Development of a New Tool for Modelling Potential Risks from Food ...
Development of a New Tool for Modelling Potential Risks from Food ...
Development of a New Tool for Modelling Potential Risks from Food ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Development</strong> <strong>of</strong> a <strong>New</strong> <strong>Tool</strong> <strong>for</strong><br />
<strong>Modelling</strong> <strong>Potential</strong> <strong>Risks</strong> <strong>from</strong><br />
<strong>Food</strong> Contact Materials<br />
P.S. Price MS, Dr. J.M. Zabik, G.M. Wiltshire, and<br />
Dr. H.M. Hollnagel<br />
The Dow Chemical Company<br />
ILSI 4 th International Symposium on<br />
<strong>Food</strong> Packaging<br />
Prague, Czech Republic<br />
1
Project Goals<br />
• Characterize risks posed by mixtures <strong>of</strong> migrants rather than single<br />
substances<br />
• Create a screening tool <strong>for</strong> non-carcinogenic effects to identify<br />
mixtures that have a low potential <strong>for</strong> toxicity<br />
• Develop a tool that:<br />
– Can be applied to a wide range <strong>of</strong> mixtures;<br />
– Does not require toxicity data on mixture components; but<br />
– Can make effective use <strong>of</strong> toxicity data when available<br />
• Build on work done in the Threshold <strong>of</strong> Toxicological Concern (TTC)<br />
• Use the latest guidance on assessing chemical toxicity (ECHA and<br />
other sources)<br />
2
Challenges Posed by Mixtures<br />
• Complexity:<br />
– Infinite number <strong>of</strong> mixtures (can’t test them all)<br />
• Little or no toxicity data <strong>for</strong> many components <strong>of</strong> mixtures.<br />
• Traditional approaches <strong>for</strong> assessing mixture toxicity<br />
– Additive models too conservative (not all effects add);<br />
– Independence models not sufficiently conservative (some might<br />
add);<br />
– Both approaches require toxicity data on all components<br />
3
Components <strong>of</strong> <strong>Tool</strong><br />
• Use existing models <strong>of</strong> mixture toxicity<br />
– To develop estimates <strong>of</strong> safe exposures under different<br />
assumptions<br />
– To identify the mixture components that drive toxicity<br />
• Use Cramer classes to fill data gaps<br />
• Monte Carlo modeling <strong>of</strong> uncertainty in estimates<br />
4
Limitations to the <strong>Tool</strong><br />
• Does not address carcinogenic / genotoxic effects<br />
– A separate assessment <strong>of</strong> carcinogenic effects is required<br />
• The approach requires in<strong>for</strong>mation on the structure <strong>of</strong> the compounds<br />
– To assign the compounds to a Cramer class<br />
– To confirm that the Munro et al. 1996 data set has similar compounds<br />
– To determine that there are no structural alerts <strong>for</strong> carcinogenicity and<br />
that they are not organophosphorous compounds<br />
• Synergy / Potentiation<br />
– Not modeled directly<br />
– Proposed approach minimizes the potential <strong>for</strong> synergistic effects by<br />
restricting exposures to mixture components to levels where synergy is<br />
unlikely to occur (Konemann and Pieters, 1996)<br />
5
Models <strong>for</strong> Mixture Toxicity<br />
Derived No Effect Level (DNEL) <strong>for</strong> single compound:<br />
DNEL =<br />
POD<br />
AF A AF H<br />
POD = Point <strong>of</strong> departure (NOAEL, NOEL, LOAEL, BMD, etc.)<br />
AF = Adjustment Factor (interspecies or interindividual uncertainty)<br />
DNEL <strong>for</strong> mixture assuming additivity:<br />
