A Framework for Risk Characterization of Environmental Pollutants

Transcription

1 Journal of the Air & Waste Management Association ISSN: (Print) (Online) Journal homepage: A Framework for Risk Characterization of Environmental Pollutants Dennis F. Naugle & Terrence K. Pierson To cite this article: Dennis F. Naugle & Terrence K. Pierson (1991) A Framework for Risk Characterization of Environmental Pollutants, Journal of the Air & Waste Management Association, 41:10, , DOI: / To link to this article: Published online: 06 Mar Submit your article to this journal Article views: 170 View related articles Citing articles: 2 View citing articles Full Terms & Conditions of access and use can be found at Download by: [ ] Date: 10 January 2018, At: 18:08

2 Dennis F. Naugle and Terrence K. Pierson Research Triangle Institute Research Triangle Park, North Carolina Risk characterization is defined by both the U.S. National Academy of Sciences and the U.S. EPA as the estimation of human health risk due to harmful (i.e., toxic or carcinogenic) substances or organisms. Risk characterization studies are accomplished by integrating quantitative exposure estimates and dose-response relationships with the qualitative results of hazard identification. A Risk Characterization Framework has been developed to encourage a systematic approach for analysis and presentation of risk estimates. This methodology subdivides the four common components of the risk assessment process into ten elements. Each of these elements is based on a term in a predictive risk equation. The equation allows independent computations of exposure, dose, lifetime individual risk, and risk to affected populations. All key assumptions in the predictive risk equation can be explicitly shown. This is important to understand the basis and inherent uncertainties of the risk estimation process. The systematic treatment of each of the ten elements in this framework aids in the difficult job of comparing risk estimates by different researchers using different methodologies. The Risk Characterization Framework has been applied to various indoor and outdoor air pollutants of a carcinogenic nature. With further development, it also promises to be applicable to noncarcinogenic effects. The Research Strategies Committee of the U.S. Environmental Protection Agency (EPA) Science Advisory Board (SAB) concluded that "EPA needs to reshape its strategy for addressing environmental problems in the next decade and beyond. In addition to the current emphasis on federallymandated controls that are put in place to clean up pollutants after they have been generated, the Agency must develop a strategy that emphasizes the reduction of pollution before it is generated. A strategic shift in emphasis from control and clean-up to anticipation and prevention is absolutely essential to our future physical, environmental, and economic health." 1 Implementation of this concept requires a shift from "end-of-pipe" controls which were used effectively in the 1970's under the Clean Air Act to alternative, strategies. While the SAB comments above are generally intended for outdoor environmental issues, pollution prevention and risk assessment techniques are just as likely (perhaps even more likely) to play a major role in future indoor air quality strategies. A recent EPA report 2 suggests that indoor air risks are higher than most other environmental risks. It suggests that priorities for EPA action should depend more on these current risks and less on the perceived environmental problems of the past. Programs used successfully to mitigate pollution outdoors may not be appropriate paradigms for indoor air quality strategies. Indoor air quality (IAQ) strategies will likely focus on decentralized programs where risk reduction is a prime objective. Such strategies emphasize materials substitution over air cleaning; localized mitigation options over rigid federal regulations; actions by many professional organizations such as the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE), and the American Society of Testing and Materials (ASTM), and federal agencies such as the Department of Energy (DOE), and the Consumer Product Safety Commission (CPSC) to supplement the federal EPA programs; and by mitigation decisions (within the above and other organizations, local governments, or by individual homeowners) motivated by the reduction of health risks. Increased emphasis on indoor air issues is likely to emerge with the realization that pollutant concentrations in private homes or other indoor environments are generally greater than air quality levels measured outdoors. 3 " 5 Sources of Indoor Air Pollutants The growing number of complaints about commercial and residential indoor air quality are associated with multiple pollutants emanating from multiple sources. Figure 1 illustrates many common sources of indoor air pollutants. This matrix shows 15 separate categories frequently identified in the literature as significant sources of the 14 types of indoor air pollutants shown. Note that several of the "sources" are very broad categories such as building mate J. Air Waste Manage. Assoc.

