MORE QUESTIONS THAN ANSWERS: Further Comments on the 2004 CDC Study (DeStefano et al 2004)
By Martin Hewitt
Dr Brian Hooker claims that the CDC suppressed findings of a higher risk of autism in Afro-American boys receiving MMR at 24 and 36 months, a claim based on the concerns of Dr Bill Thompson, co-author breaking ranks with the authors of the 2004 CDC paper. In response to the controversy, and Hooker’s reworking of the CDC data(Hooker 2014), the CDC issued a statement (here) claiming it had only published data on race from a subsample of children with birth certificates because this provided richer data than the larger sample used. Hooker has countered saying the statement raises more questions than answers. The questions and answers the CDC raises and the claim of suppressed evidence can be seen more clearly if we revisit the 2004 paper (DeStefano 2004) and focus specifically on its research design, which will reveal a gaping omission where questions should have been asked but weren't. The question is what is the risk of autism faced by black children in the US who receive the MMR vaccine. We will see that it is inconceivable that the CDC researchers didn't use its total sample to ask this question. It would also be illogical if the relevant data required to answer this question weren't part of the CDC's design.
Normally one looks to research to identify answers. But more important in this case are the questions not asked and the CDC’s attempt to answer Hooker’s charge. But first we need to describe as simply as possible the design of the study, which we will do in three steps. Inevitably technical details and nuances will be overlooked to concentrate on the main issue of the missing question.
First, the design compared two groups: 1) case children with autism who received the MMR and 2) control children without autism who also received MMR. The two groups, all from Metropolitan Atlanta Georgia, were matched on demographic criteria such as age, gender and race. The study took the case group from autistic children who fitted its clinical criteria and the control group from children attending 'regular' (in the UK 'mainstream') schools. The objective of the research was to see if there was a higher likelihood or probability of autism occurring in the first group compared with the second at ages 18, 24 and 36 months. If the MMR had contributed to the case children becoming autistic, we would expect to find a higher proportion or percentage of children receiving MMR in the case group at 18, 24 or 36 months. This would show a higher relative risk for the first group. If there was no difference in the risk for the two groups, and MMR was not associated with the autism of the first group, the ratio of the first to the second group would be 1. A higher ratio than 1, if statistically significant, would indicate that the first group has a higher relative risk compared with the second of being affected by the MMR.
The two groups formed the total sample. Second, from the total sample the researchers produced a subsample of children for whom Georgia state birth certificates were available. This subsample was also broken into case and controls. This is because the certificates provided details on children and their mothers not available in the total sample and included data on race, maternal age, maternal education and birth weight. The subsample constitutes 56% of the total sample, ie just over half the total sample of children were drawn into the subsample because of birth certificate details available. The more detailed texture of the subsample has to be set against its reduced statistical power due to size. It is the way the CDC researchers use the subsample that lies at the core of Hooker’s concerns – an issue we will return to. So we now have a total and a subsample each divided into case and control groups, a total of four groups, with more detail in the subsample of case and control groups.
Third, using data available in the total sample and subsample, the study asks a series of questions to establish the relative risks of autism occurring in the case group and control groups in the two samples (I’ll refer to two samples as shorthand for what is one total sample with its subsample of birth certificated children). These questions then are applied to the four groups and relative risks compared between case and control groups in the total sample and in the subsample. At this step we begin to see there are more questions than answers. For listing the questions points to the fact that the questions are not applied evenly across the two samples. Let’s first list the questions (1 to 3) that are asked even-handedly of case and control groups in the two samples, before identifying the unasked question (4). Data elicited by each question are summarised in the paper in a series of tables.
- How do the characteristics of cases and control subjects in the total sample and the birth certificate sample compare (see table 2), comparing for factors such as age, gender and race in the two samples, and for maternal age, maternal education, birth weight, etc in the subsample?
- What is the association between age at first MMR vaccination and autism case status for the total sample and the birth certificate sample and according to gender and age (table 3)?
- What are the associations between age at first MMR vaccination and autism case status within selected clinical subgroups of cases for the total sample and the birth certificate sample (table 4)?
So we have three questions applied even-handedly to case and control groups in each of the two samples. These questions produced broad findings that there were no differences in risk between the two groups on each of the factors studied apart from a few differences that were not statistically significant or the researchers explained by reference to external factors unrelated to the interests of the study.* However the fourth question, also addressing race, is addressed only to the subsample:
- What are the associations between age at first MMR vaccination and autism case status according to race, birth weight, and maternal characteristics in the birth certificate sample (table 5)?
