SCIENTIFIC REPORT submitted to EFSA

Size: px
Start display at page:

Download "SCIENTIFIC REPORT submitted to EFSA"

Transcription

1 Supporting Publications 2012:EN-232 SCIENTIFIC REPORT submitted to EFSA Investigation of the state of the science on combined actions of chemicals in food through dissimilar modes of action and proposal for science-based approach for performing related cumulative risk assessment 1 Prepared by: Andreas Kortenkamp (ULSOP), Richard Evans (ULSOP), Michael Faust (F+B), Fritz Kalberlah (FoBiG), Martin Scholze (ULSOP), Ulrike Schuhmacher-Wolz (FoBiG) (ULSOP: School of Pharmacy, University of London; F+B: Faust and Backhaus Environmental Consulting GbR; FoBiG: Forschungs- und Beratungsinstitut Gefahrstoffe GmbH) 1 Question No EFSA-Q Accepted for Publication on 23 January Any enquiries related to this output should be addressed to pesticides.pprprocurement@efsa.europa.eu Suggested citation: Kortenkamp A, Evans R, Faust M, Kalberlah F, Scholze M and Schuhmacher-Wolz U. Investigation of the state of the science on combined actions of chemicals in food through dissimilar modes of action and proposal for science-based approach for performing related cumulative risk assessment. Supporting Publications 2012:EN-232. [233 pp.]. Available online: European Food Safety Authority, 2012

2 Abstract The purpose of this project was to summarise the state of the science on combined actions of chemicals in food through dissimilar modes of action and to propose a science basedapproach for performing the related cumulative risk assessment (CRA). A systematic literature search was carried out (Task 1) to identify relevant experimental studies of mixtures, and this served as the basis for a summary of the state of the science (Task 2). Of central importance was the confirmation that there is no current example of a situation in which the concept of independent action (IA) provides an accurate prediction that is also more conservative than dose addition (DA), supporting the use of DA as a conservative default in CRA. The quantitative difference between predictions based on DA or IA were analysed in detail, and this analysis suggested that the differences that might be expected in practice are small. Currently used approaches to grouping (Task 3) and to CRA (Task 4) were reviewed to identify their critical features. An approach is proposed (Task 5) that unifies the assessment of similarly and dissimilarly acting chemicals based on pragmatic use of assessment approaches derived from the concept of DA. The approach incorporates a tiered framework. At lower tiers, the grouping of chemicals is driven by their co-occurrence in the exposure scenarios under investigation. At higher tiers, chemicals that evoke a common adverse outcome should be grouped together. The proposal is illustrated with three case studies. Conclusions and recommendations It is feasible and justified to utilise CRA methods and tiered framework analyses derived from DA also for combinations of dissimilarly acting chemicals. There can be one unified approach for dealing with mixtures in regulatory practice, irrespective of (often presumed) modes of action. Key words: Combined actions of chemicals, cumulative risk assessment, dissimilar modes of action, dose addition, science-based approach, independent action Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

3 Summary The aims of this project were to 1) summarise the state of the science on combined actions of chemicals in food through dissimilar modes of action, and 2) to propose a science basedapproach for performing the related cumulative risk assessment (CRA). The project progressed through six project tasks, culminating in this report (Task 6). The report is broadly structured according to the project tasks but flexibility was used to allow the report content to be arranged suitably for the scientific content. Task 1 was to collect and scrutinize the relevant scientific literature. A search strategy was refined to identify relevant experimental mixture studies and the literature was compiled into a Microsoft Access database, termed the CRADIS (Cumulative Risk Assessment of Dissimilarly acting chemicals) database (sections 4-7). The CRADIS database accompanies this report and contains 173 experimental mixture studies (section 8). The search was performed in Web of Knowledge (Thomson Reuters) to give the best coverage of the literature. Experimental studies were classified and analysed within the database. All chemicals included in the experimental studies were indexed allowing an overview of the types of chemicals being studied in the literature. A deep analysis of experimental design and results was restricted to studies considered to have relevance to two issues: 1) low dose issues and 2) the issue of chemical dissimilarity. The literature identified in this task served as a basis for Task 2. Task 2 was to summarize and assess the state of the science. Relevant mixture studies were analysed and revealed that the majority of the literature deals with studies comprising only a few components, often only two components were studied (section 9). The literature relating to low dose issues and to chemical dissimilarity was reviewed in detail (section 10). In this task definitions of dissimilar mode of action were compiled and compared (section 11) and legislation relating to cumulative risk assessment was reviewed (section 12). The available assessment concepts for mixture toxicity, including independent action (IA) and dose addition (DA), were reviewed and this included considerations of the theoretical background, applicability, data requirements and empirical evidence for each concept (section 13). A thorough analysis of the quantitative difference between DA and IA predictions was considered pivotal to this project (section 13.4) and revealed that prediction might be expected to differ by less than one order of magnitude, even for mixtures with a high number of components. Finally, cumulative risk assessment approaches based on the available mixture concepts were reviewed (section 13.6) and the data requirements set in legislation were analysed (section 13.9). The state of the science summary arising from this task underpins the work in the subsequent Tasks 3, 4 and 5. Task 3 was to propose scientific criteria for establishing cumulative assessment groups of pesticides and other types of chemicals when dissimilar mode of action is a relevant mechanism leading to a common effect, in order to perform dietary cumulative risk assessment for regulatory purposes. The review of literature relevant to this task was included in the report section for task 2, with a review of the definitions that are relevant to dissimilar modes of action. In this section a physiologically based approach to grouping was reviewed. The development and use of grouping criteria are integral to the approach that is proposed in task 5, therefore proposals relevant to this are detailed in that task (section 15). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

4 Task 4 was to assess the available approaches and methods for risk assessment of mixtures of pesticides and other chemicals in or on foods showing dissimilar mode of action. The approaches discussed by various bodies (US EPA, EFSA, WHO/IPCS) are described and analysed (section 14). The use of tiered approaches in cumulative risk assessment were considered in detail (section 14.3) and integrated into the final proposal. Task 5 was to propose a science-based approach for performing cumulative risk assessment of chemicals in food acting through dissimilar modes of action. An approach is proposed that unifies the assessment of similarly and dissimilarly acting chemicals based on pragmatic use of assessment approaches derived from the concept of DA (section 15). The approach incorporates a tiered framework in line with other proposals such as that developed by IPCS (2009). The approach suggests to abandon distinctions according to (presumed) modes of action of chemicals at lower tiers of the analysis and to assess all the chemicals that occur together in the exposure scenario under investigation. At higher tiers, and only if the risks identified at lower tiers are deemed unacceptable, should chemicals be grouped together according to their ability to evoke a common adverse outcome. Features of the proposed approach are illustrated in three case studies (section 15.8). Conclusions. The overall conclusion of this project is that it is feasible and justified to utilise CRA methods and tiered framework analyses originally developed for similarly acting mixtures also for combinations of dissimilarly acting chemicals. There can be one unified approach for dealing with mixtures in regulatory practice, irrespective of (often presumed) modes of action. Recommendations. It is recommended to use a tiered framework analysis with CRA derived from DA also for the assessment of mixtures of dissimilarly acting chemicals. At lower tiers of the analysis, all chemicals deemed relevant for the exposure scenario under investigation should be assessed, irrespective of their presumed modes of action. At higher tiers, when the risk estimates at lower tiers are deemed unacceptable, chemicals known not to contribute to a relevant common adverse outcome can be excluded from the analysis. By way of further refining the analysis, criteria for the grouping of chemicals into common assessment groups based on their capability of affecting a common adverse endpoint, should be applied. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

5 Table of Contents Abstract... 2 Summary... 3 Table of Contents... 5 Background Terms of reference Acknowledgements Introduction and Objectives Introduction Project scope Guiding questions Project tasks Objectives Materials and Methods Data collection Systematic literature search Description of the search strategy that was adopted Development and implementation of the literature search strategy Aim of the search strategy Search strategy details Search sources Document types: primary articles and reviews Use of MeSH terms Filtering by journal or research field Synonym selection, preliminary searches Reasons for differences in number of results between PubMed and WoK Duplicate detection and removal Relevance of results to the topic of this report Import of results into the project database (CRADIS) Ad hoc approaches The CRADIS database Initial analysis (linking to chemicals) Mixture analysis Deep analysis Results TASK 1: information collection CRADIS database, current status and contents TASK2: state of the science summary Results of CRADIS analysis Number of mixture components in experimental studies Publication date of included studies Chemicals included in experimental mixture studies Primary literature Experimental studies relevant to chemical dissimilarity Non-chemical stressors Experimental studies relevant to low dose Experimental studies relevant to both dissimilar chemicals and low dose ( dual relevance )43 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

6 Analysis of chemicals and mode/mechanism of action information in dual relevance studies Method development Risk assessment approaches Mathematical models Definitions of dissimilar mode of action UK Committee on Toxicity (2002) European Food Safety Authority (2008) US Environmental Protection Agency (2002) Agency for Toxic Substances and Disease Registry (2004) WHO/IPCS (2009) Norwegian Scientific Committee for Food Safety (2008) Legislation Consideration of possible cumulative risk in EU pesticide regulations Legal background Consideration of cumulative risk in EU legislation Refusal of the authorisation of PPP Assessment concepts for mixture toxicity and empirical evidence for their validity Independent action (IA) Applicability of IA to mixtures composed of agents with dissimilar modes of action Data requirements for using IA Under IA, when is a mixture risk acceptable? Correlation assumptions in IA and their consequences Empirical evidence for IA Dose addition (DA) Data requirements for using DA Under DA, when is a mixture risk acceptable? The role of mode and mechanism of action information in the choice between IA and DA Definitions of mode and mechanism of action Use of empirical evidence to infer similarity or dissimilarity Conclusions Quantitative differences between DA and IA predictions Problem formulation Empirical evidence Limiting factors and resulting maximal differences The number of mixture components (n) The dose ratio of mixture components The slopes of dose response curves The effect of threshold assumptions Sample calculations and simulations Deterministic simulations of extreme prediction differences Probabilistic simulations of distributions of prediction differences Conclusions Toxicological interactions Approaches to cumulative risk assessment methods Approaches based on dose addition (DA) Hazard Index Toxic Unit Summation Point of Departure Index Relative Potency Factors Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

7 Toxic Equivalency Factors Data requirements and applicability of the cumulative risk assessment methods Approaches based on independent action (IA) Use of the TTC in CRA approaches The TTC concept Use of the TTC in cumulative risk assessment Cumulative risk from sub-adi levels Extrapolation from NOAEL or Benchmark Dose to ADI x 10 Assessment Factors Interspecies Assessment Factor Intraspecies Assessment Factor Continuous Responses Quantal (Dichotomous) Responses Conclusion The issue of non-adverse effects at levels below ADIs Effects below a critical effect size Issues arising from the dichotomization of continuous effect variables Precursor Effects Biochemical Alterations Conclusions Data requirements in legislation and availability of such data Toxicology data available to regulators Legal requirements for the provision of toxicity data for food improvement agents Legal requirements for the provision of toxicity data for substances for nutritional purposes Legal requirements for the provision of toxicity data for food contact materials Legal requirements for the provision of toxicity data for novel foods and novel food ingredients Legal requirements for the provision of toxicity data for residues Legal requirements for the provision of toxicity data for contaminants Legal requirements for the provision of toxicity data for undesirable substances in animal nutrition Legal requirements for the provision of toxicity data for feed additives Summary and conclusions Exposure data available to regulators Legal requirements for the provision of exposure data for pesticide residues Legal requirements for the provision of exposure data for substances in live animals and products of animal origin Legal requirements for the provision of exposure data for food improvement agents Legal requirements for the provision of exposure data for food supplements Legal requirements for the provision of exposure data for food contact materials Legal requirements for the provision of exposure data for food contaminants Summary and conclusions Cumulative exposure assessment (CEA) TASK 3: Grouping criteria TASK 4: Assess approaches to CRA for dissimilarly acting chemicals, evaluate the dose additivity approach Existing approaches to CRA for dissimilarly acting chemicals Commonalities of approaches Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

8 14.2. Practical implications for CRA of chemicals with diverse uses and properties and present in foods Assessment frameworks and tiering TASK5: propose an approach to CRA (dissimilarity) A tiered framework analysis for combinations of dissimilarly acting chemicals A unified approach to cumulative risk assessment for similarly and dissimilarly acting chemicals Principal assumptions, simplifications and requirements Rules for step-wise refinements as the analysis moves to the next higher tier Elements of cumulative risk assessment in a tiered framework analysis Framework analysis: initial considerations Tier Tier Tier Tier When should the risk characterisation step result in risk management measures? Criteria for the grouping of chemicals in CRA at higher tiers Case studies Case 1: pesticides and contaminants in lettuce Assessment Scenario A Assessment Scenario B Assessment Scenario C Higher Tier Assessments Conclusions Case 2: Mycotoxins in food commodities Case 3: dietary exposure to pesticides Description of dataset Tier 1: HI analysis (all ADIs) Interpretation of HI and HQ values HI analysis using TTC values (pseudo tier 0 analysis) PODI (NOAELs) Tier 2 (consideration of effects) Breakdown of HI analysis using PPDB health issues Tier 3 (group according to known or plausible toxicological independence) Conclusions Conclusions and Recommendations Conclusions Conclusions Identification of data gaps Recommendations References References cited in this report Legislation and regulations cited (sections 12 and 13.9) References included in the CRADIS database Appendices Appendix A Appendix B Appendix C Glossary / Abbreviations Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

9 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

10 Background As provided by EFSA, section I.1.2 BACKGROUND INFORMATION, Tender specifications for project CFT/EFSA/PPR/2010/02) Regulation (EC) No. 396/2005 on Maximum Residue Levels (MRLs) of pesticides in or on food and feed emphasises the importance of carrying out further work to develop a methodology to take into account cumulative and synergistic effects of pesticides for dietary risk assessment. As there are currently no harmonised or internationally agreed methodologies to assess risks from exposure to more than one compound, EFSA is developing methodologies to assess risks arising from exposures to more than one active compound in food. On 28/29 November 2006, EFSA started working on cumulative risk assessment of pesticides by organising a colloquium on Cumulative risk assessment of pesticides to human health: the way forward. Based on the results from this international event, EFSAs Scientific Panel on Plant Protection Products and their Residues (PPR Panel) elaborated an opinion to evaluate the suitability of existing methodologies and, if appropriate, identify new approaches to assess cumulative and synergistic risk from pesticides to human health with a view to setting MRLs for those pesticides within the framework of Regulation (EC) No. 336/2005 (EFSA, 2008). Following the general opinion of the PPR Panel, a worked example was developed for a group of triazole compounds and the results were reported in a separate opinion with suggested additional refinements of the methodology (EFSA, 2009). In this opinion the PPR Panel also proposed criteria for inclusion of compounds in a cumulative assessment group (CAG), highlighting the possibility of different levels of refinement in a step-wise approach. Before practical implementation of cumulative risk assessments for regulatory purposes, the PPR Panel is currently working on a third opinion identifying pesticides that can, based on their structure and effects, be grouped together for cumulative risk assessment. In the preparatory phase of this opinion, the PPR Panel outsourced the information collection aimed at establishing a database with cumulative assessment groups. As an additional progress in the hazard assessment for cumulative risk assessment, the state of the science on combined actions of pesticides and chemicals with dissimilar mode of action, including endocrine disruptors, should be investigated. Evidence in recently published scientific literature shows that certain endocrine disruptors show a dose additive effect even if they do not share the same primary molecular target (Kortenkamp, 2007 and the papers there reviewed; Christiansen et al. 2008; Moretto, 2008; Anonymous, 2009; Jacobsen et al, 2010; Reffstrup et al, 2010). Considering these results, the criterion for establishment of cumulative assessment groups of pesticides and chemicals in or on foods with dissimilar modes of action leading to a common effect, should be further elaborated. This should apply in particular, but not only, to pesticides interacting with or affecting the endocrine system. It has also been postulated that this type of combined toxicity can be addressed by the doseaddition approach (EFSA, 2008; University of London, 2009). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

11 References: Anonymous, Expert workshop on combination effect of chemicals, Hornbaeck, Denmark, organized by Danish Ministry of the Environment and the Danish Environmental Protection Agency. Christiansen S., Scholze M., Axelstad M., Boberg J., Kortenkamp A., Hass U., Combined exposure to anti-androgens causes markedly increased frequencies of hypospadias in the rat. International Journal of Andrology, 31, EFSA (European Food Safety Authority), Scientific Opinion of the Panel on Plant Protection Products and their Residues (PPR Panel) on a request from the EFSA evaluate the suitability of existing methodologies and, if appropriate, the identification of new approaches to assess cumulative and synergistic risks from pesticides to human health with a view to set MRLs for those pesticides in the frame of Regulation (EC) 396/2005. The EFSA Journal (2008) 704, EFSA (European Food Safety Authority), Scientific Opinion of the Panel on Plant Protection Products and their Residues (PPR Panel) on risk assessment for a selected group of pesticides from the triazole group to test possible methodologies to assess cumulative effects from exposure through food from these pesticides on human health. The AFSA Journal (2009) 7(9), Jacobsen P. R., Christiansen S., Boberg J., Nellemann C., Hass U., Combined exposure to endocrine disrupting pesticides impairs parturition, causes pup mortality and affects sexual differentiation in rats. International Journal of Andrology, 33, Kortenkamp, A., Ten years of mixing cocktails: a review of combination effects of endocrine-disrupting chemicals. Environmental Health Perspectives, 115, Moretto A., Exposure to multiple chemicals, when and how to assess the risk from pesticide residues in food. Trends in Food Science & Technology, 19, S56-S63. Refstrup T. K., Larsen C. J., Meyer O., Risk assessment of mixtures of pesticides, Current approaches and future strategies. Regulatory toxicology and Pharmacology 56, University of London, 22 December State of the Art Report on mixture toxicity, Final Report. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

12 Terms of reference As provided by EFSA, tender specification to project CFT/EFSA/PPR/2010/0, sections I.1.3 and I.1.4 PURPOSE OF THE ASSIGNMENT Overall objective(s): The overall objective of the contract resulting from the present procurement procedure is to provide scientific information on different aspects of combined actions of chemicals in food acting through dissimilar modes of action and to define criteria regarding the elaboration of cumulative assessment groups of pesticides which do not necessarily share a common mechanism or mode of action. The review should be applied to chemicals present in food in general and not restricted to pesticides in order to obtain wider information. The information will be used by the EFSA PPR Panel to further refine the hazard assessment for dietary cumulative risk assessment. Specific objectives: The specific objectives of the contract resulting from the present procurement procedure are as follows: Information gathering on the state of the science on combined actions of chemicals with dissimilar mode of action, including endocrine disruptors. The starting point for literature search should be the State of the Art Report on Mixture Toxicity published by the European Commission (EC) in 2009, available on the EC Environment website. The contractor should also consider the report on Risk Assessment of Mixtures of Pesticides and Similar Substances (Committee on Toxicity, UK, 2002) and the Opinion of the Scientific Steering Committee of the Norwegian Scientific Committee for Food Safety, Combined Toxic Effects of Multiple Chemical Exposures (VKM Norway, 2008) to avoid any duplication of works. Elaboration of general criteria for establishment of cumulative assessment groups of pesticides and chemicals in or on food acting through dissimilar modes of action leading to a common effect. This should apply in particular, but not only, to pesticides interacting with or affecting the endocrine system. Proposal for science-based approach for performing cumulative risk assessment of chemicals in food acting through dissimilar modes of action, including endocrine active substances. Identification of research needs to improve the understanding of the relevance of independent action for mixture toxicity. SCOPE OF THE WORK, EXPECTED OUTCOMES AND DELIVERABLES, TIMELINE AND PAYMENTS Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

13 The contractor selected as a result of this tendering procedure is expected to carry out the following tasks: Collect and scrutinize all recently published information from scientific literature relevant to combined actions of chemicals showing dissimilar mode of action and not covered by the EC report. Particular emphasis will be given to data at low dose (below the NOAEL). Provide a comprehensive summary and assessment of the collected information in view of understanding o the nature and site of actions, o possible targets, o interactions with different subsystems of the affected organisms and the different types of the possible combined actions when the dissimilar mode of action is to be considered. During this activity the references valid for low dose mixtures in EC, COT and VKM reports should be identified. The identified studies in these reports should be deeply analysed if -the compounds have common phenomenological end points, -the tested concentrations were appropriate to trigger observable effects, -the protocols had really the aptitude to validate the independent action theory, -the observed effects were analysed considering the need of different approaches to different types of dose-response data (continuous, quantal or ordinal). The toxicological endpoints and type of effects relevant for the cumulative risk assessment based on a dissimilar mode of action should be identified in particular, but not only, for chemicals interacting with or affecting the endocrine system. Propose scientific criteria for establishing cumulative assessment groups of pesticides and other types of chemicals when dissimilar mode of action is a relevant mechanism leading to a common effect, in order to perform dietary cumulative risk assessment for regulatory purposes. Assess the available approaches and methods for risk assessment of mixtures of pesticides and other chemicals in or on foods showing dissimilar mode of action. In the report on mixture toxicity (EC, 2009) the dose addition approach was considered to be probably conservative enough in most of the cases. Scrutinizing the collected information the contractor should identify cases, if any, where this approach would be underconservative. Propose a science-based approach for performing cumulative risk assessment of chemicals in food acting through dissimilar modes of action. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

14 Provide a structured report of the requested activities. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

15 Acknowledgements This contract was awarded by EFSA to: The School of Pharmacy, University of London (ULSOP) Contract/grant title: Investigation of the state of the science on combined actions of chemicals in food through dissimilar modes of action and proposal for science-based approach for performing related cumulative risk assessment Contract/grant number: CT/EFSA/PPR/2010/04 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

16 Introduction and Objectives INTRODUCTION This is the final report for the project titled Investigation of the state of the science on combined actions of chemicals in food through dissimilar modes of action and proposal for science-based approach for performing related cumulative risk assessment (CFT/EFSA/PPR/2010/02). 1. Project scope The project was specified so as to have relevance for: Dietary cumulative risk assessment; which could include for example pesticides, contaminants, additives, contact materials and toxins Regulatory purposes; which encompasses regulatory requirements and legal background Current practice; for example having consideration for the EFSA practice on CRA for similarly acting chemicals, as shown by the 2008 Opinion (EFSA 2008b) and the 2009 triazole case (EFSA 2009c). 2. Guiding questions In addition to the specified project tasks and objectives, we have identified a number of guiding questions which we have used to direct and structure our work on the project and this report. The guiding questions were: What theories and definitions are available or applied to considerations of dissimilar action in the context of cumulative risk assessment? Is there an approach to mixture assessment based on the independent action (IA) approach? In 2009, there was no available mammalian example of an experimental situation when IA gave a different effect prediction to dose addition (DA) and was also accurate (EC 2009). Has this position changed in scientific literature since 2009? Is there an experimental example in the scientific literature when IA provided the more conservative prediction, compared to DA? When using the IA concept, when is risk considered acceptable? Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

17 3. Project tasks The project was organized into 6 tasks, the tasks and their relationship to each specific objective (SO) of the project are shown in Figure 1, and are listed here: Task 1 was to collect and scrutinize the relevant scientific literature. The major output of this task is a Microsoft Access database containing the relevant literature. Task 2 was to summarize and assess the state of the science. Particular attention has been paid to definitions of dissimilar mode of action and to legislation relating to cumulative risk assessment. The state of the science summary underpins the subsequent Tasks 3, 4 and 5; which feed directly into Task 6 (Final Report). Task 3 was to propose scientific criteria for establishing cumulative assessment groups of pesticides and other types of chemicals when dissimilar mode of action is a relevant mechanism leading to a common effect, in order to perform dietary cumulative risk assessment for regulatory purposes. Task 4 was to assess the available approaches and methods for risk assessment of mixtures of pesticides and other chemicals in or on foods showing dissimilar mode of action. In the report on mixture toxicity (EC, 2009) the dose addition approach was considered to be probably conservative enough in most of the cases. Scrutinizing the collected information the contractor should identify cases, if any, where this approach would be underconservative. Task 5 was to propose a science-based approach for performing cumulative risk assessment of chemicals in food acting through dissimilar modes of action. Task 6 was to provide a structured report of the requested activities OBJECTIVES The specific objectives listed in the project tender specifications from EFSA are listed in Figure 1, which also shows how the objectives are related to the project tasks. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

18 SO1. Information gathering Task 1: collect recent studies, focus on dissimilar mode of action, low dose issues Task 2: summarise and assess collected data SO2. Elaboration of general criteria Task 3: propose grouping criteria Task 4: assess available approaches for dissimilar mode of action, assess appropriateness of dose additivity approach SO3. Proposal for science based approach Task 5: propose approach for cumulative risk assessment of chemicals acting through dissimilar modes of action SO4. Identification of research needs Task 6: provide report on requested activities Figure 1: Relationship between project tasks and the specific objective (SO) of the project Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

19 Materials and Methods 4. Data collection Data collection (Task 1) was performed in a systematic fashion, and was organised around a literature database (Microsoft Access), referred to as the CRADIS database, which stands for Cumulative Risk Assessment of DISsimilarly acting chemicals. The literature database was populated with two strategies: firstly, a systematic literature search was carried out with the aim of identifying peer-reviewed literature relevant to the application of independent action (IA) as a concept in mixture toxicology; secondly, ad hoc approaches were used to identify documents pertinent to the project but not identified in the systematic search. In particular, studies relating to mixtures of chemicals considered to be dissimilar, and studies relating to low-dose issues were sought. The data collection strategy is illustrated in Figure 2. In the remainder of this methods section the systematic search is described in section 5, ad hoc approaches in section 6 and the resulting literature database is described in section 7. Literature search focused on independent action, and synonyms PubMed & Web Of Knowledge Relevant articles from key reports COT, 2002; VKM, 2008; ULSOP, 2009 Ad hoc searches Author names Keywords Expert input Selected journals CRADIS Bibliographic data Abstract Article annotation Screen for relevance, document decision Obtain full text Link chemicals Deep analysis Mixtures, dissimilarity, low dose Figure 2: Schematic diagram showing the data collection strategy Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

20 5. Systematic literature search 5.1. Description of the search strategy that was adopted This section describes the important details of the search strategy that was adopted. The following section (section 5.2) provides more detail about the development and refinement of the strategy. The search was performed in Web of Knowledge (Thompson Reuters). A cut-off date of March 2011 was agreed at the First Coordination Meeting (Parma, 17 th January 2011) and the search was updated in April and May 2011 to ensure coverage of literature published, and indexed, up to the cut-off date. The search phrases used are listed in Box 1. Box 1: search strategy The following quoted phrases were used to query Web of Knowledge (Thomson Reuters, wok.mimas.ac.uk): independent action independent joint action effect summation effect addition response addition response summation bliss independence simple similar action effect multiplication * response additivity ** Target databases included: Medline (1950 present), Web of Science (1970 present), BIOSIS Previews (1969 present) All phrases were joined with an OR. Accepted document types were: article, review *note that searches showed all results (6) for this term were. ** Inclusion of this term added 18 articles, of which 6 were considered relevant. The most recent article was published in 2009, preceded by 2 published in 2005; suggesting that response additivity is not a commonly used term in current parlance. The search had a deliberately wide focus to reduce the chance of omitting relevant studies, consequently a large number of articles (approximately 1,150) were identified. The abstract of each article was obtained and scanned to assess the relevance of the article to the project. Articles were classified as follows: : articles that are not relevant, and that were returned due to the use of search phrases in other contexts than mixture toxicology Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

21 ON domain: articles that had relevance to the project. These results were further classified as o Trivial: articles that had some relevance but that were of minor use to the project aims o Experimental: articles that described an experimental, usually of chemical mixtures, in which mixture toxicity concepts, including independent action, were employed. o Method development: articles that developed methodology for dealing with mixture, or frameworks and approaches to cumulative risk assessment o Reviews The classification was performed in a secondary Microsoft Access database, and a report from that database, showing the classification of each article, is included as Appendix A. Around 79% of the search results were classified as off domain, this was due to the deliberately wide focus of the search. The use of an Access database allowed this sorting of many irrelevant hits to be done efficiently and with full documentation of the process (Appendix A). The remaining documents, around 250, were considered potentially relevant to the project, however around 17% were classified as having minor relevance, and were not analysed further. This produced 203 documents for further consideration, of which 80% dealt with experimental studies, 9% dealt with method development and 10% were relevant reviews. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

22 5.2. Development and implementation of the literature search strategy This section provides further detail of the development and refinement of the search strategy stated in section Aim of the search strategy The aim was to identify a suitable search strategy that could be used prior to a manual inspection and filtering of the results, ultimately producing a relevant set of documents without major omissions. For this purpose it is acceptable if the search initially includes false positives (i.e. results include articles that are not relevant) but not if there are many false negatives (i.e. results do not include relevant articles). This is because manual checking will allow for the elimination of false positives, but cannot compensate for documents that are missing due to being false negatives. False negatives can be dealt with by expert input of domain specific knowledge or by serendipitous/ad hoc browsing through the literature, for example looking at other articles in the same journal issue as a relevant article can identify other relevant articles that were not returned by the search process used. That aim of refining the search strategy was to identify which search terms should be included to give the best chance of retrieving documents relevant to the independent action model used in mixture toxicology. This includes identification of useful synonyms, and suitable combinations thereof, for example are relevant documents found by one important keyword or is the use of several synonyms necessary. The use of synonyms is not consistent in the field of mixture toxicity, therefore some terms may not be true synonyms, in that some authors may use them to refer to another concept, however they must be included in the search strategy so as not to discount the instances where they are used to refer to the concept of interest here. For those cases when the synonym is phrase is used but is not used as a synonym then manual inspection will identify them as off-domain Search strategy details Search sources Two sources were evaluated: PubMed (public MEDLINE) which is publically available from the US National Library of Medicine and Web of Knowledge, which is available through subscription (Thomson Reuters). Searches were carried out using the web interfaces provided, for PubMed ( the main search box on the home page was used. For WoK ( the search box for the topic field was used. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

23 Document types: primary articles and reviews Both data sources, PubMed and WoK, tag articles according to the document type, for example primary article, review, conference proceeding etc. We considered the possible use of these tags to limit the search strategy. In PubMed the use of Review[pt], where [pt] indicates publication type, in the search string should limit the results to review articles, however this approach was not adopted since visual inspection of the results showed that some articles tagged in this way were in fact primary research papers. In WoK, selection of document type = REVIEW, also included primary research papers. In WoK, a greater number of document types are present, for example conference proceedings and book chapters, that are not generally present in PubMed. It was desirable to remove these types of results since they increase the apparent number of results without being likely to contain relevant information, for example, although conference proceedings can be an indicator of ongoing research, the level of detail provided is not usually sufficient for thorough analysis, and the item may not have undergone full peer review. Consequently, tags were not used in PubMed and the counts shown in the table below are for all results and, in WoK, document types of ARTICLE and REVIEW were included/selected but MEETING, ABSTRACT and other types were excluded Use of MeSH terms MeSH terms can be a useful way to separate articles that include a search phrase being used in comm. on usage from those that use the phrase as specialist terminology. A preliminary search for important terms ( concentration addition, dose addition and independent action ) indicated that the MeSH vocabulary lacks these terms for mixture concepts and that therefore database curators (PubMed, MEDLINE and WoK) will not have been able to use MeSH terms to tag articles as relating to specific mixture concepts. MeSH terms were therefore not used as part of the search strategy Filtering by journal or research field Certain databases provide a field for indicating the area of research, and this may be a way to limit the number of irrelevant documents that are found in a search. However preliminary investigation of this option suggested that the process is not transparent (for example, it is not clear how journals and/or articles are assigned to the research field), and it is not easy to identify journals that definitely will or will not contain relevant documents. Consequently this option was not utilised since it might have led to the omission of relevant documents Synonym selection, preliminary searches To identify relevant synonyms and the extent to which searches in PubMed and WoK produce overlapping results, a series of searches were carried out using selected search strings in both sources and the number of results ( hits ) was compared. The results of searches for independent action and synonyms are shown in Table 1. For the purposes of Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

24 comparison, similar searches to those listed above were carried for concentration addition and synonyms thereof. These results are shown in Table 2. This analysis indicated that few synonyms were used redundantly, and that all synonyms should be used to give the best chance of the search identifying relevant documents Reasons for differences in number of results between PubMed and WoK Initial searches indicated that WoK and PubMed searches typically produce different numbers of results. Generally, WoK produced more results than PubMed. Visual inspection suggests that some of these additional results are indeed relevant, i.e. they are not a consequence of including irrelevant results, and that they should not be omitted. A potential reason for these differences is the journals that are indexed, in particular PubMed is directed towards the medical literature, although it is by no means limited to medical journals. PubMed have also adopted an updating model that allows journals to submit entries which are almost immediately available online with minimal checking by PubMed/NLM, in contrast WoK does their own curation of database records, for example indexing the references that are cited by an article, and this process takes time. The additional curation by WoK provides additional features at the cost of immediacy. WoK was adopted as the main source of searches in this project because of its greater coverage, and because it also includes MEDLINE results Duplicate detection and removal The presence of duplicate results increases the redundancy of a database, and can have an impact of the amount of work involved in processing a set of database records. We therefore considered whether the search strategy resulted in duplicate records. We have not noticed the presence of duplicate records in PubMed, however WoK does include duplicate records. We were able to remove some of these by use of the duplicate identification algorithm included in the Reference Manager software that was used to process the results of the search strategy. The remainder were identified by sorting the database in alphabetical order by title and visual scanning to identify neighbouring identical titles and author lists. One explanation of duplicates could be the presence of an advance or e publication as well as the final authoritative print version, which often has a substantially later date of appearance. Around 30 documents were found to be duplicated, and these duplicate entries were removed. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

25 Relevance of results to the topic of this report After duplicate scanning, titles were read to decide if they were relevant to the topic: IA as a mathematical concept for the joint action of chemicals, usually in toxicology For this analysis articles were loaded into an access database for classification of their relevance (, ON domain (trivial, experimental, method development, review)). Classification was made on the basis of title, abstract (if available) and publication date (if an abstract was unavailable, articles before 2000 were not usually pursued unless the title/date indicated high relevance). Database name: classification of IA search results.accdb Table 3 shows the result of this analysis. Out of 1153 results from the literature search, 203 were considered relevant; this is shown in the table as ON domain and not TRIVIAL. 32 of these results were not found in PubMed (so 171 were in PubMed, and were linked to their PMIDs); so around 16% of the articles derived from WoK only, and using WoK in addition to PubMed identified 18% extra articles. Table xx lists the journals that articles came from that were only retrieved via WoK Import of results into the project database (CRADIS) The 203 articles that were relevant (ON domain and NOT trivial) were then imported to the CRADIS database for further analysis. Articles with a PMID (PubMed identifier) were imported automatically. A Visual Basic code was written that used the PMID to download citation details and abstract from the PubMed server in the tagged MEDLINE format and parsed the results into the appropriate CRADIS database fields), articles that did not have PMIDs were manually entered into the CRADIS database. 26 of the identified articles had already been entered into CRADIS and 32 articles did not have PMIDs (they were unique to Web of Knowledge, see Table 4); therefore 142 articles were loaded automatically using their PMID and 32 articles were manually loaded. The CRADIS database is fully described in section 7. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

26 Table 1: results of searches in PubMed and WoK for independent action and synonyms thereof PubMed Web of Knowledge Search string Hits (number Hits (number on Notes on domain *) domain *)*** Independent action / independent action /672 simple dissimilar action Not found** 2 Simple 11 (1) 29/23 (6) AND dissimilar AND action simple independent action 7 (6) 7/7 (6) PM: phrase used once in 2011, then 1998; focus of 6 articles was epidemiological and mathematical WoK: phrase last used in 1998 PubMed and WoK results differ by one paper each Independent joint action 12 (11) 23/22 (19) Response addition /156 Response summation 16 45/41 response addition 0 1 AND response summation Effect addition /272 WoK: 262 if language =English Effect summation 10 19/17 effect addition 0 0 AND effect summation response addition OR response summation OR effect addition OR effect summation /485 PM: results show there is no overlap There is only one paper with more than one of these phrases in it; a search within these results for independent action OR independent joint action OR effect summation OR effect addition OR response addition OR response summation / 1163 independent action = 20 hits bliss independence 28 55/30 effect multiplication 0 6 (0) WoK: All 6 results were OFF- DOMAIN Not a useful term *question used to decide relevance was did the paper deal with the mathematical concept used in mixture toxicology? **search returned no results. Search engine response was quoted phrase not found (PubMed) or Search Error: Invalid query. Please check syntax (WoK). *** shows aaa/bbb (n), when aaa = count of all results, bbb= count of results for ARTICLES and REVIEW document types, n=number of documents that were relevant. All database searches were performed on Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

27 Table 2: results of searches in PubMed and WoK for concentration addition and synonyms thereof PubMed Web of Knowledge Search string Hits (n) Notes Hits (n)*** Notes Dose addition /107 dose additivity 35 52/42 dose addition AND 4 6/6 Shows 6 papers overlap dose additivity Dose additive 56 73/56 Concentration /479 addition Concentration 11 22/18 additivity Concentration paper overlaps addition AND Concentration additivity concentration 21 45/44 additive dose addition AND concentration addition 0 PM: Note that this result shows there is NO overlap simple similar action 7 14/12 dose addition OR /645 dose additivity OR concentration addition OR concentration additivity OR simple similar action dose addition OR dose additivity OR dose additive OR concentration addition OR concentration additivity OR concentration additive OR simple similar action /714 *question used to decide relevance was did the paper deal with the mathematical concept used in mixture toxicology? **search returned no results. Search engine response was quoted phrase not found (PubMed) or Search Error: Invalid query. Please check syntax (WoK). *** shows aaa/bbb, when aaa = count of all results, bbb= count of results for ARTICLES and REVIEW document types All database searches were performed on Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

28 Table 3: results of classifying search results Notes on definition N All results Did not match topic, e.g. glucose 907 independent action of drug A, response summation and psychology, independent action of two genes (no mathematical concept implied) 2. there was only a title (no abstract) and the title did not have enough information to deem the article relevant 3. article in a foreign language (n=1) ON domain TRIVIAL 1. Title only, appears relevant from 43 the information in the title. If the title appeared very relevant then efforts were made to find the abstract or full text, especially if the document was recent (post 2000) EXPERIMENTAL STUDY 163 METHOD DEVELOPMENT 19 REVIEW 21 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

29 Table 4: list of journals from which articles came that were unique to WoK searches Journal title Number of articles Environmental Toxicology And Chemistry 7 Asian Journal of Ecotoxicology 4 Human And Ecological Risk Assessment 3 Biometrics 2 Journal Of Applied Ecology 2 Canadian Journal Of Fisheries And Aquatic Sciences 1 Chinese Science Bulletin 1 Continental Shelf Research 1 Crop Protection 1 Crop Research (Hisar) 1 Fresenius Environmental Bulletin 1 Green Chemistry 1 Journal Of Economic Entomology 1 Journal Of Environmental Quality 1 Phytopathology 1 Scandinavian Journal Of Statistics 1 Umweltwissenschaften Und Schadstoff-Forschung 1 Weed Research 1 Weed Technology 1 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

30 6. Ad hoc approaches Ad hoc approaches were used to allow flexibility in identifying literature that was not found in the systematic literature search but that was identified, for example, by expert input. Approaches included: identification of relevant documents from the core reports: COT, 2002; VKM, 2008 and EC, 2009 targeted literature searches for key authors searches of specific journals, issues expert input citation searches (Web of Knowledge, Thomson Reuters) We evaluated the utility of using citation searches, in which articles citing a key paper are located, rather than a search strategy based on keywords. However we found that this approach generated large numbers of results, was dependent on the seed key paper selected and relied on complete, comprehensive citation of prior papers by authors which is not usually done. Dependence on the seed paper can be reduced by using several or many seeds, however this increases the size of the output above manageable levels. Consequently citation searches were used as part of the ad hoc strategy, not used systematically. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

31 7. The CRADIS database Results from the search strategies described above were compiled into a Microsoft Access database, referred to as the CRADIS (Cumulative Risk Assessment of DISsimilarly acting chemicals) database. The database accompanies the Final Report and the filename is CRADIS.accdb. The database contains two main tables, one containing bibliographic information and analysis of articles ( ARTICLES ) and the other containing chemical information ( CHEMICALS ). Each table has a unique ID field that is used for linking between the tables. The ARTICLES table is linked to a sub table called MIXTURES which details a mixture experiment. The tables are presented to the database user through a series of forms, starting with a menu form that allows the user to access the database in three ways: 1. from an articles perspective, when the user starts from an article or list of articles and can then click-through to linked information in the MIXTURES and CHEMICALS tables 2. from a chemicals perspective, when the user start from a chemical of interest and can click through to linked information in the ARTICLES and MIXTURES tables 3. from a list of database reports. The user is presented with a list of preformatted reports that are built from the database contents. Initially all articles identified were listed in the database, then bibliographic, abstract and linking information (linking to PubMed through PubMed identifiers (PMID), linking to full text sources through Digital object identifiers (DOI)) was added. Articles were then classified as to their area of relevance to the key issues for this project: mixture studies, primarily those studies with relevance to chemical dissimilarity, and lowdose. Full text was sourced for relevant articles. Relevant documents were then annotated further, and subjected to deeper analysis as described next Initial analysis (linking to chemicals) Links to related chemicals were made by connecting entries on the ARTICLES table to entries on the CHEMICALS table. The use of linking allows chemical information, such as abbreviations, full names, mw, and ID numbers to be entered once only Mixture analysis For all studies, the number of mixture components, n, was documented. This result is shown on the reports page of the CRADIS database in a PivotTable showing the size of n in all of the mixture studies. When the included multiple mixtures, the highest n was entered. For mixtures including nonchemical stressors, the nonchemical stressor was included in the calculation of n. Selected studies, those with relevance to both dissimilarity and low dose Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