3 Sources AC Systems Outdoor Air Building Materials gss SSSi iib 'J5K5! 333! Copying Machines Earth or Ground (gases and particles) Furnishings Kerosene Heaters Gas Stoves Gas Heaters Consumer Products Insulation Moist Materials Tobacco Smoke Vehicle Exhaust Woodstoves Figure 1. Some major sources of indoor air pollution. rials and consumer products. Likewise, some of the "pollutants" such as volatile organics and biological agents cover a broad range which could include hundreds of chemical species or biological organisms. Figure 1 helps illustrate the dilemma of someone analyzing a "sick building" complaint or the cancer-causing potential of an indoor environment. A scientifically thorough investigation of an individual's environment might entail extensive chemical and biological analyses to determine the principal sources causing elevated concentrations which should be mitigated. Add to this complexity the fact that individual exposures have contributions from multiple microenvironments (e.g., indoors in home and work settings, transportation settings, etc.). Furthermore, temporal as well as spatial considerations are important since day to day variabilities are inevitable as people vary their microenvironments and as the microenvironments themselves change with different weather and source conditions. Even under similar environmental conditions, health responses will vary with individual susceptibility. The characterization of health risks is therefore a technically complex and multifactorial process. The Role of Risk Characterizations The relationships between basic environmental research, integration studies, such as risk assessments, and risk management, are shown in Figure 2. A few of the key research areas which provide the foundation to manage environmental risks are illustrated in the left column. This research is appropriately conducted in great technical depth and replicated by independent projects. For this reason, *e.g., Molds, spores, and mites. integration studies are also needed to combine individual research projects covering different technical disciplines and to draw general technical conclusions on both the risk to people and the alternatives for mitigation. These integration studies provide technical inputs for risk management options and decisions. Other inputs to risk management decisions include economic considerations and technical and political feasibility for implementing a particular option. Note in Figure 2 that any level decision-maker is considered "management," depending on the specific IAQ decisions at issue. For example, the most important decisionmakers for radon and many other indoor air pollutants are individual homeowners and state and local environmental managers. Involvement by federal officials is also required, particularly in research and risk communication areas. Mitigation of naturally occurring radon will be done only if millions of individual decision-makers are convinced the health risks are serious enough to motivate some sort of mitigation action. On the other hand, federal and state regulations have been enacted to address several indoor air pollutants, such as asbestos and lead, through controls on the sources for these pollutants. Risk characterizations can have a major impact on (and should be integral to) the risk management process. To fulfill this potential, risk characterization studies must be: 1. Available. Numerical risk characterization studies, as defined by the National Research Council 6 and in the EPA Risk Assessment Guidelines Document of 1986, 7 are relatively new and are often not available for keyenvironmental decisions at all levels. October 1991 Volume 41, No

4 RESEARCH AREA»,«INTEGRATION STUDIES» RISK MANAGEMENT Source measurements and models Mitigation Alternatives ~- Laboratory and field observations of adverse health effects and exposures Extrapolation methods for high to low dose and animal to human Source and environmental field measurements, personal exposure monitors, estimated exposures, characterization of populations I RISK ASSESSMENTS _ ^- Management* Development \ \ 1 r 1 Consideration c>f health, economic, soci. al, political consequences of management* c>ptions A i r Management* decisions and actions Figure 2. Relationship between research and IAQ risk management. Management can be within federal agencies, state and local agencies, industries or even households. 2. Technically defensible. A decision-maker must be convinced that the complex technical components of a risk characterization study are based on sound scientific data. Simplifying assumptions made in the absence of scientific data should be thoroughly considered. Uncertainties in the risk estimates must be communicated to decision makers and to the public. 3. Understandable. Efforts to avoid simplifying assumptions to make risk estimates more realistic will inevitably lead to greater technical complexity. This complexity can increase to a point where only a handful of scientists are likely to judge the study results. A balance between 1, 2, and 3 is therefore needed for risk characterizations to be best utilized. Traditional Risk Assessments The U.S. Environmental Protection agency and other federal agencies have been in the forefront of developing risk assessment methodologies that provide a mechanism for incorporating scientific evidence and scientific judgement into the decision-making process. In order to insure high technical quality and consistency in Agency-sponsored risk assessment efforts, the EPA has issued guidelines for risk assessment relating to carcinogenicity, mutagenicity, chemical mixtures, suspect developmental toxicants, and estimating exposures. 7 These guidelines outline a general approach to be used in developing analyses of health risks. They emphasize that such analyses will be conducted on a case-by-case basis with full consideration given to all pertinent scientific and technical information. Risk assessment is defined as the overall procedure by which potential adverse health effects of human exposure to toxic agents are characterized. It includes the following four components: hazard identification, exposure assessment, dose-response assessment, and risk characterization. 