Phillipe Stoop, our Director of Research and Innovation, has published an article on the parliamentary report on health expertise.
An analysis of inter-agency discrepancies
A parliamentary report published on 16 May 2019 studies the operation of agencies responsible for assessing health risks, often criticised in the press for their supposed complacency towards industrial lobbies, particularly in the case of the glyphosate study. This report explains in a very pedagogical way the functioning of these agencies and the evaluation processes, and makes constructive proposals to find “the path of trust”. However, by only making proposals for improving the agencies, without ever questioning the relevance of the criticisms addressed to them, this report risks having the opposite effect to the stated aim: to lead the agencies into an infinite spiral of justifications, in order to answer questions that are not necessarily scientifically founded. This is all the more so because the report does not address an issue that falls entirely within the legislative domain: the need to amend certain regulatory texts, which are inconsistent with the complexity of health risk assessment.
In this first article, we will look at the first two chapters of this report, which explain the foundations of health expertise and the way in which the agencies operate, and analyse the reasons for potential divergences between agencies. In a second article, we will look at Chapter III of the report, which analyses the reasons for citizens’ doubts about the agencies, and makes proposals for returning to the “path of trust” advocated by parliamentarians… and we will see how this report has only come halfway down this path.
OPECST (Office Parlementaire pour l’Evaluation des Choix Scientifiques et Technologique) presented on Thursday, 16 May 2019, a long-awaited report entitled: “ASSESSMENT OF HEALTH AND ENVIRONMENTAL RISKS BY AGENCIES: Finding the Path of Confidence[1] “. This report has often been presented by the press as a report “on the objectivity and independence of health agencies in the examination of glyphosate”, which is simplistic, but not totally false: this was indeed the motivation of the initial referral letter from the President of the European Affairs Commission Sabine Thillaye (LREM), later confirmed by a 2nd referral from the President of the Economic Affairs Commission Roland Lescure (LREM). Subsequently, OPECST rightly extended its mission to include the examination of all health and environmental risk assessment procedures in France and Europe. However, the specific case of glyphosate has of course been the subject of a particular examination, which is justified both by its media coverage and its legal importance: in fact, within the framework of European regulations, the differences in evaluation between the organizations that have expressed their views on this weed killer could lead to opposing decisions as to whether to maintain or ban the use of the product. It is therefore useful to recall its history as a good example of the type of controversy that OPECST would like to avoid in the future…but remembering that this is not the only problem that needs to be addressed in order to restore public confidence in the agencies.
The textbook case of glyphosate
Glyphosate is one of the most widely used herbicides in the world and has always been considered of low concern (and, in particular, non-carcinogenic) by the health agencies responsible for assessing the health and environmental risks of pesticides. In 2015, the IARC (International Agency for Research on Cancer, a research organisation under the World Health Organisation) therefore created a great stir by classifying it as a “Probable Carcinogen” [2].
The IARC classification has a somewhat special status in the field of risk assessment: it is the world’s authority in identifying carcinogenic substances, whether they are regulated artificial products or natural substances (to give two often cited examples in comparison with glyphosate, it has also classified red meat as a probable carcinogen, and charcuterie as a carcinogen). On the other hand, it makes a statement only on the danger caused by these substances (capacity to cause cancer, regardless of the dose at which this effect occurs), and not on the risk (capacity to cause cancer, at the doses to which the users or consumers of the substance are actually exposed). Moreover, it has no mandatory regulatory value, as each national health agency carries out its own hazard and risk assessment for the authorisation of regulated products (or delegates it to other agencies, in the case of European agencies collaborating with each other).
Following IARC’s opinion, the national health agencies have of course reviewed its monograph on glyphosate, and none of them have confirmed its carcinogenicity, even in terms of hazard.
In the European regulatory context, this divergence between IARC and the other agencies has particularly important legal implications. Indeed, in general, the authorisation of regulated chemicals is based on the demonstration of the absence of a risk, and not of a hazard. On the other hand, in the case of pesticides, European Regulation 1107/2009 [3] prohibits the authorisation of a product presenting a proven or probable danger of carcinogenesis or endocrine disruption, even if there is no demonstrated risk. However, this regulation refers to the European hazard classification defined by ECHA (European Chemical Agency), and not to that of IARC.
