The public consultation on the ”opinion concerning Toxicity and Assessment of Chemical Mixtures” by the Scientific Committee on Consumer Safety (SCCS), the Scientific Committee on Health and Environmental Risks (SCHER) and the Scientific Committee on Emerging and Newly Identified Health Risks (SCENIHR) ended on the 15th of September. Below are the comments that I submitted. For the ease of reading things in context, I first provide the inital question, then the reaction of the Scientific Committees (all of which are also given in the draft opinion itself) and then the comment(s) that I submitted.
If you don’t fancy reading on the screen, please click this link – which should give you the possibility to generate a PD or print the text. You can even delete paragraphs before printing, e.g. if you do not want the complete quotes from the original opinion draft.
There was a word limit of 4000 characters per comments, which was certainly an impediment. I could have used some more space at least for the answers to question 1 (science) and question 4 (possible approaches). Oh well, there will be more possibilities for exchanging views and ideas.
Now I only hope that the Committees make all the comments that they received during the consultation period publically available. Should be an interesting read…
Question 1: Is there scientific evidence that when organisms are exposed to a number of different chemical substances, that these substances may act jointly in a way (addition, antagonism, potentiation, synergies, etc.) that affects the overall level of toxicity?
Yes, under certain conditions, chemicals may act jointly in a way that the overall level of toxicity is influenced.
Chemicals with common modes of action may act jointly to produce combination effects that are larger than the effects of each mixture component applied singly. These effects can be described by dose/concentration addition.
For chemicals with different modes of action (independently acting), no robust evidence is available that exposure to a mixture of such substances is of health concern if the individual chemicals are present at or below their zero-effect levels. It is important to note that these zero-effect levels are not represented by the NOELs or NOECs. NOEL(C)s or PNECs are derived from experimental studies and may be associated with effect levels of up to 20%. Chemicals with different modes of action may however also affect the same endpoint, for instance, acute toxicity or carcinogenicity (effect addition).
For ecological effects, the exposure to mixtures of dissimilarly acting substances at low but potentially relevant concentrations should be considered, even if all substances are below the individual PNECs.
In the examples in which independent action provided a more accurate prediction, dose (concentration) addition slightly overestimated the actual mixture toxicity, which suggests that the use of the dose/concentration concept for risk assessment of chemicals of unknown toxic mechanisms is sufficiently protective.
Interactions (including antagonism, potentiation, synergies) usually occur at medium or high dose levels (relative to the lowest effect levels). At low exposure levels they are either not occurring or toxicologically insignificant.
- There seems to be an emerging consensus that individual NOELs/NOECs do not safeguard against unwanted mixture effects, neither for similarly nor for dissimilarly acting substances and neither in a human health nor ecotoxicological context. I fully support this conclusion.
- In the part of the Kortenkamp et al report (2009) that concerns the low-dose issue, it is stated (as also quoted in the draft opinion, p. 12) that “In demonstrating that dissimilarly acting chemicals too have the propensity to produce significant mixture effects when combined at levels below NOECs…”. This statement is contested in the following text of the draft opinion (“The Working Group has evaluated these studies and concludes that they do not allow such interpretation:”). However, the study details that are provided in the following (page 12/13 [of the opinion]) do not seem to contradict the initial statement that NOECs, but also certain fractions of NOECs, might contribute to the toxicity of a mixture, even if composed of dissimilarly acting compounds. It should perhaps be emphasized that the Kortenkamp et al report only makes the statement that NOECs are not safe (“levels below NOECs can contribute”), which is completely in line with the draft opinion. The Kortenkamp report does not state that ALL concentrations below NOECs (e.g. DNELs, TDIs) always contribute to the mixture toxicity. In fact, the concentration range below NOECs and other PODIs was termed a “grey area” in which we have only very limited experimental evidence (Kortenkamp et al report, executive summary, pages 6/7).
- The term “no effect level” (or “zero effect level”) plays a critical role in the text of the opinion (not only with respect to this question 1, but also in the core text, e.g. page 31). However, no precise definition of this central term is given in either the core text or the accompanying glossar. On page 11 it is only stated what no effect levels are not (NOECs). Later in the document zero effect levels seem to be equated to the DNEL or TDI (page 30), which would limit the use of the term to human toxicology.
- The Michaelis-Menten Kinetics that is provided earlier as background information on receptor interactions and hence the nature of concentration-response relationships (page 13 of the draft) is actually a non-threshold model. Implying, that at least on a molecular level a “no effect level” is never reached (as long as the concentration is > 0).
