Food and Chemical Toxicology xxx (2013) xxx–xxx Contents lists available at ScienceDirect Food and Chemical Toxicology journal homepage: www.elsevier.com/locate/foodchemtox Selection of appropriate tumour data sets for Benchmark Dose Modelling (BMD) and derivation of a Margin of Exposure (MoE) for substances that are genotoxic and carcinogenic: Considerations of biological relevance of tumour type, data quality and uncertainty assessment Lutz Edler a, Andy Hart b, Peter Greaves c, Philip Carthew d, Myriam Coulet e, Alan Boobis f, ⇑ Gary M. Williams g, Benjamin Smith h, a German Cancer Research Centre (DKFZ), Im Neuenheimer Feld 280, 69120 Heidelberg, Germany b The Food and Environment Research Agency – FERA, Sand Hutton, YO41 1LZ York, United Kingdom c Department of Cancer Studies and Molecular Medicine, University of Leicester, LE2 7LX Leicester, United Kingdom d Unilever, Colworth House Sharnbrook, MK44 1LQ Bedfordshire, United Kingdom e Nestlé Research Centre, Vers-Chez-Les-Blanc, 1000 Lausanne, Switzerland f Imperial College, Hammersmith Campus, Ducane Road, W12 0NN London, United Kingdom g New York Medical College, Basic Science Building, Room 413, Valhalla, NY 10595, United States h Firmenich, Rue de la Bergere 7, 1217-Meyrin 2, Switzerland article info abstract Article history: This article addresses a number of concepts related to the selection and modelling of carcinogenicity data Available online xxxx for the calculation of a Margin of Exposure. It follows up on the recommendations put forward by the International Life Sciences Institute – European branch in 2010 on the application of the Margin of Expo- Keywords: sure (MoE) approach to substances in food that are genotoxic and carcinogenic. The aims are to provide Genotoxic carcinogen practical guidance on the relevance of animal tumour data for human carcinogenic hazard assessment, Tumour relevance appropriate selection of tumour data for Benchmark Dose Modelling, and approaches for dealing with Data selection the uncertainty associated with the selection of data for modelling and, consequently, the derived Point Benchmark Dose Modelling of Departure (PoD) used to calculate the MoE. Although the concepts outlined in this article are interre- Margin of Exposure Uncertainty assessment lated, the background expertise needed to address each topic varies. For instance, the expertise needed to make a judgement on biological relevance of a specific tumour type is clearly different to that needed to determine the statistical uncertainty around the data used for modelling a benchmark dose. As such, each topic is dealt with separately to allow those with specialised knowledge to target key areas of guidance and provide a more in-depth discussion on each subject for those new to the concept of the Margin of Exposure approach. Ó 2013 ILSI Europe. Published by Elsevier Ltd. All rights reserved. Abbreviations: AIC, Akaike Information Criterion; BMD, Benchmark Dose; BMDL, 1. Introduction Lower Benchmark Dose; BMDU, Upper Benchmark Dose; BMR, Benchmark Response; COC, Committee On Carcinogenicity; ECHA, European Chemicals Agency; In recent years, the Margin of Exposure (MoE) has been pro- EFSA, European Food Safety Authority; EPA, Environmental Protection Agency; EPRI, European Parliaments Research Initiative; FDA, Food and Drug Administration; posed as an alternative to low dose extrapolation as a means of IARC, International Agency for Research on Cancer; IPCC, Intergovernmental Panel providing advice to risk managers on the potential level of concern on Climate Change; IPCS, International Programme on Chemical Safety; MoA, Mode from exposure to chemicals that are genotoxic and carcinogenic of Action; MoE, Margin of Exposure; NCI, National Cancer Institute; NIOSH, National and to help prioritise risk management actions (Benford et al., Institute for Occupational Safety and Health; NOAEL, Not Observed Adverse Effect 2010; JECFA, 2005). The MoE is expressed as the ratio between Level; NTP, National Toxicology Program; OECD, Organization for Economic Cooperation and Development; PoD, Point of Departure; PoD/RP, Point of Depar- an appropriate Point of Departure (PoD) on the dose–response ture/Reference Point; WHO, World Health Organisation. curve for a tumour response and a relevant estimate of human ⇑ Corresponding author. Address: ILSI Europe, Avenue E. Mounier 83, Box 6, 1200 exposure. In 2010, a special supplement volume was published Brussels, Belgium. Tel.: +32 2 771 00 14. in Food and Chemical Toxicology (Benford et al., 2010) entitled E-mail addresses: [email protected] (L. Edler), [email protected] (A. Hart), ‘‘Application of the Margin of Exposure (MoE) Approach to Sub- [email protected] (P. Greaves), [email protected] (P. Carthew), [email protected] (M. Coulet), [email protected] (A. Boobis), stances in Food that are Genotoxic and Carcinogenic’’. In 12 case [email protected] (G.M. Williams), benjamin.smith@firmenich.com, studies, different chemicals found in food and acknowledged to [email protected] (B. Smith). 