II. Our perspective and framing assumptions

Before advancing our critique, several preliminaries should be made explicit. This paper is predicated on key well-grounded assumptions about the characteristics of the issues under discussion. The topic is located at the intersection of science and policymaking. Whenever expert advisory bodies provide “risk assessments” to policymakers, both scientific and policy considerations are unavoidably involved and interconnected in important ways. Examples are provided below. Nonetheless, the influences of both sets of considerations and their interactions are often not explicitly acknowledged. While many scientific advisory bodies and the policymakers they advise claim that their judgments are based on – and only on – scientific considerations, this is almost always misleading. As academic philosophers frequently observe: you cannot derive an “ought” from an “is”.

If scientific advice to policymakers was entirely derived just from scientific considerations, it should always take a plural and conditional form rather than a monolithic and prescriptive form.Footnote ³ For example, it could indicate what is known and not known about the consequences of following – or failing to follow – a range of possible policy options that policymakers have enumerated. If, on the other hand, the advice recommends one particular course of action (eg “adopt and implement these measures”), value judgments must have influenced and contributed to that advice, even if those value judgments remain implicit and unacknowledged.Footnote ⁴

Deciding, for example, what does and does not count as a relevant risk, which types of evidence are to be deemed relevant and how much of which types of evidence are deemed variously necessary and/or sufficient to sustain judgments in favour of accepting, restricting or banning some product or process all presuppose or imply value judgments; they are not empirical findings. Similarly, how the reliability of evidence of various sorts of studies is characterised and adjudicated depends on socially contestable value judgments. Those value judgments are increasingly – and informatively – coming to be termed “risk assessment policy” judgments.Footnote ⁵

The approach adopted in this paper will, therefore, involve making our own key value judgments explicit and indicating how they contribute to our understanding and interpretation of the science and to our policy judgments and recommendations. We shall, moreover, also endeavour to make explicit some of the value judgments that have been implicit in the advice and conclusions of the institutions that we shall be discussing, in particular the European Commission and its CSS and the BfR.

Chemical risk assessment policy decision-making operates in a contestable and often contested space because different groups of stakeholders have different and often conflicting interests. Firms that manufacture, sell and use chemicals have commercial interests that depend on, for example, national governments and the European Commission categorising their compounds as acceptably safe for sale and use. If, however, those products do or might pose risks to, for example, consumers, third parties or the natural environment, then the organisations focused on protecting public and environmental health often favour tight restrictions or even bans on the production, sale and use of those materials.

Biomedical sciences in general – and toxicology in particular – often have to deal with biological variability within and between species and populations, and therefore they often can provide only estimates of probabilities rather than certainty or precision. This is evident, for example, whenever official bodies chose to use a “weight of evidence approach”. Such decisions inevitably reflect contestable value judgments when assigning the “weights”. In practice, when official bodies claim to have applied a “weight of evidence” approach, it has often involved assigning weights to different pieces of evidence in ways that were neither consistent (as, for example, between putative false positives and putative false negatives) nor transparent or justified.Footnote ⁶

Whenever risk assessors provide advice to risk managers that recommends one particular course of action, those recommendations invariably presuppose some value judgments, which provoke questions concerning whether those judgments favour the protection of public and environmental health or the interests of commercial organisations. Our values presume prioritising the protection of public and environmental health over commercial interests, or even over seeking to balance equally between those two conflicting interests. Those values inform our analysis and assessment of the status quo, which we judge to be inadequate, and consequently we interpret the CSS as a welcome step in the right direction.

III. The scientific case for the Commission’s Chemical Strategy for Sustainability

Herzler et al criticised the CSS by arguing that “without a detailed assessment of which risks are currently deemed to be insufficiently addressed, it is hard to establish whether additional regulation might be necessary or existing regulation might need to be improved, and for which part of the population”.Footnote ⁷ But that approximates to arguing that, until a quantitative risk assessment has been completed, it would be premature to regulate on the basis of partial incomplete information, which implies that precaution should never be exercised. While the CSS did not provide a detailed assessment of the hazards or risks that it highlighted, it did not claim that it would or had. The CSS did not purport to provide a purely scientific argument for the directional policy shift that it was proposing. The CSS drew attention to the limitations of the available science; it complained about what it referred to politely as “data gaps” concerning, for example, levels of exposure and underexamined toxicological endpoints. The CSS was making a value judgment about the policy implications of such data gaps and scientific uncertainties. The approach adopted in the CSS is entirely consistent with Regulation (EC) No 1107/2009 that covers pesticidal products, which states: “The precautionary principle should be applied and this Regulation should ensure that industry demonstrates that substances or products produced or placed on the market do not have any harmful effect on human or animal health or any unacceptable effects on the environment.”Footnote ⁸

