Policy Analysis. Introduction

Transcription

1 Policy Analysis Policies for Chemical Hazard and Risk Priority Setting: Can Persistence, Bioaccumulation, Toxicity, and Quantity Information Be Combined? JON A. ARNOT* AND DON MACKAY The Canadian Environmental Modelling Centre, Trent University, 1600 West Bank Drive, Peterborough, ON, Canada, K9H 7B8 Received January 16, Revised manuscript received March 30, Accepted March 31, Existing methods used to screen chemical inventories for hazardous substances that may pose risks to humans and the environment are evaluated with a holistic mass balance modeling approach. The model integrates persistence (P), bioaccumulation (B), toxicity (T), and quantity (Q) information for a specific substance to assess chemical exposure, hazard, and risk. P and B are combined in an exposure assessment factor (EAF), P, B, and T in a hazard assessment factor (HAF), and P, B, T, and Q in a risk assessment factor (RAF) providing single values for transparent comparisons of exposure, hazard, and risk for priority setting. This holistic approach is illustrated using 200 Canadian Domestic Substances List (DSL) chemicals and 12 United Nations listed Persistent Organic Pollutants (POPs). Priority setting results are evaluated with those of multiple category-based screening methods employed by Environment Canada and applied elsewhere that use cutoff criteria in multiple categories (P, B, and T) to identify hazardous chemicals for more comprehensive evaluations. Existing methods have categorized the DSL chemicals as either higher priority (requiring further assessment; screened in) or lower priority (requiring no further action at this time; screened out). The priority setting results of the cutoff-based categorization are largely inconsistent with the proposed integrated method, and reasons for these discrepancies are discussed. Many chemicals screened out using existing methods have equivalent or greater risk potential than chemicals screened in. Decisions for screening assessments using binary classification on the basis of cutoff criteria can be flawed, and complementary holistic methods for priority setting evaluations such as the one proposed should be considered. Introduction International and national chemical management programs seek to protect human health and the environment from the potential risks of hazardous chemicals as identified by the four separate criteria of persistence (P), bioaccumulation (B), and toxic (T) properties, and quantity (Q) information (1 4). Hazard is an intrinsic or intensive property of the * Corresponding author tel: extn 7645; fax: ; jonarnot@trentu.ca. chemical, whereas risk is an extensive property that requires information on the quantity of chemical released and the resulting exposure (5). Hazard is thus a function of P, B, and T, and risk is a function of P, B, T, and Q. Monitoring data are available for only a small percentage of the estimated chemicals that require evaluation, and there is a general lack of measured data on chemical properties (6). For example, of the approximately organic chemicals on Canada s Domestic Substances List (DSL), measured bioaccumulation factors (BAF) in fish exist for <0.5% and measurements for vapor pressure and acute toxicity in fish exist for <10% (7, 8). It is thus necessary to rely on estimation methods for key physical-chemical information (e.g., degradability, toxicity) and the application of models (7, 9). There is a need to focus the most immediate effort on chemicals of greatest concern; however, setting such priorities is challenging. While reducing risk is the fundamental objective, for practical reasons most current methods first screen for hazards (P, B, and T properties) followed by exposure and risk assessment. Numeric bright-line or cutoff criteria are used for chemical hazard categories such that a chemical receives a tentative binary assignment as satisfying or not satisfying a selected cutoff criterion (e.g., B or not B ) (see (6) for examples of criteria). Using three hazard categories a chemical can fall into one of eight classes (e.g., P, B, T, or P, not B, T, or not P, not B, not T ). A fundamental problem arises when comparing these classes since it involves subjective judgment of the relative importance of P, B, and T (10). The effectiveness of existing criteria to adequately identify all potentially hazardous chemicals is often questioned and uncertainty in preliminary assessments is neglected (5, 11). In this analysis we evaluate existing chemical screening methods with a holistic mass balance modeling approach. We first outline a holistic method by which quantitative information on P and B are combined to assess exposure in an exposure assessment factor (EAF); P, B, and T are combined to assess hazard in a hazard assessment factor (HAF), and P, B, T and Q are combined in a risk assessment factor (RAF) for a screening level metric of risk. A mass balance model simulates fate and transport processes (P), aquatic and terrestrial food