No difference between “too many too soon” and “too few too late” when it comes to vaccines

ResearchBlogging.orgRegular readers know that I have a tendency every so often to whine about when writing about the antics of the anti-vaccine movement seems to engulf this blog. Yes, it’s true. Every so often I get really, really tired of the bad science, pseudoscience, magical thinking, misinformation, and even outright lies that emanate from various anti-vaccine websites and blogs. This week, I promised myself I would try not to do it. There are times when duty calls, and this is one of those times. For better or for worse, as hard as I still find it to believe, somehow I’ve become one of the top bloggers defending vaccines, and there are times when I have to stop whining and just embrace this role. Given the one-two-three triple whammy of the start of the anti-vaccine autism quackfest Autism One, Andrew Wakefield having had his license to practice medicine in the U.K. yanked, and the hilariously insane anti-vaccine rally to take place in Grant Park in Chicago tomorrow, it’s time just to go with the flow and do what needs to be done.

Today this is something I’m more than happy to do, at least today.

If there’s one thing about the anti-vaccine movement that I’ve learned over the last several years, it’s that it’s nothing if not, for lack of a better word, nearly infinitely pliable. To put it more simply, anti-vaccine activists are experts at throwing out as much stuff as they can and seeing if anything sticks, adjusting their stories, and moving the goalposts every time each of their successive demands for more evidence are met by scientists. Although there has always been an anti-vaccine movement, its most recent incarnation is built primarily around the idea that vaccines cause autism somehow. First it was Andrew Wakefield presenting dubious, trial lawyer-funded “research” purporting to show that the MMR vaccine causes “autistic enterocolitis” and even autism itself. Then, not long after that, the U.S. version of this manufactrovery showed up in the form of the concept that mercury in the thimerosal preservative that used to be in vaccines cause autism, promoted initially by David Kirby and Robert F. Kennedy, Jr. Fortunately, more than a decade’s worth of research consisting of large epidemiological studies has utterly failed to find even a whiff of a hint of a link between either the MMR vaccine or thimerosal-containing vaccines and autism. Unfortunately, the anti-vaccine movement simply moved the goalposts to the “toxins” gambit, in which it is claimed that vaccines are laced with “toxins” such as formaldehyde, antifreeze, and tissue from aborted fetuses. Never mind that there are no parts from aborted fetuses or antifreeze in vaccines, and scary-sounding chemicals like formaldehyde are present in concentrations far too low to be a problem. Even so, “Green our Vaccines” sure sounds like a slogan that means something, even though it doesn’t.

The latest gambit, and arguably one of the most successful because it’s the most vague and difficult to falsify, is the “Too Many Too Soon” slogan. Under this idea, anti-vaccine propagandists claim that infants are getting too many vaccines too soon (hence the slogan) and that all those nasty vaccines being given to such young infants is somehow messing up their immune system, attacking their brains, and giving them autism. This is a tough one to combat because the only definitive way to refute it would involve studying unvaccinated versus vaccinated children (which, not surprisingly, is the latest demand of the anti-vaccine movement). Doing such a study in a rigorous prospective randomized fashion would be completely unethical because it would leave the control group unprotected, while retrospective studies would be prone to a lot of confounders, given that there are likely to be other factors besides vaccination status that make populations who aren’t vaccinated different from those who are.

That’s why a study hot off the presses yesterday (well, hot off the web, as it were, given that it’s an E-pub ahead of print) is so well timed. Released right as Autism One starts and Andrew Wakefield tanks, what better time for a study to look right at the very question that anti-vaccinationists seem to want answered? The title of the article is even an arrow aimed right at the heart of the “too many too soon” mantra, namely On-time Vaccine Receipt in the First Year Does Not Adversely Affect Neuropsychological Outcomes.

This study, which comes from investigators at the University of Louisville, couldn’t compare unvaccinated children with vaccinated children. What it could do, however, was to provide about as good evidence shy of a prospective study to put the lie to Dr. Sears “alternative” vaccine schedule and the claims that, if parents just “spaced out” vaccines and “vaccinated more slowly,” it would be safer and wouldn’t cause autism. In this study, the investigators in essence compared children who had received all their vaccines on time with children who did not, looking for differences in neurodevelopmental outcomes.