32 issues, were analysed in more detail and parameters such as mixture design, prediction models used, results observed, analyses used etc were documented, see section Deep analysis Those articles deemed relevant to both of the key issues for the project, low dose and dissimilarity, were subject to a deeper analysis of the mixture design and results. This is documented on the mixture and deep analysis tabs in the CRADIS database. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

33 Results TASK 1: INFORMATION COLLECTION Task 1 entailed the design, creation and population of an annotated literature database CRADIS (Microsoft Access). The database forms the basis of the scientific summary (Task 2) and provides source material for subsequent tasks. 8. CRADIS database, current status and contents The CRADIS database contains 266 articles, including primary literature, reviews and scientific reports. These are listed in the reference list (section 18). Virtually all of the articles have abstracts, and full text was sourced for 168 articles. All articles have been classified as to their area of relevance to the project. Chemicals tested in 173 mixture studies have been indexed to allow direct retrieval of all the studies relevant to a given chemical. In doing so, 631 chemicals have been indexed and linked to the experimental studies that included them. Mixture studies identified as relevant to both chemical dissimilarity and to low-dose issues have been subjected to a deep analysis, as specified in the project proposal and the Tender Specifications. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

34 TASK2: STATE OF THE SCIENCE SUMMARY This section presents a summary of the state of the science, which is the primary output of Task 2. The summary focuses on an analysis of the published mixtures studies that were compiled into the CRADIS database (see Task 1) and a review of the critical issues in the field. This review served as the basis for the work performed in fulfilment of project tasks 3, 4 and Results of CRADIS analysis 9.1. Number of mixture components in experimental studies Figure 3 presents a pie chart showing the number of components examined in each of the experimental mixture studies contained within the CRADIS database. This analysis shows that the majority (52%) of studies examined binary mixtures (2 components), and that less than 25% of all the studies examined mixtures with more than six components. The highest number of components studied was 33 (Hermens et al. 1985). This focus on binary mixtures most likely reflects the intention of developing mixture concepts rather than being a practical attempt to screen chemicals, since it is evident that the systematic examination of binary mixtures of all possible chemical combinations would be unfeasible. Mixture studies with few components have a low ability to determine whether DA and IA are better predictors of observed data because the two predictions are more likely to be close to each other as the number of components decreases (see section 13.4). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

35 Figure 3: pie chart showing the number of components investigated in experimental studies of mixtures Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

36 9.2. Publication date of included studies The bar chart below indicates the publication year of the articles and reviews included in the CRADIS database. The upward trend reflects both the growing number of publications in this field and the intentional focus on recent literature in this state of the science summary. Figure 4: bar chart showing publication year of articles included in CRADIS Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

37 9.3. Chemicals included in experimental mixture studies In addition to the analysis required by the tender specifications, the chemicals tested in all mixture studies (173) for which details were available were indexed and linked to the experimental studies in which they were tested. The indexed chemicals are listed in Appendix B, and the frequency with which chemicals were tested is indicated graphically in a Wordle diagram below. Figure 5: Wordle diagram showing a visual representation of the chemicals examined, and the frequency of their inclusion, in experimental studies Figure shows the chemicals tested in mixture studies in a Wordle plot ( in which word size is proportional to the number of studies that the chemical was tested in. For orientation to the relationship between word size and inclusion, copper was included in 12 studies whilst bisphenol A was included in 6. The full list of chemicals is provided in the CRADIS database and in Appendix B of this report. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

38 10. Primary literature The CRADIS database contains 174 experimental studies of mixtures, of which 94 were classified as having relevance to dissimilar chemicals and 26 were classified as having relevance to low dose issues. 13 studies were classified as relevant to both dissimilar chemicals and low dose issues. Each of these literature groups is now summarized Experimental studies relevant to chemical dissimilarity We identified and listed the criteria used by authors of mixture studies relevant to dissimilarity to assign similarity or dissimilarity. Note that at this stage the term dissimilarity, and to some extent by implication also similarity, encompasses most possible usages of the term (s). These results are found in the MixGroupingCriteria field in the Articles table of the CRADIS database. The main approaches were: Mode/mechanism of action/toxicity. In most cases a framework for establishing the mode/mechanism of action was not stated, and we expect that different authors performed this analysis in different ways. This may be unsurprising since it has been often noted that there is no current framework or established vocabulary for describing mechanism of toxicity. Chemical grouping, for example as metals, pesticides, herbicides, insecticides and ionic liquids. We note that terms such as pesticide, herbicides, and insecticides refer to a common use of chemicals with potentially diverse structures and probably should not be considered as groups of structurally related chemicals unless further clarification is provided, for example neonicotinoid insecticides. Opposing actions, mixtures of chemicals with directly opposing (cytochrome p450 inducers and inhibitors (Scott and Hodson 2008)) or indirectly opposing (mixtures of estrogens and antiandrogens (Eustache et al. 2009)) actions. Approximately half of the studies used some version of mode/mechanism of action whilst around one third used an approach based on some kind of chemical grouping Non-chemical stressors In addition to studies of combinations of chemicals, seven studies that included chemicals with a nonchemical stressor were identified and considered as being relevant to dissimilar action. These studies are analysed in Table 5. All of the studies were ecotoxicological in focus. All of the combinations studied were binary and included combinations of one chemical (including carbaryl, carbendazim, cadmium, fluoranthene, imidacloprid, nickel, nonylphenol or potassium dichromate) and one nonchemical stressor (including low/high temperature; low dissolved oxygen levels, low soil moisture, perceived predation risk/threat, parasitism and low photon flux). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

39 All of the studies compared experimental results with a prediction based on IA, whilst around half of the studies also compared to a prediction based on DA. In most cases IA was considered as the reference model, as opposed to the situation for studies of chemicals only when DA is often considered as the reference model, even if IA is also frequently evaluated. The inclusion of nonchemical stressors in CRA may therefore change the appropriate position of IA in CRA approaches, as described in later project tasks. The inclusion of nonchemical stressors in CRA has been called for, but has not been done to date (NRC 2009). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

40 Table 5: analysis of mixture studies that included nonchemical stressors Study chemical Nonchemical stressor Organism Mixture concepts used (Ferreira et al. 2010) Nickel Low/high temperature; Daphnia magna Straus IA, DA low dissolved oxygen levels (water flea) (Jensen et al. 2009) Nonylphenol Low/high temperature Dendrobaena octaedra (earthworm) (Long et al. 2009) Fluoranthene low soil moisture Lumbricus rubellus (earthworm) (Pestana et al. 2009) Imidacloprid perceived predation risk Sericostoma vittatum (caddisfly); Chironomus riparius (midge) (Coors and De Carbaryl predation threat; Daphnia magna (water Meester 2008) (Ferreira et al. 2008) (Cleuvers et al. 2002) Cadmium, carbendazim Potassium dichromate parasitism low dissolved oxygen levels low photon flux flea) Daphnia magna Straus (water flea) Scenedesmus subspicatus (green alga) IA IA IA, DA IA IA, DA IA, DA, effect summation Conclusions of authors Deviations from IA were noted (both synergy and antagonism) whilst DA could be a poor fit to the observed data. IA was used as the reference model, the authors observed strong synergism with nonylphenol and high but not low temperature. IA provided a good description of the combined stressor data and was the most parsimonious model describing joint effect The majority of parameters showed no deviation from IA or DA IA was found to provide useful, quantitative predictions of effect Observed adequate prediction by IA in some cases, underprediction by both IA and DA in other cases, and also observed antagonism from IA. Combination effects could be calculated well only by IA Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

41 10.2. Experimental studies relevant to low dose We compiled the definition of low dose used by 27 studies, and this information can be found in the LowDoseDefinition field of the Articles table in the CRADIS database and is also shown in Table 6. Ten studies used no observed effect concentrations (NOEC), sometimes with an estimation that this was equivalent to an EC 01, and a further seven used less than NOEC or a fraction of NOEC (one third, one quarter or one tenth). These studies are listed in the following table. Two studies used a set effect level, of EC 05 (Deng et al. 2007) or EC 10 (Merino-Garcia et al. 2003), and one used 0.04 times the LC 50 (Hermens et al. 1985). Other studies used the MRL (Wade et al. 2002) or ADI (Ito et al. 1995) and two studies used a measure of human levels, human serum concentrations (van Meeuwen et al. 2007) or background human daily intake (Crofton et al. 2005). One used a set concentration, 1mg/kg/day (Eustache et al. 2009). This analysis shows that various definitions of low dose are in use in the literature, and that their meanings are sufficiently different to complicate comparison of studies described as low dose without further information and analysis. Table 6: Low dose definitions used in the primary literature Low dose definition Authors Pub. year NOEC, EC01 NOEC Backhaus T, Scholze M, Grimme LH 2000 NOEC Walter H, Consolaro F, Gramatica P, Scholze M, Altenburger R 2002 NOEC Kunz PY, Fent K 2006 NOEC, EC01 Kunz PY, Fent K 2009 NOEL, EC01 Manzo S, Buono S, Cremisini C 2010 chemical doses that had no observable effect when tested alone Christiansen S, Scholze M, Axelstad M, Boberg J, Kortenkamp A, Hass U NOEC (=EC01) Silva E, Rajapakse N, Kortenkamp A 2002 sub-effective concentrations, expected to be similar to EC01 ED01 EC01 Less than, or fraction of, NOEC <NOEC 2008 Belden JB, Lydy MJ 2006 van Meeuwen JA, van den Berg M, Sanderson JT, Verhoef A, Piersma AH Wang Z, Chen J, Huang L, Wang Y, Cai X, Qiao X, Dong Y Broderius SJ, Kahl MD, Elonen GE, Hammermeister DE, Hoglund MD <NOEC Broerse M, van Gestel CA Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

42 <NOEC, EC01 <NOEC, EC01 Faust M, Altenburger R, Backhaus T, Blanck H, Boedeker W, Gramatica P, Hamer V, Scholze M, Vighi M, Grimme LH Faust M, Altenburger R, Backhaus T, Blanck H, Boedeker W, Gramatica P, Hamer V, Scholze M, Vighi M, Grimme LH 1/10 NOAEL Jonker D, Woutersen RA, van Bladeren PJ, Til HP, Feron VJ 1/4 NOAEL Jonker D, Woutersen RA, van Bladeren PJ, Til HP, Feron VJ MinimumOAEL (MOAEL), NOAEL, 1/3 NOAEL ECx Groten JP, Schoen ED, van Bladeren PJ, Kuper CF, van Zorge JA, Feron VJ EC05 Deng Fc;Liu Ss;Liu Hl;Mo Ly; 2007 EC10 Merino-Garcia D, Kusk KO, Christensen ER 2003 fraction of LC x LC50 Hermens J, Leeuwangh P, Musch A 1985 MRL, ADI MRL, TDI, NOEL Wade MG, Foster WG, Younglai EV, McMahon A, Leingartner K, Yagminas A, Blakey D, Fournier M, Desaulniers D, Hughes CL ADI, 100xADI Ito N, Hasegawa R, Imaida K, Kurata Y, Hagiwara A, Shirai T Human levels human serum concentrations background human daily intake set concentration van Meeuwen JA, Ter Burg W, Piersma AH, van den Berg M, Sanderson JT Crofton KM, Craft ES, Hedge JM, Gennings C, Simmons JE, Carchman RA, Carter WH Jr, DeVito MJ 1 mg/kg/day Eustache F, Mondon F, Canivenc-Lavier MC, Lesaffre C, Fulla Y, Berges R, Cravedi JP, Vaiman D, Auger J Not clearly defined not clearly defined Pavlaki MD, Pereira R, Loureiro S, Soares AM 2011 "low concentration-response range" Petersen K, Tollefsen KE Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

43 10.3. Experimental studies relevant to both dissimilar chemicals and low dose ( dual relevance ) 14 studies were considered to have relevance to both of the main topics of this project, dissimilar chemicals and low dose. These studies were subject to a full analysis of their experimental content in the CRADIS database. Four of the studies were published in the review period of this report (2008 March 2011) and are reviewed below. A review of the experimental literature prior to this period can be found in the State of the Art Report on Mixture Toxicity (EC 2009). Key features of all fourteen studies, including those prior to the review period, are summarised in Table 7, and chemical and mode/mechanism information from all fourteen studies was analysed, see section Our analysis included comparing each experimental to five quality criteria that have been previously applied to mixture studies (EC 2009;Kortenkamp et al. 2007) This analysis is included in Table 7. The quality criteria are: A: toxicity of individual mixture components was experimentally determined under identical conditions as the mixture; B: stability of test concentrations under test conditions was checked by analytical methods (does not apply to animal experiments with direct dosing); C: uncertainty of experimentally determined effects, effect concentrations, or effective doses was estimated by statistical methods; D: uncertainty of mixture toxicity predictions was estimated by statistical methods; and E: no observed effect concentrations (NOECs) or NOELs were determined for every individual substance, and individual concentrations or doses resulting in the given joint effect were demonstrated to be at or below these NOECs or NOELs, or insignificance of individual effects was demonstrated by other statistical approaches (Kortenkamp et al. 2007). We note that the recent experimental studies published since the 2009 EC report have not significantly altered the state of the mixture toxicology field however they are reviewed briefly here because they were identified by the exhaustive search strategy that was adopted for the project. An important question for this project was whether there is empirical evidence of relevance to the use of IA in risk assessment, this question was not answered by these recent studies and the state of the science on this issue is reviewed separately in section Chronic dietary exposure to a low-dose mixture of genistein and vinclozolin modifies the reproductive axis, testis transcriptome, and fertility (Eustache et al. 2009). Eustache et al examined chronic exposure to low doses of endocrine disruptors and administered binary mixtures of a phytoestrogen, genistein, and an antiandrogen, vinclozolin Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

44 to male rats from conception to adulthood at doses of 1, 10 and 30 mg/kg/day. We cannot analyse this in detail because the authors did not use mixture toxicity concepts to formulate quantitative effect predictions, which is an important quality criteria for mixture studies (Kortenkamp et al. 2007). Integrated fuzzy concentration addition-independent action (IFCA-IA) model outperforms two-stage prediction (TSP) for predicting mixture toxicity (Wang et al. 2009). Wang et al. studied 12 industrial organic chemicals (Benzene, toluene, chlorobenzene, phenol, aniline, nitrobenzene, 3-nitrotoluene, 2,4-Dichlorophenol, hydroquinone, 3- nitrochlorobenzene, 4-nitrophenol, 2,4,6-trichlorophenol) with four different modes of toxic action (nonpolar narcotic, polar narcotic, pro-electrophile, oxidative phosphorylation uncoupling) with the aim of comparing the ability of an integrated fuzzy concentration addition-independent action (IFCA-IA) model with the two stage prediction (TSP) model. In the TSP model chemicals are grouped by their model of action, then firstly the effect of each group is calculated using DA (referred to by Wang at al as concentration addition (CA)), and secondly these group effects are cumulated using IA. Conversely, the IFCA-IA model does not use MOA information and instead uses fuzzy logic to calculate weight coefficients for DA and IA from molecule structure descriptors. Wang et al. exposed Vibrio fischeri to three mixtures: (1) in ratio of individual EC01 concentrations; (2) in ratio of individual EC50 concentrations; (3) in ratio of equimolar concentrations. The IFCA-IA model must be trained on a data set, and Wang et al. used the experimental results of their EC01 mixture, see next, as the training set. Wang et al. found that the IFCA-IA model outperformed the TSP model with prediction errors for the EC01, EC50 and equimolar mixtures of 0.3, 6 and 0.6% for IFCA-IA and 2.8, 19 and 24% for TSP. The requirement for training data means that the IFCA-IA approach is likely to have practical obstacles to application in human risk assessment when it is likely that most mixtures of interest will not have been tested experimentally. This is also reviewed in section , concerning recent method developments. Mixture effects of nickel and chlorpyrifos on Folsomia candida (Collembola) explained from development of toxicity in time (Broerse and van Gestel 2010). Broerse and van Gestel used IA and DA to the effects of a seven week exposure to binary mixtures of nickel and chlorpyrifos on Folsomia candida, a soil-dwelling Arthropod (collembola). The authors found that chlorpyrifos toxicity exhibited an extremely steep doseresponse curve that meant the data was unsuitable for use in IA or DA models, however they found that careful monitoring of toxicity allowed the determination of mixture effects, and reported that, even at exposure levels below the No Effect Concentration, chlorpyrifos was able to reduce the toxicity of nickel. This indicated the difficulty of confounding toxicity in mixture studies, but had few wider implications for the field. Effects of binary mixtures on the life traits of Daphnia magna (Pavlaki et al. 2011). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

45 Pavlaki et al. compared the toxicity of binary mixtures of imidacloprid with either thiacloprid or nickel in Daphnia magna. Imidacloprid and thiacloprid are both neonicotinoid insecticides, for which a common mode of action may be likely. For the mixture of two insecticides, the authors observed synergy (neonatal number) and conformance to DA (body length). For the mixture of insecticide with metal, the authors observed conformance to IA (neonatal number) and a dose dependent effect on body length with synergy at low dose and antagonism at higher doses. This provides an example of when IA can predict the mixture effect (neonatal number), however in this case the experimental data was also compatible with the DA prediction. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

46 Table 7: Summary and comparison of studies with dual relevance to the project. Title of paper Joint toxicity of mixtures of groups of organic aquatic pollutants to the guppy (Poecilia reticulata). 4-week oral toxicity of a combination of eight chemicals in rats: comparison with the toxicity of the individual compounds. Effect of ingestion of 20 pesticides in combination at acceptable daily intake levels on rat liver carcinogenesis. Subacute toxicity of a mixture of nine chemicals in rats: detecting interactive effects with a fractionated two-level factorial design. Effects of subchronic exposure to a complex mixture of persistent contaminants in male rats: systemic, immune, and reproductive effects. Mixture toxicity of priority pollutants at no observed effect concentrations (NOECs). Joint toxicity of similarly and dissimilarly acting chemicals to Daphnia Author list (Pub. Date) Hermens J, Leeuwangh P, Musch A (1985) Jonker D, Woutersen RA, van Bladeren PJ, Til HP, Feron VJ (1990) Ito N, Hasegawa R, Imaida K, Kurata Y, Hagiwara A, Shirai T (1995) Groten JP, Schoen ED, van Bladeren PJ, Kuper CF, van Zorge JA, Feron VJ (1997) Wade MG, Foster WG, Younglai EV, McMahon A, Leingartner K, Yagminas A, Blakey D, Fournier M, Desaulniers D, Hughes CL (2002) Walter H, Consolaro F, Gramatica P, Scholze M, Altenburger R (2002) Merino-Garcia D, Kusk KO, Christensen ER Notes Analysed in table 6.4 (EC, 2009) Analysed in table 6.5 (EC, 2009) Analysed in table 6.5 (EC, 2009) Analysed in table 6.5 (EC, 2009) Analysed in table 6.5 (EC, 2009) effects: systemic, immune, reproductive Analysed in table 6.4 (EC, 2009) "These findings strongly indicate, that the modes of action of mixture components are indeed dissimilar in a way, that is sufficient to predict the mixture toxicity with the concept of independent action." Quality criteria fulfilled A A C E C A B C E C E A B C E A C Low dose definition used 0.04 x LC50 1/10 NOAEL ADI, 100xADI MinimumOAEL (MOAEL), NOAEL, 1/3 NOAEL MRL, TDI, NOEL NOEC EC10 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

47 magna at different response levels. (2003) Joint algal toxicity of 16 dissimilarly acting chemicals is predictable by the concept of independent action. A comparison of the lethal and sublethal toxicity of organic chemical mixtures to the fathead minnow (Pimephales promelas). Thyroid-hormonedisrupting chemicals: evidence for dosedependent additivity or synergism. Integrated fuzzy concentration additionindependent action (IFCA-IA) model outperforms two-stage prediction (TSP) for predicting mixture toxicity. Chronic dietary exposure to a low-dose mixture of genistein and vinclozolin modifies the reproductive axis, testis transcriptome, and fertility. Mixture effects of nickel and chlorpyrifos on Folsomia candida (Collembola) explained Faust M, Altenburger R, Backhaus T, Blanck H, Boedeker W, Gramatica P, Hamer V, Scholze M, Vighi M, Grimme LH (2003) Broderius SJ, Kahl MD, Elonen GE, Hammermeister DE, Hoglund MD (2005) Crofton KM, Craft ES, Hedge JM, Gennings C, Simmons JE, Carchman RA, Carter WH Jr, DeVito MJ (2005) Wang Z, Chen J, Huang L, Wang Y, Cai X, Qiao X, Dong Y (2009) Eustache F, Mondon F, Canivenc-Lavier MC, Lesaffre C, Fulla Y, Berges R, Cravedi JP, Vaiman D, Auger J (2009) Broerse M, van Gestel CA (2010) Analysed in table 6.4 (EC, 2009) Effects: lethality, sublethality (32d growth) 30 individual components were tested, and several mixtures were tested: the largest mixture was 12 components but all had the same mode of toxic action. The largest mixture of chemicals with a different MTOA w DISSIM: estrogen plus antiandrogen effects: standard reproductive toxicology end points, testicular mrna expression profiles A B C D E A B C D A C A C A B <NOEC, EC01 <NOEC background human daily intake EC01 1 mg/kg/day <NOEC Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

48 from development of toxicity in time. Effects of binary mixtures on the life traits of Daphnia magna. Pavlaki MD, Pereira R, Loureiro S, Soares AM (2011) effects: reproduction, survival, body length A B not clearly defined Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

49 Analysis of chemicals and mode/mechanism of action information in dual relevance studies In fulfilment of Task 2, we have analysed the identified studies for information on the chemicals tested in mixture studies, and the presence of mode of action information. The chemicals tested in the 14 identified studies, see section 10.3 above, were listed and any information regarding mode/mechanism of action provided by the authors of the papers, for example as a basis for grouping, was compiled. 142 chemicals were included over the 14 studies, this information is listed in Appendix C. Our analysis revealed that only limited information on mode of action was included in any, and that no consistent approach is used in the experimental literature to describe mode of action information. Comparison of studies on the basis of mode of action is virtually impossible. 64 chemicals had some mode/mechanism stated, although this includes 8 that were stated as not known. The information for 56 chemicals, with 26 modes/mechanisms is presented in Table 8. None of the chemicals were tested in more than two studies and only 17 chemicals were included in two different studies. Of these 17 chemicals, 11 had no mode/mechanism information in either, and 6 had information from one only. No chemicals had mode/mechanism from both studies, and consequently the consistency of the schemes being used to assign this information cannot be assessed. None of the studies explicitly stated a scheme. The available information is presented in Table 9. It should be noted that the studies included both mammalian toxicology and ecotoxicology studies, therefore the modes/mechanisms listed derive from a wide range of species. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

50 Table 8: list of 26 modes/mechanisms and the 56 associated chemicals Mode/mechanism Chemical name Mixture details: First author, Pub. Year amino acid biosynthesis Metsulfuron-methyl Faust, 2003 Carotenoid biosynthesis Norflurazon Faust, 2003 DNA synthesis and function Nalidixic acid Faust, 2003 enzyme inhibitor ethyl-parathion Merino-Garcia, 2003 inhibition of photosynthesis Tributyltin chloride Walter, 2002 Triphenyltin chloride Walter, 2002 lipid biosynthesis Metazachlor Faust, 2003 Membrane functions DTMAC Faust, 2003 Narcosis I pentachloroethane Broderius, 2005 Narcosis II 2,4-dimethylphenol Broderius, 2005 narcotic nonylamine Merino-Garcia, 2003 decylamine Merino-Garcia, 2003 Nonpolar narcotic Benzene Wang, 2009 Toluene Wang, 2009 chlorobenzene Wang, 2009 Nucleotide biosynthesis Azaserine Faust, 2003 Oxidative phosphorylation 4-nitrophenol Wang, 2009 uncoupling 2,4,6-trichlorophenol Wang, 2009 Photosynthetic electron transport Paraquat dichloride Faust, 2003 Terbuthylazine Faust, 2003 photosystem II inhibitor Atrazine Walter, 2002 Polar narcotic Aniline Wang, ,4-Dichlorophenol Wang, 2009 Phenol Wang, 2009 nitrobenzene Wang, nitrotoluene Wang, 2009 porphyrin biosynthesis Aclonifen Faust, 2003 Pro-electrophile hydroquinone Wang, nitrochlorobenzene Wang, 2009 protein biosynthesis Chloramphenicol Faust, 2003 Proton translocation and ATP CCCP Faust, 2003 synthesis reactive toxicant Hexachlorobutadiene Broderius, 2005 Respiratory electron transport Fenfuram Faust, 2003 Kresoxim-methyl Faust, 2003 RNA synthesis and function 8-azaguanine Faust, 2003 Metalaxyl Faust, 2003 steroid biosynthesis Triadimenol Faust, 2003 uncoupler of oxidative phosphorylation 2,4-dinitrophenol Broderius, 2005 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

51 Table continued... Upregulation of hepatic catabolism of thyroid hormones TCDD Crofton, 2005 PCDD Crofton, 2005 TCDF Crofton, PCDF Crofton, PCDF Crofton, 2005 OCDF Crofton, 2005 PCB-28 Crofton, 2005 PCB-52 Crofton, 2005 PCB-77 Crofton, 2005 PCB-101 Crofton, 2005 PCB-105 Crofton, 2005 PCB-126 Crofton, 2005 PCB-138 Crofton, 2005 PCB-153 Crofton, 2005 PCB-156 Crofton, 2005 PCB-169 Crofton, 2005 PCB-180 Crofton, 2005 PCB-118 Crofton, 2005 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

52 Table 9: chemicals tested in more than one dual relevance mixture with mode/mechanism information when stated Chemical name Mode/mechanism (if stated) Study details: First author, Pub. Year 1,2,3-Trichlorobenzene Wade, ,2,3-Trichlorobenzene Hermens, ,2,3,4-tetrachlorobenzene Wade, ,2,3,4-tetrachlorobenzene Hermens, 1985 Benzene Nonpolar narcotic Wang, 2009 Benzene Hermens, 1985 Pentachlorobenzene Hermens, 1985 Pentachlorobenzene Wade, 2002 Toluene Nonpolar narcotic Wang, 2009 Toluene Hermens, ,4-Dichlorotoluene Hermens, ,4-Dichlorotoluene Hermens, 1985 Aniline Hermens, 1985 Aniline Polar narcotic Wang, ,4-Dichlorophenol Polar narcotic Wang, ,4-Dichlorophenol Hermens, 1985 Phenol Polar narcotic Wang, 2009 Phenol Hermens, 1985 Mirex Wade, 2002 Mirex Jonker, 1990 Loperamide Jonker, 1990 Loperamide Groten, 1997 Stannous chloride Jonker, 1990 Stannous chloride Groten, 1997 Chlorpyrifos Broerse, 2010 Chlorpyrifos Ito, 1995 Endosulfan Ito, 1995 Endosulfan Wade, 2002 Cadmium chloride Wade, 2002 Cadmium chloride Groten, 1997 TCDD Wade, 2002 TCDD Upregulation of hepatic Crofton, 2005 catabolism of thyroid hormones nickel Pavlaki, 2011 nickel Broerse, 2010 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

53 10.4. Method development Some 30 articles in the CRADIS database deal with recent method developments in mixture toxicology. We will now briefly summarize the studies organized by two topics: 1) developments with potential relevance to risk assessment approaches and 2) development of mathematical models for predicting mixture effects, especially mixed models of mixture toxicity Risk assessment approaches Method developments include one that used Cramer classes and quantitative models of uncertainty to model the effects of mixtures of migrant chemicals (Price et al. 2009). This approach utilizes the existing IA and DA models with pragmatic approaches to data gaps. Another extended the approaches used in mixture toxicology to address chemical exposures to the arena of multiple insect resistance genes (Wolt 2011), thus widening the nature of mixtures for which regulatory approaches can be considered. Of relevance to the use of mode of action to define similarity or dissimilarity is an IPCS framework for assessing the relevance of a non-cancer mode of action that has been published (Boobis et al. 2008), complementing a framework for mode of action in cancer (Boobis et al. 2006). These frameworks are discussed in more detail in section Mathematical models A number of articles have proposed new modelling or mathematical approaches to mixture toxicity, often based on combinations or extensions of the existing IA and DA models. These include: A new log Kow-based model to predict the combined toxicities of antifouling chemicals, which the authors found was not accurately predicted by IA or DA (Wang et al. 2011). A novel bio-concentration factor-based model to predict toxicity in the sea urchin embryo-larval bioassay, developed after the authors found only weak correlation of the observed toxicity with predictions using IA or DA (Xu et al. 2011). A novel model of integrated concentration addition with independent action based on multiple linear regression (ICIM, (Qin et al. 2011)). The authors consider that ICIM may be useful when mixture components do not have a strictly similar or dissimilar mode of action. The Effect Residual Ratio (ERR) method (Wang et al. 2010) which can be used to assess deviations from IA and DA predictions. The MIXTOX model, developed in EU project ENV4-CT , was evaluated using binary mixture studies of nickel, irgasan (an antibiotic) and diclofenac (an antiinflammatory)(rudzok et al. 2010). The model incorporates the IA and DA models, Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

54 and provides a framework for significance testing of deviations from reference models (Jonker et al. 2005). An Integrated Fuzzy Concentration Addition-Independent Action (IFCA-IA) model which was evaluated in a of 12 industrial organic chemicals with four different modes of toxic action (Wang et al. 2009). The IFCA-IA model uses fuzzy logic to derive weight coefficients from molecular structural descriptors, and uses these coefficients to weigh the contribution of IA and DA to the model. Two of these studies propose mixed models of mixture toxicity which combine the DA and IA models and each is now examined in turn. Mixed models recognise that both the DA and IA concepts can apply to mixture effects, but that realistic scenarios are unlikely to conform strictly to either a pure DA or a pure IA prediction. Integrated Fuzzy Concentration Addition-Independent Action (IFCA-IA) Model Outperforms Two-Stage Prediction (TSP) For Predicting Mixture Toxicity (Wang et al. 2009) Wang et al. proposed an Integrated fuzzy concentration addition-independent action (IFCA- IA) model as an alternative to the two stage prediction (TSP) model. In the TSP model chemicals are first grouped by their mode of action, then the effect of each group is calculated using DA, and finally these group effects are cumulated using IA. Conversely, the IFCA-IA model does not use MOA information and instead uses fuzzy logic to calculate weight coefficients from molecule structure descriptors. The coefficients are used to set the contribution of DA and IA to the overall model. The IFCA-IA model must be trained on an experimental data set, which may not be available in the regulatory context. The chief benefit of this model, that MOA information is not required, may not be sufficient to outweigh the practical obstacle of q requirement of suitable training data. However this endeavour may chime with other efforts to increase the use of read-across from well studied chemicals to less studied chemicals, and may be a useful approach if it can be adapted from its current position in experimental mixture studies into a practical tool in cumulative risk assessment. A Novel Model Of Integrated Concentration Addition With Independent Action Based On Multiple Linear Regression (ICIM) (Qin et al. 2011) Qin et al proposed a model called Integrated Concentration addition with Independent action based on Multiple linear regression (ICIM). ICIM uses multiple linear regression to derive a relationship between observed mixture effects and the separate DA and IA predictions of those effects. This approach, using multiple linear regression, contrasts with the use of fuzzy logic based on molecular descriptors in models such as IFCA-IA. Qin et al found that ICIM had a strong predictive power, and that it provided better predictions than either DA or IA. However, like the IFCA-IA model discussed above, the ICIM model also requires training; Qin et al used experimental results of mixtures from a ray design to predict the results of Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

55 mixtures with an equipotent design. Consequently the ICIM may have similar limitations to the IFCA-IA method and cannot be considered as a practical model for use in CRA. To our knowledge, a practical CRA approach that allows a commonly used implementation of DA, namely the Hazard index, to be combined with some implementation of IA is not in use. This is a potentially important method gap. The common feature of these mixed models is the use of DA and IA as the underlying concepts. It follows logically that predictions made using these mixed models must fall within the predictions made using pure DA and pure IA. If it is accepted that both data and method gaps currently limit the use of a mixed model, then the use of this simplification, that the mixed model prediction lies between DA and IA, may prove useful, especially if the extent of the difference between DA and IA predictions is understood. The extent of, and factors driving, the differences between DA and IA predictions is the topic of section Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

56 11. Definitions of dissimilar mode of action In fulfilment of work specified for Task 3, we have assembled the definitions used in major reports or by regulatory organisations that relate to the issue of dissimilar modes of action. In most cases the related terms were simple similar action contrasted with simple dissimilar action. We note that the definitions are used to refer to the same broad concepts, but differ in the amount of detail that is specified within the definition. The definitions provided for simple similar action and simple dissimilar action are summarised in Table 10. The definitions are basically identical but differ in the amount of detail given, e.g. whether they mention the target site or the relative potency UK Committee on Toxicity (2002) The UK Committee on Toxicity (COT) uses the term simple similar action in its report on Risk Assessment of Pesticides and Similar Substances (COT 2002) for chemicals that act in the same way, by the same mechanism(s), and differ only in their potencies. Dose additivity is a consequence of simple similar action. In the case of simple dissimilar action it is assumed that the modes of action and possibly the nature and site of action differ among the chemicals in the mixture which exert their individual effects, but do not modulate the effect of other constituents of the mixture. In the COT report dose addition is used as synonym for simple similar action and independent action is used as synonym for simple dissimilar action European Food Safety Authority (2008) The Panel on Plant Protection Products and their Residues (PPR) of the European Food Safety Authority states that simple similar action occurs when chemicals in a mixture act in the same way, by the same mechanism/mode of action, and differ only in their potencies (EFSA 2008b). The effects of the mixture are equivalent to the sum of the potencycorrected doses of each component of the mixture. According to EFSA (2008) simple similar action is a synonym for dose-addition ( dose-addition, also referred to as simple similar action ). On the other hand simple dissimilar action is a synonym for response-addition ( response-addition, also referred to as simple dissimilar action ) and occurs where the modes of action and possibly, but not necessarily, the nature and sites of toxic effects differ between the chemicals in a mixture, and one chemical does not influence the toxicity of another. The effects of the mixture are equivalent to the combination of the effects of each component of the mixture. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

57 11.3. US Environmental Protection Agency (2002) The US Environmental Protection Agency does not provide a final definition of simple similar action in its Guidance on cumulative risk assessment of pesticide chemicals that have a common mechanism of toxicity (EPA 2002). But they characterise toxicologically similar chemicals as chemicals that share a common toxic effect or act at the same target site. In contrast to the PPR/EFSA and the UK COT the US EPA does not use the term dose-addition as a synonym for similar action. The US EPA recommends the assumption of dose-additivity in the risk assessment procedure in case of chemicals which are toxicologically similar. No definition of dissimilar action is presented in this EPA guidance Agency for Toxic Substances and Disease Registry (2004) In the Guidance Manual for the Assessment of Joint Toxic Action of Chemical Mixtures of the Agency for Toxic Substances and Disease Registry (ATSDR 2004) simple similar action is equated with dose addition Dose Addition, also known as simple similar action, assumes that the components of a mixture behave as concentrations or dilutions of one another, differing only in their potencies. Dose addition is considered most appropriate for mixtures with components that affect the same endpoint by the same mechanism of action. Response Addition, also known as simple independent action assumes that the chemicals act independently and by different modes of action WHO/IPCS (2009) The World Health Organisation/International Program on Chemical Safety presented in the Harmonization Project Document 7: Assessment of Combined Exposures to Multiple Chemicals: Report of a WHO/IPCS International Workshop on Aggregate/Cumulative Risk Assessment (IPCS/WHO 2009) the following definitions: Simple similar action refers to chemicals that cause toxicity through a common toxic mode of action (MOA) and are thus evaluated using dose addition approaches. Simple dissimilar action is assumed when chemicals cause a common health effect, but by a different toxic MOA. In this case, the toxic responses are thought of as biologically and statistically independent events Norwegian Scientific Committee for Food Safety (2008) The Norwegian Scientific Committee for Food Safety uses the following definition for simple similar action in its report on Combined toxic effects of multiple chemical exposures (VKM 2008): simple similar action assumes that the compounds act on the same biological site (e.g. receptor or target organ), by the same mechanism and that they differ only in their potencies. Each chemical contributes to the toxicity of the mixture in proportion to its dose, and their relative toxicities are assumed to be constant at all dose levels. The Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

58 effect would be a result of the sum of the contributing dose of each chemical. Exposure to chemicals with simple similar action will result in dose addition. The committee defines simple dissimilar action as follows The chemicals contribute to a common result, but the mechanisms by which the chemicals act are always different. Also, the nature and site of action may possibly, but not necessarily, differ among the chemicals in the mixture. Therefore, the presence of one chemical will not affect the toxicity of another chemical. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

59 Table 10: Comparison of the definitions given for simple similar action and simple dissimilar action COT (2002) EPA (2002) ATSDR (2004) EFSA (2008) VKM (2008) WHO/IPCS (2009) Simple similar action chemicals act in the same way, by the same mechanisms and differ only in their potencies Dose Additivity chemicals that are toxicologically similar and act at the same target site components affect the same endpoint by the same mechanism of action and differ only in their potencies chemicals in a mixture act in the same way, by the same mechanism/mode of action and differ only in their potencies same biological site (e.g. receptor or target organ), by the same mechanism and differ only in their potencies chemicals that cause toxicity through a common toxic mode of action Simple dissimilar action modes of action and possibly the nature and site of action differ chemicals act independently and by different modes of action the modes of action and possibly, the nature and sites of toxic effects differ between the chemicals in a mixture, chemicals common result, but mechanisms are always different; nature and site of action may differ chemicals cause a common health effect, but by a different toxic mode of action Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

60 12. Legislation Consideration of possible cumulative risk in EU pesticide regulations Legal background EU legislation on pesticides has been substantially revised and harmonised in the recent years. The evaluation, authorisation, placing on the market and control of plant protection products in the EU are regulated by Council Directive 91/414/EEC (EU, 1991). This Council Directive will be repealed by Regulation (EC) No 1107/2009 (EU, 2009) which will apply from 14 June These laws on the placing of plant protection products (PPP) on the market are complemented by Regulation (EC) No 396/2005 (EU, 2005) on maximum residue levels (MRLs) of pesticides in or on food, which aims at the harmonisation of MRL in the EU. Regulation (EC) No 396/2005 covers the setting, monitoring and control of pesticide residues. There is a direct interrelation between Regulation (EC) No 1107/2009 on the authorization of PPP and Regulation (EC) No 396/2005 on pesticide residues: where appropriate the setting of MRL is a prerequisite for granting authorization of a PPP (article 8, paragraph 1(g)). The same interrelation exists with Council Directive 91/414/EEC (article 4, paragraph 1(f)). In general, PPP have to be authorised before they can be put on the market. Only those PPP can be authorised which contain an approved active substance, safener or synergist, i.e. which are on the EU positive list. Before PPP are placed on the market it should be demonstrated that they present a clear benefit for plant production and that they and their residues do not have any harmful effect on human or animal health, including that of vulnerable groups, or any unacceptable effects on the environment. Authorisation of PPP by an EU member state should follow the uniform principles for evaluation and authorisation of PPP as presently described in Annex VI of Council Directive 91/414/EEC, which still applies in the context of Regulation (EC) No 1107/ Consideration of cumulative risk in EU legislation In the context of the approval procedure the risk assessment of the ingredients of PPP predominantly addresses the risk arising from single substances. It is a basic requirement that the information needed to establish, where relevant, an Acceptable Daily Intake (ADI), Acceptable Operator Exposure Level (AOEL) and Acute Reference Dose (ARfD) is provided by the manufacturer for an active ingredient, safener or synergist (see Annex II of Regulation (EC) No 1107/2009). Whereas Council Directive 91/414/EEC did not explicitly require cumulative risk assessment this was introduced by the new Regulation (EC) No 1107/2009. Consideration of potential mixture effects was a clear requirement for human or animal risk assessment of PPP and residues. Plant protection products shall have no immediate or delayed harmful effect on human health, including that of vulnerable groups, or animal health, directly or through drinking water (taking into account substances resulting from water treatment), food, feed or Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