6 ' 7 Figure 3 shows the relationship of these traditional components of risk assessment. Hazard identification is a qualitative determination of whether human exposure to an agent has the potential to produce adverse health effects. It involves an evaluation of all available toxicology data and other relevant biological and chemical information for the agent under consideration. The results of any hazard identification should include a summary of all key findings of the qualitative assessment and the rationale that forms the basis for the conclusion reached. All assumptions, uncertainties in the evidence, and other factors that may affect the relevance of evidence to humans should be presented. The EPA has developed a classification system based on the strength or weight of the underlying evidence for carcinogenicity, systemic toxicity and reproductive toxicity. Exposure assessment involves the characterization of the nature and site of various populations exposed to a toxic agent, and the quantitative estimation of the level and duration of their exposures. Current methods of exposure assessment are primarily medium- or route-specific, so that at present there is no single approach that is appropriate for all cases. There is a need to evaluate (at least to a first order) all routes and media of exposure. However, there is not much point in elaborate risk studies of a route and medium which contributes only 1 percent or 10 percent of the total exposure. Appropriate methods must be selected on a case-by-case basis, depending on the available data and the level of sophistication required. In general, exposure assessment consists of four steps: 1. Determination of environmental concentrations or personal monitoring; 2. Estimation of the magnitude, duration and frequency of human exposure, which are representative of the subject populations. 3. Estimation of dose received, which can be expressed as Maximum Daily Dose (MDD) for acute, subchronic or chronic exposures to noncarcinogens, or as Lifetime Average Daily Dose (LADD) for carcinogens; and 4. Characterization of exposed populations and individuals (e.g., by magnitude of dose received, age, sex, health status, time in life of exposure, other concurrent exposures) and identification of subpopulations with heightened sensitivity. Dose-response assessment is a quantitative process that involves: 1. Modeling the relation between the administered or 1300 J. Air Waste Manage. Assoc.

5 HAZARD IDENTIFICATION (Does the agent cause adverse effects?) Data analysis relating chemical and exposure to disease produced. Characterization of chemical behavior within body. Inference whether toxic effects in one setting (e.g., animals) will occur in other settings (e.g.. humans). DOSE-RESPONSE EVALUATION (What is the relationship between dose and incidence in humans?) A quantitative description relating the amount of exposure (or delivered dose) to the extent of injury or disease. RISK CHARACTERIZATION (What is the estimated Incidence of the adverse effect in a given population?) A numerical estimate of the individual probabilities of an adverse effect based on estimated exposure and dose-response factors. A numerical estimate of the number of cases of the adverse effect in the exposed population. A discussion of assumptions and uncertainties in the risk estimate. EXPOSURE ASSESSMENT (What exposures are currently experienced or anticipated under different conditions?) A quantitative description relating the magnitude and duration of concentrations to the size and nature of the population exposed. Figure 3. Traditional components of risk assessment. received dose of a substance and the incidence of an adverse health effect in an exposed population; and 2. Using the dose-response model to estimate the probability of occurrence of the effect based on human exposure to the substance. If available, dose-response estimates based on adequate human data are preferable to those derived from animal data. In the absence of appropriate human studies (which is usually the case), data from studies of animal species that respond most like humans should also be used. Risk characterization is denned as the quantitative estimation of human health risk due to exposure to harmful substances or organisms. This characterization builds upon each of the first three components of the risk assessment process. Ideally, all components are evaluated in the most rigorous and scientifically defensible way, with all assumptions and uncertainties clearly documented in order to obtain the most reliable quantitative estimates of human health risk. The most complete risk characterization is a four-step procedure involving: 1. Derivation of a point estimate of risk for individuals (i.e., estimate of risk based on the exposure assessment); 2. Derivation of a point estimate or range of risk for a given population; 3. Estimation of the uncertainty ranges associated with point estimates of risk; and 4. Aggregation of point estimates of risk (and estimates of uncertainty) across populations, routes of exposure, and/or pollutant type. The existing methods for risk characterization are far from ideal, largely because of data limitations and incomplete knowledge of the action of toxic agents on the human body. Sources of uncertainty in estimating risks of adverse health effects include those due to statistical sampling issues, exposure or dose-response models, and input parameters for these models. Predictive Risk Equation A detailed review of studies which characterize risk attributed to indoor air pollutants has been conducted. 8 Not surprisingly, these studies utilized a variety of methodologies. Analysis of the common aspects of these studies leads to the formulating of a simple predictive risk equation which could be used to express most if not all of the key elements utilized. The four general equations that comprise the Predictive Risk Equation are presented in Figure 4. Representative units are shown and are useful to illustrate dimensional consistency. Although the general equations are most relevant to carcinogenic risk assessments, they can be modified to address noncarcinogenic effects. Each of the elements of the Predictive Risk Equation are discussed in the next section. A Framework For Risk Characterization The traditional Risk Assessment approaches were combined with elements of the Predictive Risk Equation to produce the Risk Characterization Framework as presented in Figure 5. It has been developed to encourage a systematic approach for analysis and presentation of risk characterization study results. This framework subdivides the four components of the risk assessment process (as defined by EPA and the National Research Council) into ten elements (Columns B through K in Figure 5) to provide a refined and more systematic way of describing a very complex risk estimation process. Brief descriptions of each column are described here and further described elsewhere. 