CDC Senior Scientist William Thompson has alleged that a 2004 study he co-authored omitted key data that would have revealed a link between autism and a commonly-required childhood vaccine, MMR (Measles, Mumps, Rubella).
“I regret that my coauthors and I omitted statistically significant information in our 2004 article published in the journal Pediatrics. The omitted data suggested that African American males who received the MMR vaccine before age 36 months were at increased risk for autism,” said Thompson in a statement issued through his attorney Wednesday.
The CDC and Thompson’s co-author Dr. Frank DeStefano, CDC Director of Immunization Safety, defend the study as originally published.
At issue are steps the researchers allegedly took when they discovered a statistically significant link between MMR vaccine and autism among African American boys. They “refined” the analysis by excluding study children for which a Georgia birth certificate could not be produced, and used birth certificate data to “adjust” the results. When they did so, the strong statistical association diminished.
In a statement, Thompson implies his CDC employer and colleagues have not always been forthright with the public about vaccine safety issues. But DeStefano argues their final analysis produced more valid results.
Below is the audio from a telephone interview I conducted with the CDC’s Dr. DeStefano this week, so you can hear his reasoning in his own words. For his part, Thompson says he is not speaking to reporters. “I am providing information to Congressman William Posey, and of course will continue to cooperate with Congress,” Thompson said.
Above: click to hear audio of Sharyl Attkisson telephone interview with CDC’s Dr. Frank DeStefano about his questioned MMR-autism study, Aug. 26, 2014
Dr. Frank DeStefano: I think what [Thompson’s] saying there was a larger, um, uh, odds ratio or association among the–the larger group and that there was not, uh, as strong an association among the birth certificate sample. And I mean, what I say to that, I think we discussed that, uh, as I recall, this was like, you know, over ten years ago, and, uh, I think at the time we had consensus among all co-authors that the birth certificate sample provided the more valid results because it could uh, it had more complete information on, uh, on race for one, and secondly, more importantly, it had information on important factors that, uh, had to be you know controlled for particularly in studies of autism, in particular, it would be things like birth weight, the mother’s age, the mother’s education. So I think for those reasons we were able to adjust for these factors and we thought, you know, we uh, our opinion was that that the results of the birth certificate sample provided the more reliable results. And I think, you know, as I recall, we all came to consensus and, uh, signed off on that, uh, in the paper.
Attkisson: Were you aware of any of his concerns of, you know, have you been aware before today of any of his concerns about this?
DeStefano: Uh, uh, yeah, I mean I’ve continued to see, uh, uh, see him for over the past ten years and we’ve interacted fairly frequently, and, uh, uh, no I wasn’t aware of this.
Attkisson: So whoever he raised his concerns to, he didn’t, he didn’t raise it to you or anybody you knew of?
DeStefano: No, I mean the last time I saw him was probably about two months ago, and he didn’t mention anything about this.
Attkisson: And at the time he didn’t seem concerned when you said there was a consensus?
DeStefano: No, yeah, I mean at the time he did these analyses he did, you know, he did point out that in one group, you know in that larger group the the the measures of association [between MMR vaccine and autism] were higher than in the, uh, birth certificate group and, you know, we discussed that and for the reasons I mentioned, uh, we came to consensus that the, uh, birth certificate uh results were more valid.
Attkisson: Um, I was looking at one of the birth certificates and it doesn’t have – maybe you could find one that has birth weight, mother’s education, the one I’m looking at doesn’t have any of that on there.
DeStefano: Ah, I mean I don’t know what, which one you’re looking at, I mean we get to these data were, uh, you know, right from the birth, birth, uh, the Georgia birth certificate files that contained those data.
Attkisson: Ok. Does is it a valid way of you know, you guys, scientists decide things before papers are published, of course, you use your own judgment on things, but isn’t there a way, is there a valid way to look at it the way Thompson is, where he thinks, apparently, that the larger group without the birth certificates was reason for concern and something that should have been reported? To me, as just as a layperson, I would like to know that– even if, even if it culled out when you, when you got the group down through the birth certificates, I would, I still think it would be pretty important to know…
DeStefano: No, I mean, I think, you know, the other, the other important consideration here is looking at what, what time period we’re talking about. We’re, you know, autism, as you probably are aware, is a condition that really probably has its start while the child is still in the womb. And, you now, it doesn’t, some of the behaviors and such don’t come apparent, become apparent until maybe the child is one, two, three years old. But, uh, uh what we know about autism that, uh, the, uh, characteristics or behavioral signs do become ava–, you know, apparent by 24 months of age, so. So we had different cut-offs, before 18 months of age, there was no difference in, in any group in terms of, uh, vaccination levels, between the cases and controls. At 24 months of age, when, uh, au—you know—behaviors of autism or some features of autism become apparent, there was no difference between the, uh, cases and controls in any group, it was at 36 months where there was a slight differen—and the difference was we’re talking about a difference between 93% versus 91%, not a, a big difference. But, so that’s at 36 months. And at 36 months, an exposure around that time period is just not biologically plausible to have a uh, uh, a causal association with autism. I mean autism would’ve already started by then. [I me?] I reiterate it probably starts in the womb, but even if you’re saying, you wanna call it starting by the time some behavioral features become apparent, it had started before 36 months. And then, you know, we, from, so I think from a biological argument, it’s implausible this was a causal association. And then I think we have, uh–pretty convince–
Attkisson: Let me just, let me just interrupt what, before I lose that thought. So you already made up your mind regardless of what the stats show that if it, certain things show that it didn’t make sense, you wouldn’t, you would try to find out a way to…
DeStefano: No, that’s not we said, I’m just saying, you know, you interpret, you interpret findings, also, you know, there’s the statistics, then you have to also interpret, bring in things like biological plausibility, how do you interpret these results? So I think we had pretty strong evidence that these results at 36 months were primarily a reflection of requirements to attend early intervention special education programs for the for the children with autism. And why do we say that? We say that because the effect was almost all seen in children 3-5 years of age and those were the ones that early education programs and 98%, you know, 98% of that of that age group was in special education programs for which vaccination was of a requirement.
Attkisson: Is there any possibility that it is biologically plausible and you just haven’t, you know, that that’s, the consensus is that it’s not, among you guys, but that it is and you’re overlooking that?
DeStefano: I’m, I’m not aware of any data would, that would s–, you know, that would say that, uh, you would have, um, onset of autism after 36 months.