So the puppies are taking up more of my time than I thought (which is why there was no post yesterday), so much so that when people started sending me a “new” (scare quotes intentional, as you will see) claiming to show that the measles-mumps-rubella vaccine is associated with autism in African-American boys, I was going to ignore it. Why? Simple, the study wasn’t new at all, but rather yet another zombie resurrection of a crappy, previously retracted antivaccine study, this time by Brian Hooker. No, it’s no better than it was when it was first published in its original form, and you’ll laugh out loud when I tell you where it was published.
Before I do, I feel obligated to provide some background for those new to the conspiracy theory that spawned this paper. Those familiar with these events can skim and skip ahead. I will briefly point out, however, that the central premise of the conspiracy theory, which is that the CDC supposedly/allegedly “covered up” data showing that vaccines cause autism. In this case, it was supposedly a finding in a 2004 study published by its scientists that found that the measles-mumps-rubella (MMR) vaccine was associated with a significantly elevated risk of autism in African-American boys (but only African-American boys). That revelation was relayed by the “CDC whistleblower” scientists named William Thompson to the parent of an autistic child (the aforementioned Brian Hooker, whose story about the relationship between vaccination and his son’s first symptoms of autism has…evolved). Hooker “reanalyzed” the data and reported it to the world. This particular conspiracy theory thus, in the minds of antivaccine cranks, confirmed what I like to call the central conspiracy theory of antivaccine movement, namely that the CDC “knows” that “vaccines cause autism” and all sorts of other health issues but is “covering it up” because, well…it’s never really clear why. (It’s a slave to big pharma or its leaders want to create more autistic children, I suppose.)
It’s yet another of a long line of antivaccine conspiracy theories, the first of which I took notice of being the “Simpsonwood” conspiracy theory cooked up by antivaccine activist Robert F. Kennedy, Jr. in which CDC scientists supposedly “covered up” data showing that the thimerosal preservative that used to be used in several childhood vaccines caused autism. (Are you beginning to see a recurring theme here?)
The “CDC Whistleblower” antivaccine conspiracy theory and Brian Hooker’s “reanalysis”
So let’s dive into the conspiracy theory. You remember Brian Hooker, don’t you? He’s the biochemical engineer turned incompetent epidemiologist and statistician who “discovered” the “CDC Whistleblower” four years ago and thus helped launch the biggest antivaccine conspiracy theory that I can recall, which led discredited former physician and scientist Andrew Wakefield to team up with a producer of bad medical TV, Del Bigtree (The Doctors), to produce an antivaccine propaganda film (VAXXED: From Cover-Up to Catastrophe) so outrageously over-the-top and obvious that Leni Reifenstahl, were she still alive, would cry, “Das ist enug!” or even “Zu viel!” (I know I use that joke a lot, but it’s just too damned appropriate.)
While it was fascinating to observe the birth of this conspiracy theory in August 2014 and watch as it grew and metastasized, unfortunately, like so many conspiracy theories, the “CDC whistleblower” conspiracy theory has proven resistant, if not impervious, to facts, science, reason, and reality, and the original version of Hooker’s paper was part of this. Now that it’s back, I feel rather as though I’m living a version of the cliché of the aging and tired Old West gunfighter reluctantly strapping on his six shooter one last time to face down a threat. Duty calls, I guess.