There are, moreover, extensive bodies of robust evidence indicating that the prevailing chemicals regulatory regime is not sufficiently protective of public health. The same is almost certainly true for the health of ecosystems, but this paper only focuses on threats to public health. An exhaustive review of the different types of harm that chemicals may be causing is beyond the scope of this paper, but several clear examples should be sufficient.

2. Endocrinological and reproductive toxicity remains insufficiently controlled

In relation to reproductive toxicology, Herzler et al implied that the status quo was unproblematic because “the prevalence of anomalies in the newborn in the EU has been practically constant (at around 200/10,000) for the past 40 years”, although they acknowledged that they did not include genetic anomalies or adjust for parental age.Footnote ¹² An examination of the database used by Herzler et al shows that their statement was misleading; the evidence for this claim is provided in Table 1.

Table 1. Prevalence of congenital anomalies per 10,000 births (five-year arithmetic means) in the European Union for the periods 1980–1984 and 2015–2019 for selected groups of anomalies.

Note: We report five-year arithmetic means as indicators of prevailing trends.

Source: <https://eu-rd-platform.jrc.ec.europa.eu/eurocat/eurocat-data/prevalence_en>, where more details can be found.

Firstly, the prevalence of all anomalies in new-borns in the EU has not been “practically constant”. On the contrary, it has increased slowly but continuously, with a 20% increase for the period of 2015–2019 (last year available) compared to 1980–1984. The data show even greater increases for four specific types of anomalies, although some declined over that interval. The reasons for those changes need to be elucidated, but the overall prevalence increased rather than remaining practically constant.

Furthermore, a major health concern with regard to reproductive toxicity has been – and remains – diminishing fertility rather than birth defects. In 1992, Carlsen et al published evidence showing that, in Europe, adult male sperm counts had fallen by approximately 50% in 50 years.Footnote ¹³ Skakkebaek et al recently reported a severe worldwide decline in fertility over the previous five decades. According to the authors, it remains to be elucidated whether this decline is due to exposure to environmental contaminants, changes in lifestyles or both, but increasing incidences of testicular cancer are widely considered to be a result of environmental exposures to contaminants.Footnote ¹⁴ Furthermore, Levine et al conducted a meta-analysis of the results of 223 such studies and reported that sperm counts have been declining on all continents and that these rates of decline have accelerated.Footnote ¹⁵

A report commissioned by the World Health Organization (WHO) and the United Nations Environment Programme (UNEP) articulated serious concerns about the rising incidence of many endocrine-related diseases and disorders and the widespread detection of endocrine-disrupting compounds (EDCs), as well as “significant uncertainties about the true extent of risks from chemicals” due to the lack of data.Footnote ¹⁶ It is sensible to assume that those problems have not diminished during the nine years that it took the European Commission to set out its “scientific criteria for the determination of endocrine disrupting properties”.Footnote ¹⁷

The EU’s process of classifying particular compounds as EDCs and banning them only started recently, and only a very few chemicals have so far been banned, despite evidence that approximately 800 chemicals are known or suspected EDCs. Two studies that employed extensive Monte Carlo simulations estimated that health costs in the EU due to EDCs exceed 150 billion euros annually.Footnote ¹⁸

IV. Herzler et al’s perspective versus precaution

While Herzler et al’s concern that “the strong push” of the CSS for a stricter regulation of chemical mixtures “is not sufficiently data driven”, our interpretation is that the BfR team’s position is antithetical to the EU’s precautionary legislation. The precautionary principle is supposed to be used in case of scientific uncertaintyFootnote ²⁶ ; the legislation mandates the introduction and/or maintenance of regulatory measures in the absence of demonstrated risks. Herzler et al simply misunderstand precaution.

