What Causes Autism?

This infographic presents some of what we know about immune activation events causing autism and states the vaccine-autism hypothesis.

Given the close association between the immune system activations and autism, supported by animal testing, the vaccine-autism hypothesis is highly plausible.

After all, the purpose of vaccines is to activate the immune system, which is something we know can trigger autism.

Sources are listed and linked below, and I have copied a line from the abstract of each study for the convenience of the reader.

Sources

  1. Hallmeyer, 2015: Autism heritability was estimated to be 38% and the shared environmental component to be 58% https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4440679/
  2. Atladottir, 2010: admission to hospital due to maternal viral infection in the first trimester and maternal bacterial infection in the second trimester were found to be associated with diagnosis of ASDs in the offspring https://www.ncbi.nlm.nih.gov/pubmed/20414802
  3. Zerbo, 2015: women with infections diagnosed during a hospital admission, particularly bacterial infections, were at increased risk of delivering a child with ASD https://europepmc.org/articles/pmc4108569
  4. Vargas, 2005: The brains of people with ASD show a marked activation of microglia and astroglia, and cytokine profiling indicated that MCP-1 and TGF- β1, derived from neuroglia, were the most prevalent cytokines. Cerebrospinal fluid showed a unique proinflammatory profile of cytokines, including a marked increase in MCP-1. https://www.ncbi.nlm.nih.gov/pubmed/15546155
  5. Li, 2015: Neonatal vaccination of rats with bacillus Calmette-Guérin and hepatitis B vaccines modulates hippocampal synaptic plasticity https://www.ncbi.nlm.nih.gov/pubmed/26531688
  6. Wei, 2011: The cerebellum of the brains of people with ASD has increased IL-6, which alters neural cell adhesion, migration and synaptic formation https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3114764/
  7. Abdallah, 2013: The amniotic fluid of mothers of children with ASD showed elevated levels of inflammatory cytokines https://www.ncbi.nlm.nih.gov/pubmed/22175527
  8. Suzuki, 2013: In multiple brain regions in people with ASD there is excessive microglial activation https://www.ncbi.nlm.nih.gov/pubmed/23404112
  9. Jones, 2017: The serum of mothers of children with ASD with ID showed increased levels of maternal cytokines and chemokines during gestation https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5122473/
  10. Tsilioni, 2019: The brains of children with ASD have increase inflammatory cytokines https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6027314/
  11. Smith, 2007: Maternal immune activation (MIA) in mice alters fetal brain development through interleukin-6 https://www.ncbi.nlm.nih.gov/pubmed/17913903
  12. Malkova, 2012: MIA in mice results in offspring that show more autism-like behaviours https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3322300/
  13. Bauman, 2014: MIA in rhesus monkeys yields offspring with abnormal repetitive behaviors, communication, and social interactionshttps://www.ncbi.nlm.nih.gov/pubmed/24011823
  14. Choi, 2016: Either MIA or direct administration to the fetal brain of mice of inflammatory cytokine IL-17a promotes abnormal cortical development and ASD-like behaviors in offspring https://www.ncbi.nlm.nih.gov/pubmed/26822608
  15. Missig, 2018: Early-life immune activation in mice can lead to long-lasting physiologic perturbations that resemble medical comorbidities often seen in ASD and other neuropsychiatric conditions https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5770773/

Aluminum in Infants

This infographic demonstrates that the majority of exposure to aluminum in infants comes from vaccines.

Children fully vaccinated using the CDC schedule for the first 6 months will receive 3,450mcg of aluminum into the blood from those vaccines, as the aluminum salts used as vaccine adjuvants slowly dissolve at the intramuscular injection site and enter the bloodstream. This process takes no more than 6 months, so 100% will have been absorbed by 1 year of age.

