Sources of Uncertainty and Current Practice for Addressing Them: Toxicological Perspective

1. Sources of Uncertainty and Current Practice for Addressing Them: Toxicological Perspective David A. Bussard U.S. Environmental Protection Agency The views presented here are those of the presenter and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency.

2. Basics of Most Current Practice:The “Question” Being Answered. • Reference Value (RfV): An estimate of an exposure for a given duration to the human population (including susceptible subgroups) that is likely to be without an appreciable risk of adverse health effects over a lifetime. It is derived from a BMDL, a NOAEL, a LOAEL, or another suitable point of departure, with uncertainty/variability factors applied to reflect limitations of the data used. • Inhalation Unit Risk: The upper-bound excess lifetime cancer risk estimated to result from continuous exposure to an agent at a concentration of 1 µg/m3 in air. The interpretation of inhalation unit risk would be as follows: if unit risk = 2 × 10-6 per µg/m3, 2 excess cancer cases (upper bound estimate) are expected to develop per 1,000,000 people if exposed daily for a lifetime to 1 µg of the chemical per m3 of air.

3. Probability and Uncertainty: Noncancer • “Reference Value (RfV): An estimate of an exposure for a given duration to the human population (including susceptible subgroups) that is likely to be without an appreciable risk of adverse health effects over a lifetime. It is derived from a BMDL, a NOAEL, a LOAEL, or another suitable point of departure, with uncertainty/variability factors applied to reflect limitations of the data used.” • No specified probability definition of “appreciable” or “likely” or rate for the BMDL, NOAEL or LOAEL. • A “conservative” estimate: • BMLD, not BMD • Uncertainty factors (which are used to lower the concentration, not increase it) • a dose “…likely without an appreciable risk…” Meant to take into account possible unstudied effects. • Hard to know what a “best estimate” analogue would be.

4. Reference Value (RfV) practice • Value generally derived from “most sensitive” adverse endpoint or precursor of an adverse endpoint. • When sampling error can be considered, use model to derive BMD and BMDL; use BMDL. (Often about 5-10% response rate used.) • Uncertainty factors address range of issues(see next slide). • Statement regarding overall “confidence” in the value.

5. Uncertainty Factors Extrapolation/scaling/uncertainty: • Subchronic data; chronic scenario. • LOAEL v NOAEL or BMDL • Animal data; human assessment. • Uncaptured human population variability. • Uncertainty that database likely captures all sensitive endpoints.

6. Judgment • “Conservative” question being asked. • Yet, answer needs to be defensible and reasonable. • A lot of thought goes into assessing studies and overall database. • Much of that is currently not quantified in formal uncertainty analysisand can be folded into choice of “a critical study”.

7. Probability and Uncertainty: Cancer • Inhalation Unit Risk: The upper-bound excess lifetime cancer risk estimated to result from continuous exposure to an agent at a concentration of 1 µg/m3 in air. The interpretation of inhalation unit risk would be as follows: if unit risk = 2 × 10-6 per µg/m3, 2 excess cancer cases (upper bound estimate) are expected to develop per 1,000,000 people if exposed daily for a lifetime to 1 µg of the chemical per m3 of air. • probability framework and language to some extent • Could frame a “best estimate” • Yet, “upper bound” roughly defined: • A characterization of database: • “Not Likely to Be Carcinogenic to Humans.” • “Inadequate Information to Assess Carcinogenic Potential,” • “Suggestive Evidence of Carcinogenic Potential.” • “Likely to Be Carcinogenic to Humans” • “Carcinogenic to Humans.” C.f. p. 2-53 EPA Cancer Guidelines (EPA 2005)

8. Cancer Uncertainty in Practice • Animal study results • uncertain result • uncertain “cause” • uncertain animal prediction of human response. • Human epi data has its own set of uncertainties. • Dose-response modeling done in range of the data; lower-confidence limit of risk-specific dose used. • Linear extrapolation typical for below data.

9. Contrasting • Noncancer: explicitly considers more unknowns and uncertainties. But, much less clear what a “best estimate” would be, how to estimate “dose-response” around RfV, and how to pose as probability statements. • Cancer: developed with a probability framework; but many uncertainties not explicitly addressed. • Both have weight-of-evidence statements; neither presents those as “probabilities”. • Currently, both have some “0 / 1” choices operationally – though some nonquantitative consideration in the WoE statements.

10. Decision and Economic Analysis Pressure To Change • Decision-makers and economists often want “best estimates”, perhaps with two-tailed confidence limits. • Often want “marginal” impacts. Full dose-response function. • The economists also want information on any high-social-cost effects that might not be “most sensitive”. • Sometimes asking about risks of intermediate duration or episodic exposures; interest in risks of cumulative exposures.

11. Advancing Science • Some noncancer effects show no evidence of thresholds; risk at classic “NOAEL”. • Detailed dosimetry and PBPK; estimates with both uncertainty and variability. Part of interspecies and within-humans variability and uncertainty. • Increasing mechanistic information starts to push towards site concordance thinking. Yet, could mis-step. • Greater use of epidemiology data for quantification of magnitude of risk per dose?

12. Advancing Science • Potential to better identify and understand subpopulations; could change questions. • Potential of high-throughput testing systems and awareness of large number of untested chemicals. • High-throughput will steer us to thinking about uncertainties for “chemicals” or a classes of them; • Many other science developments drive us to thinking about “the specific chemical”. • Need both.

13. Improvements • Forest plots and meta-analysis. • Multiple endpoints reported or shown, not just the most sensitive. [Reduces “0 / 1”; could address “high impact/less sensitive”; info re different durations.] “cRfCs” • Careful analysis of uncertainty in dosimetry and PBPK; potentially variability analysis. • Showing values for “general population” and values for understood “subpopulations”.

14. Potential / Controversial Developments • Probabilities (“subjective”?) regarding judgment calls, and carrying forward several alternative choices. • Hard thinking about when dose-response reflects population variability, when it reflects graded processes or stochastic processes. Can they be well-differentiated? • Conceptual models to illustrate competing mode of action hypotheses and available data. May be multiple MOAs that contribute to response.

15. “Toxicologist’s Perspective”? • Have discussed risk assessment synthesis and evaluation of toxicology data for risk assessment. • Toxicologists’ perspectives differ • Strong interest in mechanistic details that sometimes don’t clearly affect “risk assessment”. • Expectations about low-dose behavior differ widely. • Statistics, modeling, and biology: different languages and disciplines though each relies on other in risk assessment. • Useful to also listen to what decision/economic analysts are trying to do with what we give them.