The BfR team implies that the reality of risk needs to be proven before any measures can be justified and that evidence that falls short of demonstrable proof should never be sufficient, except perhaps for carcinogenic, mutagenic or reprotoxic substances (CMRs) or EDCs. Firstly, that is self-evidently a value judgment, not an empirical finding, but no less importantly it is a value judgment that implies consistently awarding the benefits of all and any uncertainties to the commercial interests of the companies that produce, trade in and use the chemicals in question rather than to the protection of, for example, occupational, public and environmental health.

Herzler at al accepted that “precaution” could be appropriate in relation to CMRs or EDCs, but only if those measures are “proportional”. But Herzler et al assume that measures should always be proportional to the magnitudes of particular risks. However, if the magnitudes of the possible risks are uncertain, then the requirement for proportionality cannot be satisfied.

The arguments outlined and the evidence adduced in this section therefore support the European Commission’s judgment that the prevailing chemical regulatory regime may not sufficiently protect public health, and consequently that it could – and should – be strengthened.

V. The Chemical Strategy for Sustainability’s strategy and tactics

To attain the strategic objectives on which the CSS focused, the Commission proposed several tactical reforms, although some of them were under-described. One important change of tactics is the proposal to broaden the scope of EU chemical risk assessments to include wider ranges of potential impacts and outcomes than those currently included. That tactical change could be – and should be – acknowledged as a proposed change to a key EU risk assessment policy.

The plan to widen the scope of risk assessments was evident when, for example, the CSS recommended developing “methodologies for chemical risk assessment that take into account the whole life cycle of substances, materials and products” rather than merely assessing the safety of the particular use for which it will be sold. While the CSS optimistically implied that the prevailing regime was already doing a good job in relation to “chemicals that cause cancers, gene mutations, affect the reproductive or the endocrine system, or are persistent and bioaccumulative”, it envisaged extending that “to further chemicals, including those affecting the immune, neurological or respiratory systems and chemicals toxic to a specific organ”.Footnote ²⁷ The CSS aspired to reduce morbidity as well as premature mortality. The CSS anticipated the EU’s chemicals regulatory regime becoming more precautionary than the prevailing regime, although the word “precaution” appeared nowhere in the document.

The second tactical change focused in particular on EDCs, which interfere with hormonal systems. The CSS said that EDCs “require specific attention”. The CSS proposedFootnote ²⁸ to:

• … establish legally binding hazard identification of endocrine disruptors … building on criteria already developed for pesticides and biocides, and apply it across all legislation;

• ensure that endocrine disruptors are banned in consumer products as soon as they are identified, allowing their use only where it is proven to be essential for society;

• strengthen workers’ protection by introducing endocrine disruptors as a category of substances of very high concern under REACH [the Registration, Evaluation, Authorisation and Restriction of Chemicals];

• ensure that sufficient and appropriate information is made available to authorities to allow the identification of endocrine disruptors by reviewing and strengthening information requirements across legislation;

• accelerate the development and uptake of methods to generate information on endocrine disruptors through screening and testing of substances. (emphases in the original)

That wording acknowledged that prevailing definitions of endocrine-disrupting hazardous chemicals had not been legally binding and had been applied unevenly, and that, even when such disruptors had been identified, they had often not promptly been banned. That is one reason why the CSS also called specifically for the “full implementation of the EU rules on chemicals” (emphasis in the original).Footnote ²⁹ It also implicitly accepted that occupational risks from endocrine disrupters have not been adequately regulated, and that far too little information was available to EU authorities on the composition and toxicology of putative endocrine disrupters. This was partly because too few sufficiently sensitive and specific endocrinological studies have been conducted, but also because the Commission suspected that industry may have withheld some or all of the data they have regarding some of the studies that have been conducted. This is why the CSS explicitly articulated a new policy of: “ZERO TOLERANCE FOR NON-COMPLIANCE” (Section 2.3.2, p 18, emphasis in the original). It will “strengthen the principles of ‘no data, no market’ and the ‘polluter-pays’” (ibid).

A third tactical change responded to the scale of the scarcity and inadequacies of the data. The CSS stated that the Commission will “amend … information requirements to enable identification of all carcinogenic substances manufactured or imported in the EU” (Section 2.4.1, p 20, emphasis in the original), and that

The fourth proposed tactical change responded to the fact that EU regulatory bodies have often relied mainly on studies designed and conducted by or on behalf of the chemical industry, and that too little attention had been paid to studies conducted independently of corporate interests, as the CSS rather coyly observed that “academic studies are not sufficiently exploited” (Section 2.3.1, p 16, emphasis in the original).