Breast milk only contains an average of 23.9 mcg/L, and infants drink an average of 0.74 L/day, for a total dietary intake of 6,455mcg of aluminum in the first 12 months. Most of this will be defecated out, but up to 0.78% could leak across the gut lining and enter the bloodstream. This would amount to 50mcg of aluminum over the year. Formula milk contains an average of 226 mcg/L and soy formula can contain up to 930 mcg/L. With the same amount drunk and same gut bioavailability, this would be 476mcg and 1959mcg respectively entering the gut. Infants on a semi-solid food diet will be getting an average of 767mcg. There is no diet that exposes infants to more aluminum through the gut than they are from vaccines.

The third biggest normal exposure is air, but the amounts here are insignificant. Multiplying average tidal volume by average respiratory rate results in 730 m3 of air being breathed by an infant by 1 year of age. Normal city air has around 0.4 mcg/m3 of aluminum, resulting in an intake of 292mcg. Lung bioavailability has been estimated at 2%, so only 6mcg of aluminum enters the bloodstream through the lungs during infancy. Even in a city with high pollution or industrial air, with up to 8 mcg/m3 of aluminum, that uptake figure only reaches 116mcg. Infant exposure to aluminum via the air is insignificant.

Sources for all these figures are listed and linked below.

Sources:

Covid-19 Deaths v Flu Shot Rates

In this post I will explore whether there is a correlation between Covid-19 deaths and elderly influenza vaccination rates across countries.

Introduction

At least two studies have found an association between flu vaccines and susceptibility to other non-influenza respiratory diseases, including other coronaviruses:

  • Cowling, 2012 – a randomised controlled trial (RCT) in which 69 children were given a flu shot and 46 a placebo. 20 vaccinated children (29%) got sick with a non-flu virus and only 3 unvaccinated children (6.5%) got sick with a non-flu virus, a statistically significant result. The most common non-flu viruses detected were rhinoviruses and coxsackie/echoviruses. For both these virus types, a significant association was found between getting a flu shot and susceptibility to the virus. The sample size was too small to identify any association between flu shots and coronaviruses.
  • Wolff, 2020 – a retrospective study of the illness and vaccination records of 9469 individuals working for the Department of Defense, of which 6541 had received the 2017/8 seasonal flu shot and 2928 had not. The flu shot was associated with reduced risk of getting the flu but an increased risk of non-influenza illnesses, including specifically coronaviruses. 507 (7.8%) of the vaccinated and 170 (5.8%) of the unvaccinated tested positive for a coronavirus, resulting a significant relationship with an odds ratio of 1.36 [1.14-1.63 95% CI].

The above studies suggest that the flu shot may increase susceptibility to coronaviruses, possibly by a mechanism known as viral interference. It is hypothesised that the flu shot may increase susceptibility to SARS-CoV-2 via the same mechanism. Hence, there are reasons to expect a correlation between flu shot rates and Covid-19 death rates.

Correlation does not imply causation, obviously. No ecological study, no matter how strong the correlation, can ever be strong evidence of causation. They can merely give us a clue about where to look and what further studies to do. Ecological studies are low on the evidence hierarchy (beneath RCTs, cohort and case-control studies) but they are quick, cheap and easy, which is why these studies are often first to emerge in cases like this.

Since we have no studies of the stronger types yet, ecological studies are the best we can do for Covid-19 at the moment. One study comparing regions of Italy can be found here; a significant negative correlation was found, but no confounders were examined. Here I will present my analysis of the international data we have for flu vaccination rates and Covid-19 death rates, and then examine six possible confounders.

Data Sources

Covid-19 death rates per million people by country are available from Our World In Data. In this study, death rates as at 31st July 2020 were used.

Influenza vaccination rates in elderly people (defined for most countries as aged 65+) are available from The OECD. However, data from within the last 5 years is only available for 31 out of the 37 OECD countries. I could find no explanation why there is no recent data for the 6 other OECD countries (Austria, Australia, Colombia, Mexico, Poland and Switzerland). In this study, latest available vaccination rate data is used for each country (for most countries, this is the 2018/9 seasonal flu vaccine uptake rates).

