Stephanie Seneff: Following the Geiers dumpster-diving in the VAERS database

As I looked over the ol’ blog last night, I was shocked to realize that I haven’t blogged about the antivaccine movement and its offenses against science in nearly three weeks. That’s right! The last time I did a vaccine post was when I examined a particularly egregiously bad paper from a couple of scientists who have drunk deeply of the antivaccine Kool Aid and as a result are trying to blame the HPV vaccine Gardasil for the death of an 18-year-old woman in Australia and a 14-year-old girl in Quebec. It’s amazing but true. Rarely do I go that long without antivaccine pseudoscience attracting my attention enough to write about it. I don’t know if this was a record, but it must be close.

Don’t get me wrong, though. I didn’t consciously go out and look for antivaccine pseudoscience to blog about. (I rarely have to.) Rather, I noticed a disturbance in the antivaccine force as my Google alerts started spitting out references to a new paper that antivaccinationists seem to like. In particular, homeopath Heidi Stevenson of Gaia Health really appears to like it, as she featured a rather long post on it on Friday. It’s an article that comes from, of all places, MIT’s Computer Science and Artificial Intelligence Laboratory, with two of its authors, Stephanie Seneff and Jingjing Liu, hailing from there, while the other author, Robert M. Davidson, appears to be a private practice internist affiliated with PhyNet, Inc. in Texas. It’s a rather odd combination in that one would not expect an internist to know very much about autism, and one would expect that researchers devoted to spoken language systems and natural language processing (Jinjing Liu) and a Senior Research Scientist interested in computer conversational systems, including speech recognition, natural language parsing, discourse and dialogue modelling, language generation, and information summarization (Stephanie Seneff) would know even less.

I was not “disappointed.” In fact, this not so triumphant trio’s magnum opus, Empirical Data Confirm Autism Symptoms Related to Aluminum and Acetaminophen Exposure, published in a journal I’d never heard of before (Entropy) but one that is appropriately named if this paper is any indication of the quality of papers it publishes. In fact, it never ceases to amaze me that after all these years I can encounter antivaccine papers by people whom I’ve never heard of before. After all this time, I still get lulled into a false sense of security that I’ve learned all the players, but I keep finding out that there are always new credulous academics and non-academics willing to embarrass themselves by making utter fools of themselves parroting old antivaccine tropes as though they were the ones who discovered them and citing long-debunked crappy studies as though they were anything but long-debunked crappy studies. You’ll see what I mean in a minute.

But first, I’d like to point out that I haven’t seen a “review” article this long and comprehensively horrible in both science and analysis since I discovered Helen Ratajczak’s Theoretical aspects of autism: Causes–A review (which, not surprisingly, is cited approvingly by Seneff et al) and some of the recent output by Christopher Shaw and Lucija Tomljenovic. In fact, Seneff’s review bears more than a passing resemblance to Ratajczak’s review article in its misunderstanding of basic science and Tomljenovic and Shaw’s recent blather in its obsession with aluminum adjuvants in vaccines as The One True Cause of Autism. The main difference is that Ratajczak concentrated on misunderstanding basic science and genetics to blame vaccines on “DNA contamination in vaccines,” while Shaw and Tomljenovic concentrate on misunderstanding epidemiology, all in an attempt to blame autism on aluminum adjuvants in vaccines.

Which is what Seneff et al, do, torture epidemiology to try to blame autism on not just aluminum adjuvants but, as the title advertises, acetaminophen, and, as the title doesn’t advertise, mercury. Check out the abstract:

Autism is a condition characterized by impaired cognitive and social skills, associated with compromised immune function. The incidence is alarmingly on the rise, and environmental factors are increasingly suspected to play a role. This paper investigates word frequency patterns in the U.S. CDC Vaccine Adverse Events Reporting System (VAERS) database. Our results provide strong evidence supporting a link between autism and the aluminum in vaccines. A literature review showing toxicity of aluminum in human physiology offers further support. Mentions of autism in VAERS increased steadily at the end of the last century, during a period when mercury was being phased out, while aluminum adjuvant burden was being increased. Using standard log-likelihood ratio techniques, we identify several signs and symptoms that are significantly more prevalent in vaccine reports after 2000, including cellulitis, seizure, depression, fatigue, pain and death, which are also significantly associated with aluminum-containing vaccines. We propose that children with the autism diagnosis are especially vulnerable to toxic metals such as aluminum and mercury due to insufficient serum sulfate and glutathione. A strong correlation between autism and the MMR (Measles, Mumps, Rubella) vaccine is also observed, which may be partially explained via an increased sensitivity to acetaminophen administered to control fever.