So what are we dealing with? The “CDC whistleblower” conspiracy theory that birthed Brian Hooker’s awful “research” began when, for whatever reason, a CDC scientist named William Thompson struck up a conversation with Brian Hooker and spoke to him multiple times on the telephone, during which time he complained about a number of things at the CDC. Most importantly, for purposes of the conspiracy theory, the complained about a specific 2004 study for which he was co-investigator and co-author with another CDC scientist, Frank DeStefano. Basically, there was an anomalous result in one subgroup suggesting a correlation between MMR and autism that was almost certainly spurious (more on that later). Thompson, however, thought that the result had been deemphasized. Unfortunately for Thompson, he was far too trusting, as Hooker recorded every call in which he vented his complaints about DeStefano and multiple other CDC scientists and employees. Hooker in turn was far too trusting as well in that he revealed the existence of the recorded phone calls to Andrew Wakefield, who wasted little time publicizing them with an inflammatory video comparing the “covered up result” in the paper to the Tuskegee syphilis experiment. (The subgroup in question was African-American boys.) Based on Thompson’s telling Hooker how to access the data set used for DeStefano et al, Hooker did his own incredibly incompetent “reanalysis” of the data and published it in Translational Neurodegeneration. So bad was it that it was ultimately retracted (here’s the journal’s retraction notice), although, for whatever reason, the full version can still be found on the Translational Neurodegeneration. (Apparently “retracted” doesn’t quite mean “retracted” to the editors of this journal.)
Why Thompson reached out to Hooker is unclear. Reading Kevin Barry’s book publishing what are represented as as transcripts of what are clearly four cherry-picked phone calls out of of thirty, I got the impression that Thompson was aggrieved with the leadership of the CDC in general and DeStefano in particular. In any event, in his conversations with Barry Thompson misinterpreted studies, including one of his own, and admitted that the only reason he was staying in his branch of the CDC was so that he could funnel information to Hooker. As I characterized it at the time, Thompson’s words were the words of a very angry man who had clearly been nursing a grudge against his former colleagues since at least 2003. As part of that “funneling” of information to Hooker, Thompson apparently told him how to get access to the dataset for the DeStefano et al paper, even though the dataset could be publicly accessed by request.
Some of the claims that developed as the retelling of the conspiracy theory went on got quite ludicrous after Thompson lawyered up and issued a statement, after which he went completely silent publicly and has remained so for over four years now. In the statement, noted that Hooker had recorded his telephone conversations without permission. He also “apologized” for leaving out “statistically significant information” in DeStefano et al, casting his “revelations” as “scientific disagreement” while insinuating a subtext of unethical behavior by his co-authors. Meanwhile, he provided documents to an antivaccine-sympathetic legislator, Rep. Bill Posey (R-FL), whom I once described as Dan Burton’s successor for the title of the biggest antivaccine loon in the US Congress.
Now here’s the hilarious thing. Science and autism advocate Matt Carey got a hold of the actual “CDC whistleblower” document dump, and he did it in the simplest way you can imagine. He just asked Rep. Posey’s office, and someone actually gave him an electronic copy of the roughly 1,000 pages that Thompson had provided Posey (which was not, as some antivaxers claimed, 100,000 pages). I, too, took a look. There were no “garbage cans full of documents” that Thompson had described as the fate of a lot of the study documents and no evidence of any data being deleted in any way that was inappropriate or not according to CDC guidelines. There were no changes in the study protocol after data collection began, another charge that, if true, could have indicated scientific misconduct. It was all basically a big nothingburger. Yet, this conspiracy theory, like Pizzagate, refuses to die and, like Pizzagate, continues to cause harm.
So now that you have this background, let’s take a look at the study.
Brian Hooker resurrects his pseudoscience in the bottomest of bottom-feeding journals
I learned of the republication of this “reanalysis” through a link sent to me by readers. It’s an article from Robert F. Kennedy, Jr.’s antivaccine group Children’s Health Defense:
After four long years, CHD Board Member, Dr. Brian Hooker‘s reanalysis of the CDC’s MMR-autism data from the original DeStefano et al. 2004 Pediatrics paper has been republished in the Winter 2018 Edition of the Journal of American Physicians and Surgeons. The data, when properly analyzed, using the CDC’s own study protocol, show a strong, statistically significant relationship between the timing of the first MMR vaccine and autism, specifically in African American males. In addition, a relationship also exists in the timing of the MMR vaccine and those individuals who were diagnosed with autism without mental retardation. These relationships call into question the conclusion of the original DeStefano et al. 2004 paper which dismissed a connection between the MMR vaccine and autism.