The question addresses race only in the case and control groups in the subsample not in the total sample. Yet the study collected data on the children’s race for the total sample as well as the subsample (see table 2). It is inconceivable that the research design didn’t involve collecting findings on the way race affected the two groups in the total sample. The design would have included calculating the incidence of Afro-American children with autism following MMR by 18, 24 and 36 months in the larger total sample. But this was omitted from the study. It would have been illogical not to have included this question in the research design.
So the question missing from the paper but covered by the design is What is the association between age at first MMR vaccination and autism case status according to race in the total sample? (Data on birth weight and maternal characteristics, etc. were not available in the total sample.)
We can locate the omission within the structure of the CDC’s research design in the following table listing the key questions asked and the central question about race for which findings were not given.
Structure of questions in the research design
|
|
|
Total sample |
Subsample |
||
|
|
|
case |
control |
case |
control |
|
CDC questions comparing case and control groups on: |
|
||||
|
age, gender and race |
a |
a |
a |
a |
|
|
age at first MMR vaccination and autism case status |
a |
a |
a |
a |
|
|
age at first MMR vaccination and autism case status within selected clinical subgroups |
a |
a |
a |
a |
|
|
age at first MMR vaccination and autism case status according to race, birth weight, and maternal characteristics |
n/a |
n/a |
a |
a |
|
|
Missing question: age at first MMR vaccination and autism case status according to race |
? |
? |
|
|
|
Whereas the CDC study found that there was no greater risk of MMR being associated with age when it allowed specifically for race in the case group compared to the control group in the subsample, it omitted to provide its findings for the two groups in the total sample. This task had to wait ten years until Hooker analysed the data and found that there was indeed a significant risk among black boys associated with MMR at ages 24 and 36 months. The missing question goes to the heart of the study in revealing the question for which an answer was available but which was not disclosed in the 2004 paper. This missing question also goes to the heart of Bill Thompson’s profound unease at the way he and his co-authors chose some findings for publication, but omitted others not made public – and the consequences for the public denied information when deciding whether or not to vaccinate with MMR. One can't but underscore the irony of this move when reminded of the CDC's strapline, saving lives, protecting people.
The CDC’s response to the allegation that it censored findings on race, MMR and autism is that:
Access to the information on the birth certificates allowed researchers to assess more complete information on race as well as other important characteristics, including possible risk factors for autism such as the child’s birth weight, mother’s age, and education. This information was not available for the children without birth certificates; hence CDC study did not present data by race on black, white, or other race children from the whole study sample. It presented the results on black and white/other race children from the group with birth certificates
In other words it plays up the advantages in publishing the greater detail available from the subsample of birth certificated children, but evades the question why it didn’t publish findings on race from the total sample. The ‘more complete information’ the CDC refers to was not in fact on race per se but on other factors such as child’s weight, mother’s age, etc.. The CDC remains silent on the key question why it did not publish the findings on race from the total sample. In the light of Hooker’s re-analysis we know that Afro-American boys had a higher probability of becoming autistic following MMR at ages 24 and 36 months.
Yet again one is left with the sense that the CDC, whether deliberately or not, manipulates research samples to get the findings it wants on vaccine safety. When confronted with this charge, the CDC plumbs for more questions than answers. But science should be exact, not vague or convenient.
References
DeStefano, F, Karapurkar Bhasin, T, Thompson, W, Yeargin-Allsopp, M and Boyle, C (2004) 'Age at first measles-mumps-rubella vaccination in children with autism and school-matched control subjects', Pediatrics 113:2.
Hooker, B (2014) 'Measles-mumps-rubella vaccination timing and autism among young African American boys: a reanalysis of CDC data', Translational Neurodegeneration, 3:16.
* For example, DeStefano e al found that "the difference between case and control children
in the proportion vaccinated before 36 months of age was strongest in the 3- to 5-year-old age group" (p. 264), but explained this in terms of the greater likelihood that case children might have been vaccinated as a requirement for enrolment in early intervention or pre-school special education programmes, rather than as evidence that black boys faced a higher risk of autism following MMR at 36 months. Hooker questions the CDC's explanation suggesting it is 'highly unlikely' this type of behaviour would be seen exclusively in African American families (2014, p.4).