- Obviously, the Michaelis-Menten equation only describes molecular events in a simple system and hence a threshold of toxicological/ecotoxicological concern might very well exist (due to e.g. compensation mechanisms of the exposed organism). This, however, implies that from a mixture perspective, joint effects cannot not be ruled out from first principle. Under the assumption that the classical model of receptor binding (pages 13/14) holds and depending on the nature of the mixture of interest (in terms of potency and number of involved compounds), toxicologically non-observable, insignificant individual events might still lead to a toxicologically relevant response to the mixture, even if independent action is assumed. Simply because each concentration could be above its molecular no effect level. Again it should be stressed the empirical evidence is almost non-existing for this situation.
- Even under the (unrealistic) assumption of a completely independent mode of action of all mixture components, a mixture can only be considered safe a priori, if it can be ensured that all components are present not above their absolute no-effect level. The question aside on how realistic such a situation might be – how would one prove the absence of an effect? As obviously the absence of empirical proof (an individual effect was undetectable in a study) is no proof of the actual absence (of an effect), which would be required from a mixture toxicity perspective.
- The term “effect addition” is used without definition or explanation (also e.g. page 9).
Question 2 – If different chemical substances to which man/environment are exposed can be expected to act jointly in a way which affects their impact/toxicity
on/for man and the environment, do the current assessment methods take proper account of these joint actions?
Risk assessment on the combined effects of chemicals in a mixture is not commonly carried out at present. However, for some purposes, toxicity testing will be applied to mixtures.
As outlined in the answer to question 1, different chemical substances may act jointly ina way which affects their toxicity for man and the environment, current assessment methods for mixtures can take account of joint actions, such as dose/concentration addition or response / effect addition generally only applied under specific circumstances. With these methods acute effects of chemical mixtures composed of either dissimilarly or similarly acting substances can be reasonably well predicted. Interactions, however, are generally more difficult to assess and require expert judgement on a case-by-case basis. Specific conditions under which synergistic actions, i.e., the most relevant of interactions with regard to the toxicological risk, might be expected are outlined in the above opinion.
The methodology for the (eco-) toxicological assessment of chemical mixtures appears, generally suitable. It is, however, often not applied in practice. Assessments of aggregated and combined exposures across different industrial and use sectors, in particular, are rarely performed.
- The referenced report by Kortenkamp et al. (2009) provides quite a few examples of mixture studies that relate to the applicability of Concentration Addition and Independent Action on chronic endpoints. This holds true especially for ecotoxicology, where several studies are available on e.g. chronic studies with the classic test organisms such as bacteria, algae, daphnids and fish. It is therefore perhaps not sufficient to relate the applicability of Concentration Addition and Independent Action to acute effects only (as stated in the second paragraph).
Question 3 – Several approaches for the assessment of the mixture effects of chemicals already exist such as dose addition and independent action. What are the advantages and disadvantages of the different approaches and is there any particular model that could be considered as sufficiently robust to be used as a default option?
In view of the huge variety of human exposures to chemical mixtures, the default assumption in human risk assessment had been that they generally acted by dissimilar modes of action. In cases, however, where information is available to indicate a similar mode of action, a dose/concentration addition approach is appropriate. A dose/concentration addition approach, if applied to chemical mixture components with unknown modes of action, may result in an over-prediction of toxicity; using the independent action approach may however underestimate toxicity. Therefore, also in this case, the dose/concentration addition approach is preferable to ensure an adequate level of protection. Different methods exist for the dose/concentration addition approach (see above methodology section for details). When using the RfP or RV, one should be aware that NOAELs/LOAELs are based on single experimental data points and the values depend on the dose-spacing used in the experiment. In contrast, BMDLs are based on all experimental points and by that provide more reliable information on the dose response.
In ecotoxicology, any approach must be referred to specific endpoints and to defined taxonomic groups of organisms. The reference values (PNECs) are derived using different sensitive organisms for any type of chemical. Therefore, a combination of PNECs may be misleading.
A significant limitation of component-based approaches is that they are only applicable to mixtures of which the major components are known.
- The summation) of PEC/PNEC ratios has been repeatedly suggested in the peer-reviewed literature. This approach is known to be slightly more conservative than the summation of toxic units (which would be the scientifically more correct approach), depending on the amount and type of available data and the toxicity profiles of the mixture components. Details are provided in a separate PDF (final research report of the BEAM EU project). Hence the sum of PEC/PNECs could be used as a first (sub)-tier when applying Concentration Addition. Summing up PEC/PNECs is a particularly attractive approach as it is straight forward to integrate in current risk assessment schemes, and is applicable in situations where different amounts and data types are available for the individual compounds (see separate report for a detailed discussion).
Question 4 – Given that it is unrealistic to assess every possible combination of chemical substances what is the most effective way to target resources on those
combinations of chemicals that constitute the highest risk for man and the environment?