I could tell right away just how bad this review article was going to be. In fact, now that I read it, I almost have to doff my cap in tribute. Rarely have I seen so much antivaccine pseudoscience packed into a single paper. It’s a long one, too, making it such a—shall we say?—target-rich environment that deconstructing every antivaccine fallacy contained therein would stretch this post well beyond even Oracian standards of logorrhea. I am, therefore, going to have to “cherry pick” the most egregious bits, which, I must admit, were a bit hard to choose because there are just so many of them, so much so that my selection might seem somewhat arbitrary to readers. Such is life (and blogging). Feel free to help me out in the comments and take on the copious and almost as egregious fallacies that I just didn’t have time for. I assure you, there’s something bizarre (sometimes several somethings that are bizarre) in each and every paragraph of this 20-page magnum opus of antivaccine woo. That’s saying a lot. It was truly tiring to read this paper and try to keep track of its sometimes self-contradicting “greatest hits” parade of the quackery that makes up the autism biomed movement’s “theories” of vaccine causation of autism.

A rule of thumb that I’ve developed over the years for reading a scientific paper—any scientific paper—is that you can tell right away how bad it’s going to be by how badly the authors botch the background and rationale for their work. Suffice to say, Seneff et al botch their background and rationale very badly indeed, beginning with the very first paragraph:

Autism, and, more broadly, autism spectrum disorder (ASD), is a condition characterized by impaired cognitive and social skills [1], along with a compromised immune function [2–5]. It can now no longer be denied that the incidence of ASD is alarmingly on the rise in the U.S. [6]. While it has been suggested that the observed increase in rates may be due mainly to a change in diagnosis criteria, the actual criteria have changed very little from 1943 to DSM-IV-TR [7–9]. Despite considerable research efforts devoted to trying to uncover the cause(s) of autism, thus far no definitive answer seems available from the research literature. However, the fact that ASD rates have been rapidly increasing over the last two decades strongly points to an environmental component. Indeed, autism is recently being reframed from being a strictly genetic disease to representing a complex interaction between genetics and environmental factors, suggesting that we should focus our attention more on “environmentally responsive genes” [10].

One notes that Seneff et al cite that execrable excuse for a review article by Ratajczak as support for the concept that ASD is due to vaccines. They also seem rather unclear on the concept of autism and ASDs themselves. The concept of the “autism spectrum disorder” did not exist in 1943. The autism described by Leo Kanner was not nearly as broad a classification as how ASDs are diagnosed now. Indeed, the concept of diagnostic substitution seems to elude the authors of this paper, and Seneff et al seem not even to realize that the antivaccine movement implicitly accepts that the DSM-IVR definition of ASDs is quite broad. If it did not, why would antivaccine activists be so opposed to the proposed refinement of the diagnostic criteria for autism in the DSM-V, hatching all sorts of conspiracy theories that the real reason for the DSM-V is to decrease the number of autism diagnoses? Be that as it may, actually they cite a lot of antivaccine greatest hits. Just to give you an idea, take a look at this:

The ASD community has maintained a long-standing conviction that vaccination plays a causative role in ASD [11], an idea that has been vehemently denied by the vaccine industry [12], but nonetheless is still hotly debated [13]. A study published in 2011 has confirmed a positive correlation between the proportion of children who received vaccinations in each state over the interval from 2001 to 2007 and the incidence of autism or speech and language impairment [14]. For each 1% increase in vaccination rate, 680 additional children were diagnosed with autism or speech delay.

Note how Seneff et al start out with the false “balance,” in which the idea that vaccines cause autism (for which they cite Andrew Wakefield’s 1998 Lancet case series, which was retracted because of professional misconduct and whose results are tainted with fraud) versus one of the Danish studies showing that vaccines are not correlated with autism, as though the two arguments were comparable in validity and there were a real controversy about vaccines causing autism instead of the manufactroversy that it is. It is not “hotly debated” in the scientific community whether vaccines cause autism. It is bleated to the credulous by antivaccine activists with far more hot air than scientific knowledge. Then, it gets even better. Seneff et al cite an even more incompetent antivaccine study by Gayle DeLong. It’s a study that’s so bad that I’m shocked that even a bottom-feeding journal would publish it.

