And global warming is caused by the decrease in the number of pirates or: Why an inorganic chemistry journal should not publish a vaccine epidemiology paper

In my eagerness to pivot back to an area of my interest after having had a little fun with anti-vaccine cranks, I ignored a paper to which several of my readers referred me over the last few days. Many of them had first become aware of it when everybody’s favorite smugly condescending anti-vaccine crank, Ginger Taylor, started pimping it on her blog. Before that, it apparently popped up on the only anti-vaccine site almost as loony as Age of Autism, namely SaneVax, and it wasn’t long before this paper started making the rounds of the anti-vaccine crankosphere, showing up at Gaia Health, and then just yesterday the anti-vaccine propaganda blog Age of Autism. It was at that point that I decided that I had made a mistake in not taking a look at this article; so I was more than happy to do so. Given that most of the blogs I’ve seen pimping this article include the entire abstract, I think I’ll do the same thing, so that there’s at least a hope that a skeptical word will show up on Google searches for this article or for the author’s name:

Do aluminum vaccine adjuvants contribute to the rising prevalence of autism?

Tomljenovic L, Shaw CA.

Neural Dynamics Research Group, Department of Ophthalmology and Visual Sciences, University of British Columbia, 828 W. 10th Ave, Vancouver, BC, Canada V5Z 1L8.

Abstract
Autism spectrum disorders (ASD) are serious multisystem developmental disorders and an urgent global public health concern. Dysfunctional immunity and impaired brain function are core deficits in ASD. Aluminum (Al), the most commonly used vaccine adjuvant, is a demonstrated neurotoxin and a strong immune stimulator. Hence, adjuvant Al has the potential to induce neuroimmune disorders. When assessing adjuvant toxicity in children, two key points ought to be considered: (i) children should not be viewed as “small adults” as their unique physiology makes them much more vulnerable to toxic insults; and (ii) if exposure to Al from only few vaccines can lead to cognitive impairment and autoimmunity in adults, is it unreasonable to question whether the current pediatric schedules, often containing 18 Al adjuvanted vaccines, are safe for children? By applying Hill’s criteria for establishing causality between exposure and outcome we investigated whether exposure to Al from vaccines could be contributing to the rise in ASD prevalence in the Western world. Our results show that: (i) children from countries with the highest ASD prevalence appear to have the highest exposure to Al from vaccines; (ii) the increase in exposure to Al adjuvants significantly correlates with the increase in ASD prevalence in the United States observed over the last two decades (Pearson r=0.92, p<0.0001); and (iii) a significant correlation exists between the amounts of Al administered to preschool children and the current prevalence of ASD in seven Western countries, particularly at 3-4months of age (Pearson r=0.89-0.94, p=0.0018-0.0248). The application of the Hill's criteria to these data indicates that the correlation between Al in vaccines and ASD may be causal. Because children represent a fraction of the population most at risk for complications following exposure to Al, a more rigorous evaluation of Al adjuvant safety seems warranted.


If you want to read the whole article, it’s available as a PDF file on a website called the Office of Medical and Scientific Justice (OMSJ). I thought I knew all the major quackery websites out there, but somehow I had never come across this one before. It appears to be a doozy, posting a glowing review of the anti-vaccine movie whose misinformation and pseudoscience I deconstructed three weeks ago, attacks on Brian Deer for his exposing Andrew Wakefield for the fraud he is, and, in a classic case of crank magnetism, a heapin’ helpin’ of anthropogenic global warming denialism.

Already, things aren’t looking too good.

Still, I pride myself on always going straight to the source when examining studies like this that are being bandied about the anti-vaccine underground. Who knows? Maybe I’ll find something to change my mind. True, it’s highly unlikely, but you never know. I was, however, curious just who the authors are. Christopher Shaw, I had heard of before. He was featured in the anti-vaccine propaganda movie The Greater Good and gave a talk at the anti-vaccine conference in Jamaica featuring Andrew Wakefield in January. His co-author Lucija Tomljenovic is apparently a postdoctoral fellow who was also a speaker at that very conference, giving two talks there.

So let’s get to the meat of the article, such as it is. Personally, after reading it a thought kept going through my head, namely that chemistry journals (particularly journals devoted to inorganic chemistry) probably shouldn’t be publishing medical articles. The editors and peer reviewers, so enamored with an apparently strong correlation, fell for the oldest crank gambit in the book: Confusing correlation with causation. Perhaps the most irritating part of the article is how Tomljenovic and Shaw misuse and abuse Hill’s criteria, a famous set of nine criteria postulated by Sir Austin Bradford-Hill for assessing the plausibility and likelihood of a particular correlation indicating causation. I discussed Bradford-Hill’s criteria before when Andrew Weil also misused and abused them.

