The Scientific Basis of Uncertainty Factors Used in Setting Occupational Exposure Limits.

The uncertainty factor concept is integrated into health risk assessments for all aspects of public health practice, including by most organizations that derive occupational exposure limits. The use of uncertainty factors is predicated on the assumption that a sufficient reduction in exposure from those at the boundary for the onset of adverse effects will yield a safe exposure level for at least the great majority of the exposed population, including vulnerable subgroups. There are differences in the application of the uncertainty factor approach among groups that conduct occupational assessments; however, there are common areas of uncertainty which are considered by all or nearly all occupational exposure limit-setting organizations. Five key uncertainties that are often examined include interspecies variability in response when extrapolating from animal studies to humans, response variability in humans, uncertainty in estimating a no-effect level from a dose where effects were observed, extrapolation from shorter duration studies to a full life-time exposure, and other insufficiencies in the overall health effects database indicating that the most sensitive adverse effect may not have been evaluated. In addition, a modifying factor is used by some organizations to account for other remaining uncertainties-typically related to exposure scenarios or accounting for the interplay among the five areas noted above. Consideration of uncertainties in occupational exposure limit derivation is a systematic process whereby the factors applied are not arbitrary, although they are mathematically imprecise. As the scientific basis for uncertainty factor application has improved, default uncertainty factors are now used only in the absence of chemical-specific data, and the trend is to replace them with chemical-specific adjustment factors whenever possible. The increased application of scientific data in the development of uncertainty factors for individual chemicals also has the benefit of increasing the transparency of occupational exposure limit derivation. Improved characterization of the scientific basis for uncertainty factors has led to increasing rigor and transparency in their application as part of the overall occupational exposure limit derivation process.

Human health risks from mercury exposure from broken compact fluorescent lamps (CFLs).

Human health risk to infants/toddlers and adults was evaluated based on two exposure scenarios from compact fluorescent lamp (CFL) breakage; first in a room with no ventilation and no clean-up, and second in a room with adequate ventilation and clean-up. Concentration data from multiple exposure scenarios tested in a study by Stahler et al. (2008) were compared to human toxicity benchmarks to calculate hazard quotients. For the no clean-up scenario, hazard quotients were generally less than 1, suggesting an unlikely health risk. When the room was ventilated and the broken CFL was cleaned-up, mercury concentrations were generally lower. A review of release scenarios, along with duration-adjusted toxicity benchmarks, indicated that few releases produced levels of concern, but some scenarios resulted in exceedance of risk targets and require further study. Uncertainties in this screening characterization include assumptions about room size, ventilation, age of lamp, the distribution of mercury in the room, and also the choice of the toxicity benchmarks used to develop the hazard quotients.

Pediatric susceptibility to 18 industrial chemicals: a comparative analysis of newborn with young animals.

We comprehensively re-analyzed the toxicity data for 18 industrial chemicals from repeated oral exposures in newborn and young rats, which were previously published. Two new toxicity endpoints specific to this comparative analysis were identified, the first, the presumed no observed adverse effect level (pNOAEL) was estimated based on results of both main and dose-finding studies, and the second, the presumed unequivocally toxic level (pUETL) was defined as a clear toxic dose giving similar severity in both newborn and young rats. Based on the analyses of both pNOAEL and pUETL ratios between the different ages, newborn rats demonstrated greater susceptibility (at most 8-fold) to nearly two thirds of these 18 chemicals (mostly phenolic substances), and less or nearly equal sensitivity to the other chemicals. Exceptionally one chemical only showed toxicity in newborn rats. In addition, Benchmark Dose Lower Bound (BMDL) estimates were calculated as an alternative endpoint. Most BMDLs were comparable to their corresponding pNOAELs and the overall correlation coefficient was 0.904. We discussed how our results can be incorporated into chemical risk assessment approaches to protect pediatric health from direct oral exposure to chemicals.

Guidelines for application of chemical-specific adjustment factors in dose/concentration-response assessment.