What follows is a listing of common antivaccine “theories” of how vaccines cause autism. Seneff et al cite Mark Geier’s quackery, including his concept that impaired “detoxification” interferes with the excretion of mercury from the thimerosal preservative in vaccines, and his ideas about glutathione pathways. This is the same Mark Geier whose science has been so bad for so long that there is quite literally nothing he has done in the last 15-20 years that I can take seriously. He dumpster dives. He does bad epidemiology. He proposes chemical castration for autistic children because, according to him, testosterone binds to mercury and decreasing testosterone levels makes chelation therapy more effective by removing the testosterone. It’s utterly ridiculous, and Mark Geier’s quackery is such that he’s had his medical license yanked in several states.

When it comes to concepts about why Seneff et al think vaccines cause autism, this paper is all over the place. It’s the “impaired detoxification”! It’s the mercury, even though the hypothesis that mercury in the thimerosal in vaccines causes autism is a hypothesis that is about as thoroughly discredited as a hypothesis can be. Seneff et al also completely buy into the antivaccine pivot that’s occurred since the mercury hypothesis has been falsified, namely the idea that it’s really the aluminum adjuvant that causes autism, repeating many of the same scientifically bankrupt and fallacious arguments that Tomljenovic and Shaw did in their misbegotten review articles. They go on ad nauseam about aluminum being toxic to neurons (at concentrations far higher than could be produced in vivo by vaccines), about aluminum being toxic in people with renal failure (never mind that the study was about parenteral aluminum in total parenteral nutrition solutions) and cite the infamous “monkey business” paper as evidence that thimerosal-containing hepatitis B vaccine causes autism.

Of course, as anyone who’s been following the antivaccine movement knows, the MMR vaccine has never contained either aluminum or thimerosal, but that doesn’t stop Seneff et al from jumping from blaming “heavy metals” like mercury and aluminum for the “autism epidemic” to blaming the MMR vaccine. They even go so far as to repeat oft-debunked attacks on the Danish study that failed to find a link between MMR and autism. Never mind that the evidence base that shows that the MMR vaccine is not correlated with autism consists of far more than one study. There are multiple other well-designed large epidemiological studies that have failed to find a link between the MMR vaccine and autism. If the Danish study were wiped off the face of the earth or were never done, there’d still be more than enough evidence exonerating the MMR. Similarly, Seneff et al claim that vaccines are associated with a higher rate of sudden infant death syndrome. They aren’t. They even cite an absolutely abysmal Goldman and Miller paper that claimed to find that vaccines were associated with increased infant mortality. I might have to take that paper on, but if it’s anything like a Goldman and Miller paper that I did take on that argued the same thing using different methods, there’s little doubt that it doesn’t show what Goldman and Miller think it shows.

So what is the “analysis” that Seneff et al tack onto their incompetent review of the evidence relating vaccines and autism (or, more properly, failing to relate vaccines and autism)? Well, they start out by following the pioneers of dumpster diving the Vaccine Adverse Event Reporting System (VAERS). It’s a time-dishonored tactic of antivaccine cranks, because VAERS is not a reliable measure of autism incidence or prevalence. The reason is that anyone can report cases to it, and the adverse reactions reported might or might not be related to vaccination. I’ve cited the example of a skeptic reporting to VAERS that vaccines turned him into the Incredible Hulk, while another reported that vaccines turned his daughter into Wonder Woman. Then there’s the issue about how the VAERS database has been distorted by litigation, with lawyers encouraging parents to report autism as an “adverse reaction” to vaccines.

So, right from the start, the authors completely misunderstand the very purpose of VAERs, which is to serve as an “early warning” system for adverse reactions to vaccines. It is not in any way even intended to provide reliable quantitative estimates of the frequency of adverse events or to track their prevalence over time. Yet that’s exactly what Seneff et al use it for as part of their argument. First, they cherry pick symptoms that they thought to be associated with autism, such as anxiety, constipation, ear infections, eczema, pneumonia, and others. They basically compared the autism reports in VAERS with other adverse reactions and concluded that these symptoms are not only associated with autism. Then they go off the deep end with speculation:

Three associated words suggestive of a weakened immune system, “infection,” “ear infection,” and “pneumonia” support the observation from the literature that autism is associated with immune dysfunction [2]. It has also been demonstrated that children with the autism diagnosis exhibit a heightened immune response to antigen stimulation [105], which we propose is caused by their global deficiency in sulfate supply. Thus, their increased vulnerability to infection in general likely parallels an increased likelihood of an adverse reaction to vaccines, particularly vaccines like MMR where the pathogen is only weakened but not killed.

Evidence for this? None.

