Does a new Danish study disprove the old Danish studies concluding that there is no link between vaccines and autism? Not so fast there…

Antivaccine warriors hate science because it does not support their fear and loathing of vaccines. At the same time, they want to use it to justify that very same fear and loathing of vaccines. However, as much as antivaccinationists hate scientific studies that fail to find a link between vaccines and autism or vaccine additives like the mercury-containing preservative thimerosal and autism, there’s a certain set of studies that they hate more than any other, and these are the so-called “Danish studies.” Indeed, I learned of this hatred a long time ago, when Kristjan Wager wrote a guest post about the Statens Serums Institut, which employed some of the investigators. In any case, one of these studies found no link between thimerosal-containing vaccines and autism; the other found no link between the MMR and autism.

Of course, there’s a lot more evidence out there than just these two studies that there is no association between vaccines and autism. These two studies could disappear completely from the scientific literature or have never been done at all, and there would still be plenty of evidence that there is no association between either thimerosal or vaccines and autism. Yet the antivaccine movement keeps hammering away at them as though discrediting these studies would fulfill their fantasy of of vindication in which all the inconvenient science that shows that vaccines do not cause autism would be discredited and scientists would have to take them seriously again. This fantasy got a boost a couple of years ago when one of the co-investigators of the Danish studies, Poul Thorsen, was charged with fraud and misuse of federal grant money, given that the Danish studies were a collaboration between the CDC and Danish investigators. Robert F. Kennedy, Jr. went crazy of the the story. So did the antivaccine crank blog Age of Autism and antivaccinationists everywhere. They all went on the attack, as though proving Thorsen spent CDC grant money on expensive cars and other things it wasn’t meant to be spent on would discredit the entire study. It wouldn’t. Thorsen was not the lead author or senior author on either paper. He was in the middle of the pack of authors, meaning that he contributed to the studies but was not key. Yet, at every turn antivaccinationists tried to represent him as the principal investigator and driving force between the Danish studies.

They’re at it again, coupling the same old attacks on Thorsen with a confused reading of a new Danish study about autism that has nothing to do with vaccines. The result is a font of napalm-grade flaming stupid typical of antivaccinationists. First up at the antivaccine crank blog AoA is a rehash of the same ol’ same ol’ about Thorsen by F. Edward Yazbak MD entitled The CDC and Denmark: The Love Fest That Fizzled. It’s such a completely unoriginal regurgitation of the same sorts of nonsense RFK, Jr. and AoA were spewing three years ago that I’ll just say this: Read my three posts on the topic to understand why this is a complete non-issue. Don’t get me wrong, though. If Thorsen is guilty, he should be punished to the fullest extent of the law. My point is simply that his fraud, if he committed it, does not invalidate the science that vaccines do not cause autism.

Next up is an article by someone named John Gilmore, who wrote this howler of a headline 2003 Danish Study on Mercury Fabricated? New Study Completely Different Results. Now, I don’t know who John Gilmore is. At least, I didn’t. I Googled and learned that he is the Executive Director of the Autism Action Network and that he testified against requiring health care workers to be vaccinated with the flu vaccine. He also pops up in news stories about vaccines and autism, always arguing the antivaccine side. Remember what I said yesterday about headlines that are written in the form of a question? Usually, the answer to such questions is a resounding “No!” (Yes, I know that I’ve occasionally used such titles, but after this revelation I won’t do it nearly as often anymore. Except for the title of this post, where the question is intentionally constructed such that the answer is definitely no. End of meta blog navel gazing.)

Regarding the new study by Grønborg et al, before I discuss it, I think it’s important to point out that this study was done by different methods than the original study by Madsen et al. Madsen et al was designed to examine the question of whether thimerosal in vaccines was related to autism by examining the effect of removing thimerosal from vaccines, which was mandated in Denmark around 1992. The investigators asked a simple question: Was there any correlation between increases or decreases in autism incidence and the presence or absence of thimerosal in childhood vaccines. The study design was ecological, meaning that correlations weren’t assessed using individual patient-level data but incidence levels. The observation that they made was that the incidence of autism remained fairly constant during the period of time that thimerosal was used in Denmark and that the rise in incidence of autism beginning in 1991 continued even in the group of children born after the discontinuation of thimerosal. The authors noted that thimerosal-containing vaccines were gradually phased out, which means that a gradual decline in autism incidence would be expected if there were a causal link between thimerosal and autism; instead they observed autism incidence continuing to increase. All in all, given the tendency of ecological studies to produce false positives, that this particular ecological study was resoundingly negative. Notice, also, how the authors don’t say that thimerosal protects against autism (after all, autism incidence climbed after thimerosal was removed from vaccines in Demark); they simply said that the data do not support a positive correlation. It is, so to speak, negative ecological evidence.