First, I was amused to see that Hooker is on the board of RFK Jr.’s antivaccine crank organization. Second, however, I laughed uproariously when I saw the journal in which Hooker had republished it, thinking that there is no more appropriate journal in existence in which to publish Hooker’s pseudoscience. The Journal of American Physicians and Surgeons (abbreviated JPANDS) is the house organ disguised as a journal of the Association of American Physicians and Scientists (AAPS). You might have heard of AAPS a couple of years ago when Tom Price was nominated for Secretary of Health and Human Services, because Price is a member. There, in JPANDS, sits Hooker’s “reanalyzed reanalysis.”
The AAPS is, basically, a far right wing doctors’ organization masquerading as a medical professional society, with a world view that rejects nearly all restrictions on physicians’ practice of medicine, purportedly for the good of the patient. Other prominent features of the organization are its support of private practice and dislike of government involvement in medicine, either financially or regulatory. Distinguishing it most of all, however, is its embrace of an Ayn Rand-style view of doctors as supermen and women whose unfettered judgment results in what’s best for patients and medicine. To the AAPS, doctors are special and “outside of the herd.” Unfortunately, as I described before 12 years ago, AAPS doctors are so far “outside the herd” that the organization promotes dangerous medical quackery, such as antivaccine pseudoscience blaming vaccines for autism, including a view that is extreme even among antivaccine activists, namely that the “shaken baby syndrome” is a “misdiagnosis” for vaccine injury; its HIV/AIDS denialism; its blaming immigrants for crime and disease; its promotion of the pseudoscience claiming that abortion causes breast cancer using some of the most execrable “science” ever; its rejection of evidence-based guidelines as an unacceptable affront on the godlike autonomy of physicians; or the way the AAPS rejects even the concept of a scientific consensus about anything. Let’s just put it this way. The AAPS has featured publications by antivaccine mercury militia “scientists” Mark and David Geier and, as of two years ago (the last time I checked) still promoting antivaccine pseudoscience.
As you can see, JPANDS is the perfect journal for Hooker’s republication, which is why I laughed so hard at him for having been reduced to publishing in such an ideologically fringe pseudoscientific journal. They deserve each other, particularly given what Hooker once said about his “reanalysis”:
So I reanalyzed the dataset using what is a very, very simple statistical technique. I think that in statistics simplicity is elegance. And I’m not really that smart; so I like to do simple, easy things rather than much more intellectually challenging things. So I did the simplest, most straightforward analysis, which is a Chi Squared analysis…
Here’s a hint: In statistics, the simplest analysis is often not the correct analysis, and, boy, was this the case for Hooker’s reanalysis of the DeStefano et al dataset! He didn’t control for simple confounders. He did a crappy statistical analysis. He botched the analysis in pretty much every way possible. What does one expect, though? Hooker is a biochemical engineer who thinks he’s an epidemiologist and statistician. (Yep, he’s the sole author on this paper.)
Let me summarize what was wrong with the original study and then we’ll take a look at the new study to see if Hooker’s learned anything. First, one has to go back to Destefano et al 2004, which was a case-control study in which 624 case children were identified from multiple sources and matched to 1,824 control children on age, gender, and school. The problem was that Hooker took that same dataset and analyzed it as cohort study. I discussed this in detail when the paper first came out. (Let’s just put it this way. Analyzing a dataset collected for one study model according to a different study model is highly problematic.) He also did a number of subset analyses without proper controls and used a statistical method prone to false positives. That doesn’t even count all th times Hooker referenced gross antivaccine pseudoscience in the introduction and discussion, papers such as those by Mark Geier and even Andrew Wakefield.
One thing I noticed right away looking at the JPANDS paper is that the description of the methods is more vague. The original Translational Neurodegeneration paper, for all its horrific scientific faults, at least described the populations in much more detail, including how the population and demographic data were examined and more about the statistical analysis. I did notice one thing. Hooker apparently didn’t analyze the case control data as a cohort study, which, I guess, is an improvement. Who knows? If there’s a statistician out there who can tell me how Hooker analyzed this data based on the JPANDS paper’s methods sections, I’ll be surprised. He throws the term “logistic regression analysis” about but doesn’t really make it clear what variables he used and what confounders he tried to control for.
