Autistic Brains Have High Aluminum Levels vaccinepapers.org

“I did not see a role for aluminum in autism. And I didn’t see a role for aluminum in vaccines in autism. I have to change my mind now on both of these. I have to change my mind that aluminum has a role in autism. I believe it now does.”
–Dr Chris Exley, 2017, on the impact of the Mold 2017 study results. 

Accumulating evidence implicates aluminum adjuvants as a cause of autism. Autism is caused by inflammation in the brain during early development (gestation and first few years of life). The specific inflammation signals (i.e. cytokines) that cause autism are interleukin-6 (IL-6) and interleukin-17a (IL-17a). Aluminum adjuvants stimulate this specific type of inflammation in the brain.

There have been no epidemiological studies of Al adjuvant exposure and autism, except for one ecological study (Tomljenovic 2011). Ecological studies analyze aggregate, population-level data (e.g. comparing disease prevalence and exposures in a region or country). Ecological studies do not utilize data on individuals. For this reason, ecological studies should be used for generating hypotheses for further research. They are not worth much and should not be used as evidence of causation.

If Al adjuvant is responsible for much of the rise in autism, then autistics should have elevated brain aluminum levels. A new study (Mold et al 2017) by Dr Chris Exley’s research group provides the first-ever measurements of aluminum in autistic brains. In this article I show that the aluminum levels in autistic brains are abnormally high. Brain aluminum concentration is highly nonuniform, whether the brain is healthy or diseased. Adjacent brain regions can have dramatically different Al concentrations. Consequently, many measurements and statistical methods are necessary to determine if aluminum is elevated or normal.

A prior study (Mohamed 2015) measured aluminum levels in hair and found aluminum to be about 5X higher in hair from autistics compared to matched controls. This difference was highly significant (P<0.0001). Of course, this is merely a correlation and accordingly it cannot prove causation. Also, it is not known if Al hair content is an indicator of brain or body content of aluminum. Further, the hair Al could come from sources other than vaccines. We don’t know where it comes from.
Full paper (Mohamed 2015): Assessment of Hair Aluminum, Lead, and Mercury in a Sample of Autistic Egyptian Children

The First-Ever Measurements of Al Levels in Autistic Brains
The study was performed by Dr Exley’s research group at the University of Keele. Dr Exley has studied aluminum toxicity for decades. Dr Exley’s research group has probably performed more measurements of brain Al content than any other group in the world. The Al was measured according to a rigorous, well-established protocol. For example, blank (Al-free) samples are processed alongside brain samples to detect and correct for contamination or systematic errors. The average Al level in the blanks is subtracted from sample levels (on a per-batch basis). This is why some Al measurements have slightly negative values in the House 2012 study (described below).
Full Paper (Mold 2017): Aluminium in brain tissue in autism

In Mold 2017, 12 1-gram samples from each of 5 autistic brains were analyzed for aluminum. 4 measurements were not obtained (apparently due to process errors or sample problems), yielding 56 aluminum measurements. Specifically, a 1-gram brain sample came from each of 4 “lobes” of the brain (the occipital, frontal, temporal and parietal lobes). Each 1-gram brain sample was divided into 3 parts that were separately analyzed for aluminum. The study was limited to 5 brains because only 5 autistic brains were available for research. Aluminum content was reported as micrograms aluminum per gram dry weight brain tissue (mcg/g). Also, aluminum was measured in the hippocampus (part of the brain) for 2 subjects, for a total of 58 measurements.

The study was not controlled. No comparisons were made to non-autistic brain samples. The lack of control data was not due to oversight or neglect. Rather, healthy matched control brains were not available. Dr Exley stated the following about the lack of controls:

“We discussed control tissues but the only available were not age-matched and nor were they true controls as the donors were individuals in their 40s and 50s who died of a certain disease or condition. No age-matched healthy donor brain tissues were available. However, we have more data on the Al content of human brain tissue than anyone else and so we are in a position to compare our autism data with other data relating to almost 100 human brains. This is how we can come to our judgement that the values measured were some of the highest values ever measured in any individuals.”
-Dr Chris Exley on the Mold 2017 aluminum measurements

The Aluminum Measurements
The table below shows the Al concentrations measured in the autistic brains.  Dr Exley and other scientists consider  Al concentrations exceeding 3.0 mcg/g  to be “of pathological significance” and likely to increase risk of disease over time. Aluminum levels exceeding 3.0 mcg/g are in red. Brain locations are indicated by O, F, T, and P, which refer to the occipital, frontal, temporal and parietal lobes.

