Basis for Risk Assessment of Carcinogenic Substances

1
Basis for Risk Assessment of Carcinogenic
Substances
2
RISK ASSESSMENT OF CHEMICAL CARCINOGENS
HAZARD IDENTIFICATION
HAZARD CHARACTERISATION
EXPOSURE ASSESSMENT
RISK CHARACTERISATION
3

• Some chemicals increase cancer incidences in
  animal studies by non-genotoxic mechanisms
• Such mechanisms are likely to show thresholds in
  their dose-response relationships
• Non-genotoxic chemicals could be allocated an
  ADI based on the NOAEL for the effect leading to
  cancer

4
The interaction between DNA and chemicals that
are both carcinogenic and genotoxic (by covalent
binding and mutagenicity DNA reactive) is
believed to be stochastic i.e. governed by the
laws of probability
This means that if the exposure is doubled, then
the risk is doubled
Theoretically, only zero exposure is associated
with zero risk.
Therefore chemicals that are both carcinogenic
and genotoxic are not allocated an ADI using the
usual NOAEL and uncertainty factor approach
5
What advice can the risk assessor give to the
risk manager for chemicals that are both
carcinogenic and genotoxic?
Risk characterization for genotoxic carcinogens
6
I shall now consider briefly the strengths and
weaknesses of each approach assuming
1. The chemical has been shown to be carcinogenic
in one or more animal cancer bioassays performed
to acceptable standards
2. The chemical has been shown to be genotoxic by
a mechanism involving direct interaction with DNA
(via DNA binding) and which is may not show a
biological threshold
7
ALARA
8
Low-dose Extrapolation
Requires a mathematical model to be fitted to the
data in the observed range

• Uses a mathematical model to extrapolate from the
  observed range down to
• actual human exposure levels or
• the exposure giving a predetermined low level of
  risk

The mathematical model used for extrapolation
could be that fitted to the data in the observed
range, BUT is normally linear in the low dose
range
The most commonly applied model is linear
extrapolation from a point near the bottom end of
the observed dose-response range the point of
departure or POD
9
BMD10
BMDL10
10
(No Transcript)
11
(No Transcript)
12
Is the dose-response relationship likely to be
linear over the range of extrapolation from an
incidence of 1 in 10 to 1 in 1,000,000?
Cancer arises as a result of a complex series of
chemical-biological interactions
The reaction of the carcinogen or its active
metabolite with DNA to form adducts is
essentially linear over a very wide range of
intakes but DNA binding does not equal cancer
Various models for the development of cancers
have been proposed
13
From Lutz WK and Kopp-Schneider A (1999)
Toxicol. Sci., 49, 110-115.
14
From - Williams GM (2001) Toxicol., 166, 3-10
15
From Schulte-Hermann R, et al. (2000) Mutat.
Res., 464, 13-18
16
From Swenberg JA, et al. (1995) Toxicol.
Lett., 82/83, 751-756
17
From Lutz WK (2001) Genetic toxicology and
cancer risk assessment. Ed. Choy WN, Marcel
Dekker pp249-271
18
From Lutz WK (1999) Hum. Exp. Toxicol., 18,
707-712
19
NL Non-linear at high exposures
NL
Overall relationship between exposure and cancer
incidence is unlikely to be linear from an
incidence of 1 in 10 to 1 in 1,000,000
NL
Each step shown NL is likely to show species
differences and human variability
NL
NL
NL
20
Species differences
Species differences in toxicokinetics can be
taken into account by a PBPK model
Some PBPK models estimate the target organ
exposure to the reactive genotoxic species by
including local tissue activation and
detoxication and the external dose-response curve
can be plotted as a human equivalent internal
dose-response curve
Data on species differences are frequently
ignored due to lack of suitable data and
therefore the mathematical extrapolation predicts
the incidence at low doses in the test species
Species differences in toxicodynamics are never
taken into account (adequately)
21
(No Transcript)
22
Human variability
Human variability in sensitivity may arise from
toxicokinetic or toxicodynamic differences
Differences may arise from environmental or
genetic factors
Variability may be unimodal or polymorphic
From Vineis P (2004) Oncogene, 23, 6477-6483
23
Human variability
Human variability is not taken into account by
the linear extrapolation model - it is assumed
that the use of a linear model is so conservative
that it does not matter!
Individuals who respond when there is a high
incidence such as 50 (TD50), 25 (T25) or 10
(BMD10) will be a mixture of both sensitive and
insensitive subjects
As the exposure is reduced so fewer and fewer
subjects will respond individuals who respond
at very low exposures will be those with the
greatest sensitivity.
Therefore human variability in sensitivity must
influence the slope of the curve
24
Cancer in both sensitive and insensitive subjects
Cancer in sensitive subjects
25
(No Transcript)
26
Assuming coefficient of variation of 50
(log-normal distribution of sensitivity)
27
In reality, genetic polymorphisms of
low-penetrant genes appear to give a maximum of
2-3 fold greater odds ratios for cancer
28
Conclusion