web bioaccumulation (B), and toxicity (T) in a coherent evaluative framework. The holistic method is compared with the Canadian categorization and priority setting method using chemicals on the DSL and listed United Nations Stockholm Convention Persistent Organic Pollutants (POPs). For transparent comparisons between the methods, we apply the same basic chemical property information available for use in the DSL categorization in the holistic modeling approach. Materials and Methods Outline of the Proposed Holistic Approach. Risk Assessment Factor (RAF). The calculation of the RAF is first described using the Risk Assessment, IDentification, And Ranking (RAIDAR) model (12). RAIDAR is a screening level evaluative model that combines information on chemical partitioning, reactivity, environmental fate and transport, food web bioaccumulation, exposure, effect end point, and emission rate in a coherent mass balance framework. Detailed descriptions of the physical compartments, biological organisms and recent revisions to RAIDAR (Ver.2.0) are outlined in the Supporting Information (SI). The RAF calculation comprises four steps. First, an arbitrary unit emission rate E U (e.g., 1 mol h -1 ) is assumed ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 13, /es800106g CCC: $ American Chemical Society Published on Web 05/30/2008

2 for chemical releases into the defined evaluative environment. Either Level II or Level III steady state chemical fate calculations can be selected to determine the transport, distribution, and concentrations in the four bulk physical compartments of the environment (air, water, soil, and sediment). Level III calculations also require mode-of-entry information specifying the percentage of total unit emissions released to each bulk compartment. The standard RAIDAR environment is of regional scale ( km 2 ) similar to the EQuilibrium Criterion (EQC) model (12). Second, food web bioaccumulation models are used to calculate chemical concentrations in representative trophic guilds and species including primary producers, invertebrates, fish, wildlife, domestic livestock, and humans. Primary producers and invertebrates bioconcentrate chemical from their surrounding environment of air, water, soil, or sediment while all other species bioaccumulate chemical from their surrounding environment and from their diet. Third, calculated unit concentrations in the representative organisms C U (mol m -3 ) are compared to whole body concentrations that represent a threshold (toxic) effect end point C T (mol m -3 ). Different end points can be selected depending on assessment objectives and data availability, but the end point must be consistent for all species when comparing chemicals (12). For each species the ratio C U/C T is calculated and the largest ratio identifies those species subject to greatest hazard. The final stage is to introduce the actual emission rate E A (mol h -1 ), which is an estimated rate of release (Q), and since all model equations are linear the concentrations change by the factor E A/E U. The RAF (dimensionless) for each chemical is then calculated as RAF ) (C U /C T )(E A /E U ) (1) The RAF is the ratio of the expected concentration in the most vulnerable organism (or a selected organism) to the critical concentration in that organism corresponding to the actual emission rate. Higher priority chemicals have higher RAFs and thus a greater likelihood of risk. Insight into how P, B, T, and Q combine to express risk (RAF) can be obtained by reformulating eq 1 as RAF ) E A DT (2) where D is the environmental delivery ratio C U/E U (h m -3 ). D quantifies the ability of the environment to deliver the specific chemical to the most vulnerable organism in the defined environment and food web. The parameters E A and T are user-specified. The toxicity (or threshold) T is 1/C T. Thus, D quantifies maximum exposure potential, E A and D together characterize exposure levels to the most vulnerable species, T characterizes effect, and E A, D, and T combine to express an integrated metric of risk. When assessing a variety of chemicals, the vulnerable organism will likely change, thus C T and T will also change. E A and T should be expressed in consistent units (e.g., g, kg, or mol). Hazard Assessment Factor (HAF). The only extensive chemical-dependent property in the RAF calculation is E A (representing Q). The hazard assessment factor (HAF; dimensionless) is C U/C T, and is an intensive property being a combined function of P, B and T. The HAF provides a single value for comparing all chemicals for these combined properties. The difference between the HAF and the RAF is the ratio of the actual quantity of chemical released to the environment to the selected unit emission rate, i.e., E A/E U. Hazard assessment factors can be compared and ranked by the most vulnerable species, or it may be desirable to select an organism of interest for chemical hazard comparisons unique to that species (or trophic guild). Exposure Assessment Factor (EAF) or