Pr. Romain Gherardi often talks about his research in the media. Recently, when asked whether vaccine adjuvants are neurotoxic, he answered (listen here at 8'30''):
"They are neurotoxic in animals, without any doubt (translated from "Chez l'animal, ils sont neurotoxiques. Il n'y a aucune espèce d'hésitation à le dire")."
He must have the best evidence to support such an unambiguous claim! To check that, let's have a look at his last paper: "Non-linear dose-response of aluminum hydroxide adjuvant particles: Selective low dose neurotoxicity", published in "Toxicology". The study deals with mice.
Gherardi is a fierce advocate for the existence of a causal relationship between containing aluminum adjuvants and a clinical condition which he first called "Macrophagic Myofasciitis (MMF)". Experts Regularly review the literature, including Gherardi's studies :
The Global Advisory Committee on Vaccine Safety (GACVS, WHO) hold their first meeting on 14-19 September 1999, and addressed the MMF topic. They still had a look at the question in 2002, 2003, 2004, and released a summary in the form of a Q&A in 2008,
The French "Académie Nationale de Médecine" in 2012,
The French "Haut Conseil de Santé Publique" in 2013,
The French "Académie Nationale de Pharmacie" in 2016.
They conclude that the symptomatology of the clinical condition is vague (chronic fatigue, neurological disorders, ...). The cases are geographically clustered, almost every case being reported in France by Gherardi's team. The case reports are also limited in time: 445 cases reported since 1995 (whereas aluminum adjuvants are used in vaccines for more than 80 years) and only one case has onset since 2012 (the reporting date may differ from the onset date). On the same period, about 160 millions doses of Al-containing vaccines have been administered and about 12 millions doses are currently administered annually in France.
Cases of « MMF » according to the reporting (« cas notifiés » ) and onset dates (« cas survenus »), Excerpt from the report of the Académie Nationale de Pharmacie.
Experts agree on the existence of MMF, but only as an histological lesion (a kind of "vaccine tattoo"). Aluminum may stay localized in the injected tissue and be detected later by muscle biopsy, as was done by Gherardi on his patients. According to Gherardi, in some predisposed individuals, on the long term, aluminum leaves the injection site to reach the brain and exert a neurotoxic effect resulting in the MMF clinical condition. Gherardi's team now speaks of ASIA ("Autoimmune/inflammatory syndrome induced by adjuvants") instead of MMF, maybe in order to avoid confusion with the vaccine lesion. Whatever denomination is used, experts conclude that available studies do not establish a link between the adjuvant and a systemic disease. They are concerned about a possible decrease in vaccine uptake, despite the lack of scientific evidence.
In fact, it is well known that aluminum is neurotoxic, as observed in hemodialysis patients. However, the doses involved are larger than the doses contained in vaccines. But Gherardi has an idea : the relationship between the adjuvant toxicity and the dose would not be linear. Does the mice study support this hypothesis? Let's have a look...
The authors do not declare any conflicts of interests. However, Gherardi has strong connections with the French association E3M. This association regroups patients with the alleged MMF clinical condition and highly convinced that their disease is caused by vaccine adjuvants. This association has financially contributed to Gherardi's studies in the past (80 000 Euros) and has lobbied (by means of huger-strikes) the ANSM (French Agence Nationale de Sécurité du Médicament) to give 150 000 Euros to Gherardi's team. This is, of course, not big money. But the connections between the Professor and the association cannot be denied.
In fact, we do not really care that this link is not disclosed in the paper. First, we already knew about it. Second, it does not make the study a bad study. We have to read before to judge. But this is amusing, when you think about E3M vision concerning conflicts of interests. In this document dated of 2015, E3M denounces the competing interests of experts who disagree with the association. They are said to have "intellectual and financial connections" with pharmaceutical industries. If we were to follow E3M vision, we might say that "the intellectual and financial connections of Gherardi with E3M association is not guarantee of independence when it comes to determining the origin of a disease that is taken for granted by those who finance part of the research".
The authors mention 3 funding sources:
ANSM (the French "Agence Nationale de Sécurité du Médicament"),
CMSRI (Children’s Medical Safety Research Institute),
British Columbia Vancouver University (Luther Allyn Dean Shourds Estate)
CMSRI is part of the antivax "Dwoskin Family Foundation" (they define themselves as "for safer vaccines", of course). Gherardi sits at the scientific advisory board of CMSRI, with Exley et Shaw, also authors of the mice study. CMSRI presents Wakefield as a victim, and not as a fraud. Gherardi and Wakefield were even invited once to talk at the same meeting (held in Jamaica... funny when you think that some complain about pharmaceutical industries paying vacations disguised in professional meetings...) financially supported by NVIC (National Vaccine Information Center) which has strong connections with Dwoskin Foundation !
