Harmonization: Developing Consistent Guidelines for Applying Mode of Action and Dosimetry Information to Cancer and Noncancer Risk Assessment

The 1983 book, Risk Assessment in the Federal Government: Managing the Process, recommended developing consistent inference guidelines for cancer risk assessment. Over the last 15 years, extensive guidance have been provided for hazard assessment for cancer and other endpoints. However, as noted in several recent reports, much less progress has occurred in developing consistent guidelines for quantitative dose response assessment methodologies. This paper proposes an approach for dose response assessment guided by consideration of mode of action (pharmacodynamics) and tissue dosimetry (pharmacokinetics). As articulated here, this systematic process involves eight steps in which available information is integrated, leading first to quantitative analyses of dose response behaviors in the test species followed by quantitative analyses of relevant human exposures. The process should be equally appropriate for both cancer and noncancer endpoints. The eight steps describe the necessary procedures for incorporating mechanistic data and provide multiple options based upon the mode of action by which the chemical causes the toxicity. Given the range of issues involved in developing such a procedure, we have simply sketched the process, focusing on major approaches for using toxicological data and on major options; many details remain to be filled in. However, consistent with the revised carcinogen risk assessment guidance (USEPA, 1996c), we propose a process that would ultimately utilize biologically based or chemical specific pharmacokinetic and pharmacodynamic models as the backbone of these analyses. In the nearer term, these approaches will be combined with analysis of data using more empirical models including options intended for use in the absence of detailed information. A major emphasis in developing any harmonized process is distinguishing policy decisions from those decisions that are affected by the quality and quantity of toxicological data. Identification of data limitations also identifies areas where further study should reduce uncertainty in the final risk evaluations. A flexible dose response assessment procedure is needed to insure that sound toxicological study results are appropriately used to influence risk management decision-making and to encourage the conduct of toxicological studies oriented toward application for dose response assessments.     

An Evaluation of Risk Drivers from a Sample of Risk Assessments Conducted for the U.S. Army

We evaluate risk drivers at selected U.S. Army installations by developing a database containing contaminant-pathway-receptor combinations that exceed regulatory thresholds for ecological (toxicity quotient greater than one), human health cancer risk (predicted incremental lifetime cancer risk greater than one in ten thousand), and noncancer human health (hazard index greater than one). We compare the risk drivers from the database to reported corrective action objectives from available decision documents. For noncancer hazards, explosives (particularly in ground water) dominate the reported exceedances of regulatory thresholds in the database. PAHs in home-grown produce show the highest number of exceedances of regulatory thresholds for cancer risk. For ecological risks, PAHs in both terrestrial and aquatic environments dominate the exceedances of regulatory thresholds. All available cleanup levels were derived based on human health exposures rather than ecological exposures, except for one site. In general, ecological risks were considered to be “more uncertain,” and that was used as a basis for not relying on backcalculated target levels on the basis of ecological risk. The reverse was true for human health risks: the “conservative” assumptions incorporated into the modeling provided the justification for backcalculating health-protective target levels.     

Studies of the mortality of atomic bomb survivors. Report 12, part II. Noncancer mortality: 1950-1990.

This report updates the data on noncancer mortality for 86,572 atomic bomb survivors with dose estimates in the Radiation Effects Research Foundation's Life Span Study cohort. The primary analyses are based on more than 27,000 noncancer disease deaths that occurred in the cohort between October 1, 1950, and December 31, 1990, 30% more than in the previous report. The present analyses strengthen earlier findings of a statistically significant increase in noncancer disease death rates with radiation dose. Increasing trends are observed for diseases of the circulatory, digestive and respiratory systems. Rates for those exposed to 1 Sv are elevated about 10%, a relative increase that is considerably smaller than that for cancer. However, estimates of the number of radiation-related noncancer deaths in the cohort to date (140 to 280) are 50 to 100% of the number for solid cancer. The data do not yet clarify the shape of the dose response. There is no significant evidence against linearity, but the data are statistically consistent with curvilinear dose-response functions that posit essentially zero risk for doses below 0.5 Sv. Similarly, while the data are consistent with substantial variation in the excess relative risk with age at exposure or attained age, there is no statistically significant dependence on these factors. In view of the small relative risks and the lack of understanding of biological mechanisms, we emphasize consideration of whether the findings could be explained by misclassification, confounding or selection effects. Based on available data, we conclude that such factors are unlikely to fully explain the observed dose response. A significant dose response is also seen for deaths from blood diseases with an excess relative risk that is several times greater than that seen for solid cancer. Particular attention is paid to the possibility that this apparent effect is a consequence of the attribution of leukemia or other cancer deaths to noncancer blood diseases. We find that misclassification does not explain this excess risk. As in earlier reports, suicide rates tend to decrease with increasing dose.     

