Vaccine Papers An Objective Look at Vaccine Dangers

Click for list of papers in this post.

Most studies investigating the relationship between vaccines and autism have focused on two things: mercury and the MMR vaccine.

Some research on the subject of mercury (thimerosal) shows that it is associated with neurodevelopmental disorders and damage, at vaccine-relevant dosages (i.e., at dosages received before it was removed from most vaccines). However, the effect is too small to explain the vaccine injuries in the USA today. Also, since autism rates have increased after thimerosal removal from vaccines, mercury cannot be the primary cause of the autism epidemic.

Almost all vaccine-autism research focuses on MMR. This is an unfortunate consequence of the hugely misleading Wakefield controversy. The design of most of these studies is such that they can only demonstrate that MMR is not uniquely hazardous with respect to causing autism. This is our interpretation. The MMR is perhaps not particularly dangerous compared to other vaccines, and it may be less dangerous than other vaccines. Since MMR does not contain aluminum, and its administered at a later age, this is perhaps to be expected. But that’s not saying much.

So what does the science say about the other vaccines? See this list of studies from the American Academy of Pediatrics. American Academy of Pediatrics, Vaccine Safety: Examine the Evidence
I welcome the invitation by the AAP to “examine the evidence”.

This list has 42 studies, on the following topics:

MMR/measles vaccine and autism: 25 studies
Thimerosal/mercury: 9 studies
Review papers: 3 (mostly on MMR and thimerosal)
Vaccines and metabolism: 1 (Klein 2011)
Vaccine timing and number: 1 (Smith and Woods, 2010)
Vaccine antigen exposure: 1 (DeStefano 2013)
Autism-GI disorders: 1 (Black, 2002)
Parental reports of autism triggers: 1 (Lingam, 2003)

There are no studies on aluminum adjuvant. There are no studies on immune activation. There are no studies comparing vaccinated with the unvaccinated (100% unvaccinated). And there is only one that looks at the timing of vaccine exposure. Concerns about aluminum and immune activation causing brain injury and autism are not mentioned in this AAP document, despite the fact that the dangers of both have been known for several years. The AAP document was last updated in April 2013, so the AAP has had plenty of time to review the studies of immune activation and aluminum adjuvant.

The AAP uses this list to assert that all vaccines have been proven to not cause autism or other harm, when only one ingredient (thimerosal) and one vaccine (MMR) have been much studied. This is not reasonable and goes far beyond what these studies can show.

However, there is one study on this list that appears applicable to concerns about vaccines in general, and aluminum: the vaccine timing study by Smith, 2010. This study is interesting because it explores the timing of vaccine exposure, and to some extent the amount of vaccine exposure. This is definitely the direction the science needs to go. I assert that vaccines and aluminum do cause autism. If this is correct, then the effect should be observable in a study on the timing and amount of vaccine exposure. Does Smith deliver some answers, one way or the other?

Lets look at the Smith 2010 paper and see what we find.

Full Paper (Smith 2010): On-time Vaccine Receipt in the First Year Does Not Adversely Affect Neuropsychological Outcomes

The Smith paper has 37 citations, so its fair to say its had substantial impact and influence. The AAP apparently agrees, since they included Smith in their list of 42 papers.

The Smith study is based on a dataset designed for a study of thimerosal. 1047 children born between 1993 and 1997 were tested for neuropsychological function at age 7-10 (in the time period 2003-2004). The dataset includes socioeconomic and medical history information and of course the total vaccine history.

Two analyses were performed. The “primary” analysis looked at all 1047 children. The “seconday” analysis looked at 112 and 310 children receiving the least and the most “timely” vaccine exposure. The least timely group received fewer vaccines, and received them at older ages.

The Smith study is mostly concerned with the “timeliness” of vaccines. Timeliness is defined as obtaining all recommended vaccines within 30 days of the scheduled dates, up to age 7 months. Vaccine schedules for 1994, 1995 and 1997 are provided at bottom.

The vaccine exposure and other information about the subjects is shown below.

Above: The vaccine exposure was similar in the two groups (10.10 vs 11.8 and 7.4 vs 11.8 at 12 months). Other critical factors (male gender, household income, maternal college degree, single parent household %), were very different. This introduces substantial confounding that must be corrected by adjustment. From Smith 2010. 

There are several big problems with the Smith study that, in combination, make the results essentially worthless:

1) Inadequate vaccine exposure difference.
A subject is allocated to the “untimely” groups if even a single vaccine is received outside the 30-day window. This accounts for the small difference in vaccine exposure between the compared groups.

In the primary analysis, the vaccine exposure in the first year was 10.1 vs 11.8 vaccines, and 8.0 vs 11.1 in the first 7 months. This small difference in vaccine exposure will not produce large, easily observable effects.

The “dose-response” curve of the effect of vaccine exposure on brain injury is not known. If the dose-response curve is relatively flat in the region of about 10-12 or 8-11 vaccines, then this study will absolutely not be able to observe any effect, no matter how good the data analysis and how good the subject matching is.