Thanks Martin great article cant be long now until the murdering and maiming of innocent healthy kids stops.
MMR RIP
Posted by: ANGUS FILES | September 03, 2014 at 05:21 PM
If the study matched other vaccine uptake then most of the following is not likely relevant, but...
One question I have: if a child didn't get the MMR until age 3 or older, what is the likelihood he or she was vaccinated according to schedule otherwise?
Not that I don't think MMR has a risk, and probably an amplified age (or proximity to other vaccines or polluted environment) related risk...
Then things are complicated by the reason for the delay. If a parent already had vaccine concerns or health concerns, i.e. their child may have already been injured, then they may have held off on the MMR, and the later MMR timing group could have more individuals with higher susceptibility to MMR injury.
But first time MMR after age 3 may also reflect a general mistrust (or education?) regarding vaccination, mainstream medicine, or general apathy or lack of access that delayed MMR and possibly other vaccinations, so any timing risks found could implicate more than MMR.
And since both groups got at least the MMR, the overall risk of MMR here still goes unstudied.
The CDC seems only able to give evidence of their reluctance to find and inform of problems with their recommendations.
Posted by: Jeannette Bishop | September 02, 2014 at 04:58 PM
Can we sue newspapers magazines who lie about the risk of vaccines? I remember consumer report in March 2001 spoke of avoiding thimerasol and are now silent. What threats do the cdc make? That they will be responsible if a child gets harmed by disease? A theoretical harm? Vs the actual 1/100 harm that a child will be harmed by vaccine? I wrote this to the u Chicago journal that
Took down the truth and replaced it with a statement that the truth is too dangerous too speak just like the truth about Vioxx was ... Or other drugs that affect older wealthy people --people who count--more than infants.
I emailed the editor of the u c paper this note:
Dear editor who withdrew dr hookers paper:
You are making the judgment that the fact boys like my son suffer from Vaccine injury is a fair price to pay for not examining vaccines or making them safer.
You are violating your HIPPA oath by letting yourself be pressured. I used to have confidence in vaccines but now realize that no one cares if one in 100 get seriously hurt. They think to admit this will discredit them. Well the coverup is worse than the mistake. And has cost years of needless preventable suffering. You have the power if you have the courage to turn this around. Asking for vaccine safety does not mean you or we are anti vaccine. For children's sales stop buying into that Big Lie. Please. Any who knowingly allow children to be hurt would be better off to be cast in the water with a millstone around their neck than face the judgment that awaits them said Jesus. You . Are . Complicit. Have courage. Speak out. This paper is the tip of the ice very like penn state just becuse it is terrible does not mean it isn't true.
Posted by: Annie Donnelly | September 02, 2014 at 04:34 PM
I think more than a few statisticians have flagged DeStefano's findings. Hinjen & DeSoto basically made it a text book case & devoted a chapter to their study(s). Cochrane Reports also flagged it & suggested they'd biased their data-essentially gaming the control group Thus, this suspicion of their nonsense and obfuscation has been out there for nearly a decade.
Wondered if you'd make a comment about data analyses done by Hooker (cohort study) vs DeStefano et al (case control study) on data that was collected for a case control study. In other words, how robust is this approach?
This is of course, notwithstanding the fact that DeStefano deliberately misrepresented their findings and remain defensive about their approach. Frankly, when a researcher states that autism starts in the womb (ie, it's bad genes) and appears to discount additional factors (ie, race, sex, vaccination times, diet, locale, etc) that might provide additional insight, it begs the question about the scientist's objectivity in data evaluation and conducing these kinds of studies. If this group or institution/agency a priori decided that "autism starts in the womb _or_ vaccines don't cause autism", therein lies a problem. It's a problem that likely exists across publicly-funded institutions, doesn't appear to be getting better, and ultimately will erode public trust in nearly any information that comes out of these agencies.
Posted by: flahute | September 02, 2014 at 03:24 PM
Re-read the piece to find that I had initially misunderstood that the subsample with the birth certificates they cherry picked in reaching their conclusions actually DID have race as one of its variables, making the omission all the more glaring.
I can't tell whether such a glaring omission demonstrates hubris or simply sloppiness in covering up the findings.
The total sample could have used subjects that ALL had birth certificates, substituting mother's occupation for maternal education>16. Data on the race of all the test subjects in a total sample shouldn't be that difficult to obtain either.