Europe

There are 23 European countries in the OECD for which flu shot data is available. I have excluded two of them – Iceland and Luxembourg – for having a population of less than a million. The remaining 21 countries are shown in the following plot:

As indicated by the slope of the red line, there is a positive correlation between these two variables. A correlation coefficient is a measure of the degree to which a pair of variables are linearly related, between -1 and +1. As shown on the chart, the correlation coefficient R is 0.67, which is considered a moderate-to-high correlation. The p-value of <0.01 shows that the null hypothesis (i.e. no relationship between these variables) is falsified by the data, with a confidence level exceeding 99%.

World

There are 8 non-European OECD countries for which data is available: Canada, USA, Chile, Turkey, Israel, Korea, Japan and New Zealand. Adding these countries to the plot, we find:

The positive correlation is still present but is weakened, the correlation coefficient now 0.49. This is due to Korea, Japan and New Zealand all being outliers, having high vaccination rates and low Covid-19 death rates. The p-value of <0.01 shows that the null hypothesis of no relationship between these variables is falsified by the data, with a confidence exceeding 99%.

Confounders

One way to improve an ecological study beyond a single-variable is to look for confounders. A confounder is a variable that influences both the dependent variable and independent variable, causing a spurious association. I have looked at all of the following six variables that have been suggested to me as possible confounders:

  • Income (as GDP-per-capita)
  • Population density
  • Elderly as a proportion of population
  • Climate (as average temperature in April)
  • Health of population (as life expectancy)
  • Healthcare system (as hospital beds per population)

Income

Slight positive correlation: richer countries had more Covid-19 deaths. Not statistically significant.

Population Density

No correlation: being more densely packed is not associated with more Covid-19 deaths.

Elderly Population

No correlation: having more elderly people is not associated with having more Covid-19 deaths.

Climate

No correlation: being a colder country is not associated with more Covid-19 deaths.

Health of Population

Slight positive correlation: countries where people live longer had more Covid-19 deaths. Not statistically significant.

Healthcare System

Here we have a weak but statistically significant correlation: countries with more capacity in the healthcare system had fewer Covid-19 deaths. The correlation coefficient is -0.37, and with a p-value of 0.049, the result is significant with 95% confidence but not 99% confidence.

Multivariate Models

Two Variables

Out of the 7 variables tested, two showed a significant relationship with Covid-19 death rates: flu shot rates and hospital beds. If we create a model based on these two variables, we obtain the following:

The p-value for the flu shot rate is still <0.05, so remains significant to a 95% confidence level, but the p-value for the hospital beds has gone to 0.068, so is no longer statistically significant at that level.

Seven Variables

If we create a model based on all seven variables that we have, we obtain the following:

Here we have an opposite result to the two-variable model. The p-value for the hospital beds is still <0.05, so remains significant to a 95% confidence level, but the p-value for the flu shot rate has gone to 0.076, so is no longer statistically significant at that level.

Conclusion

Does the flu shot make us more vulnerable to Covid-19? Until we have better kinds of studies, ecological studies of Covid-19 death rates against flu vaccination rates are the best way to get an idea of whether the flu shot makes people more susceptible to Covid-19 due to viral interference, as seems to occur with other coronaviruses.

This ecological analysis found a correlation coefficient of 0.67 when only European OECD countries are included. This would be classified as a moderate-to-high positive correlation. With the addition of non-European OECD countries the correlation coefficient is 0.49, a low-to-moderate positive correlation. This contradicts the finding of the study of Italian regions, which found a negative correlation of -0.58, without looking for confounders. Confounders between Italian regions should be examined, and similar studies of regions within countries should be done to try and resolve this apparent contradiction.

Six possible confounders have been analysed and one (hospital beds per thousand population) was found to have a significant association to Covid-19 deaths, just like the flu shot rates. When combined into a multivariate model, these two variables seem to cancel out, to the extent that one of them becomes statistically significant, but which one is dependent on what other variables are included in the model. I cannot explain this behaviour – if you think you can, let me know!