University British Columbia Vancouver is mentioned because Christopher Shaw, one of the co-authors, works therein (in the opthalmology department!). Shaw is financially supported by the private and antivax fund "Luther Allyn Dean Shourds Estate".
The acceptation time for papers in Toxicology is about 3 months. This is the time needed for the manuscript to be assigned to a member of the editorial board, for this editor to find two reviewers, for the reviewers to read the paper and send their comments to the editor, for the editor to send the comments to the authors, for the authors to answer to the editor, for the editor to take a decision or ask the reviewers to assess the authors' response, ...
On the other hand, some papers take longer or are accepted very quickly, as the mice study (only 18 days!). Such a short time raises question. Was the review thorough? Or was everyone simply available at the good time ?
The study consists in animal experiments. Precautions must be taken in extrapolating results to human. The authors and Gherardi are well aware of the issue.
Groups and doses
The authors indicate they use 4 groups of 10 mice each, the first group being control (0 µg Al/kg) and the others groups receiving different amounts of adjuvant (200, 400 and 800 µg Al/kg, respectively), equivalent to 2, 4 and 8 human doses of Al-containing vaccine.
A few remarks :
The dose conversion from human to mice is not related to the weight, but to the surface body area. This is not a mistake as some might think. The authors have followed experts' recommendations.
Originally, there was a fifth group who received 133 µg Al/kg, equivalent to about 1 human dose of Al-containing vaccine. That group is not mentioned anywhere in the paper submitted in November 2016, but you can find it in an talk given at the "6ème journée d'étude du pôle Santé & Société", at Paris-Est university in March 2016.
Mice receive their total amount of adjuvant in three injections, while humans would receive it in 2, 3 or 8 injections. Put another way, mice receive for each injection the equivalent of 2/3, 4/3 or 8/3 of human doses of Al-containing vaccine. This would be perfectly fine if it was the total amount of Al that matters. But...
... Spoiler alert : The authors' conclusion is that group 200 is more affected by Al than groups 400 and 800, indicating a non linear relationship between the total amount and the effect. They explain that the amount of adjuvant present in one injection determines the size of Al agglomerates and therefore their capture by macrophages and subsequent transport to the brain. In light of this, the protocol which aimed to assess the effect of the total amount of adjuvant was not adapted. Of course, it is easy to say a posteriori. The experiment could be done again following a protocol better suited to assess the effect of the agglomerates' size, as suggested in the end of the paper's abstract.
the time interval between injections may not be equivalent to the human case. However, if it is not the total amount of adjuvant that matters, but the amount in one injection, that remark is likely not relevant.
Behavioural and motor testing
180 days after the last injection, an experimenter (always the same) performs a battery of 8 behavioural or physical tests. It is not specified whether the experimenter was blinded for the identity of the group. If not, observation bias in the results cannot be excluded. At the end of the behavioural tests, mice were sacrificed for further analyses.
We do not describe the details of every test (go and read the paper). All you need to know is that each test is subdivided into multiple subtests. For instance, the first one ("open field") consists of 7 subtests and is devoted to assess motor activity and anxiety. In the picture below, we represent each of the 8 tests by a large rectangle, and each subtest by a small square. In total, there are 36 subtests (for 40 mice, this means 1440 observations performed by the experimenter). Here, to refer to the second subtest of first test, for instance, we will use the notation T1-2.
How do we know that the results of a subtest are significant or not?
1) The first step is to assess if the four group are similar to each other in the overall or not. This is what is behind the statistical gibberish "Anova". For instance, for test T1-1 the p value is below the threshold of 0.05 and indicates that the 4 groups are significantly not similar to each other. For T1-2, p > 0.05 indicates that the four groups are similar to each other, or that the difference is not significant (n.s.). Among the 36 subtests, 30 are not significant.
2) It remains 6 significant subtests (T1-1, T1-3, T1-4, T1-5, T1-7 et T5-2) to examine in more details. Now, it is necessary to compare the groups 2 by 2, to know precisely what makes them different in the overall when using "Anova". Are all three Al groups different from group control? Or only one or two or them?