Posted by: Handthatrocksthecradle | September 02, 2014 at 02:43 PM
I wonder how many kids have been saved from a descent into autism by the revelation of William Thompson's confessions. I know I have informed more than one young mother heading to a well baby checkup with a slew of vaccines scheduled that they needed to forestall the vaccines. Many young mothers are very nervous about vaccines, but they simply succumb to the pediatrician's bullying. Now the strong ones can take a stand and refuse until the CDC fraud matter is dealt with. For the sake of those babies headed to well baby visits last week and this week, I'm glad the exposers didn't wait a day longer.
Posted by: Betty Bona | September 02, 2014 at 02:14 PM
As has been stated, it would have been better to leak the information without Drs. Hooker and Wakefield being spearheads and instead following up later. A small anonymous leak may have focused the vitriolic loons on Dr. Thompson. But of course, they'd have no way of predicting such a response in advance. Anymore than say 3 spitting, rabid TV Network anchors crucifying Dr. Wakefield. Let's hope someone, somewhere with honor in the CDC - or of stature elsewhere - is preparing to jump into the fray.
Posted by: James Butler | September 02, 2014 at 12:23 PM
What was their rationale for not making sure that all test subject in the total sample had birth certificates showing all the variables? Instead, this suggests that they selectively and deliberately dissected the subsample, all of which did not have info with respect to race, maternal data, because birth certificates were conveniently "unavailable" so that a correlation between race and autism could be established?
How did they justify not comparing different samples if ALL the variables (some had birth certificates, some didn't)from the TOTAL sample
were NOT equal and available from the outset?
Is this usual protocol? Is such sloppiness typical or would it be considered beyond the pale for this type of study outside or inside the CDC?
Anyone know?
Will they now claim that this was just an "oversight"? Will their defense be later that they were just being sloppy, certainly (wink, wink) not criminal or unethical?
Posted by: Handthatrocksthecradle | September 02, 2014 at 12:20 PM
Despite the importance of Hooker's reanalysis of DeStefano et al 2004 study, 'Translational Degeneration', the journal publishing Hooker's paper, has now retracted it and is 'reviewing' whether or not to re-install it.
'Translational Neurodegeneration' has put out two different statements justifying retracting the Hooker article.
1. http://www.translationalneurodegeneration.com/content/3/1/16/abstract
"This article has been removed from the public domain because of serious concerns about the validity of its conclusions. The journal and publisher believe that its continued availability may not be in the public interest. Definitive editorial action will be pending further investigation."
2. http://www.translationalneurodegeneration.com/content/pdf/2047-9158-3-18.pdf
"The Publisher of this article has serious concerns about the validity of its conclusions because of possible undeclared competing interests of the author and peer reviewers. The matter is undergoing investigation. In the meantime, readers are advised to treat the reported conclusions of this study with caution."
Which is it: invalidity or conflict of interest? It's perfectly possible for someone with competing interests to produce a scientifically valid research paper. Presumably the CDC, who have an interest in advancing vaccines as the bedrock of public health, would accept this.
Readers of AoA can write to the editorial team about this matter
[email protected]
Posted by: Martin Hewitt | September 02, 2014 at 11:58 AM
Martin, Thank you for your clear analyses of this fraud. And thank you Dr. Brian Hooker for taking the lead on working with the CDC whistleblower Dr. William Thompson.
In terms of the potential impact of this fraud here is a comment submitted to Sharyl Attkisson's site at http://sharylattkisson.com/audio-cdc-addresses-allegations-on-vaccine-autism-link-omission :
"...Since this study in 2004, over six million African American babies were born. See table 1 at http://www.cdc.gov/nchs/data/nvsr/nvsr62/nvsr62_09.pdf .
The CDC estimates that the current MMR immunization rate is at 91 percent. See http://www.cdc.gov/nchs/fastats/immunize.htm .
So that means for those African American children that received the MMR vaccine before three years of age– the CDC had statistical evidence that over three million boys born during that next ten year period could have a risk of Autism that is 3.36 times greater than those that did not receive the vaccine before three years of age."
Posted by: Jim Thompson | September 02, 2014 at 07:28 AM
Thanks for a sober analysis. This CDC technique for burying bad news ought to be enshrined in stats texts as the "DeStefano Correction".
Posted by: Nick Swift | September 02, 2014 at 06:17 AM