Default Assumptions

1
Past and Future Use of Default Assumptionsand
Uncertainty Factors
Dr. Michael Dourson Toxicology Excellence for
Risk Assessment (TERA)

• Default Assumptions
• Misunderstandings
• New Concepts

2
Default Assumptions
Adverse effects are either biochemical change,
functional impairment, or pathologic lesion which
impairs performance and reduces the ability of an
organism to respond to additional challenge.

• An adaptive effect enhances an organism's
  performance as a whole and/ or its ability to
  withstand a challenge. An increase in hepatic
  smooth endoplasmic reticulum is an example of an
  adaptive effect, if hepatic metabolism reduces
  the chemical's toxicity.

A compensatory effect maintains overall function
without enhancement or significant cost.
Increased respiration due to metabolic acidosis
is an example of a compensatory effect.
The critical effect is the first adverse effect
or its known precursor that occur as dose rate
increases.
Severity is the degree to which an effect changes
and impairs the functional capacity of an organ
system
EPA, IRIS Glossary Haber et al., 2001
3
Default Assumptions

• The estimation of these Safe doses involves
  several judgments
• such as...
• the choice of the most appropriate No Observed
  Adverse Effect Level (NOAEL) or Benchmark Dose
  (BMD) of the critical effect, usually from
  experimental animal data, and
• the choice of the appropriate Uncertainty
  Factors based on a review of the entire database.

EPA, multiple references
4
Default Assumptions
SLIGHT BODY WEIGHT DECREASE
FAT IN LIVER CELLS
(CRITICAL EFFECT)
CONVULSIONS
UF x MF
ENZYME CHANGE
RfD
NOEL
NOAEL
FEL
LOAEL
Dose
EPA, multiple references
5
Default Assumption Using Uncertainty Factors
Meek et al., 1994 IPCS, 1994, 2001 Rademaker
and Linders, 1994 Pohl and Abdin, 1995 EPA
multiple references
6
As Defined by EPA a ReferenceDose (RfD) Is

• ...an estimate (with uncertainty spanning perhaps
  an order of magnitude) of
• 
  
• a daily exposure to the human population
  (including sensitive subgroups)
• that is likely to be without an appreciable risk
  of deleterious effects during a lifetime.

EPA, multiple references
7
Default Assumptions

• The resulting range of an RfD has been defined
  as "perhaps an order of magnitude." This range
  is expected due to the imprecision of uncertainty
  factors.
• Thus, environmental exposures falling into the
  range of the subthreshold estimate generally
  cannot be scientifically distinguished from the
  estimate.

Felter and Dourson, 1998
8
Assumptions, Strengths and Limitations

• Major assumptions
• population threshold exists
• estimates represents subthreshold doses
• prevent the critical effect protects all
• Major strengths
• all data are reviewed for critical effect
• uncertainties addressed with factors based on
  judgment
• Major limitations
• NOAEL ignores many data
• uncertainty factors are imprecise
• risks above the RfD is not estimated

EPA, multiple references
9
Misunderstandings in Safe Dose Assessment

• Studies with small n are not useful
• studies with even one subject are important
• Uncertainty factors are arbitrary
• factors are imprecise
• An uncertainty factor of 10 is not enough
• analysis indicates that it is most often
  conservative
• Animal-based RfDs are habitually protective
• human-based RfDs doses are sometimes lower
• RfDs do not protect children
• Thats intent, but study design must be improved

10
Factor of 10 Enough?
Casarett and Doull, 6th Edition, page 19
11
Factor of 10 Enough?
(Dourson et al., 2002)
12
Factor of 10 Enough?
13
Factor of 10 Enough?
5b.
14
Factor of 10 Enough?
6
8
4
Dourson and Stara, 1983
10
2
12
Frequency
16
22
30
1.4
65
Intraspecies Adjustment Factor
Frequency vs an intraspecies adjustment factor
obtained by raising 10 to the power (3 standard
deviations the probit, log-dose slope).
Probit, log-dose slopes are shown within the
figure
15

Animal RfDs Habitually Protective?

• For 36, EPAs human-based safe doses are
  within 3-fold of animal-based values.
• For 41, EPAs human-based safe doses are more
  than 3-fold higher than animal-based values.
• For 23, EPAs human-based safe doses are more
  than 3-fold lower than animal-based values.
• An animal-based safe dose was not estimated 6
  out of 28 times, since data were judged
  insufficient or inappropriate.

(Dourson et al., 2001)
16
Animal RfDs Habitually Protective?
RfDs are the same
17
Do RfDs Protect Children?

• Does the uncertainty factor used for
  inter-human variability to toxicity protect
  children as well as adults?
• Is the uncertainty factor for database
  deficiencies, e.g., absence of kid-specific
  tests, adequate to protect children?
• Are these two factors together adequate to
  protect children?

18
(No Transcript)
19
(Dourson et al., 2002)
20
0.3
21
Interhuman Uncertainty Factor Protective of
Children?

