wrote a guest post for Orac at his old blog, in which I promised a follow-up post. This isn’t that follow-up post. Instead this is a post dealing with yet another common criticism of the Danish study Thimerosal and the Occurrence of Autism: Negative Ecological Evidence From Danish Population-Based Data by Madsen et al. (link to study).

The study looks at the claim that thimerosal is a risk factor for the development of autism by looking at Danish population-based data, to see if the 1992 discontinuation of thimerosal-containing vaccines in Denmark led to a drop in the number of new cases of autism.

Danish studies can be fairly big, as Danish health data is centrally registered and every person born or living in Denmark is uniquely identifiable by a personal code (similar to the US social security number). This study is no exception, and it looked at every psychiatric admission since 1971, and every outpatient contact in psychiatric departments in Denmark since 1995 (see Figure 1).

What the study found, was that there “was no trend toward an increase in the incidence of autism during that period when thimerosal was used in Denmark, up through 1990.” After 1990 there was an increase, even after thimerosal was removed in 1992. This lead the authors to conclude: “Our ecological data do not support a correlation between thimerosal-containing vaccines and the incidence of autism.”

Together with a few other studies, that has reached similar conclusions, the Danish study has been widely referred to in debates about the possibility of a thimerosal-autism link. Since this is the case, it has come under heavy fire from groups like Safe Minds, which in their paper Danish Thimerosal-Autism Study in Pediatrics: Misleading and Uninformative on Autism-Mercury Link (link .pdf) raises their objections to the study, which again are often quoted by other people as data points against the study. I will only be addressing one of these. They write:

Autism counts were first based on hospitalized, inpatient records and then changed in the middle of the study period to add in outpatient records. This new outpatient registry was introduced in 1995. Therefore, their purported increases after 1994 can be explained entirely by the registration of an existing autism population that did not require hospitalization. The authors minimize this discrepancy and do not adjust for it in their chart (Figure 1), yet in a prior study using the same Danish data, outpatients exceeded the inpatients by a ratio of 13.5 times, and represented over 93% of total cases. This huge gap clearly invalidates their inpatient data, the corresponding time period from 1970-94, and any evidence for a rising trend of autism in Denmark. The authors claim that inpatient admissions were rising also, but the „data [were] not shown‰. They did not explain this omission, the only bit of credible data in their possession, since it compared equivalent populations.

First of all, let me start with stating that I agree with SafeMinds on one thing; it would have been better if the figure that only contained inpatient cases after 1994 as well. Changing data-sources like this is sloppy and only leaves the authors open for this sort of criticism. Having said that, Safe Minds does seem to ignore the fact that the study at least disproves that thimerosal is the cause of autism, since if this was the case, no matter how many other data-sources were included, there would still be a drastic drop. Also, Safe Minds ignore the fact that there should be a decrease in the number of 2-4 year olds in 1994 rather than an increase.

Now, that out of the way, I want to address something else SafeMinds either misses or ignore. Even when one ignores the data after 1990, the study still shows a lack of correlation between autism and thimerosal. How is that, you ask? Well, simply put, SafeMinds looks at the wrong end of the figure.

Let’s start with looking at the figure again and ignoring everything after 1990. Unfortunately the figure is not that big, so it’s hard to see the details, but it’s fairly clear that the number of 2-4 year olds increases from 1971 to 1990, while the 5-6 year olds first drops from in the period from 1971 to 1985, and then increases again. The 7-9 year olds are more or less constant from 1971 to 1990.

This would seem consistent with both the thimerosal-autism link and the lack of same. However, as the study makes clear (but doesn’t otherwise focus on), there was a change in the vaccination program in 1970. The changes was both in the number of vaccinations, which went from 4 to 3, and in the amount of thimerosal/ethyl mercury, which went from 400 μg/200 μg to 250 μg/125 μg. In other words, a drop of 25% of the vaccinations and 37,5% of the thimerosal and ethyl mercury. If there is any kind of link between thimerosal and autism, one would expect this change to be reflected in the numbers in figure 1. In other words, we would expect a drop in the 2-4 year age bracket in 1972-1974, in the 5-6 year age bracket in 1975-1976 and the 7-9 year age bracket in 1977-1979. As we can see on the figure this is not the case in either age bracket.

In other words, this shows a lack of any correlation between autism and thimerosal. And unlike the other end of the graph, this is a comparison between equivalent populations.