Some commentators suggested that IARC’s advice was in fact consistent with that of the health agencies, on the grounds that IARC would decide on the danger, and the health agencies on the risk. This is of course not true in the case of the European agencies: the mere fact that they have maintained the authorisation of glyphosate to date shows that they do not consider it to be a probable carcinogen, unless they have forgotten the existence of Regulation 1107/2009. In France, the situation was even clearer, since ANSES had issued an opinion in 2016 in which it confirmed that it considered that there was no risk of carcinogenesis for glyphosate and that it would follow ECHA’s opinion as regards the hazard [4]. As ECHA subsequently confirmed that it did not consider glyphosate to be carcinogenic (in terms of hazard) [5], the matter should be closed if the legal arguments are strictly adhered to: glyphosate should not be banned for health reasons according to French regulations, as in the rest of Europe. However, it is of course difficult for citizens to understand why glyphosate could at the same time be classified by a world-renowned organisation as justifying a ban, if it had been issued by a European agency. Moreover, the IARC opinion is based, among other things, on epidemiological studies carried out on farmers, i.e. under conditions of real exposure of users. Even if it does not officially pronounce on the risk, its opinion therefore necessarily implies that it considers that there is indeed a probable risk to users. Despite attempts to gloss over the goat and the cabbage, it is clear that the opinions of the IARC on the one hand, and of other agencies on the other, are indeed irreconcilable: there is therefore a contradiction before which legislators should take a stand, in the most transparent and comprehensible way possible for citizens.
In this controversial context, the initial referral letter to OPECST was therefore surprising, to say the least, since it only asked for a review of the functioning of the European agencies, which are unanimously in favour of glyphosate, without including the review of the IARC (under the responsibility of the WHO and not of Europe, it was not covered by the initial referral). It was therefore judicious that OPECST decided to broaden the scope of its mission…but we shall see that this broadening did not really go in the direction of clarifying the current controversies.
The OPECST report: a clear and exhaustive inventory… within the limits it set itself
As usual, OPECST delivered a very comprehensive and clear analysis of its subject matter, despite its complexity. One can at most regret the recurrent use of language elements that guide the reading in a more or less discreet way: the term “lobby” is as usual only used to describe the influence strategies of industrialists, never those of political parties and NGOs mobilized against pesticides. Similarly, the expression “manufactures doubt” (or ignorance) comes up several times, always to denounce the practices of economic actors. This notion of “factory of doubt”, which is the subject of very official social science works [6], could however have been profitably mobilized on this subject, to examine the way in which the opinions of health agencies are continually questioned by their opponents, who are not the industrialists!
Chapter I, devoted to the institutional framework of risk assessment, presents a very pedagogical and objective presentation of the problem of implementing the precautionary principle, followed by a presentation of the attributions and functioning of the (numerous) organizations involved in this assessment in France and Europe: EFSA, ECHA, ANSES, INRS, INERIS. These chapters 1A. and 1B. are so well written that they only regret their most surprising shortcoming: the IARC, although an essential part of the ongoing controversy, is not studied in them. It was not until Chapter 1C, on the international framework for evaluation, that its name finally appeared. But, at this stage, OPECST simply explains its method of classifying carcinogenic substances and the differences with the European classification. Unlike the European agencies, OPECST has not attempted to describe its mode of operation, although its particularities undoubtedly play a major role in its recurring divergences with the other agencies: while the latter are specialized in evaluation, IARC is above all, as its name indicates, a research organization, which also has an evaluation mission. However, as we shall come back to in the following, research has objectives and a scientific culture that are very different from those of health expertise [7]. Moreover, since it is not involved in pesticide marketing authorizations, the IARC does not have access to the regulatory data submitted for the registration of plant protection products, which deprives it of much of the data useful for evaluation that other agencies have at their disposal.
Inter-agency discrepancy: a very benevolent analysis for IARC
At the end of this first chapter, OPECST comes to the study of the case of glyphosate, and analyses the causes of the discrepancies between IARC and the other agencies (in particular EFSA, which ultimately synthesises the information provided by the other European agencies). This part is still as clear and synthetic as ever, but this time there are some somewhat surprising approximations:
- On page 70, the report states that “IARC relies on the results of the Agricultural Health Study (AHS) which has been tracking the health status of 90,000 farmers and their spouses in Iowa and North Carolina since 1993”. It is difficult to see the point of highlighting the AHS as a key reference for IARC, when other agencies have of course also included its publications in their studies… but have not drawn the same conclusions as IARC.