Above: Aluminum levels in autistic brain tissue samples. 16 of 56 samples (28.6%) exceeded 3.0 mcg/g aluminum per dry weight brain tissue (shown in red). Aluminum exceeding  3.0 mcg/g is considered to be “of pathological significance”. The Al levels are highly variable, suggesting that the Al distribution is very nonuniform and concentrated in specific locations. The 2 measurements from hippocampus samples are not shown. From Mold 2017. 

In the opinion of Dr Exley and other specialists in the field,  the Al levels in the autistic brain samples are alarmingly high. Their extensive experience with aluminum measurements in human brains should be sufficient reason to take the results seriously.  Critics and skeptics reject this opinion about the Mold 2017 results, regardless of the experience of the researchers. Thats fair enough I suppose. A quantitative comparison is necessary to prove that the autistic brains have elevated Al levels.

Notice that the Al levels vary enormously, even in the same brain, and even between the 3 parts of a single 1-gram sample. For example, the 50/M subject had 18.57, 0.01, and 0.64 mcg/g in the 3 adjacent parts of a 1-gram sample. It looks like the Al is localized in small regions, and not elevated in most places. Because the Al levels vary so much, many measurements are necessary for a meaningful comparison with a control group.

The critical question is whether the aluminum levels in the autistic brains are unusually high compared to normal/healthy brains. But since we do not have comparable Al measurements from normal/healthy brain samples, a different strategy is needed.

Comparison With Non-Matched “Positive Controls”
House 2012 is a paper by Exleys group reporting Al measurements of 713 tissue samples from 60 brains. The 60 subjects were elderly (age 70-103 years), and 39 of the 60 subjects (65%) had Alzheimers disease or dementia. The 60 subjects in House 2012 are not at all matched (e.g. by age) with the autistic subjects in Mold 2017. House 2012 and Mold 2017 Al measurements were performed in exactly the same way. Specifically, 1-gram samples were divided into 3 parts, and each part was analyzed (by graphite furnace atomic absorption spectroscopy for the nerds out there). Blank samples were processed in parallel and used to detect and correct for Al contamination.  (Citation for 65% demented rate: Exley 2012, https://www.ncbi.nlm.nih.gov/pubmed/22699848).
Full Paper (House et al 2012) Aluminium, iron and copper in human brain tissues donated to the medical research council’s cognitive function and ageing study

It is acceptable and meaningful to use the elderly subjects in House 2012 as controls. There are two reasons for this:

1) Al brain levels increase with age and

2) Al brain levels are elevated in Alzheimers disease.

The 60 elderly (and mostly demented) subjects therefore must have higher Al brain levels than healthy, younger people. Using the 60 elderly subjects as controls is therefore conservative in that it is biased towards finding lower Al in the autistic subjects. Accordingly, a finding of higher Al in the autistic brains will have heightened validity. In other words, if the autistic brains have elevated Al compared to the mostly demented elderly brains, then we can be extra-confident that the autistic brains have elevated Al levels.

The elderly subjects of House 2012 are a “positive control” group. The use of positive controls is well known and accepted. For example, this paper on experiment design (Citation: https://academic.oup.com/ilarjournal/article/43/4/202/981687) describes the use of positive controls. It states: “The positive control acts as a standard against which to measure difference in severity among experimental groups.” Though neither Mold 2017 nor House 2012 are experimental studies (they are observational), the concept is the same.

Comparing Frequency Of High Al Levels
The table below shows the Al measurements in the 60 elderly brains from House 2012. 39/60 (65%) had dementia or Alzheimers, but we do not know which subjects were demented (the brain sample supplier refused to disclose this information to Dr Exley). Al levels exceeding 3.0mcg/g are outlined in red.