Treatments of Uncertainty and Variability in Ecological Risk Assessment of Single-Species Populations

The selection of the most appropriate model for an ecological risk assessment depends on the application, the data and resources available, the knowledge base of the assessor, the relevant endpoints, and the extent to which the model deals with uncertainty. Since ecological systems are highly variable and our knowledge of model input parameters is uncertain, it is important that models include treatments of uncertainty and variability, and that results are reported in this light. In this paper we discuss treatments of variation and uncertainty in a variety of population models. In ecological risk assessments, the risk relates to the probability of an adverse event in the context of environmental variation. Uncertainty relates to ignorance about parameter values, e.g., measurement error and systematic error. An assessment of the full distribution of risks, under variability and parameter uncertainty, will give the most comprehensive and flexible endpoint. In this paper we present the rationale behind probabilistic risk assessment, identify the sources of uncertainty relevant for risk assessment and provide an overview of a range of population models. While all of the models reviewed have some utility in ecology, some have more comprehensive treatments of uncertainty than others. We identify the models that allow probabilistic assessments and sensitivity analyses, and we offer recommendations for further developments that aim towards more comprehensive and reliable ecological risk assessments for populations.     

Noncancer Risk Assessment: A Probabilistic Alternative to Current Practice

Based on imperfect data and theory, agencies such as the United States Environmental Protection Agency (USEPA) currently derive “reference doses” (RfDs) to guide risk managers charged with ensuring that human exposures to chemicals are below population thresholds. The RfD for a chemical is typically reported as a single number, even though it is widely acknowledged that there are significant uncertainties inherent in the derivation of this number.

In this article, the authors propose a probabilistic alternative to the EPA's method that expresses the human population threshold as a probability distribution of values (rather than a single RfD value), taking into account the major sources of scientific uncertainty in such estimates. The approach is illustrated using much of the same data that USEPA uses to justify their current RfD procedure.

Like the EPA's approach, our approach recognizes the four key extrapolations that are necessary to define the human population threshold based on animal data: animal to human, human heterogeneity, LOAEL to NOAEL, and subchronic to chronic. Rather than using available data to define point estimates of “uncertainty factors” for these extrapolations, the proposed approach uses available data to define a probability distribution of adjustment factors. These initial characterizations of uncertainty can then be refined when more robust or specific data become available for a particular chemical or class of chemicals.

Quantitative characterization of uncertainty in noncancer risk assessment will be useful to risk managers who face complex trade-offs between control costs and protection of public health. The new approach can help decision-makers understand how much extra control cost must be expended to achieve a specified increase in confidence that the human population threshold is not being exceeded.     

Improving the Use of Toxicity Reference Values in Wildlife Food Chain Modeling and Ecological Risk Assessment

Ecological risk assessments often include mechanistic food chain models based on toxicity reference values (TRVs) and a hazard quotient approach. TRVs intended for screening purposes or as part of a larger weight-of-evidence (WOE) assessment are readily available. However, our experience suggests that food chain models using screening-level TRVs often form the primary basis for risk management at smaller industrial sites being redeveloped for residential or urban parkland uses. Iterative improvement of a food chain model or the incorporation of multiple lines of evidence for these sites are often impractical from a cost-benefit perspective when compared to remedial alternatives. We recommend risk assessors examine the assumptions and factors in the TRV derivation process, and where appropriate, modify the TRVs to improve their ecological relevance. Five areas where uncertainty likely contributes to excessively conservative hazard quotients are identified for consideration.     