Before I describe the study a bit more and, more importantly, its results, strengths and weaknesses, let me just predict what one of the criticisms of the anti-vaccine movement will be. Here is the paper’s disclosures:

Drs Smith and Woods are or have been unfunded subinvestigators for cross-coverage purposes on vaccine clinical trials for which their colleagues receive funding from Wyeth, Sanofi Pasteur, GSK, MedImmune, and Novartis; and Dr Woods has received honoraria for speaking engagements from Merck, Sanofi Pasteur, Pfizer, and MedImmune and has received research funding from Wyeth and Sanofi Pasteur.

Yes, anti-vaccinationists will be screaming “big pharma bought this study.” Of course, big pharma didn’t actually pay for this study, and Dr. Smith hasn’t even received any pharmaceutical funding. All an “unfunded unfunded subinvestigator for cross-coverage purposes on vaccine clinical trials” is is a physician whose partner is a funded investigator for a vaccine trial and who thus might have to cover for his from time to time for purposes of taking care of patients who are on the trial when his partner is out of town. As for the rest, I can only point out one huge difference: This potential conflict of interest (COI) is listed right up front. Andrew Wakefield did not disclose his COI of being funded by trial lawyers for his 1998 Lancet paper, nor did Laura Hewitson initially reveal that she is the mother of an autistic child who was a complainant in the Autism Omnibus trial the first time she tried to report her results.

Smith and Woods then did something ingeniously simple in concept but devilishly difficult in execution. Basically, they took publicly available data from the VaccineSafety DataLink Study of thimerosal exposure and 42 neuropsychological outcomes. This is the same dataset that I wrote about when its results were published in the New England Journal of Medicine nearly three years ago. Basically, the study looked at a whole bunch of neurodevelopmental outcomes other than autism and tried to correlate them with thimerosal-containing vaccines. There were no associations other than those likely to be due to random chance alone. More interestingly, Sallie Bernard of SAFEMINDS participated in the design of the study and then turned on it when it didn’t show what she expected that it would show, thus providing more evidence of the futility of trying to include anti-vaccine zealots in such activities is pointless and only leads to trouble.

This time around, Smith and Woods decided to look at the same dataset, which includes 1,047 children, and try to correlate on-time vaccination status versus delayed vaccines with neurodevelopmental outcomes. These are the same neurodevelopmental outcomes that Thomson et al reported on in the NEJM in 2007; so what this study does not provide us with is an answer to the question of whether “too many too soon” affects autism incidence. However it can answer a lot about the 42 neurodevelopmental measures studied in the original paper in 2007. The methods were as follows:

Publicly available data, including age at vaccination, from a previous VaccineSafety Datalink study of thimerosal exposure and 42 neuropsychological outcomes were analyzed. Vaccine receipt was defined as timely when each vaccine was received within 30 days of the recommended age. Associations between timeliness and each outcome were tested in univariate analyses. Multivariable regression models were constructed for further assessment of the impact of timeliness on neuropsychological outcomes after adjustment for potential confounders. Secondary analyses were performed on a subset of children with the highest and lowest vaccine exposures during the first 7 months of life.

These analyses were controlled for potential confounders, including age, gender, birth weight, poverty status, maternal IQ, maternal education, study site, cumulative ethyl mercury exposure during the first 7 months of life, and Home Observation for Measurement of the Environment score (an objective assessment of stimulation and emotional support in the home environment, which has been associated with developmental outcomes). Additional confounders examined included day care versus home care, language (English only versus multiple languages), duration fo breast feeding, Apgar scores, maternal age, and others. Both univariate (one variable at a time) and multivariate (correcting for multiple comparisons) were performed.

Now here’s where it gets interesting (and sometimes difficult to explain):

In the primary analyses, timely receipt was significantly associated with better performance on 12 of 42 outcomes in univariate analyses (Table 2). Specifically, children with timely receipt scored statistically better on the Boston Naming Test, grooved pegboard, metacognition, and teacher Connor’s ratings for hyperactivity and inattentiveness. They also had higher verbal, performance, and full-scale IQs and were reported by parents to stutter less than children with untimely receipt. Children with untimely receipt did not perform better (no clinically or statistically significant differences) on any of the outcomes.