This manuscript addresses guidance in the use of kinetic and dynamic data to inform quantitatively extrapolations for interspecies differences and human variability in dose-response assessment developed in a project of the International Programme on Chemical Safety (IPCS) initiative on Harmonisation of Approaches to the Assessment of Risk from Exposure to Chemicals. The guidance has been developed and refined through a series of planning and technical meetings and larger workshops of a broad range of participants from academia, government agencies and the private sector. The guidance for adequacy of data for replacement of common defaults for interspecies differences and human variability is presented in the context of several generic categories including: determination of the active chemical species, choice of the appropriate metric (kinetic components) or endpoint (dynamic components) and nature of experimental data, the latter which includes reference to the relevance of population, route and dose and the adequacy of the number of subjects/samples. The principal objective of this guidance developed primarily as a resource for risk assessors, is to foster better understanding of the components of and criteria for adequacy of chemical-specific data to quantitate interspecies differences and human variability in kinetics and dynamics. It is anticipated that this guidance will also encourage the development of appropriate data and facilitate their incorporation in a consistent fashion in dose-response assessment for regulatory purposes (IPCS, 2001).

Genetic polymorphisms in assessing interindividual variability in delivered dose.

Increasing sophistication in methods used to account for human variability in susceptibility to toxicants has been one of the success stories in the continuing evolution of risk assessment science. Genetic polymorphisms have been suggested as an important contributor to overall human variability. Recently, data on polymorphisms in metabolic enzymes have been integrated with physiologically based pharmacokinetic (PBPK) modeling as an approach to determining the resulting overall variability. We present an analysis of the potential contribution of polymorphisms in enzymes modulating the disposition of four diverse compounds: methylene chloride, warfarin, parathion, and dichloroacetic acid. Through these case studies, we identify key uncertainties likely to be encountered in the use of polymorphism data and highlight potential simplifying assumptions that might be required to test the hypothesis that genetic factors are a substantive source of human variability in susceptibility to environmental toxicants. These uncertainties include (1) the relative contribution of multiple enzyme systems, (2) the extent of induction/inhibition through coexposure, (3) allelic frequencies of major ethnic groups, (4) the absence of chemical-specific data on the kinetic parameters for the different allelic forms of key enzymes, (5) large numbers of low-frequency alleles, and (6) uncertainty regarding differences between in vitro and in vivo kinetic data. Our effort sets the stage for the acquisition of critical data and further integration of polymorphism data with PBPK modeling as a means to quantitate population variability.

Mathematical modelling and quantitative methods.

The present review reports on the mathematical methods and statistical techniques presently available for hazard characterisation. The state of the art of mathematical modelling and quantitative methods used currently for regulatory decision-making in Europe and additional potential methods for risk assessment of chemicals in food and diet are described. Existing practices of JECFA, FDA, EPA, etc., are examined for their similarities and differences. A framework is established for the development of new and improved quantitative methodologies. Areas for refinement, improvement and increase of efficiency of each method are identified in a gap analysis. Based on this critical evaluation, needs for future research are defined. It is concluded from our work that mathematical modelling of the dose-response relationship would improve the risk assessment process. An adequate characterisation of the dose-response relationship by mathematical modelling clearly requires the use of a sufficient number of dose groups to achieve a range of different response levels. This need not necessarily lead to an increase in the total number of animals in the study if an appropriate design is used. Chemical-specific data relating to the mode or mechanism of action and/or the toxicokinetics of the chemical should be used for dose-response characterisation whenever possible. It is concluded that a single method of hazard characterisation would not be suitable for all kinds of risk assessments, and that a range of different approaches is necessary so that the method used is the most appropriate for the data available and for the risk characterisation issue. Future refinements to dose-response characterisation should incorporate more clearly the extent of uncertainty and variability in the resulting output.

Use of human data for the derivation of a reference dose for chlorpyrifos.