The secondary analysis at 7 months is better for observing an effect (4.2 vs 11.2 vaccines), but this difference declines by 12 months (7.4 vs 11.8 vaccines). Severe vaccine reactions and brain injury have been observed at many ages, so its not known if ages 0-7 months are more sensitive than say 7-12 months (though this seems likely). In the CDC’s internal studies of thimerosal exposure, the earliest vaccines (at ages 1-2 months) seemed to cause the most damage, and the risk declined after about age 2-4 months. But this is for thimerosal only, and not necessarily vaccines generally. We don’t know how the risk for aluminum changes with age, for example.

Vaccine critics have said for many years that studies with small differences in vaccine exposure are not adequate for establishing safety, or quantifying the risks. The only way safety can be determined is by comparing the vaccinated with the completely unvaccinated.

The small vaccine exposure difference between the groups is a fatal flaw that renders the Smith results unconvincing.

2) Groups are unreasonably mismatched for important factors (confounders) known to strongly influence neuropsychological health.
The less-timely vaccinated subjects had much lower socioeconomic status than the timely-vaccinated. Income and maternal college degree % were lower and single parent households were higher in the untimely vaccinated. This difference is even greater in the secondary analysis, which strongly suggests that the less timely vaccinated were in this group due to neglect, poverty and lack of access to medical services. All these factors are known to strongly influence neuropsychological health, intelligence and other outcomes measured in this study.

Smith performed “multivariate analysis” (MA) to attempt to control for these confounders. MA is a known and accepted method for (hopefully) eliminating the effect of mismatches between compared groups.

But MA has limitations. When the mismatches are large or numerous, then large adjustments must be made, and this introduces error than can conceal an effect.

Also, large mismatches tend to give researchers undue and inappropriate power to determine the outcome of a study. The amount of adjustment provided by MA can be selected by researchers, based on numerous assumptions and subjective decisions on how to fit trendlines and the like. MA can be a free license for researchers to determine the study outcome, instead of the data determining the outcome. This is a reasonable basis for rejection when the study authors have conflicts of interest, which Smith and Woods do.

In the Smith study, the entire outcome seems to depend critically on how much adjustment is introduced to negate the socioeconomic mismatches. In other words, the large mismatch between groups gives Smith and Wood enormous freedom to adjust the analysis to obtain whatever outcome they desire.

Consider, for example, if the groups were accurately matched. Little MA would be required, and there would be little opportunity for Smith and Woods to sway the outcome. I am not asserting that Smith and Woods did do such things, just that they had the opportunity to do so without detection, if they wanted to.

Also, its important to note that factors #1 and #2 interact in a way that further undermines confidence in the results. Factor #1 (small difference in vaccine exposure) makes the effect of vaccines small and difficult to observe. Factor #2 (large MA adjustment) produces large errors that can overwhelm the small effect of vaccines. In combination, these two problems can synergistically conceal the effect of vaccines.

For these reasons, the large mismatches create concern about the reliability of the results, even though multivariate analysis was used to attempt to correct them. This is especially so due to the stated conflicts of interest of Smith and Woods.

3) The largest confounders bias the results in the same direction: toward a finding of no effect.
There is an additional confounder that does not seem to be connected to socioeconomic status: there are more males in the “untimely” group. The male% mismatch is 51 vs 45.8% in the primary and 58 vs 46.5% in the secondary analysis. Why male babies would be less likely to be vaccinated on time is not discussed by Smith, and perhaps is not known. But I assert that this is because males are more likely to suffer vaccine reactions, and parents are less likely to vaccinate on time if they observe such reactions or signs of autism. There is evidence that autism diagnosis or neurodevelopmental problems have a strong influence on parental vaccination decisions. See below*.

Males are known to be more susceptible to autism and aluminum adjuvant toxicity. In experiments performed by the Shaw laboratory at UBC (LINK: Experiment #3), it was found that males are more sensitive to the neurotoxic effect of aluminum adjuvant. So the hypothesis of vaccine-induced brain damage and autism asserts that males are more sensitive. Therefore, if the vaccines are producing adverse effects, the “untimely” groups will be more strongly affected due to the overabundance of males. Consequently, the reduced number of males in the “timely” group will conceal the adverse effects of vaccines.

The socioeconomic mismatches operate in the same direction. The “untimely” groups had lower socioeconomic scores, and this will also result in lower neuropsychological scores. This will also conceal vaccine injury.

So, all these significant mismatches and confounders operate in the same direction: in opposition to the asserted effect of vaccination.

Now, multivariate analysis (MA) is supposed to correct for all this stuff. But as explained above, MA introduces error when the corrections must be large. And since all these effects move the outcomes in the same direction (toward a finding of no adverse effects), then the MA adjustment must be that much larger to counteract all of them. And large MA corrections introduce large errors, and allow for unscrupulous researchers to determine the outcome, instead of the data determining the outcome.