In view of the almost infinite number of possible combinations of chemicals to which humans and environmental species are exposed some form of initial filter to allow a focuson mixtures of potential concern is necessary. The following criteria are proposed for consideration:
- Human and/or environmental exposure at significant levels (e.g. approaching the NOEL/NOEC or PNEC for several components).
- Chemicals that are produced and/or marketed as multi-constituent substances or commercial mixtures with several components and/or active ingredients (i.e., as
- defined by EU legislation, e.g., REACH, CLP, pesticides and biocidal products legislation, food law, etc.).
- Potential serious adverse effects of one or more chemicals at the likely exposure levels.
- Likelihood of frequent or large scale exposure of the human population or the environment.
- Persistence of chemicals in the body and/or in the environment. High persistence/bioaccumulation would be a property of importance.
- Known information of potential interaction at levels of human and environmental exposure.
- Predictive information that chemicals act similarly such as (quantitative) structure activity relationships and structural alerts.
- Particular attention should be paid to mixtures for which one or more components are assumed to have no threshold for its effects such as genotoxic carcinogens; a MOE or a lifetime cancer risk approach could be applied.
- Exposure to one or more components approaching the threshold levels for adverse effects would mean that the mixture should be given priority for assessment. A TTC like approach can be used to eliminate combinations that are of concern (for details on the applicability of a TTC approach for the assessment of chemical mixtures see Boobis et al., 2011 and Price et al., 2009).
For the environment, attention should be paid to mixtures of chemicals, individual components of which approach the PNEC.
In view of the difficulty and time needed to retrieve or generate an appropriate dataset for hazard characterisation and exposure estimates, a tiered approach, such as proposed by the WHO/IPCS (2009b) or EFSA (2008), may be considered. (For details on the tiered approach, see above text.) The identification of the data gaps after the application of the tiered approach should determine the extent of testing of chemical mixtures and study design.
- The TTC approach is certainly a strategy with substantial potential. It should be stressed, however, that to my knowledge there is no experimental evidence at hand that demonstrates that a mixture assessment approach based on the TTC concept is protective. In view of such a fundamental lack of data it might be too far fetching if the TTC concept is suggested already now in a regulatory setting. It requires validation first.
- In this context it is interesting to note that the corresponding opinion (“Use of the Threshold of Toxicological Concern (TTC) Approach for the Safety Assessment of Chemical Substances”, by SCHER, SCCP and SCENIHR) still does not seem to be finalized even for individual compounds – although the public consultation period ended already in January 2009. Also the opinion on “Exploring options for providing preliminary advice about possible human health risks based on the concept of Threshold of Toxicological Concern (TTC)” is not finalized yet.
- There seems to be a consensus that mixtures of compounds with a similar mode or mechanism of action follow the basic principle of Concentration Addition. This implies that under these circumstances even concentrations at or below the TTC add to the joint toxicity of a mixture and hence mixture effects cannot be ruled out. This situation is, however, not adequately considered in the provided decision tree (page 36 of the draft [of the opinion, that is]). According to the decision tree further action is never required if the individual compounds are present below their individual TTCs, even if all compounds are similarly acting. This does not only violate the basic principle of Concentration Addition, but is also in contradiction to e.g. the TEF/TEQ approach for dioxin mixtures.
- Even if it is assumed that the TTC is a valid approximation of a true zero effect concentration, a mixture can only be regarded a priori as causing no reason for concern, if all compounds are present below their individual TTC and if all compounds are completely dissimilarly acting. It is currently completely unknown whether such a scenario is realistic or whether it is only a very special case of more theoretical relevance.
- The decision tree on page 36 of the preliminary opinion is supposed to capture both, the toxicological as well as the ecotoxicological assessment of a mixture. However, the TTC is a concept that is so far rooted strongly only in human toxicology. Although there is a one publication by de Wolf et al (Mode of action and aquatic exposure thresholds of no concern, ET&C, 24(2), 479-485), there is currently no evidence (either conceptual or empirical) on how such an approach relates to common ecotoxicological descriptors of low effects or thresholds (e.g. PNECs, EQs). Furthermore, the publication focuses on a few well studied mechanisms of action only. Again, in view of the lack of conceptual and empirical models/data it seems somewhat premature to use the TTC criterion as a broad and general indicator for “no further action required”.
- Finally, the draft TTC-opinion lists classes of compounds and endpoints for which the TTC concept is not suited (e.g. nano-materials, the endpoints “pharmacological or microbiological effects, page 30 of the TTC draft opinion). These exceptions could be integrated in the decision tree for mixture assessment or its supporting text.