Delivery Ratio (D). The model calculates an exposure assessment factor (EAF; h m -3 ), or C U/E U, for each species. The highest EAFs identify the species that are most vulnerable to contamination (D ratio) but not necessarily vulnerable to toxic effects. The combined fate and bioaccumulation models calculate the EAF for each chemical and for each representative species in the model. Thus, the EAF is an intensive hazard property of the chemical being a combined function of P and B and provides for transparent assessments of chemical exposure potential to all species. Further insight into the factors controlling D and the EAF and their dependence on P and B are provided in the SI. It may be desirable to select an organism of interest in addition to the most vulnerable organisms when setting priorities. Exposure assessment factors or D ratios can be compared with those of known POPs to inform the potential for exposures to specific species or trophic guilds of interest. In summary, three types of chemical assessments factors in this study are proposed and calculated as EAF ) C U /E U f(p, B) (3) HAF ) E U (EAF)T ) C U /C T f(p, B, T) (4) RAF ) HAF(E A /E U ) ) EAF(E A )T ) (C U /C T )(E A /E U ) f(q, P, B, T) (5) Evaluating Existing Screening Level Methods and Criteria. The Canadian Environmental Protection Act (1999) declared that all DSL chemicals must be categorized for hazard by September 2006 using P, B, and T cutoff criteria (2, 9). Hazardous substances may then be evaluated for potential risk (2). Of the 200 DSL chemicals used in this case study, 100 were categorized in, i.e., DSL(I), and are considered to require further assessment, and 100 were categorized out, i.e., DSL(O), and are considered to not require further action at this time. This categorization or priority setting method is similar to that employed or planned in other jurisdictions. Details of the chemicals and their physical-chemical properties are listed in the SI. The DSL chemicals used in this study are not expected to appreciably ionize at environmental ph. Level II and Level III RAIDAR calculations were performed for 12 Stockholm Convention POPs and 200 Canadian DSL chemicals as a case study to illustrate insights obtained by applying the EAF, HAF, RAF approach. For simplicity in this example, equal chemical emission rates were assumed to air, water, and soil for Level III calculations, i.e., 1/3-unit emissions to each. Negligible advective losses in air and water and negligible biotransformation rates in the food webs were assumed for these calculations. Threshold (toxicity) effect concentrations and actual emission rates for HAF and RAF calculations were based on Environment Canada s selected pivotal inherent Toxicity (it) values and quantity ranges used for the DSL categorization (7). The it values for the 200 DSL chemicals were either measured or estimated aquatic exposure lethal effect concentrations. Thus, acute lethality was chosen as the threshold (toxicity) effect end point for HAF and RAF calculations. For each chemical, an internal effect concentration (IEC, mmol. kg -1 ) was calculated as a product of the pivotal it value (mmol L -1 ) and the bioconcentration factor (BCF, L kg -1 ). The toxic ratio (TR) approach (13, 14) was used to differentiate selective chemicals from baseline (narcotic) chemicals (5 mmol kg -1 assumed) (5). For POPs, the IEC values were taken from Mackay (5) when data were available (these IEC values were also calculated from acute it and BCF data), and when data were not available those chemicals were assumed to exert a narcotic mode of acute toxic action, i.e., IEC ) 5 mmol kg -1. Estimates for the BCF and emission rates are described in the SI. VOL. 42, NO. 13, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

3 FIGURE 1. Environmental delivery ratios (D), or maximum exposure assessment factor (EAF) values (h m -3 ), for Persistent Organic Pollutants (b), Domestic Substances List chemicals screened in using current methods (DSL(I), +), and DSL chemicals screened out using current methods (DSL(I), -) as calculated using Level II (top) and Level III (bottom) RAIDAR fate calculations. Chemicals are not consistently numbered between figures and are ordered from highest to lowest. Results and Discussion RAIDAR Assessment Factors for DSL Chemicals and POPs. Figure 1 compares Level II and Level III environmental delivery ratios (D ratios) or EAFs (h m -3 ) for the benchmark POPs and DSL chemicals. Environmental delivery ratios for POPs range from to and to for Level II and Level III calculations, respectively. Delivery ratios for DSL(I) chemicals range from to and to for Level II and Level III calculations, respectively. Delivery ratios for DSL(O) chemicals range from to and to for Level II and Level III calculations, respectively. Figure 1 highlights that approximately 10% (Level III) to 20% (Level II) of DSL(O) chemicals have exposure potentials equivalent to current POPs, and that approximately 15% (Level II) to 20% (Level III) of