Above: Aluminum levels in brain samples from 60 elderly control subjects, aged 70-103 years. 39 of the 60 had dementia or Alzheimers disease. 97 of the 713 measurements (13.6%) exceeded 3.0 mcg/g aluminum. By comparison, 28.6% of the autistic brain measurements exceeded 3.0mcg/g. The difference is statistically significant (P=0.005).  Adapted from House et al 2012. 

Al levels exceeding 3.0 mcg/g were more than twice as frequent in autistic brain samples compared to the elderly controls (28.6% vs 13.6%). But is this difference statistically significant, or is it possibly due to random variations? This question is answered by “Fishers exact test” (two-tailed). Fishers exact test is the most appropriate and conservative way to calculate a P-value for the comparison. I used an online calculator for Fishers exact test (https://www.graphpad.com/quickcalcs/contingency1.cfm). The P-value is 0.005, indicating high statistical significance according to the conventionally-accepted threshold of P=0.05 (lower P-values indicate greater significance). The P=0.005 means that the difference between the autistic and elderly controls is very unlikely to be due to chance. In other words, we can be confident that the autistic brains really do have high frequency of high Al levels (>3.0 mcg/g).

Above: Mold 2017 and House 2012 Al brain measurements are compared using a threshold value of 3.0 mcg/g . Analysis by Fishers exact test (two-tailed) gives a P-value of  0.005, which indicates high statistical significance. We can confidently conclude that autistic brains have greater frequency of high Al levels compared to elderly and mostly demented brains. 

Could Results Be Explained by Male-Female Balance?
Notice that the female autistic subject had only one elevated aluminum measurement. This suggests the possibility that male brains tend to accumulate more aluminum than female brains. Of the 60 House 2012 subjects, only 22 were male and 38 were female. So, could the higher Al among the autistics be explained merely by the fact that 4/5 autistics were male? To answer this question, I repeated the analysis for male subjects only. For the male autistics in Mold 2017, 15 of 45 measurements (33%) exceeded 3.0 mcg/g. For the male subjects in House 2012, 38 of 260 measurements (14.6%) exceeded 3.0 mcg/g. The difference was statistically significant with P=0.0047. So, sex differences in aluminum accumulation do not explain the results of Mold 2017.

Above: Limiting the comparison to only male subjects produces the same result: autistics have greater frequency of high Al levels (33%) compared to elderly and mostly demented subjects (14.6%). P=0.0047.

A Caveat: Measurement Independence
The above analysis requires an assumption: that all the Al measurements are independent of one another. However, each brain contributed 12 measurements, and so the set of 12 measurements for each brain may be correlated and therefore not completely independent. Most of the measurements are independent from one another because most come from different brains. Non-independence will erroneously decrease the P-value, and falsely make the result more statistically significant that it really is. There are ways to avoid or correct for this problem, but that is beyond the scope of this article. I looked at how the correction is performed, and concluded that the effect on the P-value will be small. I mention this issue so the critics are not the first to bring it up!

Comments
The threshold of 3.0 mcg/g was not cherry picked or contrived to obtain a statistically significant result. I also analyzed the data with a threshold value of 3.5 mcg/g (an old threshold of pathological concern), and obtained a P-value of 0.015. These are the only two threshold values I explored.

I excluded the 2 Al measurements from the hippocampus in Mold 2017 (0.02 and 1.42 mcg/g), because none of the House 2012 Al measurements were from the hippocampus. Including the hippocampus measurements has an inconsequential effect, changing the P value to 0.0066.

Analyzing the data “dichotomously” by applying a threshold test (> or < than 3.0mcg/g) is a simple but inferior analysis method because it does not utilize all the information in the data. Notice for example that there would be no difference detected if the threshold was 1mcg/g or 40mcg/g. A more rigorous approach is to analyze the Al measurements as a continuous variable. This is beyond the scope of this article. I invite anyone interested to perform this analysis. .