Mixture DNEL =<br />
-1 -1<br />
POD<br />
Σ F i<br />
i<br />
AF Ai<br />
AF Hi<br />
F i = Weight fraction <strong>of</strong> the i th chemical in the mixture<br />
6
Models <strong>for</strong> Mixture Toxicity (Cont.)<br />
DNEL <strong>for</strong> mixture assuming independence:<br />
Mixture DNEL =<br />
Min<br />
POD i<br />
AF Ai AF Hi<br />
F i<br />
Contributions to Additive models <strong>of</strong> mixture toxicity can be<br />
ranked based on:<br />
Toxicity<br />
Weight <strong>of</strong><br />
Compound<br />
=<br />
F i<br />
AF Ai AF Hi<br />
POD i<br />
Mixture components with the highest weights will drive the<br />
mixture’s toxicity and can be investigated further.<br />
7
Using the Munro et al. Analysis <strong>of</strong><br />
Cramer Classes to Fill Data Gaps<br />
• Cramer 1 classes associate structural properties with<br />
toxicological potency<br />
• Assignment to the three Cramer classes is made based on<br />
structure<br />
• Use the lower 5 th percentile <strong>of</strong> the distribution or NOELs<br />
as the POD<br />
• Use 100 fold adjustment factor proposed by Munro et al.<br />
1996 2 .<br />
1<br />
Cramer et al. 1978 <strong>Food</strong> Cosmetic Toxicology Vol. 16. pp.255-278<br />
2<br />
Munroe et al. 1996 <strong>Food</strong> and Chemical Toxicology Vol 34. pp 829-867<br />
8
The Cramer Class Approach<br />
Class<br />
II<br />
Class<br />
I<br />
5%ile NOEL<br />
(mg/kg/day)<br />
3.0<br />
Class<br />
III<br />
Class I<br />
II<br />
0.91<br />
III<br />
0.15<br />
9
Probabilistic Non-cancer Risk<br />
Assessment<br />
• Concept first raised in the early 1990s<br />
• Groups that have published in this area<br />
– RIVM, TNO, FoBiG, and BAuG<br />
– U.S. EPA<br />
– Harvard Center <strong>for</strong> Risk Analysis<br />
• Cited in ECHA guidance <strong>for</strong> REACH<br />
• Benefits<br />
– Allows the investigation <strong>of</strong> the impact <strong>of</strong> making multiple conservative<br />
assumptions <strong>for</strong> each component<br />
– The approach gives:<br />
• A single value (lower 5 th percentile)<br />
• Complete distribution to provide a context to the standard<br />
10
Why Per<strong>for</strong>m a Probabilistic<br />
Analysis?<br />
• The use <strong>of</strong> the Cramer classes/Munro et al. approach <strong>for</strong> a<br />
single chemical gives a conservative assumption <strong>of</strong><br />
toxicity<br />
• Applying the same approach to multiple chemicals can<br />
result in an over-estimation<br />
– It is unlikely that all chemicals in a mixture will be as toxic as the<br />
5 th percentile<br />
• The conservative adjustment factors have a similar<br />
problem<br />
11
Probabilistic Noncancer Risk<br />
Assessment<br />
• Uses Monte Carlo analysis<br />
• Based on the same equations<br />
• Requires uncertainty distributions <strong>for</strong> POD and each<br />
adjustment factor<br />
• Currently ECHA guidance on DNELs is most closely<br />
aligned to published probabilistic assessments<br />
12
Probabilistic Additive Model <strong>of</strong><br />
Mixture Risk<br />
-1 -1<br />
Uncertainty<br />
distribution <strong>of</strong><br />
NOAEL in<br />
Sensitive Humans<br />
=<br />
Σ F i<br />
AF Ai<br />
POD i<br />
AF Hi<br />
Fifth percentile<br />
provides a basis <strong>for</strong> a<br />
point estimate <strong>of</strong> the<br />
Mixture DNEL<br />
13
Application <strong>of</strong> the <strong>Tool</strong> to <strong>Food</strong><br />
Contact Materials<br />
• Chemicals detected in water stored in<br />
polypropylene bottles (Skjevrak et al. 2005)<br />