In 1998 a panel of experts met to discuss the data available on chlorpyrifos, both human and animal, and to determine the most appropriate endpoints to be used for the derivation of the reference dose (RfD). Since that time, additional data have become available on chlorpyrifos from an experimental study involving humans. Moreover, Food Quality Protection Act (FQPA) considerations need to be addressed, and the appropriate cholinesterase endpoint, whether plasma, red blood cell, peripheral nerve, or brain, has become highly debated. Therefore, Dow AgroSciences, one of the manufacturers of chlorpyrifos, convened a second panel of toxicology and medical experts on June 21, 1999, to consider the presently available scientific literature both published and unpublished on chlorpyrifos and to determine the acute and chronic toxicological RfDs for chlorpyrifos. Four questions were posed to this second panel of experts concerning the available data on chlorpyrifos. (1) Should the RfD for chlorpyrifos be based on acetylcholinesterase (AChE) inhibition or butyrylcholinesterase (BuChE) inhibition as an endpoint for adverse effect? (2) Should the RfDs for chlorpyrifos be based on the data set from three human studies, which are supported by animal data? (3) Should the FQPA safety factor be reduced to 1xbased on animal studies of pre- or postnatal toxicity? (4) If an RfD for chlorpyrifos were to be based on animal data, then is a 10-fold interspecies uncertainty factor necessary? The panel of experts concluded that: (1) inhibition of BuChE is not an adverse effect, and the RfD for chlorpyrifos should be based on AChE inhibition; (2) the RfD for chlorpyrifos should be based on the three available human studies, which are also supported by animal data; (3) the extra FQPA safety factor should be reduced to 1x, because chlorpyrifos shows no pre- or postnatal toxicity of concern at relevant human exposure conditions; and (4) the extra 10-fold safety factor for interspecies variation appears overly conservative because no differences in species sensitivity to chlorpyrifos is evident.

Using human data to protect the public's health.

The value of using human data in the assessment and management of risk is evaluated. Although the use of such data has a long and successful history with environmental contaminants and the development of drugs and commercial chemicals, recent deliberations within the Environmental Protection Agency (EPA) have questioned this practice in part. Specifically, we evaluate the degree to which reference doses (RfDs) and reference concentrations (RfCs) derived from human data on EPA's Integrated Risk Information System (IRIS) differ with RfDs and RfCs that we estimate from experimental animal data. We also use several minimal risk levels of the Agency for Toxic Substances and Disease Registry (ATSDR) and tolerable intakes of Health Canada in this comparison. Human-based RfDs are more than threefold lower than the corresponding animal-based RfDs for 23% of the comparisons. Human- based RfDs or RfCs are lower than corresponding animal-based RfDs or RfCs for 36% of the comparisons. Furthermore, for 10 of 43 possible comparisons, insufficient experimental animal data are readily available or data are inappropriate to estimate either RfDs or RfCs. We also discuss human pharmacokinetic data from volunteer studies and mechanistic studies with human tissues in vitro and demonstrate through a series of case discussions that utilization of such data is important when making decisions to protect exposed individuals. Moreover, physiologically based pharmacokinetic (PBPK) modeling evaluates critical information in assessing interindividual variability and identifying at-risk populations. Within the limits of our analysis, we conclude that the direct use and interpretation of human data, in conjunction with data gathered from experimental animals, are public health protective policies that should be encouraged.

Applications of mechanistic data in risk assessment: the past, present, and future.

Mechanistic data, when available, have long been considered in risk assessment, such as in the development of the nitrate RfD based on effects in a sensitive group (infants). Recent advances in biology and risk assessment methods have led to a tremendous increase in the use of mechanistic data in risk assessment. Toxicokinetic data can improve extrapolation from animals to humans and characterization of human variability. This is done by the development of improved tissue dosimetry, by the use of uncertainty factors based on chemical-specific data, and in the development of physiologically based pharmacokinetic (PBPK) models. The development of the boron RfD illustrates the use of chemical-specific data in the improved choice of uncertainty factors. The draft cancer guidelines of the U.S. Environmental Protection Agency emphasize the use of mode of action data. The first choice under the guidelines is to use a chemical-specific, biologically based dose-response (BBDR) model. In the absence of a BBDR model, mode of action data are used to determine whether low-dose extrapolation is done using a linear or nonlinear (margin of exposure) approach. Considerations involved in evaluating a hypothesized mode of action are illustrated using 1,3-dichloropropene, and use of a BBDR model is illustrated using formaldehyde. Recent developments in molecular biology, including transgenic animals, microarrays, and the characterization of genetic polymorphisms, have significant potential for improving risk assessments, although further methods development is needed. Overall, use of mechanistic data has significant potential for reducing the uncertainty in assessments, while at the same time highlighting the areas of uncertainty.

Uncertainties in the reference dose for methylmercury.