Before I go on, let me just say that there was no other subgroup in which Hooker found a correlation between MMR vaccination and autism. As I like to say, once again, he proved Andrew Wakefield wrong, failing to find a relationship between MMR and autism for the vast majority of children. No matter how much he tortured the data, he still couldn’t find a statistically significant relationship, other than in one small subgroup.
As for the analysis, unsurprisingly Hooker finds what he describes as a statistically significant correlation between MMR vaccination and autism diagnosis, but it depends on age of first MMR vaccination. For instance, for receipt of first MMR dose at 18 months or earlier, there is no statistically significant relationship. At 24 months, there is a weakly statistically significant (p=0.03) relationship with an odds ratio of 1.82. However, there were only 23 cases, which is a small number for a case-control study like this. At 36 months, the odds ratio was 3.86, but this is based on only 7 cases. You can see the problem with this. It’s the same problem that Hooker had with his previous result and the same reason that DeStefano et al didn’t think the correlation was real; the numbers are so small that this is almost certainly a spurious result, particularly given that it wasn’t seen in any other subgroup and there is no biologically plausible mechanism why MMR would increase the risk of autism in only African-American males who received the MMR after 24 months.
As Reuben over at The Poxes Blog put it:
The nail in the coffin for the Hooker paper is that autism is usually diagnosed by the time a child is three years old. There was no increased risk at 18 months, higher but not by a whole lot at 24, and then the three-fold increase at 36 months. Gee, was it the MMR vaccine, mister? No, the effect is being modified by age. It’s as if I asked you if your shoe size was bigger at 36 months because you drank milk vs because you were 36 months. It’s age. It’s the way that autism is diagnosed. You’re going to have more children diagnosed as autistic at 36 months than you will at 18 months or at 24 months.
Hooker claims to have corrected for this, but he’s demonstrated his incompetence, and the fact that he got a similar result suggests to me that he failed.
Ironically, Hooker’s result only occurs when he expands the number of children by including children for whom there was no Georgia birth certificate. The reason is that the original DeStefano study analyzed children with birth certificates because that allowed the inclusion of confounders. Basically, in DeStefano et al, vaccination data were abstracted from immunization forms required for school entry, and records of children who were born in Georgia were linked to Georgia birth certificates for information on maternal and birth factors. The original correlation disappeared when the analysis was carried out only on children with birth certificates. Why? Because confounders could be properly considered for these children.
That’s why I found this paragraph in the Results section so telling:
Unlike the analysis by DeStefano et al, the first analysis in this study treated age of first MMR vaccination as a continues independent variable reported in years of age. Also, unlike the original DeStefano et al analysis, in this study the presence of autism diagnoses was evaluated without using covariates. Cases were matched to controls based on birth year and schools attended.
In other words, it sounds as though Hooker, faced with annoying potential corrections for confounders (covariates), simply dispensed with correcting for them in his reanalysis. His work is, as is the case for so many antivaccine studies, a zombie study that keeps returning from the dead, although this time it returned in the clown suit that is JPANDS.
The bottom line
Basically, Hooker is a biochemical engineer who’s convinced himself that vaccines cause autism and has further convinced himself that he’s competent at epidemiology and statistics. I’ve said it before on numerous occasions: If you’re going to do an epidemiological study, unless you’re an epidemiologist you need a competent epidemiologist collaborating with you as co-investigator. Even if you are an epidemiologist, you need a competent statistician to work with you from the very beginning to design the study. If you don’t, you’re likely to produce crap. Hooker did neither, and produced crap.
I’ll give Hooker minimal credit. He did apparently realize that analyzing case control data as a cohort study was so awful that even JPANDS might not have bought it. So he tried to switch his methodology. He still ended up proving why biochemical engineers should not try to do sophisticated epidemiology and statistics without skilled epidemiologists and statisticians.