Appendix: Paul’s Chart

This picture has been shared on social media:

It comes from this blog, authored by “Paul”, who created it in response to a “whimsical suggestion” in the BMJ (here) by Dr Allan Cunningham: to correlate influenza vaccine uptake with Covid-19 death rates. My post above was inspired by Dr Cunningham’s challenge and by seeing the flaws in Paul’s chart and wanting to dig into the data myself.

In his note, Dr Cunningham provided data on influenza vaccine coverage rates in the elderly and covid-19 death rates per million for 20 selected European countries. His source for covid deaths was Worldometers, accessed 21st May 2020. His source for flu shot rates was the OECD. There are 26 European members of the OECD… Dr Cunningham did not include Belgium, Greece or Iceland in his list of 20 countries despite the data being available at the same source he used – with no explanation given. As we have seen, data from Austria, Poland, and Switzerland are not available from the OECD.

Here is a plot of the 20-countries data provided by Dr Cunningham:

A coefficient of determination, or R-squared, is the proportion of the variance in the dependent variable (covid deaths) that is predictable from the independent variable (flu shot rate). Using just Dr Cunningham’s 20 countries, we find an R-squared of 0.5327 under the assumption of a linear relationship, which corresponds to a correlation coefficient of 0.7299.

On Paul’s chart, this correlation coefficient is displayed prominently… but as we have seen, this value comes from Cunningham’s data as shown on the chart above. Paul’s chart is completely different, so the value of R he shows has no relation to his chart! Paul’s chart displays 27 data points (seven extra), the values plotted are different from Cunningham’s (due to a change to the data source), and the line drawn on the chart is exponential rather than linear (so it has nothing to do with the correlation coefficient of 0.7299, which assumes a linear relationship).

The extra seven countries that Paul added are curious. From among the three countries mysteriously omitted by Dr Cunningham, Paul rightly added back Belgium and Iceland, but not Greece, which would be a significant outlier, weaking the association. Paul adds Poland, which isn’t in the OECD data, and Romania and Croatia, which aren’t even in the OECD. Paul refers to the ECDC as a data source for these three and all other European countries. He may be referring to this publication, but I could not find the exact numbers he used. Adding Poland, Romania and Croatia strengthens the association because apparently they all have low vaccination rates and few Covid-19 deaths.

Strangest of all, Paul has added Canada and USA, being the only two countries on the chart outside of Europe. He uses OECD data for these countries, which makes it strange why he would omit other non-European OECD countries like Korea, New Zealand and Japan. These countries would all weaken the association, and their omission seems somewhat convenient for someone wishing to make the case that there is a strong correlation.

It is misleading to display on the chart an R value that has nothing to do with the data in the chart. It is misleading to cherry-pick countries and omit significant outliers without explanation. I think my charts give a more complete and honest picture than this chart by Paul.

The Flu Shot and COVID-19

The flu shot makes people more vulnerable to non-influenza viruses, including coronaviruses.

Thus, they are likely more vulnerable to COVID-19 too.

Source:

“Influenza vaccination and respiratory virus interference among Department of Defense personnel during the 2017–2018 influenza season”

Wolff, 2020

https://www.ncbi.nlm.nih.gov/pubmed/31607599

Mistakes of Mitkus (#1)

The argument that the amount of aluminum in vaccines is safe relies on a 2011 paper by Mitkus, which presents an aluminum pharmacokinetic model. The paper is flawed. The model is flawed. Here is an example of an error in the Mitkus 2011 paper.

Priest 2004 provided the equation for aluminum retention used in Mitkus’s model. The equation and the half-lives stated in this paper are inconsistent.

1.4 days half-life => exponent of 0.495
40 days half-life => exponent of 0.0172
1727 days half life => exponent of 0.000401

0.595 as 0.495 and 0.172 as 0.0172 look like typos.