Perhaps the silliest aspect of this article is Table I, in which Tomljenovic and Shaw try to convince you that the inflammatory aspects of various autoimmune diseases share aspects with inflammation provoked by aluminum adjuvants. Of course, I’d be shocked if some autoimmune diseases didn’t share some aspects of inflammation provoked by aluminum adjuvants or even vaccines in general. Inflammation is a common process that can be provoked by many things. I could tell you that the cytokine profiles that Tomljenovic and Shaw point to as being so “similar” to cytokine profiles due to aluminum adjuvants are the same sorts of cytokine profiles that result from almost any sort of injury. If, for example, as a surgeon I cut open your abdomen in order to rearrange your anatomy for therapeutic intent, I bet I could find studies with cytokine profiles that I could tenuously compare to cytokine profiles due to vaccination with aluminum-containing vaccines. For example, Tomlijenovic and Shaw point out that inflammatory bowel disease results in “Increased NLRP3 inflammasome complex signaling and NLRP3-dependent over- production of IL- 1β, IL-6, IL-18, TNF- α and reactive oxygen species (ROS) in MS, EAE, Type 1 diabetes mellitus [164-166] and animal models of IBD [167],” while the aluminum adjuvant can similarly “activate the NLRP3 inflammasome complex and NLRP3- dependent cytokines [33,34,172].” This is not in the least bit surprising, given that the NLRP3 system is activated by a wide range of “danger signals.”

In fact, perusing the chart I’m struck by how tenuous the resemblances between inflammation due to autoimmune diseases and inflammation due to aluminum adjuvants is. Presumably this is the best these two could come up with, and their best just isn’t all that convincing. None of this stops the not-so-dynamic duo from including autism and Gulf War Syndrome on their list. The latter they characterize as being “specifically recognized as ‘Autoimmune/inflammatory syndrome induced by adjuvants,” which is news to me given that I thought the emerging consensus was that Gulf War Syndrome probably doesn’t exist as a single distinct syndrome but rather as many health problems with different etiologes, much less is it recognized as some sort of autoimmune syndrome caused by vaccine adjuvants. After all, none of the anthrax vaccines soldiers received prior to going to the Gulf used squalene adjuvants. Meanwhile, autism spectrum disorders are listed in the chart as being “linked to Al-adjuvanted vaccines.” I suppose that’s true in the literal sense in that anti-vaccine activists have linked ASDs to Al-adjuvanted vaccines, but what Tomlijenovic and Shaw are doing is what lawyers like to call assuming “facts not in evidence.” Again, there is no solid evidence linking vaccines, whether Al-adjvanted or not, to autism, and several large epidemiological studies that have utterly failed to find a link between vaccines and autism. Where were the peer reviewers here?

The “evidence” in the paper consists mainly of Tomlijenovic and Shaw comparing increasing ASD prevalence to the increasing number of vaccines in vaccine schedules in various countries, their argument being that increasing doses of aluminum through vaccines correlates with increasing prevalence of ASD. Basically, they collected data on ASD diagnoses for children from ages 6-21, from 1991-2008 from the US Department of Education Annual Reports for ASD prevalence. Next, they tried to correlate the autism prevalence in this group with the cumulative aluminum dosage they received before age 6 through the pediatric vaccination schedule. They then basically did the most simplistic analysis imaginable, plotting the minimum, mean, and maximum aluminum exposures against ASD prevalence. Can you say “ecological fallacy”? Sure, I knew you could.

To recap, because I haven’t had to discuss it in a while, the ecological fallacy can occur when an epidemiological analysis is carried out on group level data rather than individual-level data. In other words, when the group is the unit of analysis, the chances of finding a false positive correlation go way, way up, as happened with a bad study trying to correlate vaccines with infant mortality. As Epiwonk once described it:

To make this jump from group-level to individual-level data is The Ecological Fallacy, which can be defined simply as thinking that relationships observed for groups necessarily hold for individuals.

The ecological fallacy was first described by the psychologist Edward Thorndike in 1938 in a paper entitled, “On the fallacy of imputing the correlations found for groups to the individuals or smaller groups composing them.” (Kind of says it all, doesn’t it.) The concept was introduced into sociology in 1950 by W.S. Robinson in 1950 in a paper entitled, “Ecological correlations and the behavior of individuals,” and the term Ecological Fallacy was coined by the sociologist H.C. Selvin in 1958. The concept of the ecological fallacy was formally introduced into epidemiology by Mervyn Susser in his 1973 text, Causal Thinking in the Health Sciences, although group-level analyses had been published in public health and epidemiology for decades.

Add to this the fact that, for all the authors’ claims that they controlled for confounding factors, by falling for the ecological fallacy they allowed huge confounders into their analysis. Even worse, they appeared to make no attempt to control for birth cohort other than to remove vaccines from their calculations that hadn’t been introduced into the schedule at the time the children were vaccinated. (How nice of them.) In any case, although the diagnostic criteria used for autism and ASDs were set in 1994 in the DSM-IV, screening in schools, increased availability of services, and decreasing stigma to a diagnosis of autism led to an explosion in autism diagnoses. The way to control for this would have been to examine much more narrowly defined birth cohorts. They didn’t. They used a single 15-year period. They also did nothing more than look for a linear correlation between aluminum dose and autism prevalence, citing r = 0.92, instead of calculating r2. The authors are incredibly impressed by this (and apparently so were the reviewers), even though it’s not so hard to produce high Pearson coefficients for a lot of seeming correlations that in fact don’t have anything to do with each other. The most heinous example I can recall is a ham-handed attempt to correlate abortion rates with breast cancer incidence.