• Chemicals identified and concentrations<br />
measured<br />
– 72 hr duration ambient temperatures<br />
– Highest values <strong>from</strong> multiple bottles reported in<br />
paper<br />
14
Residue Data<br />
Compound<br />
Concentration in Water ug/l<br />
Di isobutyl phthalate 36<br />
Dibutyl phthalate 9<br />
Ethyl-4-Ethoxybenzoate 101<br />
2,4-di-tert-butylphenol 25<br />
Ethylbenzoate 15<br />
4-Methylbenzaldehyde 37<br />
Toxicity Data<br />
Compound Cramer Class Safe Level <strong>of</strong> Intake Reference<br />
Di isobutyl phthalate 1<br />
Dibutyl phthalate 1 1 EPA IRIS 1990<br />
Ethyl-4-Ethoxybenzoate 2<br />
2,4-di-tert-butylphenol 1<br />
Ethylbenzoate 1<br />
4-Methylbenzaldehyde 1<br />
15
Revised Toxicity Data<br />
Compound<br />
Estimate <strong>of</strong> Safe Dose mg/kg/d<br />
Di isobutyl phthalate 0.03<br />
Dibutyl phthalate 0.2<br />
Ethyl-4-Ethoxybenzoate 0.0091<br />
2,4-di-tert-butylphenol 0.03<br />
Ethylbenzoate 0.03<br />
4-Methylbenzaldehyde 0.03<br />
16
Results<br />
0.05<br />
Safe Dose mg/kg/d<br />
0.04<br />
0.03<br />
0.02<br />
0.01<br />
0.00<br />
Additive Independence MC Additive MC Independence<br />
Estimates <strong>of</strong> Mixture Toxicity<br />
• Estimates are similar <strong>for</strong> independence and additive models – one<br />
compound drives the mixture’s toxicity<br />
• Probabilistic models give values 2-3 fold higher than deterministic<br />
17
Drivers <strong>of</strong> Mixture Toxicity<br />
• One compound, Ethyl-4-Ethoxybenzoate, drives toxicity<br />
Compound<br />
Toxicity<br />
Contribution<br />
Weight<br />
Di isobutyl phthalate 5.4<br />
Dibutyl phthalate 0.2<br />
Ethyl-4-Ethoxybenzoate 49.8<br />
2,4-di-tert-butylphenol 3.7<br />
Ethylbenzoate 2.2<br />
4-Methylbenzaldehyde 5.5<br />
Toxicity Driver<br />
• Additional data on this compound could have a major<br />
impact on the estimate <strong>of</strong> mixture toxicity<br />
• Removal <strong>of</strong> this compound would raise the estimate <strong>of</strong> a<br />
safe dose <strong>of</strong> the mixture 4 fold!<br />
18
Estimates <strong>of</strong> Safe Doses in Bottled Water<br />
• Assuming 2 l/d intake and 60 kg body weight, the estimates <strong>of</strong> safe levels <strong>of</strong> the<br />
mixture in the water range <strong>from</strong> 0.5 – 1.5 mg/l<br />
• The highest level <strong>of</strong> the mixture found in the study is less than 0.25 mg/l<br />
3.00<br />
Water Conc. Of Mixture mg/l<br />
2.50<br />
2.00<br />
1.50<br />
1.00<br />
0.50<br />
Estimates <strong>of</strong> Safe Level <strong>of</strong> Mixture in Water<br />
0.00<br />
Additive Independence MC Additive MC<br />
Independence<br />
Higest<br />
Reported<br />
Level <strong>of</strong><br />
Mixture<br />
There<strong>for</strong>e, total toxicity <strong>of</strong> the migrants at the concentrations reported in<br />
this study does not pose a risk <strong>for</strong> individuals consuming water <strong>from</strong> the<br />
polypropylene bottles.<br />
19
Summary<br />
• The approach is allows estimates to be made on any<br />
mixture <strong>of</strong> migrants where:<br />
– the relative amounts, and<br />
– structures <strong>of</strong> the migrants are known;<br />
• Identifies the migrants that drive mixtures’ toxicities<br />
– Provides guidance <strong>for</strong> risk management <strong>of</strong> migrants<br />
• Estimates take advantage <strong>of</strong> toxicity data when available<br />
• Probabilistic approaches show that deterministic values<br />
overestimated toxicity by a factor <strong>of</strong> 2-3 in this example<br />
20
Questions<br />
21