The fifth proposed tactical change responded to the fact that, in relation to pharmaceutical products, pesticides and food ingredients, EU rules have recently been amended to stipulate that no study can be used in support of an application for market authorisation unless the protocol for that study had been notified to the Commission before the study commenced.Footnote ³⁰ This is designed to ensure that companies no longer have the discretion to decide – for themselves and confidentially – whether or not to provide the EU’s regulatory institutions with all data from all studies.

The sixth proposed tactical change focused on additional risks potentially arising from the combination effects of chemicals. It stated:

The seventh proposed tactical reform is to strengthen the regime covering chemical pollutants in the natural environment, so that environmental health as well as public health are properly protected. The CSS provided particular comments on one type of pollutant, namely the per- and poly-fluoroalkyl substances (PFASs)Footnote ³¹ , which it said

Overall, the CSS recognised many of the achievements of the EU’s chemicals regulatory regimes, but also many of its limitations and shortcomings, and consequently it highlighted at least seven key respects in which the Commission judged that those regimes could and should be strengthened, not just in the sense of being more restrictive, but also in the sense of being more comprehensive, better informed and more accountable by being more transparent.

VI. The BfR team’s response to the CSS

Herzler et al’s critical response to the CSS provided a complacent narrative, asserting that toxicological risks from chemicals had already been minimised and that the prevailing regulatory regimes had been “optimised”.Footnote ³² It implicitly assumed that all toxicological risks have been exhaustively identified, documented, quantified and subjected to controls that are comprehensive, effective and sufficient. That judgment combined both factual and normative considerations, many of which we do not share and are contesting.

The BfR team argued that, while there might be a role for the Commission to consult with a broad range of social, environmental and commercial stakeholders, decisions about the composition and implementation of a chemicals sustainability strategy should be taken only by reference to scientific considerations. We reiterate, however, that science alone can never settle policy questions. The BfR team implied that its own narrative was an entirely scientific one; we shall, however, provide evidence showing that it was replete with unacknowledged value judgments.

The BfR team asserted that

VII. A critique of Herzler et al and of the status quo

The above is a surprisingly bold assertion from the BfR given its profoundly flawed approach to producing a risk assessment of the controversial herbicide glyphosate. One would at least have expected that all applicable regulations and guidelines had been properly followed by the BfR when it prepared for the EFSA and the European Commission its draft Renewal Assessment Report (RAR) of glyphosate in 2015. However, several flaws remained in the BfR’s RAR, though some of these were corrected in an Addendum,Footnote ³⁴ which rapidly emerged at the end of August 2015 in response to the International Agency for Research on Cancer (IARC) monograph on glyphosate that had been published a few weeks earlier. Other flaws in the BfR’s RAR, however, remained unaddressed.

One of the flaws that was corrected was a reassessment of statistical significances using a proper statistical procedure. In the Addendum, BfR acknowledged that “initially, the RMS [Rapporteur Member State] [had] relied on the statistical evaluation provided with the study reports”.Footnote ³⁵ That wording reveals that the BfR failed to conduct an independent statistical assessment of the data provided by the industry’s Glyphosate Taskforce until after it came under political pressure, once IARC asserted that glyphosate is “probably carcinogenic to humans”.Footnote ³⁶ Only then did the BfR conduct its own statistical evaluation, which used a statistical method appropriate for evaluating tumour incidences, namely a trend test, rather than the pairwise comparisons that were reported in Glyphosate Taskforce documents.

The BfR also initially ignored available evidence of glyphosate exposure causing oxidative stress, which is a potentially relevant mechanism that could explain the development of tumours. That evidence had not been acknowledged in its original assessment. Only after IARC had identified “strong evidence that glyphosate … can act to induce oxidative stress”Footnote ³⁷ did the BfR acknowledge this in the RAR Addendum.