61 air, or consequences in the workplace or through other indirect effects, taking into account known cumulative and synergistic effects where the scientific methods accepted by the Authority to assess such effects are available; or on groundwater (article 4, paragraph 3(b)). The same applies to PPP residues (article 4, paragraph 2(a)): they shall not have any harmful effects on human health, including that of vulnerable groups, or animal health, taking into account known cumulative and synergistic effects where the scientific methods accepted by the Authority to assess such effects are available, or on groundwater. For ecotoxicological effects however, there is no requirement for taking into account known cumulative and synergistic effects. These more general terms in Regulation (EC) No 1107/2009 provide the basic legal background for the consideration of potential cumulative risk of PPP ingredients without addressing the necessary next steps. But Regulation (EC) No 396/2005 on maximum residue levels clearly specifies in article 36, paragraph 1(c) the necessity to develop a methodology for assessing aggregate, cumulative and synergistic effects ( studies and other measures necessary for the preparation and development of legislation and of technical guidelines on pesticide residues, aimed, in particular, at developing and using methods of assessing aggregate, cumulative and synergistic effects ). Based on these requirements the Panel on Plant Protection Products and their residues (PPR) developed and published a methodology for the assessment of cumulative and synergistic risks of pesticides to human health (EFSA, 2008) which applies to substances with similar mode of action. It is the aim of this ongoing project to develop a methodology for the cumulative risk assessment of substances with dissimilar mode of action Refusal of the authorisation of PPP It is clearly stated in Regulation (EC) No 1107/2009 that active substances or products placed on the market should not adversely affect human or animal health or the environment (article 1, paragraph 4). As a basic requirement for the authorization all information necessary for establishing guidance values such as ADI or AOEL have to be submitted to the authority. In the context of the human health risk assessment for the general population or workers the guidance values are compared with the estimated exposure towards the PPP or its residues under the recommended conditions of use. If the exposure is lower than the ADI or AOEL, the application of the PPP can be regarded as safe. However, if the authority concludes that the application of the PPP as requested by the manufacturer is not safe for workers or the general population the authorisation can take place but the authority can link it to certain restrictions (article 6 of Regulation (EC) No 1107/2009). Those restrictions may comprise for example: manner and conditions of application, designation of categories of users (professional or not professional), designation of areas where the use may be authorised under specific conditions, the need to impose risk mitigation measures and monitoring after use. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

62 Which restrictions are appropriate has to be considered on a case by case basis and will also depend on the importance of a PPP (risk benefit analysis). A similar procedure is applied for environmental health risk assessment. After authorisation of a PPP national authorities check in their monitoring programmes whether the existing MRL are maintained. If these control programmes give any advice that existing MRL cannot be maintained under normal conditions of use this might result in a change of the authorisation and restrictions as discussed above or even the removal of the authorisation may be the consequence (article 44 of Regulation (EC) No 1107/2009). Neither Regulation (EC) No 1107/2009 nor Regulation (EC) No 396/2005 provides any information about the legal consequences if a cumulative health risk assessment points to a possible risk of the cumulative exposure. Even the EFSA opinion (EFSA, 2008) does not provide any further information. Risk management measures to be taken in case of a cumulative risk have not been worked out until now. It has to be discussed how to proceed if a cumulative risk assessment (CRA) has indicated a potential cumulative risk. Should all PPP, which were part of the CRA, be re-evaluated? Should only those, which contribute most to the overall risk of the mixture, be reevaluated? Should some PPP of the mixture be banned? Should the manner of applications be changed? Should the simultaneous use of certain PPP be banned? Further guidance from the authorities on risk management measures in case of a risk due to cumulative exposure is necessary. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

63 13. Assessment concepts for mixture toxicity and empirical evidence for their validity Methods for experimental mixture studies can be divided into two major classes, whole mixture approaches and component based approaches. In whole mixture approaches, direct toxicological assessments of a given chemical mixture, such as a complex environmental sample are conducted by treating a mixture as if it were a single chemical. The composition of the mixture is not the topic of investigation, and whole mixture approaches do not require new, mixture-specific assessment concepts. In component-based approaches, efforts are made to anticipate the effects of a mixture on the basis of the toxicity of its components. This makes it possible to draw more general conclusions about the relationship between the effects of single substances and those of their combinations. Numerous methods for this purpose have been described in the literature (see the overview in (EC 2009). Such methods allow quantitative predictions of mixture toxicities, without the need to test different mixture ratios, mixture concentrations or overwhelmingly large numbers of permutations of mixture components. They require information about the effects of the mixture components after administration as single chemicals and about their levels in the mixture (mixture ratio). The effects of all components must have been measured under the same conditions as the experimental mixture, using the same toxicological endpoint. The experimentally observed mixture effects can then be compared with those expected on the basis of the effects of the components. Among the component-based approaches, two fundamentally different concepts exist for the calculation of mixture effects on the basis of the toxicity of its components, independent action (IA) and dose or concentration addition (DA or CA). Both concepts rely on an additivity assumption, which is based on the expectation that all chemicals in the mixture exert their effects without influencing each other s action. The additivity assumptions are not fulfilled when components of the mixture interact with one another, e.g. by undergoing chemical reactions with each other, or by inducing (de)toxifying metabolic conversions that target some or all of the mixture components. The difference between IA and DA is in the way in which each concept constructs its additivity assumption. IA derives additivity assumptions from probabilistic considerations of the effects of the mixture components. In contrast, DA is based on the idea that all components in the mixture behave as if they are simple dilutions of one another, which is often taken to mean that DA describes the joint action of compounds with an identical mechanism of action. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

64 13.1. Independent action (IA) Independent action (sometimes also termed Effect Addition, Effect Multiplication or Abbotts Rule) conceptualises mixture effects by assuming that combined effects can be calculated from the effects caused by the individual mixture components on the basis of the statistical concept of independent random events (Bliss 1939). This can be mathematically expressed as: E( c ) = 1 [1 E( Mix c i i= 1 n )] (Eq. 1a) if the effect increases with increasing concentrations (e.g. when mortality data are considered) and E( c ) = n E( mix c i i= 1 ) (Eq. 1b) when the effect decreases with increasing concentrations (when e.g. survival rates are observed). In both equations E(c Mix ) denotes the effect provoked by the total mixture at a concentration n c Mix = c i i= 1. E(c i ) are the effects that the individual components would cause if applied singly at that concentration at which they are present in the mixture. Due to this probabilistic background, IA assumes strictly monotonic concentration-response curves of the individual mixture components and an Euclidian-type effect parameter scaled to an effect range of 0-1 (0-100%) Applicability of IA to mixtures composed of agents with dissimilar modes of action Theoretically, the stochastic principles of IA are also valid when one and the same agent is administered sequentially and irreversible events such as mortality are investigated. Because organisms cannot die twice, the probability expressed in equation 1 a,b applies, despite the fact that the mechanism by which the chemical provokes mortality is identical. In the case of simultaneous administration of many chemicals the principle of independent events can only be realised by making the additional assumption that all components in the mixture exert their effects by activating different effector chains that converge to produce a common effect. This is commonly thought to apply in cases where the chemicals in the mixture exert their effects through strictly independent, i.e. dissimilar mechanisms. By activating differing effector chains every component of a mixture of dissimilarly acting chemicals provokes effects independent of all other agents that might also be present, and this feature lends itself to statistical concepts of independent events. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

65 Data requirements for using IA IA uses single substance effects, E(c i ), for predicting a mixture effect (equation 1 a, b). This means that the information need for utilising IA changes substantially with increasing numbers of mixture components. For example, according to IA a binary mixture of two agents that individually produce a 30% effect will lead to a 50% mixture effect. Thus, the application of IA means that effects of 30% have to be measured with reliability which usually does not present problems. In a 10-component mixture producing a 50% mixture effect, however, each component has to be present at a concentration that produces only a 6.7% individual effect. But effects of that magnitude are already at the borderline of what can be measured reliably in many in vivo toxicological experiments. The more compounds are present in a mixture, the lower the individual E(c i ) s become that are required as input values for estimating a 50% mixture effect. The fact that increasingly lower E(c i )-values for each component need to be measured for calculating IA-predictions is a serious drawback, as this increases experimental demands beyond what is technically achievable with the number of animals per does group normally used in toxicity studies. NOAELs are not readily suited as input data for IA. NOAELs denote the highest tested doses that produced effects not statistically significantly different from those in untreated controls, but they do not describe effect magnitudes. Depending on the resolving power of the chosen experimental arrangement, the effects associated with NOAELs can be quite large, but cannot be measured directly, and are only accessible through regression modelling in dose-response analyses. However, the number of doses tested in studies that establish a NOAEL is often rather limited and does not permit regression analysis. As a result, it is normally not possible to establish whether a NOAEL is associated with a 5%, 10% or 20% effect. Consequently, the input data required for using IA are not accessible through reporting a NOAEL Under IA, when is a mixture risk acceptable? Equations 1 a and b imply that agents present at doses associated with zero effects will not contribute to the joint effect of the mixture. If this condition is fulfilled for all components in the mixture no combination effect is expected under IA. This means that claims of absence of mixture responses can only be substantiated if very small effects can be distinguished with reliability from zero effects. However, especially with mixtures composed of large numbers of components, this demands exceeds the resolving power of most toxicological studies which struggle to resolve 5% effects. According to IA, 10 components at doses associated with 5% effect will already produce a combination effect of 45%. Correspondingly, 100 agents with a 1% effect are expected to produce a mixture effect of 63%, and with 100 chemicals of 0.1% effect the expected joint response under IA will still be 9.5 %. Such small effects can only be demonstrated with astronomically large numbers of animals. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

66 Correlation assumptions in IA and their consequences Consideration of the correlation of susceptibility of organisms in the IA model was often present in early discussions, but has become less mentioned. Correlation of susceptibility refers to the extent to which the individuals most susceptible to one chemical are also those most susceptible to a second chemical. Correlation can be expressed as a correlation coefficient, r, which takes values between 1 and -1. A value of 1 indicates that there is complete correlation in susceptibility, i.e. that the sensitivity of individuals to either component is the same, for example the individual most sensitive to the first chemical is also the most sensitive to the second; conversely, a value of -1 indicates complete negative correlation, in that the individuals most sensitive to one chemical will be the least sensitive to the second chemical and vice versa. Finally, a value of 0 indicates that the susceptibility to one chemical is unrelated to susceptibility to the other. A generalised model of correlated independent action was introduced in the 1940s by Plackett and Hewlett (reviewed in (Boedeker and Backhaus 2010)). Three subtypes of IA were formulated for binary combinations. Firstly, when the susceptibility is uncorrelated, i.e. the correlation coefficient (r) =0, the combined effect is given by the following equation, which is equivalent to the IA equation presented in equation 1a (section 13.1) for a binary mixture: When r=0, P 1,2 = P 1 + P 2 P 1 P 2 (Eq. 2a) In this case, the mixture probability (P 1,2 ) is given by adding the probability of an effect by one chemical (P 1 ) to the probability of an effect of a second chemical (P 2 ) and subtracting the product of the two probabilities (P 1 P 2 ). Secondly, when there is total positive correlation of susceptibility (r=1), the effect of the combination is that of the most potent component because the individuals that might have been affected by the weaker component will already have been affected by the most potent component. This situation is sometimes termed no addition (Boedeker and Backhaus 2010). In this case the IA equation is: When r=1, P 1,2 = max(p 1, P 2 ) (Eq. 2b) Thirdly, when there is total negative correlation (r=-1), the effect of the combination is the numerical summation of both components, which is sometimes termed effect summation (Boedeker and Backhaus 2010). In this case the individuals affected by either chemical are different and so there is no need to consider effects of the two chemicals on one individual, only on separate individuals: When r=-1, P 1,2 = P 1 +P 2 (Eq. 2c) These formulations for binary mixtures depend on assumptions or knowledge about the correlation between the two components in the binary mixture. Such knowledge is less likely to be available, and assumptions harder to substantiate, for multi component mixtures with Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

67 tens or hundreds of components. For instance the concept of total negative correlation when applied to a mixture of more than two components may not even be meaningful (EPA 2000). The original formulation of these concepts, over 60 years ago, may underlie some suggested pragmatic applications of IA. For example, the notion that the toxicity of a mixture is that of the toxicity of the most potent component matches the formulation of IA under the assumption of total positive correlation (Eq. 2b), whilst the notion that the mixture toxicity is given by the summed effects of the components (effect summation, ES) matches the formulation of IA under total negative correlation (Eq. 2c). Boedeker and Backhaus considered that the subtype of IA with total negative correlation was one of the rare theoretical foundations of ES (Boedeker and Backhaus 2010), although total negative correlation is not always stated as a requirement for the use of ES. The chief requirement usually given for using ES is the presence of linear dose-response relationships (EC 2009). The mathematical form of ES may also provide an approximation of the effect under IA when certain other assumptions hold, for example that there are only two components and that their individual effects are small (EPA 2000). In this case however ES is used as an approximation to the effect under IA for ease of computation, not because it is considered innately valid. In typical human toxicology situations, it seems unlikely that there will be sufficient information about the correlation of susceptibilities to the mixture components. In the absence of this information, the assumption of no correlation may be more reasonable that assuming the extremes of either total negative or total positive correlation. In the event that assumptions of correlation are made, this should be clearly stated and the rationale for doing so should be justified Empirical evidence for IA The results of the systematic literature search were carefully examined for empirical evidence for the validity of IA. In particular we sought examples when: IA provided an accurate prediction of a mixture effect, in a situation in which the predictions under DA and IA were separable. IA provided a prediction that was more conservative than the DA prediction. IA provided a more conservative prediction AND was also accurate. We found that experimental studies published since the State Of The Art Report on Mixture Toxicity (EC 2009) have not altered the literature summary given in that report. The current position is now summarised. The ecotoxicology literature contains a few examples when the effects of carefully designed mixtures were shown to validate the IA model (EC 2009). Of importance is a of 16 biocides whose combination effect on algal toxicity was accurately predicted by IA (Faust et al. 2003). These components were selected on the basis of their strictly different specific mechanism of action, which are listed in the CRADIS database analysis of this. Faust et Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

68 al. also found that whilst IA was the accurate model, in this case the predicted effect under DA was greater than that under IA, and so DA could be considered the conservative model. If this finding is generally applicable then there would be reassurance that even in cases when IA is valid, DA can still provide a conservative risk estimate, supporting the use of DA as the default concept in mixture risk assessment. A proof-of-principle example of the validity of IA has not been identified in the mammalian toxicology literature (this report, (EC 2009)). The reason for this is likely to be the difficulties, including costs and ethical considerations, of performing mammalian studies with a sufficiently large number of components (which is required to allow the predictions of IA and DA to be distinguished from each other) and the difficulty in selecting an appropriate effect for which there are enough well characterised chemicals with strictly different specific mechanisms of action. The amount and level of knowledge required to design a mixture experiment that is suitable to test the hypothesis that IA is accurate in a mammalian system appears to be far greater than the knowledge that is typically available for chemicals. A situation when IA was both more conservative than DA, and also accurate was not identified in the literature. The factors that lead to IA producing a more conservative prediction than DA are discussed in section 13.4.This putative situation is an important one to consider, since it would have implications for the use of DA as a conservative default. Consequently the factors discussed in section 13.4 can be evaluated to consider how likely it is that this situation could occur, given that it has not been observed in experimental studies to date Dose addition (DA) DA (also known as concentration addition, CA) is based on the idea that all components in the mixture behave as if they are simple dilutions of one another, which is often taken to mean that DA describes the joint action of compounds with an identical mechanism of action. When chemicals interact with an identical, well-defined molecular target, it is thought that one chemical can be replaced totally or in part by an equal fraction of an equi-effective concentration (e.g. an EC50) of another, without changing the overall combined effect. If the assumption of dose addition holds true, these fractions of equi-effective single substances concentrations also called toxic units simply sum up to an overall toxic unit of the mixture. Therefore, DA is also known as Toxic Unit Summation. The concept can be mathematically formulated as: ECx 1 n p i Mix = i= 1 ECxi (Eq. 3) with n denoting the number of mixture components, p i the relative fraction of chemical i in the mixture, and x a common effect level, provoked by an exposure to a single substance or mixture concentration ECx Mix resp. ECx i. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

69 In general, no explicit formulation of the DA-expected mixture effect E(c Mix ) is possible. Direct calculations are restricted to the effect levels associated with the effect concentrations (ECx-values) (Faust et al. 2003). The requirement of parallel dose response curves has often been used as a decision criterion about the application of DA to a specific mixture. However, the general formulation of DA in equation 3 does neither assume a specific shape of each concentration-response curve of the components, nor a specific relationship between the curves, such as parallelism. Even if all chemicals in a mixture share an identical receptor binding site, differences e.g. in the toxicokinetic behaviour of the substances might lead to concentration-response curves that are not parallel, yet DA may still apply Data requirements for using DA To predict mixture effects by using DA, information about the doses that induce the same specific effect are required for both the mixture and all single components. For example, if the effect dose of a mixture leading to a 50% effect is known, then the equivalent effect doses (ED50) for all mixture components need to be available to reach decisions whether the combined effect is dose additive (see equation 3). In this case, the sum of toxic units in equation 3 will be 1. The same requirements need to be met for any other effect level. Usually, information about effect doses is accessible through dose-response analyses of the individual components in a mixture. As with IA, NOAELs are strictly speaking not suited as input values for using DA, because NOAELs represent different (but unknown) effect doses. However, unlike IA, the measurement precision required for using the concept as the number of mixture components increases, does not change. This feature makes DA generally easier to use in most situations. It is obvious from equation 3 that DA represents the weighted harmonic mean of the individual ECx values, with the weights being the fractions p i of the components in the mixture. This has important favourable consequences for the statistical uncertainty of the DApredicted joint toxicity. As the statistical uncertainty of the DA-predicted ECx for the mixture is the result of averaging the uncertainties of the single substance ECx-values, the stochastic uncertainty of the DA prediction is always smaller than the highest uncertainty found in all individual ECx-values. Perhaps contrary to intuition, the consideration of mixtures composed of a large number of agents actually reduces the overall stochastic uncertainty. This feature renders DA predictions quite reliable and robust Under DA, when is a mixture risk acceptable? DA implies that every toxicant in the mixtures contributes in proportion to its toxic unit (i.e. its concentration and individual potency) to the mixture toxicity. Whether the individual doses are also effective on their own does not matter. Thus, combination effects should result from toxicants at or below effect thresholds, provided sufficiently large numbers of components sum up to a sufficiently high total dose. Unfortunately, that is often misunderstood to mean that mixture effects will arise with any combination of agents, if the principles of DA are fulfilled. However, this is not the case. For example, the joint effect of Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

70 two agents combined at 1/10 of their ADI is expected to be considerably smaller than the effect (if any) associated with the ADI of each of the chemicals on their own. Similarly, 100 chemicals combined at 1/100 of their ADI will not produce a mixture effect greater than the effects provoked at the ADI s of each of the single components (see equation 3). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

71 13.3. The role of mode and mechanism of action information in the choice between IA and DA We searched and analysed several sources of guidance on mixture risk assessment for insights into criteria relevant to the choice between similar and dissimilar action, dose addition or independent action. An overview of the guidance from regulatory bodies is provided in Task 4 (section 14), and this section concentrates on the potential role of mode of action in mixture risk assessment Definitions of mode and mechanism of action Two concepts that can feature in guidance on the application of mixture concepts are mode of action and mechanism of action. These concepts have been defined as follows: Mode of action (MOA): a biologically plausible sequence of key events leading to an observed effect supported by robust experimental observations and mechanistic data (Boobis et al. 2006). Mechanism of action: a sufficient understanding of the molecular basis for an effect and its detailed description so causation can be established in molecular terms (Boobis et al. 2006) or a detailed explanation of the individual biochemical and physiological events leading to a toxic effect (EFSA 2008b). It has been noted that the US EPA use the term mechanism of action to accompany the definition of mode of action given above (EFSA, 2008b). Care must therefore be taken not to confuse these concepts when comparing international, European and American activities. As an example of an MOA definition, the key events making up the MOA of a DNA-reactive chemical are: 1. Exposure of target cells (e.g., stem cells) to ultimate DNA-reactive and mutagenic species in some cases this requires metabolism. 2. Reaction with DNA in target cells to produce DNA damage. 3. Misreplication on damaged DNA template or misrepair of DNA damage. 4. Mutations in critical genes in replicating target cell. 5. These mutations result in initiation of new DNA/cell replication. 6. New cell replication leads to clonal expansion of mutant cells. 7. DNA replication can lead to further mutations in critical genes. 8. Imbalanced and uncontrolled clonal growth of mutant cells may lead to preneoplastic lesions. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

72 9. Progression of preneoplastic cells results in emergence of overt neoplasms, solid tumours (which require neoangiogenesis), or leukaemia. 10. Additional mutations in critical genes as a result of uncontrolled cell division results in malignant behaviour (Preston and Williams 2005). This example shows that even an MOA, compared to a full mechanism, can contain a significant amount of detail, and illustrates the data demands that would be required to systematically analyse multiple chemicals with multiple modes of action. A fully elucidated mechanism of toxicity is only available for a few chemicals, one example being cyanide, whilst more chemicals have a known MOA, with key events that are known, measurable, necessary and consistent (Carmichael et al. 2011). The studies required to identify an MOA are not currently a regulatory requirement for industrial chemicals, perhaps because such studies have a uncertain time scale and costs, with inconsistent interpretation and regulatory impact (Carmichael et al. 2011). Our analysis (sections 10.1 and ) showed that when authors of experimental studies have made reference to mode or mechanism of action, they are not using a common framework to make these assignments. Approaches using agreed MOAs, or at least MOAs defined in accepted ways, could provide a common framework that, if used, would improve comparison of results from mixture assessments and experimental studies. A number of frameworks are being proposed and may prove useful if they achieve consensus. However, their suitability for use in assigning mixtures to DA or IA must also be assessed. IPCS frameworks for assessing the relevance of cancer (Boobis et al. 2006) and noncancer (Boobis et al. 2008) modes of action have been developed. The frameworks apply a weight of evidence approach based on the Bradford Hill criteria for causality to evaluate a proposed MOA and its relevance to humans. Such frameworks provide a mechanism to establish whether the MOA of a chemical has been described previously for other chemicals or whether it is novel. Of relevance to CRA, the systematic use of such frameworks, and MOAs arising from or validated by them, might allow the development of a set of MOAs founded on the same principles, thus allowing them to be compared, and that can be used for more reliable grouping or related activities. It was suggested that a database of generally accepted MOAs and informative cases should be constructed and maintained (Boobis et al. 2006). This need has been recognised by others, for example McCarty and Borgert considered that the absence of any generally accepted classification scheme for either modes/mechanisms of toxic action or of mechanisms of toxicity interactions is problematic as it produces a cycle in which research and policy are interdependent and mutually limiting. (McCarty and Borgert 2006). Clearly, this issue constitutes a substantial knowledge gap. The advantage of these approaches is that their international nature may favour international acceptance and agreement, however are they suitable for the purpose of assigning DA or IA? We note that the definitions of mode or mechanism of action given above, do not define when a mode/mechanism is novel, or separate or distinct from another mode/mechanism. This is a critical need if the mode/mechanism information is to be used to inform the selection of a mixture assessment concept, DA or IA. If the modes/mechanisms are not independent the Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

73 case for IA will not be supportable, and if no reference to independence is made then the definition will not be applicable to the choice between concepts, since independence is the key criteria for choosing IA over DA Use of empirical evidence to infer similarity or dissimilarity Some authors have concluded that, rather than selecting a mixture concept, DA or IA, and using it to predict the effect of a mixture, in fact one can perform a mixture experiment and then infer knowledge about the similarity or dissimilarity of the components of a mixture based on whether its results are closer to the DA prediction (infer that components are similar) or to the IA prediction (infer that the components are dissimilar).example of this usage are now provided: Walter et al. stated "These findings [that experimental effect was predicted by IA] strongly indicate, that the modes of action of mixture components are indeed dissimilar in a way, that is sufficient to predict the mixture toxicity with the concept of independent action." (Walter et al. 2002). Bellas concluded that the toxicity of a mixtures of zinc pyrithione and Sea-Nine...was accurately predicted by the IA concept, suggesting a dissimilar mode of action of those substances (Bellas 2008) Hodges et al. titled their paper Defining the toxic mode of action of ester sulphonates using the joint toxicity of mixtures. and concluded that...data indicated that ES substances exhibit concentration addition with linear alkylbenzene sulphonate (LAS) and phenols and response addition with alcohols. This suggests that ES behave with a similar mode of action to phenol and LAS which are known polar narcotics and with a dissimilar mode of action to alcohols which are known baseline narcotics. (Hodges et al. 2006). Escher et al. proposed an in vitro assessment of modes of toxic action, relating to pharmaceutics in aquatic environments and used mixture experiments as a diagnostic tool to analyse the mode of toxic action (Escher et al. 2005). Lutz et al claimed that Distinction between dose addition and response addition for the analysis of the toxicity of mixtures may allow differentiation of the components regarding similar versus independent mode of action (Lutz et al. 2005). This approach reverses the use of mixture concepts to predict effects, and may therefore be less relevant to the needs of CRA. However, it illustrates one possible response to the difficulties in assigning similarity and dissimilarity, namely that the assignment is not done. This post hoc assignment of similarity or dissimilarity would have an impact in the methods used to assess the relative impact of individual components in a mixture to identify targets for risk management. On consequence to CRA, is that this approach would require experimental data for the mixture of concern. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

74 Conclusions Existing guidance can depend heavily on information, e.g. MOA, which is often not available or is available in incompatible forms. Frameworks are being developed, but international consensus is required. A clear requirement for the framework to be suitable for the choice between DA and IA would be advantageous. Criteria based on concepts such as MOA were not framed in order to be useful for guiding the choice between mixture concepts, such as DA and AI; consequently they may indeed have limited usefulness for such purposes. Known or plausible independence is the key criteria for using IA instead of DA. We also note that, for the IA model, non-similarity is not the same as dissimilarity. This is reflected in some guidance (e.g. (EPA 2000)) by the use of similarity vs. independence, rather than similarity vs. dissimilarity. This is a helpful distinction. The clear difficulties with using the available frameworks and definitions (lack of consensus, lack of suitability for choosing (independence), high data requirements and poor data availability for the world of chemicals) lead to the questions: how important is it to be able to choose between DA and IA, given that it is not easy to do so? And what is the impact of having to assume one or other concept? These questions are addressed in detail in the following section (section 13.4). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

75 13.4. Quantitative differences between DA and IA predictions Problem formulation In general, the assumption of dose additivity (DA) is considered to provide a more conservative estimate of mixture toxicity than the alternative assumption of independent action (IA). Synergistic effects that exceed the DA expectation are exemptions and not the rule, at least for multi-component mixtures. In addition, data requirements for a proper application of DA are easier to fulfil than for IA. With these three arguments, DA has been recommended as a reasonable worst case assumption for the purpose of regulatory hazard and risk assessments (EC 2009), in particular in situations, where modes and mechanisms of action of mixture components are not fully independent and not fully dissimilar, or insufficiently known or unknown. Similarly, the recently published WHO/IPCS framework (Meek et al. 2011) suggests DA as a default tier zero assumption for all components cooccurring in an exposure scenario and potentially contributing to a common adverse health outcome. The main concern that may be raised against such a use of DA as a pragmatic default assumption, irrespective of toxicants modes and mechanisms of action, is that the use of a conceptually unsound model may potentially result in vastly over-protective mixture toxicity assessments, not scientifically justified and conflicting with the principle of proportionality in the regulatory management of chemicals risks. This raises the question about the quantitative differences between independent action and dose addition: What is the maximal quantitative error that may result, if DA is applied in a situation where in fact IA or a mixed model would provide the correct mixture toxicity estimate? To explore this question further, it is necessary to define a suitable measure for the differences in mixture toxicity predictions by DA and IA. In general, toxicity may be quantified either in terms of the strength or frequency of an effect (E) at a given dose (d) or concentration (c), or in terms of the dose or concentration causing a specific effect level (x), so-called effect doses (EDx) or concentrations (ECx), such as ED50 for instance. In a regulatory context, usually the latter approach is used. Correspondingly, effect doses (or concentrations) can be calculated for mixtures that contain a given set of toxicants in a given dose ratio (EDx mix ). To distinguish between the models used for their calculation, either DA or IA, they are in the following denoted by EDx DA and EDx IA, respectively. Such effect doses may potentially vary over orders of magnitude. It is therefore more convenient to describe the differences between EDx DA and EDx IA in terms of a relative figure rather than in absolute terms. Considering further that EDx DA is usually expected to be a lower value (i.e. the higher mixture toxicity estimate) than EDx IA, the differences between both predictions are in the following quantified my means of the ratio: EDx IA / EDx DA Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

76 This quotient is 1 in the case that both concepts give identical predictions, it is >1 in the typical case that IA predicts a lower toxicity (i.e. a higher effect dose), and conversely is between 0 and 1 if IA predicts higher toxicity. In other words: The ratio EDx IA / EDx DA gives the factor by which dose addition overestimates the actual toxicity in a situation where independent action would in fact provide the correct estimate Empirical evidence In published experimental studies on multi-component mixture toxicity the quantitative differences between both predictions, IA and DA, have been reported to be remarkably small, at least from a regulatory perspective. For different types of mixtures with up to 20 components, predictions of EC50 or ED50 values derived from the two models differed in no case by more than a factor of 5 (EC 2009). From these pieces of evidence the quantitative differences between IA and DA may be assessed to be of only minor relevance in a regulatory context. However, it may be questioned, whether the available observations do really represent typical situations for realistic exposure and assessment situations. Their validity may be restricted to the special mixtures, conditions and toxicity endpoints which have actually been investigated. Therefore, a consensual acceptance of DA as a generally justifiable default assumption would require strong arguments against this suspicion. Hence, in the light of available experimental evidence, the question to be clarified can also be formulated as follows: Do existing experimental findings of relatively small differences between DA and IA just represent special situations or do they reflect a general rule? This problem cannot be solved by experimentation only but needs complementary approaches, such as mathematical and statistical analyses and simulation studies. The basic questions to be addressed by such approaches are: Which are the factors that determine the quantitative differences between independent action and dose addition? May the ratio EDx IA / EDx DA take any value or do limiting cases exist? Which ratios between both predictions may occur under realistic scenarios? Limiting factors and resulting maximal differences For calculating predictions of effect doses or concentrations of mixtures, the general mathematical definitions of DA and IA can be transformed into the following expressions, as explained in detail in Faust et al. (2003). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

77 DA: EDx DA 1 n p i = 1 (Eq. 4) i= 1 Fi ( xi) IA: x% = 1 ( 1 F( p ( EDx ))) n i i IA (Eq. 5) i= 1 With these two formulae, the parameters that determine the ratio between predictions of effect doses by IA and DA are completely defined. These are simply the variables for which input data have to be entered into the formulae for obtaining the alternative mixture toxicity predictions EDx DA and EDx IA. There are four crucial factors: the number of mixture components n, the slope of the individual dose response curves defined by functions F i, the mixture ratio p 1 : p 2 :... : p i, with p i denoting the individual dose d i of component i expressed as percentage of the total dose Σd i of a mixture (p i = d i /Σd i ; Σp i = 1; i = 1 to n), and the effect level X under consideration. If no restrictions apply to these parameters, there is no fixed type of relation between both predictions of mixture toxicity. The effect dose calculated under the assumption of independent action (EDx IA ) may be larger, equal to, or even lower than the corresponding prediction by DA (EDx DA ) (Drescher and Bodeker 1995). Practical relevance has been demonstrated for the first two situations (EC 2009) but appears to be questionable for the lastmentioned theoretical case. To our knowledge, there is no convincing experimental example, where IA does not only predict a significantly higher toxicity than DA, but where this prediction is also the more accurate one. Although the ratio between both predictions is fully determined by the four parameters, it is unfortunately not possible to express the ratio EDx IA / EDx DA as an explicit function of these parameters. This results from the fact, that the formulation of IA as given in Eq. 5 provides only an implicit prediction of the effect dose of a mixture, which cannot be turned into an explicit function. Hence, the same applies to the ratio EDx IA / EDx DA. Nevertheless, it is possible to examine the ratio EDx IA / EDx DA for the existence of limit values that cannot be exceeded under a certain constellation of determining factors. To this end, the only general presumption to be made is that the dose response functions F i of individual toxicants are monotonously increasing, and that any given effect E i can be definitely related to a single dose, the so-called effect dose EDx, which can only take positive values (EDx 0). If dose response functions include lower or upper thresholds below or above which the effect is constantly zero or 100 %, these thresholds must therefore be defined as ED0 and ED100, respectively. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

78 The number of mixture components (n) Given the aforementioned premises, limiting cases exist for independent action: the predicted effect dose of the mixture can never be larger than the sum of individual effect doses of mixture components that cause the same effect level X: n 0 EDx EDx (Eq. 6) IA As a consequence, limiting cases also exist for the ratio between both predictions, EDx IA / EDx DA : 0 EDx i= 1 i IA i= 1 1 EDx n DA pi 1 i= 1 Fi ( xi) n i EDx (Eq. 7) As was shown in (Faust 1999), this expression can be solved to EDxIA 0 n (Eq. 8) EDx This means, for a mixture with a given number of components n, the ratio EDx IA / EDx DA cannot take any value, but is generally delimited by zero and n, irrespective of all other parameter values. In other words: Effect doses predicted by independent action can never exceed the corresponding prediction by dose addition by a factor that is greater than n, i.e. the number of mixture components. This means, at least theoretically, that the ratio can become infinitely small. But for a given number of components it cannot become infinitely large. For a two-compound mixture a maximal factor of 2 applies, and for mixtures with up to ten components it can never exceed an order of magnitude. For multi-component mixtures with very large numbers of components, however, it may still become very large, at least theoretically if no restrictions apply to the other determining factors The dose ratio of mixture components Further analysis reveals that the maximum ratio of n can only occur under the condition that all components of a mixture are present in equal fractions of equi-effective doses. In any other situation the ratio will always be smaller than n. When doses or concentrations of mixture components are expressed as fractions of equieffective doses or concentrations, the resulting dimensionless dose units have been termed toxic units (TU) (Sprague 1970). For practical applications, mostly the 50% effect level is used as the reference point (ED50 i ). In principle, however, TU values can be calculated for any effect level X: DA Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

79 TUx d i i = (Eq. 9) EDxi Given these definitions, it can be shown that the maximum ratio between the predictions based on IA and DA is defined by the ratio between the sum of toxic units and the maximal toxic unit value of the single substances in a mixture (Junghans et al. 2006): 0 EDx EDx i IA i= 1 DA n max i (1,..., n) TU { TU } i n (Eq. 10) This has important practical implications for situations where one or a few compounds dominate a mixture in terms of toxic units. For example, if one mixture component contributes already 50% to the total sum of toxic units, the quotient ECx IA / ECx DA can never exceed a value of 2, no matter what the total number of components and their toxic units may be The slopes of dose response curves Independent of the limiting effect of a specific dose ratio, the slopes of individual dose response curves of a given set of toxicants in a mixture have a general limiting effect on the possibility range of the prediction ratio EDx IA / EDx DA. The relationship between slope and prediction differences is visualized in Figure 6 for a hypothetical 10-compound mixture. The ratio EDx IA / EDx DA takes the value of 1, i.e. both IA and DA give exactly the same prediction (Figure 6, B), if the dose response curves of all mixture components can be described by the following specific form of the Weibull function (Drescher and Bodeker 1995) Ed ( ) = 1 exp( exp( α + ln(10) log( d))), (Eq. 11) i i i in which the general slope parameter βi has the special value of ln(10) (= ), while αi is a location parameter that has no effect on the ratio EDxIA / ECxDA. If all dose response curves are steeper, DA always predicts a higher toxicity than IA (Figure 6, A) and the maximum possible ratio EDx IA / EDx DA tends towards the maximum value of n (= number of components) if all the curves become infinitely steep. Conversely, DA predicts a lower mixture toxicity than IA (Figure 6, C), and the lowest possible ratio tends towards zero if the individual dose response curves of all mixture components become infinitely flat, and if no threshold assumptions are made (see section below). In a mixed situation with both relatively steep and flat curves (as compared to Eq. 11), large prediction differences are unlikely to occur (see section below) and it strongly depends on the dose ratio and the effect level, whether DA or IA predicts the higher mixture toxicity (Figure 6, D). The finding, that the steepness of individual dose response curves is a crucial limiting factor for the possible quantitative differences between mixture toxicity predictions by IA and DA, leads to the question, what are realistic scenarios for the distribution of slope values for dose response curves? In experimental reality, dose response curves become neither infinitely steep nor infinitely flat, but their slopes typically vary in a certain range, which unfortunately, Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

80 however, is neither very well understood nor documented. Therefore, this point needs further research that goes beyond the tasks of this report. However, as an explorative step in this direction, we performed an initial simulation with a sample set of eco-toxicological dose response data for algal reproduction (section below). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

81 100 Individual Dose Response Curves A All β i =10 Mixture Toxicity Predictions (ED50 Ratio) DA IA % Effect B All β i =LN(10) % Effect 50 IA congruent with DA C All β i = DA IA % Effect D Different β i DA IA % Effect Dose (Log Scale) Total Dose (Log Scale) Figure 6: Effect of the slope of dose response curves (DRCs) of individual substances (left) on the ratio between predictions of mixture toxicity by IA and DA (right). Dose response curves of 10 hypothetical mixture components are described by the Weibull model E i = F i (d i ) = 1-exp(-exp(α+β i log 10 (d i )))). Different scenarios are depicted for the slope parameter β i : steep curves (A), intermediate steepness (B), very shallow curves (C), and a mixed situation (D). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

82 Resulting mixture toxicity predictions by DA and IA are shown for mixtures containing the 10 components in the ratio of their individual ED50 values The effect of threshold assumptions The situation, that a general upper limit (= n) can be defined for the ratio EDx IA / EDx DA, but no general lower limit other than zero, results from the fact that a priori no restrictions were assumed for the individual dose response curves other than that they are monotonously increasing. The situation changes, when threshold assumptions are made. If threshold doses or concentrations (ED 0 ), at and below which the response is assumed to be either zero or constant and indifferent from controls, are assumed to exist for individual mixture components, than both models, IA and DA, imply that such thresholds should also exist for mixtures. Under the assumption of IA, no mixture toxicity occurs, if doses of all components are at or below individual threshold levels. Under the assumption of DA, in contrast, this is only a necessary but no sufficient condition. DA means that mixture toxicity will in no case occur, if the doses of all components are not higher than 1/n of the individual threshold (with n being the number of components). If, in contrast, individual doses are in the range between 1/n and 1/1 of an individual ED 0, DA implies that it depends on the dose ratio whether an effect occurs or not. As a consequence of these different model assumptions, effect doses predicted by IA can never be lower than a threshold for mixture toxicity calculated under the assumption of DA (ED 0,DA ). Hence, threshold assumptions result in a lower limit for the generally possible range of ratios EDx IA / EDx DA (Figure 7) which is then defined by ED0 DA EDx IA n (Eq. 12) EDx EDx DA Hence, the theoretical possibilities for situations where IA predicts a considerably higher mixture toxicity than DA are substantially narrowed down by the introduction of threshold assumptions and largely confined to higher effect levels. In addition to defining a lower boundary for the ratio the EDx IA / EDx DA, threshold assumptions may also have a quantitative impact on the calculation of effect doses under the assumption of IA. These predicted effect doses may become higher and consequently also the difference between IA- and DA-based predictions may be increased. This is the consequence of methodological constraints of statistical dose-response modelling as explained in the following. Application of the IA model for multi-component mixtures requires reliable statistical estimates of low effects of individual mixture components. If a statistical methodology is used which tends to over- (or under-) estimate low effects, then it is very likely that the overall mixture effect that is expectable under the assumption of IA is in the same way over- (or under-) estimated. In a strict quantitative sense, classical regression-based dose-response models always estimate effect changes in comparison to a control or reference value, independent of the dose range. They assume that with decreasing doses or concentrations, effect changes are becoming microscopically small and thus negligible, but not zero (unless Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. DA

83 the dose is also zero). This model assumption obviously conflicts with the threshold concept in toxicology, which assumes that below a certain threshold dose no effect changes occur at all. Continuous regression models still provide estimates of effect changes (although very small) at dose ranges below the assumed threshold doses. As a result of this methodological constraint, it is possible that a combination effect is overestimated by applying the IA model with dose response estimates from continuous regression models, when at least one compound in the mixture is present below its toxicological threshold. Figure 7: The potential range of ratios between predictions of effect doses for mixtures by IA and DA. A: Under the assumption of continuous dose response curves (DRCs) for single substances. B: Under the assumption of thresholds for individual DRCs. The problem of quantifying toxicological thresholds is complex, both conceptually and methodologically. If the term threshold is understood as a zero effect dose, its quantitative determination from experimental data by statistical means is for various reasons impossible (see (Scholze and Kortenkamp 2007;Slob 1999). So-called mathematical threshold models are therefore sometimes used as a pragmatic approximation for the toxicological threshold (Cox 1987). They are derived from commonly used nonlinear regression models by additional inclusion of a model parameter which describes the threshold dose. These threshold models are fitted to the data by the same statistical procedures as non-threshold models and provide an estimate of the most likely dose-response curve (Figure 8). The basic mathematical model may be chosen from a large class of models and different methods are also available for the estimation process. There is no model and method that is the universally most powerful and Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