As noted above, the elderly, demented controls bias the results towards a finding of no elevated aluminum in the autistics. Therefore, a comparison with healthy, age-matched controls would almost certainly show an even greater excess of elevated Al levels.

Critics argue that the Mold 2017 study is worthless because it lacks Al measurements from controls. While it is reasonable to criticize the absence of controls, it is not reasonable to say that the data is worthless or cannot be used to support opinions from researchers with many years experience in measuring Al in brain samples. Dr Exley and other experienced scientists are firm that the autistic brains had alarmingly high levels of Al. It is not reasonable for critics to dismiss their opinions, especially if the critics (e.g. Dr Gorski of Respectful Insolence) have no experience or expertise in aluminum toxicity or measurement. However, I wish that the study authors had included a comparison to something.

Readers may be wondering why the Mold 2017 paper did not use the 60 elderly subjects as controls, as I have done here. Dr Exley did not want to use controls that were not matched to the autistic subjects, because doing so could invite criticism of this choice of controls. This is the primary reason why non-matched controls were not included in the paper. In my opinion, comparing with non-matched controls is a lot better than no comparison at all.

There are two possible concerns about using the elderly subjects as controls:
1) The House 2012 Al measurements were performed at a different time (presumably about 5 years prior), as part of a different research project. Although the Al measurement procedures in the two studies are nominally identical, there may be subtle or systematic differences (e.g. slight variations in equipment calibrations or chemical reagents) that could bias the results, since they were done years apart. The proper way to generate control data is to analyze the control and autistic samples at the same time.  This is the best way to reduce the risk of errors or bias.
2) The control and autistic subjects were not age-matched. Although it is known that Al brain content increases with age, the age difference reduces the value of the comparison because it does not allow a calculation of exactly how elevated the Al is in the autistic brains. We can be confident that Al levels are elevated, but we cannot say how elevated they are.

The 2 issues above might be concerning to a peer reviewed journal. However, I believe that using the 60 elderly subjects as positive controls is informative and provides powerful evidence that autistic brains have an abnormally high frequency of high-level Al deposits.

Fluorescence Microscopy and Cell Identification
Mold 2017 reported that the aluminum was concentrated within immune system cells associated with capillary blood vessels. This is a highly provocative finding because, if true, it suggests that the observed Al is from aluminum adjuvant, and that the Al adjuvant was transported to the brain by the immune system cells (e.g. macrophages). This “trojan horse” mechanism of Al adjuvant particle transport is explained here: Vaccine Aluminum Travels Into The Brain. 

However, critics note that visual identification of cell type is subjective and susceptible to confirmation bias. The researchers may be “seeing what they want to see”. This is a reasonable objection and accordingly the cell identification must be considered to be preliminary, even though the researchers are well experienced in this task. Unequivocal identification of cell types requires the use of selective stains, which were not employed in Mold 2017. An obvious next step is to definitively identify the types of cells containing the aluminum with cell-selective stains.

Dr Exley believes the intracellular location of the aluminum is a particularly important finding, because it has not been observed in dementia, Alzheimers or other brain diseases associated with aluminum. In those diseases, the aluminum is present in the space outside of cells, and not within immune cells. Dr Exley states:

Rebutting Orac (Dr Gorski)
Well known vaccine promoter and blogger Dr David Gorski criticized the Mold 2017 study: https://respectfulinsolence.com/2017/11/29/christopher-exley-using-bad-science-to-demonize-aluminum-adjuvants-in-vaccines/

Orac is lazy and bombastic, a terrible combination. His writing is filled with ad hominem attacks, name-calling and snark. His scientific arguments are superficial and inadequately cited. He seems to not read the scientific literature or to understand the context of research he writes about. His criticism of Mold 2017 fits this pattern. Nevertheless, Orac occasionally makes points deserving of discussion or rebuttal, and rarely he says something correct and insightful. Below I respond to Orac’s statements about Mold 2017:

Orac: “There is really no good evidence that aluminum in vaccines causes ASDs, try as as antivaxers might to try to “prove” that there is.“