9 Column B (Source Factors) is the starting point for the predictive risk equation. The estimation of risk can be based on the study of a single source emitting one or more pollutants of concern or the study of a pollutant or mixture that is emitted from one or more sources. Column B, although descriptive in nature, provides information on the source(s) of the pollutant(s) under study. Column C (Pollutant Concentration) of the framework records the numerical data of exposure concentrations for each pollutant under study. Pollutants in the indoor environment include respirable particles, environmental tobacco smoke, radon, asbestos, certain organic and inorganic compounds, and biological agents. Some of these exist as gases or vapors, some as fibers or particulate material, some are absorbed onto suspended particulate material, and October 1991 Volume 41, No

6 Source _K ' Pollutant Exposure ' Dosimetry I Response Lifetime > Exposed Risk to Exposed Factors "^Concentration x Duration/Setting=Exposure x Factors = Dose x Factor = Individual Risk x Population = Population 1. Exposure Pollutant Concentration x Exposure Duration = Exposure 2. Dose Exposure (mg/m 3 x days) (mglm 3 ) x (days) Dosimetry Factors (contact rate: m 3 /day) (absorption rate: %) (inverse average body weight: -r ) or Kg inverse days in lifetime: days (mglm 3 x days) = Dose / rng I kg x day 3. Lifetime Individual Risk Dose x Dose-Response Relationship = Lifetime Individual Risk mg kg x day mg ( probability lifetime probability lifetime Total exposure over a lifetime Average dose per day over a lifetime Individual risk over a lifetime 4. Risk to Exposed Population Lifetime Individual Risk Figure 4. I probability \ lifetime I Exposed Population # people x 1 lifetime # years Elements of the predictive risk equation. some are distributed between a vapor-phase and a particlebound state. 10 The estimation of the pollutant concentrations will be based on different types of data depending on the focus of the study. Methods for estimating the concentration of indoor pollutants include direct measurement through sampling and analysis, modeling, analysis of biological markers, and questionnaires. Measurements can be taken at specific points in a house to determine a time weighted averaged concentration for a particular room or the house as a whole. Alternatively, individuals can wear personal exposure monitors (PEM) for some specified period of time. When used in combination with stationary monitors, PEMs can provide a profile of total exposure partitioned into separate microenvironments. Modeling of exposure requires data on source emission rates, ventilation and infiltration, removal by adsorption onto surfaces, mixing, volume of space in which exposure occurs, and of activity patterns of individuals in each of the environments being modeled. Models are particularly useful in making exposure estimates in temporal and spatial regimes where measured concentrations are not available and in relating exposure concentrations to particular sources. Thorough evaluation ("validation") of new models is extremely important. Biological markers result from the analysis of the physiological fluids of exposed persons. For example, the presence of nicotine and its major metabolite, cotinine, in biological fluids is typically a good indicator of exposure to tobacco, tobacco smoke, or environmental tobacco smoke. 11 The determination of nicotine and cotinine in saliva, blood, or urine of active or passive smokers can be used effectively to estimate the level of exposure to these substances. Questionnaires are useful tools for assessment of exposure to indoor air pollutants, especially with respect to time activity patterns of individuals in specific environments and their estimate of exposure (e.g., number of cigarettes smoked at home each day). Developing questions that elicit unambiguous replies and using replies to properly make quantitative estimates of exposure are especially important. Risk to Exposed Population # cases year Estimated cases in exposed population per year Column D (Exposure Duration and Setting) of the framework combines the identification of each environmental setting in which exposure occurs and an estimation of the time spent in that environment. The duration of exposure is expressed differently for different health effects of concern. For carcinogenic effects, duration is typically expressed as total hours or days of exposure during one's lifetime. For example, if an individual is exposed every day of their life then exposure would be 25,550 days, assuming a 70-year lifetime. For noncarcinogenic effects, one is also concerned about acute health effects from short-term exposures at elevated concentrations. Short-term durations of hours or even minutes may be important. Whenever possible, Column E {Exposure) should be expressed as the product of the concentration of pollutant to which one is exposed in a particular setting times some specific time period. This presentation follows the familiar equation used in other recent exposure models 12 which states: i = 2 where: Ej = integrated pollutant exposure of person i (Framework Column E), Cj = pollutant concentration (Framework Column C) encountered in microenvironment j, tjj = time spent by person i in microenvironment j (Framework Column D), and J = number of microenvironments considered. Expressed in the Framework with a new row of numbers for each population (Framework Column J). A chronic exposure is assumed to be relatively constant over a long period of time and thus, what is typically calculated is an average daily pollutant concentration to which one is exposed. This approach is standard for calculating lifetime exposures to carcinogens and assumes sources and patterns of living do not change over a lifetime J. Air Waste Manage. Assoc.