DSL(I) chemicals have exposure potentials that are orders of magnitude lower than current POPs. Figure 2 compares Level II and Level III HAFs for the benchmark POPs and DSL chemicals. Hazard assessment factors for POPs range from to and to for Level II and Level III calculations, respectively. Hazard assessment factors for DSL(I) chemicals range from to and to for Level II and Level III calculations, respectively. Hazard assessment factors for DSL(O) chemicals range from to and to for Level II and Level III calculations, respectively. This suggests approximately 20% of the DSL(O) chemicals have HAF values that are equivalent to HAF values for current POPs based on either Level II or Level III calculations. There is about a 70% (Level FIGURE 2. Hazard assessment factor (HAF; black) and risk assessment factor (RAF; red) values for Persistent Organic Pollutants (b), Domestic Substances List chemicals screened in using current methods (DSL(I), +), and DSL chemicals screened out using current methods (DSL(I), -) as calculated using Level II (top) and Level III (bottom) RAIDAR fate calculations. Chemicals are not consistently numbered between figures and are ordered from highest to lowest. III) to 80% (Level II) overlap between HAF values for DSL(O) and DSL(I) chemicals. Figure 2 also compares Level II and Level III HAFs and RAFs for the DSL chemicals. Risk assessment factors for DSL chemicals can increase or decrease compared to HAFs for the same chemical as a function of E A. Risk assessment factors for DSL(I) chemicals range from to and to for Level II and Level III calculations, respectively. Risk assessment factors for DSL(O) chemicals range from to and to for Level II and Level III calculations, respectively. Approximately 80% (Level III) to 90% (Level II) of DSL(O) chemicals have RAF values that are equivalent to RAF values for DSL(I) chemicals. Some RAFs for DSL(O) chemicals are about 8 orders of magnitude higher than RAFs for DSL(I) chemicals. Level II and Level III results are fairly similar with regards to the relative RAF rankings of chemicals (Figure S-4), suggesting that Level II calculations may be useful if information on mode-of-entry is not available. In summary, the exposure potential of some DSL(O) chemicals can be high compared to current POPs and some DSL(I) chemicals can have comparatively low exposure potential. There is also substantial overlap in HAFs between DSL(O) and DSL(I) chemicals. These results suggest that applying P, B, and T criteria individually can lead to less discriminating priority setting and that the combined exposure and hazard indicators may be preferable. As expected, including Q for the calculation of risk can result ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 13, 2008

4 FIGURE 3. An overview of current and holistic chemical screening and priority setting methods. TABLE 1. A Case Study Comparison of Hazard and Risk Information for Four Canadian Domestic Substances List (DSL) Chemicals Based on Level III Fate Calculations a chemical DSL categorization decision P OV (d) log K OW it (mg L -1 ) IEC (mmol kg -1 ) E A (kt y -1 ) HAF RAF hexachloroethane P, it, not B ,3,5-triBB P, B, it propenamide not P nor B, it ,4-benzenediol not P nor B, it a P OV, overall persistence; K OW, octanol-water partition coefficient; it, pivotal exposure toxicity value selected in the DSL categorization (7); IEC, internal effect concentration calculated from it as described in the text; E A, estimated actual emission rate for risk calculations derived from production volume quantity information (Q) reported for DSL chemicals (7) as described in the text; HAF, hazard assessment factor; RAF, risk assessment factor; 1,3,5-triBB, 1,3,5-tribromobenzene. in a large change in priorities compared with combined hazard rankings. RAF rankings are significantly different from HAF rankings. Based on current data many DSL chemicals that are not currently considered for further evaluations are shown to have equivalent or greater risk potential than DSL chemicals currently identified for further evaluations. Outlining Potential Pitfalls of Existing Methods and P, B, T Cutoff Criteria. Figure 3 summarizes key differences between existing and holistic methods for screening chemical substances for exposure, hazard, and risk. The same basic information used in the DSL categorization was used in this study, yet there are substantial differences in the identification of chemicals requiring more comprehensive evaluation. These differences do not preclude potential errors in the proposed alternative; rather they highlight discrepancies with current practices and the need to evaluate the effectiveness of all methods. Methods used to prioritize chemical risk assessments and to identify POPs may result in type I (false positives) and type II (false negatives) errors at preliminary stages of chemical evaluation. Multiple category cutoff criteria methods may be particularly susceptible to potentially high levels of type I and type II errors resulting in the