• Linear extrapolation from an incidence of 1 in 10
  to 1 in 1,000,000 does not reflect
• Inherent non-linearities (and possible
  thresholds) in the underlying biology of the
  numerous sequential steps
• Interspecies differences in the underlying
  biology of the numerous sequential steps
• Human variability in the underlying biology of
  the numerous sequential steps

29
Margin of Exposure
30
Margin of Exposure
MOE intake giving a predetermined incidence of
cancer human intake
Predetermined level of response incidence of
cancer
Could be the NOAEL or the LOAEL - but these will
vary from study to study as well as compound to
compound
Application of a consistent method of
dose-response analysis allows valid comparisons
across compounds
Level of response should be close to the bottom
end of the observed range e.g. T25 or BMD10 or
BMDL10
31
Human intake

• Should be that relevant to the compound under
  assessment
• Population average intake
• Average intake in consumers
• Intakes by high consumers
• A number of MOE values could be calculated which
  would give more useful advice for the risk manager

32
Interpretation of the calculated MOE
The MOE has to allow for the uncertainties and
variability discussed above
There is no MOE that gives assurance of zero risk
(apart from infinity)
Existing advice on interpretation of the
magnitude of MOE 75-300 a human health
concern (acrylamide JECFA) 3,800 is of
concern (ethyl carbamate JECFA) 20,000 of
low concern (ethyl carbamate
JECFA) 10,000-25,000 of low concern (PAHs
JECFA) 10,000 or more low priority (EFSA draft
opinion) gt500,000 low priority (Health
Canada) 50,000 used by ANZ to produce a
guideline value for soil contaminants
33
Decision Tree for the Risk Characterization of
Chemicals that are Carcinogenic and Genotoxic
34
Decision Tree for the Risk Characterization of
Chemicals that are Carcinogenic and Genotoxic
Is the chemical carcinogenic and genotoxic?
You have not been listening!
Should advice take into account human exposure
and potency?
Use ALARA
Apply for full membership of NAPPRAS
Nero and Pontius Pilate Risk Assessment Society
35
(No Transcript)
36
Should a mathematical model be fitted to the
animal cancer dose-response data in the observed
range (if appropriate)
Why not?
Apply for associate membership of NAPPRAS
Nero and Pontius Pilate Risk Assessment Society
37
(No Transcript)
38
(No Transcript)
39
Should the dose-response data be extrapolated to
a low or negligible risk - such as 1 in 1,000,000?

• Does the mathematical model take into account -
• Species differences in toxicokinetics and DNA
  repair?
• Human variability in toxicokinetics and DNA
  repair?
• Inherent non-linearities in the biological
  processes?

Contact Nobel Prize Committee
Do you believe the results?
(or found a political party)
Why bother?
Why?
40
Should the dose-response data be extrapolated to
a low or negligible risk - such as 1 in 1,000,000?

• Calculate Margin of Exposure (MOE) using
• a defined point on the cancer dose-response
  curve such as the BMD10, BMDL10, T25
• human intakes reflecting different exposure
  scenarios

41
Should advice to the risk manager describe of the
adequacy of the estimated MOE to allow for the
variabilities and uncertainties associated with
the use of the animal data?
Should a general MOE indicative of a low health
concern and a low priority for risk management
action be determined?
Apply for JECFA membership
Apply for EFSA Panel membership
42
(No Transcript)
43
Should a general MOE indicative of a low health
concern and a low priority for risk management
action be determined?
Apply for JECFA membership
Apply for EFSA Panel membership
44
Decision Tree for the Risk Characterization of
Chemicals that are Carcinogenic and Genotoxic