7 ** i sgenic ncy O B n o c r cinoge shold c ie o c 5 o a o% P c «S (5 UJ 1a "?.i O E COQ 2 2??' W J X o III JE. ^ *I >5 Ilfllpiil I* Its Q.»!! ll o«"8? October 1991 Volume 41, No

8 However, when dealing with less than lifetime exposures (i.e., acute or subchronic) and an array of noncarcinogenic endpoints which are dependent on both pollutant concentration and duration of exposure, it is important to make both Column C and Column D factors explicit. The terms exposure and dose are sometimes used interchangeably. However, the framework distinguishes between these two terms by specifically incorporating exposure duration and by providing a set of dosimetry factors that relate the estimated exposure to the dose received by an individual. As shown in Column F (Dosimetry Factor) of the framework, and previously illustrated in Figure 4, these factors include: contact rate, absorption rate, average body weight, average lifetime, regional surface area of the lung, and regional deposited dose ratio. Other factors may also be required in specific analyses. Column G (Dose) of the framework is the product of exposure estimates and the variety of modifiers discussed above. Dose is expressed as pollutant mass per kilogram of body weight per day (mg/kg-d). For toxicological studies involving animals under controlled conditions, the pollutant dose can be directly measured. For epidemiological studies correlating human disorders with exposures, the dose must be estimated. The Column H of the framework (Response Factor) describes the magnitude of the response of an individual to a given dose of the substance. For practical reasons, human observational data are usually available for few of the different possible chemicals and exposure routes. It is therefore necessary to derive mathematical models of the dose-response relationship based on the best understanding of the mechanism of action of the toxic substance. The dose-response relationship for carcinogens is usually expressed as a potency factor defined as the 95 percent upper confidence limit of the human excess lifetime cancer risk associated with one unit of lifetime exposure to the carcinogen measured in units of (mg/kg-day)" 1. Potency factors have been estimated by EPA's Human Health Assessment Group (HHAG) for many carcinogens. However, there are also many substances suspected of being human carcinogens for which HHAG has not estimated a potency factor due primarily to a lack of adequate data. Methodologies to estimate the chance of one or more toxic endpoints are much less developed than carcinogenic risk assessments. The U.S. EPA has used the concept of a reference dose (RfD) rather than potency factors for noncarcinogenic health effects. 13 The RfD 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 appreciable risk of deleterious effects during a lifetime. 14 The RfD is determined by use of the following equation: RfD = NOAEL/CUF x MF) where: NOAEL = the no observed adverse effect level, using either human or animal data, UF = uncertainty factors that reflect the uncertainty in various types of data used to estimate RfDs, and MF = modifying factors which reflect qualitative professional judgements regarding scientific uncertainties of the entire database of the chemical. The basis for developing an RfD is the determination of the NOAEL in adequately designed and well conducted epidemiology or toxicology studies. In the absence of sufficient data for determining the NOAEL for the chosen toxic endpoint, a lowest observed adverse effect level (LOAEL) can be used. Although RfDs are currently used for chronic exposures via a particular route of exposure, similar thresholds could be developed for subchronic and acute exposures. Additionally, dose-response relationships at exposure levels above the threshold value have also been estimated and used to derive estimates of adverse health risks or cases. 15 ~ 18 - In their simplest form, carcinogenic risks are estimated by multiplying a lifetime average daily dose (Column G) in units of milligrams per kilogram of body weight per day, by a potency factor (Column H), in units of lifetime risk per unit of exposure, (mg/kg-day)" 1. Column I (Lifetime Individual Risk) of the framework gives estimates of the lifetime excess risk of cancer for an individual exposed at the given lifetime exposure. Alternatively, biologically based dose-response relationships can be used to estimate the cancer risks from different lifetime exposure profiles, taking into account ages at different exposure levels. For noncarcinogenic agents, it is usually assumed that there is a threshold dose below which there is no effect. The ratio of the exposure level to the threshold dose gives some indication of the likelihood of occurrence