Madsen et al was not a perfect study. By itself, it was good but not conclusive evidence that thimerosal-containing vaccines are not associated with autism. Taken together, Madsen et al and all the other studies done since the 1990s that have failed to find a link between thimerosal and autism constitute very strong evidence that thimerosal-containing vaccines almost certainly do not cause autism. So how on earth can Gilmore claim that Grønborg et al is evidence that suggests that the authors of Madsen et al falsified data? I’m half-tempted to say because he’s an ignoramus and leave it at that, but you, my readers, expect more, as satisfying as calling an AoA denizen an ignoramus can be. One thing that is a dead giveaway that this is pure antivaccine nonsense is that Gilmore refers to Madsen et al as the “Thorsen paper.” Any time you see Madsen et al referred to as the “Thorsen paper,” even though Thorsen was neither the first author nor the senior author, it’s a very good indication that antivaccine nonsense will follow.

One reason that Denmark is so attractive for epidemiological studies is because it has a national health database that tracks its citizens from birth until death, each citizen having a unique identifier number. So it’s not surprising that Grønborg et al would take advantage of this data source to ask a different question, namely what is the risk of autism or ASD in children who have an older sibling affected by autism/ASD. They examined all children born between January 1, 1980 and December 31, 2004, a total of 1.5 million children. In this case, the objectives were to estimate the relative recurrence risk for ASDs, including recurrence in full and half siblings, and to examine the time trends in ASDs. The main outcome was the hazard ratio for ASDs among children having an older sibling with ASDs compared to children not having an older sibling with ASDs. It’s a deceptively simple question, but it requires complex statistics to get the answer. It’s also a different design. Instead of an ecological design, it’s a cohort study looking at cohorts with and without older siblings with autism/ASD. You don’t need to understand the difference to understand that different study designs can produce different answers.

So what did the authors find? This:

The overall relative recurrence risk for ASDs was 6.9 (95% CI, 6.1-7.8), and it did not change significantly over time; similar risks were observed in maternal and paternal full-siblings. The relative recurrence risks were 2.4 (95% CI, 1.4-4.1) for maternal half-siblings and 1.5 (95% CI, 0.7-3.4) for paternal half-siblings.

In other words, in these cohorts, if a child has an older sibling with autism/ASD, he has an approximately seven-fold increased risk of developing autism/ASD himself. Moreover, this relative risk was pretty constant over the study period, suggesting a genetic predisposition rather than an environmental risk factor, as the authors put it:

The difference in the recurrence risk between full- and half-siblings supports the role of genetics in ASDs, while the significant recurrence risk in maternal half-siblings may support the role of factors associated with pregnancy and the maternal intrauterine environment in ASDs.

In other words, there isn’t a hint of a whiff of anything in this study that might lead one to think that an environmental factor after birth (for example, vaccines) has anything to do with autism. Of course, looking for such factors wasn’t the purpose of this study and therefore we can’t say for sure just based on this study that environmental factors don’t have a role in autism/ASD, but this study is certainly not evidence that there is a role for environmental factors.

None of this stops Gilmore from disingenuously claiming:

Obviously the autism rate was decreasing in the 1995-2000 period when Thorsen and colleagues claimed in was increasing. In 2003, when there was a great deal of urgency to show that mercury in vaccines had nothing to do with autism, the numbers went up. But in 2013, when the topic has nothing to do with vaccines, and Poul Thorsen was not involved, the numbers went down.


In the 2003 Thorsen paper, the claim is made that there were only 956 children diagnosed with autism in Denmark between 1971 and 2000. Yet the 2013 study claims that from 1980 through 1999 there were 12,698 cases of autism spectrum disorders, with 2321 of those cases being full-syndrome autism. How a more than 1300% discrepancy could arise from researchers looking at the same data from the same database is difficult to imagine.

Let’s take the first point first. Actually, I don’t have to that (much) because Matt Carey addressed that question nearly a month ago, when Grønborg et al first published their study. In fact, he addressed the very canard that Gilmore brought up about the different numbers of diagnoses of autism spectrum disorders in the two studies. The first point that needs to be made is that Madsen et al studied autism incidence, not prevalence, and Grønborg et al studied prevalence, not incidence. Incidence is defined as the proportion of a population that is diagnosed with a condition in a given time period (usually a year). In the case of Madsen et al, what was studied was age-specific incidence; i.e., the proportions of the population diagnosed at different age ranges. Prevalence is defined as the proportion of a population that has a given condition. There are two kinds of prevalence: Point prevalence, defined prevalence at a specific point in time, and period prevalence, defined as the proportion of a population that has the condition at some time during a given period. What this means is that prevalence includes new diagnoses of a condition as well as all of the population diagnosed before. It’s not surprising that studies examining incidence and prevalence could come up with different numbers.