Both the BfR and subsequently EFSA discounted evidence of glyphosate’s carcinogenicity by claiming that the dose levels used in some of the mouse studies exceeded what they alleged was a maximum “limit dose” of 1,000 mg/kg.Footnote ³⁸ But no such limit has been officially specified for carcinogenicity studies. The Organisation for Economic Co-operation and Development (OECD) guidance documents, to which the RAR referred, had recommended a maximum dose of 1,000 mg/kg for chronic toxicity studies but not for carcinogenicity studies.Footnote ³⁹ The spurious tactic of citing the OECD was invoked by the BfR in its 2015 report on glyphosate, by EFSA in its conclusions on glyphosate, in the Harmonised Classification and Labelling (CLH) report on glyphosate and again in a recent draft report on glyphosate of the EU’s Assessment Group on Glyphosate,Footnote ⁴⁰ even after the limit-dose argument had repeatedly been criticised in peer-reviewed journals.Footnote ⁴¹

Instead, OECD guidance document No. 116 recommended avoiding dietary concentrations of test substances higher than 50,000 mg/kg in long-term bioassays.Footnote ⁴² This concentration is well above the highest concentrations of 20,000 and 30,000 mg/kg used in two of the mouse carcinogenicity studies on glyphosate. This was, however, never acknowledged by the BfR, EFSA or the European Chemicals Agency (ECHA).

VIII. Further problems with the status quo

Another problematic feature of the status quo in official EU toxicological assessments of the carcinogenic potential of pesticides has been the tactic of invoking historical control data (HCD) as grounds for discounting evidence of adverse effects, especially increased tumour incidences. Often when corporate interests want to discount putative evidence of toxicity that emerges from comparing evidence of adverse effects in groups of test animals with evidence from the concurrent control group, they argue that the tumour incidences seen in the concurrent control group were unusually low when compared to control groups examined previously in earlier studies. They argue that if the test animals were compared to historical control groups the differences would cease to be statistically and/or toxicologically significant.

OECD guidance document No. 116Footnote ⁴³ and Guidance of the European Chemicals AgencyFootnote ⁴⁴ both provided very similar recommendations for the conditions under which HCD can legitimately be used to interpret new data. While HCD on tumour incidences in earlier control groups may sometimes be relevant when assessing the findings of recent studies, OECD emphasised “that the concurrent control group is always the most important consideration in the testing for increased tumour rates”.Footnote ⁴⁵ Similarly, ECHA stipulated that “historical data must be from the same animal strain/species, and ideally, be from the same laboratory to minimise any potential confounding due to variations in laboratory conditions, study conditions, animal suppliers, husbandry”, and that historical data should be contemporary to the study being evaluated (eg within a period of up to around five years of the study).Footnote ⁴⁶

The BfR’s glyphosate RAR represented a flagrant example of violating those rules when citing HCD to discount adverse findings. In practice, the BfR used a compilation of HCD stretching over a timespan of twenty-two years from seven different laboratories, comparing the results from a stressful housing condition (wire-bottom cages) of the historical controls with those of the glyphosate studies housed under less stressful conditions (eg solid-floor cages), while also neglecting the duration of studies (eighteen versus twenty-four months), which had an important influence on spontaneous tumour incidences.

Other official advisory bodies in EU Member States have contributed to the maintenance of the status quo. For example, when acting as rapporteurs, they have often invoked HCDs to discount positive toxicological findings in ways that violated the OECD and ECHA criteria, and dismissed evidence of increased tumour incidences when compared with concurrent controls. Specifically, such transgressions can be found in the carcinogenicity assessments in four out of nine other draft RARs reviewed by Clausing,Footnote ⁴⁷ namely those for the active ingredients dimoxystrobin, folpet, phosmet and pirimicarb. Official institutions have frequently acted in ways that are inconsistent with the rules with which they claimed to comply.

Another important type of the failure in practically applying “the principles and methods of the science of regulatory risk assessment”Footnote ⁴⁸ has been the willingness of European authorities to accept and rely on studies with high mortality rates during the course of those studies. The premature deaths of the animals entail, firstly, that many died too young to have provided sufficient data, and secondly, that the numbers of animals available for detailed end-of-study post-mortem examination are too small to provide sufficient data from which statistically significant effects might be derivable.