• Protection of the children with the use of an
  uncertainty factor of either 3 or 10 is between
  67 and 100 of chemicals or population.
• Newborns or premature infants are more sensitive
  to the toxicity of parent compounds when compared
  to adults older children are more resistant.
• Protection can be as low as 60 of the population
  of individuals with severe disease.
• Studies in large populations suggest that near
  100 of population protection is achieved.

22
Database Uncertainty Factor Protective of
Children?
Animal lifespan to toxicity tests (time frames
are not to scale).
Organogenesis
Germ Cell
Birth
Sexual Maturity
Death
Weaning
Conception
2 Generation Reproduction
2 Year Chronic Bioassay
Developmental toxicity
(From Dourson et al., 2002)
23
(No Transcript)
24
Database Uncertainty Factor Protective?

• A factor of 3 or 10 protects for either 92 or 98
  of occurrences of lower NOAEL. Thus, a factor is
  needed when developmental or reproductive effects
  may be critical, but specific data are lacking.
• RfDs and RfCs should be based on these endpoints
  when specific data suggest they are critical, and
  no reduction in their value should be used for
  less than lifetime exposures.
• These conclusions are based on pesticides, but
  should be checked with data from other types of
  chemicals (Dourson et al., 1992).

25
Conclusions on Misunderstandings

• Studies with small numbers are still important
  study severity of effect as well as number.
• Uncertainty factors are imprecise, not arbitrary.
• The normal use of the uncertainty factor for
  human data results in protection of between 100
  and 1000-fold variation.
• Ignoring data is not protective of human health.
  Human and laboratory animal data together give
  the best picture of the overall toxicity of a
  chemical.

26
Conclusions on Misunderstandings continued

• Combined likelihood that the interhuman and
  database factors protect children is very
  probable.
• Use of an additional factor (e.g., FQPA) unlikely
  to provide greater protection to kids older than
  6 months adequate testing needed for earlier.
• Current toxicity testing protocols useful, but
  can be improved for young animals.
• Avoid misunderstandings. Scientists offering
  opinions should be queried as to their
  understanding.

27
New Concepts

• Over the last several years, scientists have
  begun using more data when choosing uncertainty
  factors
• Scientists use a number of approaches
• Methods range from default (presumed
  protective) to those incorporating more
  biological data (biologically-based protective)

Meek et al, 2001
28
New ConceptsExample of Default UF Hg Oral RfD
EPA, 2001
29
New Concepts Categorical Default

• Renwick (1993) proposed breaking the interspecies
  and intraspecies UFs into toxicokinetic (TK) and
  toxicodynamic (TD) components
• This approach was modified by World Health
  Organization- International Programme on Chemical
  Safety as follows

IPCS, 1994 2001
30
New Concepts Interindividual Variability in
Ingested Hg
a Corresponding to either 1ppm Hg in hair or 1ppb
in blood
NAS, 2001
31
New ConceptsExample of Compound Specific
Adjustment Factors (CSAF) Categorical Default
for Hg
Total UF 5.8 Methyl Hg RfD 2 E-4 mg/kg-day
Dourson et al., 2001
32
New Concepts Ingested Hg
NAS, 2001
Predicted mean probability of MeHg intake
corresponding to 11ppm MeHg in hair.
33
New Concepts Compare ingestion rates
(?g/kg/day) of Seychelles studies from either
deterministic approach or Monte Carlo approach
(ICF Kaiser. 1998)
34
New ConceptsRfD Based on a Monte Carlo
1st
2nd 3 UF for database
3rd
35
Summary of Hg Example

• Limited scientific data supports use of default
  uncertainty factor for intra-species variability
  of 10 (Hg RfD 1 E-4) (EPA, 2001).
• Newer methods on categorical defaults allow
  replacement of defaults with compound specific
  data (Hg RfD 2 E-4) (Dourson et al., 2001).
• Monte Carlo Analysis requires more data but
  allows probabilistic approach to RfD
  determination (1st percentile Hg RfD 3 E-4)
  (ICF Kaiser. 1998).

36
(No Transcript)
37
IPCS (2001) CSAF Guidance

• Choice of the appropriate endpoint must be
  critical effect based on understanding mode of
  action.
• Data for application of the framework must relate
  to the active form of the chemical.
• The metric for toxicokinetics or measure of
  effects for toxicodynamics needs careful
  consideration in relation to the delivery of
  chemical to target organ.
• Relevance of population, route of exposure,
  dose/concentration and adequacy of numbers of
  subjects/samples must be considered and impact on
  validity of calculated ratio addressed.

38
Conclusions On New Concepts

• Agencies are using other than 10-fold factors on
  a more regular basis based on data.
• Compound Specific Adjustment Factors (CSAFs) are
  justified when adequate and specific data exist.
• Monte Carlo methods are available should be
  used.
• As a result
• In developing subthreshold doses, the first
  choice should be use data to generate a
  distribution or CSAF a second choice would be
  the default factor.
• Use of distributions CSAF will lead to better
  data and fewer uncertainties.