of the adverse health effects associated with exposure to the toxic substance. Threshold-based doses are most commonly established for chronic exposures, but may also be established for acute and subchronic exposures. Risk characterization also involves a description of the uncertainty associated with any risk estimates. This description may be qualitative, quantitative or both depending on the scope of the project and the availability of resources. Sources of uncertainty in risk assessment include statistical sampling issues concerning environmental data, doseresponse models and their input parameters, and incomplete understanding of the biological cause and effect relationship with a pollutant or mixture. Column J (Exposed Population) of the framework is provided for the types and number of affected subpopulations included in a risk analysis. A numerical estimate of the number of individuals in each subpopulation is therefore required. Also, each subpopulation must be linked to a specific exposure scenario. Ideally, in characterizing an exposed population or subpopulation, a distribution should be obtained that incorporates variability associated with age of the population, exposure levels associated with different activity patterns and microenvironments, and the susceptibility within the population to a specific effect. Column K of the framework (Risk to Exposed Population) is typically expressed as the expected or observed number of cases in the population. Risks estimated can be in a deterministic fashion with a predictive risk equation as described in Figure 4. Alternatively, when sufficient human data allows, a statistical analysis of epidemiologic data can be used to obtain a risk estimate in Column K. In the latter case, exposures (Column E) related to sources (Column B) must then be "backward calculated" (going from the right side of the framework to the left side) in order to suggest a cause and effect relationship. Discussion of Uncertainties Characterization of risk often depends on the data available and the way the analyst collects and organizes these data. The more common risk characterization estimates require a two-step procedure whereby a single point estimate is calculated for individual lifetime risk for an exposed population and point estimates aggregated across all exposed populations to get a population risk. Uncertainty in estimates of both individual lifetime risk and population risk should be addressed either qualitatively or quantitatively. The existing methods for risk characterization are far from ideal, largely because of data limitations and incomplete knowledge of the biological mechanisms of action of toxic agents on the human body. There are four major 1304 J. Air Waste Manage. Assoc.

9 sources of uncertainty in estimating point risks of adverse health effects: 1. Uncertainty due to statistical sampling issues; 2. Uncertainty in the exposure or dose-response models; 3. Uncertainty in the input parameters for these models; and, 4. Uncertainty due to lack of completeness in the models. At each point in the risk characterization process where an uncertainty exists, an assumption or scientific judgement must be made in lieu of firm scientific evidence. Specific examples of uncertainties that require the insertion of some assumption into the risk characterization process include those related to: Specification of an exposure scenario; Extrapolation from animal to human exposures; Extrapolation from high to low doses; and Extrapolation from one route of exposure to another. Under ideal circumstances each uncertainty should be assigned individual and joint probability distributions, from which average or worst case point estimates of adverse health effects are generated. In addition, these probability distributions should be used to perform a sensitivity analysis for the point estimates for each input parameter and to generate overall probability distribution for the estimated risks. Although considerable uncertainties exist in risk assessment methodologies and their applications, risk assessment has been adopted by government agencies in the United States and elsewhere to provide a quantitative and consistent framework for systematically evaluating environmental health risks and options for their control. The major criticisms of the methodology are that the data requirements are both time and resource intensive and that scientific judgement is often presented as scientific fact. The information needed to quantify exposure does not exist and, in some instances, the availability of data may not even support logical assumptions to be made regarding exposure. Second, scientific judgement embedded in all risk assessments needs to be understood and communicated