For T1-1, group 200 is significantly different from group control (it is also different from groups 400 and 800 which are similar to group control). This result is presented in Fig. 1 (a) of the paper. We have denoted the small square in our illustration with (a) to help you to follow. The situation is the same for the three other subtests of T1 denoted by (b, c, d) in Fig. 1 of the paper. Those four subtests (T1-1, T1-4, T1-5, T1-7) are indicative of an effect of the adjuvant received at low dose, but not at middle or high dose.
T1-3 is significant, but is not discussed in the paper. We plot the results in the picture below. It is reasonable to assume that the test is significant in the overall because of the difference between groups 200/400 and group 800. It is not possible to say whether the difference between groups 200/400 and group control is significant or not.
Finally, let's have a very close look at T5-2 (Fig. 1 (e) of the paper). It is significant in the overall, but NOT because of the difference between group 200 and group control. It was significant in the overall because of the difference between group 200 and group 400! This is why the authors mention a "tendency" between group 200 and group control, but not a significant difference.
Why highlight the results of T5-2 and not those of T1-3 ? We cannot speak for the authors. However, we can not help but notice that this “tendency” conveniently adds to the results of T1-1, T1-4, T1-5 et T1-7 in an attempt to make the conclusion that ONLY group 200 is affected look stronger, whereas the results of T1-3 go against the conclusion. The conclusion is not so robust as it seems.
Finally, there are 30 subtests saying “nothing to report”, only 4 subtests saying “beware of the low dose”, and 2 subtests difficult to interpret. We cannot even exclude the possibility that all or part of the 6 significant results out of 36 are statistical artifacts resulting from chance. This is the multiple comparison problem or look-elsewhere effect. It could become even worse when the groups are compared 2 by 2 within a subtest (6 possible comparisons since there are 4 groups). Fortunately, for that step, the authors use the (beware the statistical gibberish) "Bonferroni's test" which is designed to avoid the multiple comparison problem. Too bad that the author do not address the issue of the 36 subtests performed in the first place.
Besides, it seems that some observations performed by the experimenter are not reported in the paper. We already told you about the "missing group 133", but some subtests' results may also be missing for all groups. For instance, in the description of T1 ("open field"), it is said that the "number of rearing" is observed. But there are no such results in Table 1 of the paper. This is strange, especially if you notice that this kind of observations is indeed reported for T2 in the "supplementary data", meaning that there is no reason to not report them for T1.
Last but not least, some results presented in the paper of November 2016 are not consistent with those presented in the talk of March 2016. In the talk, there was significant overall differences for T2-2 ("elevated O-maze, time spent in open arms") and T7-3 ("Tail suspension test, latency to be immobile"). Why did these effects get lost on the way?
Brain analysis : Microglia
At first, it is said that half of the mice (5 from each group) would be sacrificed for histological examination of brain tissues. In the end, we learn that only 3 brains out of 5 have been used. Why? Interestingly, it is specified this time that the examiners are blinded for the identity of the group. They observe the microglial density and the microglial cell body size in four brain regions. So there are 8 subtests, with the inherent risk of « look-elsewhere effect » ! The microglia are macrophages resident in the brain.
The microglial cell body size was similar in all group, for each brain region. The microglial density is significantly different in the overall for the 4 groups for only one brain region (ventral forebrain), involved in anxiety. Group 200 reacts differently from group control, as can be seen on Fig. 2 (c) of the paper. But the authors failed to show the values for groups 400 and 800. We believe that if the conclusion of the study is that the adjuvant is easily transported to the brain when it is at lowest concentration in the syringe, the authors should give the readers the values for the other groups.
In the talk of March 2016, one can find the value for group 800 (and also for the missing group 133, for which the microglial density seems to be similar to the one of group control, by the way). Strangely, in March's talk, the p-value between control and group 200 was 0.056, which is indicative of a tendency strictly speaking, and not of a significant difference. In the published paper, the p-value is said to be <0.05. What happened between March and November?
Brain analysis : Al level
The other half of the mice are sacrificed for a determination of the cerebral Al level. This time, the 4 mice are indeed used. The results are given as µg (microgram) of Al per gram of dried brain tissue. Only one test here. And the results are given for the four groups. So far so good.
Group 200 has significantly more Al in the brain than group control. Groups 400/800 (and also 133, by the way...) do not significantly differ from control.
For 3 of the 10 mice in each group (again, why only 3?), analysis of muscle tissue is performed to assess the presence of inflammation/granuloma (macrophage aggregates). The possible outcomes are defined semi-quantitatively : no inflammation (0 granuloma), inflammation + (1 to 3 small granulomas), ++ (at least 3 small granulomas) or +++ (at least 3 large granulomas). The groups 200, 400, and 800 are tested. Group control is not. Too bad. But it is unlikely to change anything.