The evidence linking autism and aluminum adjuvants is powerful and growing. The scientific case is based on the fact that autism is caused by neuroinflammation during early development, and that Al adjuvants travel into the brain and cause the specific type of neuroinflammation shown to cause autism. The scientific case is described in this article, which has 97 citations: http://vaccinepapers.org/wp-content/uploads/Autism-and-aluminum-adjuvants-in-vaccines.pdf

In fact, there is no good evidence that Al adjuvants are safe, do not cause autism or other neurological/psychiatric disorders. Long term health outcomes have never been studied, especially for exposure resulting from the full vaccine schedule. In 2015, CDC researcher DeStefano (in collaboration with others), explored the feasibility of studying Al adjuvant safety. They searched the literature and concluded Al adjuvant has not been studied by epidemiology: “…no large epidemiological studies have specifically examined associations between health outcomes and vaccine ingredients, other than thimerosal.” Citation: https://www.ncbi.nlm.nih.gov/pubmed/26518400

There are no epidemiological studies of autism and Al adjuvants (or Al-containing vaccines) that show safety. There are a couple that shown harm, though these studies have limitations or flaws (which is why I do not cite them). Only MMR has been much studied in relation to autism, and MMR does not contain Al adjuvant. So the MMR-autism studies are obviously irrelevant to Al adjuvant safety. Despite this fact, vaccine promoters describe the MMR-autism studies as relevant to ALL vaccines. For example, vaccine promoters often make claims about vaccines in general (e.g. “vaccines do not cause autism”), and then cite studies limited to MMR. This is a logical error because studies of MMR cannot be cited as evidence for the safety of Al-containing vaccines.

Orac: “I could see a number of problems with the methods used. First, there are no controls. Certainly, if Exley were accessing the Oxford Brain Bank to obtain tissue from autistic people, then why couldn’t he have also obtained brain tissue from age-matched normal controls? If the bank doesn’t have such tissue, then it should have been justified.”

I agree with Orac that the lack of controls is a problem. The lack of controls or comparison with other Al measurements weakens the paper. Thats why I did the above analysis. Mold 2017 should have at least explained that controls were sought but not available. Also, Mold 2017 is an important study even though it did not include controls. Mold 2017 provides the first-ever measurements of aluminum levels in autistic brains. The comparison with the House 2012 data shows that autistic brains have a high frequency of elevated Al compared to elderly and mostly demented positive controls.

Orac: “Exley then cites a previous study of his, which he refers to as “our 60 brain study,” that has “allowed us to define loose categories of brain aluminium content beginning with ≤1.00 μg/g dry wt. as pathologically benign (as opposed to ‘normal’).” I looked up the study, reference 15, and was puzzled. In it Exley reported measuring aluminum levels in the brain of one man, a 69-year-old who had died of Alzheimer’s disease. Either Exley screwed up the reference, or he hoped that no one noticed that he had referred to a reference that doesn’t describe what he claims it described.“

That was a typographical error in the accepted manuscript (the preliminary version, before final editing). It was corrected in the final paper. The 60-brain study (Ref 15) is House 2012. It’s not reasonable to make a big issue about a typo error (especially in an accepted manuscript, which is not in final form), and it is nasty to use a typo error to speculate about deceptive intent or dishonesty. But such is the nasty personality of Orac.

“I also notice some rather extreme variability in measurements. For instance, the sample replicates for one specimen were 1.71, 1,64, and 17.10 μg/g dry weight. For those not familiar with scientific terminology, “replicate” means repeated measurements performed on the same sample, although in this case the replicates the samples were divided into three, and each portion subjected to the same procedure to measure aluminum. The replicates were then averaged and a standard deviation calculated. There were other replicates as questionable. For example, for another sample the replicates measured 2.44, 1.66, and 22.11 μg/g dry weight. For still another sample, replicates were 0.01, 0.64, and 18.57 μg/g dry weight. Remember, again, these are replicate measurements of the same samples…. these values do not give me a lot of confidence in the reproducibility of Exley’s methodology.”