Perhaps my favorite part is how Seneff et al suffer from multiple muscle strains and tears trying to handwave a relationship. It is rather amusing how they argue. Basically, they plot the number of adverse events reported to VAERS that contained the word “autism” or “autistic” and noted that the number of reports increased rapidly until 1998, after which it leveled off. Their explanation?


Given that autism incidence in the VAERS database continued to rise after a small dip around 200 and given the knowledge that the total thimerosal burden had been significantly reduced by that time, we developed the hypothesis that an enhancement of aluminum adjuvant in the vaccine might have been the reason for the observed continued high rates of autism. Mercury was phased out of vaccines around 1999 [110]. However, four doses of a new aluminum-containing pneumococcal vaccine were added to the vaccine schedule within the next few years [110, 111], increasing the total aluminum burden by 20%. This could have masked any drops in autism rates consequential to the total mercury burden.

Yes, that’s right. According to Seneff et al, the reason we didn’t see a drop in autism incidence after mercury was removed from vaccines was because the amount of aluminum in vaccines was increased by the addition of more vaccines. Of course, while it is true that it was in 1999 that the CDC recommended that thimerosal be removed from vaccines, it took more than two years for this removal to happen and in fact thimerosal-free vaccines didn’t supplant thimerosal-containing vaccines until 2001, and the last lots of thimerosal-containing vaccines didn’t expire until 2002. Be that as it may, the above paragraph is what is known as “making excuses.” Autism prevalence didn’t fall 3-5 years after there were no more than trace amounts of thimerosal in childhood vaccines other than the flu vaccine, as would be predicted by the hypothesis that mercury in vaccines causes autism.

Next, the authors looked at adverse reactions reported before 2000 and then after 2000 and tried to relate symptoms to aluminum-containing vaccines. These included injection site reactions, infection, swelling, pain, cellulitis, depression, death, fatigue, and insomnia. The assumption, apparently, was that more common events after 2000 would be related to aluminum. I kid you not. In any case, the authors then plotted the total number of all these symptoms versus time in a couple of different ways and found a steady increase, with a peak around 2003, followed by a slow decline. Of course, by 2003, there was no more than trace mercury in vaccines, so how do Seneff et al explain this observation?

It’s as bad as you think:

Figure 3 shows the ratio of the total number of aluminum-related adverse reactions associated with a particular year to the total number of aluminum-containing events occurring during that year. This number rises steadily over the turn of the century, reaching above 1.0 starting in the year 2000, and peaking at 1.4 in 2003. This means that multiple adverse reactions―cellulitis and reaction site macule, for example, are occurring in association with a single adverse event. The peak at 2003 corresponds well with the peak in autism-related reports. Thus there are introduced both an increase in the number of events around 2000 as well as an increase in the potency of each event to induce a reaction. This could be explained as an increased sensitivity to aluminum in the population, possibly due to a synergistic effect of cumulative exposure to multiple toxins [113].

One notes that reference 113 is a Boyd Haley reference. Yes, that Boyd Haley, people. In any case, this is nonsense of the highest order. Indeed, I find it highly amusing that the very reason that reports of adverse reactions due to aluminum-containing vaccines peak at around 2003 is because, as they said, that was probably the peak of mercury-autism fear mongering; so it makes sense that reports like this might also peak. It certainly says absolutely nothing about whether aluminum-containing vaccines cause autism.

I could go on, but it’s late and I’m tired. There’s still a whole lot more there trying to relate MMR and Acetaminophen to autism, not to mention looking at several individual vaccines. The “quality” of the “analysis” is comparable to the one I just described. Worse, our intrepid trio of antivaccine researchers have big plans:

In future work, we plan to create and maintain a web site where users can intelligently search the VAERS database, asking questions in spoken or typed natural language, such as, “Is there an association between miscarriage and the Gardasil vaccine?” An intuitive graphical interface will also help users easily find adverse event reports relevant to their personal experiences. This system will be modeled after a similar system we have already constructed for prescription drugs [121]. We believe that the VAERS database is a rich resource, many of whose secrets are yet to be revealed.

Yes, that’s what I’m afraid of. Bad analysis leading to spurious “correlations” without biological plausibility will be the fruits of Seneff’s dumpster diving of the VAERS database with the fixed belief that vaccines cause autism. If it’s not the mercury, she’ll find that it’s the aluminum. If it’s not the aluminum, she’ll find that it’s the MMR. If it’s not the MMR, she’ll find that it’s something to do with vaccines.

That’s how antivaccine crank researchers work.