The results are presented in a Table 4 of the paper, but no statistical analysis is done. So we do not really know what to conclude. All we can say is that group 200 TEND to be granuloma free (which would be consistent with a previous migration of Al to the brain, as indicated by the analysis of brain Al level) , whereas groups 400 and 800 tend to have granulomas.
The paper ends as it should with a discussion presenting conclusions drawn from the results and possible interpretations to explain them.
The 4 behavioural subtests allowing to conclude to an effect at low dose only on motor activity and anxiety (T1: "open field"), whereas nothing happens at middle and high doses, are highlighted, as well as the subtest with the "non significant tendency" regarding muscular strength (T5: "grip strength"). The subtest demonstrating and effect for both low and middle doses is consigned to oblivion.
Apart from this, we think that the results of the different tests are not particularly consistent with each other. If the "grip strength" test demonstrates something regarding muscular strength, why is it not the case for the "wire mesh hang" test? If the "open field" test demonstrates something about anxiety and/or motor activity, why is it not the case for the "o-maze " and/or "rotarod" tests?
To move on, let's admit for a moment that an effect arises at low dose only regarding motor activity, anxiety, and muscular strength. The authors claim that this is consistent with :
The microglial density cell in the "ventral forebrain". But due to the lack of presented results for middle and high doses, we must believe them on their word. Besides, this brain region is said to be related to anxiety, which is indeed assessed in the open field test. But once again, why the o-maze test, also assessing anxiety, do not show anything? Moreover, what would be the link between an effect on muscular strength and motor activity and the ventral forebrain, whereas there is nothing significant to observe in the motor cortex? When we look closer, it does not seem so consistent.
The analysis of Al brain level (there is Al in the brain in group 200, but little if any in groups 400 and 800) , which is itself consistent with the lack (presence) of granuloma in group 200 (groups 400 and 800).
The discussion that brings possible explanations for the transport mechanism of Al from the injection site to the brain, only for the low dose, in interesting. But, for us, the question of the neurotoxicity (that is to say, the effect on behaviour, motor activity... ) remains. To say that the adjuvant is neurotoxic seems premature to us in the light of behavioural tests which are less conclusive than it seems and their lack of consistency with the microglia analysis.
Our conclusions ?
Missing fifth group supposed to be equivalent to about 1 human dose of Al-containing vaccine.
Protocol devoted to study the effect of the cumulated amount of adjuvant received, whereas the authors' concllusions are about the effect of the amount received in a single dose.
Possible observation bias in the behavioural tests.
Only 6 out of the 36 subtests performed on the 40 mice demonstrate an overall significant difference between the four groups. Possibility of statistical artifacts.
Only 4 out of the 6 subtests are consistent with the author's claim for a significant effect only at the low dose.
The fifth subtest is highlighted by the authors because of a "statistical tendency" and not because of a significant difference.
The sixth subtest test go against the conclusion, and is conveniently not discussed by the authors.
Some behavioural test results may be omitted in the paper.
The results for groups 400 and 800 on microglial cell density in the ventral forebrain are not communicated.
Inconsistencies between some published results (behavioural tests, microglial cell analysis) and results presented a few month before in a talk.
Only 3 mice out of 5 (or 10) are used for microglial cell analysis (or muscle analysis).
Inconsistency between the brain region affected by the adjuvant (ventral forebrain, related to anxiety) and the behavioural tests demonstrating varied effects regarding anxiety, motor activity and muscular strength.
So, no reason to go crazy about the study. In place of the authors' recommendation to reevaluate the adjuvant safety, or the request by the association E3M to remove it from vaccines, it would be just as good to recommend the use of high doses of Al-adjuvant in vaccines.
But whatever, the media campaign continues. Gherardi sells a book ("Toxic story: Two or three embarrassing truths about the vaccine adjuvants/Toxic Story: Deux ou trois vérités embarrassantes sur les adjuvants des vaccins"), is invited on TV, radio, ... His collaborator (Guillemette Crépeaux), leading author of the mice study, even tries to make you believe that the French drug agency (ANSM) is hiding reports about Gherardi's team work (whereas they have funded their study, and whereas the study are published in international journals? Please...). ANSM has reacted quickly and strongly.
In the meantime, the fears of French parents grow with possible consequences on vaccine uptakes... which does not seem to be a big concern for the association E3M or Gherardi.
A French version of the text is available here.