84 most appropriate one. Consequently, the resulting threshold estimates depend crucially on the selected model and method. Moreover, if not bound to positive ranges, the estimation process can indeed also result in a negative threshold, simply because the model would then describe the data best. It is also not well understood how the quality and quantity of data (such as dose or concentration spacing, replicate number, and data variation) influences the resulting threshold estimates; however, to some degree data uncertainty is reflected by the confidence belts of the threshold estimates. With the aim of developing ideas about the potential quantitative effect that the use of threshold models may have on the ratio between alternative predictions of mixture toxicity by IA and DA, we applied such models as part of the following simulation with a sample set of ecotoxicological dose response data. As can be seen in Figure 8, threshold models may predict zero effect levels for doses that are associated with considerable effects in continuous dose-response models. Figure 8: Classical continuous dose-response model (black line) and threshold model (red line), both fitted to the same set of data (grey dots) Sample calculations and simulations The mathematical considerations in the preceding sections demonstrate that, if no specific dose ratio is fixed, the range of possible differences between IA and DA is limited by the number of mixture components in the first place, and that it may be further restricted by the actual slope of dose response curves of single substances in the second place. Thus, slope turns out to be a crucial factor, but unfortunately, in the open literature, results of toxicity testing are mostly reported in terms of effect doses or concentrations (e.g. ED50) and NOEL or LOEL values only. Although the situation has somewhat improved in recent years, typically, neither slope parameters nor original experimental data, allowing a re-fitting to appropriate regression models, are documented. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

85 The following sample calculations and simulations therefore make use of a high quality sample set of ecotoxicological data for the inhibition of algal reproduction. These of course have no direct relevance for human toxicology. However, according to the experience of the authors with the statistical analysis of actual data from chronic mammalian toxicity studies (Christiansen et al. 2008;Christiansen et al. 2009;Hass et al. 2007;Metzdorff et al. 2007), it can be stated that the dose response curves resulting from such studies are typically not steeper than those observed for algal reproduction. On the contrary, dose response curves for reproductive toxicity in rodents often appear to be less steep. As explained in the preceding sections, this would mean that quantitative differences between predictions of mixture toxicity by IA and DA can only be expected to be typically smaller than those observed in simulations with algal toxicity data, not larger. Thus, if relatively small differences between mixture toxicity predictions by IA and DA are observed with the algal toxicity data, this gives good reason to hypothesize, that the same might apply to estimates of chronic mixture toxicity in rodents. In this sense, the sample simulations were intended to provide a proof-of-principle type of evidence. The data set comprised original experimental data for the algal toxicity of 106 different chemicals, mainly pesticides, few anti-foulings, anti-biotics, and surfactants and some industrial chemicals. The data were compiled and generated as part the EU FP5 Project BEAM (Bridging effect assessment of mixtures to ecosystem situations and regulations, EVK1-CT , ), specifically with the aim of obtaining high quality descriptions of dose response curves that allow statistically valid estimates of low effect concentrations for the purpose of low dose mixture toxicity studies (Faust et al. 2001;Faust et al. 2003;Walter et al. 2002). To this end, they were all generated under identical testing conditions and with a considerably higher number of test concentrations than in usual regulatory testing protocols. Summary statistics derived from these data sets have been published in variety of papers; for the purpose of the simulations performed for this report, however, exclusive use was made of the original experimental raw data, which were all refitted to a variety of both continuous and threshold response models. To the potential range of ratios between predictions of effect concentrations for the algal toxicity of mixtures that can be generated from 106 different chemicals, computer simulations were performed in two steps: first a deterministic simulation on the possible extreme values for the prediction differences was conducted, and secondly a probabilistic simulation on the distribution of ratios between predictions based on IA and DA was performed, which also includes a comparison of the use of continuous dose response models with threshold models Deterministic simulations of extreme prediction differences Deterministic simulations were performed to calculate the extreme ratios between alternative predictions of effect concentrations (ECx IA / ECx DA ), both maximum and minimum, that may occur with any possible combination of toxicants that are included in the sample data set. In case of 106 toxicants, these are more than different combinations with 2 to 106 components. With the concentration ratio of mixture components, the ratio between predictions (ECx IA / ECx DA ) varies within a limited range. The deterministic simulation Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

86 results in an exact definition of these limits for any possible number of mixture components and for any preset effect level x. However, when all dose or concentration response functions of individual substances are described by using the Weibull model, these limits become even independent of the magnitude of the effect level x. For this simulation step, we therefore fitted all data sets to the Weibull model. The resulting values for the slope parameter β varied between β = 1.6 and β = 8.9 for all chemicals but one outlier, which had an extremely steep curve with β = A graphical impression of the meaning of these figures can be obtained by comparison with Figure 6, which shows Weibull sample curves with slope values between β = 1 and β = 10. In the sample data set, the mean was β = 3.7 and 80 % of the values exceeded β = ln 10 2,3. This means, that for most of the possible combinations of these toxicants, DA can be expected to predict a higher mixture toxicity than IA. Feeding these slope data into the computer simulations of maximum possible prediction differences yielded the following results (Figure 9): For any mixture that could be generated from the specific set of 106 chemicals the corresponding predictions of effect concentrations by IA and DA will never differ by more than a factor of 8.3, which is considerably smaller than the theoretical maximum value of 106. If mixtures are composed only from those of the 106 chemicals which have the concentration response curves with the lowest gradient, then situations occur, where IA predicts a higher mixture toxicity than DA. However, the ratio ECx IA / ECx DA will in no case drop below a value of 0.4, which means that the effect concentration predicted by IA will never be smaller than 40 % of the corresponding value predicted by DA. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

87 106 Endpoint: Algal Reproduction ECx mix IA / ECxmix CA Any Mixture Any Mixture that can be generated from a specific set of 106 toxicants: Maximum ratio = 8.3 Minimum ratio = number of mixture components Figure 9: Potential range of ratios between predictions of effect concentrations for the algal toxicity of mixtures that could be generated from a data set for 106 chemicals. Note that ratios between predictions ECx IA / ECx CA are plotted on a logarithmic scale Probabilistic simulations of distributions of prediction differences The results of the deterministic simulation studies are in agreement with empirical mixture studies and hence support the hypothesis that quantitative differences between IA and DA predictions are typically relatively small, at least in comparison to other uncertainties in regulatory risk assessments. However, these results are based on the use of a non-threshold regression model, and might therefore be misleading with respect to mixture risk assessment for non-genotoxic chemicals. For instance, if all compounds in the mixture are present at doses below their thresholds, but are estimated to be still effective, then IA erroneously predicts a mixture effect although no one should be expected. Hence, the question arises whether and to what extent the consideration of dose thresholds may affect the validity of the assumption that small quantitative differences between IA and DA predictions are the rule. In order to verify or falsify the validity of this assumption we performed probabilistic simulations studies on the expected differences between IA and DA with the same data set. In contrast to the deterministic approach, these simulations did not focus on the determination of those mixture compositions that result in maximum differences between DA and IA Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

88 predictions, but possible dose ratios were considered evenly. This was done to gain insight into the actual likelihood of significant over- or under-estimations that may result from the assumption of DA in cases where in fact independent joint action occurs. We investigated four scenarios with mixtures composed of 10, 20, 50 and 100 compounds, respectively. In a first step we re-analysed the toxicity data for all 106 chemicals by adopting the best-fit approach (Scholze et al. 2001) to a class of five non-linear regression models (weibull, logit, probit, generalized logit I and II) that were extended to include a threshold parameter. In short, all models were separately fitted to every data set, then for every chemical the best fitting model was selected according to statistical criteria, and only this one was then used for the subsequent simulation studies. In the second step, a hypothetical mixture was simulated with compounds sampled by random (without replacement) from the pool of 106 chemicals, and with mixture ratios also generated by random, so that every possible mixture composition had an equal chance to occur (toxic units were generated from the uniform distribution on the interval [0,1] and normalised in such a way that the sum of the components fractions was one). Finally, the corresponding ratios between effect concentrations predicted by DA and IA were calculated on the basis of the best-fit curves derived from the threshold models. This simulation step was repeated times in order to generate distributions of the ratios between predictions. For comparison, this procedure was finally performed a second time, by not using the threshold versions of the models, but the original versions which assume continuous dose response curves. As a result, we obtained probabilistic information about the likelihood of over-estimations (or under-estimations) of algal toxicity by the assumption of DA in situations of independent joint action for four different mixture scenarios and two different approaches to low-effect modelling. These distributions are depicted in Figure 10 and provide the following evidence: It is confirmed, that DA usually provides the more conservative mixture toxicity estimate. The likelihood for the reverse situation, i.e. that IA predicts a higher joint toxicity than DA, is very low and tends to zero with increasing numbers of mixture components. After simulations we never observed a ratio between the two predictions of mixture-ec50 values that was greater than 4.2. This is considerably lower than the maximum possible value of 8.3 that was determined for this data set in the preceding deterministic simulation (see section above). The difference results from the fact that all possible mixture compositions had the same chance to be included in the simulations, and it thus demonstrates how unlikely the occurrence of the possible extreme case is. An enormously high number of simulations would have obviously been required to cover also the highest possible values. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

89 Figure 10: Probabilistic simulations on predictions of algal toxicity of mixtures with 10, 20, 50 and 100 components, randomly composed from a data set of 106 different chemicals. Shown are the distributions of ratios between DA- and IA-based predictions of EC 50 values of the mixtures, resulting from simulations. Experimental data sets for individual compounds were either fitted to continuous regression models (black) or threshold models (red). Dotted vertical lines refer to the 95% percentiles of the distributions. Figures in the upper left corners provide the percentages of simulations in which IA estimated a higher joint toxicity than DA (ratio <1). As expected, the analysis of data sets by means of threshold-models leads to higher differences between the alternative predictions than the use of non-threshold models. In fact, this can be attributed to a resulting increase of the EC50 values of the mixtures that are predicted under the assumption of IA, because the corresponding predictions on the basis DA were unaffected by the inclusion of a threshold parameter in the regression modeling (data not shown). The differences between the outcomes of both modeling approaches increase with increasing numbers of compounds. However, even with 100 mixture components, they did not differ by more than a factor of 2.2. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

90 In additional simulation studies we also compared predictions for EC10 values of the mixtures instead of EC50 values. The results were similar (data not shown) Conclusions Existing experimental evidences, mathematical analyses and results from simulation studies give reasons to assume that quantitative differences between predictions of effect doses or concentrations for multi-component mixtures derived from the alternative models of independent action and dose additivity are relatively small for realistic assessment situations, not exceeding an order of magnitude, and typically only differing by a factor of less than 5 for mixtures with up to 100 components. Evidence for the validity of this assumption mostly results from ecotoxicological test data. There are reasons to assume that the principle also holds with data from mammalian toxicology, but further research on this point is required. Thus, in general the current status of knowledge about quantitative prediction differences supports the use of dose additivity as a pragmatic and precautious default approach to the predictive hazard assessment of chemical mixtures. Any lower protection level should be justified by specific toxicological knowledge about the mixture of concern. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

91 13.5. Toxicological interactions Toxicological interactions have been defined as any toxic responses that are greater than or less than what is observed under an assumption of additivity. (EPA 2000). In this context, additivity is typically the most appropriate concept from a choice of DA or IA. Interactions that results in an effect less than expected under additivity are referred to as antagonism, subadditivity or inhibition, and those resulting in an effect greater than predicted are described as synergy or supra-additivity. Consideration of interactions is important because the use of component-based approaches, such as DA or IA, assumes the absence of interactions. A very relevant question is how much the occurrence of an interaction could alter the mixture effect from that predicted under additivity. Of most concern is a possible synergy, or supra-additive interaction, that would increase the level of toxicity compared to that expected. The literature relating to this question has been critically reviewed (Boobis et al. 2011). Boobis et al. reviewed the experimental evidence for synergies at low doses (defined as doses close to points of departure for individual chemicals) in mixture studies and identified 90 studies, of which only 11 studies reported a quantitative estimate of a low-dose synergy. Three criteria were identified to make the quantification of synergy more consistent: Synergy should be defined as departure from the mixture prediction using DA A uniform procedure should be developed/used to assess synergy at low doses The method used to define the POD used to assess synergy should be standardised. Only 6 studies were considered to provide a useful quantitative estimate of synergy and these comprised three studies of binary mixtures, two studies of five component mixtures and one of an 18 component mixture. When the magnitude of synergy was calculated based on the ratio of observed to predicted dose for a fixed response ( Method A ) or the ratio of observed to predicted response for a fixed dose ( Method B ) it was found that the magnitude of synergy at low doses did not exceed that of the prediction made using DA by greater than a factor of 4. The number of studies identified was not great enough to allow comparison of the effect of using method A or B on the observed synergy. Boobis et al. noted that the role of interactions is the subject of continued debate amongst scientists and risk assessors, and that there is incomplete agreement on the impact of interactions following exposure to a chemical mixture. Given the results of their review, Boobis et al. considered that there is probably merit in the default regulatory approaches that assume toxicological interactions are not likely to occur at the low dose permitted under existing exposure standards however they acknowledged that this could not be a firm conclusion, especially for the effects of cumulative and low-level chronic exposure, and that, although the magnitude of observed synergies appears to be low, more work is required to determine the frequency of synergy in real world situations (Boobis et al. 2011). The presence of significant, unpredictable synergies could have questioned the use of any additivity concept, including DA, as a default in cumulative risk assessment. However if Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

92 significant synergies can be considered unlikely, as is indicated, then the suitability of DA as a conservative default is unaffected. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

93 13.6. Approaches to cumulative risk assessment methods The evaluation of experimental data describing the combined effects of chemicals, in this report referred to as mixture effect assessment, has to be distinguished clearly from approaches employed for conducting cumulative risk assessment in practice, here termed cumulative risk assessment methods. The application of cumulative risk assessment methods requires clarity about the goal of the assessment. The aim can be to arrive at a risk estimate, an estimation of safe levels, of margins of exposure, or can consist of ways of prioritizing certain mixtures, for further or for regulatory interventions. Estimations of safe levels or margins of exposure may be based on worst-case-assumptions, but the prioritization of mixtures (or affected sites) has to rely on fairly accurate quantitations of risk. Almost all cumulative risk assessment methods in current use are applications of the concept of dose addition. These include the Hazard Index (HI), Toxic Unit Summation (TUS), Point of Departure Index (PODI), Relative Potency Factors and the TEQ concept. Methods explicitly derived from independent action are not developed. An implicit application of independent action is the assumption that mixture effects will not arise when all chemicals in question are present at levels below their ADIs, with the additional implicit assumption that ADIs represent true zero effect levels. It should be emphasised that the implicit application of independent action can only be used for chemicals for which ADIs have been derived. However, this is only the case for a small minority of chemicals in current use Approaches based on dose addition (DA) Hazard Index The Hazard Index (HI) (Teuschler and Hertzberg 1995) is a regulatory approach to component-based mixture risk assessment derived from DA and which can be generally defined by the formula HI = n i= 1 EL AL i i where EL is the exposure level, AL is the acceptable level, and n is the number of chemicals in the mixture. Various measures for exposure levels and expectable levels may be applied; the only constraint is that EL and AL must be expressed in the same unit. Input values for AL can be ADIs or reference doses (RfD) for specific endpoints. If HI > 1, the total concentration (or dose) of mixture components exceeds the level considered to be acceptable. The method offers flexibility in applying different UFs when defining AL for the individual substances. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

94 An assumption implicit in the use of the HI approach, and one that derives from the principles of the DA concept, is that the acceptable levels AL for each individual chemical represent exposures associated with the same (small or negligible) effect. In most cases, this is not proven in practice, and will remain unproven in the foreseeable future. For most practical applications, however, the error in making this assumption can be considered small Toxic Unit Summation The method of Toxic Unit Summation (TUS) (Sprague 1970) is a direct application of the DA concept and defined by the formula TUS = n i= 1 TU i = n i= 1 c i ECx i where c i are the actual concentrations (or doses) of the individual substances in a mixture and ECx i denote equi-effective concentrations (or doses) of these substances if present singly (e.g. EC50 i ). The quotients c i / ECx i are termed Toxic Units (TU). Toxic Units rescale absolute concentrations (or doses) of substances to their different individual toxic potencies. They express the concentrations (or doses) of mixture components as fractions of equi-effective individual concentrations (or doses) ECx i. Typically, x = 50 % (EC50 i ) is chosen as the reference level, but TUS can also be calculated for any other effect level x. If TUS = 1, the mixture is expected to elicit the total effect x. If the sum of Toxic Units is smaller or larger than 1, the mixture is expected to elicit effects smaller or larger than x, respectively Point of Departure Index The Point of Departure Index (PODI) is an approach to component-based mixture risk assessment which is similar to the HI and TUS. In contrast to the HI, however, exposure levels (EL) of chemicals in a mixture are not expressed as fractions of individually acceptable levels (AL) but as fractions of their respective points of departure (PODs) such as NOAELs or benchmark concentrations or doses (BML). In this way, different uncertainty factors that may be included in AL values (see HI) are removed from the calculation (Wilkinson et al. 2000): PODI = n i= 1 EL i POD i A PODI lends itself to the estimation of margins of exposure for the mixture of interest. Similar to the HI, there is the implicit assumption that all PODs are associated with the same effect magnitude, a principle derived from the features of DA Relative Potency Factors The Relative Potency Factor (RPF) approach is another application of the DA concept for mixtures of chemical substances that are assumed to be toxicologically similar (EPA 2000). The concentrations (or doses) of mixture components are scaled relatively to the Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

95 concentration (or dose) of an index compound, and then summed up. The scaling factor is called RPF. The total toxicity of the mixture is assessed in terms of the toxicity of an equivalent concentration of the index compound. In general, the mixture concentration C m expressed in terms of the index compound for n compounds is C m = n i= 1 ( c RPF ) i i where c i is the concentration of the i th mixture component, and RPF 1 = 1, as i = 1 indicates the index chemical Toxic Equivalency Factors The Toxic Equivalence Factor (TEF) is a specific type of RPF formed through a scientific consensus procedure (EPA 2000). Based on the assumptions of a similar mechanism of action of structurally related chemicals and parallel concentration (or dose) response curves, they were first developed for dioxins. The total toxicity of the mixture is assessed in terms of the toxicity of an equivalent concentration of an index compound. The total equivalent quantity TEQ is estimated by summation of the concentrations (or doses) of mixture components c i multiplied by the respective TEF i : TEQ = n i= 1 ( c i TEF i ) Data requirements and applicability of the cumulative risk assessment methods All of the above cumulative risk assessment methods require at least rudimentary doseresponse information of individual mixture components which is used to derive the input values, be they ADIs, RfDs, POD or information about relative potencies such as RPF or TEF. Information about exposures must also be available. The HI sums up ratios of exposure levels and ADIs or RfDs over chemicals. These estimates can be arrived at by utilizing different uncertainty factors (UF) for each mixture component, in order to deal with differences in data quality and sources of uncertainty. If this is perceived to be inadequate, the PODI method can be used. PODI is based not on reference doses, but on points of departure (NOAELs, benchmark doses). Extrapolation issues (e.g. animal to human) are dealt with either by using one overall UF, or by estimating margins of exposure. The TEQ concept is predicated on the choice of a reference chemical and requires parallel dose-response curves for all components. Both these requirements are often not met by chemicals, but the method has been validated for dioxins and dioxin-like substances. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

96 Approaches based on independent action (IA) In general, CRA approaches based on IA are much less available than approaches based on DA, see section Approaches such as the Hazard Index or Toxic Equivalency Factors do not have counterparts founded on IA principles and, because of the difference in formulation of DA and IA, similar approaches may not be conceivable. One pragmatic application of IA is the stance that a mixture effect will not occur if each component is present at or below its individual zero effect level. However, this rests on the use of true zero effect levels, whose identification may be controversial, and should not be applied when effects are present but cannot be measured (when they are below the statistical detection limits of the assay). This issue is discussed in detail in sections 13.7 and 13.8, and has been thoroughly reviewed (EC 2009). Simplified IA approaches that are sometimes mooted are the notions that 1) the mixture effect is equal to the effect of the most potent component or that 2) the mixture effect is equal to the summation of the effects of the components. As discussed in sections and , these approaches appear to rely on assumptions about the correlation of susceptibility to the mixture components, these assumptions being rarely stated and potentially hard to substantiate. There would appear to be no practical approach for the use of IA in CRA, other than the assumption that mixture effects will not occur if the individual components are without effect. As discussed here and in section 13.8, this may not be a reasonable assumption. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

97 Use of the TTC in CRA approaches The TTC concept The TTC represents a level of human intake or exposure to a chemical that is considered to be of negligible risk despite the absence of any chemical-specific toxicity data (Munro et al. 2008). Use of the TTC is an approach to risk characterisation that balances the uncertainties in using data for other compounds against low levels of exposure. The TTC approach is not intended for chemicals with an established risk assessment procedure, such as food additives, pesticides and therapeutic drugs (Munro et al. 2008). The TTC was initially developed by the FDA for packaging migrants and used a single value, the threshold of regulation, of 1.5 µg/day. The rationale for this was summarised by Munro et al.; briefly, this is the level at which most carcinogens have a less than one in a million lifetime risk, and is times lower than the level at which other toxic effects occur (Munro et al. 2008). Subsequently TTCs were set by dividing the 5th percentile of the NOAEL distribution for e.g. Cramer class I chemicals, by what was termed the usual 100- fold uncertainty factor and multiplied by 60Kg. Use of the 5 th percentile gives a 95% chance that the NOAEL for an unknown is higher than the NOAEL assumed in the TTC. Three structural classes (see Table 11) of chemical were assigned TTC values of 1,800, 540 and 90 µg/day, followed by a TTC of 18µg/day for organophosphates and 0.15µg/day for chemicals with structural alerts for genotoxicity. TTCs for Cramer classes were originally intended for food flavours, the toxicities of other chemicals may have different NOAEL distributions that affect the intended precautionary value of the TTCs if this is not adjusted for. The separate TTCs for organophosphates and genotoxic structural alerts reflect this. Compounds for which TTCs are not appropriate include proteins, non-essential metals compounds and dioxin-like chemicals (Munro et al. 2008). The TTC approach may need modification (i.e. a modified input distribution of NOAELs or modified conversion process from NOAEL to TTC) for chemicals with specific routes of exposure (e.g. topical application of cosmetics), unless it is validated for these situations. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

98 Table 11: TTC values TTC, µg/person/day Cramer class Description 1800 Cramer class I Simple structures, efficiently metabolised, low potential toxicity 540 Cramer class II Less clearly innocuous than class I, no positive indications of toxicity, no lack of data for toxicity 90 Cramer class III Structural features that preclude a strong initial presumption of safety, or that indicate toxicity 18 Organophosphates 0.15 Structural alerts for genotoxicity A recent EFSA Opinion suggested that the TTC value for Cramer class II structures was not well supported by the underlying toxicological data, and proposed that class II structures should be classified as if they were Class III structures (EFSA 2011a). The Opinion listed categories of substances for which the TTC approach should not be used as: High potency carcinogens (i.e. aflatoxin-like, azoxy- or N-nitroso-compounds) organic substances Metals Proteins Substances that are known or predicted to bioaccumulate Substances with structures that are not adequately represented in the original databases from which the TTC values have been derived, e.g. nanomaterials and radioactive substances Substances likely to have the potential for local effects on the gastro-intestinal tract Use of the TTC in cumulative risk assessment A concept such as the TTC may be valuable in cumulative risk assessment since in most cases there will be a lack of toxicological data for some of the mixture components. The concept may allow such gaps to be bridged with conservative assumptions, rather than causing an assessment to founder on a lack of data. Price et al. explored the use of the TTC in estimating mixture toxicity, and concluded that use of the TTC led to conservative estimates that could be suitable for screening mixture assessments (Price et al. 2009). Price et al. found that the use of the TTC could overestimate the toxicity of a mixture if it was applied to a mixture component that was the dominant contributor of the mixture risk, but not if it was applied to a component that was not a Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

99 dominant component. Consequently it was suggested that the impact of using the TTC should be evaluated in each case, and if the overall risk estimate was driven by components for which TTC values were used then the estimate should be taken with caution (Price et al. 2009). In the unlikely, or perhaps undesirable, event that the TTC approach is used for all components in a mixture, then risk could still be deemed acceptable on this basis, providing exposure levels are low, but an indication of risk would indicate the need for refinement (rather than immediate concern or action). In this case refinement would mean the collection of toxicological data to replace the TTC. If a mixture assessment based on the TTC is used to prioritise components for data collection, then this is equivalent to prioritising on exposure levels (if all the TTC classifications are the same) or on exposure levels weighted by structural features (if more than one TTC classification is included). The use of TTCs in a mixture assessment is explored further in case 3 (section ). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

100 13.7. Cumulative risk from sub-adi levels A key reference point in single substances and mixture toxicology is the acceptable daily intake (ADI). ADI stands for an estimate of the amount of a substance in food expressed on a body weight basis that can be ingested daily over a lifetime, without appreciable risk to any consumer on the basis of all known facts at the time of evaluation, taking into account sensitive groups within the population (e.g., children and the unborn). (Regulation (EC) No 396/2005). Other reference values, such as the tolerable daily intake (TDI) from JECFA or the reference dose from U.S.EPA (RfD) are similarly defined. ADIs are derived by considering one substance at a time. It is frequently argued, most recently in the opinion of the three EU Scientific Committees SSCP, SCHER and SCENIHR on Toxicity and Assessment of Chemical Mixtures (DG Health and Consumer Protection 2011), that combination effects are not to be expected for dissimilarly acting substances when each chemical is present at levels around its ADI. The Committees stated that: The TDIs, DNELs or equivalent values are expected to represent a value at which no effects are produced; thus for threshold substances, the assumption is that this value is equal to or lower than the no-effect level; thus an E(C i )=0 should be assumed for exposures at the TDI or DNEL level. Consequently, the co-exposure to several substances all below the estimated TDI, DNEL or equivalent value should be assumed to be negligible if all substances have dissimilar modes of action. (emphasis added) As summarized in section , under IA mixture effects can be ruled out with certainty only if all components are present at doses equivalent to zero-effect levels. If, however, some compounds in the mixture produce small, but statistically not significant effects, joint effects may cumulate to significant levels, particularly when large numbers of chemicals are present that all affect a common endpoint. It is obvious that the principles of IA place a heavy burden on the ability to distinguish zero effect levels from small effects that may be impossible to verify experimentally. Thus, in connection with ADIs, three specific questions arise from the viewpoint of mixture risk assessment and mixture toxicology: Is it reasonable to associate an ADI with a zero-effect level in all cases? Are there effects below ADIs, which have no impact on health after exposure to single substances but may be relevant in the case of exposure to mixtures? Is it valid to extrapolate observations made with single substances or their mixtures in the dose range equivalent to no observed adverse effect levels (NOAELs) to arrive at risk estimates regarding sub-adi exposure levels? These considerations are of principal interest with regard to the assessment of mixture toxicity: The principles of mixture toxicity (DA, IA or interaction) are generic and apply to all endpoints, not only to adverse health effects. Therefore, the question of possible Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

101 effects at levels below ADIs, here referred to as sub-adi effects, has to be extended to non-adverse effects. A discussion of these issues is presented in the following sections. We will first deal with the question as to whether ADIs and TDIs can with certainty be associated with zero-effect levels, before turning to the issue of effects at sub-adi levels, so-called precursor effects Extrapolation from NOAEL or Benchmark Dose to ADI x 10 Assessment Factors To derive an ADI, it is customary to use a NOAEL from experimental animal studies for endpoints representative of long-term toxicity. The NOAEL is divided by an assessment factor (also referred to as an uncertainty factor) to arrive at an ADI. A default assessment factor of 100 is normally used. Instead of NOAELs, benchmark doses can also form the basis of ADIs. In this case, the lower confidence limit of a benchmark dose is combined with an assessment factor. The default factor of 100 that is used in these assessments is the composite of two separate assessment factors: a factor of 10 for interspecies extrapolation, and a factor of 10 for intraspecies extrapolation. For the purposes of this report, we disregard further assessment factors for LOAEL to NOAEL extrapolations, acute to chronic or for route-to-route extrapolation. EFSA recently initiated a project to re-evaluate and, potentially update, these assessment factors in the future. In specific cases, where data of high quality are present, assessment factors with lower numerical values can be used, although this has only been done in rare instances, as evidenced by numerous PRAPeR documents. NOAELs are not necessarily doses without adverse health effects. Especially because of the small number of animals used per dose group and the resulting limited statistical power, NOAEL may be associated with effects. EFSA have stated: The size of the effect at the NOAEL is, on average, over a number of studies, close to 10% (quantal responses) or 5% (continuous responses) (EFSA 2009a) Interspecies Assessment Factor The interspecies assessment factor of 10 is the composite of two factors, one for allometric scaling to take account of different body sizes between species, the other to adjust for differences in toxicokinetics. For allometric scaling from the rat to other species, a factor of 4 is generally used, and this is well supported by numerous evaluations (Kalberlah and Schneider 1998;Schneider et al. 2004). If the experimental studies are performed with mice, a Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

102 factor of 7 would be more appropriate. Conversely, a smaller factor of 1.7 is justified in the case of studies with dogs. Interspecies differences are not always fully addressed by the allometric scaling factor. Although justified in most cases, there may be deviations (towards smaller or larger species differences) due to toxicokinetic and/or toxicodynamic reasons. For this reason, an additional factor of 2.5 is introduced, which, however, is less well supported by experimental evidence. It may have been chosen for reasons of convenience: the product of 4 x 2.5 yields the traditional interspecies factor of Intraspecies Assessment Factor To cover differences in sensitivity between human subjects, an intraspecies variability factor of 10 is used. There is only limited empirical evidence to support the numerical value of this factor. WHO proposed to separate this factor into a toxicokinetic and a toxicodynamic component, with two subfactors of 3.2 (WHO 1999). Although not demonstrated on the basis of large datasets, it is generally assumed that among healthy human subjects there is greater inter-individual variation in susceptibility towards toxic substances than in experimental animals. The reason for this is that highly inbred, genetically homogenous strains of laboratory animals (mostly rodent species) are used for toxicological investigations, and these strains show a considerably smaller variability in susceptibility than human populations. A second reason for the higher inter-individual variability in susceptibility among humans is the presence of vulnerable groups in the general population such as the elderly, children (although for certain effects children should not generally be considered to be more susceptible than adults), or individuals with compromised health. The quantitative assessment of this variability is complicated. Hattis and colleagues have compiled a large database of individual studies, which investigated differences in effectrelated observations or toxicokinetic parameters in humans (Hattis et al. 1999a;Hattis et al. 1999b;Hattis and Anderson 1999). As a first attempt to use such data to put extrapolations on a sound footing, the database was utilized by Schneider et al. to derive distributions to be used in a probabilistic model for effect assessment (Schneider et al. 2006). For the data sets about differences in effect-related doses in adults retrieved from the Hattis database these distributions describe the ratios between the group mean effective dose and the dose causing the effect in a specified (low) quantile of the examined population. The following ratios of the medians of distributions were obtained by Schneider et al. from the Hattis-database (Schneider et al. 2006): dose ratio average susceptible (median) / 10% quantile of groups investigated: 3.31 dose ratio average susceptible (median) / 5% quantile of groups investigated: 4.82 dose ratio average susceptible (median) / 1% quantile of groups investigated: Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

103 This should be considered a proof of concept further efforts are needed to substantiate the numerical values by expanding and differentiating the database. WHO recommended that the intraspecies factor of 10, if possible, should be replaced by a data-derived factor for a particular substance (WHO 1994). It is further mentioned that in specific cases this factor of 10 might not be sufficiently protective. Reference is made to the case of genetic polymorphisms, which may cause large differences in the internal dose between population subgroups that carry different alleles of enzymes responsible for xenobiotic metabolism. In a recent paper, Dorne reviewed data on the inter-individual variability due to toxicokinetic factors (but omitting considerations of toxicodynamic differences between individuals which should also be taken into account) (Dorne 2010). When evaluating pharmacokinetic data the highest variability was observed with substances which are metabolised via pathways including polymorphically expressed xenobiotica-metabolising enzymes. According to this analysis, factors of up to 4.7 (99 th percentile, for CYP2D6-dependent pathways) would be necessary to consider resulting differences in internal doses. Relevant variability could also be caused by age (neonates with immature metabolic capacity for some specific pathways, and elderly people). In the subsequent paragraphs we discuss the level of protection which can be attributed to an ADI or equivalent value. In this context, effect data which are measured on a continuous, numerical scale (e.g. body weights, protein concentration in urine, etc.) have to be distinguished from dichotomous variables (categorical or quantal) which provide information about incidences, e.g. the number of affected subjects in a group of exposed individuals Continuous Responses In the context of risk assessments, a certain level has to be defined as the threshold for adversity for continuous effect variables (e.g. activity of transaminases in blood, protein concentration in urine, etc.). These critical effect levels have to be determined individually for each effect, as adversity depends on the nature of the effect, physiological background ranges, etc. In the hypothetical example in Figure 11 the critical effect level for excretion of protein x in urine was determined as a urinary concentration of 5 µg/l. Higher urinary protein excretion would be considered an adverse effect. The dose group showing a protein concentration of 5 µg/l (group mean) in urine was considered to represent the NOAEL. An ADI can be derived by application of assessment factors with the implicit assumption that in susceptible humans this concentration would also not be exceeded. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

104 Figure 11: Schematic presentation of ADI derivation from continuous variables from experimental animal data (dose-response curve on the right): POD is the dose at which the critical effect size (here: 5 µg/l) is not exceeded. By application of assessment factors this level of protection is transferred to the average human population and the subpopulation of susceptible humans Figure 12 translates this into a probability distribution showing the incidence for exceeding the critical effect level of 5 µg/l in the general population. Theoretically, at each dose level a distribution of urinary concentrations in the population can be assumed, with the percentage lying above the critical effect level decreasing with decreasing dose. The percentage of the population that exceeds this level is shown in Figure 12 as an incidence curve. In this figure the assessment factors applied are large enough to ensure that also for the most susceptible individuals the critical effect level is not exceeded at the ADI (incidence zero or close to zero %). In Figure 13 the population dose-response curve is moved to the left, with the consequence that, starting from the POD, the assessment factors are not large enough to cover also the most vulnerable individuals. The dose-response curve in the general population is rarely known. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

105 Figure 12: Schematic presentation of ADI derivation from continuously measured experimental animal data (dose-response curve on the right): Hypothetical case A: assessment factor sufficient to prevent exceeding a critical effect size in susceptible individuals Figure 13: Schematic presentation of ADI derivation from continuously measured experimental animal data (dose-response curve on the right): Hypothetical Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

106 case B: assessment factor insufficient to prevent exceeding a critical effect size in most susceptible individuals With continuous data, an increase of the number of animals per group will not automatically alter the group mean values measured. But larger experimental groups will have an influence on the uncertainty of the measured values (e.g. expressed by the standard deviations to the mean values). Thus, if the critical effect level is defined as absolute value (example: concentration of methaemoglobin not to exceed 5%) larger experimental groups will reduce the overall uncertainty of the measurement data, but will not tend to change the mean dose designated as POD Quantal (Dichotomous) Responses With quantal variables, the number of affected individuals within an exposure group is counted. The aim is to identify a dose group in which the frequency of the affected animals is significantly higher than with the next lower dose. The highest dose without significant effects is called NOAEL. With results from a standard repeated-dose experiment (e.g. a 90- day oral rodent according to OECD Test Guideline 408, with 10 animals per sex and dose group) it can be assumed that effects in the 10% incidence range might be detectable, depending on the type of effect, background rate and other factors. Alternatively, by dose-response modeling a dose associated with a pre-determined low effect level (e.g. 10% incidence above background) can be estimated as benchmark dose (BMD; for a 10% effect level: BMD 10 ). The lower confidence limit of the BMD is called BMDL. The following figures schematically present the ADI derivation for categorical data. Figure 14 exemplifies the situation where assessment factors lead to an ADI below a population threshold (incidence at ADI: 0%). In Figure 15 application of assessment factors results in an ADI connected with a relevant incidence for the adverse effect under investigation (approx. 5% in this hypothetical example). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

107 Figure 14: Schematic presentation of ADI derivation for categorical experimental animal data (dose-response curve on the right): Hypothetical case A: assessment factor sufficient to include most susceptible individuals in general population Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

108 Figure 15: Schematic presentation of ADI derivation for categorical experimental animal data (dose-response curve on the right): Hypothetical case B: assessment factor insufficient to include most susceptible individuals in general population As the dose-response curve of the human population is unknown, the actual level of protection provided by the ADI is also indeterminate. If the assessment factors applied are higher than that needed to take account of inter- and intraspecies differences, the corresponding incidence level in the human population will be well below the 10% level (in case of a BMD 10 used as POD). If the NOAEL is associated with a lower effect size, the ADI may well be equivalent to a zero effect level. In other cases, where the animal model used for a specific substance and effect is insensitive (which might lead to overlooking an effect) or where there is an especially sensitive population subgroup (e.g. with a yet unknown enzyme polymorphism), then the effect incidence in the sensitive population group (or depending on the size of this sensitive group in the general population) may even be higher than that observed in the experimental animal group at the POD. It should be mentioned that in the case of categorical data, those animals showing effects at the POD will be the more susceptible ones. Even when it is agreed that the variability in inbred experimental animals is much lower than that between human individuals it must be recognised that by applying an intraspecies factor to a dose derived from a categorical NOAEL, intraspecies variability is counted twice At the lower end of the dose-response curve the uncertainty in determining the dose-response relationship increases sharply (due to the intrinsic difficulty to determine low-level effects), with the consequence that even for the best investigated substances the dose-response Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

109 relationship in this area and the existence or non-existence of population thresholds becomes indeterminate. Primarily for gross apical endpoints, thresholds (i.e. dose level associated with 0 increased incidence) might also be observed at a population level. But in many cases even for substances with an assumed threshold, due to poorly understood toxicity mechanisms and lacking information on inter-individual differences (which, in case of a threshold mechanism means lacking knowledge on individual thresholds) dose- and effect estimates in the low-dose area will become uncertain Conclusion The quantification of residual risks associated with ADIs, especially for sensitive subgroups in the general population, is impossible. Major uncertainties arise from the (generally unknown) inter-individual variability in the human population. The assumption that the usually applied assessment factor of 100 sufficiently accounts for interspecies differences in vulnerability, as well as intraspecies variability to yield zero effect levels in all cases is not supported by empirical evidence, and largely unverifiable. The uncertainties associated with ADIs and TDIs are recognized by risk assessors. The three EU Scientific Committees (DG Health and Consumer Protection 2011) pointed out that there is obvious uncertainty in setting the TDI or DNEL. Furthermore, risk assessors do not exclude some residual health risk if exposures to doses equivalent to ADIs occur. There is regulatory guidance which advises as follows: when the critical effect is judged of particular significance, such as developmental neurotoxic or immunotoxic effects, an increased margin of safety shall be considered (Regulation (EC) No 1107/2009). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

110 13.8. The issue of non-adverse effects at levels below ADIs This section deals with evidence for first biological changes that might occur at dosages below ADIs. Such effects are not necessarily to be classed as adverse, but they too might sum up to significant levels when there is exposure to multiple chemicals, with consequences that are largely unexplored. The following aspects are of relevance: Effects below a critical effect size defined to separate adverse responses from those assumed to be of no harm, for continuous response variables. Issues arising from the dichotomisation of continuous variables into quantal variables that define individuals as affected or not affected. The possibility that upstream events leading to overt disease might cumulate to produce adverse effects, although the upstream events themselves, so-called precursor effects, may not be adverse. Biochemical alterations that occur in low dose ranges Effects below a critical effect size From a biological point of view, an effect may be irrelevant even though it is not strictly zero. Biological systems are capable of correcting certain disturbances provoked by exposure to chemicals, however the challenge lies in establishing a relevant effect size that defines the borderline between effect and no-effect and adversity and lack of harm in a biological or toxicological, and not a statistical sense. Several approaches exist to defining an effect of relevance quantitatively. In criterionreferenced evaluations, the importance of an effect is judged in relation to a clear biological or clinical criterion. An example would be the procedure for defining a critical sperm count below which fertility experts recommend assisted fertilisation. This is based on information about correlations between sperm count and fertilisation success in human populations. The criterion used was the point below which fertilisation rates began to decrease with lower sperm counts (Jorgensen et al. 2006). In many cases, however, straightforward biological or clinical criteria are not available and in these situations effects of relevance are defined by norm-referenced evaluations. This involves establishing a critical effect by considering the variance of the effect parameter in the population under investigation. There are numerous examples that follow this approach. Cut-off points for elevated cholesterol levels, low birth weights or late onset of walking in children are all derived by determining certain percentiles often the 95 th - of cumulative population frequency distributions of the selected effect variable. In mutagenicity testing, critical mutation frequencies are defined in terms of multiples of standard deviations of background mutation rates. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