misapplication of limited resources for the evaluation of substances that are of low potential risk while neglecting those substances that are of high potential risk. The main objective of this policy analysis is to evaluate the application of current knowledge used for chemical screening assessments to inform decisions and not to discover truth (15). Table 1 summarizes hazard and risk information and discrepancies between the existing methods and the holistic methods for four selected chemicals. In this analysis a Type I error is considered to occur when a DSL chemical is screened in, i.e., DSL(I), but the holistic approach indicates it is of low priority for risk assessment. A Type II error is considered to occur when a DSL chemical is screened out, i.e., DSL(O), but the holistic approach indicates it is of high priority for risk assessment. Two chemicals (hexachloroethane and 1,3,5-tribromobenzene) were categorized in, i.e., DSL(I), but have low HAF and RAF values. An explanation for the discrepancy is the misclassification of toxic hazard. These two chemicals were considered to be it in the DSL categorization, i.e., LC(EC) 50 e1 mg L -1 ; however, the chemicals are baseline narcotics ( mmol kg -1 )(13, 14, 16). Thus they are of comparatively low toxic hazard. Further, on the basis of the DSL categorization data, these chemicals have low estimated emissions and low RAFs and are of low risk in comparison to other RAF estimates for the 200 chemicals. Two chemicals (2-propenamide and 1,4-benzenediol) were categorized out, i.e., DSL(O), but have high HAF and RAF values. An explanation for this discrepancy is the apparent potency (low body burdens for toxic effects) and high emission rates for these two chemicals. These two chemicals were also identified as it in the DSL categorization; however, the IEC values suggest a selective mode of action, i.e., more potent than baseline toxicity (13, 14, 16). Thus, 2-propenamide and 1,4-benzenediol are toxic hazards, and despite the low hydrophobicity and low persistence (not B or P hazards) in the environment, these two chemicals have comparatively higher HAF values than hexachloroethane and 1,3,5-tribromobenzene. Further, on the basis of the DSL categorization information, these chemicals have higher estimated emissions than hexachloroethane and 1,3,5- tribromobenzene and have comparatively higher RAFs than many of the other 200 case study chemicals. Toxicity. It is well established that there are inherent problems associated with using exposure concentrations in water (ecotoxicology) or air or food (mammalian toxicology) as expressions of relative toxicities of organic chemicals and that IECs or body burdens can be more reliable for expressing relative toxicities (14, 16). As long ago as 1939, Ferguson illustrated how the disturbing effect of phase distribution obscures the interpretation of toxicity data (17). There is a high degree of overlap for the DSL(I) and DSL(O) pivotal it VOL. 42, NO. 13, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

5 values, and approximately 25% of the DSL(O) chemicals are expected to be more potent than narcotics, i.e., possess a selective mode of action (Figure S-5) (13, 14). Further, bioassays using exposure data cannot explicitly account for possible increases in internal chemical concentrations that can occur for chemicals that biomagnify. The proposed method seeks to address this issue by estimating internal concentrations associated with toxic effects. The end points and toxicity data selected in this study are the same as those used in the DSL categorization; however, internal concentrations will not always be appropriate (e.g., carcinogenic end points require special treatment). The limitations of existing toxicity information and the expectation that approximately 65% or more of high production volume and commercial chemicals are acutely narcotic (18, 19) justifies the proposed methods. Different thresholds can be included as more reliable toxicity data become available. Persistence. Many existing methods for assessing persistence define media specific cutoff criteria half-lives; however, there are inherent difficulties in setting media specific cutoff criteria for assessing persistence (20). The overall persistence (P OV) in an evaluative environment has been proposed as an alternative to media-specific half-lives and is calculated by RAIDAR and other models (21). Table 1 indicates that non P categorized chemicals can still pose high combined hazards and risks and that P OV alone is not an adequate indicator for risk. The holistic method avoids the use of media specific halflife cutoff criteria and combines persistence with other information providing single transparent indicators of exposure, hazard, and risk. Level II fate calculations were selected for consistent analysis with the DSL categorization for persistence decisions (9). Level III is often used in screening level assessments, and an example of these calculations is provided for comparisons (21). RAIDAR