So in this analysis, vaccines actually appear protective. But is this real, or is this just a bunch of spurious results from multiple comparisons? To test this, Smith and Woods subjected the data to multivariate analysis. When they did, what they found is that two of the 42 measures actually remained significantly different and continued to favor the group who received vaccinations on time according to schedule. Specifically, children who received vaccines on time scored 1 point higher on the Developmental Neuropsychological Assessment (NEPSY) speeded naming test. They also scored 2.7 points higher on the Wechsler Abbreviated Scale of Intelligence performance IQ, a standard test used by school psychologists and child psychologists.

Of course, the apparent better results noted in the group vaccinated on time could well be due to confounders not yet identified, but one thing’s for sure. Not a single statistically significant negative neurodevelopmental outcome was found in the group vaccinated on time compared to the group that received some or all of its vaccines late. Moreover, the investigators identified two subsets, a group who had the maximal exposure to vaccines during their first seven months of life, and compared them to a matched “least timely” group, who had, true to their name, the least timely exposure to vaccines and recieved almost half fewer than the most vaccinated group. Comparison of these groups showed that the more vaccinated group did better on 15/42 of the neurodevelopmental outcome measurements on univariate analysis, but these differences didn’t persist in multivariate analysis. In other words, there was no statistically significant difference in any of the measured outcomes between the two groups.

Overall, this study was fairly cleverly done, and quite timely. It has the strength of using a dataset that has a lot of socioeconomic indicators that might be confounders, allowing the investigators to correct for them. It is also a rigorous dataset that doesn’t rely on parental reporting but rather actual measures by health care professionals. I can, however, see one criticism that anti-vaccine zealots will level at it. Specifically, the children in this study were born between 1993 and 1997. The results reported in Thompson et al in 2007 examined neurodevelopmental outcomes between 2003 and 2004. Thus, the vaccine schedule was different. The authors answer this criticism quite convincingly:

Because the children in this study were born between 1993 and 1997, these results may not be generalizable to the current infant immunization schedule, which now includes 3 doses of heptavalent pneumococcal conjugate vaccine, 3 doses of oral rotavirus vaccine, and 1 or 2 doses of influenza vaccine in the first year of life (earliest eligibility at 6 months of age). This limitation is presently unavoidable in any vaccine safety study with long-term follow-up. However, most of the children in this study received DTP rather than DTaP, so the total antigenic burden to which children in this study cohort were exposed was actually higher than that encountered by children today.

It’s good to see a study like this begin to address the latest, most pernicious of antivaccine myths. Even though it doesn’t look at quite the same vaccine schedule as what we have today, the schedule that it did study was comparable. Moreover, if, as anti-vaccinationists say, “too many too soon” is bad and “spreading out” vaccines is safer, then one would expect to see differences between the most vaccinated and least vaccinated groups in a study like this. No such differences were observed that held up to multivariate analysis, and all the differences observed in univariate analyses favored the more highly vaccinated group. Finally, even though this dataset is from the “thimerosal era” of vaccination, when several childhood vaccines contained thimerosal and thus appears to be yet another weakness that anti-vaccinationists can point to. However, it’s fairly easy to answer that criticism by simply pointing out Thompson et al, which looked specifically at thimerosal exposure during that time period using exactly the same dataset. No differences due to thimerosal exposure was found in these neurodevelopmental outcome measures other than what could easily be ascribed to random chance. Taken together, these two studies provide a great deal of reassurance that vaccine exposure does not correlate with long-term developmental outcomes or that, if it does, less vaccinated or unvaccinated children actually do worse. Truly, maybe when anti-vaccine activists shout “too many too soon” at us we should should back “too few too late” is dangerous.

Actually, though, as much as I like seeing a study like Smith and Woods’, seeing such a well-designed study see print profoundly depresses me in a way. Even though Smith and Woods have provided a very strong bit of reassuring evidence that vaccines do not cause neurodevelopmental harm in a large number of outcomes, it’s depressing that such a study is even still considered necessary in 2010 and even more depressing that a lot more money and a lot more effort will be poured into similar studies. That’s money that gets diverted from much more fruitful potentially fruitful research into the causes and potential treatments for autism.


Michael J. Smith, MD, MSCE, Charles R. Woods, MD, MS (2010). On-time Vaccine Receipt in the First Year Does Not Adversely Affect Neuropsychological Outcomes Pediatrics