111 If a critical effect size is used for deriving an ADI by defining a dose threshold for the experimental data there will always be a sub-critical effect size above zero, which is classed as not adverse and which is associated with doses below the chosen NOAEL or BMDL. A pertinent example is the critical effect size used by the Dutch RIVM for white blood cell counts during the assessment of mycotoxins (T-2/HT-2). Reductions of white blood cell counts of more than 5% were classed as critical. Conversely, reductions in counts of less than 4% are interpreted as non-adverse effects, and doses associated with these smaller effects are assessed as presenting no concern. Similarly, decreases of acetylcholine esterase activity of more than 20% are regarded as adverse (Bokkers et al. 2009). While these procedures may make sense when exposure to single chemicals is considered, problems might occur with exposures to multiple chemicals that also affect the endpoint under investigation. In such cases, effects below the critical effect size may well sum up to a combination effects above the cut-off effect sizes used for the assessment of single chemicals Issues arising from the dichotomization of continuous effect variables Issues equivalent to those discussed in connection with critical effect sizes may arise when continuous effect variables are dichotomized to classify individuals as affected or not affected, and when dose thresholds are subsequently defined on the basis of these classifications. In such cases, biochemical or precursor effects that might be present in individuals assigned as not affect might also build up to effects above the line regarded as affected, when exposure is to multiple chemicals Precursor Effects Biological events that precede overt adverse outcomes measured in chronic toxicity assays are often referred to as precursor effects. With regard to certain disease states, the scientific understanding of the molecular events, precursor effects, that advance pathological processes to a level where they become manifest as disease or adverse effect has improved considerably. It has been recognized that there is great potential in using these precursor effects, instead of adverse outcomes themselves, as tools for the screening of chemicals (Woodruff et al. 2008). In certain instances, it has been suggested to regard the precursor effects themselves as adverse. They occur at doses lower than those required to elicit overt effects in long-term animal studies. Examples for this stem from advancing knowledge about thyroid disruption, and from disruption of the function of male sex hormones. A prominent example is thyroid disruption. It is well established that thyroid hormone insufficiency during pregnancy can cause lasting neuro-developmental deficits in the affected child. Thyroid hormone levels are determined through a complex interplay between several factors, including uptake of iodine from dietary sources which is required for the synthesis of the hormone, transport of iodine to the thyroid gland, synthesis of the precursor of active thyroid hormone, transport of the precursor to tissues and deiodinisation of the precursor to the active thyroid hormone. Feedback processes exist to ensure that declining thyroid Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

112 hormone levels in the blood are compensated through increased synthesis in the thyroid gland. Thyroid hormone levels can serve as an early indicator of downstream adverse effects on neurodevelopment, and it has been suggested to regard diminutions of thyroid hormone as an adverse effect in itself, although it is distinct from the induction of overt neurodevelopmental effects (Woodruff et al. 2008). Suppression of thyroid hormone levels in experimental animals occur at lower doses than those necessary to produce overt toxicity. For example, perchlorate inhibits iodide uptake in the thyroid at rather low concentrations. Much higher levels of perchlorate are required to induce adverse effects on development or reproduction, secondary to a decline in serum thyroid hormone levels and an increase in thyroid stimulating hormone. This may lead to thyroid hypertrophy, hypothyroidism and finally to reproductive (prenatal or postnatal) effects. However, compensatory mechanisms may prevent an effect on circulating hormone levels. The NAS stated that one would need a 75% inhibition in iodide uptake for the perchlorate effect to be adverse (Murray 2005). There is good evidence that several chemicals can work together to induce reductions in thyroid hormone levels at doses where the individual chemicals are ineffective (Crofton et al. 2005). (Crofton 2008) demonstrated that chemical substances may interact with the thyroid hormone pathways at different sites and via different mechanisms. Woodruff et al. reported from a workshop in 2007, that the participants agreed that early biological perturbations and precursor effects (like hormonal changes) should be regarded as adverse outcomes in experimental studies (Woodruff et al. 2008). However, in current risk assessment practice, usually only the overt outcome, but not the precursor effect, is considered adverse, and the ADI (or TDI/RfD) is usually derived from such a late downstream end point. To illustrate the point, Woodruff et al. demonstrated a continuum between an adverse and a non-adverse outcome with precursor effects which may occur well below the apparent adverse effect threshold, which is the ADI/TDI after (interspecies- and intraspecies-) transformation as shown in Figure 16 (Woodruff et al. 2008). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

113 Figure 16: Distribution of a typical physiological parameter within the population and how that may vary depending on the influence of chemical and biological background (adopted from Woodruff et al.(woodruff et al. 2008) and reprinted with kind permission of the editor) Biochemical Alterations Biochemical changes are specific precursor effects at the molecular level. A no observed effect level for many biochemical changes may be well below the doses required for the induction of adverse effects which are used to derive ADIs. For example, the no observed transcription effect level (NOTEL) is frequently observed well below NOAELs for toxicity. Even if some substances in a mixture act via dissimilar modes of action, some of the isolated biochemical changes may be identical at very low concentrations. Zarbl et al. described the NOTEL as a much more sensitive indicator of (not necessarily adverse) effects and demonstrated the relationship in a schematic way (see Figure 17) (Zarbl et al. 2010). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

114 Linear extrapolation s e n o s p e R NOEL NOAEL NOTEL Experimentally observed Log dose Figure 17: Hypothetical dose-response curve as applied to risk assessment. The solid line indicates a typical response curve that shows a threshold for the experimentally measured endpoint. These data are then used to determine a NOAEL and/or a NOAEL. Data are then used to determine a reference dose. The use of genomics provides the opportunity to define much more sensitive NOTEL (indicated on graph), NOAEL or a mechanistically based benchmark dose for use in risk assessment. Citation from, and graph redrawn from, (Zarbl et al. 2010). On the basis of empirical data it is not always straightforward to demonstrate relevant biochemical changes at very low concentrations below a NOAEL or even below an ADI. Below, some examples from in vivo and in vitro studies are provided. Zheng et al. studied the hepatotoxic potential of four different substances with regard to biochemical marker proteins and gene expression (Zheng et al. 2011). The substances acted through different modes of toxic action. Dose ranges up to 1000-fold below those needed to affect conventional toxicity endpoints were chosen for this test. Effects on changes of gene expression of cytochrome P450 enzymes were observed. Judson et al. analysed perturbations of a biological pathway as an early event leading to adverse (apical effects) typically observed at higher concentrations, which they referred to as a biological pathway altering dose (BPAD) (Judson et al. 2011). Below a BPAD an Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

115 adaptive effect is assumed. Although derived from in vitro-studies results were transformed into an in vivo-dose estimate for precursor effects also covering variability and uncertainty of this estimate. Figure 18 (adopted from this publication) clearly points to BPADs well below the apical No Effect Level (NEL) and the Low Effect Level (LEL) for liver hypertrophy. The estimated BPADLs (red circles) are located below the NEL/100 (red triangle), which would correspond to an ADI (NEL/100), for 7 out of 14 examples for pesticides (conazoles). In the case of cumulative exposures some substances may well lead to identical perturbations of pathways which might affect apical endpoints if critical levels are reached. These examples provide additional indications for the possibility of effects below ADI for the single substances (in this case: pesticides), which should be considered in the assessment of combined exposures to mixtures. There are further examples of biochemical changes at very low concentrations (Abril et al. 2011;Ludwig et al. 2011;Schulpen et al. 2011;Thomas et al. 2011;Zheng et al. 2011). Figure 18: Comparison of high throughput chemical risk assessments with LEL and NEL values for liver hypertrophy from animal studies on the 14 conazole fungicides in Phase 1 of ToxCast. For each chemical, the black box gives the population-variability-derived (1%,99%) confidence intervals about the median BPAD. The whiskers indicate uncertainty-derived 95% confidence intervals about the extremes of the variability confidence interval. The BPADL99 is indicated by a red circle: the LEL by a blue box: the NEL by a grey triangle and the NEL/100 by a red triangle. Estimated chronic exposure Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

116 levels from food residues are indicated by vertical red lines. All values are in mg/kg/day (cited from Judson et al. (2011); for further explanation see description in text). Reprinted with permission from (Judson et al. 2011) Copyright (2011) American Chemical Society Conclusions The observations and issues presented in this chapter show that biochemical changes and precursor effects have been reported in dose ranges below those associated with ADIs or TDIs. Together with the uncertainties discussed in section 13.7, these reports cast further doubt on the notion that ADIs are true zero effects levels in all cases. The uncertainty is introduced because the dose-response relationship is rarely known for an adverse effect in the sensitive human population. This means that the ADI may not be a zeroadverse effect level for the critical effect observed above NOAEL. Extrapolations to the sub- ADI range do not necessarily mean a principle change from observations at/above NOAELlevel or a switch from high dose to low dose considerations and assumptions. The issues examined in this chapter support the possibility of mixture effects including those driven by IA at a sub-adi exposure level. The principles of IA may also hold true for nonadverse effects, which may cumulate to exceed homoeostasis and become adverse in case of exposure to mixtures. In conclusion, simultaneous exposure to multiple chemicals may result in cumulative effects even when the chemicals are present at levels below ADIs. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

117 13.9. Data requirements in legislation and availability of such data For (cumulative) risk assessment both hazard assessment and exposure data are needed. The complexity and quality of these data determines the quality of the risk assessment. A comprehensive legal framework exists in the EU which sets out the data requirements for toxicological and exposure data for substances in food. The following section summarises which data are legally required for hazard and exposure assessment in the EU and discusses the consequences for risk assessment Toxicology data available to regulators Food, whether from vegetable or animal origin, may contain ingredients, added intentionally or unintentionally, apart from its intrinsic constituent parts. It is EFSAs remit to perform scientific risk assessment for food and feed which form the basis for risk management decisions which should protect consumers from health risks from the food chain. In the following section we will analyse the possible differences in the amount and quality of data available on health risks associated with intentionally or unintentionally added food ingredients. Ingredients added intentionally to food should improve its properties such as taste, odour, stability or impart a certain property to food. These food improvement agents are food additives food flavourings food enzymes Their use is only permitted as far as no health risk is associated with their use or a special beneficial effect is achieved. As with food, some substances, so called feed additives, are added intentionally also to feed to improve its properties. These might be consumed by humans who eat food from animal origin. Furthermore, substances may be added for nutritional purposes like vitamins, minerals or certain amino acids in e.g. foods for particular nutritional uses or as ingredients in food supplements. Another group of food ingredients are substances which are used intentionally during the production or packaging of food. Their presence in food is usually not wanted but cannot totally be avoided. Therefore, concentration limits which should be without appreciable risk for the consumers have been established for these substances food contact materials residues of plant protection products Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

118 residues of veterinary drugs residues of biocidal products used in animal husbandry Other substances may be detectable in food or feed, although their presence is not wanted but cannot be avoided due to e.g. their overall presence in the environment: food contaminants feed contaminants ( undesirable substances in animal nutrition ) Legal requirements for the provision of toxicity data for food improvement agents Food additives, food enzymes and food flavourings make up the group of the so called food improvement agents. A common authorisation procedure has been laid down for these food improvement agents. A central part of the authorisation process is the safety evaluation of these substances to show that these substances do not pose a safety concern to consumers. The legal framework encompasses Commission Regulation (EU) No 234/2011 (EU, 2011) and Regulation (EC) No 1331/2008 (EU, 2008a) on a common procedure for evaluation and authorization of these substances, Regulation (EC) No 1332/2008 (EU, 2008g) on food enzymes, Regulation (EC) No 1333/2008 (EU, 2008f) on food additives, and Regulation (EC) No 1334/2008 (EU, 2008e) flavourings. Specific data requirements for risk assessment of food additives, food enzymes and flavourings are listed in Commission Regulation (EU) No 234/2011 and are summarised in the table below. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

119 Table 12: Specific data requirements (core data set) for risk assessment of food improvement agents Food additives Food enzymes Food flavourings examination for structural/metabolic similarity to flavouring substances in an existing flavouring group evaluation (FGE) Toxicokinetics Subchronic toxicity Subchronic toxicity Subchronic toxicity a Genotoxicity Genotoxicity b Genotoxicity b Chronic toxicity / Chronic toxicity and carcinogenicity a carcinogenicity Reproductive and Developmental toxicity a developmental toxicity a: where applicable b: specified in the guidance (see below): information on the ability to induce gene mutations and structural and numerical chromosomal aberrations Additional guidance on the evaluation of the food improvement agents are presented in three guidance documents (EFSA, 2009a; b; 2010). The guidance on food additives from 2009, which is currently under re-evaluation (finalisation announced for summer 2011), does not provide a specification of the requirements on toxicological data additionally to those requirements already addressed in the regulation. But in the former guidance of the Scientific Committee on Food (Guidance on submission for food additive evaluations; a detailed list of toxicological data necessary for the evaluation of food additives was presented: metabolism / toxicokinetics subchronic toxicity genotoxicity (gene mutations in bacteria, gene mutations in mammalian cells in vitro, (preferably the mouse lymphoma tk assay), induction of chromosomal aberrations in mammalian cells in vitro; positive results in any of the above in vitro tests will normally require further assessment of genotoxicity in vivo. chronic toxicity and carcinogenicity in two species reproduction and developmental toxicity (at least a two generation in one species and developmental toxicity data in two species) Additional studies (e.g. immunotoxicity, allergenicity, food intolerance, neurotoxicity, human volunteer studies, in vitro studies as alternatives to in vivo studies, special studies, acute toxicity, skin and eye irritation and skin sensitisation) may be indicated if necessary. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

120 The guidance on flavourings specified the needs for data on genotoxicity (information on the ability to induce gene mutations and structural and numerical chromosomal aberrations in vitro), developmental toxicity (a developmental in rodents) and repeated dose toxicity (a 90-day feeding in rodents, preferably in rats). Further toxicity tests may be requested by the Panel if a need for additional testing would arise from the submitted data. Also the guidance on enzymes specified the needs for data on genotoxicity (information on the ability to induce gene mutations and structural and numerical chromosomal aberrations in vitro, verification of positive results in vitro with in vivo tests) and an oral subchronic toxicity test. Additional studies to the core set may be requested. But it might also be possible to reduce or completely wave toxicological test, e.g. in case the safe use of the enzymes is documented. In summary, comprehensive toxicity data have to be submitted in the context of the authorization process of food additives. For food additives data on toxicokinetics, subchronic and chronic toxicity have to be generated as well as data on specific endpoints like mutagenicity, carcinogenicity or reproductive toxicity. Usually information on the NOAEL and LOAEL for all these endpoints will be available from these studies. The core set of toxicological requirements for food enzymes and food flavourings encompasses much less studies. Only if there are concerns about the toxicological hazards of these substances more data may be requested. This highlights the difficulties in cumulative risk assessment. If for example the chronic cumulative exposure to some food additives and some flavourings should be assessed, there might be the difficulty that for the food additives NOAEL-values from chronic studies are available, but only NOAEL from sub-chronic studies may be available for the flavourings. Due to the different designs, these NOAEL values are not directly comparable to each other. For further discussion see section Legal requirements for the provision of toxicity data for substances for nutritional purposes This group encompasses substances added for nutritional purposes in foods for particular nutritional uses, fortified foods or as ingredients in food supplements. Food for particular nutritional uses is used as a synonym for dietetic foods or dietary foods, i.e. these are foodstuffs for e.g. infants or young children in good health, or for persons whose digestive processes or metabolism are disturbed (Directive 2009/39/EC (EU, 2009)). Fortified foods means food to which micronutrients like vitamins or minerals have been added. The legal framework for addition of vitamins, minerals or certain other substances is described in Regulation (EC) No 1925/2006 (EU, 2006a). Food Supplements are defined as concentrated sources of nutrients or other substances with a nutritional or physiological effect whose purpose is to supplement the normal diet. They are Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

121 marketed 'in dose' form i.e. as pills, tablets, capsules, liquids in measured doses etc. ( Nutrients stands for vitamins and minerals (Directive 2002/46/EC (EU, 2002b)). Activities of EFSA have mainly concentrated on the regulation of minerals and vitamins. At the moment, the substances that can be added for nutritional purposes are controlled through positive lists (see Administrative guidance on submission for safety evaluation of substances added for specific nutritional purposes in the manufacture of foods ( _en.pdf). Before any substance is added to such a positive list an application for authorization has to be performed. Thus far, no formal procedure for the toxicological risk assessment of these substances for nutritional purposes has been published. For vitamins and minerals present in food supplements upper safe limits for the daily intake as well as minimum amounts per daily portion of consumption have to be established. Furthermore, reference intakes for the population (so called recommended daily allowances; Directive 2008/100/EC (EU, 2008d)) are requested (Article 5 of Directive 2002/46/EC). The maximum amounts should be based on scientific risk assessment based on generally accepted scientific data. Regarding the scientific opinions for the vitamins and minerals it becomes obvious that toxicological effects have been investigated comprehensively, i.e. whole data set from acute to chronic toxicity as well as reproductive toxicity are available for nearly all substances under consideration Legal requirements for the provision of toxicity data for food contact materials Food contact materials are defined as materials and articles intended to come into contact with foods such as packaging materials, cutlery and dishes, processing machines, containers, materials and articles in contact with water for human consumption. The legislative framework for food contact materials is laid down in Regulation (EC) No 1935/2004 (EU, 2004a). Food contact materials should under normal or foreseeable conditions of use not transfer their constituents to food in quantities which could endanger human health. According to Regulation (EC) No 1935/2004 Article 5 an authorisation procedure is required for: substances for use in food contact materials and articles, especially substances used in plastic materials and articles, substances used for active or intelligent functions in active and intelligent materials, and substances used in regenerated cellulose films materials and the articles themselves manufacturing processes of food contact materials and articles specifically chemically recycling processes for plastics to be used in plastic food contact materials Toxicological requirements during the authorisation process are described in the Guidance document on the submission of a dossier on a substance to be used in Food Contact Materials for evaluation by EFSA by the Panel on additives, flavourings, processing aids and materials in contact with food (AFC) ( The Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

122 amount of toxicological data which have to be submitted for authorisation depends on the quantity of the substance which might migrate into the food: high migration (5-60 mg/kg/food): extensive data set migration between mg/kg food: reduced data set low migration (<0.05 mg/kg food): limited data set Core extensive data set - 3 mutagenicity studies in vitro i) test for induction of gene mutations in bacteria ii) test for induction of gene mutations in mammalian cells in vitro (preferably the mouse lymphoma tk assay) iii) test for induction of chromosomal aberrations in mammalian cells in vitro - 90-day oral toxicity studies, normally in two species - studies on absorption, distribution, metabolism and excretion - studies on reproduction in one species, and developmental toxicity, normally in two species - studies on long-term toxicity/carcinogenicity, normally in two species Reduced core set - 3 mutagenicity tests - a 90-day oral toxicity - Data to demonstrate the absence of potential for accumulation in man Limited data set - 3 mutagenicity tests Additional studies or special investigations may be necessary in case if prior knowledge or structural considerations reveals that these studies are indicated. To prevent a risk to human health migration limits have been established for some components of food contact materials. For example for plastic materials an overall migration limit of 10 mg of substances/dm 2 of the food contact surface for all substances that can migrate from food contact materials to foods exists. For individual authorized substances specific migration limits (SML) have been fixed on the basis of a toxicological evaluation, i.e. with regard to the ADI or TDI of a specific substance. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

123 Legal requirements for the provision of toxicity data for novel foods and novel food ingredients Novel foods and novel food ingredients are foods and food ingredients that have not been used for human consumption to a significant degree in the EU before 15 May 1997 ( The legal background is laid down in Regulation (EC) No 258/97. Novel foods and novel food ingredients have to be notified before they can be placed onto the market. A risk assessment has to be performed for the notification. If there are any concerns or objections an authorisation process is initiated. Currently there are no obligatory recommendations on the toxicological data to be submitted for the risk assessment (see Recommendation 97/618/EC (EU, 1997b)) Legal requirements for the provision of toxicity data for residues Food from animal or plant origin may contain residues of plant protection products pharmacological active ingredients including pharmacological active biocidal products used in animal husbandry. Those residues should not be harmful to the consumer. Plant protection products, pharmacological active ingredients and biocidal products have to be authorised before they can be placed on the market. Comprehensive toxicological investigations and risk assessment is a central part of the authorisation procedure. If necessary, maximum residue levels (MRL) are established or the application of the substances is prohibited. According to Annex II of Council Directive 91/414/EEC (EU, 1991) the following toxicological tests are stipulated for the authorisation procedure of plant protection products: data on metabolism and toxicokinetics acute toxicity (at least after oral and dermal exposure) skin and eye irritation skin sensitisation toxicity after repeated exposure (90-day in rats and dogs) genotoxicity (gene mutation in bacteria, in vitro mammalian cell gene test, in vitro mammalian cytogenicity test plus additionally in vivo testing if necessary) long term oral toxicity in the rat carcinogenicity test in the rat and mouse two generation in rats Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

124 on developmental toxicity in rats and rabbits additional studies e.g. delayed neurotoxicity can be performed if necessary Testing procedures for pharmacological active ingredients and biocidal products are comparable to those of plant protection products, i.e. comprehensive toxicological data are available for these groups but are not reported here in detail Legal requirements for the provision of toxicity data for contaminants Food contaminants have not been added intentionally to food. These substances, like mycotoxins, heavy metals, polycyclic aromatic hydrocarbons or acrylamide may be present in food as a consequence of the production procedure, storage conditions, transport or packaging. Another large group of contaminants cannot be avoided in food due to their overall presence in the environment. Food contaminants shall be kept as low as can be reasonably achieved following good working practices. Maximum levels have been set for certain food contaminants in order to protect human health ( These maximum levels have been set on basis of toxicological risk assessments as well as considering socioeconomic and risk management considerations. There are no formal requirements for the toxicological investigations to be performed. But, for most of these substances large numbers of toxicological studies exist. If sufficient toxicological data are available Tolerable Weekly Intakes (TWI) are derived for food contaminants. The procedure for derivation of TWIs is basically comparable to the procedure for setting Acceptable Daily Intakes (ADI). These values are expressed as weekly intake, because the daily intake may show great variations in contrast to the weekly intake. To indicate that contaminants are not really wanted in food the guidance values are called tolerable instead of acceptable Legal requirements for the provision of toxicity data for undesirable substances in animal nutrition Undesirable substances in animal nutrition are the counterparts to food contaminants. Undesirable substances are any substance or product in and/or on the product intended for animal feed which presents a potential danger to human health, animal health or the environment or do not adversely affect livestock production ( Like for food contaminants maximum limits for e.g. heavy metals, dioxin, aflatoxin, certain pesticides have been established for feed materials, feed additives and feedingstuffs (see Directive 2002/32/EC (EU, 2002a) and its amendments Commission Directive 2003/57/EC (EU, 2003b) and Commission Directive 2003/100/EC (EU, 2003c)). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

125 Legal requirements for the provision of toxicity data for feed additives Feed additives should improve the quality of feed and the quality of food from animal origin, or improve the animals performance and health ( Feed additives have been classified in the following categories: technological additives (e.g. preservatives, antioxidants, emulsifiers) sensory additives (e.g. flavours, colorants) nutritional additives (e.g. vitamins, minerals, amino acids, trace elements) zootechnical additives (e.g. digestibility enhancers, gut flora stabilizers) coccidostats and histomonostats. The rules for authorisation, supervision and labeling of feed additives are set out in Regulation (EC) No 1831/2003 (EU, 2003a). Detailed rules for the implementation of Regulation (EC) No 1831/2003 of the European Parliament and of the Council are laid down in Commission Regulation (EC) No 429/2008 (EU, 2008b). As a basic requirement for the authorisation of feed additives studies concerning the safety of its use for the target animal and for the consumer are required. Based on the provided studies the safety for the consumer shall be evaluated and potential residues of the additive or its metabolites in food derived from animals given feed or water containing or treated with the additive shall be established. The aim of the safety evaluation is to establish an ADI. Toxicological data generally required are guideline studies on toxicokinetics; acute toxicity; genotoxicity (mutagenicity, clastogenicity; at least gene mutations in bacteria and/or in mammalian cells, induction of chromosomal aberrations in mammalian cells, in vivo test in mammalian species); sub-chronic oral toxicity; chronic oral toxicity/carcinogenicity; reproduction toxicity including teratogenicity (two generation and developmental toxicity ); This set of toxicological studies can be supplemented by other studies e.g. on immunotoxicity or neurotoxicity as far as necessary. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

126 Summary and conclusions The legal data requirements for the different substance groups are summarised in the following table. Table 13: Summary of the legal requirements for toxicological data for different substances added to food in the European Union (further details are presented in the text) Substance type Study type Feed additives Food additives Food enzymes Food flavourings Substances for nutritional purposes b Food contaminants c Undesirable substances in animal nutrition c Food contact materials with high migration d Food contact materials with medium Food contact materials with low migration d Novel foods and novel food ingredients b Residues from plant protection products Toxicokinetics/metabolism X X (X) X X X Acute toxicity X (X) X Skin and eye irritation X Skin sensitisation X Genotoxicity a X X X X (X) X X X X Subchronic toxicity X X X X (X) X X X Chronic toxicity/carcinogenicity X X X (X) X X Reproductive toxicity, X X (X) X X endpoint fertility Reproductive toxicity, X X X (X) X X endpoint developmental toxicity Additional studies (XX ) (XX ) (XX ) (XX ) (XX ) (XX ) (XX ) (XX ) (X X) (X): test not formally required but available in most cases (XX): further tests may be requested if scientifically justified a: usually a core set of in vitro tests is required which may be supplemented with additional tests if necessary b: no formal procedure for the toxicological risk assessment have been published c: substance not added intentionally to food or feed, no formal requirements on toxicological data d: high migration : 5-60 mg/kg/food; medium migration: mg/kg food; low migration: < 0.05 mg/kg food There is a substantial difference between substances added intentionally or unintentionally to food or feed. Substances added intentionally to food or feed have to fulfill certain defined data requirements which should enable the authority to assess the potential toxicological risk associated with the uptake of these substances. Besides the core set of data additional data can be requested in case there are any concerns about the safety of these substances. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

127 Data requirements may differ substantially between the different substance groups. Complete data sets are required for example for pesticides, nearly complete datasets for food and feed additives and food contact materials with high migration. But only rudimental toxicological data may be available for example for food contact materials with low migration if there are no further concerns. Up to now formal data requirements are missing for novel food and novel food ingredients. There are also no formal data requirements for food or feed contaminants, because their presence in food or feed is not wanted but cannot be avoided e.g. due to the overall presence of in the environment or their formation during the manufacturing process. But there is no manufacturer or importer who is responsible for these substances and who can be forced to perform toxicological studies. These potential differences in data availability may limit the options of cumulative risk assessment of chemical substances in food. Cumulative risk assessment may be performed with respect to acute or chronic toxicity or regarding certain endpoints like reproductive toxicity. At a higher CRA tier, endpoint specific points of departure (e.g. NOAELs, BMDLs) may be used. Table 13 obviously demonstrates the limits of such an approach. Based on the legal requirements it cannot be expected that for all substances present in food comparable toxicity data are available. For example, data on acute or chronic toxicity are not requested for all substance groups, even data on subchronic toxicity are not available for all groups. The problem for the risk assessor arising from this situation may be that no information on the NOAEL or LOAEL for a certain endpoint such as hepatotoxicity is available or that only information from toxicity studies with a different design is available, e.g.: For substance A a NOAEL for hepatotoxicity from a chronic toxicity with rats is available and for substance B a NOAEL (or even only a LOAEL) from a subchronic with dogs is available. For certain contaminants it is even likely that there are human toxicity data which have to be compared to toxic effects in experimental animals. The NOAELs or LOAELs from these different studies are not directly comparable to each other, and this may compromise efforts to aggregate such heterogeneous data by using methods for cumulative risk assessment. It would need further expert judgement to transfer these data by regarding species differences and differences in exposure time to comparable reference values. This illustrates the basic difficulties in cumulative risk assessment. Comparable data (e.g. NOAELs from toxicity studies with identical design) should be available for compounds from one substance group. But even this is not guaranteed, because certain studies have only be performed if there is a specific concern. But such studies are not legally required, which means that even within one substance group the quality and nature of the available data may be very heterogeneous Exposure data available to regulators Exposure assessment is the second important factor besides hazard characterization for risk assessment. Generally, exposure via diet ( dietary exposure or dietary intake ) can be calculated by multiplication of the concentration of a chemical in food with the food consumption (WHO, 2009). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

128 Dietary exposure = As in hazard assessment an acute and a chronic dietary exposure can be calculated, depending on whether an acute or chronic risk shall be evaluated. In the following section we will discuss the basic principles and the legal background for dietary exposure assessment mainly focusing on pesticide regulations. The other sections will only shortly be addressed Legal requirements for the provision of exposure data for pesticide residues Obviously, for a detailed exposure assessment sufficient and qualified data on the concentration of chemicals in food as well as on food consumption are necessary. The general legal basis for the collection of food exposure data in the EU is laid down in Regulation (EC) No 178/2002 (EU, 2002c). Regulation (EC) No 882/2004 (EU, 2004b) established the general legal provisions for food inspections and monitoring on official controls performed to ensure the verification of compliance with feed and food law, animal health and animal welfare. Regulation (EC) No 882/2004 applies to all food ingredients. In the context of plant protection products the preparation of annual reports on pesticide residues is based on the legal provisions laid down in Regulation (EC) No 396/2005 (EU, 2005), Chapter 5: Member states are required to establish national control programmes and to carry out regular official controls on pesticide residues in food commodities in order to check compliance with the MRL for pesticide residues and to assess the consumers exposure. The information has to be submitted to the European Commission, to EFSA and the other Member States. Apart from the national control/monitoring programmes (designed by each country) a second programme, a coordinate European programme, has been established (Regulation (EC) No 1213/2008 (EU, 2008c)) for pesticide residues. It is the aim of the EU coordinated programme to provide statistically representative data regarding pesticide residues in food available to European consumers. In November 2011 the 2009 EU Report on Pesticide Residues has been published by EFSA (EFSA, 2011). The report delivers a summary of the national monitoring data and inter alia an analysis of the results of the monitoring data on pesticide residues provided; some ideas about the reasons why certain MRLs were exceeded; an analysis of chronic and acute risks to the health of consumers from pesticide residues; an assessment of consumer exposure to pesticide residues based on the information provided under the first bullet point and any other relevant information available, including reports submitted under Directive 96/23/EC. In 2009 ten food commodities have been investigated in the EU coordinated programme. In a three year cycle investigating about 10 different commodities every year, the major Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

129 components of the European diet, which are represented by products, can be investigated by this procedure (the commodities analysed in the three year cycle represent between 39% and 95% of the diets of the different EU nations). These data will provide a representative basis for the estimation of the exposure to pesticide residues in food of European consumers (EFSA, 2011). In 2009 food from plant origin was analysed for 120 different pesticides (100 of them mandatory) and food from animal origin for 32 pesticides (29 of them mandatory). For risk assessment purposes the data on pesticide residues in the different food commodities were combined with the data on food consumption as presented in the EFSA Pesticide Residue Intake Model (PRIMo). The consumption data of PRIMo are based e.g. on national food surveys which collected long term and short term consumption data for the whole population or more differentiated for different age groups. Apart from the national food consumption data also data from the WHO cluster diets are incorporated in PRIMo which allows the consideration of different food patterns in the individual EU member states or a calculation for larger than national groups. Food consumption data for the EU member states are also available from the Concise European Food Consumption Database in Exposure Assessment ( However, this database only provides consumption data on a limited number (15) of broad food categories, which are only suitable for a preliminary exposure assessment. Apart from PRIMo which is based on national and international data on food consumption, national databases and models also exist. For example, the German Federal Institute for Risk Assessment (BfR) first developed the VELS model in 2005, which is based on the consumption data of children aged 2-5 years, which are considered to be the most sensitive subgroup due to the relative high food intake in comparison to body weight. In 2011 a new model for the calculation of dietary exposure has been established, which also includes the food consumption data of adults (aged years) and the subgroup of women in the child bearing age (aged years). These data are based on the results of a national food survey performed in 2006 (BfR, 2011). In the Netherlands a probabilistic model has been developed to calculate the acute or chronic intake of pesticide residues, the so called MCRA (Monte Carlo Risk Assessment; (Pieters et al., 2005). Probabilistic models are also available from Ireland (CREMe 2), United Kingdom (CSL), the University of Bremen, Germany, and the USA (DEEM/Calendex, CARES, Lifeline and SHEDS) (EFSA, 2008). Discussion Monitoring programmes deliver numerous data on pesticide residue concentrations in food. Apart from the data from surveillance samples which are collected without any particular suspicion there are also enforcement samples. These are taken if there is suspicion about the safety on non-compliance with the legal limits (i.e. these data are not representative for the food available on the EU market). They are collected as part of the national food monitoring programmes and compiled by EFSA for the EU member states. In 2009 about 120 pesticides have been monitored in 10 commodities in the EU coordinated programme. Within the 3 year circle of the EU coordinated programme data on the most relevant 30 food commodities are collected. The EU coordinated and national data on chemical concentrations in food are Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

130 combined with food consumption data using excellent models which reflect different food patterns from several European countries. Thus, dietary intake of pesticide residues in the EU is very well documented due to the legal requirements in the EU. Nevertheless, there are some shortcomings in this system. Currently about 500 pesticides are authorized in the EU, and probably 1000 worldwide (EFSA, 2011). Although the EU the three-year-monitoring programme delivers an enormous data base, not all pesticides on the market can be covered due to costs and efficiency reasons. Furthermore, not all food commodities can be considered: MRLs have been set for about 400 food commodities. These are partially covered by the national monitoring programmes, but data gaps still remain, because not all residue data in one commodity can be transferred or extrapolated to another. An additional difficulty in pesticide monitoring is that MRLs are established for Raw Agricultural Commodities (RAC) of plant or animal origin. In some cases the MRL refer not only to the edible parts of the plant but also comprise inedible parts (e.g. bananas with peel). Food monitoring data are established in a way that the compliance with the legal requirements can be checked. But this does not necessarily mean that these data deliver information of the real pesticide residue concentration taken up by the processed food Legal requirements for the provision of exposure data for substances in live animals and products of animal origin The legal requirements for substances in live animals and products of animal origin are comparable to those for pesticide residues: The general legal basis for the collection of food exposure data is laid down in Regulation (EC) No 178/2002 and Regulation (EC) No 882/2004. The requirements for substance control planning for live animals or products of animal origin are laid down in Council Directive 96/23/EC (EU, 1996). This is amended by Commission Decision 97/747/EC (EU, 1997a) which establishes levels and frequencies of sampling for certain animal products. As for pesticide residues, EFSA generates an annual summary report on the results of residue monitoring in food of animal monitoring in the Member States which comprises not only results on residues from veterinary products but also other substances (e.g. pesticides) and contaminants (EC, 2010). The underlying data on substance concentrations in the different food stuffs can together with the existing data on food consumption be used for a calculation of the dietary intake of substances from live animals and products of animal origin Legal requirements for the provision of exposure data for food improvement agents Intake of food additives and flavourings has to be monitored by the member states which have to report their findings to the Commission and Authority (Article 27 of Regulation (EC) No 1333/2008; Article 20 of Regulation (EC) No 1334/2008). There are no specific requirements for enzymes regarding food monitoring. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

131 Legal requirements for the provision of exposure data for food supplements Food supplements are provided in a dosed form which contains a recommendation for the daily intake and a warning not to exceed the stated recommended dose. Effective monitoring should be guaranteed (see Directive 2002/46/EC). Maximum levels and minimum effective doses have also been set for vitamins and minerals added to foodstuff. But no monitoring programme like for the other food ingredients has been established Legal requirements for the provision of exposure data for food contact materials During e.g. processing, manufacturing, or packaging food comes into contact with numerous materials (plastics, metals, ceramics, etc.). To ensure that these materials do no liberate substances into food in an amount that would impair food safety official controls have to be performed in order to perform compliance with Regulation (EC) No 1935/2004. The European Reference Laboratory for Food Contact Materials (EURL-FCM) has been established in 2003 which is responsible for the production and dissemination of internationally accepted quality assurance tools, including validated methods, reference materials, reference measurements, interlaboratory comparisons and training. The EURL- FCM cooperates with a supporting network of European National Reference Laboratories Legal requirements for the provision of exposure data for food contaminants Data on contaminant concentrations in food are collected in Europe on basis of Regulation (EC) No 882/2004 and Regulation (EC) No 1881/2006 (EU, 2006b) which are amended by several other regulations like e.g. Commission Regulation (EC) No 333/2007 (EU, 2007) where the methods of sampling and analysis for certain food contaminants are described. Coordinated data from Member States can be collected within a scientific cooperation (SCOOP) task. The legal basis for this procedure is Council Directive 93/5/EEC (EC, 1993). This directive lays down a procedure whereby Member States can focus their scientific resources in a co-ordinated manner on problems in the area of food. Pooled data on particular issues of concern regarding food safety can be collected for different substance groups. Beyond these EU wide activities, the WHO Global Environment Monitoring System Food Contamination Monitoring and Assessment Programme (GEMS/Food) contributes to a relevant amount to the measurement of contaminant concentrations in food worldwide. But GEMS/Food also established important data on the food consumption worldwide (GEMS/Food cluster diets) Summary and conclusions Monitoring programmes deliver numerous data on the concentration of pesticide residues, food contaminants and other ingredients of food. Apart from the data from surveillance samples which are collected without any particular suspicion, also enforcement samples which are taken if there is suspicion about the safety on non-compliance with the legal limits (i.e. these data are not representative for the food available on the EU market) are collected in Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

132 the food monitoring programmes. These data have to be submitted to EFSA and many of these data are published by EFSA. Apart from the data on substance concentrations in food in the last years several programmes became publicly available which provide information on food consumption and can be used for the calculation of dietary intake. While qualified monitoring data are necessary for a skilled and realistic exposure assessment, the existing models also allow a worst case consideration: Based on the existing food consumption data and the legal MRL for the individual commodities maximal intakes can be calculated for a lower TIER in a tiered risk assessment approach. Although the existing monitoring system provides large datasets on dietary exposure there are still some relevant shortcomings in the assessment of chemicals via dietary intake. For example monitoring programmes can never encompass all substances possibly contained in food and will never cover all food commodities. Data on substance concentrations as collected by food monitoring must not necessarily represent the substance uptake by food due to further processing of the food, i.e. they usually represent a worst case. This means that an uncertainty analysis of the data, whether the calculated intake is rather an over- or underestimation of the real intake, should be performed. Today generally qualified monitoring data are available which can be used for a skilled exposure assessment. But due to the fact that monitoring is always limited due to cost and effectiveness reasons there will always be data gaps as not all substances potentially added to food (un-intentionally or intentionally) can be monitored in all commodities. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

133 Cumulative exposure assessment (CEA) The data requirements for use of IA or DA were considered in the relevant sections on each additivity concept; however these sections concentrated on the requirements on effect data, rather than the requirements on the corresponding exposure data that is also required for a risk assessment. We therefore considered whether the choice of IA or DA alters the data requirements for cumulative exposure assessment (CEA) or if the exposure assessment is essentially independent of the model. Consideration of the mathematical and practical features of the concepts, sections 13.1 and 13.2, suggests that the latter situation applies, in that a risk assessment using either IA or DA concepts requires exposure data but the choice of model does not alter the type or amount of information required. The availability of data for CEA is currently a major knowledge gap in that most studies do not measure exposure to multiple chemicals simultaneously and therefore assumptions must be made as to how the exposure to multiple chemicals should best be estimated when the information for each chemical is derived from different samples and, often, different studies. Knowledge of variations in cumulative exposure within populations is also lacking, for example whether certain members or sections of the population are highly exposed to multiple chemicals or if exposure is more balanced. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

134 TASK 3: GROUPING CRITERIA The literature areas relevant to this task were the definition of dissimilar action, mode and mechanism of action; the literature was collected and summarised in Tasks 1 and 2, and presented in sections 10.1, 11 and The insights from these activities are directly relevant to the proposed approach (Task 5) and, to avoid duplication, further description has been incorporated into that section of this report (section 15.7). A 2008 report Cumulative risk assessment for phthalates - the tasks ahead" (NRC 2008) contains several insights into the grouping of chemicals with varying mechanism of action which now are reviewed here. The report noted that there are ambiguities in the criteria that are proposed for grouping chemicals in CRA, and that the use of grouping criteria that are inappropriately narrow may prevent chemicals that do indeed cause a common toxic effect from being assessed together. The report strongly recommends a physiologically based approach to grouping in which chemicals that cause any or all of the end effects attributed to a selected physiological process would be grouped together (NRC 2008). For example, during development the action of androgens drives male sexual differentiation, and this can be disrupted through a variety of molecular mechanisms, such as androgen receptor antagonism and reduced testosterone synthesis, producing end effects in males such as nipple retention and hypospadias. A physiologically based approach would group all chemicals that caused either end effect, nipple retention or hypospadias, together for assessment. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