Quantifying Uncertainty in Predictions of Invasiveness

Using the Australian weed risk assessment (WRA) model as an example, we applied a combination of bootstrapping and Bayesian techniques as a means for explicitly estimating the posterior probability of weediness as a function of an import risk assessment model screening score. Our approach provides estimates of uncertainty around model predictions, after correcting for verification bias arising from the original training dataset having a higher proportion of weed species than would be the norm, and incorporates uncertainty in current knowledge of the prior (base-rate) probability of weediness. The results confirm the high sensitivity of the posterior probability of weediness to the base-rate probability of weediness of plants proposed for importation, and demonstrate how uncertainty in this base-rate probability manifests itself in uncertainty surrounding predicted probabilities of weediness. This quantitative estimate of the weediness probability posed by taxa classified using the WRA model, including estimates of uncertainty around this probability for a given WRA score, would enable bio-economic modelling to contribute to the decision process, should this avenue be pursued. Regardless of whether or not this avenue is explored, the explicit estimates of uncertainty around weed classifications will enable managers to make better informed decisions regarding risk. When viewed in terms of likelihood of weed introduction, the current WRA model outcomes of ‘accept’, ‘further evaluate’, or ‘reject’, whilst not always accurate in terms of weed classification, appear consistent with a high expected cost of mistakenly introducing a weed. The methods presented have wider application to the quantitative prediction of invasive species for situations where the base-rate probability of invasiveness is subject to uncertainty, and the accuracy of the screening test imperfect     

The Joy Before Cooking: Preparing Ourselves to Write a Risk Research Recipe

Rather than the conventional practice of compiling a list of interesting research projects and then attempting to make the case that each represents a high priority, I will attempt an approach rooted in decision analysis. The information of greatest value, according to decision theory, is that which most enables us to make more reliable, transparent, and cost-effective decisions. Therefore, I begin with a brief discussion of how and why typical decisions relying on cancer dose-response information can fall short, in an attempt to assess where and how this aspect of risk assessment is “broken” before generating a list of research projects to “fix” it. I discuss the problem of model uncertainty in dose-response assessment, and conclude it is impossible to gauge how valuable it might be to know the correct model until we agree on guidelines for how to make decisions given imperfect information in this regard. After discussing four broad research areas that arguably represent particularly high priorities given this framework, I conclude by identifying three overarching areas of risk assessment and management that, if not given commensurate attention, threaten to render even perfect dose-response information of dubious value.     

Analysis and Portrayal of Uncertainty in a Food-Web Exposure Model

The results of quantitative risk assessments are key factors in a risk manager's decision of the necessity to implement actions to reduce risk. The extent of the uncertainty in the assessment will play a large part in the degree of confidence a risk manager has in the reported significance and probability of a given risk. The two main sources of uncertainty in such risk assessments are variability and incertitude. In this paper we use two methods, a second-order two-dimensional Monte Carlo analysis and probability bounds analysis, to investigate the impact of both types of uncertainty on the results of a food-web exposure model. We demonstrate how the full extent of uncertainty in a risk estimate can be fully portrayed in a way that is useful to risk managers. We show that probability bounds analysis is a useful tool for identifying the parameters that contribute the most to uncertainty in a risk estimate and how it can be used to complement established practices in risk assessment. We conclude by promoting the use of probability analysis in conjunction with Monte Carlo analyses as a method for checking how plausible Monte Carlo results are in the full context of uncertainty.     