- On page 74, the report cites the famous “Monsanto Papers” as arguments that can be used against the health agencies, without mentioning that these “Papers” do not reveal any irregularity concerning the publications used by the agencies (in the latter, the possible involvement of Monsanto employees or its providers was well mentioned, in accordance with good practice concerning conflict of interest declarations [8]). The disputed practices cited concern only non-regulatory communication. These “Papers” therefore only call into question Monsanto’s communication practices, and not the work of the agencies, despite the amalgamation that was made by a large part of the press in the wake of the glossary produced by Le Monde.
- Finally, the chapter closes with a summary of the methodological differences that may explain differences of interpretation between IARC and EFSA. This inventory is intended to be exhaustive and neutral, … but, by the same token, tends to water down the problem of these differences of appreciation, which OPECST nevertheless points out are recurrent. The accumulation of small methodological differences of little practical importance gives the impression that the differences between IARC and health agencies concern a host of esoteric technical details, and that in the end both opinions have the same legitimacy. In fact, and particularly in the case of glyphosate, the differences are mostly due to one thing: the level of evidence each side gives to the same data. However, OPECST’s analysis on this subject is surprising: the report begins by stating that, unlike IARC, EFSA (and the other agencies) only takes into account GLP (Good Laboratory Practice) approved laboratory experiments, which is false: like IARC, EFSA takes into account experiments from public research, which generally do not have GLP approval. Further on, he considers that EFSA “attaches greater importance … to studies conducted in accordance with OECD guidelines”. These clearly defined differences in methodology are accompanied by more vague considerations whose scientific justification is difficult to see: “EFSA tends to give less weight to epidemiological studies (cohort and case-control studies) than IARC”, while “IARC gives more importance to statistically weak signs”.
In the end, OPECST concludes that “EFSA appears to require stronger evidence than IARC to conclude that there is a significant link between cancer and glyphosate, which may explain two assessments with different conclusions”. This cautious neutrality gives the impression that EFSA (and the other agencies with them) is certainly more rigorous than IARC, but at the same time risks allowing dangers (that are difficult to detect) to escape. As with the Monsanto Papers, we find here a discreet echo of the strange theories formulated in Le Monde by Stéphane Foucart, according to whom “In matters of public health, scientific rigorism is a dangerous posture” [9]. Non-insider readers may understand that too meticulous compliance with GLP and OECD guidelines may hinder the detection of the famous “weak signals”, which OPECST has repeatedly reminded us of the importance of detecting them early. This is not at all the case, quite the contrary. In scientific terms, the ability to detect weak signals is determined by the statistical power of the experimental protocol used. The OECD guidelines are intended (among other things) to ensure good statistical power in regulated experiments. GLP, on the other hand, simply relates to the quality of the methods used, and the transparency and traceability of the experimental work carried out; they therefore provide no direct guarantee of the statistical power of the experimental protocol. However, they do guarantee at least a good quality of the work carried out according to the protocol, which limits the risk that the published result will be “noisy” due to deficient or even fraudulent methods. It is therefore surprising that OPECST seems to consider compliance with these two safeguards as optional, and it is in any case wrong to think that breaking them reduces the risk of ignoring a real danger.
On this subject, we would also like to know what scientific justification there is for the assertion that “IARC gives more importance to weak statistical signals”: in scientific terms, this would be tantamount to favouring experiments with the lowest risk of 2nd species [10] (i.e. the highest statistical power). However, the IARC monographs do not cite any information on the risk of 2nd species in the references they have retained (nor do the health agencies, but at least this “weak signals” argument has not been used in their favour). Moreover, it is enough to look at the history of the divergences between the IARC and the EFSA to see that this theory of a greater sensitivity of the IARC to weak signals does not stand up to the facts: OPECST itself reminds us on page 78 that, out of the 25 molecules for which the EFSA and IARC classifications diverge, it is most often the EFSA which proposed the strictest classification (in 14 cases against 11).