RAF rankings from four different Level III mode-of-entry scenarios were previously compared with Level II calculations and illustrated that relative RAF rankings were generally robust to mode-of-entry scenarios (12). As discussed previously (12), and elsewhere (22, 23), mode-of-entry can be important for certain chemicals and should be considered in comprehensive evaluations; however, these differences do not change the general findings in this policy analysis. Level IV time dependent calculations may ultimately be more appropriate for chemicals that are very persistent and for which the time to steady state in the environment may be long. Steady state calculations (Level II and Level III) may be conservative for very persistent chemicals if they are not used in commerce for long periods of time; however, these chemicals have historically been of greatest concern, and thus a conservative approach for screening assessments seems justifiable. Bioaccumulation. Chemicals not captured by current B criteria (e.g., log K OW < 5) can be of equivalent or greater hazard and risk as those that are captured by current B criteria. Evidence suggests that air-breathing organisms (birds, mammals) have the capacity to biomagnify chemicals with log K OW < 5 if the octanol-air partition coefficient (K OA) is high and the chemical is not biotransformed at a rate sufficient to lower the body burden (11, 24). Table 1 illustrates that chemicals that are not expected to biomagnify, or even appreciably bioaccumulate, may pose high risks if emissions and toxicity are high (see also Figure S-6). The proposed method addresses these concerns by including mechanistic food web bioaccumulation models for both aquatic and terrestrial species in the calculation of EAF, HAF, and RAF. The food web models capture biomagnification potential, and biotransformation rates can be included in the mass balance equations as data become available. The EAF, HAF, and RAF estimates will be reduced for chemicals that are biotransformed in food webs. It is important to obtain and include reliable estimates for biotransformation half-lives for assessments of exposure, hazard, and risk. The availability of biotransformation data is limited, but research in this area is advancing (25). Notably, DSL(I) and DSL(O) categorization decisions based on existing methods and criteria generally assumed negligible rates of biotransformation, and thus the holistic methods are consistent with existing methods for comparisons in this analysis. Exposure. The use of P and B criteria in screening assessments is presumably to identify chemicals with the greatest potential for high exposure levels; however, current methods and criteria may not always achieve this objective. The proposed method includes exposure estimates for humans and nonhuman organisms by combining elements of P and B into a single value (EAF). Other combined exposure indicators exist (22, 26). For example, McKone and colleagues have shown that the human intake fraction is a valuable indicator of human exposure (27). The EAF calculation differs from the human intake fraction because the EAF includes absorption, biotransformation potential, and excretion processes in the organism of interest allowing for the calculation of a total body burden or concentration value. The EAF estimates are calculated in this study at steady state whereas the human intake fraction can be calculated at steady state or under dynamic conditions. Notably, although far-field exposures to humans are included in the current RAIDAR model framework, for the chemicals included in this case study, humans were not identified as the most vulnerable species and thus did not determine the illustrated HAF and RAF estimates (Table S-9). Human intake fractions are useful indicators for human exposure potentials and by definition they do not include ecological receptors. Uncertainty and Cutoff Criteria. Existing methods and cutoff criteria do not address the inherent uncertainty associated with the values selected for the chemical in the assessment and the cutoff value. For example, if a BCF or BAF in fish g 5000 is considered to be bioaccumulative (1, 9), then a chemical with a measured or estimated BCF or BAF of 5100 is considered B, and a chemical with a measured or estimated BCF or BAF of 4900 is considered not B. The difference between these values is essentially inconsequential in the final calculation of a chemical s risk potential. Uncertainty (and variability) in bioaccumulation measurements is typically about a factor of 3 after rigorous data quality reviews (8), and models cannot be expected to provide more accurate values since models also require physical-chemical property values (e.g., K OW). Similar arguments about actual values being near cutoff criteria could be made for P and T criteria. These problems become acute if the P, B, and T properties are close to the cutoff criteria values; however, as discussed earlier, actual values do not always have to be near cutoff criteria (Table 1). Combining binary scores from multiple categories further requires subjective judgment. A holistic approach permits the use of multiple cutoff criteria, but in addition it provides single values for transparent comparisons of exposure, hazard, and risk potential. Inspection of eq 2 shows the RAF is particularly sensitive to uncertainties in E A and T and reliable information for these parameters is required for reliable RAF calculations. RAIDAR RAFs can vary over 13 orders of magnitude (12); therefore, high accuracy is not always required, especially for lower priority chemicals. For example, even seemingly large errors or uncertainties in actual emission rates (e.g., 2-3 orders of magnitude) have minimal impacts in priority rankings for large numbers of chemicals using the holistic approach. Conversely, quantity estimates may result in entirely different requirements for chemical evaluations using category-based methods for Q, i.e., REACH (3). Emission estimates E A included in this study illustrate the importance of Q in the ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 13, 2008

6 ultimate assessment of risk and the large difference between hazard and risk rankings. Uncertainty analyses using mass balance models have been illustrated (15, 23, 28) and can readily be included in the proposed holistic methods to address uncertainty, identify key data gaps, and prioritize the required information to reduce the uncertainty in assessing hazards and risks. The general findings of the present analysis will not be affected by conducting specific uncertainty analyses for each chemical. Considerations for the Proposed Methods. The EAF, HAF, and RAF calculations are functions of the evaluative environment. Thus changes to these parameters may result in differences in ranking EAF, HAF, and RAF. Different regions have different physical and biological characteristics (e.g., arctic, temperate, or tropical); therefore, modifications must be made to the RAIDAR environment to adjust for such factors if assessments for other representative conditions are desired. For chemical comparisons, a consistent evaluative environment is required. The RAF is a screening level metric for risk that can be used to screen and prioritize chemicals for comprehensive risk assessments; it is not a substitute for risk assessment (see SI for further clarification). The current assumptions and limitations using the RAIDAR model and some of the research required for improvements to exposure and risk models in general have been outlined (12). There are other models that calculate different metrics for hazard and risk potential (e.g., intake fraction, toxicity potential, and impact assessment) that may also provide similar evaluations of current regulatory methods and criteria (22, 23, 26, 29). Advancing Screening Level Hazard and Risk Assessments. This policy analysis shows that existing methods used in regulatory programs and international treaties reflect the state of the science some decades ago and may not adequately identify chemicals that pose the greatest risks to humans and the environment. The categorization of the substances on the Canadian DSL was a significant task and a pioneering effort in the assessment of commercial chemicals; however, methods and criteria should evolve as the science progresses. Ultimately monitoring data and robust toxicity information are needed for decisions regarding POP and PBT designations, but these data are only available for screening few or small fractions of the chemicals requiring assessment. The holistic method provides a scientifically defensible strategy based on the current science for screening and transparently prioritizing many existing, new, and premanufacture chemicals for both hazard and risk. This information can be used directly or complement information from existing methods. The holistic approach provides a framework for prioritizing data needs for risk assessments and can guide monitoring programs by identifying environmental compartments that are likely to have the highest concentrations. Acknowledgments The authors thank the Natural Sciences and Engineering Research Council of Canada, Environment Canada, and the consortium of companies that support research at the Canadian Environmental Modelling Centre. Supporting Information Available A description of RAIDAR Ver.2.0 model input parameters for the 212 case study chemicals and summary output for the Level II and Level III calculations. This information is available free of charge via the Internet at Literature Cited (1) UNEP. Final Act of the Conference of Plenipotentiaries on the Stockholm Convention on Persistent Organic Pollutants, United Nations Environment Program: Geneva, Switzerland, 2001; p44. (2) Government of Canada. Canadian Environmental Protection Act, 1999, Government of Canada: Ottawa, ON, (3) European Commission. Technical Guidance Document on Risk Assessment, Joint Research Centre, Institute for Health and Consumer Protection, European Chemicals Bureau: Ispra, Italy, (4) U.S. EPA. Proposed category for persistent, bioaccumulative, and toxic chemicals. Federal Register, 1998, Vol. 63, pp (5) Mackay, D.; McCarthy, L. S.; MacLeod, M. On the validity of classifying chemicals for persistence, bioaccumulation, toxicity and potential for long-range transport. Environ. Toxicol. Chem. 2001, 20, (6) Muir, D. C. G.; Howard, P. H. Are there other persistent organic pollutants? A challenge for environmental chemists. Environ. Sci. Technol. 