Mitkus took the equation and half-lives from Priest’s paper, and apparently did not notice the inconsistency. He not only failed to correct the typos in the equation, he added one of his own, mis-stating 11.4 as 11.

Newton, the original source of the equation and co-author of the 2004 paper by Priest, later authored a paper with the same equation… but with the none of the typos.

Mitkus used the wrong equation

How could Mitkus have made such a basic error?

How did none of the co-authors or peer reviewers spot it?

Why has the paper not been retracted in light of this error?

Sources:

Priest 2004: The biological behaviour and bioavailability of aluminium in man, with special reference to studies employing aluminium-26 as a tracer: review and study update  https://www.ncbi.nlm.nih.gov/pubmed/15152306

Mitkus 2011: Updated aluminum pharmacokinetics following infant exposures through diet and vaccination https://www.ncbi.nlm.nih.gov/pubmed/22001122

Newton 2012: Long-term retention of injected aluminium-26  https://www.ncbi.nlm.nih.gov/pubmed/22549096

MMR-Autism Association in DeStefano 2004 Study

William Thompson is the CDC whistleblower who revealed that he had been involved in a cover-up of a key result in the vaccine-autism debate.

He was referring to the DeStefano 2004 study of MMR and autism, on which Thompson was a co-author, conducting the statistical analysis. Thompson claimed that an association between MMR and autism in African American boys was identified in the data, but that the finding was omitted from the final paper. He cited the pressure to show no association between MMR and autism, and explained how they tried various statistical techniques to try to hide the association.

The infographic above presents the data behind the debate. Brian Hooker’s 2014 re-analysis of the data shows there is indeed an association between MMR and autism in African American boys in the data.

Forget the politics; the science here is telling us there is an association between a vaccine and autism.

Sources:

Vaccine Autism Studies Inadequate

In 2012, the Institute of Medicine (IOM) released a comprehensive evidence review entitled “Adverse Effects of Vaccines: Evidence and Causality”.

They looked at 8 different vaccines and 76 different adverse events. One of these adverse events was autism.

  • For 1 vaccine (MMR), the IOM favored rejection of a causal relationship.
  • For 1 vaccine (DTaP), the IOM declared the evidence inadequate to accept or reject a causal relationship.
  • For the other 6 vaccines in the review, the IOM did not look for any evidence regarding a causal relationship.

Clearly then, the correct conclusion of this evidence is NOT that “vaccines do not cause autism”. There is not enough evidence to make that conclusion.

Even if a causal relationship between MMR and autism is rejected, it does not follow that “vaccine do not cause autism” because MMR is only one of 8 or more vaccines, and the evidence is inadequate to accept or reject a causal relationship for them. There have also been no studies looking for associations between cumulative vaccinations, or different timings, or different combinations of vaccines, and autism.

The CDC cites this IOM report for its claim that “vaccines do not cause autism” and yet this report does not support this claim.

Maternal Immune Activation and Autism

The diagram in the infographic above comes from:

The CDC Vaccine-Autism Studies

The studies cited by the CDC on their “Vaccine Do Not Cause Autism” page cannot possibly support that claim. The CDC’s conclusion is invalid.

See the infographic above for details about why that is.

The Vaccine-Autism Hypothesis

Here are the sources for each point in the above infographic.

Genetic / Environmental Etiology

The Claim: Autism is partly but not wholly genetic; it can be triggered in genetically susceptible individuals by environmental factors that alter neurodevelopment

The Sources:

Infections in Pregnancy

The Claim: A wide range of bacterial and viral infections during pregnancy are associated with autism, so one trigger is a serious immune activation event during a critical stage of neurodevelopment

The Sources:

  • Atladottir, 2010: admission to hospital due to maternal viral infection in the first trimester and maternal bacterial infection in the second trimester were found to be associated with diagnosis of ASDs in the offspring  https://www.ncbi.nlm.nih.gov/pubmed/20414802
  • Zerbo, 2015: women with infections diagnosed during a hospital admission, particularly bacterial infections, were at increased risk of delivering a child with ASD  https://europepmc.org/articles/pmc4108569

Immune Activations in Animals

The Claim: Immune activations in pregnant or newborn animals alter neurodevelopment and increase autistic-like behaviors in the pups and infants; multiple immune challenges multiply the effect.