This paper critically examines the National Academy of Sciences and the National Research Council report on the toxicological effects of methyl mercury and the recently published US Environmental Protection Agency Reference Dose (RfD) for Methylmercury. Particular scrutiny is placed on the choice of the critical study and the underlining assumptions utilized in the selection of specific uncertainty factors (UFs) and the rationale for using a less-than-default factor of 10. The UFs that were utilized or considered by other agencies and organizations are also critically examined, explained and compared to one another. Based on these analyses, the authors suggest research that could be performed that would ameliorate the uncertainty of choosing a more precise partial UFor that may even provide completeness of database to allow for selecting of a UF for unity, thus improving the precision of the current published RfD.

Hazard identification and dose response of inhaled nickel-soluble salts.

A substantial body of occupational epidemiology data has shown that exposure to mixed soluble and insoluble nickel causes the development of lung and nasal cancer. However, due to coexposure of these populations to soluble and insoluble forms of nickel, and limitations in exposure measurements, the contribution of soluble nickel is difficult to determine. Soluble nickel was negative in an NTP inhalation bioassay, while there was some evidence for tumorigenicity in rats for less soluble nickel oxide, and there was clear evidence for tumorigenicity of insoluble nickel subsulfide in rats. Results of parenteral assays follow a similar pattern, but provide evidence of weak carcinogenicity of soluble nickel. Kinetic factors also indicate that exposure to soluble nickel alone has a low carcinogenic potential. Overall, we conclude that the carcinogenic activity of insoluble nickel compounds should not be used to predict the carcinogenic potential of water-soluble nickel salts. The overall data suggest a nonlinear dose-response relationship for carcinogenicity, but the data are insufficient to determine the doses at which such nonlinearities occur. Under the U.S. EPA's 1996 proposed "Guidelines for Carcinogen Risk Assessment," inhaled soluble nickel compounds would be classified as "cannot be determined," because the existing evidence is composed of conflicting data. A reference concentration of 2 x 10(-4) mg Ni/cu x m was calculated, based on lung fibrosis in male rats observed in the NTP study.

Hazard identification and dose response of ingested nickel-soluble salts.

People can ingest soluble nickel compounds as a normal constituent of food or as a contaminant in drinking water. This paper presents an assessment of the noncancer and cancer human health risks from ingestion of soluble nickel compounds. A reference dose (RfD) of 8 x 10(-3) mg Ni/kg/day in addition to the amount in food was calculated, based on albuminuria in female rats exposed to nickel sulfate in drinking water for 6 months (A. Vyskocil et al., 1994, Hum. Exp. Toxicol. 13, 689-693). This RfD is comparable to the current RfD based on decreased body weight in a chronic feeding study in rats (A. M. Ambrose et al., 1976, J. Food Sci. Technol. 13, 181-187). The potential for nickel-induced reproductive toxicity was also taken into account in the derivation of the RfD. There are a number of negative animal bioassays with soluble nickel salts, but all of them have deficiencies that preclude a definitive conclusion. According to EPA's 1996 draft cancer guidelines, the carcinogenic potential of oral exposure to soluble nickel "cannot be determined because there are inadequate data to perform an assessment."

Health risk above the reference dose for multiple chemicals.

Recent work indicates that the regression of toxicity data viewed as categories of pathological staging is useful for exploring the likely health risk at doses above a Reference Dose (RfD), which 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. Toxic effects, which may include both quantal and continuous data, are classified into ordered categories of total toxic severity (e.g., none, mild, adverse, severe). These severity categories are regressed on explanatory variables, such as dose or exposure duration, to estimate the probability of observing an adverse or severe effect. In this paper, categorical regression has been expanded to compare the likely risks across multiple chemicals when exposures are above their RfDs. Existing health risk data for diazinon, disulfoton, S-ethyl dipropylthiocarbamate, fenamiphos, and lindane were analyzed. As expected, the estimated risks of adverse effects above the RfD varied among the chemicals. For example, at 10-fold above the RfD these risks were modeled to be 0.002, 0.0001, 0.0007, 0.002, and 0.02, respectively. The results and impacts of this analysis indicate that categorical regression is a useful screening tool to analyze risks above the RfD for specific chemicals and suggest its application in evaluating comparative risks where multiple chemical exposures exist.