2006, 40, (7) Environment Canada. Existing Substances Program at Environment Canada (CD-ROM). Ecological Categorization of Substances on the Domestic Substances List (DSL), Existing Substances Branch, Environment Canada: Ottawa, ON, (8) Arnot, J. A.; Gobas, F.A.P.C. A review of bioconcentration factor (BCF) and bioaccumulation factor (BAF) assessments for organic chemicals in fish. Environ. Rev. 2006, 14 (4), (9) Environment Canada. Guidance Manual for the Categorization of Organic and Inorganic Substances on Canada s Domestic Substances List, Existing Substances Branch, Environment Canada: Ottawa, ON, (10) Swanson, M. B.; Socha, A. C., Eds. Chemical Ranking and Scoring: Guidelines for Relative Assessment of Chemicals. Proceedings of the Pellston Workshop on Chemical Ranking and Scoring, Sandestin, FL, February 12-16,1995; Society of Environmental Toxicology and Chemistry: Pensacola, FL, (11) Kelly, B. C.; Ikonomou, M.; Blair, J. D.; Morin, A. E.; Gobas, F.A.P.C. Food web-specific biomagnification of persistent organic pollutants. Science 2007, 317 (5835), (12) Arnot, J. A.; Mackay, D.; Webster, E.; Southwood, J. M. Screening level risk assessment model for chemical fate and effects in the environment. Environ. Sci. Technol. 2006, 40 (7), (13) Verhaar, H. J. M.; Van Leeuwen, C. J.; Hermens, J. L. M. Classifying environmental pollutants. 1. Structure-activity relationships for prediction of aquatic toxicity. Chemosphere 1992, 25, (14) Maeder, V.; Escher, B. I.; Scheringer, M.; Hungerbuhler, K. Toxic ratio as an indicator of the intrinsic toxicity in the assessment of persistent, bioaccumulative, and toxic chemicals. Environ. Sci. Technol. 2004, 38 (13), (15) Hertwich, E. G.; McKone, T. E.; Pease, W. S. A systematic uncertainty analysis of an evaluative fate and exposure model. Risk Anal. 2000, 20 (4), (16) McCarty, L. S.; Mackay, D. Enhancing ecotoxicological modeling and assessment. Environ. Sci. Technol. 1993, 27 (9), (17) Ferguson, J. The use of chemical potentials as indices of toxicity. Proc. R. Soc. London, Ser. B: 1939, 127, (18) Veith, G. D.; Call, D. J.; Brooke, L. T. Structure toxicity relationships for the fathead minnow. Pimephales promelas: Narcotic industrial chemicals. Can. J. Fish. Aquat. Sci. 1983, 40, (19) Verhaar, H. J.; van Leeuwen, C. J.; Bol, J.; Hermens, J. L. M. Application of QSARs in risk management of existing chemicals. SAR QSAR Environ. Res. 1994, 2, (20) Webster, E.; Mackay, D.; Wania, F. Evaluating environmental persistence. Environ. Toxicol. Chem. 1998, 17, (21) Klasmeier, J.; Matthies, M.; MacLeod, M.; Fenner, K.; Scheringer, M.; Stroebe, M.; Le Gall, A. C.; McKone, T. E.; van de Meent, D.; Wania, F. Application of multimedia models for screening assessment of long-range transport potential and overall persistence. Environ. Sci. Technol. 2006, 40, (22) Bennett, D. H.; Margni, M. D.; McKone, T. E.; Jolliet, O. Intake fraction for multimedia pollutants: A tool for life cycle analysis and comparative risk assessment. Risk Anal. 2002, 22 (5), (23) Huijbregts, M. A. J.; Thissen, U.; Jager, T.; van de Meent, D.; Ragas, A. M. J. Priority assessment of toxic substances in life cycle assessment. Part II: Assessing parameter uncertainty and human variability in the calculation of toxicity potentials. Chemosphere 2000, 41 (4), VOL. 42, NO. 13, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

7 (24) Gobas, F.A.P.C.; Kelly, B. C.; Arnot, J. A. Quantitative structureactivity relationships for predicting the bioaccumulation of POPs in terrestrial food webs. QSAR Comb. Sci. 2003, 22 (3), (25) Arnot, J. A.; Mackay, D.; Bonnell, M. Estimating metabolic biotransformation rates in fish from laboratory data. Environ. Toxicol. Chem. 2008, 27 (2), (26) Pennington, D. W.; Margni, M.; Payet, J.; Jolliet, O. Risk and regulatory hazard-based toxicological effect indicators in lifecycle assessment (LCA). Hum. Ecol. Risk Assess. 2006, 12, (27) Bennett, D. H.; McKone, T. E.; Evans, J. S.; Nazaroff, W. W.; Margni, M. D.; Jolliet, O.; Smith, K. R. Defining intake fraction. Environ. Sci. Technol. 2002, 36 (9), 206a 211a. (28) MacLeod, M.; Fraser, A.; Mackay, D. Evaluating and expressing the propagation of uncertainty in chemical fate and bioaccumulation models. Environ. Toxicol. Chem. 2002, 21, (29) Hertwich, E. G.; Mateles, S. F.; Pease, W. S.; McKone, T. E. Human toxicity potentials for life-cycle assessment and toxics release inventory risk screening. Environ. Toxicol. Chem. 2001, 20 (4), ES800106G ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 13, 2008