Orac is correct to notice this problem, but also wrong because the samples were not replicates. Mold 2017 erred by using the word “replicate”, which is defined as “an exact copy”. The 3 measurements for each 1-gram samples were not replicates. The 1-gram samples were not homogenized or blended. Rather, each 1-gram sample was cut into 3 parts of 0.3g each, and the 0.3-gram parts were analyzed separately. The extreme variability reflects actual variation in Al levels, not measurement imprecision. I confirmed with Dr Exley that the samples were not homogenized.

Orac: “In that study (house 2012), he found that the median aluminum content in 713 samples was 1 µg/g dry tissue and that 75% of the measurements were less than 2 µg/g dry tissue. Some things to note. First, these were all brains from elderly people. Second, Exley basically says that 75% of all values were less than 2 µg/g dry tissue.”
AND
“Heck, even compared to Exley’s “60 brain paper” (House 2012) I have a hard time being alarmed by the numbers in this new study.”

Orac is comparing the Mold 2017 data with the House 2012 data, but in a lazy, superficial way. As noted above, the fact that the House 2012 samples were from elderly subjects increases concern about the autistic brain measurements. To be fair, the burden for doing the statistical comparison is not Orac’s. The burden rests with the Mold 2017 authors. The Mold 2017 findings would be better understood by non-experts if some basis for comparison had been included.

Orac: “…if only we had an idea of the aluminum concentration in normal brain tissue. If only…Oh, wait, we do  (Andrasi 2005) . It’s between 1.4 to 2.5 µg/g dry tissue, as measured in a study that compared aluminum concentration in the brains of patients with Alzheimer’s disease. Oops! That’s not very different from the values that Exley found,”

Calculating the average of the Mold 2017 measurements is easy, but Orac didn’t bother. It’s 3.08 mcg/g (including the 2 hippocampus measurements), which of course is higher than aluminum levels in all the Andrasi 2005 controls (see below). Note that the 3.08 mcg/g average is outside the SD range of 4/5 results from Andrasi 2005. So Orac’s own citation contradicts his claim that the Al measurements are “not very different” from Andrasi 2005.

But average Al levels are likely not as important as the distribution and frequency of pathologically high levels exceeding 3.0mcg/g. It’s the pathologically high concentrations that cause inflammation and injury, not Al levels slightly above normal (e.g. in the range of about 1-3 mcg/g).

Also, the Andrasi control subjects (n=3) were ages 55-71, with an average age of 61. The autistic subjects in Mold 2017 had an average age of 32.8 years (subjects were 44, 50, 22, 15 and 33 years). This is important because Al accumulates in the brain with age. The elderly controls in Andrasi should therefore have higher Al levels, but instead it’s the opposite. The young autistic subjects had the higher average Al level. This fact supports a conclusion that autistic brains have abnormally high average Al levels.

Finally, Andrasi 2005 did not use blanks for correcting for Al contamination. Consequently, Andrasi’s measurements may be erroneously elevated because of undetected contamination. This is a serious concern because very low concentrations and quantities of Al are being measured. Dr Exley’s method uses blanks and subtracts contamination levels for each batch of samples (see House 2012 for discussion of this). Dr Exley’s measurements from House 2012 are therefore more reliable than the measurements from Andrasi 2005.

Above: Average Al levels in brains of healthy controls (ages 55-71, median age = 61), and Alzheimers disease (AD) subjects. Orac cited the Al levels in the controls, on left. The average of the Al measurements in Mold 2017 (including hippocampus) is 3.08 mcg/g, which exceeds average Al levels in the Andrasi controls.  From Andrasi et al 2005. 

Conclusion
The Mold 2017 study is a respectable first exploratory investigation. The results add to the compelling evidence that aluminum adjuvants cause autism.

I eagerly await follow-up research looking at the amount, distribution, chemical form and microscopic location (e.g. intracellular or extracellular) of aluminum in autistic brains. Discovering that the Al comprises Al compounds used as adjuvants (AlOH, AlPO4 or AlSO4), or that it resides inside immune cells (e.g. macrophages) would be particularly provocative because such findings would suggest the Al is from vaccines.