Question 5 – Where are the major knowledge gaps with regard to the assessment of the toxicity of chemical mixtures?
With regard to the assessment of chemical mixtures (as defined in the mandate), a major knowledge gap at the present time is the rather limited number of chemicals for which there is good mode of action information. Currently there is neither an agreed inventory of mode of actions, nor a defined set of criteria how to characterise a mode of action for data-poor chemicals.
Much of the work on interactions relates to enzyme inducers and inhibitors, to promoters of carcinogenic effects. The dose/concentration approach requires information on the dose response shape for the chemicals to be considered. This information is rarely available in sufficient quality. Research is needed to define criteria that predict dose additivity.
In ecotoxicology, the problem is even more complex. A knowledge of all possible modes of actions that may occur in the different types of organisms of a complex biological community is difficult (if not impossible) to be attained. On the other hand, it must be considered that ecologically relevant endpoints are generally broader and not so specific (e.g. toxicity on specific organs, etc.) as in human toxicology.
Other major knowledge gaps are:
- The general lack of robust and validated tools for the prediction of interactions.
- How exposure and/or effects may change over time
- It has been repeatedly shown that Independent Action and Concentration Addition often predict virtually indistinguishable mixture toxicities (see e.g. the Kortenkamp et al. report). Also in the recent publication by Brice et al that is actually cited in the draft opinion (“Characterizing the Noncancer Toxicity of Mixtures Using Concepts from the TTC and Quantitative Models of Uncertainty in Mixture Toxicity”, Risk Analysis, 29(11), 1534-1548, 2009), it can be clearly seen that the difference in toxicity (risk) predictions are usually negligible, far below any regulatory relevance. In view of this body of evidence the call of more mechanism / mode of action information might be misleading. The major knowledge gap hampering mixture assessment is the lack of knowledge on where, how often and to what extent humans and the environment are exposed to what types of chemical mixtures.
Question 6 – Does current knowledge constitute a sufficiently solid foundation upon which to address the toxicity of chemical mixtures in a more systematic wayin the context of EU legislations?
In many cases, knowledge is insufficient for a robust scientific analysis.
If toxicologically significant interactions can be excluded, the components of a mixture are identified and known mode of action information is available, either a dose addition
or independent action model should be applied. This set of information, in human toxicology, is however rarely available and, in most cases, very cost- and labourintensive
to generate. Often, it may not be possible to obtain the required data due, e.g., to limitations in existing study designs and analytical methods.
In ecotoxicology, the mode of action should be known for all the relevant taxonomic groups of aquatic and terrestrial ecosystems. So, the availability of information is even
more difficult; in addition, modes of actions considered dissimilar at the individual level may affect the same population relevant endpoint, and therefore, the dose/concentration
addition model may be more appropriate for predicting effects at the population level.
However, in most cases, when applying a dose/concentration addition approach, it is necessary to rely on assumptions such as mode of action, shape and slope of dose response curves of the individual components. These assumptions may be generated by grouping of chemicals into categories and assessment groups. However, no generally agreed criteria for the grouping of substances exist, adding to the uncertainties associated with this approach. Choosing independent action approach may however underestimate combined effects of similarly acting chemicals. Hence, if no mode of action information is available, the dose/concentration addition method should be preferred over the independent action approach.
Prediction of possible interaction requires expert judgement and hence needs to be considered on a case-by-case basis.
In future, pathway-based toxicity evaluations (e.g. inflammation – oxidative stress – genotoxicity) based on in silico and in vitro methodology will become more feasible, enabling these methods to identify common effects. However, the report of a recent meeting of the US National Academic’s Standing Committee on Use of Emerging Science for Environmental Health Decision concluded that “many challenges remain to be addressed before the findings from high-throughput screens and in silico models may be considered sufficiently robust and informative” (Rusin and Daston, 2010). The Working Group agrees with this conclusion.
In ecotoxicology, a relevant issue may be related to combined effects capable to affect reproduction, population dynamics and ecosystem’s health. For some chemicals these effects may become evident even some time after exposure stopped.
Having reviewed the available evidence, the Committees recommend that a mixture dependent approach is used for the assessment of chemical mixtures as outlined in the following diagram:
- I fully support the conclusion that in the absence of mode of action information, the dose/concentration addition method should be preferred. This constitutes a sufficiently robust method for the first tier assessment of mixtures in a regulatory setting. Empirical evidence as well as conceptual considerations indicates that the additional consideration of Independent Action often does not add much information (see discussion in the report by Kortenkamp et al. 2009 and Brice et al 2009 (see full reference above)). Please refer to specific comments to the diagram in from page 36 of the draft opinion in my comments to question 4.