Default Assumptions, Uncertainty Factors EPAs Approach

1
Default Assumptions, Uncertainty Factors EPAs
Approach

• BOSC Risk Assessment Workshop
• February 2-3, 2005

Rita Schoeny, Ph.D., Senior Science
Advisor, Office of Water, U.S. EPA
2
Outline

• What are defaults? Why are they used? When?
• Defaults and Science Policy
• Examples
• Uncertainty factors, non-linear low dose
  extrapolation

3
Defaults What and Why

• the option chosen on the basis of risk
  assessment policy that appears to be the best
  choice in the absence of data to the contrary
  NRC 1983
• Because the data are not to be had
• Science and Judgment (NRC 1994) says its OK
• As do the most recent version of the Cancer
  Guidelines and GAO (2001)

4
Science and Judgment Principles for Defaults

• Protect public health,
• Ensure scientific validity,
• Minimize serious errors in risk estimation,
• Maximize incentives for research,
• Orderly and predictable process,
• Foster openness and trustworthiness

5
Biggest Change Over Time

• Old days, one used defaults unless there were
  data indicating a departure
• e.g. LMS procedure for cancer risk
• Now EPA uses defaults only if existing data are
  determined to be inadequate or not usable in the
  assessment.

6
Use Data Before Invoking Defaults
Analyze the available data

Is there too much uncertainty or is critical
information lacking?
Invoke a default option
Y
N
The primary goal of EPA actions is public
health protection, accordingly, as an agency
policy, the defaults used in the absence of
scientific data to the contrary should be health
protective (SAB 1999).
7
Defaults Are Not Arbitrary

• Usually based on data and a scientific consensus
• Defaults based on published studies, empirical
  observation, extrapolation from related
  observations, scientific theory
• Not site or chemical specific -- generalizable
• Usually have been peer reviewed
• Are identified in risk characterization

8
Use of Defaults Is Science Policy

• Assumptions are used to ensure that risk is not
  underestimated EPAs overall goal is to protect
  public health. but not overestimate either.
• Defaults are re-examined as more data are
  available and methodologies are improved.
• EPA addresses chemical-specific use of a default
  during peer review of the assessment

9
Example Animal Data Are Relevant to Humans

• All data? Adverse vs.adaptive effects severity.
• Specific guidance on some lesions as to their
  relevance to humans -- Kerry will discuss
• Most emphasis on mode of action
• Benign tumors have potential to progress to
  malignant tumors?
• Not always depends on tumor type, MOA,
  scientific judgment (e.g. NTP process)
• Target organ concordance? Kerry will discuss
• Case by case, MOA

10
Example Effects at High Dose Are Relevant to Low
Dose

• Maximum Tolerated Dose
• Standard NTP bioassay uses MTD
• EPA Guidelines say consider case by case (in
  context of other studies and lines of evidence)
  use guidance for particular lesions (thyroid
  tumors)
• Guidelines say inspect data, use results at MTD
  -- but not above for extrapolation
• Establish MOA

11
Example the 70 Kg Spherical Human . . .

• Who lives for 70 years and drinks 2 l water per
  day
• Analyses by Office of Water
• Used CSFII 94-96, n 15,000
• Generated distributions specific to age and
  gender
• 99th percentile all ages, women and men
• 2 l/day mean 1 l/day

12
There Is a Clear Pattern to EPA and Defaults

• Usable data trump defaults
• As our understanding of MOA, pharmacokinetics,
  etc. evolve, more data become usable.
• But, well be using defaults for a long time to
  come.

13
Low Dose Extrapolation UF
14
Use of Uncertainty Factors

• EPA follows the RfD/ RfC Technical Panel Report
  (U.S. EPA 2002)
• Default UF are generally 10 fold
• Intra-human, species to species,
  less-than-lifetime, LOAEL to NOAEL, data base
  insuf.
• 5 UF is inappropriate (why are we doing RfD?)
  no UF gt 3000.
• No Modifying factor

15
Same Rule for UF Data First, then Default

• Example Inhalation, RfCs use RfC methodology
  guidance (U.S. EPA 1994) in determining
  interspecies UF. (Generally use UF 3 when
  dosimetric adjustment of animal data).
• Example Methylmercury PK UF of 3 (based on
  analyses of interindividual variability) and
  default PD UF of 3
• Risk Assessment Forum working on Guidance for
  Data-derived UF.
• NCEA analyses of scientific foundation for
  estimating uncertainty in reference values

16
Quo Vadis RfD?

• Non-linear extrapolation could be
• Probabilistic, estimation of risk at an exposure
  level,
• An estimate based on a Benchmark Dose approach,
  using uncertainty factors,
• A traditional RfD based on N/LOAEL,
• Calculation of margin between human exposure
  level and a point of departure for some effect,
• Or . . .

17
(No Transcript)
18
Example Route to Route

• Default assumption chemicals which cause
  internal toxicity by one route will be toxic by
  another
• Corollary, if chemical is absorbed by one route,
  absorbed by all others
• Based on principle that target tissue dose is
  determinant of toxicity
• So if route does not result in internal dose to
  target, then assumption doesnt hold
• Chemicals very active at point-of-entry are less
  likely to result in toxic internal dose