39
Extra Slides
40
Animal Safe Doses Habitually Protective?
41
Based on Dourson et al., 1992
42
Benchmark Dose

• What is a Benchmark Dose (BMD)?
• The statistical lower confidence limit on the
  dose producing a predetermined level of change in
  an adverse effect compared with the response in
  untreated animals.
• In practice...
• The 95 lower confidence on the dose that
  causes, for example, a 10 increase in the number
  of animals developing fatty liver compared with
  untreated animals.
• A BMD is calculated by fitting a mathematical
  dose-response model to data.

43
Advantages of BMD Approach

• The BMD is not limited to the doses tested
  experimentally and is less dependent on dose
  spacing.
• The BMD takes into account the shape of the
  dose-response curve.
• The approach provides flexibility in determining
  biologically significant rates (e.g., a 10
  increase may be appropriate for one response
  while a 1 increase is appropriate for a
  different response). Alternatively, a different
  UF might be used to address this issue.
• Use of the BMD gives incentive to conduct better
  studies because more rigorous studies result in
  tighter uncertainty bands, and thus, higher BMDs.

44
Limitations to the BMD

• Possible to introduce error in the model
  prediction of BMD if the models are used to
  extrapolate to low doses without incorporating
  information on mechanism.
• Quantal data (e.g., tumor incidence or number of
  pups with a deformity) and continuous data (e.g.,
  changes in body/organ weight or serum enzyme
  levels) are handled differently.
• Unless the raw data from a study are available,
  the ability to estimate a BMD may be limited by
  the format of the data presented.

45
Dealing with multiple BMDs

• Since each biologically, statistically
  significant response in a single sex and species
  is modeled separately, it is almost certain that
  even one study will produce several BMDs.
• Chemicals may also have studies for which BMDs
  are determinable, and others where only a NOAEL
  can be judged.
• Which BMD should be used to calculate an RfD?
• choosing the smallest is consistent with the
  concept of critical effect, and, thus, may be an
  appropriate default
• using a geometric mean can be a reasonable
  approach

46
RfD/RfC Exceeders
ABOVE RfD DOSE
FOG OF UNCERTAINTY
"SAFE"
REGION OF NO EFFECTS
"NOT SAFE"
REGION OF ADVERSE EFFECTS
RfD DOSE
INCREASING DOSE
47
Categorical Regression
RfD Definition Regression
model "without appreciable risk" r lt
10-2 "is likely to be" P()
gt 0.95 "deleterious effect"
severity moderate or frank New RfD
Definition P ( r lt 10-2 at doseltRfD ) gt 0.95
where r P (severity gt1)
48
Advantages Limitations of Categorical Regression

• Advantages
• provides a consistent basis for calculating risk
  above the RfD
• all useful data can be categorized
• accounts for severity of toxic effect
• Limitations
• animal to human extrapolation is still needed
• data are transformed into categories which loses
  information

49
Summary of Traditional Approach New Methods

• Misunderstandings abound ask questions.
• Estimates of safe doses are accurate but
  imprecise. They are believed to be without risk,
  but cannot be used to estimate risk.
• Benchmark dose (BMD), categorical regression, and
  chemical specific adjustment factors (CSAF) ask
  different questions of the data, and are not
  alternatives to each other, nor the current
  safe dose approach.
• Methods of combining two or more of these methods
  are routinely used. For example, in its
  estimation of safe dose, Health Canada EPA use
  both BMD and CSAF.

50
Developmental Neurotoxicity (DN) NOAELs Compared

• According to Makris et al. (1998) for 9
  pesticides
• 8 DN NOAELs were lower than developmental values
• 6 DN NOAELs were lower than or equal to
  neurotoxicity values
• 7 DN NOAELs were between 1 and 93-fold higher
  than chronic values, for 2 DN NOAELs values were
  70 and 90 of chronic values (approximately
  equal)
• Peer review by SAP (1999) suggested that maternal
  or developmental toxicity NOAELs were same or
  lower than DN values for 10 of 12 pesticides DN
  study was not more sensitive than developmental
  or reproductive.

51
Risk for Chemicals with Common Mechanism of
Toxicity

• Simple mixtures of similarly-acting chemicals
  can be described by the Hazard Index
• Simple a few defined chemical components
  (nlt25?)
• Hazard Index assumes similar toxicity and dose
  addition
• i.e., assumes NO INTERACTIONS
• Toxicological Interactions Exist
• Most interactions data are for binary (n2)
  mixtures
• Strong evidence of interaction currently used
  to change priority of the risk assessment
• Qualitative use of the information does not
  change the risk or the cleanup goals

52
inadequate
Only Qualitative Assessment
adequate
on Whole Mixture
on Components
yes
no
yes
no
WOE-based Hazard Index (for similar toxicity)
Mixture RfD/C Slope Factor
Independence
Toxicological Similarity
Mixed
Qualitative Assessment
yes
no
Comparative Potency Mixture RfD/C Slope Factor
yes
no
Response Addition

RPFs TEFs Environmental Factors
Qualitative Assessment
Hazard Index
Adapted from Hertzberg, Rice and Teuschler,
1999. Methods for health risk assessment of
combustion mixtures. In Hazardous Waste
Incineration Roberts, Teaf and Bean, eds.
Lewis Boca Raton. 105-148.