4) Multivariate analysis adjustment appears to be inadequate.
The mismatches described above, if not corrected for, will result in the timely-vaccinated group having far superior neurospychological test scores. The goal of MA is to adjust the scores such that the effects of the socioeconomic and gender mismatches (and other confounders) are eliminated, thereby leaving only the effect of the vaccines. If the vaccines have no effect, and all the confounders and mismatches are properly corrected, then the adjusted test scores in the two groups will be equal. In other words, MA attempts to isolate the effect of the vaccines on the test scores.

However, Smith observed that the timely-vaccinated children performed better on the neuropsychological tests, even after MA adjustment. This suggests that the MA adjustment was too small (unless one believes that vaccines cause better neuropsychological functioning). A too-small MA adjustment will conceal adverse effects of vaccines (because the mismatches bias the results in favor of no effect).

5) Study had only 9 children with zero vaccine exposure
Studies of adverse effects of vaccines are typically fatally flawed due to the failure to include a group of unvaccinated controls, and this study by Smith and Woods is no different. Only 0.86% of study subjects received no vaccines. Until some quality research is published comparing vaccinated with unvaccinated controls, the best available science on vaccine safety are controlled experiments with animals. And so far, the controlled experiments with animals show that vaccines cause brain damage and autism.

6) No attempt to control for healthy user bias.
Healthy user bias is a type of selection bias resulting from unhealthy, high risk children being concentrated in the unvaccinated group. This occurs because unhealthy children (e.g. with signs of neurodevelopmental or immune disorders) are often not vaccinated. Smith does not explain why the untimely-vaccinated children were not vaccinated on-time. But we know from other research (e.g. Jain et al. 2015) that parents avoid vaccinating sick children. I assert that healthy user bias is the reason why Smith observed better outcomes in the timely-vaccinated; those children were healthier to begin with. Smith states:

“Children in the most timely group performed statistically better than children in the least timely group for 15 of the 42 outcomes, including 10 of the 12 outcomes associated with better outcome in the primary analysis.“

Healthy user bias is explained in detail here: http://vaccinepapers.org/healthy-user-bias-why-most-vaccine-safety-studies-are-wrong/

Conclusion
The problems with the Smith and Woods study render the findings unreliable and essentially worthless. Vaccine safety is critical, so we must demand good quality rigorous science on the issue. The Smith study is not it. Its very unfortunate that we don’t have quality science to guide rational decision-making regarding vaccine safety.

The problems with Smith 2010 highlight the need for studies of the health outcomes of the unvaccinated.

A Word About Conflict of Interest
I don’t think its reasonable to reject scientific results merely because the researchers have conflicts of interest. Reasonable rejection of a study requires more than this, because even researchers funded by biased interests can do good science.

However, some factors, when combined with a conflict of interest, are cause for concern and even rejection:
1) poor research methods
2) study design that allows outcome to be determined by researchers instead of the data

The Smith 2010 study has both problems, in combination with a high conflict of interest (Smith and Woods receive funding from 5 and 6 vaccine manufacturers, respectively).

The use of groups with such similar vaccine exposure, and dissimilar socioeconomic factors and dissimilar gender ratios is poor science. Good quality biomedical science requires closely matched controls and substantial differences in the intervention under investigation (in this case the intervention is vaccine exposure).

The large MA corrections necessitated by the mismatches introduce large subjective influence by Smith and Woods. This makes the final results critically dependent on unknowable decisions by Smith and Woods. They can get any result they want.

These facts, in combination with their long relationship with numerous vaccine manufacturers, make it reasonable to reject the results. I reject the findings of Smith and Woods as worthless.

Arithmetic
There are a few arithmetic errors in the Smith paper. It states that 311 subjects comprise 20% of the study population (its actually 30%), and it states that 235 received all recommended vaccines (its actually 245). Smith and Wood have acknowledged these errors.

Other vaccine critics have made much of these arithmetic errors, and assert that this is reason enough to reject the paper. This is not reasonable. Arithmetic and typographical errors happen.  I don’t worry much about these arithmetic errors. The arithmetic errors are not nearly enough reason to reject the findings.
______________________________________________
NOTES
* I speculate (with some evidence) that males may have received delayed vaccinations due to a greater sensitivity to vaccine reactions and aluminum. A parent observing an adverse vaccine reaction will likely delay or avoid further vaccines, thereby placing vaccine-injured children into the “untimely” group. This will introduce a severe selection bias: the “untimely” group will have a greater proportion of vaccine-damaged children.  This bias will conceal adverse effects of the vaccines. And this effect provides an explanation of why the timely-vaccinated children had superior neuropsychological test scores. There is good evidence that autism diagnosis strongly affects parental vaccination choices, in Jain et al: http://vaccinepapers.org/review-of-jain-et-al-jama-2015-and-comments-on-mmr-autism/

CDC vaccine schedules for 1994, 1995 and 1997.

---

Papers in this post: