Derivation of Reference Doses
THIS CHAPTER contains a brief description of the methods used by toxicologists at Oak Ridge National Laboratory (ORNL) to derive the U.S. Army's interim reference doses (RfDs) for GA, GB, GD, VX, sulfur mustard, and lewisite. Those methods were based on the procedures outlined by the U.S. Environmental Protection Agency for Superfund risk assessment guidelines (EPA 1989) and for reference concentrations (EPA 1994). An alternative method, the benchmark-dose (BD) approach (Crump 1984) is also described. Because uncertainty factors are integral to both approaches, further consideration is also given to the statistical distribution and confidence associated with them.
Because sulfur mustard is the only agent identified in this report as a carcinogen, a description of the derivation of the carcinogenic slope factor is presented in the chapter on sulfur mustard (see Chapter 7 ).
The RfD is an estimate (with uncertainty spanning perhaps an order of magnitude or greater) of a daily dose to the human population (including susceptible subpopulations) that is likely to be without an appreciable risk of deleterious health effects during a lifetime (EPA 1989). Numerical estimates of risk (probability of an adverse health effect) are not provided by the RfD process. The RfD process only describes the
exposure conditions that are unlikely to cause noncancer health effects, which are typically assumed to have a threshold dose above which deleterious health effects would be expected to occur. The major issues in the calculation of the RfD are identifying the most sensitive effects that are relevant to the human for selecting the no-observed-adverse-effect level (NOAEL) or the lowest-observed-adverse-effect level (LOAEL), the determination of the NOAEL or LOAEL from the most appropriate study, and the magnitudes of the uncertainty factors and the modifying factor used in the process.
ORNL has published a detailed description of the method for deriving the RfD (see Opresko et al. 1998). Briefly, the RfD is calculated by determining the most sensitive and significant NOAEL or LOAEL for noncancer effects and dividing that by the product of a series of uncertainty factors and a modifying factor:
where UFA represents the uncertainty of using experimental animal data for human effects, UFH represents the variable susceptibilities in the human population (e.g., genetic, nutritional, age), UFL represents the expected ratio of the LOAEL to the NOAEL when a LOAEL is used instead of a NOAEL, UFS represents the uncertainty of predicting chronic exposure effects on the basis of subchronic exposure studies, UFD represents the uncertainty assigned to an inadequate data base, and MF is a modifying factor to account for any additional uncertainty not addressed by the standard uncertainty factors.
Typically, uncertainty factors are assigned values ranging from 1 to 10. If information concerning a factor is sparse and uncertainty is high, a default value of 10 generally is used. If information is available, the uncertainty factor might be reduced to 1. For example, UFA would be 1 if the NOAEL or LOAEL is based on human data. A value of less than 1 for the UFA would be used if the NOAEL for the biological end point of concern is more sensitive in animals than in humans. UFH would be less than 10 if the NOAEL or LOAEL is based on a susceptible human subpopulation. UFS would be 1 if adequate chronic exposure studies have been conducted or chronic effects are not expected. UFL would be 1 if a NOAEL is available. UFD would be 1 if an adequate array of human
and animal data is available for various biological effects. For an uncertainty factor that falls between 1 and 10, a factor of 3 is typically assigned, because 3 is the approximate logarithmic mean of 1 and 10, and the assumption is made that the uncertainty factor is distributed lognormally (EPA 1994).
A modifying factor between 1 and 10 is used when the five uncertainty factors do not explicitly account for all scientific uncertainties that exist. The default value for the modifying factor is 1.
The main shortcoming of the traditional RfD approach is the use of the NOAEL. Weaknesses with that use include the following: (1) the NOAEL does not incorporate information on the slope of the dose-response curve or the variability in the data, (2) the NOAEL is likely to be higher with smaller sample sizes or an inadequate study, (3) the NOAEL is limited to one of the experimental doses, (4) the number and spacing of doses in a study influence the dose chosen for the NOAEL, and (5) because the NOAEL is defined as a dose that does not produce an observed increase in adverse effects compared with control levels and is dependent on the power of the study, theoretically, the risk associated with the NOAEL might fall anywhere between zero and 10% for quantal data (EPA 1991). The true risk at a NOAEL can vary from zero to over 20% depending on the end point, spacing of doses, and numbers of animals used (Leisenring and Ryan 1992).
Because of shortcomings in the use of NOAELs to determine doses with low risk, Crump (1984) proposed that the NOAEL be replaced by a benchmark dose (BD) associated with a biological effect. The BD is a dose with a specified low level of excess health risk, generally in the range of 1% to 10%, that can be estimated from data with little or no extrapolation outside the experimental dose range. Specifically, the BD is derived by modeling the data in the observed experimental range, selecting an incidence level within or near the observed range (e.g., the effective dose producing a 10% increased incidence of response), and determining the upper confidence limit on the model. Because the method does not involve extrapolation far below the experimental range of doses, the BD is less dependent on the choice of the mathematical form of the dose-response model than it is on the choice of the uncer-
tainty factors. To account for experimental variation, a lower confidence limit or uncertainty factors on the BD are recommended to assure that the specified excess risk (e.g., 10%) is not likely to be exceeded.
The BD approach uses more of the available toxicity data (such as the number of animals, dose-response data, and data variability) than the traditional RfD method and provides a consistent basis for calculating the RfD. However, selecting the magnitudes of the uncertainty factors remains an issue in the BD approach, as in the traditional RfD approach. Depending on the level of risk selected at the BD, the uncertainty factor would be comparable to the ratio of the LOAEL to the NOAEL (Dourson et al. 1985).
In the proposed cancer risk-assessment guidelines of the U.S. Environmental Protection Agency (EPA 1996), a BD approach based on uncertainty factors is one of the methods suggested when a nonlinear dose response is expected (e.g., for nongenotoxic carcinogens or tumor promoters). Low-dose linear extrapolation below the BD would still be used for direct-acting (genotoxic) carcinogens. For carcinogens that act indirectly (e.g., nongenotoxic carcinogens, such as tumor promoters), it might be possible to predict doses below which no appreciable risk is expected. However, it must be shown that exposure to such carcinogenic substances does not augment existing background processes that lead to low-dose linearity (Crawford and Wilson 1996). Hence, EPA (1996) is proposing that techniques for risk assessment be based upon the mode of action of the carcinogen. It has been suggested that the lower confidence limit on an excess risk of 10% or less for the BD is the point of departure (the point at which low-dose extrapolation occurs for either linear or nonlinear extrapolation) for low-risk assessment.
Multiplying several uncertainty factors together possibly can lead to a large overall uncertainty factor that results in an unnecessarily small RfD. It is unlikely that each uncertainty factor simultaneously needs to be at its maximal value. Several investigators recently addressed the issue of compounding conservatism resulting from using the upper bounds on each of the uncertainty factors in the calculation of a RfD or reference concentration (e.g., Burmaster and Harris 1993; Calabrese and Gilbert 1993; Bogen 1994; Nair et al. 1995; Baird et al. 1996; Swartout
et al. 1998). EPA typically uses a maximum of 3,000 for the product of four uncertainty factors that individually are greater than 1, and a maximum of 10,000 with five uncertainty factors (Dourson 1994).
The product of the uncertainty factors (U = UFA × UFH × UFL × UFS × UFD) must produce high confidence that the overall product (U) is large enough to protect susceptible subpopulations adequately from long-term exposures. In the absence of information to select a specific value for an uncertainty factor, default values of 10 generally are used. The value assigned to an uncertainty factor can be considered a random variable, the default value being larger for most end points and chemicals. For example, Swartout (1996) examined the ratio of doses that produced equivalent adverse effects, NOAELs or LOAELs, from subchronic and chronic exposures to about 100 substances. Swartout (1996) observed that the median ratio was 2. If data were available only from a subchronic study, on average, the NOAEL or LOAEL dose for chronic exposures should be reduced by a factor of 2. Swartout (1996) observed that the 95th percentile for the ratio of subchronic to chronic doses for NOAELs or LOAELs was 17. Hence, to cover only the uncertainty of estimating chronic effects from subchronic data, the UFS should be 17 for 95% coverage. A different factor might be necessary for some classes of chemicals or specific end points. The currently used default factor of 10 provides about 89% coverage in general. If an adverse effect is proportional to the total dose (i.e., the dose rate times the duration of exposure), extrapolation of results from a 90-day subchronic exposure to a 2-year chronic exposure would use an average UFS of 730 days ÷ 90 days = 8.
From selected databases providing the distributions of various uncertainty factors, Baird et al. (1996) used computer simulations to obtain the overall product of uncertainty factors required to achieve selected levels of confidence. They indicated that dividing the NOAEL from a chronic exposure study in animals by the product of default values of 10 for UFA and UFH (10 × 10 = 100) gives about 95% confidence that RfD = NOAEL ÷ 100 is adequately conservative. Baird et al. (1996) found that a total uncertainty factor of 1,000 tends to provide about 99% confidence of adequate conservatism when three default values of 10 are used, and a total of 3,000 provides a similar result when four uncertainty factors are used. Thus, it appears that the conventional products of 10 for default values of uncertainty factors provide reasonable assurance of safety.
The subcommittee concludes that the method used by ORNL to derive the Army's interim RfDs is scientifically sound and is consistent with the guidelines and process used by EPA. It must be emphasized that scientific judgment is often a key overriding factor in that method. Because the process involves a series of extrapolations, each with its own degree of uncertainty, emphasis should be placed on establishing doses that are judged to be safe for human exposure based on the best scientific information. In addition, the subcommittee believes that the BD approach should also be considered in establishing RfDs in the future. Similar risk estimates from the conventional and the BD methods would provide greater confidence in the proposed RfDs.
Baird, S.J.S., J.T. Cohen, J.D. Graham, A.I. Shlyakhter, and J.S. Evans. 1996. Noncancer risk assessment: A probabilistic alternative to current practice. Hum. Ecol. Risk Assess. 2:79–102.
Barnes, D.G., and M. Dourson. 1988. Reference dose (RfD): Description and use in health risk assessments. Regul. Toxicol. Pharmacol. 8:471–486.
Bogen, K.T. 1994. A note on compounded conservatism. Risk Anal. 14:379–381.
Burmaster, D.E., and R.H. Harris. 1993. The magnitude of compounding conservatisms in Superfund risk assessments. Risk Anal. 13:131–134.
Calabrese, E.J., and C.E. Gilbert. 1993. Lack of total independence of uncertainty factors (UFs): Implications for the size of the total uncertainty factor. Regul. Toxicol. Pharmacol. 17:44–51.
Crawford, M., and R. Wilson. 1996. Low-dose linearity: The rule or the exception? Hum. Ecol. Risk Assess. 2:305–330.
Crump, K.S. 1984. A new method for determining allowable daily intakes. Fundam. Appl. Toxicol. 4:854–871.
Dourson, M.L., R.C. Hertzberg, R. Hartung, and K. Blackburn. 1985. Novel methods for the estimation of acceptable daily intake. Toxicol. Ind. Health 1:23–41.
Dourson, M.L. 1994. Methods for establishing oral reference doses (RfDs). Pp. 51–61 in Risk Assessment of Essential Elements, W. Mertz, C.O. Abernathy, and S.S. Olin, eds. Washington, D.C.: ILSI Press.
EPA (U.S. Environmental Protection Agency). 1989. Risk Assessment Guidance for Superfund, Vol. I, Human Health Evaluation Manual (Part A), Interim
Final. EPA/540/1-89/002. U.S. Environmental Protection Agency, Office of Emergency and Remedial Response, Washington, D.C.
EPA (U.S. Environmental Protection Agency). 1991. Guidelines for Developmental Toxicity Risk Assessment. Notice. Fed. Regist. 56(234):63798–63826.
EPA (U.S. Environmental Protection Agency. 1994. Methods for the Derivation of Inhalation Reference Concentrations and Application of Inhalation Dosimetry. EPA/600/8-90/066F. Office of Health and Environmental Assessment, National Center for Environmental Assessment, Cincinnati, Ohio.
EPA (U.S. Environmental Protection Agency). 1996. Proposed Guidelines for Carcinogen Risk Assessment. Notice. Fed. Regist. 61(79):17960–18011.
Leisenring, W., and L. Ryan. 1992. Statistical properties of the NOAEL. Regul. Toxicol. Pharmacol. 15:161–171.
Nair, R.S., J.H. Sherman, M.W. Stevens, and F.R. Johannsen. 1995. Selecting a more realistic uncertainty factor: Reducing compounding effects of multiple uncertainties. Hum. Ecol. Risk Assess. 1:576–589.
Opresko, D.M., R.A. Young, R.A. Faust, S.S. Talmage, A.P. Watson, R.H. Ross, K.A. Davidson, and J. King. 1998. Chemical warfare agents: Estimating oral reference doses. Rev. Environ. Contam. Toxicol. 156:1–183.
Swartout, J. 1996. Subchronic-to-Chronic Uncertainty Factor for the Reference Dose. Abstract F2.03. Society of Risk Analysis Annual Meeting, New Orleans, La.
Swartout, J.C., P.S. Price, M.L. Dourson, H.L. Carlson-Lynch, and R.E. Keenan. 1998. A probabilistic framework for the reference dose (Probabilistic RfD). Risk Anal. 18:271–282.