Bayesian methods for uncertainty factor application for derivation of reference values

• Published in: Regulatory toxicology and pharmacology Vol. 80; pp. 9 - 24
• Main Authors: Simon, Ted W., Zhu, Yiliang, Dourson, Michael L., Beck, Nancy B.
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier Inc 01-10-2016
• Subjects: Administration, Oral; Animals; Bayes Theorem; Bayesian methods; Computer Simulation; Dose-Response Relationship, Drug; Epidemiologic Methods; Hazard assessment; Hazardous Substances - pharmacokinetics; Hazardous Substances - toxicity; High-Throughput Screening Assays; Humans; Inhalation Exposure; Models, Statistical; Monte Carlo Method; Reference concentration; Reference dose; Reference Values; Risk Assessment; Toxicity Tests - methods; Toxicity Tests - standards; Uncertainty; Uncertainty factor; Uncertainty factor; Reference dose; Bayesian methods; Hazard assessment; Risk assessment; Reference concentration
• ISSN: 0273-2300; 1096-0295
• DOI: 10.1016/j.yrtph.2016.05.018

In 2014, the National Research Council (NRC) published Review of EPA’s Integrated Risk Information System (IRIS) Process that considers methods EPA uses for developing toxicity criteria for non-carcinogens. These criteria are the Reference Dose (RfD) for oral exposure and Reference Concentration (RfC) for inhalation exposure. The NRC Review suggested using Bayesian methods for application of uncertainty factors (UFs) to adjust the point of departure dose or concentration to a level considered to be without adverse effects for the human population. The NRC foresaw Bayesian methods would be potentially useful for combining toxicity data from disparate sources—high throughput assays, animal testing, and observational epidemiology. UFs represent five distinct areas for which both adjustment and consideration of uncertainty may be needed. NRC suggested UFs could be represented as Bayesian prior distributions, illustrated the use of a log-normal distribution to represent the composite UF, and combined this distribution with a log-normal distribution representing uncertainty in the point of departure (POD) to reflect the overall uncertainty. Here, we explore these suggestions and present a refinement of the methodology suggested by NRC that considers each individual UF as a distribution. From an examination of 24 evaluations from EPA’s IRIS program, when individual UFs were represented using this approach, the geometric mean fold change in the value of the RfD or RfC increased from 3 to over 30, depending on the number of individual UFs used and the sophistication of the assessment. We present example calculations and recommendations for implementing the refined NRC methodology.
[Display omitted]
•NRC suggested EPA use Bayesian methods in developing Reference Values (RfVs).•24 chemicals assessed by EPA’s IRIS program were examined as a proof-of-concept.•Bayesian methods applied to individual UFs increased the RfV from 2.5 to 150 fold over the 24 individual chemicals considered.•Fold-change increased with the composite UF and decreased with use of PBPK models.•We propose EPA begin using this method now to gain experience and spur development.