135 TASK 4: ASSESS APPROACHES TO CRA FOR DISSIMILARLY ACTING CHEMICALS, EVALUATE THE DOSE ADDITIVITY APPROACH We have assessed the available approaches and guidance for risk assessment of mixtures, including European approaches ((EFSA 2008b;EFSA 2009c), American (EPA 1986;EPA 2000) and international (IPCS/WHO 2009;Meek et al. 2011) approaches. We have concentrated on areas of the guidance relevant to a dissimilar mode of action. These approaches are described below. The use of DA as a conservative default assumption was reviewed by (EC 2009) and is discussed in this report in sections 13.15, , 13.5 and The situations where the DA approach might not be sufficiently conservative, i.e. if IA was the more appropriate model and DA under-predicted risk, were reviewed in section To summarise, an experimental situation in which DA produced a less conservative mixture effect prediction than IA was not identified in our systematic literature search (section ). 14. Existing approaches to CRA for dissimilarly acting chemicals Opinion of the Scientific Panel on Plant Protection products and their Residues to evaluate the suitability of existing methodologies and, if appropriate, the identification of new approaches to assess cumulative and synergistic risks from pesticides to human health with a view to set MRLs for those pesticides in the frame of Regulation (EC) 396/2005 (EFSA 2008b). The Panel limited their considerations to plant protection products for pragmatic reasons (the general lack of data for other products) and also limited their considerations to the impact of dose addition, deliberately omitting response addition (IA) and interactions. The Panel proposed criteria for including chemicals in a common assessment group (CAG) and highlighted the possibility of refining the group in a stepwise fashion. Criteria for grouping included general criteria (chemical structure, mechanism of pesticidal action) or more refined criteria (common toxic effect, toxic mode of action). The approaches considered for CRA were, in order of increasing complexity and refinement, the hazard index, the reference point index, the relative potency factor method and physiological based toxicokinetic (PBTK) modelling. Specific Panel recommendations of relevance to this project task were: To use a tiered approach to toxicological evaluation and intake estimation To evaluate assumptions and uncertainties at least qualitatively, with quantitative examination of those factors potentially critical to the outcome Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

136 To increase the usefulness of exposure data for CRA by, for example, designing monitoring programs and establishing reporting levels with the needs of CRA in mind Scientific Opinion on Risk Assessment for a Selected Group of Pesticides from the Triazole Group to Test Possible Methodologies to Assess Cumulative Effects from Exposure through Food from these Pesticides on Human Health (EFSA 2009c). The 2009 Opinion carried out a cumulative risk assessment for triazole fungicides using the methods proposed by the earlier opinion (EFSA 2008b). As a result of this experience, the panel proposed a simplification of the tiered approach, as follows: From an early stage, the CAG should be as refined as the data allows. Exposure assessments should be limited to one deterministic and one probabilistic tier The panel concluded that although a tiered approach is an appropriate way to address cumulative dietary risk assessment it cannot yet be applied on a routine basis The PPR Panel identified the following issues: the basis for and establishment of CAGs on a European level, definition and agreement on desired levels of protection, improvement of the robustness of methodologies of cumulative exposure assessment and development of guidance on their appropriate use. The panel also noted that: when assessment of a CAG based on relatively broad criteria, due the absence of information on mode or mechanism of action for the common toxicological effect, fails to give adequate reassurance, this may serve as a trigger for further research, to enable the assessment to be completed. The establishment of relevant CAGs is the starting point for all cumulative risk assessments. Consensus should be reached at an international level on the criteria and compounds that should be used to create a CAG, to avoid differences between national cumulative risk assessments. An important issue is that a first tier should be more conservative compared to the next tiers. In itself, the hazard assessment tiers are clear and could be performed for any CAG. An important lesson from the case could be that a variety of methods are now available but that data on exposure and toxicological effects are not readily available for use with the methods. The greater availability of data, on standardised platforms, would actually assist in Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

137 the comparison of methods, since there could be less focus on compromises that tolerate data gaps and the risk estimates from different methods can only be compared when the method is able to run through to completion. EPA: Guidelines for the Health Risk Assessment of Chemical Mixtures (EPA, 1986) and Supplementary Guidance (EPA, 2000). EPA guidance has several areas of relevance. If a component based approach is selected, the guidance indicates a choice between toxicologically similar mixtures, with assessment by the hazard index or relative potency factor approaches, and toxicologically independent mixtures to be assessed by response addition (i.e. IA). (Figure 2-1, EPA, 2000). The Guidance notes that this choice may rely on data that is not readily available or may require scientific judgement. It is also noted that the true toxicologic mechanism of action...is rarely known for a given mixture or even for most of its components and the judgements that are made of toxicologically similar action or independence of action, for example, will be uncertain. The Guidance proposes that the assessor should deal with this by implementing several of the assessment approaches, and evaluating the range of health risk estimates that are produced. The Guidance also states that: response addition (IA) is the default approach when the component chemicals are functionally independent. dose addition and response addition... represent default approaches for toxicologically similar and toxicologically independent chemicals, respectively. Guidance section titled Criteria for Dose Addition [DA] vs. Response Addition [IA] The key criterion for choosing [DA or IA] is the similarity or independence among the chemicals in the mixture. Thus, in the EPA Guidance the overall distinction is between similarity and independence, for use of DA and IA respectively. The concept of dissimilarity is not emphasised. WHO/IPCS: framework for the risk assessment of combined exposure to multiple chemicals (Meek et al. 2011) The WHO/IPCS framework is described in more detail in section The framework presents dose additivity as the default assumption for estimating risk in all tiers (Meek et al. 2011) and notes that The use of dose additivity is considered conservative based on analysis of empirical results for effects of combined exposure including to chemicals that induce critical effects by different modes of action (US EPA, 2007; EFSA, 2008; European Commission, 2010). The use of independent action is not considered in any tier of this framework, consequently no guidance is given, or needed, to address the question of similar and dissimilar chemicals. However in higher tiers, use of MOA information is proposed but not qualified by consideration of the use of independent action to cumulate refined groups. Tier 2 allows for the refinement of the assessment group through consideration of increasingly more specific Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

138 information on mode of action or other factors on which to base the grouping (e.g., molecular modeling). For some substances (e.g., cholinesterase inhibitors), this may be at the level of a molecular target. or the use of relative potency factors based on a selected index chemical. Tier 3 can incorporate increasingly refined information on mode of action, including both kinetic and dynamic aspects. These can include both physiologically based pharmacokinetic (PBPK) and biologically based dose response (BBDR) models If the refinement in Tier 2 or 3 results in one group being replaced by two or more subgroups, it becomes possible for risk to be underestimated unless there is a means of aggregating the risk of the subgroups. If the risk is aggregated by DA, then the use of subgrouping becomes trivial since DA would be used within AND between the subgroups and would be mathematically equivalent to the use of DA on the whole group. IA could be used between subgroups, with DA used within the subgroup, if there are experimental or theoretical grounds to expect independence of effects of each subgroup, but the methods for doing this within CRA are not currently used or generally available. This chance of underestimating risk would not occur when risk estimates are refined through refined exposure assessments, consequently we consider that exposure assessments should be as refined as possible before grouping on the basis of mode/mechanism information is introduced. In this way the risk estimate may become acceptable before reliance on contentious subgrouping becomes necessary Commonalities of approaches Several commonalities emerge from consideration of the various approaches to risk assessment: Use of a tiered approach to deal with data availability, and a recognition that data availability is a potentially limiting factor. The use of tiered approaches is discussed in detail in section The need for a default assumption to allow assessment to proceed in the absence of detailed mechanistic data Agreement on the choice of additivity concepts: DA and IA Practical implications for CRA of chemicals with diverse uses and properties and present in foods Practical mixture assessment approaches have generally been formulated for pesticides rather than other chemical groups. The need to extend methods developed with pesticides in mind to other groups of chemicals in order to examine many, or perhaps all, types of chemical within one CRA could be affected by the difference between pesticides, which are relatively well studied and have a clear regulatory approach, and other groups of chemicals that are less well studied. Issues include assumptions about typical exposure levels that may be different for pesticides than for, for example, food additives which are deliberately added to food and food Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

139 contaminants which arise from environmental contamination. Environmental contaminants may be less responsive to regulation and harder to manage, and may exceed their regulatory limits, if they exist, more often than pesticides. Their chemical properties are also likely to differ from chemicals designed for a purpose and intended to have a low human hazard, for example the persistent organic pollutants (POPs) compared to modern pesticides, with rapid human metabolism. The data available for pesticides on e.g. mode of toxic action, although by no means complete, is also likely to be more substantial than that for other chemical groups. Finally, although pesticides constitute a diverse group of chemicals, the inclusion of nonpesticides in an assessment is likely to increase diversity on most metrics, for example chemical properties, uses, human exposure and toxicological effects etc. The inclusion of all relevant chemicals in a CRA is scientifically important, since a division of chemicals according to usage (e.g. pesticide, food additive etc) is unlikely to mirror a division in toxicological risk. Thus, the scientific credibility of a CRA will be increased by the inclusion of all relevant chemicals (those to which human exposure occurs) irrespective of regulatory divisions Assessment frameworks and tiering Several ways of dealing with mixtures in chemicals risk assessment have recently been proposed and discussed (IPCS 2009;Kortenkamp and Hass 2009;NRC 2008). Depending on the quality of the data and the level of detail available for mixtures risk assessment (data poor or data rich), tiering methods are useful for exploring the problem to be assessed. When appropriate, more sophisticated models and associated supporting data can be used. As a representative example for the approaches discussed in the literature we describe the framework analysis for cumulative risk assessment proposed by (IPCS 2009). The IPCS framework analysis begins by considering the nature of the exposures in question and assesses whether the key mixture components that make up combined exposures are actually known. This first step of the analysis may already expose knowledge gaps that preclude continuation of a framework analysis. Next, IPCS proposes to analyse whether exposure to specific chemicals is likely in the setting relevant to the assessment. In some cases, chemicals may not be released into the environment, or degradation is known to be very rapid, thus minimising the chances of exposure. If exposure is judged to be likely, the analysis can proceed by considering whether there are opportunities for co-exposure within a relevant timeframe, and which substances are likely to occur together. If the chances of co-exposure are negligible, the analysis may be terminated. In cases where combined exposures are anticipated to take place, the next logical question to be answered concerns the identity of chemicals that should be considered together in cumulative risk assessment. At this stage, there are several options for proceeding with the analysis: Some chemicals may be subjected to cumulative risk assessment because they occur together in relevant exposure scenarios. In other cases it may be appropriate to select only Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

140 those that affect a common endpoint. The analysis can be further narrowed by restriction to chemicals that affect a common endpoint by a common mode of action. To avoid that unnecessary resources are invested to reach decisions about the risk assessment issue under investigation, the IPCS framework has adopted a hierarchical structure, with several tiers, depending on data quality, data demand and level of detail that is accessible. At the lower tiers, relatively crude assumptions are made about exposures and potencies of the chemicals that make up the mixture to be assessed. If the lower tiers already indicate negligible risks, no further efforts are investigated to refine the analysis, and the assessment is terminated. At the lowest tier, Tier 0, crude and semi-quantitative exposure estimates may be used for the development of a HI. Accordingly, the ADIs entered into the calculation of a HI may not necessarily be comparable, having been derived from a variety of different endpoints, or by applying UFs in inconsistent ways. At this stage, it may also be assumed that all chemicals in the mixture are as potent as the most potent chemical present, if more appropriate data are not available. Although these simplifications and inconsistencies may not be in accord with the assumptions of the DA concept, they reflect the realities of chemical safety testing which is geared towards identifying the most sensitive toxicities of a chemical. As a result, potency information in relation to the same adverse outcome, as required by DA, is often not available and this may limit the analysis. The procedure is considered to be sufficiently conservative and protective, given that the input values are based on critical effects that occurred at the lowest dose for each chemical. In a risk characterisation step, the margin between estimated exposure and hazard is considered as a decision basis for determining whether a more refined analysis is required. If the Tier 0 HI is judged to be too large, the analysis can proceed to Tier 1, with the aim of introducing refinements wherever possible. At this tier, the assumptions on exposures may still be deterministic and may reflect worst case assumptions. For hazard assessments, the IPCS document judges it permissible to assume that chemicals have the same potency as the most potent known chemical present in the mixture, should better data not be available. Where possible, more accurate estimates of potency may be incorporated, such as benchmark doses. Alternatively, PODIs may be calculated. The risk characterisation step determines whether further refinements of analysis should be conducted. In Tier 2 analyses, deterministic exposure assessments are further refined, by incorporation of measured data, additional parameters and more realistic scenarios. Hazard assessments may focus on more restrictive assumptions about chemicals to be included in the analysis, e.g. by including more information about modes of action, or the nature of endpoints that are affected by the mixture components. Only PODs in relation to common adverse outcomes may be included in Tier 2, or an analysis based on RPFs, with an index chemical, may be attempted. IPCS suggests adopting the PODI method at this stage, with a consideration of the margin between exposure and hazard as a basis for decisions as to whether higher tier assessment are required. As an example for a Tier 3 analysis, IPCS proposes the use of probabilistic exposure assessments. Tier 3 assessments for hazard incorporate increasingly refined information about Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

141 the mode of action of the chemicals in question. According to the IPCS proposal, Tier 3 risk characterisation steps still only decide whether a refined analysis is required. The stage where risk management measures, rather than further refinements of the analysis, should be envisaged is left open in the IPCS document. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

142 TASK5: PROPOSE AN APPROACH TO CRA (DISSIMILARITY) 15. A tiered framework analysis for combinations of dissimilarly acting chemicals In fulfillment of Task 5 of the project, the purpose of this section is to propose a sciencebased approach for performing cumulative risk assessments of chemicals in food that act through dissimilar modes of action. The science base, and our analysis, of it was presented in the preceding tasks. The proposal is based on three elements: Cumulative risk assessment methods derived from DA, A tiered approach following the broad principles that underpin the IPCS framework analysis (IPCS 2009), and An orientation towards criteria for the grouping of dissimilarly acting chemicals that are based on common adverse outcomes. The application of cumulative risk assessment methods derived from DA to combinations of chemicals with dissimilar modes of action is essentially driven by pragmatism. As shown in chapter 13.4, the predictions for combination effects that can be made by using DA and IA do not differ significantly in most cases encountered in practice, and the factors that drive these prediction differences are well understood. DA also generally yields the more conservative mixture effect predictions, and can be regarded as sufficiently protective in most situations. Examples where IA was more conservative, and produced an accurate prediction, could not be located in the scientific literature (see section ). Furthermore, risk assessment methods based on IA are currently not in use and are yet to be developed (see section ). This, together with the fact that the data requirements for applying IA are hard to meet in practice, argues for an extension of DA based cumulative risk assessment methods also to mixtures composed of chemicals with dissimilar modes of action. Tiered approaches to cumulative risk assessment can avoid unnecessary expenditure of resources by offering the possibility of discontinuing the analysis on the basis of crude and simple assumptions about exposures and hazards when margins of safety (or exposure) are judged to be sufficient. In this way, lengthy, but largely unproductive efforts of refining the analysis can be avoided. It is clear that mixture toxicology cannot aggregate disparate effects. It is impossible to define the joint effect of e.g. a liver toxicant and a pulmonary toxicant, if both these toxicities do not find expression in a common adverse outcome, even though the adverse outcome may arise through very different mechanisms. For this reason, cumulative risk assessments for dissimilarly acting combinations of chemicals should be based on common adverse outcomes as the organising principle of grouping for the purpose of defining common assessment groups. At lower tiers of the analysis, chemicals with ADIs derived from disparate effect endpoints may be considered together, as suggested by IPCS (2009). But as the tiers become more refined, the mixture components should be grouped according to common Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

143 adverse outcomes, as much as possible. This may lead to the exclusion of some substances as the analysis moves to the next higher tier A unified approach to cumulative risk assessment for similarly and dissimilarly acting chemicals Our proposed application of cumulative risk assessment methods derived from DA also to combinations of dissimilarly acting chemicals opens up the possibility of adopting a unified approach to analysing combinations of chemicals, irrespective of their mode of action, especially at lower tiers of the analysis. This can successfully overcome difficulties that may prevent continuation of the analysis in situations where uncertainties about the modes of action of chemicals block further efforts Principal assumptions, simplifications and requirements In common with the cumulative risk assessment approaches applied in regulatory practice (EPA 2000), and the IPCS framework analysis (IPCS 2009), our proposal is based on a number of assumptions which we make explicit, as follows: 1. The possibility of synergisms or antagonisms is disregarded. This assumption is the direct consequence of the fact that the degree of synergism or antagonism cannot be predicted quantitatively on the basis of the toxicity of the mixture components. All mixture effect prediction methods and accordingly, all cumulative risk assessment methods, assume additivity. Considering that the likelihood of synergisms is relatively small (Boobis et al. 2011;EC 2009), the disregard for toxic interactions may be regarded as sufficiently protective. 2. Simultaneous exposure to multiple chemicals is assumed. In numerous settings encountered by humans there is simultaneous exposure to multiple chemicals. For example, there is consumption of single food items that contain multiple chemicals and, even when food items are consumed sequentially, the subsequent exposure of body tissues to the chemicals contained within the items may be simultaneous. Strictly sequential exposures are also a reality, but the risk assessment methods available for cumulative risk assessment are not applicable to sequential exposure to multiple chemicals. In theory and concept, methods for sequential exposures have yet to be developed. 3. Exposures from non-food sources are not taken into account. It is estimated that exposure via food accounts for the majority of chemical exposures currently experienced by humans. However, exposures from non-food sources, including air, consumer items, personal care products and pharmaceuticals also play a role. In view of EFSA s mission, the analysis can be restricted to food items as the source of Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

144 exposure, but may be extended to other exposure routes and sources, if deemed necessary. 4. Potency estimates for mixture components are considered together at low tier analyses, although they may be derived from different endpoints. Application of DA (and IA) requires the use of potency estimates for the same adverse outcome as input values. However, such input values are often not available because chemical safety testing is geared towards identifying critical toxic effects. In practice, this means that toxicity information for chemicals that occur together in mixtures often derives from disparate endpoints. To enable assessments of cumulative risks, the demand for potency estimates for the same endpoints is therefore relaxed, especially for analyses at lower tiers. This simplification is in line with the principles of the framework analysis suggested by (IPCS 2009). 5. It is assumed that the potency estimates entered into cumulative risk assessment methods (e.g. ADIs, POD) describe doses associated with the same effect magnitude. As discussed in section 13.2, the equations for DA are based on single chemical effect doses for identical effect magnitudes. When applied to the PODs that enter the mathematical expressions used in many cumulative risk assessment methods such as HI or PODI, this means that all PODs should describe effect doses for the same effect levels. In practice however, this demand cannot always be met, except in the case of benchmark doses which are defined in relation to the same effect levels. The effects associated with NOAELs that form the basis for ADIs by combination with UFs are normally not known. To make cumulative risk assessment methods workable despite these knowledge gaps, ADIs and PODs are taken as if they described effect doses for the same effect magnitude. 6. Potency estimates can be derived from different tests, performed under different conditions. In the interest of consistency, the evaluation of experimental mixture effects by using the concepts of DA or IA should utilise effect data for all the mixture components that were gathered under the same experimental conditions, with the same animal strains. If this condition is not fulfilled, a bias may be introduced into the analysis, leading to erroneous determinations of mixture effects in terms of additivity, synergy or antagonism. Cumulative risk assessment however has to rely on data that were produced in the context of single chemical testing, under widely varying experimental conditions, even when the same strains were used, so that the demand of consistency of data cannot be realised in practice. To allow continuation of cumulative risk assessments, this demand is therefore relaxed. 7. Data on exposures and potency must be recorded by using the same dose metric. The same dose metric (e.g. intake and potency as mg/kg d) must be used to allow utilization of the formula for HI or PODI. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

145 15.3. Rules for step-wise refinements as the analysis moves to the next higher tier Cumulative risk assessment is based on at least three components: (1) selection of a defined number of chemicals, on the basis of criteria derived from the exposure setting under investigation or other grouping criteria, (2) data about exposures of the chemicals in the mixture to be assessed, and (3) estimates about their potency. Should application of a cumulative risk assessment approach indicate risks that are deemed to be unacceptable, the analysis is refined in a step-wise fashion. As a general rule, it is proposed that only one of the above three components is refined at a time, as the analysis moves to the next higher tier. The most practical component should be refined first, for example the component that will require the fewest pragmatic assumptions Elements of cumulative risk assessment in a tiered framework analysis In common with single chemical risk assessment, there are three elements of cumulative risk assessment: Exposure assessment, hazard assessment and risk characterization. Exposure assessment compiles exposure information about all the chemicals considered together. Depending on the refinement of the analysis, this can be based on quite crude assessments at lower tiers or on probabilistic data at higher tiers. Hazard assessment is based on potency estimates of the chemicals in question. According to the level of detail possible, this will rely on ADIs, or on PODs for specific endpoints, depending on the aim of the analysis and the grouping criteria for the common assessment group. The risk characterization step ends with a decision about continuation of the analysis, with further refinement. The decision criterion is whether the HI exceeds 1, or whether the margin between hazard and exposure is judged to be sufficient. At some stage of the analysis the risk characterization step should yield a decision criterion for risk reduction measures (see below) Framework analysis: initial considerations As suggested in the IPCS framework (IPCS 2009), the analysis begins with considerations of the nature of the exposures in question and assesses whether the key mixture components that make up combined exposures are actually known. This first step may already expose knowledge gaps that preclude continuation of the analysis.risk assessment approaches that start from known exposures may have limitations for use in the setting of regulatory limits, which may need to be set without knowledge of exposures. In particular, although the regulatory limits for single agents may be set by reversing the risk assessment procedure, the same is not true for the regulation of mixtures. Approaches to this issue have been discussed elsewhere (EFSA 2009c). Continuation of the framework analysis should be driven by the likelihood with which coexposures to chemicals can occur. Accordingly, the analysis should proceed by considering which chemicals may occur together in the setting under investigation, initially without Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

146 consideration of their toxicological profile. It is conceivable that the analysis reveals that the complete array of chemicals of concern present in e.g. a food item is not known. In such cases, the analysis may proceed on the basis of known chemicals, but the uncertainty introduced by knowledge gaps may have an impact on the magnitude of margins of exposure that are deemed to be acceptable Tier 0 At the lowest tier, Tier 0, all chemicals that occur together in the exposure setting under investigation are considered, irrespective of the effects they elicit. Commensurate with the (low) quality of data that find entry in the analysis at this stage, it is suggested that a hazard index (HI) should be constructed. Crude and semi-quantitative exposure estimates may be used for the development of a HI. Similarly, the potency estimates entered into the calculation of a HI may also be quite basic. For example, in line with the IPCS (2009) proposal, it may be assumed that all chemicals in the mixture are as potent as the most potent constituent, if more appropriate data are lacking. Thresholds of Toxicological Concern (TTC) may also be used at this stage, with the aim of bridging data gaps. If available, ADIs may be entered into the calculation of a HI, but there is no need to enter ADIs consistently for all chemicals considered in Tier 0. The ADI values may be derived from a variety of different endpoints and species, and may include different UFs. During the risk characterisation step, the margin between estimated exposure and hazard is considered as the decision basis for determining whether a more refined analysis is required. If the HI exceeds 1, the analysis should proceed to Tier Tier 1 In agreement with the proposed IPCS framework analysis, all chemicals relevant to the exposure scenario under investigation are considered in Tier 1, irrespective of the effects they produce, and without consideration of the modes of action involved. No chemical that is a component part of the exposure scenario under investigation is excluded from the analysis at this stage. The assumptions on exposures may still be deterministic and may reflect worst case assumptions, but they should rely on measured values as much as possible. For hazard assessments, estimates of potency for each chemical are incorporated, such as ADIs or benchmark doses. Simplifying assumptions, such as TTC or potencies similar to the most toxic known substance present in the mixture should be abandoned. The potency estimates may be for a variety of different endpoints, and can be derived from studies with a variety of different test species. Alternatively, and if possible, hazard assessments may be based on PODs for all the chemicals considered, with application of the PODI in the risk characterisation step (for further details see below). Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

147 The risk characterisation step determines whether further refinements of analysis should be conducted. This should be pursued if the HI (or PODI) is larger than 1, or if the margin of safety is judged to be inappropriate Tier 2 In the preceding tiers of the analysis, no chemical present in the mixture under consideration was excluded from the analysis. In Tier 2, this restriction may be relaxed, by considering for the first time the effect profile of chemicals, with the intention of creating assessment groups of chemicals. Since this step of the analysis is by no means trivial, and may be hampered severely by data gaps, we propose to deal with this topic initially by excluding those chemicals that are known not to produce a chosen common adverse outcome (see case 3, section15.8.3). Consequently, the assessment group may include chemicals where there is a degree of uncertainty as to whether they can contribute to a common effect. This is to avoid a situation in which the analysis lacks conservatism by defining too narrow common assessment groups on the basis of positive effect criteria. As before, the exposure assessment should rely on measured data. The hazard assessment may utilise ADIs (or, alternatively PODs) that were derived for common endpoints, irrespective of any consideration of mode of action. If the risk estimates exceed acceptable levels, the analysis may be refined and proceed to Tier Tier 3 At this stage, the analysis may adopt more restrictive criteria about common adverse outcomes, and may define groupings of chemicals for assessment on the basis of phenomenological effect criteria. As is commensurate with an analysis for dissimilarly acting chemicals, these groupings will still be made irrespective of any modes of action or mechanisms that may underlie the induction of the effects of interest. The exposure assessment element may utilize probabilistic data, if available. Tier 3 assessments for hazard may incorporate increasingly refined potency estimates for the specific endpoints that form the basis for defining groups of chemicals for assessment. At this stage, PODs may be used, with the aim of constructing a PODI. In the interest of consistency of analysis, the PODs should derive from the same animal species. In this way, the analysis approaches a level of detail similar to that exercised during the evaluation of experimental mixture studies, where all single chemicals and the mixture were studied under similar conditions. Nevertheless, the PODs may still reflect some differences in terms of data quality and experimental standards. In Tier 3, construction of a PODI, rather than a HI, may be regarded as more appropriate, for the following reasons: It is assumed that the PODs that form the basis of the analysis are of similar quality. In this case, it is appropriate that the aggregation for mixture effects should be conducted at the level of experimental data, by calculating a PODI. This practice achieves a high level of Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

148 consistency that comes nearer to the application of DA as a mixture assessment concept in experimental mixture studies. It realises a high level of transparency, by avoiding the introduction of too many assumptions (e.g. use of different uncertainty factors in inconsistent ways). In the risk characterisation stage of the analysis, an uncertainty factor can be used for the entire group of chemicals, or, alternatively, a margin of exposure can be determined When should the risk characterisation step result in risk management measures? As described in section 14.3, the IPCS framework analysis does not define the stage where the risk characterisation step should result in proposing risk management measures, rather than further refinements of the analysis. In our proposed framework analysis for combinations of dissimilarly acting chemicals, the only option for introducing further refinements of the grouping of chemicals beyond Tier 3 is in considering the modes of action of the mixture components under consideration. However, such refinements are called for when the analysis is for combinations of similarly acting chemicals, and would not be appropriate for dissimilarly acting chemicals. For this reason, it is proposed to discontinue the analysis at Tier 3, with a recommendation for risk management measures if the margin of exposure or the size of the PODI is deemed unacceptable. By default, that should be the case with margins of exposure smaller than 100, or with a PODI larger than 1 (if constructed by using the default UF of 100). However, in view of the data gaps that might preclude proceeding to Tier 3, it is a point of serious consideration whether risk management measures should also be envisaged at the end of Tier 2. This may be justified in cases where further refinements of analysis are not possible in the foreseeable future, because existing data gaps are unlikely to be filled Criteria for the grouping of chemicals in CRA at higher tiers Key features of our proposed unified approach are that all chemicals relevant to the exposure scenario under investigation are considered together at lower tiers of the analysis, irrespective of their presumed mode of action, and that the grouping of chemicals according to their ability to affect common adverse endpoints, with exclusion of substances judged to be irrelevant, only takes place at higher tiers. This raises the question as to how such groupings should be conducted, and whether the data necessary to achieve well-founded groupings are available. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

149 Here, we approach this issue by first considering debates and proposals about groupings for similarly acting mixtures in the context of tiered approaches to CRA. Several case studies have been published where the applicability of the IPCS framework analysis was investigated. These studies have considered chemicals with quite similar structural features, e.g. polybrominated biphenyl ethers or carbamates (IPCS 2009). EFSA has published a document about grouping criteria for triazoles (EFSA 2009c). It has been noted that cumulative risk assessment based on quite narrow criteria of chemical similarity may underestimate risks by excluding other substances from the analysis that might nevertheless contribute to cumulative effects, even by differing modes of action (EC 2009;NRC 2008). There is experimental evidence that the concept of DA produces reliable estimations of combination effects with mixtures composed of chemicals with diverse modes of action (Christiansen et al. 2009;Crofton et al. 2005;Rider et al. 2010). For this reason, broader grouping criteria, based on similarity of effect in a phenomenological sense, have been proposed for similarly acting combinations, e.g. for the assessment of antiandrogenic substances (NRC 2008). Such grouping criteria can also be utilised for dissimilarly acting chemicals, e.g. from Tier 2 onwards. However, the development of phenomenological grouping criteria based on common adverse outcomes has to be done with due consideration of physiological processes. As was discussed in the NRC report, phenomenological grouping criteria should be based on a sound understanding of the physiological processes that lead to a common adverse outcome. In the case of combinations of antiandrogenic substances, the decisive factor is disruption of androgen action in fetal life, which can occur by different mechanisms and modes of action. Blocking of the androgen receptor, suppression of androgen synthesis and disruption of steroid-metabolising enzyme systems are processes of relevance, which lend themselves as criteria for the grouping of chemicals that should be considered together for an assessment of cumulative anti-androgenic effects. Accordingly, substances shown to antagonise the androgen receptor, suppress androgen synthesis, interfere with steroid-converting enzymes, and suppress androgen receptor expression should be considered as candidates for a common assessment group for chemicals affecting male sexual differentiation. The identification of candidate chemicals can be based on both in vitro and/or in vivo data, and suitable assays are available. While these are relatively clear criteria for the identification of candidate chemicals that should be subjected to CRA, the implementation of CRA in practice has to rely on the provision of further data about exposures and potency estimates. Both these elements currently present bottlenecks, as discussed by (Kortenkamp and Faust 2010). Exposure estimates for many chemicals identified as in vitro androgen receptor antagonists are rudimentary, or not available at all. The input data for the hazard assessment step of CRA should be derived from in vivo studies where the ability of candidate chemicals to affect the above anti-androgenic endpoints has been assessed. Such data are currently available only for a very limited set of chemicals. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

150 A comparable level of understanding is currently achieved perhaps with thyroid-disrupting chemicals, but the bottlenecks that prevent meaningful application of CRA are similar. Thyroid hormone levels are determined through a complex interplay between several factors, including uptake of iodine from dietary sources which is required for the synthesis of the hormone, transport of iodine to the thyroid gland, synthesis of the precursor of active thyroid hormone, transport of the precursor to tissues and deiodinisation of the precursor to the active thyroid hormone. Suppression of thyroid hormone levels by disruption of any of these processes is regarded as the decisive factor. There is experimental evidence that chemicals capable of interfering with these processes through multiple mechanisms elicit combined effects well predicted by DA (Crofton et al. 2005;Crofton 2008). Accordingly, elements of these physiological steps can serve as criteria for the inclusion of chemicals in a common assessment group. Thus, substances shown to Interfere with iodine uptake, Suppress synthesis of the precursor hormone, Block deiodinisation of the precursor, or Antagonize hormone action at the receptor level are candidates for inclusion in a common assessment group, and suitable in vitro and in vivo assays for the identification of such chemicals are available. However, as with antiandrogenic chemicals, the implementation of CRA is hampered by a lack of data for the critical input values, i.e. exposure estimates and potency estimates for the above thyroidrelevant endpoints. The physiological processes that lead to other common effects are comparatively poorly understood. As a way of dealing with these knowledge gaps in the context of cumulative risk assessment, discussions are underway to group chemicals together according to the effects they exert on target organs, or on systems, such as the reproductive system. Such grouping strategies have to be applied judiciously and with great care. For example, all carcinogens, irrespective of the tissues in which they induce tumours, could be considered together. While such a grouping is sensible at lower tiers of the analysis, a better understanding of the physiological processes involved will have to drive the grouping at higher tiers. Similarly, all substances known to produce reproductive and developmental toxicity could be considered together at lower tiers. However, reproductive toxicity encompasses a broad range of different effects, ranging from reductions in fertility to malformations of organs. At higher tiers, such diverse effects will have to be differentiated. It is beyond the scope of this project to define detailed phenomenological grouping criteria. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

151 15.8. Case studies To illustrate various aspects of the proposed approach to cumulative risk assessment we provide three case studies in this section. Each is intended to illustrate the approach and should not be considered as an attempted cumulative risk assessment. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

152 Case 1: pesticides and contaminants in lettuce In this case, we explore the application of CRA to the analysis of specific food items. In a monitoring program in the year 2007 the German Federal Agency for Risk Assessment (BfR) investigated pesticide residues and environmental contaminants such as heavy metals and nitrate in various food items including salad. Butterhead lettuce was analysed for potential residues of 510 different pesticides. Most of the samples were from Germany. However, some samples came from other European countries. A total of 63 samples were taken. There are no published data on specific substances and their levels that can be attributed to a single sample. Arithmetic mean levels (MW) of pesticides and contaminants in lettuce are documented and it is indicated, in how many samples (absolute/ percentage of all) the levels were below, at, or above the analytical detection limit. For an assessment of potential health effects due to the residues and contaminants, we compared the levels in lettuce with the substance specific reference values (ADI for pesticides and TDI for contaminants). ADI-values were mostly reported by EFSA (EFSA 2011b). Where necessary, those were supplemented by earlier listings from BfR (BfR 2007). For lead the former provisionally tolerable weekly intake (PTWI) of 25 µg/kg x w was used (3.5 µg/kg x d), not accounting for the uncertainties as documented by EFSA (EFSA 2010b). Similarly, for cadmium the tolerable weekly intake (TWI) of 2.5 µg/kg x d was used and transformed into a daily dose of 0.4 µg/kg x d (EFSA 2009b). For copper there is an acceptable range of oral intake (AROI) with the upper limit of more than 2-3 mg per adult and day (WHO 1998), which was accounted for by a reference value of 5 mg/d x 60 kg = 0.08 mg/kg x d. For nitrate, an ADI was available (EFSA 2008a) Table 14 provides a list of the residues and contaminants in lettuce, the number of samples, where those substances were found above detection limit, the content as arithmetic means (MW) and the respective reference values (ADI). We used these survey results to analyse the data for the impact of mixture effects as assessed according to the tiered approach proposed in this report (section 15). Three assessment scenarios (A, B, C) were assumed alternatively: Assessment scenario Scenario A Lettuce only Scenario B Lettuce plus food background Scenario C Exposure characterization Isolated analysis of lettuce consumption; no other sources for identified residues and contaminants in food considered; reference point: ADI Analysis of lettuce consumption assuming that also other sources for all substances within the mixture contribute to intake; EFSAs data on pesticide uptake via different food commodities (expressed as % of ADI and calculated by using mean pesticide concentrations and mean intake values for each commodity) were used as background values; reference point: flexible (background dependent) fraction of ADI Analysis of lettuce consumption assuming that only a fixed minor fraction Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

153 Lettuce allocated fraction of the ADIs for all mixture components is allocated to this food commodity as permissible maximum intake; reference point: fixed fraction of ADI (quote of ADI) For all scenarios we assumed that the average consumption of salad in Germany is 23 grams per day (EFSA 2008a) Assessment Scenario A In this scenario the analysis focuses exclusively on lettuce as the sole source of intake (butterhead lettuce; lettuce only ). Other food items are not included in the analysis. Tier 1 of the proposed framework analysis is applied. The results are shown in Table pesticides or nitrate or metal contaminants were found with a HQ of 0.1% (shaded in Table 15), with the assumption of a mean consumption of lettuce 23 grams/person/day and a body weight of 60 kg). The maximum number of pesticides identified in one and the same sample was 13, but the identity of the substances was not reported in the original report (BVL 2008a). Under the (conservative) assumption that all of the 53 different substances are present in a single sample, and using the arithmetic mean of the concentrations [mg/kg] found from all 63 samples for each of the substances, a HI of 28% can be calculated. The main contribution is from nitrate, with a HQ of 22.8%. About 2% of the HI may be attributed to pesticide residues. Some of the substances were not analysed in all of the 63 samples. The data for pirimicarb were not included as only 10 samples were analysed for this pesticide. As is appropriate for a low tier analysis, there was no grouping of the pesticides and contaminants in terms of common adverse outcomes. It may be assumed that the number of contaminants and pesticides present in one and the same lettuce sample is well below the maximum number of 13 found in one sample, and this will reduce the hazard index accordingly. However, certain substances may have been present in single samples with concentrations well above the arithmetic mean, on which the above estimate was based. In any case, the analysis demonstrates that most of the HI is due to a single substance exposure (nitrate). All other analysed pesticides and contaminants contribute only to a minor degree to the HI. Therefore, in this case no higher tier analysis is necessary. However, it may be argued that scenario A is insufficiently conservative, and that the risk characterization step leads to a false negative decision as it does not cover pesticides exposure due to other food items Assessment Scenario B To deal with the shortcomings of the previous scenario A, it was now assumed that the pesticides and contaminants found in lettuce are also taken up via other food items ( lettuce plus food background ). To account for uptake via other food sources (for details see next paragraph), the ADIs (TDIs) for each substance were corrected accordingly. The HQ were calculated by using lower ADIs, corrected for the intake from other food items (ADI x (1- calculated exposure via other food items in %). As the amount of each residue and contaminant in other food items differed, substance specific allocations were calculated. For Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

154 example, the ADI for dimethoate was lowered by 14.8% because total uptake of dimethoate via different food commodities was 14.8% of the ADI as calculated by EFSA (EFSA 2011b) and accordingly, the allocated ADI therefore was 85.2% of the original ADI. Similarly, 72% auf the TDI for cadmium was attributed to other sources with a remaining allocated TDI of 28% of the original TDI. Table 16 shows the results of this calculation. The data on the fraction of ADI assigned to exposure to the respective pesticides via other food commodities was taken from The 2009 European Union Report on Pesticide Residues in Food (EFSA 2011b). EFSA has performed an analysis on the total uptake of pesticide residues via food based on monitoring data from the Member States and consumption data from the PRIMo database. These data were transformed into a figure representing a fraction of the respective ADI. We used the data for GEMS/Food Cluster E diet for further analysis. Note that this fractional ADI includes exposure to all food items, also including lettuce. Therefore, there is some double counting: the fraction of the ADI available for salad exposure depends on the fraction for all other food items, which again includes salad. Within this analysis it was not possible to adjust for this imprecision. However, the bias introduced by this omission is negligible. For lead, an average uptake via food of µg/kg x w is reported (EFSA 2010a) corresponding to up to 35% of the PTWI. Cadmium exposure via food may be very high and may already reach or exceed the TWI (EFSA 2009b). A fraction of 72% as a minimum assumption was assigned to other food items but salad. For copper, 1-2 mg/adult intake is via drinking water and food (WHO 1998), which corresponds to about 40% of the AROI reported above. For nitrate, a high percentage of the ADI is taken up via drinking water and cured meat (close to 20%). More than 66% of the ADI has to be assigned to fruits and vegetables, in this case after exclusion of salad (calculated from (EFSA 2008a)). From this analysis we assigned a fraction of 86.4% of the ADI to other food items. Calculation of the HI with the modified ADIs or TDIs gave a value of 180%, clearly exceeding 100%. The majority of the HI can be attributed to nitrate (167%), 10% to cadmium and the pesticide residues only make up 2% of the HI. For scenario B pesticide uptakes via other commodities were taken from the 2009 European Union report on pesticide residues in food (EFSA 2011b). The individual allocations of the ADI of each pesticide with respect to GEMS/Food cluster diet E was regarded in this calculation. Again, nitrate was the most relevant single substance in this scenario B as it was already in scenario A. Nitrate uptake via other fruits and vegetables besides lettuce and the uptake via drinking water and cured meat accounts for 86.4% of ADI and only 13.6% of the ADI remain for lettuce in scenario B. Based on the mean concentration of nitrate in lettuce and a mean consumption of 23 g/d the respective hazard quotient for nitrate (167.4%) clearly exceeded 100%. Because the high hazard index is driven by one single substance, no higher tier analysis was conducted. Increases of residues or contaminants in other food will simultaneously decrease the available fraction for lettuce and would therefore change the hazard index. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

