EFSA Margin of Exposure Approach

1. EFSAMargin of Exposure Approach Susan Barlow EFSA Scientific Committee

2. EFSA OPINION

3. BACKGROUND TO THE EFSA OPINION • What are the possible approachesfor risk assessment of genotoxic carcinogens ingested in low amounts? • How should the outcomes of the various approaches be interpretedin terms of risks to human health? • To what extent do the approaches available meet the needs of risk managers? • e.g. Do they give practical optionsin situations where exposure cannot be completely eliminated and the magnitude of risk cannot be readily determined? • Do they enable prioritiesfor risk management actions to be set?

4. WHAT ARE THE POSSIBLE APPROACHES? • ALARA – as low as reasonably achievable • TTC – the threshold of toxicological concern • Dose-response modelling/Numerical calculation of range of risk – by linear low-dose extrapolation • Dose-response modelling/MOE – margin of exposure

5. ALARA • Needs only hazard identification data (animal or human) • Widely used • Scientifically supportable • Clearly a protective option BUT • Does not make use of all the data (e.g. potency, exposure) • Not helpful for gauging magnitude of risks in relation to exposures (urgency, extent of RM measures needed) • Not helpful for prioritising risks to be managed

6. TTC • Needs no toxicological data • Does need good exposure data (or worst case scenario) • Coming into use for very low level contaminants BUT • Very low TTC value for potentially genotoxic substances (0.15µg/day) • Not suitable for extremely potent carcinogens (e.g. aflatoxins, azoxy, nitrosamines)

7. LINEAR EXTRAPOLATION • Provides quantitative upper bound estimate of risk (very conservative) • Widely used BUT • How uncertain are the estimates? • Estimates very model-dependent • Can be misunderstood and misused

8. MOE • Pragmatic approach that uses both potency and human exposure data • Does not extrapolate below range of observations • Can be complementary to other approaches • Useful as one input to the whole RA process BUT ......

9. DERIVING THE MOE • Select cancer data set(s) to be used • Traditional approach using NOAELs not appropriate for DNA-reactive substances (cannot infer thresholds) • Use mathematical and statistical modelling to fit best curve to the data in the observed range • Define a benchmark response (e.g. 10% increase in tumour incidence compared to background) and derive corresponding benchmark dose (e.g. BMD10) • To take account of uncertainties in BMD, use lower 95% confidence interval on BMD as reference point or point of departure (e.g. BMDL10)

10. DERIVING THE MOE: BMD APPROACH BMD10 10% Response = BMR Model fitted to data points BMDL10 Lower 95% confidence interval on dose giving a 10% response

11. CALCULATING THE MOE BMDL10 MOE = ______________ Human exposure

12. RECOMMENDATIONS FROM 2005 EFSA/WHO/ILSI CONFERENCE • Need for criteria to assess the quality/adequacy of data for MOE approach(EFSA) • Guidance on dose-response modelling, derivation of Reference Point (WHO, EFSA) • Guidance on explanation and communication of uncertainties (EFSA) • Consider ranges of MOE rather than single values • Consider how to band MOEs with respect to levels of concern (to reduce misinterpretation) • Develop example case studies on specific substances (ILSI)

13. Published papers from 2005 EFSA/WHO/ILSI Conference

14. APPLICATION OF THE MOE APPROACH • BY JECFA • Acrylamide (64th M February 2005) • 1,3-dichloro-2-propanol (67th M June 2006) • Ethyl carbamate (64th M February 2005) • PAHs (64th M February 2005) • By EFSA • Aflatoxins (CONTAM Panel January 2007) • Ethyl carbamate (CONTAM Panel Sept 2007) • PAHs (CONTAM Panel June 2008)

15. USE OF MOE APPROACH BY EFSA Ethyl carbamate in food and beverages • Animal cancer bioassay data • Reference point: BMDL10 • 3 exposure scenarios • Whole population • Consumers of alcoholic beverages • 95th percentile of consumers of alcoholic beverages MoE=18 000MoE=5 000MoE=600

16. USE OF MOE APPROACH BY EFSA Polycyclic Aromatic Hydrocarbons in Food • Type of data: 2-year carcinogenicity study on coal tar mixtures • Reference point: BMDL10 • Mean and 97.5th percentile exposure to benzo-α-pyrene, PAH2, PAH4 & PAH8

17. USE OF MOE APPROACH BY EFSA Aflatoxins in almonds, hazelnuts, pistachios and derived products • Type of data: - rat carcinogenicity study - human epi data (chronic hepatitis B) • Reference point: - BMDL10 A (animal data) - BMDL10 H and BMDL1 H (human data) ALARA recommended

18. JECFA COMPARISON OF SUBSTANCES Daily intake MOE Level of concern Acrylamide 1 g/kg mean 300 of concern 4 g/kg high 75 of concern Ethylcarbamate 15 ng/kg (background) 20,000 low 80 ng/kg (+ alcoholic bev) 3,800 of concern PAHs (benzo[a]pyrene as marker) 4 ng/kg mean 25,000 low 10 ng/kg high 10,000 low DCP 51 ng/kg mean 65,000 low 136 ng/kg high 24,000 low

19. LIMITATIONS of MOE APPROACH • MOE is a ratio – it does not quantify the risk • Requires good data on food intake and chemical concentration • Risk of misinterpretation – context and characterisation of uncertainties is needed • Assignment of ‘concern levels’ (banding of MOE values) needs further discussion between RAs and RMs

20. CONCLUSIONS OF EFSA SC 2005 • MOE can be used as a harmonised approach for the RA of genotoxic carcinogens • Can be used to rank and compare substances • Provides RM with a basis for setting priorities for action (and does not exclude applying ALARA) • Use the BMD approach to derive the Reference Point (BMDL10) (If data unsuitable, use T25) • Communicate all assumptions and uncertainties • MOE of 10,000 or higher considered as of low concern