Distribution of exposure concentrations and doses for constituents of environmental tobacco smoke.

The ultimate goal of the research reported in this series of three articles is to derive distributions of doses of selected environmental tobacco smoke (ETS)-related chemicals for nonsmoking workers. This analysis uses data from the 16-City Study collected with personal monitors over the course of one workday in workplaces where smoking occurred. In this article, we describe distributions of ETS chemical concentrations and the characteristics of those distributions (e.g., whether the distribution was log normal for a given constituent) for the workplace exposure. Next, we present population parameters relevant for estimating dose distributions and the methods used for estimating those dose distributions. Finally, we derive distributions of doses of selected ETS-related constituents obtained in the workplace for people in smoking work environments. Estimating dose distributions provided information beyond the usual point estimate of dose and showed that the preponderance of individuals exposed to ETS in the workplace were exposed at the low end of the dose distribution curve. The results of this analysis include estimations of hourly maxima and time-weighted average (TWA) doses of nicotine from workplace exposures to ETS (extrapolated from 1 day to 1 week) and doses derived from modeled lung burdens of ultraviolet-absorbing particulate matter (UVPM) and solanesol resulting from workplace exposures to ETS (extrapolated from 1 day to 1 year).

A probabilistic framework for the reference dose (probabilistic RfD).

Determining the probabilistic limits for the uncertainty factors used in the derivation of the Reference Dose (RfD) is an important step toward the goal of characterizing the risk of noncarcinogenic effects from exposure to environmental pollutants. If uncertainty factors are seen, individually, as "upper bounds" on the dose-scaling factor for sources of uncertainty, then determining comparable upper bounds for combinations of uncertainty factors can be accomplished by treating uncertainty factors as distributions, which can be combined by probabilistic techniques. This paper presents a conceptual approach to probabilistic uncertainty factors based on the definition and use of RfDs by the U.S. EPA. The approach does not attempt to distinguish one uncertainty factor from another based on empirical data or biological mechanisms but rather uses a simple displaced lognormal distribution as a generic representation of all uncertainty factors. Monte Carlo analyses show that the upper bounds for combinations of this distribution can vary by factors of two to four when compared to the fixed-value uncertainty factor approach. The probabilistic approach is demonstrated in the comparison of Hazard Quotients based on RfDs with differing number of uncertainty factors.

Boron tolerable intake: re-evaluation of toxicokinetics for data-derived uncertainty factors.

Boron, which is ubiquitous in the environment, causes developmental and reproductive effects in experimental animals. This observation has led to efforts to establish a Tolerable Intake value for boron. Although risk assessors agree on the use of fetal weight decreases observed in rats as an appropriate critical effect, consensus on the adequacy of toxicokinetic data as a basis for replacement of default uncertainty factors remains to be reached. A critical analysis of the existing data on boron toxicokinetics was conducted to clarify the appropriateness of replacing default uncertainty factors (10-fold for interspecies differences and 10-fold for intraspecies differences) with data-derived values. The default uncertainty factor for variability in response from animals to humans of 10-fold (default values of 4-fold for kinetics and 2.5-fold for dynamics) was recommended, since clearance of boron is 3- to 4-fold higher in rats than in humans and data on dynamic differences--in order to modify the default value--are unavailable. A data-derived adjustment of 6-fold (1.8 for kinetics and 3.1 for dynamics) rather than the default uncertainty factor of 10-fold was considered appropriate for intrahuman variability, based on variability in glomerular filtration rate during pregnancy in humans and the lack of available data on dynamic differences. Additional studies to investigate the toxicokinetics of boron in rats would be useful to provide a stronger basis for replacement of default uncertainty factors for interspecies variation.

An approach for modeling noncancer dose responses with an emphasis on uncertainty.

This paper presents an approach for characterizing the probability of adverse effects occurring in a population exposed to dose rates in excess of the Reference Dose (RfD). The approach uses a linear threshold (hockey stick) model of response and is based on the current system of uncertainty factors used in setting RfDs. The approach requires generally available toxicological estimates such as No-Observed-Adverse-Effect Levels (NOAELs) or Benchmark Doses and doses at which adverse effects are observed in 50% of the test animals (ED50s). In this approach, Monte Carlo analysis is used to characterize the uncertainty in the dose response slope based on the range and magnitude of the key sources of uncertainty in setting protective doses. The method does not require information on the shape of the dose response curve for specific chemicals, but is amenable to the inclusion of such data. The approach is applied to four compounds to produce estimates of response rates for dose rates greater than the RfD.