Tomlijenovic and Shaw’s article does exactly the same thing, only slightly slicker.

Given how common papers like this are from anti-vaccinationists are, I sometimes think it would be fun to play a game I’d like to call “Name That Correlation!” What other correlations with the increase in autism diagnoses can we find over the last 20 or 30 years? Let’s see. Personal computer use has been rising since the 1980s. Perhaps that’s the cause of autism! More similar to Tomlijenovic and Shaw’s time frame, Internet use has exploded since the early 1990s. Back in 1990, few people had Internet access or e-mail addresses. (As hooked in as I am now, believe it or not, I didn’t have Internet access back then, either.) Now almost everyone does, and Internet access has become truly mobile via smartphones like the iPhone, Blackberry, and Android handsets. I bet a nice correlation between Internet usage and autism diagnoses could be constructed. Come to think of it, mobile phones, although first introduced in the 1980s, didn’t really begin to take off until the mid-1990s. In the early 1990s, mobile phones were uncommon because they were so expensive and coverage was very spotty. Now, the nearly everybody owns one. The time frame of Tomlijenovic and Shaw’s study fits the time frame of the rise of mobile phone use almost perfectly!

Finally, Tomlijenovic and Shaw misuse and abuse Bradford-Hill’s criteria. For example, they list criteria numbers one and two as being satisfied for aluminum and autism. Those are strength and consistency. The problem with these criteria is that they aren’t supposed to be evaluated by one study. They conclude their association is strong because they have a high Pearson correlation coefficient, but their study is an outlier. It’s not correct to say that the correlation is strong based on the totality of the evidence. Ditto for consistency, as, again, their study is an outlier, and, quite frankly, citing DeLong’s execrably embarrassing study as a study that found a correlation between vaccine uptake and autism does not help their case. They also try to convince readers that one of Bradford-Hill’s other criteria, such as biological rationale and coherence, have been met because of their attempt to make the tortuous vague resemblances between cytokine profiles they constructed seem like strong evidence for biological plausibility. Even worse, they try to use their confusion of correlation with causation as an argument that there is a temporal relationship between the purported cause and the effect. No, it’s not. In fact, they have not convincingly met any of Bradford-Hill’s criteria, much less eight out of nine.

None of this stops them from calling for–you guessed it!–more research:

Clearly, we cannot draw definite conclusions regarding the link between Al adjuvants and autism based on an ecological study such as the present one and hence the validity of our results remains to be confirmed. A case control study with detailed examination of vaccination records and Al body burden measurements (i.e., hair, urine, blood) in autistic and a control group of children would be one step toward this goal. Nonetheless, given that the scientific evidence appears to indicate that vaccine safety is not as firmly established as often believed, it would seem ill advised to exclude pediatric vaccinations as a possible cause of adverse long-term neurodevelopmental outcomes, including those associated with autism.

Notice the wording. “It would seem ill-advised to exclude vaccines” and “vaccine safety is not as firmly established as often believed.” It’s the old anti-vaccine tactic of sowing fear, uncertainty, and doubt about vaccines based on highly dubious science and then calling for more studies of a hypothesis that scientists, despite having looked extensively, have been unable to find any convincing evidence to support. The authors then communicate with sympathetic anti-vaccine bloggers in order to try to use their “science” to persuade government officials that more research is needed:

This study by Tomljenovic and Shaw provides convincing evidence of a strong relationship between the aluminum adjuvant frequently used in vaccinations and the increasing burden of childhood vaccinations. Even more significantly, it makes a highly plausible case for vaccines’ aluminum adjuvant as a cause of autistic spectrum disorders.

The only question is whether it should be considered a final determination. If one takes the view that only a double-blind placebo-controlled study can demonstrate the truth, then it will never be found. Conventional medicine refuses to do such studies in vaccinations. The claim is that it would be unethical to refuse the benefits of a vaccination. Of course, the consideration of harm from vaccinations is somehow not factored into the scenario. The results of this study, of course, demonstrate that the views of medical ethicists are strongly biased towards assumptions of the infallibility of their treatments.

Ah, I love the way anti-vaccine propagandists whine about basic ethical considerations in clinical trials and try to paint vaccines as somehow “privileged” in how they are treated.

Finally:

In personal communcation with Tomljenovic, I have been informed that they are doing animal experiments to further test the aluminum-adjuvant-as-an-ASD cause hypothesis. They would also like to do a human investigation involving comparison of individual medical records of vaccinations with aluminum body burdens. However, this is very costly and the funds are not yet available.

Great. It looks as though we have a new Mark and David Geier, only substituting aluminum instead of mercury. I guess we can look forward to a lot more bad science. Oh, well. I guess it will guarantee that I’ll have blogging material for years to come. Unfortunately.