A Strategy for Using Weight-of-Evidence Methods in Ecological Risk Assessments

Our review of existing approaches and regulatory uses of weight-of-evidence (WOE) methods suggested the need for a practical strategy for deploying WOE within a predictive ecological risk assessment (ERA). WOE is the process of considering strengths and weaknesses of various pieces of information in order to inform a decision being made among competing alternatives. A predictive ERA uses existing information relating cause and effect to estimate the probability that today's action X will lead to tomorrow's adverse outcome Y. There appears to be no practical guidance for use of WOE in predictive assessments. We therefore propose a strategy for using a WOE approach, within an ERA framework, to weigh and integrate outcomes from various lines of evidence to estimate the probability of an adverse outcome in an assessment endpoint. An ERA framework is necessary to connect the results of an assessment to the management goals of concern to decision-makers and stakeholders. Within that framework, a WOE approach provides a consistent and transparent means of interpreting the myriad types of data and information gathered during a complex ecological assessment. Impediments to application of WOE are discussed, including limited regulatory guidance, limited prior regulatory use, and persistent reliance on threshold-based decision-making.     

Probabilistic Environmental Risk Assessment for Dibutylphthalate (DBP)

In the risk assessment methods for new and existing chemicals in the European Union (EU), environmental “risk” is characterized by the deterministic quotient of exposure and effects (PEC/PNEC). From a scientific viewpoint, the uncertainty in the risk quotient should be accounted for explicitly in the decision making, which can be done in a probabilistic risk framework. To demonstrate the feasibility and benefits of such a framework, a sample risk assessment for an existing chemical (dibutylphthalate, DBP) is presented in this paper. The example shows a probabilistic framework to be feasible with relatively little extra effort; such a framework also provides more relevant information. The deterministic risk quotients turned out to be worst cases at generally higher than the 95th percentile of the probability distributions. Sensitivity analysis proves to be a powerful tool in identifying the main sources of uncertainty and thus will be effective for efficient further testing. The distributions assigned to the assess ment factors (derivation of the PNEC) dominate the total uncertainty in the risk assessment; uncertainties in the release estimates come second. Large uncertainties are an inherent part of risk assessment that we have to deal with quantitatively. However, the most appropriate way to characterise effects and risks requires further attention. Recommendations for further study are identified.     

Probabilistic Analysis of Human Health Risks Associated with Background Concentrations of Inorganic Arsenic: Use of a Margin of Exposure Approach

Substantial evidence exists from epidemiological and mechanistic studies supporting a sublinear or threshold dose–response relationship for the carcinogenicity of ingested arsenic; nonetheless, current regulatory agency evaluations have quantified arsenic risks using default, generic risk assessment procedures that assume a linear, no-threshold dose–response relationship. The resulting slope factors predict risks from U.S. background arsenic exposures that exceed certain regulatory levels of concern, an outcome that presents challenges for risk communication and risk management decisions. To better reflect the available scientific evidence, this article presents the results of a Margin of Exposure (MOE) analysis to characterize risks associated with typical and high-end background exposures of the U.S. population to arsenic from food, water, and soil. MOE values were calculated by comparing a no-observable-adverse-effect-level (NOAEL) derived from the epidemiological literature with exposure estimates generated using a probabilistic (Monte Carlo) model. The plausibility and conservative nature of the exposure and risk estimates evaluated in this analysis are supported by sensitivity and uncertainty analyses and by comparing predicted urinary arsenic concentrations with empirical data. Using the more scientifically supported MOE approach, the analysis presented in this article indicates that typical and high-end background exposures to inorganic arsenic in U.S. populations do not present elevated risks of carcinogenicity.     

Quantifying uncertainty in predictions of invasiveness, with emphasis on weed risk assessment

Using the Australian Weed Risk Assessment (WRA) model as an example, we applied a combination of bootstrapping and Bayesian techniques as a means of explicitly estimating the posterior probability of weediness as a function of an import risk assessment model screening score. Our approach provides estimates of uncertainty around model predictions, after correcting for verification bias arising from the original training dataset having a higher proportion of weed species than would be the norm, and incorporates uncertainty in current knowledge of the prior (base-rate) probability of weediness. The results confirm the high sensitivity of the posterior probability of weediness to the base-rate probability of weediness of plants proposed for importation, and demonstrate how uncertainty in this base-rate probability manifests itself in uncertainty surrounding predicted probabilities of weediness. This quantitative estimate of the weediness probability posed by taxa classified using the WRA model, including estimates of uncertainty around this probability for a given WRA score, would enable bio-economic modelling to contribute to the decision process, should this avenue be pursued. Regardless of whether or not this avenue is explored, the explicit estimates of uncertainty around weed classifications will enable managers to make better informed decisions regarding risk. When viewed in terms of likelihood of weed introduction, the current WRA model outcomes of ‘accept’, ‘further evaluate’ or ‘reject’, whilst not always accurate in terms of weed classification, appear consistent with a high-expected cost of mistakenly introducing a weed. The methods presented have wider application to the quantitative prediction of invasive species for situations where the base-rate probability of invasiveness is subject to uncertainty, and the accuracy of the screening test imperfect.     