0278-6915/$ - see front matter Ó 2013 ILSI Europe. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.fct.2013.10.030 Please cite this article in press as: Edler, L., et al. Selection of appropriate tumour data sets for Benchmark Dose Modelling (BMD) and derivation of a Margin of Exposure (MoE) for substances that are genotoxic and carcinogenic: Considerations of biological relevance of tumour type, data quality and uncertainty assessment. Food Chem. Toxicol. (2013), http://dx.doi.org/10.1016/j.fct.2013.10.030 2 L. Edler et al. / Food and Chemical Toxicology xxx (2013) xxx–xxx be acting via a direct DNA-reactive mode of action were considered ucts and the Environment (UK Committee on Mutagenicity of in order to provide guidance on how to apply the MoE approach to Chemicals in Food, 2000) and the European Food Safety Authority genotoxic and carcinogenic substances. A key conclusion, based on Scientific Committee’s opinion on genotoxicity testing strategies this work, was that ‘‘depending on the tumour endpoint that is se- applicable to food and feed safety assessment (EFSA, 2011a)do lected to be used in this approach and the ways in which the data provide solid guidance on appropriate tiered testing strategies for are analysed, it is possible to generate very different PoDs to be identifying genotoxic substances, including assessment of their di- used in calculating the MoE, and, hence, in the value of the MoE it- rect DNA-reactivity. self. It is, therefore, essential that the selection of the cancer end- In addition to establishing a direct DNA-reactive MoA, consider- point and mathematical treatment of the data are clearly ation of the biological relevance of animal tumours to human can- described and justified if the results of the MoE approach are to cer risk is key in the selection of tumour data for Benchmark Dose be trusted and to be of value to risk managers’’ (Benford et al., Modelling. This utilises what is commonly referred to as the ‘‘hu- 2010). man relevance framework approach’’ (Boobis et al., 2006), which In the article by Benford et al. (2010), the PoD, or reference has evolved over the years from a simple consideration of the mor- point, used for calculation of the MoE is in the lower 95% confi- phological and biological comparisons of rodent and human tu- dence limit (BMDL) of the Benchmark Dose (BMD) producing the mours to encompass a greater understanding of toxicity specified response (Benchmark Response or BMR) in a study in pathways and MoAs relevant to human carcinogenesis. Both the experimental animals. For substances that are both genotoxic genotoxic MoA and the relevance to humans of an animal tumour and carcinogenic, the BMR is typically taken as a 10% increase in response should be considered when selecting appropriate data the incidence of tumours above the modelled background. The sets for modelling and derivation of the BMDL. BMDL is used to establish a PoD that assures high (95%) confidence Since the selection of data requires expert opinion regarding that the specified response will not be greater at that dose level both the genotoxic MoA and relevance to human carcinogenesis, (EFSA, 2009b). As pointed out by Benford and co-authors, however, uncertainty enters into the choice of any data used to derive the the BMDL is heavily dependent on the risk assessors’ judgement as BMDL. This uncertainty needs to be characterised by the risk asses- to the selection and treatment of the data modelled and on how sor in such a way that it is transparent and can be taken into ac- the models are selected and employed. To address this issue, and count by the risk manager (Codex Alimentarius Commission, further build on the framework for applying the MoE approach to 2011; EFSA, 2009b, 2011a). In the case of data selection for deriv- substances that are genotoxic and carcinogenic, a new expert ing a MoE for a genotoxic carcinogen, the source of uncertainty is group was formed by ILSI Europe. The task of this group was to de- reflected in two key decisions: velop practical guidance for risk assessors on the factors to con- sider when selecting appropriate tumour data sets for – Decisions on which tumours/studies should be considered for Benchmark Dose Modelling and the subsequent derivation of a deriving the MOE. MoE. Specifically, the expert group focussed on the relevance of – Decisions on the value of the PoD to be selected from the animal tumour data for human carcinogenic risk assessment, tumours/studies. appropriate selection of tumour data for Benchmark Dose Model- ling, and approaches for dealing with the uncertainty associated In the first of these, the uncertainties relate to a categorical with the selection of data for modelling and, consequently, the de- question to include or exclude a particular study or tumour type. rived PoD used to calculate the MoE. In the second, the uncertainty relates to a quantitative question, A key consideration before deriving the MoE of any substance is in particular, what value to take for the PoD.