to decision makers and other users of the risk information. This requires those knowledgeable in risk assessment to be accessible and to have access to the decision makers. Framework Applications The Risk Characterization Framework can be applied in a number of ways. The systematic treatment of each of the ten framework elements aids the difficult job of comparing risk estimates made by different researchers using different methodologies. An extensive review of the indoor air technical literature revealed very few risk characterization studies which adequately reported all of the key data elements from source concentrations to either lifetime or population risks. Studies were typically strong in the exposure assessment components (Columns B, C, D, and E) but weak in the health oriented components (Columns G through K) or vice-versa. Fourteen studies were analyzed using the Risk Characterization Framework and are described elsewhere. 8 Not surprisingly, risk estimates made by various investigators under non-identical input assumptions will produce differing results. How risk estimates vary and what key components produce such differences are both of interest. Figure 6 illustrates how key numerical differences between studies by three investigators can be compared in the Risk Characterization Framework format. Benzene lifetime risk estimates are shown in Column I of this figure as computed in studies by Tancrede, Wallace, and McCann. Note that risk estimates by Tancrede and McCann differ by over a factor of ten even for situations where average concentration (Column C) and dose (Column G) are similar. In this case, variance in response factors (Column H) caused the risk estimate differences. Note also that both the mean concentration from the Wallace example (about 50 percent higher) and the response factor shown in Column H accounted for risk estimate differences. The review study mentioned previously 8 provides further details on risk characterization comparisons among studies focusing on indoor air pollutants such as radon, environmental tobacco smoke, volatile organic compounds, and asbestos. A second and very different type of Risk Characterization Framework application is illustrated in Figure 7. Two simple calculations using benzene examples are made. The "Benchmark Estimates" in Column A arbitrarily assume a lifetime average daily concentration of one-tenth the OSHA Permissable Exposure Limit. It is combined with the EPA unit risk (potency) factor for benzene. The "Working Adults" example in Column A assumes a concentration of 0.1 ppm as used above but for only 6.6 hours of "occupational" exposure per day. Outdoor and indoor residential concentrations assume the mean concentrations from Total Exposure Assessment Methodology (TEAM) study data. The effect of time activity patterns is thus illustrated. Key numerical risk data points and assumptions are explicitly shown. If a reviewer wants to see the effect of changing any parameter (e.g., using the 95th percentile concentrations from the Figure 7 Footnote), a quick "sensitivity study" can easily be done with a hand calculator. These examples are not intended as complete risk characterizations as defined by EPA and NAS guidelines since they do not include discussions of hazard identification or other components as discussed earlier. The estimated risks are higher than typically expected (see Figure 6) due to the high, but not unrealistic, concentration of 0.1 ppm assumed in the occupational setting. This approach can be useful as a training tool to illustrate how key assumptions and conversion factors impact the risk estimate. This paper is based on several of the five related EPA reports on indoor air quality risk characterization studies of carcinogenic compounds. This series of EPA/Environmen- Pollutant Exposure Concentration x Duration/Setting = Exposure (C) (D), (E) 1 Benzene (ng/m 3 ) (mean levels) days per lifetime fig/m 3 x days Dosimetry Response Lifetime ' K Factors = : Dose x Factor = Individual Risk x (F) (G) (H) I (1) See Fig 4 mg kg x day kg x day mg Lung cancer incidence Annual Risk Exposed to Exposed Population = Population (J) (K) # People CoLmns I x J 70yrs Tancrede et al Wallace 20 McCann et al lifetime 25,550 25, E E E E E E E E E E E E-4 "98th Percentile Estimate" 1.6 E-4 "Upperbound Estimates" 4.6 E-5 "Maximum Likelihood" Los Angeles 178 E+6 All metropolitan U.S. 400 Figure 6. comparisons). Example use of risk characterization framework for study comparisons (see Reference 8 for details of these and other October 1991 Volume 41, No

10 a, DC C o g.ox l IUJ EL in CM CO in CO ell r= co x~ -g 70kg ; 25, sdays ime II T-* -EiS IQ. IK a eg cc nn co FLU CD 1 il o o o u S T3 T3 XI 3 = J= O > o J. Air Waste Manage. Assoc.