Improving Risk Assessment: Priorities for Epidemiologic Research

The Epidemiology Work Group at the Workshop on Future Research for Improving Risk Assessment Methods, Of Mice, Men, and Models, held August 16 to 18, 2000, at Snowmass Village, Aspen, Colorado, concluded that in order to improve the utility of epidemiologic studies for risk assessment, methodologic research is needed in the following areas: (1) aspects of epidemiologic study designs that affect doseresponse estimation; (2) alternative methods for estimating dose in human studies; and (3) refined methods for dose-response modeling for epidemiologic data. Needed research in aspects of epidemiologic study design includes recognition and control of study biases, identification of susceptible subpopulations, choice of exposure metrics, and choice of epidemiologic risk parameters. Much of this research can be done with existing data. Research needed to improve determinants of dose in human studies includes additional individual-level data (e.g., diet, co-morbidity), development of more extensive human data for physiologically based pharmacokinetic (PBPK) dose modeling, tissue registries to increase the availability of tissue for studies of exposure/dose and susceptibility biomarkers, and biomarker data to assess exposures in humans and animals. Research needed on dose-response modeling of human studies includes more widespread application of flexible statistical methods (e.g., general additive models), development of methods to compensate for epidemiologic bias in dose-response models, improved biological models using human data, and evaluation of the benchmark dose using human data. There was consensus among the Work Group that, whereas most prior risk assessments have focused on cancer, there is a growing need for applications to other health outcomes. Developmental and reproductive effects, injuries, respiratory disease, and cardiovascular disease were identified as especially high priorities for research. It was also a consensus view that epidemiologists, industrial hygienists, and other scientists focusing on human data need to play a stronger role throughout the risk assessment process. Finally, the group agreed that there was a need to improve risk communication, particularly on uncertainty inherent in risk assessments that use epidemiologic data.     

Probabilistic Sensitivity Analysis of Public Health Risk Assessment from Contaminated Soil

Recently, there has been a growing trend toward using stochastic (probabilistic) methods in ecological and public health risk assessment. These methods are favored because they overcome the problem of compounded conservatism and allow the systematic consideration of uncertainty and variability typically encountered in risk assessment. This article demonstrates a new methodology for the analysis of uncertainty in risk assessment using the first-order reliability method (FORM). The reliability method is formulated such that the probability that incremental lifetime cancer risk exceeds a predefined threshold level is calculated. Furthermore, the stochastic sensitivity of this probability with respect to the random variables is provided. The emphasis is on exploring the different types of probabilistic sensitivity obtained through the reliability analysis. The method is applied to a case study given by Thompson et al. (1992) on cancer risk resulting from dermal contact with benzo(a)pyrene (BaP)-contaminated soils. The reliability results matched those of the Monte Carlo simulation method. On average, the Monte Carlo simulation method required about 35 times as many function evaluations as that of FORM to calculate the probability of exceeding the target risk level. The analysis emphasizes the significant impact that the uncertainty in cancer potency factor has on the probabilistic modeling results compared with other parameters.     