155 Assessment Scenario C This analysis is based on lettuce consumption assuming that only a fixed minor fraction of the ADIs for all mixture components is allocated to this food commodity as permissible maximum intake. This fixed fraction of ADI is used as reference point. For the calculations in Scenario C we assumed that 5% of the ADI of each individual pesticide or contaminant is allocated for the uptake via lettuce. This percentage is based on the average lettuce consumption (23 g) in relation to consumption of vegetables and fruits, which may contain residues of pesticides (about 400g/d). Table 17 provides the results of this assessment including the hazard ratios. The hazard index with 5% of the respective ADIs as reference points clearly exceeds 100% and reaches a value of 559% or However, as in scenarios A and B, only one substance (nitrate) contributes to most of this load. After excluding nitrate and the metal contaminants a HI of only 40% can be calculated, based on the pesticides alone. Because of the disproportionate influence of nitrate the result would not change significantly if the analysis proceeded to a higher tier assessment. The lettuce scenario indicates that there is an issue with one single, or perhaps two substances, nitrate and cadmium. The procedure adopted in this scenario has been used for drinking water, where the WHO proposes 10% of the ADI for the respective substances as a reference point Higher Tier Assessments In this case we have presented variations of a Tier 1 assessment for lettuce as food item. Despite the fact that there are only small incentives to proceed to higher tier analyses, we would like to sketch out some elements that might be useful for Tier 2 analyses. In Table 18 we illustrate a first grouping step. To this end, we selected randomly a small subset of 5 pesticides, with the aim of exploring methodological aspects of the approach. We grouped the pesticides and contaminants according to common adverse outcomes by considering potential target organs of the substances within the mixture. As an example for the proposed procedure we analysed the dossiers for the 5 chosen pesticides (PRAPeR, DAR, EFSA substance specific scientific opinions) with the aim of deriving further information about common adverse outcomes and their points of departure. The points of departure and the corresponding endpoints that formed the basis for the estimation of an ADI were listed, and the ADI then taken for the calculation of HI (see Table 18). For toxicities that were not critical for the estimation of ADIs, reference doses (RfD) were estimated as follows: Bifenthrin is a clear neurotoxicant (tremors in dogs and rats at elevated doses) and the NOAEL for this effect was used to establish the ADI. Bifenthrin also shows carcinogenic effects in animals at much higher doses by a nongenotoxic mode of action and is classified as a low potency suspected carcinogen (classified Canc. Cat. 2, CLP). The carcinogenicity is observed at doses more than 10 times higher than those associated with neurotoxicity. Accordingly, a RfD for carcinogenicity was assumed to be 10 times higher than the ADI, derived by multiplying the ADI with a factor of 10. These factors are listed in Table 18 for Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

156 other toxicities. If, for a specific organ, no toxicity was detected, the multiplier was set to zero. As is apparent from Table 18, there are 3 neurotoxicants and 3 agents leading to sensitisation among the selected 5 pesticides. At higher tiers of the analysis, these target organ effects would have to be considered in more detail. For nephrotoxicity, glomerular and tubular nephrotoxic substances should be distinguished if this distinction is clearly supported by the data Conclusions The main findings of this case can be summarized as follows: The number of substances occurring together in one food item and therefore contributing to a mixture effect is often limited (in this case the maximum number of pesticides in one lettuce sample was 13).However, some substances might have been missed, due to inappropriate analytical detection limits. On the basis of more sensitive detection limits, the number of substances to be considered in CRA would increase accordingly, but the relevance of this in terms of CRA is unclear, because only relatively small HQ would result, with limited impact on the HI. A few single substances (such as nitrate or cadmium) with widespread environmental occurrence may constitute the majority of a HI and are of health concern on their own, independent of any combination effects. Reducing exposures to these substances will have a strong impact on the overall risk estimate. A higher tier approach in mixture risk assessment is not warranted, if one single substance contributes disproportionately to the HI. We demonstrated the feasibility in principle of a tiered approach by considering a small subset of substances of this case. In fact, PRAPeR- and DAR-documents provide useful quantitative information to be used in higher tiers mixture risk assessment. However, we did not pursue such an analysis here, because the HI could be explained in terms of a few single chemicals. Nevertheless, some detailed guidance should be established in the future on how to derive RfD for specific common adverse outcomes different from those that form the basis for estimating ADIs. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

157 Table 14: Residues and contaminants in butterhead lettuce (BVL 2008a;BVL 2008b) n > detect. Limit % of samples above detect limit Content Arithmetic mean MW [mg/kg] EFSA 2011 ADI *) [mg/kg bw x d] Acetamiprid Azoxystrobin Bifenthrin Boscalid (Nicobifen) Methyl bromide Folpet Carbendazim Chlorpyrifos-methyl Chlorpyrifos Clothianidin Cypermethrin [ISO] Cyprodinil Deltamethrin Diethofencarb Dimethoate Dimethomorph Thiram**) Endosulfan (-A, -B, sulfate) Fenhexamide Fludioxonil Indoxacarb Imidacloprid Iprodione Iprovalicarb Cyhalothrin, l Metalaxyl-M Linuron Methomyl Myclobutanil Oxadixyl Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

158 Pendimethalin Pirimicarb Procymidone Propamocarb Propyzamide Pymetrozine Pyraclostrobin Pyrimethanil Tebufenpyrad Terbuthylazine Thiamethoxam Tolylfluanid Tolclofos-Methyl Trifluralin Vinclozolin Pencycuron Spinosad Spinosyn A Spinosyn D Lead Cadmium Copper***) Nitrate *) for some substances EFSA-ADIs were not available. In those cases other ADI or TDI reported in the literature were used. See text for further references. **) Only Dithiocarbamate listed in report - Thiram used as Dithiocarbamate with lowest ADI from those permitted for use in Germany ***) further contaminants (Thallium, Zinc) were excluded because of low concentrations and no adequate reference values (TDI) available Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

159 Table 15: Lettuce only assessment scenario (tier 0). Hazard Index with no influence of exposure to other food items MW [mg/kg] ADI (mg/kg bw) Hazard quotient % (MWx23g/d) / (ADI*60kg) Acetamiprid Azoxystrobin Bifenthrin Boscalid (Nicobifen) Methyl bromide Folpet Carbendazim Chlorpyrifos-methyl Chlorpyrifos Clothianidin Cypermethrin [ISO] Cyprodinil Deltamethrin Diethofencarb Dimethoate Dimethomorph Thiram Endosulfan (-A, -B, sulfate) Fenhexamide Fludioxonil Indoxacarb Imidacloprid Iprodione Iprovalicarb Cyhalothrin, l Metalaxyl-M Linuron Methomyl Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

160 Myclobutanil Oxadixyl Pendimethalin Pirimicarb Procymidone Propamocarb Propyzamide Pymetrozine Pyraclostrobin Pyrimethanil Tebufenpyrad Terbuthylazine Thiamethoxam Tolylfluanid Tolclofos-Methyl Trifluralin Vinclozolin Pencycuron Spinosad Spinosyn A Spinosyn D Lead Cadmium Copper Nitrate HI 27.96% Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

161 Table 16: Assessment Scenario B: Lettuce plus food background. Calculation of Hazard Index (tier 0) MW [mg/kg] ADI (mg/kg bw) Exposure via all food commodities analysed (%ADI, WHO Food Cluster E)* HQ % (MWx23g/d ) / (ADI*60kg) HQ % based on adjusted ADI Acetamiprid Azoxystrobin Bifenthrin Boscalid (Nicobifen) Methyl bromide Folpet Carbendazim Chlorpyrifos-methyl Chlorpyrifos Clothianidin Cypermethrin [ISO] Cyprodinil Deltamethrin Diethofencarb Dimethoate Dimethomorph Thiram Endosulfan (-A, -B, sulfate) Fenhexamide Fludioxonil Indoxacarb Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

162 Imidacloprid Iprodione Iprovalicarb Cyhalothrin, l Metalaxyl-M Linuron Methomyl Myclobutanil Oxadixyl Pendimethalin Pirimicarb Procymidone Propamocarb Propyzamide Pymetrozine Pyraclostrobin Pyrimethanil Tebufenpyrad Terbuthylazine Thiamethoxam Tolylfluanid Tolclofos-Methyl Trifluralin Vinclozolin Pencycuron Spinosad Spinosyn A Spinosyn D Lead Cadmium Copper Nitrate HI 27.96% % * Data from (EFSA 2011b) Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

163 Table 17: Assessment Scenario C (fixed fraction ). Calculation of hazard index (tier 0) MW [mg/kg] ADI (mg/kg bw) HQ % (MWx23g/d) / (0,05*ADI*60kg) Acetamiprid Azoxystrobin Bifenthrin Boscalid (Nicobifen) Methyl bromide Folpet Carbendazim Chlorpyrifos-methyl Chlorpyrifos Clothianidin Cypermethrin [ISO] Cyprodinil Deltamethrin Diethofencarb Dimethoate Dimethomorph Thiram Endosulfan (-A, -B, sulfate) Fenhexamide Fludioxonil Indoxacarb Imidacloprid Iprodione Iprovalicarb Cyhalothrin, l Metalaxyl-M Linuron Methomyl Myclobutanil Oxadixyl Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

164 Pendimethalin Pirimicarb Procymidone Propamocarb Propyzamide Pymetrozine Pyraclostrobin Pyrimethanil Tebufenpyrad Terbuthylazine Thiamethoxam Tolylfluanid Tolclofos-Methyl Trifluralin Vinclozolin Pencycuron Spinosad Spinosyn A Spinosyn D Lead Cadmium Copper Nitrate HI % Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

165 Table 18: Higher tier mixture risk assessment example: subset of pesticide data Substance Name CAS No Respiratory Sensitisation Neurotox Hepatotox Nephrotox Bones Hematotox Developmental Reproductive Cancer ADI [mg/kg x d] Bifenthrin Fenhexamid Fludioxonil Dimethoate Cypermethrin TIER 2 n Calculate hazard ratio and hazard index only for those substances, for which the value in the respective column is > 0. Use original ADIs as reference points (e.g. for kidney effects three substances are regarded as relevant). Highest hazard index characterises the mixture effect TIER 3 Calculate hazard quotients and hazard index only for those substances, for which the value in the respective column is > 0 (relevant endpoint; see TIER 2). Use ADI x value given in the respective matrix field above as your reference point (e.g., use 4 x ADI(Bifenthrin) = 4 x = 0.6 for kidney effects of Bifenthrin) of endpoint specific relevant substances Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

166 Case 2: Mycotoxins in food commodities Mycotoxins are toxic secondary metabolites produced from several species of fungi. They are widely distributed food and feed contaminants which pose a serious problem in food and feed safety. Several mycotoxins may occur simultaneously in food and feed which results in a combined intake of mycotoxins by humans and animals. Maximum levels for mycotoxins in food have only been derived for single mycotoxins or mycotoxins belonging to the same group like aflatoxins B1, B2, G1, G2, fumonisins B1 and B2 or T2-toxin and HT-2 toxins (see Until now, risk assessment focuses on single substances or closely related mycotoxins such as T2-toxin and its metabolite HT-2 toxin. Whether the combined intake of numerous mycotoxins from structurally similar (e.g. trichothecenes) or different groups may possibly result in an increased risk to human health shall be investigated in the following case. While the previous case on pesticides and contaminants present in lettuce focused on one single food item, this case will investigate the cumulative exposure to several mycotoxins from different commodities. In the past several scientific cooperation (SCOOP) tasks have focused on the collection of data about the levels of mycotoxins in food and on the assessment of dietary intake of these substances. For this case the SCOOP data on fusarium toxins, patulin and ochratoxin A (EC 2002a;EC 2002b;EC 2003) have been evaluated. Data reported from the United Kingdom (UK) on the dietary mycotoxin concentrations, food consumption and dietary intake have been selected for this case due to the comprehensive measurements and detailed analysis of consumption data for several groups of the population. Food commodities were analysed for the following trichothecene mycotoxins: Group A trichothecenes Group B trichothecenes T-2 toxin Deoxynivalenol HT-2 toxin Nivalenol T2-triol 3-Acetyldeoxynivalenol Diacetoxyscirpenol 15-Acetyldeoxynivalenol Neosolaniol Fusarenon-X Monoacetoxyscirpenol Verrucarol Deoxynivalenol (DON), nivalenol (NIV), T-2 toxin and HT-2 toxin are the most relevant substances from this group. They are included in the further analysis because they are the only members from this group for which reference values exist (see below). Additionally the fusarium toxins zearalenone and fumonisin B1 and B2 as well as ochratoxin A and patulin were considered for cumulative risk assessment. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

167 The tolerable daily intakes (TDIs) as derived by the Scientific Committee on Food (SCF), the Joint FAO/WHO Expert Committee on Food Additives (JECFA) or EFSA are reported in Table 19. In Table 20 the total dietary intake of DON, NIV, T-2 toxin, HT-2 toxin, zearalenone, patulin and OTA for adults and toddlers as calculated by UK in the SCOOP reports are summarized (EC 2002a;EC 2002b;EC 2003). All food commodities monitored for mycotoxins were used for this calculation (for details see SCOOP reports). Total dietary intake was calculated using consumption data of the whole population differentiated by age groups and not only of consumers. For this example the adults (16-64 years) and toddlers ( years) were selected, the UK analysis was much more differentiated. Further, mean values for mycotoxin concentrations in food commodities have been used. The following principles have been applied for the calculation of the mean mycotoxin concentrations in food: All data provided were used. If limit of detection (LOD) and limit of quantification (LOQ) were available, LOD/2 were used for results lower than the LOD. For results between LOD and LOQ, numerical values, if available, were used. If only LOQ was available, or if numerical values between LOD and LOQ were not available, LOQ/6 was used for values below the LOQ. Here, we provide the results of a Tier 1 analysis for mycotoxins. In this analysis, TDIs derived from a variety of endpoints, in a variety of species were used (see Table 19). The total dietary intake values were then compared with the TDIs for each mycotoxin to calculate HQ for the individual mycotoxins. The HQs were then summed up to the hazard indices (HI) for each age group regarding all mycotoxins. Whereas the HQs were all < 1 for the individual mycotoxins, a HI of 0.84 or 1.01 was calculated for females or males, respectively. Total intake data for fumonisin B1 and B2 are missing in this table, because the UK did not report these data to SCOOP. But it was stated in the SCOOP report that fumonisin intake can reach up to 14% of the TDI in adults (regarding the results from the other countries), which would result in HIs exceeding 1. Even without taking account of fumonisin, a HI of 2.5 was calculated for toddlers. Had it been possible to consider fumonisin, the HI would have been even larger, because fumonisin intake may reach 22% of the TDI according to SCOOP. The trichothecenes DON, T-2 toxin and HT-2 toxin contribute strongly to the HI. There is therefore the need to refine the analysis further by proceeding to Tier 2. In Tier 2, it would be necessary to group mycotoxins according to common effects or common target organs. As shown in Table 21, all selected mycotoxins cause reproductive and immunotoxic effects. Additional endpoints affected by these mycotoxins are e.g. growth retardation, haematotoxicity, neuro-, renal- and liver toxicity. These are the most relevant endpoints affected by the mycotoxins but do not include all toxicological effects elicited by these substance. For further information see (EFSA 2006;SCF 2000a;SCF 2000b;SCF 2002;WHO 1995). Regarding the two apical endpoints reproductive toxicity and immunotoxicity it becomes obvious that none of these mycotoxins can per se be excluded from the cumulative Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

168 risk assessment. Calculation of a HI by using ADI with respect to immunotoxic or reproductive toxic effects in Tier 2 would end up with the identical result as obtained in Tier 1. A further refinement of this CRA would be desirable, e.g. by using NOAELs or BMDs for immune- or reproductive toxicity. For a better comparison POD for the same reproductive or immunotoxic effects should be selected. As these values are not easily available from the toxicity reviews without further evaluation of the scientific literature we stopped the elaboration of this example at this point. This example demonstrates that the proposed CRA method can also be applied to situations where cumulative exposure from several commodities takes place. It may be a useful tool to investigate possible risks which might be lost in single substance evaluations. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

169 Table 19: Tolerable daily intakes for mycotoxins (source: (EC 2002a;EC 2002b;EC 2003) TDI (µg/kg bw) Deoxynivalenol a Key Most sensitive endpoint Point of departure 1 µg/kg bw 2-year feeding in mice Nivalenol a 0.7 µg/kg bw Long term dietary in mice T-2 toxin/ht-2 toxin a 0.06 µg/kg bw 3-week feeding with T-2 toxin in swine reduced growth reduced growth and haematotoxic effects haematotoxic effects 0.1 mg/kg bw/day (NOAEL) 0.7 mg/kg bw/day (LOAEL) 0.03 mg/kg bw/day (LOAEL) Zearalenone a 0.2 µg/kg bw c 15-day in pigs hormonal effects 40 µg/kg bw (NOEL) Fumonisin B1+B2 a 2 µg/kg bw chronic toxicity/carcinogenicit y in rats Patulin b 0.4 µg/kg bw d combined reproductive toxicity, long term toxicity/carcinogenicit y in rats Ochratoxin A e µg/kg bw/day e 90-day feeding in female pigs Effects in rats and equine leukoencephalomalacia Not specified in SCOOP report Nephrotoxicity 0.2 mg FB1/kg bw/day (NOAEL) 43 µg/kg bw/day (NOEL) 8 µg/kg bw/day (LOAEL) Extrapolation factor Notes. a: evaluation by the SCF; b: evaluation by JECFA; c: temporary TDI; d: PMTDI: provisional maximum TDI; e: evaluation by EFSA (EFSA 2006); f: calculated on basis of a tolerable weekly intake (TWI) of µg/kg bw/week Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

170 Table 20: Mycotoxin uptake in the population of the United Kingdom (further explanations in the text) and cumulative risk assessment Age Group Mycotoxin Total dietary intake TDI HQ (Intake/TDI) (ng/kg bw/day) (ng/kg bw/day) Average male adult (16-64 years) DON NIV T-2 + HT-2 toxin Zearalenone Patulin OTA HI: Average female adult (16-64 years) DON NIV T-2 + HT-2 toxin Zearalenone Patulin OTA HI: Average toddler ( years) DON NIV T-2 + HT-2 toxin Zearalenone Patulin OTA HI: 2.52 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

171 Table 21: Most relevant apical toxicological effects affected by the selected mycotoxins Reproductive toxicity Growth retardation Immunotoxicity Haematotoxicity DON x x x NIV x x x X T-2 toxin x x x X HT-2 x x x X toxin ZEA x x FB1+2 (x) x x x x PAT x x x x OTA x x x x Neurotoxicity Renal toxicity Liver toxicity Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

172 Case 3: dietary exposure to pesticides Description of dataset For this case we illustrate the stages of a cumulative risk assessment using the exposure and risk data provided for 34 pesticides in JMPR reports from 2009 and 2010 (JMPR 2009;JMPR 2010), available online rep/en/. Rather than providing realistic risk estimates for exposure scenarios involving pesticides, the aim of this case is to explore whether, and to what degree, lack of data might hamper continuation of the analysis. Accordingly, the chemicals included in this case were selected on practical grounds, that data about them were available in either of the two most recent JMPR reports. This does not indicate that chemicals that were not included in this analysis should not be considered, nor that the included chemicals are considered priorities for assessment. The guiding approach used is the hazard index (HI). Annex 3 of each JMPR report contains international estimated daily intakes (IEDI) which are expressed as the percentage of the respective ADI, for the 13 cluster diets in the Global Environment Monitoring System-Food contamination and assessment programme (GEMS/Food, These numbers are equivalent to the hazard quotients (HQ) required for the calculation of a hazard index (HI), see section , but depicted on a percentage scale so that the range from 0 to 1 is expressed as 0 to 100%. The compiled data is presented in Table 22 and has been rescaled so that the HQ and HI are on a unitary scale, not percentage, in which values greater than 1 indicate a greater than acceptable risk for the single chemical (HQ) or mixture (HI) Tier 1: HI analysis (all ADIs) A hazard index (HI) analysis of the full dataset is presented in Table 23. In line with a low tier analysis, all available ADIs, irrespective of the endpoints they are based on, are considered together. In the main body of the table cells highlighted in red indicate when a single substance has a HQ in excess of 1, indicating a greater than acceptable risk for that chemical alone. HQ values exceeded one in two cases (1.1, 1.4), both for chlorpyrifos methyl. The JMPR note that values above 1 (or 100, when expressed at percentages) should not necessarily be interpreted as giving risk to a health concern because of the conservative assumptions used in the assessments (JMPR 2009). However, for CRA it is important to note that if any single chemical has a HQ of >1 then the HI will exceed 1 irrespective of the contribution of any other chemicals. The final row of the table shows the HI, the sum of all individual HQs. This row is coloured according to the extent by which the HI exceeds 1. The table shows that the HI exceeds 1 for all thirteen GEMS diets, and ranged from 1.3 to 5.2. In no case was the HI solely driven by a Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

173 single chemical exceeding a HQ of 1. In low tiers, this analysis would indicate the need for further refinement and precludes the conclusion of acceptable risk. Further approaches to interpreting the HI are shown in the remainder of the case, which use the data subset for GEMS diet E, which includes Austria, Belgium, Croatia, Czech Republic, Denmark, France, Germany, Hungary, Ireland, Luxembourg, Malta, Netherlands, Poland, Slovakia, Slovenia, Switzerland and United Kingdom of Great Britain and Northern Ireland Interpretation of HI and HQ values Table 23 shows the ADI, exposure (JMPR IEDI) and HQ of each chemical in the GEMS diet E. In this case the HI, the sum of all HQs, is 2.9, indicating that an acceptable level of risk cannot be concluded and that refinement of the assessment is required. This data is shown graphically in Figure 19. The graph shows the distribution of exposure (IEDI, panel A) and risk (ADI, panel B) values that produce the HQ distribution (panel C and panel D, log scale). The graphs are ordered along the x-axis according to decreasing HQ values, and thus the cumulative HQ graph (panel E) shows the build up of contributions towards the HI of The HI is indicated by a solid vertical line. The cumulative HQ graph (panel E) can be used to explore the amount of refinement that would be required to enable the HI to attain values below 1 (indicating acceptable risk). The graph provides access to the two extreme approaches to the refinement, in the best case the chemicals that contribute most to the estimate would be refined first, and the dotted line at 1.88 indicates the chemicals (all those with bars below and the first bar to exceed the value of 1.88) that would need to be refined for the HI to drop below 1. In this case, the first seven chemicals would need to be refined and if their individual HQs still contributed significantly after refinement then additional chemicals would need refinement. Conversely in the worst, most inefficient, case, if refinement was directed at those chemicals not making a significant contribution (those on the right of the graphs) then the dotted line at 1 indicates the chemicals (all those chemicals with bars crossing the dotted line) that would need to be refined for the HI to be below 1. In this worst case, 32 chemicals out of 34 would need to be refined, indicating that if the first two chemicals in the graphs, haloxyfop and chlorothalonil, are not refined then the HI is very likely to exceed 1. Consideration of the HQ spectrum can reveal where refinement could have the most effect and is likely to be more efficient than refining all chemicals or taking a random approach. When the spectrum is skewed (as can be seen in panel C) then focused refinement is even more likely to be more efficient than other approaches, since a few chemicals are disproportionately contributing to the risk estimate, and should be targeted by refinement. An understanding of the extent to which refinement might be effective could be used to decide if refinement of this approach is practical (for example how many chemicals must undergo risk refinement) or if an exposure refinement might have more impact. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

174 HI analysis using TTC values (pseudo tier 0 analysis) The preceding HI analysis was possible because all of the included chemicals have been assigned ADIs. If a chemical has not been assigned an ADI it has been suggested that the TTC approach (see section ) could be applied. We have therefore applied the TTC approach to the data set used in this case to explore its impact. We note that the TTC concept is not intended for use when chemicals have been assigned ADIs, as is true here; however we are using the approach here to assess its impact rather than to perform an assessment. We have also used the extreme case when TTC values are used for all of the chemicals in the assessment. SMILES codes for each pesticide were retrieved from the PPDB (pesticide properties database, and ToxTree software (version 2.5.0, Ideaconsult Ltd.; was used to assign chemicals to the appropriate Cramer class based on their SMILES code. All of the chemicals were assigned to Cramer class III, which receives a TTC value of 90 ug/person per day. To perform a HI assessment using TTC values, abbreviated to HI (TTC), HQs are calculated by dividing IEDI values by TTC values rather than ADIs, and are termed here HQ (TTC). This analysis is presented in Table 23 which compares the HI values calculated using ADIs (previous section) or TTC values (this section). Impact of TTC values on the risk estimate The HI for GEMS diet E was 2.9 when calculated using the ADI, use of the TTC values results in a HI (TTC) of The TTC might typically be applied to only a few components in a mixture assessment, and the impact of the use of TTC values should be assessed whenever they are used. This is because risk estimates driven by the use of TTC values are clear candidates for refinement before action is taken. In this case, the mean HQ (TTC) was 1.1, indicating that if a TTC value was applied to a single chemical, with the remainder assessed using their ADIs, then on average, the HI would be expected to exceed one in all cases. HQ (TTC) values actually ranged from to 5.6 and the impact of smaller values would not greatly affect an overall HI, whilst larger values could be fully responsible for a risk estimate exceeding 1. The distribution of these values would only apply generally if other chemicals follow the same risk (ADI) and exposure distribution of the pesticides used in this case. Use of the conservative TTC approach to bridge data gaps in higher tiers may be impractical when the exposure level of a mixture component is close to the TTC and it may be that either refinement of the exposure assessment or the acquisition of toxicological data is preferred over the propagation of TTC values though to higher tiers. The interpretation of HQ (TTC) values is less valuable than interpretation of HQ (ADI) values because the TTC does not represent a measure of risk in the same way that the ADI does. Consequently HI (TTC) and HQ (TTC) are a measure of exposure with an approximation of risk based solely on chemical structure. HQ and HI values derived from ADI or TTC are also not proportional to each other, see Figure 20. Therefore HQ (TTC) Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

175 values should not be strictly used for targeting refinement since it is probably the TTC values themselves that require refinement, i.e., they would ideally be replaced by ADIs once a suitable toxicity data set was available PODI (NOAELs) The use of a point of departure index (PODI) approach is shown in Table 24. Table 24 lists the NOAEL value for each chemical together with supporting information, including uncertainty factor (UF) values. Ideally a PODI approach would use values for the point of departure (POD) for the same endpoint, however in this example we have applied the approach irrespective of the endpoint for illustration purposes. For this case PODI = (Table 24). Interpretation of a PODI requires multiplication by a group uncertainty factor (UF G ), which for this case might be 100, since it was the UF applied to 33 out of 34 of the pesticides. However the assignment of a group uncertainty factor may require expert judgement and further deliberation rather than a crude averaging approach, as used in this case for pragmatic reasons. Using an uncertainty factor of 100, gives a PODI x UF G value of 3.05, which exceeds 1 and would be considered to indicate a greater than acceptable risk. The PODI x UF G value of 3.05 is greater than the value of the HI (2.88) because one pesticide, triazephos, had a single chemical UF of 10 (because human toxicology data was available), consequently the use of an UF of 100 in the PODI approach increases the contribution of triazephos. This effect illustrates the inability to apply chemical specific UFs in the PODI approach, which might be considered a limitation. However some commentators consider that the ability to remove policy- driven UFs is in fact an advantage, allowing subjective uncertainty considerations to be dealt with by the selection and application of a UF G, and separating these calculations from calculations dealing with data-based potency values, PODs (Wilkinson et al. 2000). Interpretation of a PODI value can also be done through determination of a combined margin of exposure (MOE T ). For this case the MOE T is (the reciprocal of the PODI value), well below a value of 100 or more which might be required for an acceptable MOE T in this situation, considering that the UF applied to 33 out of 34 of the pesticides was 100. In low tiers, when data may not be available on specific endpoints there may be little advantage in using a PODI approach. In cases when the chemicals being considered have the same UF, the HI, PODI and MOE T approaches should produce numerically identical results (Wilkinson et al. 2000); and in cases when the UF differs there may need to be an expert judgment as to the appropriate group UF to apply. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

176 Tier 2 (consideration of effects) In the previous tier, all HQs were summed despite the fact that the ADI on which the HQ is based derive from different endpoints and not all of the chemicals cause every possible toxic effect. It seems logical that the next level of refinement would be to break down the HI so as to only include chemicals that affect a common adverse outcome, but irrespective of their mode of action. The implicit assumption made at this stage is that lack of consideration of mode of action will capture those chemicals that exert their effects through a variety of mechanisms. However, realisation of this level of detail is less trivial than it initially seems and has substantial data requirements that may be virtually impossible to satisfy. This is now illustrated by reference to the dataset considered in this case. The data in Table 24 shows that the NOAEL for each chemical was derived from a different endpoint. Unless the NOAEL was assigned for the same endpoint then further refinement will require substantially more data. If the same endpoint was used then the risk would be calculated using the ADI, the most sensitive endpoints, and the risk from any other endpoint, occurring by definition at higher doses, could be disregarded because the risk would always be less than that for the most sensitive endpoint. When different effects lead to NOAELs, the data for all endpoints becomes required as chemicals can only be excluded from further calculations if they are shown not to cause an effect. This is because if, e.g. all 34 pesticides in this case had a different effect responsible for the NOAEL then a number of the chemicals could still share a common toxic effect at a dose only slightly above the dose producing the NOAEL. If this were the case for a sufficient number of chemicals then the cumulative risk of that effect might be unacceptable but would have been undetected if the HI was broken down using groups based on the effect leading to NOAEL. For this case the data that is required would be the highest dose tested without effect for every endpoint of toxicological interest. This is a substantial data requirement, both in the experimental work required and the effort involved in compilation of the data. It was not feasible to perform this data collection for this case, since the data, if available, are present in free text or tables in reports with an often inconsistent structure and published in pdf format. None of these features facilitate data extraction and compilation. This case shows that a database of toxicological endpoints without summarisation to the most sensitive end point is required for meaningful cumulative risk assessments. Failure of the CRA Since compilation of the data was not feasible, the assessment of this case would fail at this stage, with low tiers having failed to conclude acceptable risk (because HI > 1) and with this tier being unable to complete due to data requirements. The data requirement of this stage is sufficiently high that many assessments could fail at this stage. It is a characteristic of our approach that this analysis should not feature in a very early stage of a CRA, since this would predispose to early failure in many cases. Positioning of this large data requirement Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

177 allows a CRA to conclude acceptable risk if that can be achieved in low tiers with conservative assumptions. The CRA will only fail in higher tiers when more data are required to conclude acceptable risk than are available. In order to illustrate how the assessment would continue we have now used a compromise data set based on human health classifications assigned to each of the chemicals in the Pesticide Properties DataBase (PPDB; Breakdown of HI analysis using PPDB health issues The PPDB includes nine human health issues, which are listed in Table 25. The issues are not regulatory categories and may not be suitable for regulatory purposes but they are used here to illustrate a grouping approach that may refine the analysis. For the purpose of this case it is assumed that the PPDB health issues cover all the effects of interest, and that they are assigned reliably. The PPDB states whether each chemical is known to cause an effect, known not to cause the effect or if the data is unclear or unavailable. Consequently, it is possible to exclude chemicals from calculation of a HI if they are known not to cause an effect, and this is the approach to the grouping of chemicals for CRA that drives the analysis at this stage. It would not be conservative to only include the chemical that are known to cause an effect, since in many cases there will many chemicals for which the data are not available or is unclear. Consequently data from all three situations ( known, unclear and no data ) are included, but their relative contribution is calculated so it can be seen how much of the revised HI is due to known risks or to the conservative assumption of risk unless the absence of risk has been stated. This analysis is shown in Figure 21. Figure 21 shows the HI calculated for each health issue. The height of each bar is 2.88, which is the HI based on all ADIs (see above). Each bar is then broken into 4 portions. The top (white) portion can be discounted because it shows the HQs that derived from chemicals that are known (according to the PPDB classification) not to cause the health issue. The height of each bar when this portion is discarded can be compared to the critical value of 1 because the graph background is colour coded (blue below 1, red above 1). For ease of viewing, this graph is also shown in Figure 22 with the white portions removed. Portions of each bar coloured in red indicate the sum of HQs for chemicals that PPDB classified as known to cause an effect, portions coloured orange indicate the sum of HQs for chemical that have an uncertain data set and those coloured grey indicate the sum of HQs for chemicals which lack data. Detailed assessment of the toxicological relevance of these breakdowns may be inappropriate since the data may not be suitable for risk assessment, however the general pictures is that this analysis approach allows the reduction of risk estimates when chemicals can be excluded (when known not to have an given effect). In this case however, most risk estimates remained above the critical value of 1, and require further refinement because the risk cannot be deemed acceptable. For acetylcholinesterase inhibition the data set contained many chemicals Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

178 that were known not to cause the effect, and consequently the HI for this issues was reduced below 1 (0.9), indicating an acceptable risk. Use of reference values for each effect in place of ADI: adjusted ADIs If the data requirements for this tier were met, then ADIs could be 1) grouped by relevant chemicals and 2) replaced by use of the actual values for the effect under consideration. This was not possible for this case using PPDB classification due to the nonnumeric nature of the classification, however this is a feature of the summary nature of the PPDB, and may not be a concern in reality, i.e. if there is knowledge about testing for all endpoints there is also likely to be dose information too Tier 3 (group according to known or plausible toxicological independence) For this case detailed mechanism or mode of action information for all of the potential toxic effects was considered unlikely to be available. It seems likely that risk estimates can be refined in other ways before this massively data intensive step is reached. In the event of such information being available, the methods used throughout this case and which are based on DA, could be supplemented with methods based on IA. This would be a high tier assessment. Use of approaches based on IA should be done in a mixed model approach, which requires development. In such an approach all chemicals causing a given effect are first grouped according to similarity and the group risk is estimated with methods based on DA, and then the risk of all groups is aggregated using methods based on IA. The case for application of approaches based on pure IA rests on assumptions that are considered highly unlikely to ever be met, for example: That the toxic effect being considered is observed in an assay and that the observed effect can be caused by multiple, toxicologically independent routes. That all components cause a common effect but each through a toxicological independent route. In the case of a scenario with 10 components, this would imply 10 independent routes to one toxic effect Conclusions In low tiers the HI approach using TTC or ADI values produced a HI greater than one, indicating a greater than acceptable risk. Further refinement was not possible due to high data requirements although the approach that could be used was illustrated using PPDB health issue classifications. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

179 The HI is a composite of hazard and exposure data; however the overall driver of the HI value would appear to be the number of chemicals included. Opinions vary as to what the realistic human scenario should be, and this constitutes an important knowledge gap. The case identifies several issues for consideration: For cumulative risk assessment toxicological data need to be accessible and collated across endpoints and across chemicals. The data for hazard and exposure should be in comparable metrics and available in an open, standardised format. Data summarisation and censoring should be avoided. The available risk assessment approaches have differing data demands, and there should be a realistic assessment of whether these requirements will be met before an approach is employed, otherwise the likely outcome is that the assessment cannot be completed due to data requirements. Assessment approaches with high data demands should be placed high in framework approaches to give the best chance of completing risk assessments at lower tiers before high tiers that are likely to fail on data requirements are reached. Refinement of exposure should be considered, and possibly applied before more complex risk assessment approaches are employed. The data requirements for the selection of risk assessment approaches based on IA limit these approaches to very high tiers. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

180 Table 22: Hazard quotients (HQs) and hazard index (HI) for pesticides from JMPR 2009, 2010 Chemical name: GEMS region JMPR year Hazard quotient (IEDI / ADI) GEMS/Food consumption cluster diets A B C D E F G H I J K L M Africa Africa/Europe /Middle East Africa/Middle East Europe/Middl e East 2,6 DICHLOROBENZAMIDE BENALAXYL BUPROFEZIN CHLOPRYRIFOS METHYL CYPERMETHRIN FENBUCONAZOLE FLUOPICOLIDE HALOXYFOP HEXYTHIAZOX INDOXACARB METAFLUMIZONE METHOXYFENOZIDE PROTHIOCONAZOLE SPIRODICLOFEN Europe Europe Far East Latin America Africa Africa Latin America Far East Europe/Latin America Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

181 ZOXAMIDE BIFENAZATE BIFENTHRIN BOSCALID CADUSAFOS CHLORANTRANILIPROLE CHLOROTHALONIL CLOTHIANIDIN CYPROCONAZOLE DICAMBA DIFENOCONAZOLE ETOXAZOLE FENPYROXIMATE FLUBENDIAMIDE FLUDIOXONIL FLUOPYRAM MEPTYLDINOCAP NOVALURON THIAMETHOXAM TRIEAZOPHOS HI(sum of HQs): Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

182 Table 23: HI analyses based on ADI or TTC values ADI (JMPR) mg/kg bw exposure (mg/person) HQ (IEDI/ADI) Cramer class (ToxTree 2.5) Chemical name BENALAXYL III MEPTYLDINOCAP III METAFLUMIZONE III ZOXAMIDE III ,6 DICHLOROBENZAMIDE III CHLORANTRANILIPROLE III CADUSAFOS III DICAMBA III ETOXAZOLE III CLOTHIANIDIN III HEXYTHIAZOX III CYPROCONAZOLE III FLUDIOXONIL III PROTHIOCONAZOLE III FENBUCONAZOLE III THIAMETHOXAM III TRIEAZOPHOS III FENPYROXIMATE III METHOXYFENOZIDE III DIFENOCONAZOLE III SPIRODICLOFEN III FLUOPYRAM III FLUOPICOLIDE III BUPROFEZIN III INDOXACARB III FLUBENDIAMIDE III BIFENTHRIN III BIFENAZATE III CYPERMETHRIN III BOSCALID III NOVALURON III TTC (mg/person/day) HQ(IEDI/TTC) Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

183 CHLOPRYRIFOS METHYL III CHLOROTHALONIL III HALOXYFOP III Sum Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

184 A IEDI (mg/person) B ADI (mg/kg bw) C HQ D HQ E Cumulative HQ HALOXYFOP CHLOROTHALONIL CHLOPRYRIFOS METHYL NOVALURON BOSCALID CYPERMETHRIN BIFENAZATE BIFENTHRIN FLUBENDIAMIDE INDOXACARB BUPROFEZIN FLUOPICOLIDE FLUOPYRAM DIFENOCONAZOLE SPIRODICLOFEN METHOXYFENOZIDE FENPYROXIMATE TRIEAZOPHOS THIAMETHOXAM FENBUCONAZOLE CYPROCONAZOLE FLUDIOXONIL PROTHIOCONAZOLE HEXYTHIAZOX CLOTHIANIDIN DICAMBA ETOXAZOLE CADUSAFOS 26-DICHLOROBENZAMIDE CHLORANTRANILIPROLE BENALAXYL MEPTYLDINOCAP METAFLUMIZONE ZOXAMIDE Chemical name Figure 19: visualisation of HQ values and their contribution to the HI Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

185 Each panel in this graph shows a different attribute of the chemicals listed along the bottom x-axis. A: international estimated daily intakes (IEDI); B: ADI (JMPR), c) HQ (IEDI/ADI), D: HQ shown on a log y axis; E: cumulative HQ (HI). The solid line in Panel E indicates the HI (at 2.88), the dotted lines at 1 and 1.88 indicate the critical values of 1 and HI minus 1, respectively. The interpretation of these values is discussed in the text. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

186 6 5 4 ) T C (T 3 Q H HQ (ADI) Figure 20: comparison of HQ values based on ADI or TTC values Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

187 Table 24: PODI (NOAEL) analysis exposure (mg/person) ADI (JMPR) mg/kg bw safety factor' (jmpr) NOAEL (mg/kg bw) Chemical name 2,6 DICHLOROBENZAMIDE rat microscopic changes in the liver BENALAXYL dog atrophy of the seminiferous tubules E 05 BIFENAZATE dog compensatory haematopoiesis, alteration in urine analysis parameters and liver toxicity BIFENTHRIN rats increased incidence of tremors in dams during days of gestation and increased fetal and litter incidences of hydroureter without hydronephros species effect assay duration (d) HQ(exp/NOAEL) BOSCALID increased gamma glutamyltransferase activity and increased incidences of hepatic eosinophilic foci in male rats BUPROFEZIN rat increases in the incidence of thyroid F cell hypertrophy CADUSAFOS rat inhibition of erythrocyte cholinesterase activity longterm 4.44E 05 CHLOPRYRIFOS METHYL rat inhibition of brain acetylcholinesterase activity and adrenal vacuolation CHLORANTRANILIPROLE mice eosinophilic foci accompanied by hepatocellular E 05 dev tox Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

188 hypertrophy and increased liver weight CHLOROTHALONIL rat kidney toxicity longterm CLOTHIANIDIN rat decreased body weight and food consumption chron 9.28E 05 ic CYPERMETHRIN dog severe clinical signs of neurotoxicity CYPROCONAZOLE rat 2 year of toxicity and carcinogenicity and the multigeneration reproduction in rats based on reduced body weight gain and liver toxicity 730/ multi gen repro DICAMBA rabbit maternal toxicity (behavioural changes) dev tox DIFENOCONAZOLE reduced body weight gains, reduced platelet counts and hepatic hypertrophy ETOXAZOLE dog liver effects (e.g., increases in serum levels of AP and triglycerides, absolute and relative liver 365/ E 05 weights and incidence of centrilobular hepatocyte hypertrophy) FENBUCONAZOLE rat NOAEL from 2yr FENPYROXIMATE rat reductions in body weight gain and plasma protein concentration FLUBENDIAMIDE rat/dog effects on the liver (both sexes), kidney, thyroid and hair loss (females) and decreased eosinophil count (males) observed in a 2 year feeding in rats, and on the basis of a NOAEL of 100 ppm (equal to 2.2 mg/kg bw per day), based on increased alkaline phosphatase levels, shortened activated prothrombin time and increased liver weights observed in a 1 year in dogs Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

189 FLUDIOXONIL rat NOAEL from 2yr FLUOPICOLIDE mice/ra ts organ weight increases and gross and microscopic changes in the liver and kidneys in an 18 month dietary of toxicity and carcinogenicity in mice, supported by the NOAEL of 8.4 mg/kg bw per day identified on the basis of histopathological changes in the liver and increased kidney weights in a 2 year dietary of toxicity and carcinogenicity in rats FLUOPYRAM rat changes in liver (hepatocellular hypertrophy, eosinophilic foci) HALOXYFOP rat low pup body weight multi gen HEXYTHIAZOX rat increases in fatty vacuolation of the adrenals in both sexes, the severity of chronic nephritis and the incidence of swollen/withdrawn testes in males INDOXACARB dog erythrocyte damage and the secondary increase in haematopoiesis in the spleen and liver E MEPTYLDINOCAP dog reduced body weight gain in males E 05 METAFLUMIZONE dog clinical signs of poor general state of health, decreased food consumption reduced bodyweight gain and body weight loss, and changes in haematological parameters E 06 Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