Categorical regression of toxicity data: a case study using aldicarb.

Categorical regression is a mathematical tool that can be adapted to estimate potential health risk from chemical exposures. By regressing ordered categories of toxic severity or pathological staging on exposure dose, this method can estimate the likelihood of observing any of the categories of severity at any dose level. Depending on the nature of the available data, these estimates can take the form of incidence rates for any of the categories in an exposed population or the probability of a new study conducted at a specified dose level being classified as one of the categories. Categorical regression is illustrated using toxicity data on aldicarb. For aldicarb, the data fall into three different groups: human clinical studies, dietary exposures in experimental animals, and accidental human exposure by contaminated crops. The U.S. EPA has assessed this literature and developed a reference dose (RfD) of 0.001 mg/kg-day. The results of applying categorical regression to data from human clinical studies suggests a maximum likelihood risk estimate of adverse effects of 0.008% at a 10-fold higher dose than the RfD when blood cholinesterase inhibition is not considered as an adverse effect. When blood cholinesterase inhibition of 20% or more is considered as an adverse effect, a maximum likelihood risk estimate of adverse effects is 0.1% at a dose 10-fold higher than the RfD.

Hexavalent chromium-contaminated soils: options for risk assessment and risk management.

Risk assessment involves establishing scientifically defensible dose-response relationships for end points of concern. For Cr(VI)-contaminated soils, this includes conducting dose-response assessments for blood, liver, and kidney toxicity following oral exposure; lung cancer following inhalation exposure; and allergic contact dermatitis following dermal exposure. This dose-response information is then integrated with a site-specific exposure assessment (or default assumptions) in order to develop a site-specific (or generic) soil criterion within the framework of a comprehensive risk characterization. Risk managers develop cleanup standards designed to protect against all possible adverse effects, taking into account these site-specific (or generic) criteria and other factors such as technical feasibility, cost-benefit analyses, and socio-political concerns. Recently a push for cost-benefit analyses of environmental decisions has occurred, further supporting the need for risk assessors to prepare a comprehensive risk characterization, with its attendant uncertainties. These risk assessment and management issues are brought to the forefront by risk assessors and risk managers dealing with Cr(VI)-contaminated soils. This article offers a review and analysis of the risk characterization of Cr(VI)-contaminated soils, showing that the differing toxicities with route of exposures do not necessarily lead to different characterizations or risk. Soil concentrations in the range of 130 to 450 ppm appear to protect against noncancer toxicity from oral exposure, cancer toxicity from inhalation exposure, and allergic contact dermatitis from dermal exposure.

Route-to-route extrapolation of the toxic potency of MTBE.

MTBE is a volatile organic compound used as an oxygenating agent in gasoline. Inhalation from fumes while refueling automobiles is the principle route of exposure for humans, and toxicity by this route has been well studied. Oral exposures to MTBE exist as well, primarily due to groundwater contamination from leaking stationary sources, such as underground storage tanks. Assessing the potential public health impacts of oral exposures to MTBE is problematic because drinking water studies do not exist for MTBE, and the few oil-gavage studies from which a risk assessment could be derived are limited. This paper evaluates the suitability of the MTBE database for conducting an inhalation route-to-oral route extrapolation of toxicity. This includes evaluating the similarity of critical effect between these two routes, quantifiable differences in absorption, distribution, metabolism, and excretion, and sufficiency of toxicity data by the inhalation route. We conclude that such an extrapolation is appropriate and have validated the extrapolation by finding comparable toxicity between a subchronic gavage oral bioassay and oral doses we extrapolate from a subchronic inhalation bioassay. Our results are extended to the 2-year inhalation toxicity study by Chun et al. (1992) in which rats were exposed to 0, 400, 3000, or 8000 ppm MTBE for 6 hr/d, 5 d/wk. We have estimated the equivalent oral doses to be 0, 130, 940, or 2700 mg/kg/d. These equivalent doses may be useful in conducting noncancer and cancer risk assessments.