This subject is all the more important as compliance with the OECD guidelines was central to the controversy over glyphosate: ECHA recalled that the animal experiments invoked by the IARC did not comply with these recommendations: the carcinogenic effects retained by the IARC were observed at concentrations higher than the BAT (Maximum Tolerable Dose). This means that animals were exposed to such massive doses of glyphosate that its short- and long-term (yet low) toxicity, or even simply its repellent effect in food, make it impossible to obtain reliable comparisons between exposed animals and the control population [11]. This argument, which is difficult to refute, is in itself sufficient to prevent classification as a probable carcinogen: in the absence of these non-compliant publications, IARC would no longer have the “sufficient evidence on the animal” required for classification as a probable carcinogen. This is in addition to the fact that the epidemiological results retained by IARC on the link between glyphosate exposure and non-Hodgkin’s lymphoma (NHL) are also very fragile. These studies are in the minority compared to those that found no link. Moreover, they also found links between NHL and the other pesticides studied, and therefore do not specifically incriminate glyphosate [12].
Of course, it is normal that OPECST did not seek to decide between the agencies’ opinions, which would be an interference in the work of the scientists. But the legislator could at least decide on the minimum level of quality of scientific work worthy of consideration in their studies. However, not only does the report not give an opinion on this subject, but it even seems to open new floodgates, since OPECST has no objection to the fact that IARC has based its classification on animal experiments that do not comply with OECD guidelines. If compliance with these guidelines is optional, does OPECST consider any experimentation admissible? For example, should the agencies have retained G. E. Séralini’s work on glyphosate, which they all rejected (including IARC)? Clarification on this delicate issue seems necessary to avoid future discussions on the criteria for the elimination of publications reviewed by the agencies. Similarly, one might at least have expected OPECST, noting that divergences between IARC and EFSA are frequent, to ask for a commitment from both agencies in this case to produce a joint document, explaining the reasons for their divergences, and the additional data that would need to be collected in order to arrive at a common opinion. However, as we shall see in our second article, OPECST makes no such proposal.
Last but not least, this first part of the report lacks one obvious element: the analysis of public research practices, which produces the publications used by the agencies when reviewing products already authorised, as in the case of glyphosate. Much more than the discrepancies between agencies, which in the end remain fairly rare, it is the announcements of scientific work, which seem to demonstrate health effects ignored by the agencies, that raise doubts among citizens. In the absence of a study of this component of the controversies, the OPECST report seems to consider that only the functioning of the agencies needs to be improved, and only makes proposals along these lines. However, as we shall see in the second part of this article, the functioning of research in this field also deserves to be examined, and the major challenge is indeed to improve understanding between agencies and researchers, even more than between agencies and citizens!
NB: this article presents the point of view of its author, and not that of the Académie d’Agriculture de France.
[1] www2.assemblee-nationale.fr/content/download/80472/893594/version/1/file/OPECST_Rapport_provisoire_agences_comp.pdf
[2] http://monographs.iarc.fr/ENG/Monographs/vol112/mono112-10.pdf
[3] https://eur-lex.europa.eu/legal-content/FR/TXT/PDF/?uri=CELEX:32009R1107&from=EN
[4] https://www.anses.fr/fr/content/avis-de-l’anses-sur-le-caractère-cancérogène-pour-l’homme-du-glyphosate
[5] https://echa.europa.eu/fr/-/glyphosate-not-classified-as-a-carcinogen-by-echa
[6] http://www.inra.fr/Chercheurs-etudiants/Economie-et-sciences-sociales/Toutes-les-actualites/Science-et-territoires-de-l-ignorance
[7] http://www.forumphyto.fr/2016/10/04/peche-aux-alphas-contre-chasse-aux-petits-betas-pourquoi-lanalyse-des-risques-environnementaux-ne-devrait-pas-etre-seulement-un-travail-de-chercheurs/
[8] http://www.forumphyto.fr/2017/10/16/glyphosate-que-disent-vraiment-les-monsanto-papers/
[9] https://www.lemonde.fr/idees/article/2018/10/27/en-matiere-de-sante-publique-le-rigorisme-scientifique-est-une-posture-dangereuse_5375460_3232.html
[10] In applied statistics to the evaluation of health risks, the risk of 2nd species amounts to not detecting a real, but low risk (as opposed to the risk of 1st species, which is to attribute to the product under study an effect that is a simple statistical hazard).
[11] https://echa.europa.eu/documents/10162/e847df76-1549-c107-0a55-c9df29caacc7
[12] http://www.forumphyto.fr/2017/11/20/glyphosate-linsoutenable-legerete-du-circ/
Predict and Decide