190 METHOXYFENOZIDE rat/dog effects on erythrocytes plus liver and thyroid hypertrophy in the long term in rats, and 300 ppm, equal to 9.8 mg/kg bw per day, for haematological effects in the 1 year in dogs NOVALURON rat erythrocyte damage and secondary splenic and liver changes PROTHIOCONAZOLE rat gross and microscopic changes in the liver and kidneys in a 2 year of toxicity and carcinogenicity SPIRODICLOFEN dog adrenal effects in males and females, and increased relative testes weights in males THIAMETHOXAM dog prolonged thromboplastin time TRIEAZOPHOS Human NOEAL from human volunteer ZOXAMIDE dog reduced body weight gain in females E 05 longterm( r)/1yr (d) Sum: Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

191 Table 25: PPDB human health classifications Chemical name carcinogen mutagen endocrine disrupter reproduction/de velopment effect acetylcholinester ase inhibitor neurotoxicant respiratory tract irritant 2,6 x nd nd x nd nd nd nd nd DICHLOROBENZAMIDE BENALAXYL x nd x? x x x x x BIFENAZATE x nd nd nd x x nd yes yes BIFENTHRIN?? yes? x yes nd x x BOSCALID? nd x? x x nd x x BUPROFEZIN? nd x? x x x x x CADUSAFOS x nd x x yes? x x x CHLOPRYRIFOS METHYL x nd x nd yes? x yes x CHLORANTRANILIPROLE x nd x x x x nd x? CHLOROTHALONIL? x x nd x x yes yes yes CLOTHIANIDIN x nd?? x yes x x x CYPERMETHRIN? x?? x x yes yes yes CYPROCONAZOLE? nd nd? x x yes x x DICAMBA? x nd? x x x yes yes DIFENOCONAZOLE? nd nd? x x x yes yes ETOXAZOLE x nd nd? x x nd?? FENBUCONAZOLE? nd nd? x x? x x FENPYROXIMATE x nd nd yes x x nd yes yes FLUBENDIAMIDE nd nd yes yes x nd nd x? FLUDIOXONIL? nd nd? x x x yes yes FLUOPICOLIDE? nd nd x x x x x x FLUOPYRAM nd nd nd nd nd nd nd nd nd HALOXYFOP? nd nd? nd nd nd yes yes HEXYTHIAZOX? nd nd nd x x yes yes yes INDOXACARB x nd nd x x yes x yes yes MEPTYLDINOCAP x nd x x x x yes yes METAFLUMIZONE nd nd nd yes x nd nd nd nd METHOXYFENOZIDE x nd? x x x nd?? NOVALURON x nd nd x x x??? PROTHIOCONAZOLE x nd nd? x nd x x x SPIRODICLOFEN? nd nd? x? nd x x THIAMETHOXAM? nd x x x x x? x skin irritant eye irritant Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

192 TRIEAZOPHOS x nd nd nd yes yes yes yes yes ZOXAMIDE x nd nd x x x nd x yes For each effect, each chemical was classified in the PPDB as yes : chemical known to cause a problem (cells filled red); x : chemical known not to cause a problem (cells filled white);? : Possibly, status of chemical not identified (cells filled orange); nd : No data (cells filled grey). A single value for which a classification as lacking in the PPDB is indicated by a cell filled black. Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

193 Figure 21: HI analysis breakdown by PPDB health issues Supporting publications 2012:EN exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the European Food Safety Authority is subject. It may not be considered as an output adopted by EFSA. EFSA reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

EFSA s perspective on risk assessment of chemical mixtures

EFSA s perspective on risk assessment of chemical mixtures EFSA s perspective on risk assessment of chemical mixtures Christer Hogstrand King s College London Jean-Lou Dorne European Food Safety Authority, Scientific Committee Past Activities on Chemical Mixtures

More information

pesticides in the EU regulatory framework Prof Andreas Kortenkamp

pesticides in the EU regulatory framework Prof Andreas Kortenkamp Addressing combined effects of pesticides in the EU regulatory framework Prof Andreas Kortenkamp Brunel University London Info session on cumulative risk assessment for o sess o o cu u at e s assess e

More information

EFSA work on Cumulative Risk Assessment of pesticides. Luc Mohimont EFSA Pesticides Unit EuroMix Project 20/05/2015

EFSA work on Cumulative Risk Assessment of pesticides. Luc Mohimont EFSA Pesticides Unit EuroMix Project 20/05/2015 EFSA work on Cumulative Risk Assessment of pesticides Luc Mohimont EFSA Pesticides Unit EuroMix Project 20/05/2015 Content State of Play with the dietary cumulative risk assessment in EFSA Establishment

More information

State of the Art Report on Mixture Toxicity. Executive Summary

State of the Art Report on Mixture Toxicity. Executive Summary State of the Art Report on Mixture Toxicity Final Report Executive Summary 22 December 2009 Study Contract Number 070307/2007/485103/ETU/D.1 Contractor The School of Pharmacy University of London (ULSOP)

More information

European Food Safety Authority (EFSA)

European Food Safety Authority (EFSA) TECHNICAL REPORT APPROVED: 06 April 2017 doi:10.2903/sp.efsa.2017.en-1210 Outcome of the preliminary pesticides peer review meeting on the assessment of endocrine disrupting properties in mammalian toxicology

More information

Thought Starter Combined Exposures to Multiple Chemicals Second International Conference on Risk Assessment

Thought Starter Combined Exposures to Multiple Chemicals Second International Conference on Risk Assessment Thought Starter Combined Exposures to Multiple Chemicals Second International Conference on Risk Assessment M.E. (Bette) Meek & A. Kortenkamp 1 Outline State of the Art Assessment of Mixtures (aka Combined

More information

FÜR RISIKOBEWERTUNG BUNDESINSTITUT

FÜR RISIKOBEWERTUNG BUNDESINSTITUT BUNDESINSTITUT FÜR RISIKOBEWERTUNG Legal and Practical Aspects of the Cut-off Criteria for Reproductive Toxic and Endocrine Disrupting Effects for Approval and Classification of Pesticides in Europe Roland

More information

Recent Developments and Future Plans in the EFSA Assessments of Pesticides. Hermine Reich Pesticides Unit

Recent Developments and Future Plans in the EFSA Assessments of Pesticides. Hermine Reich Pesticides Unit Recent Developments and Future Plans in the EFSA Assessments of Pesticides Hermine Reich Pesticides Unit Pesticides Unit and Panel activities Scientific Panel on Plant Protection Product and their Residues

More information

Cumulative Risk Assessment

Cumulative Risk Assessment Cumulative Risk Assessment Acropolis - Better tools Jørgen Schlundt Bodil Hamborg Jensen Risk Analysis - DK Risk Assessment National Food Institute Independent science Risk Management Danish Vet. And Food

More information

Cumulative risk assessment legal

Cumulative risk assessment legal Technical Meeting with stakeholders on Cumulative Risk Assessment Cumulative risk assessment legal framework and perspective of DG SANCO Veerle Vanheusden Almut Bitterhof European Commission DG Overview

More information

STUDIES TO EVALUATE THE SAFETY OF RESIDUES OF VETERINARY DRUGS IN HUMAN FOOD: GENERAL APPROACH TO ESTABLISH AN ACUTE REFERENCE DOSE

STUDIES TO EVALUATE THE SAFETY OF RESIDUES OF VETERINARY DRUGS IN HUMAN FOOD: GENERAL APPROACH TO ESTABLISH AN ACUTE REFERENCE DOSE VICH GL54 (SAFETY) ARfD November 2016 For Implementation at Step 7 STUDIES TO EVALUATE THE SAFETY OF RESIDUES OF VETERINARY DRUGS IN HUMAN FOOD: GENERAL APPROACH TO ESTABLISH AN ACUTE REFERENCE DOSE (ARfD)

More information

CUMULATIVE RISK ASSESSMENT OF PESTICIDES TO HUMAN HEALTH: THE WAY FORWARD

CUMULATIVE RISK ASSESSMENT OF PESTICIDES TO HUMAN HEALTH: THE WAY FORWARD EFSA SCIENTIFIC COLLOQUIUM SUMMARY REPORT 7 CUMULATIVE RISK ASSESSMENT OF PESTICIDES TO HUMAN HEALTH: THE WAY FORWARD ISSN 1830-4737 28-29 November 2006 - Parma, Italy 2. Summary Report EFSA Scientific

More information

Evaluation of active substances in plant protection products Residues Anja Friel European Food Safetey Authority, Parma/ Italy

Evaluation of active substances in plant protection products Residues Anja Friel European Food Safetey Authority, Parma/ Italy Evaluation of active substances in plant protection products Residues Anja Friel European Food Safetey Authority, Parma/ Italy European Conference on MRL-Setting for Biocides Berlin, 18-19 March 2014 Legal

More information

EFSA Info Session Pesticides 26/27 September Anja Friel EFSA Pesticides Unit (Residues team)

EFSA Info Session Pesticides 26/27 September Anja Friel EFSA Pesticides Unit (Residues team) Scientific Guidance Document of the PPR Panel on the establishment of the residue definition to be used for dietary risk assessment (EFSA-Q-2013-01001) EFSA Info Session Pesticides 26/27 September 2016

More information

Preparatory work to support the re-evaluation of technological feed additives

Preparatory work to support the re-evaluation of technological feed additives EXTERNAL SCIENTIFIC REPORT APPROVED: 26 March 2015 PUBLISHED: 10 April 2015 Preparatory work to support the re-evaluation of technological feed additives IRTA 1, ACSA 2 N. Tous 1, J. Brufau 1, A. Pérez-Vendrell

More information

Guidance on the review, revision and development of EFSA s cross-cutting guidance documents

Guidance on the review, revision and development of EFSA s cross-cutting guidance documents SCIENTIFIC OPINION ADOPTED: 1 April 2015 PUBLISHED: 16 April 2015 AMENDED: 20 July 2016 doi:10.2903/j.efsa.2015.4080 Guidance on the review, revision and development of EFSA s cross-cutting guidance documents

More information

COMMISSION REGULATION (EU)

COMMISSION REGULATION (EU) 11.3.2011 Official Journal of the European Union L 64/15 COMMISSION REGULATION (EU) No 234/2011 of 10 March 2011 implementing Regulation (EC) No 1331/2008 of the European Parliament and of the Council

More information

Questions and Answers on Candidates for Substitution

Questions and Answers on Candidates for Substitution Questions and Answers on Candidates for Substitution Rev. 1, January 2015 Background The European Commission is required by Regulation (EC) No 1107/2009 ( the Regulation ) to establish a list of substances

More information

Pesticide risk assessment: changes and perspectives for mammalian toxicology in the new EC regulation 1107/2009

Pesticide risk assessment: changes and perspectives for mammalian toxicology in the new EC regulation 1107/2009 Pesticide risk assessment: changes and perspectives for mammalian toxicology in the new EC regulation 1107/2009 M.Tiramani Pesticide Risk Assessment Peer Review (PRAPeR) Mammalian toxicology New Pesticide

More information

Practical guidance for applicants on the submission of applications on food additives, food enzymes and food flavourings

Practical guidance for applicants on the submission of applications on food additives, food enzymes and food flavourings Version 2 Updated on 29/11/2011 Practical guidance for applicants on the submission of applications on food additives, food enzymes and food flavourings Valid as of: 11 September 2011 Disclaimer: This

More information

DRAFT COMMISSION DELEGATED REGULATION (EU) /... of XXX

DRAFT COMMISSION DELEGATED REGULATION (EU) /... of XXX EUROPEAN COMMISSION Brussels, XXX C(2016) 3752 projet DRAFT COMMISSION DELEGATED REGULATION (EU) /... of XXX setting out scientific criteria for the determination of endocrine-disrupting properties pursuant

More information

Harmonisation of human and ecological risk assessment of combined exposure to multiple chemicals

Harmonisation of human and ecological risk assessment of combined exposure to multiple chemicals EFSA Scientific Colloquium Harmonisation of human and ecological risk assessment of combined exposure to multiple chemicals Andreas Kortenkamp Institute of Environment, Health and Societies, Brunel University

More information

COMMISSION REGULATION (EU) / of XXX

COMMISSION REGULATION (EU) / of XXX EUROPEAN COMMISSION Brussels, XXX SANTE/11992/2017 Rev. 0 [ ](2017) XXX draft COMMISSION REGULATION (EU) / of XXX amending Annex II to Regulation (EC) No 1107/2009 by setting out scientific criteria for

More information

Action plan for improving the peer review process. European Food Safety Authority (EFSA)

Action plan for improving the peer review process. European Food Safety Authority (EFSA) TECHNICAL REPORT APPROVED: 29 November 2017 doi:10.2903/sp.efsa.2017.en-1349 Action plan for improving the peer review process European Food Safety Authority (EFSA) Abstract This document reflects on the

More information

REASONED OPINION. European Food Safety Authority 2, 3. European Food Safety Authority (EFSA), Parma, Italy

REASONED OPINION. European Food Safety Authority 2, 3. European Food Safety Authority (EFSA), Parma, Italy EFSA Journal 2012;10(7):2841 REASONED OPINION Reasoned opinion on the review of the existing maximum residue levels (MRLs) for paraffin oil (CAS 64742-54-7) according to Article 12 of Regulation (EC) No

More information

Risk Assessment to Risk Management Terminology of Risk Assessment. EFSA Project on Risk Assessment Terminology

Risk Assessment to Risk Management Terminology of Risk Assessment. EFSA Project on Risk Assessment Terminology Risk Assessment to Risk Management Terminology of Risk Assessment EFSA Project on Risk Assessment Terminology Professor Tony Hardy, Chair of the Panel on Plant Protection Products and their Residues 2

More information

SCIENTIFIC / TECHNICAL REPORT submitted to EFSA

SCIENTIFIC / TECHNICAL REPORT submitted to EFSA SCIENTIFIC / TECHNICAL REPORT submitted to EFSA Applicability of thresholds of toxicological concern in the dietary risk assessment of metabolites, degradation and reaction products of pesticides 1 Prepared

More information

Potential Impacts Regarding Human Health Risk Assessment

Potential Impacts Regarding Human Health Risk Assessment FEDERAL INSTITUTE FOR RISK ASSESSMENT EU CONFERENCE ON ENDOCRINE DISRUPTORS Criteria for Identification and Related Impacts Potential Impacts Regarding Human Health Risk Assessment Andreas Hensel One Substance

More information

SCIENTIFIC OPINION. Scientific Opinion of the Panel on Plant Protection Products and their Residues (PPR) (Question No EFSA-Q )

SCIENTIFIC OPINION. Scientific Opinion of the Panel on Plant Protection Products and their Residues (PPR) (Question No EFSA-Q ) The EFSA Journal (2009) 1171, 1-6 SCIENTIFIC OPINION Updating the opinion related to the revision of Annexes II and III to Council Directive 91/414/EEC concerning the placing of plant protection products

More information

European Food Safety Authority (EFSA)

European Food Safety Authority (EFSA) TECHNICAL REPORT APPROVED: 03/05/2017 doi:10.2903/sp.efsa.2017.en-1223 Outcome of the consultation with Member States, the applicant and EFSA on the pesticide risk assessment for L-ascorbic acid in light

More information

Overview of the procedures currently used at EFSA for the assessment of dietary exposure to chemical substances

Overview of the procedures currently used at EFSA for the assessment of dietary exposure to chemical substances Overview of the procedures currently used at EFSA for the assessment of dietary exposure to chemical substances Davide Arcella Dietary and Chemical Monitoring (DCM) Unit Inputs Chemical Occurrence Hazard

More information

First Phase of Impact Assessment on Endocrine Disruptors:

First Phase of Impact Assessment on Endocrine Disruptors: First Phase of Impact Assessment on Endocrine Disruptors: How to screen which chemicals would fall under different options for criteria to identify endocrine disruptors EU Conference on Endocrine Disruptors,

More information

EFSA and Member States

EFSA and Member States Scientific Cooperation between EFSA and Member States Taking Stock and Looking Ahead Committed to ensuring that Europe s food is safe Scientific Cooperation between EFSA and Member States Scientific cooperation

More information

EUROPEAN COMMISSION HEALTH & CONSUMER PROTECTION DIRECTORATE-GENERAL. Joint meetrag of the Earopean Commission Scieatäfíc Committees and the Еигореав

EUROPEAN COMMISSION HEALTH & CONSUMER PROTECTION DIRECTORATE-GENERAL. Joint meetrag of the Earopean Commission Scieatäfíc Committees and the Еигореав Ref. Ares(2012)123978-03/02/2012 EUROPEAN COMMISSION HEALTH & CONSUMER PROTECTION DIRECTORATE-GENERAL Directorate D - Public Health and Risk Assessment Unit D5 - Risk Assessment Brussels, 11 July 2011

More information

Welcome and introduction to EFSA

Welcome and introduction to EFSA Committed since 2002 to ensuring that Europe s food is safe Welcome and introduction to EFSA Claudia Heppner Head - Food Ingredients and Packaging Unit (FIP) Scientific Evaluation of Regulated Products

More information

Food additives and nutrient sources added to food: developments since the creation of EFSA

Food additives and nutrient sources added to food: developments since the creation of EFSA EFSA Journal 2012;10(10):s1006 SPECIAL ISSUE Food additives and nutrient sources added to food: developments since the creation of EFSA Birgit Dusemund, John Gilbert, David Gott, Hugues Kenigswald, Jürgen

More information

The regulatory landscape. The now and the not yet

The regulatory landscape. The now and the not yet The regulatory landscape The now and the not yet Perspectives Aims Promote common understanding Anticipate the coming changes Prepare for afternoon sessions Who governs pesticides? All EU legislation comes

More information

The "Cocktail-effect" Do pesticides play a role?

The Cocktail-effect Do pesticides play a role? The "Cocktail-effect" Do pesticides play a role? /////////// Martin Larsson Researcher & Regulatory Scientist Bayer Crop Science, Denmark 9 November 2018 Talking points Why are we discussing potential

More information

Endocrine Disruptors

Endocrine Disruptors Endocrine Disruptors Dr. Bettina Hrdina-Zödl Institute for Plant Protection Products, Department of Toxicology Antragstellerkonferenz Vienna, 27.03.2014 www.ages.at Österreichische Agentur für Gesundheit

More information

Guidance on the Use of Probabilistic Methodology for Modelling Dietary Exposure to Pesticide Residues 1

Guidance on the Use of Probabilistic Methodology for Modelling Dietary Exposure to Pesticide Residues 1 DRAFT SCIENTIFIC OPINION Guidance on the Use of Probabilistic Methodology for Modelling Dietary Exposure to Pesticide Residues 1 EFSA Panel on Plant Protection Products and their Residues (PPR) 2, 3 European

More information

Overview of the procedures currently used at EFSA for the assessment of dietary exposure to different chemical substances 1

Overview of the procedures currently used at EFSA for the assessment of dietary exposure to different chemical substances 1 EFSA Journal 2011;9(12):2490 SCIENTIFIC REPORT OF EFSA Overview of the procedures currently used at EFSA for the assessment of dietary exposure to different chemical substances 1 ABSTRACT European Food

More information

TNsG on Annex I Inclusion Revision of Chapter 4.1: Quantitative Human Health Risk Characterisation

TNsG on Annex I Inclusion Revision of Chapter 4.1: Quantitative Human Health Risk Characterisation TNsG on Annex I Inclusion Revision of Chapter 4.1: Quantitative Human Health Risk Characterisation These Technical Notes for Guidance were adopted during the 34 th meeting of representatives of Members

More information

Classification of developmentally toxic pesticides, low dose effects, mixtures perspective of the industry

Classification of developmentally toxic pesticides, low dose effects, mixtures perspective of the industry Classification of developmentally toxic pesticides, low dose effects, mixtures perspective of the industry Steffen Schneider 9th Berlin Workshop on Developmental Toxicology Berlin September 13, 2018 What

More information

Challenges in environmental risk assessment (ERA) for birds and mammals and link to endocrine disruption (ED) Katharina Ott, BASF SE, Crop Protection

Challenges in environmental risk assessment (ERA) for birds and mammals and link to endocrine disruption (ED) Katharina Ott, BASF SE, Crop Protection Challenges in environmental risk assessment (ERA) for birds and mammals and link to endocrine disruption (ED) Katharina Ott, BASF SE, Crop Protection Charles River Symposium, Den Bosch, 3rd October 2017

More information

Guidance on the Use of Probabilistic Methodology for Modelling Dietary Exposure to Pesticide Residues 1

Guidance on the Use of Probabilistic Methodology for Modelling Dietary Exposure to Pesticide Residues 1 SCIENTIFIC OPINION Guidance on the Use of Probabilistic Methodology for Modelling Dietary Exposure to Pesticide Residues 1 EFSA Panel on Plant Protection Products and their Residues (PPR) 2, 3 European

More information

REPORT OF THE SPECIAL ADVISORY FORUM MEETING ON EU GMO RISK ASSESSMENT

REPORT OF THE SPECIAL ADVISORY FORUM MEETING ON EU GMO RISK ASSESSMENT REPORT OF THE SPECIAL ADVISORY FORUM MEETING ON EU GMO RISK ASSESSMENT 13 November 2007 Introduction Over 60 EU GMO risk assessment experts, nominated by the Advisory Forum members and representing the

More information

TTC and science. Hans Muilerman, PAN Europe

TTC and science. Hans Muilerman, PAN Europe TTC and science. Hans Muilerman, PAN Europe www.pan-europe.info TTC, science or politics? How does a pragmatic, US non-precautionary principle style, tool fit in the EU toolbox? EFSA opinion: Safe level

More information

Genotoxicity Testing Strategies: application of the EFSA SC opinion to different legal frameworks in the food and feed area

Genotoxicity Testing Strategies: application of the EFSA SC opinion to different legal frameworks in the food and feed area Genotoxicity Testing Strategies: application of the EFSA SC opinion to different legal frameworks in the food and feed area Juan Manuel Parra Morte. Pesticides Unit. EFSA. 19th Annual Conference of the

More information

EFSA s Catalogue of support initiatives during the lifecycle of applications for regulated products

EFSA s Catalogue of support initiatives during the lifecycle of applications for regulated products TECHNICAL REPORT APPROVED: 12 March 2015 PUBLISHED: 16 03 2015 EFSA s Catalogue of support initiatives during the lifecycle of applications for regulated products Abstract European Food Safety Authority

More information

Scientific Panel on Plant Protection Products and their Residues

Scientific Panel on Plant Protection Products and their Residues PESTICIDES UNIT Scientific Panel on Plant Protection Products and their Residues Minutes of the 6 th meeting of the PPR Working Group on pesticides in foods for infants and young children Held on 20-21

More information

ECPA position paper on the criteria for the determination of endocrine disrupting properties under Regulation

ECPA position paper on the criteria for the determination of endocrine disrupting properties under Regulation POSITION PAPER 09/06/2016 PP/14/PD/23734 ECPA position paper on the criteria for the determination of endocrine disrupting properties under Regulation The European Commission is currently developing new

More information

European Food Safety Authority

European Food Safety Authority MINUTES OF THE 24 th PLENARY MEETING OF THE SCIENTIFIC PANEL ON PLANT PROTECTION PRODUCTS AND THEIR RESIDUES held in Parma on 31 January 1 February 2007 (adopted by written procedure on 21 February 2007)

More information

EU policy on acrylamide in food reducing human exposure to ensure a high level of human health protection

EU policy on acrylamide in food reducing human exposure to ensure a high level of human health protection Directorate-General for Health & Food Safety EU policy on acrylamide in food reducing human exposure to ensure a high level of human health protection Frans Verstraete Principles for regulating contaminants

More information

Application of human epidemiological studies to pesticide risk assessment

Application of human epidemiological studies to pesticide risk assessment Workshop What does the future hold for harmonised human health risk assessment of plant protection products? Application of human epidemiological studies to pesticide risk assessment Antonio F. Hernández,

More information

Risk Assessment Terminology, Expression of Nature and level of Risk and Uncertainties

Risk Assessment Terminology, Expression of Nature and level of Risk and Uncertainties Risk Assessment Terminology, Expression of Nature and level of Risk and Uncertainties Professor Tony Hardy, Chair of EFSA PPR Panel Brussels 13 th November 2008 Comparative Review of Risk Terminology of

More information

Parma, 21/09/2015 SCIENTIFIC EVALUATION OF REGULATED PRODUCTS DEPARTMENT

Parma, 21/09/2015 SCIENTIFIC EVALUATION OF REGULATED PRODUCTS DEPARTMENT SCIENTIFIC EVALUATION OF REGULATED PRODUCTS DEPARTMENT Parma, 21/09/2015 Note on the establishment of a Standing Working Group on Dietary Reference Values for vitamins of the Scientific on Dietetic Products,

More information

Draft Concept Paper. 05. Mai Seite 1 von 20

Draft Concept Paper. 05. Mai Seite 1 von 20 Draft Concept Paper Development of a Stepwise Procedure for the Assessment of Substances with Endocrine Disrupting Properties According to the Plant Protection Products Regulation (Reg. (EC) No 1107/2009)

More information

EFSA cross-cutting guidance lifecycle. European Food Safety Authority (EFSA), Daniela Maurici, Raquel Garcia Matas, Andrea Gervelmeyer

EFSA cross-cutting guidance lifecycle. European Food Safety Authority (EFSA), Daniela Maurici, Raquel Garcia Matas, Andrea Gervelmeyer TECHNICAL REPORT APPROVED: 29 June 2018 doi:10.2903/sp.efsa.2018.en-1446 Abstract EFSA cross-cutting guidance lifecycle European Food Safety Authority (EFSA), Daniela Maurici, Raquel Garcia Matas, Andrea

More information

DOSE SELECTION FOR CARCINOGENICITY STUDIES OF PHARMACEUTICALS *)

DOSE SELECTION FOR CARCINOGENICITY STUDIES OF PHARMACEUTICALS *) DOSE SELECTION FOR CARCINOGENICITY STUDIES OF PHARMACEUTICALS *) Guideline Title Dose Selection for Carcinogenicity Studies of Pharmaceuticals *) Legislative basis Directive 75/318/EEC as amended Date

More information

EU REFERENCE LABORATORIES FOR AVIAN INFLUENZA AND NEWCASTLE DISEASE

EU REFERENCE LABORATORIES FOR AVIAN INFLUENZA AND NEWCASTLE DISEASE EUROPEAN COMMISSION HEALTH & CONSUMERS DIRECTORATE-GENERAL G2- Animal Health 04 Veterinary Control programmes SANCO/7048/204 Draft Working document EU REFERENCE LABORATORIES FOR AVIAN INFLUENZA AND NEWCASTLE

More information

Consultation Response

Consultation Response Consultation Response CHEM Trust response to EFSA consultation on a draft scientific opinion on: Recent developments in the risk assessment of chemicals in food and their potential impact on the safety

More information

Contribution to the European Commission s Public Consultation on. 15 January 2015

Contribution to the European Commission s Public Consultation on. 15 January 2015 Contribution to the European Commission s Public Consultation on Defining criteria for identifying Endocrine Disruptors in the context of the implementation of the Plant Protection Product Regulation and

More information

Survey results - Analysis of higher tier studies submitted without testing proposals

Survey results - Analysis of higher tier studies submitted without testing proposals Survey results - Analysis of higher tier studies submitted without testing proposals Submission of higher tier studies on vertebrate animals for REACH registration without a regulatory decision on testing

More information

What are the challenges in addressing adjustments for data uncertainty?

What are the challenges in addressing adjustments for data uncertainty? What are the challenges in addressing adjustments for data uncertainty? Hildegard Przyrembel, Berlin Federal Institute for Risk Assessment (BfR), Berlin (retired) Scientific Panel for Dietetic Foods, Nutrition

More information

WORLD TRADE ORGANIZATION

WORLD TRADE ORGANIZATION WORLD TRADE ORGANIZATION WT/DS321/R/Add.5 31 March 2008 (08-0912) Original: English CANADA CONTINUED SUSPENSION OF OBLIGATIONS IN THE EC HORMONES DISPUTE Report of the Panel Addendum This addendum contains

More information

Sample size calculation tool for monitoring stunning at slaughter 1 2

Sample size calculation tool for monitoring stunning at slaughter 1 2 EFSA supporting publication 2013:EN-541 TECHNICAL REPORT Sample size calculation tool for monitoring stunning at slaughter 1 2 European Food Safety Authority 3, 4 European Food Safety Authority (EFSA),

More information

Who are we? What are we doing? Why are we doing it?

Who are we? What are we doing? Why are we doing it? European Food Safety Authority Advising the EU on TSE: A discussion of the way in which the EU seeks scientific advice in relation to TSEs Bart Goossens Scientific Coordinator Panel on Biological Hazards

More information

Science Policy Notice

Science Policy Notice Science Policy Notice SPN2002-01 Children s Health Priorities within the Pest Management Regulatory Agency (publié aussi en français) January 3, 2002 This document is published by the Submission Coordination

More information

Title: Scientific principles for the identification of endocrine disrupting chemicals a consensus statement

Title: Scientific principles for the identification of endocrine disrupting chemicals a consensus statement Title: Scientific principles for the identification of endocrine disrupting chemicals a consensus statement Outcome of an international expert meeting organized by the German Federal Institute for Risk

More information

COMMISSION OF THE EUROPEAN COMMUNITIES REPORT FROM THE COMMISSION TO THE COUNCIL AND THE EUROPEAN PARLIAMENT

COMMISSION OF THE EUROPEAN COMMUNITIES REPORT FROM THE COMMISSION TO THE COUNCIL AND THE EUROPEAN PARLIAMENT EN EN EN COMMISSION OF THE EUROPEAN COMMUNITIES Brussels, 5.12.2008 COM(2008) 824 final REPORT FROM THE COMMISSION TO THE COUNCIL AND THE EUROPEAN PARLIAMENT on the use of substances other than vitamins

More information

Software tool for calculating the predicted environmental concentrations (PEC) of plant protection products (PPP) in soil:

Software tool for calculating the predicted environmental concentrations (PEC) of plant protection products (PPP) in soil: EXTERNAL SCIENTIFIC REPORT APPROVED: 12/10/2016 Software tool for calculating the predicted environmental concentrations (PEC) of plant protection products (PPP) in soil: Abstract Final report VITO NV

More information

EFSA Statement regarding the EU assessment of glyphosate and the socalled

EFSA Statement regarding the EU assessment of glyphosate and the socalled EFSA Statement regarding the EU assessment of glyphosate and the socalled Monsanto papers Background On 29 May 2017, EFSA received a request from the European Commission to produce a statement concerning

More information

CRD03. Introduction 1

CRD03. Introduction 1 Introduction 1 JOINT FAO/WHO FOOD STANDARDS PROGRAMME CODEX COMMITTEE ON PESTICIDE RESIDUES 48 th Session Chongqing, P.R. China, 25-30 April 2016 DISCUSSION PAPER REVISITING THE INTERNATIONAL ESTIMATE

More information

2008 Public Status Report on the Implementation of the European Risk Management Strategy. Executive Summary

2008 Public Status Report on the Implementation of the European Risk Management Strategy. Executive Summary European Medicines Agency London, 17 March 2009 Doc. Ref. EMEA/43556/2009 2008 Status Report on the Implementation of the European Risk Management Strategy Executive Summary The European Risk Management

More information

COUNCIL OF THE EUROPEAN UNION. Brussels, 7 September 2009 (OR. en) 11261/09 Interinstitutional File: 2008/0002 (COD) DENLEG 51 CODEC 893

COUNCIL OF THE EUROPEAN UNION. Brussels, 7 September 2009 (OR. en) 11261/09 Interinstitutional File: 2008/0002 (COD) DENLEG 51 CODEC 893 COUNCIL OF THE EUROPEAN UNION Brussels, 7 September 2009 (OR. en) 11261/09 Interinstitutional File: 2008/0002 (COD) DLEG 51 CODEC 893 LEGISLATIVE ACTS AND OTHER INSTRUMTS Subject: Common Position with

More information

CHAPTER 2: RISK ANALYSIS

CHAPTER 2: RISK ANALYSIS Update Project Chapter : Risk Analysis Draft May 00 0 0 0 0 0 PRINCIPLES AND METHODS FOR THE RISK ASSESSMENT OF CHEMICALS IN FOOD CHAPTER : RISK ANALYSIS Contents CHAPTER : RISK ANALYSIS.... INTRODUCTION....

More information

Module 34: Legal aspects, ADI and GRAS status of food additives

Module 34: Legal aspects, ADI and GRAS status of food additives Paper No.: 13 Paper Title: FOOD ADDITIVES Module 34: Legal aspects, ADI and GRAS status of food additives 34.1 Legal Aspects of Food Additives The data provided by Joint Expert Committee on Food Additives

More information

Scientific Cooperation and Networking between EFSA and Member States

Scientific Cooperation and Networking between EFSA and Member States Scientific Cooperation and Networking between EFSA and Member States Carola Sondermann EFSA Scientific Cooperation Unit EFSA Workplan 2010 EFSA in 2010: an established player in Europe s food safety system,

More information

Prediction of mixture effects?

Prediction of mixture effects? Swedish Society of Toxicology Is mixtures risk assessment necessary and what are obstacles to making progress? Andreas Kortenkamp Institue for the Environment, Brunel University London 23 March 2012, Stockholm,

More information

Putting thresholds into practice: where are we now?

Putting thresholds into practice: where are we now? Putting thresholds into practice: where are we now? Anaphylaxis Campaign Corporate Members Conference, The Brewery, London Allergen Thresholds: the complete picture René Crevel René Crevel Consulting Limited

More information

of the French Agency for Food, Environmental and Occupational Health & Safety

of the French Agency for Food, Environmental and Occupational Health & Safety The Director General Maisons-Alfort, 27 March 2012 of the French Agency for Food, Environmental and Occupational Health & Safety regarding a request for scientific and technical support for the revising

More information

Assessing and Managing Health Risks from Chemical Constituents and Contaminants of Food

Assessing and Managing Health Risks from Chemical Constituents and Contaminants of Food 16 17 September 2013 Assessing and Managing Health Risks from Chemical Constituents and Contaminants of Food Workshop on A Framework for Assessing the Health, Environmental and Social Effects of the Food

More information

Ongoing review of legislation on cadmium in food in the EU: Background and current state of play

Ongoing review of legislation on cadmium in food in the EU: Background and current state of play Directorate-General for Health & Ongoing review of legislation on cadmium in food in the EU: Background and current state of play - International ICCO workshop, London, 3-4 May 2012 Michael Flüh bind the

More information

Action Levels and Allergen Thresholds What they will mean for the Food Industry Dr. Rachel WARD r.ward consultancy limited

Action Levels and Allergen Thresholds What they will mean for the Food Industry Dr. Rachel WARD r.ward consultancy limited Action Levels and Allergen Thresholds What they will mean for the Food Industry Dr. Rachel WARD r.ward consultancy limited 1 Allergenic Foods Are Unique! More than 160 foods are known to provoke allergic

More information

OECD QSAR Toolbox v.4.2. An example illustrating RAAF scenario 6 and related assessment elements

OECD QSAR Toolbox v.4.2. An example illustrating RAAF scenario 6 and related assessment elements OECD QSAR Toolbox v.4.2 An example illustrating RAAF scenario 6 and related assessment elements Outlook Background Objectives Specific Aims Read Across Assessment Framework (RAAF) The exercise Workflow

More information

Secretary-General of the European Commission, signed by Mr Jordi AYET PUIGARNAU, Director

Secretary-General of the European Commission, signed by Mr Jordi AYET PUIGARNAU, Director COUNCIL OF THE EUROPEAN UNION Brussels, 13 February 2014 (OR. en) 6438/14 COVER NOTE From: date of receipt: 3 February 2014 To: No. Cion doc.: PHARM 14 SAN 72 MI 161 COMPET 107 DELACT 29 Secretary-General

More information

APPROVED: 17 March 2015 PUBLISHED: 27 March 2015

APPROVED: 17 March 2015 PUBLISHED: 27 March 2015 TECHNICAL REPORT APPROVED: 17 March 2015 PUBLISHED: 27 March 2015 Outcome of the consultation with Member States, the applicant and EFSA on the pesticide risk assessment for tall oil crude in light of

More information

Feedback on the Training on the Risk Assessment Model ImproRisk

Feedback on the Training on the Risk Assessment Model ImproRisk Feedback on the Training on the Risk Assessment Model ImproRisk Dr. Georgios Stavroulakis, Scientific Fellow of EFSA FP of Cyprus and Member of Risk Assessment (RA) Unit of the State General Laboratory

More information

Use of TTC and Human Relevance George E. N. Kass, PhD

Use of TTC and Human Relevance George E. N. Kass, PhD Use of TTC and Human Relevance George E. N. Kass, PhD Future Challenges in Developing Assessment Methodologies for Human Health Effects Tokyo, 14 November 2018 Disclaimer The views, thoughts and opinions

More information

Emanuela Turla Scientific Officer Nutrition Unit - EFSA

Emanuela Turla Scientific Officer Nutrition Unit - EFSA EFSA s role, experiences with the evaluation of the applications for authorisation of Novel Food or notification of Traditional Food from the third country Emanuela Turla Scientific Officer Nutrition Unit

More information

Risk Management Option Analysis Conclusion Document

Risk Management Option Analysis Conclusion Document Risk Management Option Analysis Conclusion Document Substance Name: tributyl O-acetylcitrate (ATBC) EC Number: 201-067-0 CAS Number: 77-90-7 Authority: France Date: August 2016 Version 2.1 October 2015

More information

Statistical analysis of comparative data on composition and agronomic characteristics: new software tool and recurring issues identified

Statistical analysis of comparative data on composition and agronomic characteristics: new software tool and recurring issues identified Statistical analysis of comparative data on composition and agronomic characteristics: new software tool and recurring issues identified C. Paoletti*, J. Perry and H. Broll* * EFSA GMO Unit; EFSA GMO Panel

More information

EUROPEAN COMMISSION. Modus Operandi for the management of new food safety incidents with a potential for extension involving a chemical substance

EUROPEAN COMMISSION. Modus Operandi for the management of new food safety incidents with a potential for extension involving a chemical substance EUROPEAN COMMISSION Modus Operandi for the management of new food safety incidents with a potential for extension involving a chemical substance The Health and Consumer Protection Directorate-General of

More information

TECHNICAL REPORT OF EFSA. List of guidance, guidelines and working documents developed or in use by EFSA 1

TECHNICAL REPORT OF EFSA. List of guidance, guidelines and working documents developed or in use by EFSA 1 EFSA Technical Report (2009) 279, 1-13 TECHNICAL REPORT OF EFSA List of guidance, guidelines and working documents developed or in use by EFSA 1 Prepared by the Secretariat of the Scientific Committee

More information

European Union legislation on Food additives, Food enzymes, Extractions solvents and Food flavourings

European Union legislation on Food additives, Food enzymes, Extractions solvents and Food flavourings European Union legislation on Food additives, Food enzymes, Extractions solvents and Food flavourings European Commission, DG, Unit E3 Chemicals, contaminants and pesticides Serbia-Screening meeting on

More information

Dietary Risk Assessment of Nitrates in Cyprus and the relevant uncertainties

Dietary Risk Assessment of Nitrates in Cyprus and the relevant uncertainties Dietary Risk Assessment of Nitrates in Cyprus and the relevant uncertainties Georgios Stavroulakis, Maria Christofidou, Maro Christodoulidou, Popi Nicolaidou Kanari Eleni Ioannou-Kakouri (ex-head of RA

More information

1 OJ L 354, , p OJ L 80, , p. 19.

1 OJ L 354, , p OJ L 80, , p. 19. Call for scientific and technical data on the permitted food additives sulphur dioxide (E 220), sodium sulphite (E 221), sodium bisulphite (E 222), sodium metabisulphite (E 223), potassium metabisulphite

More information

Foundations of Mixture Toxicology and Their Regulatory Implications

Foundations of Mixture Toxicology and Their Regulatory Implications BfR - Mehrfachrueckstaende von Pestiziden in Lebensmitteln Foundations of Mixture Toxicology and Their Regulatory Implications Andreas Kortenkamp Institute for the Environment, Brunel University London

More information

Official Journal of the European Union L 109/11

Official Journal of the European Union L 109/11 19.4.2008 Official Journal of the European Union L 109/11 COMMISSION REGULATION (EC) No 353/2008 of 18 April 2008 establishing implementing rules for applications for authorisation of health claims as

More information

COMMISSION IMPLEMENTING DECISION. of

COMMISSION IMPLEMENTING DECISION. of EUROPEAN COMMISSION Brussels, 4.7.2017 C(2017) 4462 final COMMISSION IMPLEMENTING DECISION of 4.7.2017 on the identification of bis(2-ethylhexyl) phthalate (DEHP), dibutyl phthalate (DBP), benzyl butyl

More information