The Sources:

Inflammation and Autism

The Claim: Inflammation is an immune system response to a challenge.  Inflammation is associated with autism.

The Sources:

Vaccines and Immune Activation

The Quote: “Immune activation is the objective of vaccines”

… is from Matheson vs. Schmitt, Plotkin Deposition, 16:27:56, 11 Jan 2018  https://www.cafepeyote.com/files/Plotkin_Deposition_-_Summary.pdf or https://youtu.be/DFTsd042M3o?t=24959

Here is the context of the quote:

  • QUESTION: Are you familiar with the study called “Maternal Immune Activation Alters Fetal Brain Development through Interleukin-6” [Smith, 2007]?
  • PLOTKIN: Vaguely, yes. Yeah.
  • QUESTION: Published in the Journal of Neuroscience?
  • PLOTKIN: Yeah, well, I don’t remember the journal.
  • QUESTION: Is that one of the journals you consider respectable?
  • PLOTKIN: Yes.
  • QUESTION: And this was out of the University of California Medical Center. This is from California Institute, CalTech. That institution did a number of studies regarding — that group did a number of studies relating to immune activation and neurological disorder, correct?
  • PLOTKIN: Yes.
  • QUESTION: And they found a connection between immune activation and neurological historical disorders, correct?
  • PLOTKIN: Yes.
  • QUESTION: Okay. And one of the study’s findings they had was that immune activation alters fetal brain development through interleukin-6, correct?
  • PLOTKIN: As I said before, IL-6 is an important cytokine. I would point out in relation to immune activation, that immune activation occurs as a result of disease and exposure to a variety of stimuli, not just vaccines.
  • QUESTION: But it can be caused by vaccines, correct?
  • PLOTKIN: Immune activation is the objective of vaccines.

The Claim: Challenges to the immune system can come from wild pathogens or from vaccines

This is confirmed by the quote above from Plotkin, as well as any immunology textbook.

The Claim: Vaccines must activate the immune system strongly in order to work

The Claim: Aluminum adjuvants are used in many vaccines specifically to increase the challenge to the immune system and boost the response

These two claims are basic to how vaccines work, as explained on the websites of the CDC and the CHOP:

 “An adjuvant is an ingredient used in some vaccines that helps create a stronger immune response in people receiving the vaccine. In other words, adjuvants help vaccines work better. Some vaccines that are made from weakened or killed germs contain naturally occurring adjuvants and help the body produce a strong protective immune response. However, most vaccines developed today include just small components of germs, such as their proteins, rather than the entire virus or bacteria. Adjuvants help the body to produce an immune response strong enough to protect the person from the disease he or she is being vaccinated against. Adjuvanted vaccines can cause more local reactions (such as redness, swelling, and pain at the injection site) and more systemic reactions (such as fever, chills and body aches) than non-adjuvanted vaccines.”

CDC https://www.cdc.gov/vaccinesafety/concerns/adjuvants.html

 “Aluminum is used in vaccines as an adjuvant. An adjuvant is a vaccine component that boosts the immune response to the vaccine. Adjuvants allow for lesser quantities of the vaccine and fewer doses. The adjuvant effects of aluminum were discovered in 1926. Aluminum adjuvants are used in vaccines such as hepatitis A, hepatitis B, diphtheria-tetanus-containing vaccines, Haemophilus influenzae type b, and pneumococcal vaccines, but they are not used in the live, viral vaccines, such as measles, mumps, rubella, varicella and rotavirus.”

CHOP https://www.chop.edu/centers-programs/vaccine-education-center/vaccine-ingredients/aluminum