11 tal Criteria and Assessment Office monographs includes: I Development of a Risk Characterization Framework 9 II Methods of Analysis for Environmental Carcinogens 10 III Overview of Indoor Concentrations of Environmental Carcinogens 19 IV Use of Benzene Measurement Data in Risk Characterization Estimates 20 V A Review of Indoor Air Quality Risk Characterization Studies. 8 Acknowledgments This work was sponsored by the U.S. EPA Environmental Criteria and Assessment Office (ECAO), Research Triangle Park, North Carolina. Views expressed are those of the authors and do not represent official policy. The authors acknowledge Dr. Michael A. Berry, Deputy Director of ECAO, who has been a research colleague throughout this effort. Additionally, suggestions by a number of people in various EPA Office of Research and Development locations were very helpful. References 1. U.S. Environmental Protection Agency Science Advisory Board, Future Risk: Research Strategies for the 1990's, EPA, SAB-EC , U.S. Environmental Protection Agency, Comparing Risks and Setting Environmental Priorities, Overview of Three Regional Projects, Office of Policy, Planning, and Evaluation, Washington, DC, Wallace, L. A.; Pellizzari, E. D.; Hartwell, T. D.; Sparacino, C. M.; Sheldon, L. S.; Zelon, H. "Personal exposures, indooroutdoor relationships and breath levels of toxic air pollutants measured for 355 persons in New Jersey," Atmos. Environ. 19:1651(1985). 4. Pellizzari, E. D.; Hartwell, T. D.; Perritt, R. L.; Sparacino, C. M.; Sheldon, L. S.; Zelon, H. S.; Whitmore, R. W.; Breen, J. J.; Wallace, L. "Comparison of indoor and outdoor residential levels of volatile organic chemicals in five U.S. geographical areas," Environ. Internat. 12:619 (1986). 5. Shah, J. J.; Singh, H. B. "Distribution of volatile organic chemicals in outdoor and indoor air," Environ. Sci. Technol. 22:1381(1988). 6. National Research Council, Risk Assessment in the Federal Government: Managing the Process, National Academy Press, Washington, DC, U.S. Environmental Protection Agency, Risk Assessment Guidelines of 1986, Office»of Health and Environmental Assessment, Washington, DC, EPA /045, Naugle, D. F.; Pierson, T. K.; Layne, M. E. "Review of Indoor Air Risk Characterization Studies," Final report for the U.S. Environmental Protection Agency, Environmental Criteria & Assessment Office, Research Triangle Park, North Carolina, EPA/600/8-90/044, (1989). 9. Naugle, D. F.; Pierson, T. K.; Mauskopf, J. A. "Development of a Risk Characterization Framework." Final report for the U.S. Environmental Protection Agency, Environmental Criteria & Assessment Office, Research Triangle Park, North Carolina, RTI No. 4479/01-F, (1990). 10. Peterson, M. R.; Naugle, D. F.; Berry, M. A. "Methods of Analysis for Environmental Carcinogens." Final report for the U.S. Environmental Protection Agency, Environmental Criteria & Assessment Office, Research Triangle Park, North Carolina, EPA/600/8-90/041, Russell, M. A. H.; Jarvis, M. J.; West, R. J. "Use of urinary nicotine concentrations to estimate exposure and mortality from passive smoking in non-smokers," British J. Addiction. 81: 275 (1986). 12. Behar, J. V.; Thomas, J.; Pandian, M. D. "Benzene Exposure Assessment Model (BEAM)," Second interim report for the, U.S. Environmental Protection Agency, Environmental Monitoring and Support Laboratory, Las Vegas, Nevada, EPA/600/ x-90/027, U.S. Environmental Protection Agency, Interim Methods for Development of Inhalation Reference Doses, Environmental Criteria and Assessment Offices, Office of Health and Environmental Assessment, Research Triangle Park, North Carolina, EPA/600/8-88/066, U.S. Environmental Protection Agency, Integrated Risk Information System Supportive Documentation Volume 1, Office of Health and Environmental Assessment, Washington, DC, EPA/600/8-86/0320, U.S. Environmental Protection Agency, "Linear Location Risk Analysis Model," Draft report, August MRI, "Comparison of Risks and Costs of Hazardous Waste Alternatives: Methods Development and Pilot Studies," Final report prepared for the U.S. Environmental Protection Agency, Wallsten, T. S.; Whitfield, R. G. "Assessing the Risks to Young Children of Three Effects Associated with Elevated Blood- Lead Levels," Argonne National Laboratory Report AWL/AA- 32, Argonne, II Keeney, R. L., et al. "Development and Application of a Risk Assessment Methodology to Study Alternative National Ambient Carbon Monoxide Standards," Prepared for the U.S. Environmental Protection Agency, Strategies and Air Standards Division, Research Triangle Park, North Carolina, May Gold, K. W.; Berry, M. A.; Naugle, D. F. "Overview of Indoor Concentrations of Environmental Carcinogens," Final report for the U.S. Environmental Protection Agency, Environmental Criteria & Assessment Office, Research Triangle Park, North Carolina, EPA/600/8-90/043, August Naugle, D. F.; Peterson, M. R. "Use of Benzene Measurement Data in Risk Characterization Estimates: A Preliminary Approach," Final report for the U.S. Environmental Protection Agency, Environmental Criteria & Assessment Office, Research Triangle Park, North Carolina, EPA/600/8-90/042, Dr. Naugle is director and Dr. Pierson is Risk Analysis manager of the Center for Environmental Analysis, Research Triangle Institute, Research Triangle Park, NC This manuscript was submitted for peer review on June 20, The revised manuscript was received on August 8,1991. October 1991 Volume 41, No