Another factor that could affect the numbers is followup time. Let’s look at the two time periods. First, Madsen et al looked at children diagnosed with autism between 1971 and 2000, while Grønborg et al studied children born from January 1, 1980 to December 31, 2004. Grønborg et al also estimated the overall prevalence of autism as 0.3% for childhood autism (the more severe “classic” autism) and 1.2% for all ASDs. As Carey explains:

The current Denmark study included individuals diagnosed until the end of 2010. I.e. there were 10 more years of followup. In those 10 years a lot more people were diagnosed. Where there were 956 diagnosed with autism by 2000 (for birth years 1971 to 2000), 2321 were diagnosed by 2010. That’s an increase of 240%. And the new study focused on birth years 1980 to 1999. I.e. the entire 1970′s birth cohort is not included in this count, and they still found over twice as many autistics. Where were they in 2000, when the previous study was performed? Living in Denmark, not identified as autistic.

There are a few factors which are likely behind this increase, but here we have a great example of “increased awareness” affecting autism prevalence.

And, those numbers were for childhood autism. For ASD, the increase is even larger. 10,377 Danes had an autism spectrum disorder diagnosis (for birth years 1980-1999) in the new study (the previous study included none). That’s a whopping 1080% increase. Again, there are a few reasons for this (including the increased awareness above), but here’s what “expanding the definition” does to autism.

Exactly. Over the years, the definition of autism has been expanded to include autism spectrum disorders, and more attention has been paid to diagnosing autism and ASDs, not just in Denmark, but in the US and other developed countries as well. As I’ve pointed out time and time and time again, whenever you look for a condition more carefully (i.e., screen for it), you will always find a lot more of it than you thought there was before. Always. Widening the definition of a condition, as was done for ASDs, will only enhance the effect. In any case, with increased awareness and screening, subclinical cases are identified, and diagnostic substitution results in people who would before have been diagnosed as having some other condition having the condition being screened for. It happened with ductal carcinoma in situ when mammographic screening became common. It happened for hypertension. And it happened for ASDs. Seek, and you shall find. Since the mid-1990s increasingly we have been seeking autism and ASDs. We have found. Moreover, in the later study, we have found more because followup times for the latest birth cohort in the study have been longer. Given that ASDs tend to be diagnosed at older ages than childhood autism, longer followup times will guarantee that more of the children in the later part of the study period will be appropriately diagnosed and their diagnoses included in the study.

Matt also did a really interesting thing and graphed the numbers from Grønborg et al:

As you can see, Matt graphed the adjusted autism prevalence from birth years 1980 to 2004. As you can see, the prevalence started increasing in the 1980s. Childhood autism prevalence plateaued around 1996-1997. In other words, the prevalence of childhood autism, which was mainly what was looked at by Madsen et al, was flat from 1996 to 2004, which is not consistent with thimerosal having a causative role in autism. After all, childhood autism continued to rise after 1992 (when thimerosal was removed from childhood vaccines) and then plateaued. As for ASDs, they peaked in 1994-1995 and then appear to have slowly decreased. One can’t help but note the deception involved in Gilmore’s graph, which begins at 1995 and only shows the decrease and not the increase before that. I tend to agree with Matt:

The prevalence of ASD’s do see a decline. That must be it! Evidence that thimerosal was causing autism in Denmark! But it isn’t. The prevalence of ASD in 2003-04 is the same as that in 1990-91, before thimerosal was removed. Why does the ASD prevalence go down? We can’t say for sure, but my strong suspicion is that it’s the same reason why the authors in 2003 were seeing a decrease: too few years of follow up. Autistic kids are typically diagnosed earlier than those with other ASD’s, but the average age was about 5 in Denmark in 2003 (as I recall). ASD kids can have an average age of diagnosis of 8. Recall that the recently released study followed kids up to the end of 2010. It’s no surprise to me that the estimated prevalence for ASD kids born in 2002 is lower than that for kids born in 2000 in this study. And this is consistent with the flat prevalence for kids with childhood autism diagnoses, as they are typically diagnosed earlier and 8-9 years would be enough to find the majority of the autistics in that population.

In other words, the short followup has a larger affect on ASD diagnoses than it does on childhood autism diagnoses.

Given that this study is over a month old, I’m rather surprised that the antivaccine movement took so long to respond to it. On the other hand, maybe it took the brain trust over at AoA nearly a month to come up with a way to link this study to Poul Thorsen and vaccines. Either way, you’ll be hard-pressed to find a better example of lying with statistics than what Gilmore did in his post for AoA. This new Danish study does not in any way invalidate the Danish studies from 10 years ago, nor is it evidence that there might be something to the now discredited notion that vaccines or thimerosal in vaccines can cause autism.