Adequacy Conditions for Reporting Uncertainty in Chemical Risk Assessments

Uncertainty may influence decision-making. A prerequisite for a decision to be well founded is thus that scientific experts inform decision-makers about all decision relevant uncertainty. A set of conditions is provided for adequate characterization of scientific uncertainty for the purposes of regulatory decision-making. These conditions require specification of (1) the character and degree of uncertainty about the assessment variables, (2) the possibility of reducing the uncertainty, and (3) the degree of agreement among experts. Furthermore, it is required that (4) the information covered by the previous conditions is presented in a clear and comprehensible way. The point of departure is that characterizing scientific uncertainty conceptually means specifying all potentially important possibilities that are consistent with the state of scientific knowledge. The conditions are intended to be applied to human health risk assessment of chemicals. However, the basic approach, to consider potentially important possibilities, should be useful also to environmental, and site-specific risk assessment.     

Scientific data and theories for salmonellosis dose–response assessment

An “expansive” risk assessment approach is illustrated, characterizing dose–response relationships for salmonellosis in light of the full body of evidence for human and murine superorganisms. Risk assessments often require analysis of costs and benefits for supporting public health decisions. Decision-makers and the public need to understand uncertainty in such analyses for two reasons. Uncertainty analyses provide a range of possibilities within a framework of present scientific knowledge, thus helping to avoid undesirable consequences associated with the selected policies. And, it encourages the risk assessors to scrutinize all available data and models, thus helping avoid subjective or systematic errors. Without the full analysis of uncertainty, decisions could be biased by judgments based solely on default assumptions, beliefs, and statistical analyses of selected correlative data. Alternative data and theories that incorporate variability and heterogeneity for the human and murine superorganisms, particularly colonization resistance, are emerging as major influences for microbial risk assessment. Salmonellosis risk assessments are often based on conservative default models derived from selected sets of outbreak data that overestimate illness. Consequently, the full extent of uncertainty of estimates of annual number of illnesses is not incorporated in risk assessments and the presently used models may be incorrect.     

A Probabilistic Human Health Risk Assessment for Environmental Exposure to Dibutylphthalate

In the European Union, Directive 92/32/EC and EC Council Regulation (EC) 793/93 require the risk assessment of industrial chemicals. In this framework, it is agreed to characterise the level of “risk” by means of the deterministic quotient of exposure and effects parameters. Decision makers require that the uncertainty in the risk assessment be accounted for as explicitly as possible. Therefore, this paper intends to show the advantages and possibilities of a probabilistic human health risk assessment of an industrial chemical, dibutylphthalate (DBP). The risk assessment is based on non-cancer endpoints assumed to have a threshold for toxicity. This example risk assessment shows that a probabilistic risk assessment in the EU framework covering both the exposure and the effects assessment is feasible with currently available techniques. It shows the possibility of comparing the various uncertainties involved in a typical risk assessment, including the uncertainty in the exposure estimate, the uncertainty in the effect parameter, and the uncertainty in assessment factors used in the extrapolation from experimental animals to sensitive human beings. The analysis done did not confirm the reasonable worst-case character of the deterministic EU-assessment of DBP. Sensitivity analysis revealed the extrapolation procedure in the human effects assessment to be the main source of uncertainty. Since the probabilistic approach allows determination of the range of possible outcomes and their likelihood, it better informs both risk assessors and risk managers.     

Noncancer risk assessment of arsenic safe and unsafe water uses: Bayesian estimation on cohort study

Cancer and noncancer risk of arsenic exposure depends on arsenic intake through drinking water and diets. The present study evaluated the probability of noncancer effects of arsenic exposure from drinking water and diets in a cohort of 82 participants in arsenic-endemic rural areas, considering arsenic-safe and arsenic-unsafe water uses for three consecutive years. The risk assessment included the collection of last 24 hours' diet replica and urine of the participants followed by total arsenic analysis of the same. Toxic dose emerging from exposure duration is a nonlinear variable. So, Bayesian estimation of the data for noncancer risk assessment of the variable arsenic consumption was performed. In spite of using arsenic-safe water, we observed arsenic consumption and release. Participants with skin lesions had more arsenic in urine than participants without skin lesions. Future risk for participants without skin lesions was twice due to less arsenic release in urine. For the first time, Bayesian simulation was used to assess noncancer risk on a cohort for a consecutive three-year study. A significant finding was the higher assessed noncancer risk of the participants without skin lesions than the participants with skin lesions.