Carcinogenicity studies are long-term studies; they typically cover approximately 75% of the average life expectancy of laboratory rodents. The validity of the findings of such studies crucially depends on sufficient animals surviving to the end of the minimum study duration of eighteen or twenty-four months for mice and rats, respectively. Therefore, OECD guidance document No. 116 recommended:

Details from two RARs serve to illustrate violations of that requirement.Footnote ⁵⁰ In the RAR of phosmet, the survival rate in a rat study was well below 50%, but it was nevertheless deemed acceptable and reliable by the authorities as indicating a lack of carcinogenicity. Survival rates at the end of the 108 weeks were 20%, 24%, 32% and 38% in males and 32%, 35%, 32% and 42% in females for the control, low-, mid- and high-dose groups, respectively.

Similarly, a rat study on pirimicarb with survival rates in males of between 35% and 48% was accepted. Ironically, for the mouse study, the following was noted: “In this study there is evidence of a carcinogenic potential of pirimicarb but with the limitations in the historical control data and reduced survival this data alone is not sufficient to conclude on carcinogenicity.”Footnote ⁵¹

In other words, in the rat study, where the Rapporteur Member State (the UK) should have taken low survival rates into account before accepting a putative negative finding, that criterion was ignored. However, in the mouse study, in a situation where survival rates were not significantly low, a positive finding was nonetheless dismissed by alleging reduced survival, even though reduced survival is not a legitimate criterion for exclusion in cases of positive findings of significantly increased tumour incidences.

Herzler et al criticised the CSS for arguing that there is “a necessity for further improving the protection of ‘vulnerable groups’”.Footnote ⁵² Their view was that “without a detailed assessment of which risks are currently deemed to be insufficiently addressed, it is hard to establish whether additional regulation might be necessary or existing regulation might need to be improved, and for which part of the population” (ibid). But that approximates to arguing that, until a quantitative risk assessment has been completed, it would be premature to regulate on the basis of partial incomplete information, which implies that precaution should never be exercised; but that is incompatible with EU legislation.

In relation to EDCs, Herzler et al implied that, before the CSS’s approach could deserve to be adopted, and before any further particular measures are implemented, the Commission should comprehensively identify all toxicologically problematic chemicals; that approach is, however, directly antithetical to precaution. Herzler et al’s perspective also irredeemably involves a value judgment – and not a scientific one – about how much of which kinds of evidence should be deemed necessary and sufficient before specific measures are adopted. Their value judgment, moreover, accords with industrial interests rather than prioritising the protection of public and environmental health.

IX. The BfR’s misunderstandings and misrepresentations of the Chemical Strategy for Sustainability

Herzler et al complained that the CSS was biased, but they were seemingly unaware of their own biases, many of which are evident from the facts that they chose to emphasise and the indicators that they chose to highlight. For example, Herzler et al criticised the CSS for failing adequately to differentiate “risks” from “hazards”. Historically, chemicals were typically restricted or banned only on the basis of a demonstrable risk given anticipated levels of exposure, not just on the basis of their intrinsic hazardous properties without reference to particular levels or frequencies of exposure. Herzler et al said correctly that

However, the BfR team was itself blurring this distinction by failing to engage properly with the fact that some hazards – under the explicit provisions of EU legislation – have been officially categorised as unacceptable. For example, EU Regulation 1107/2009 states that a genotoxic, carcinogenic or reprotoxic pesticides “shall only be approved … if it is not or has not to be classified … as category 1A or 1B” (for mutagenicity, carcinogenicity or reproductive toxicity). This pivotal piece of legislation focuses on hazard categories, referring to their intrinsic toxicological properties.

The BfR itself risked disseminating “conceptual misunderstandings” by repeatedly stating that, in the case of glyphosate, IARC “only” provided a hazard assessment, while implying that the BfR had been more thorough by providing a risk assessment.Footnote ⁵⁴ In practice, moreover, the BfR’s judgment focused in particular on an important hazard characteristic when concluding that glyphosate is not carcinogenic. If the BfR had come to the same conclusion as IARC, under EU legislation no risk assessment would have been needed.

The Commission did not fail to distinguish risks from hazards; rather, it articulated an evaluative policy judgment that both sets of considerations can – and under certain conditions should – be adequate grounds for introducing or maintaining regulatory measures. The BfR team simplistically suggested that policy decisions can be made solely on scientific grounds and assumed that their chosen version of “the science” was incontestable. Claiming or even just assuming that only risks (which the BfR team interpret as quantified proven damage but not evidence of hazards) should be sufficient grounds for regulatory controls is unambiguously a normative policy judgment rather than an empirical finding.