Screening Level Risk Assessment Model for Chemical Fate and

Environ. Sci. Technol. 2006, 40, 2316-2323

Screening Level Risk Assessment Model for Chemical Fate and Effects in the Environment JON A. ARNOT,* DON MACKAY, AND EVA WEBSTER Canadian Environmental Modelling Centre, 1600 West Bank Drive, Trent University, Peterborough, Ontario, K9J 7B8 JEANETTE M. SOUTHWOOD Golder Associates, 32 Steacie Drive, Kanata, Ontario, K2K 2A9

A screening level risk assessment model is developed and described to assess and prioritize chemicals by estimating environmental fate and transport, bioaccumulation, and exposure to humans and wildlife for a unit emission rate. The most sensitive risk endpoint is identified and a critical emission rate is then calculated as a result of that endpoint being reached. Finally, this estimated critical emission rate is compared with the estimated actual emission rate as a risk assessment factor. This “back-tracking” process avoids the use of highly uncertain emission rate data as model input. The application of the model is demonstrated in detail for three diverse chemicals and in less detail for a group of 70 chemicals drawn from the Canadian Domestic Substances List. The simple Level II and the more complex Level III fate calculations are used to “bin” substances into categories of similar probable risk. The essential role of the model is to synthesize information on chemical and environmental properties within a consistent mass balance framework to yield an overall estimate of screening level risk with respect to the defined endpoint. The approach may be useful to identify and prioritize those chemicals of commerce that are of greatest potential concern and require more comprehensive modeling and monitoring evaluations in actual regional environments and food webs.

Introduction It is generally accepted that existing and new chemicals of commerce should be assessed for the risk they may pose to the environment and human health. These assessments are intended to prevent repetition of situations in which the synthesis and use of chemicals resulted in unacceptable effects. The assessment is a challenging task because of the large number of chemicals in commerce, the lack of information on the substances’ partitioning and reactivity properties, the quantities that may be emitted to the environment, and the use patterns. Often, the only information available is molecular structure and an indication of likely uses. A two-tier assessment system is often utilized in which the chemical’s intrinsic hazard is first estimated followed by a risk assessment. Hazard criteria assess intensive * Corresponding author e-mail: [email protected]; phone: 705748-1011, ext. 1645; fax: 705-748-1080. 2316

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 7, 2006

properties of the substance and are thus independent of quantity used; examples being persistence, bioaccumulation factors, inherent toxicity, and potential for long range transport. Risk of adverse effects may depend on these criteria but also on the quantity emitted. Whereas hazard ranking is useful to identify those substances that may pose a high risk, it may erroneously include substances which pose little or no risk because of the low quantities emitted, and it may exclude high volume chemicals which are relatively nonhazardous. The ultimate regulatory goal is to determine and control risk rather than hazard. The lack of monitoring data for most chemicals of commerce necessitates the use of models which purport to predict environmental fate and transport, bioaccumulation, and exposure. Examples are Simplebox/EUSES (1, 2), ChemCAN (3), CalTox (4), and the PBT profiler (5). These models require a variety of input parameters, ideally empirical in origin. If only molecular structure is known Quantitative Structure-Activity Relationships (QSARs) must be used, such as Estimation Programs Interface (EPI) Suite (6) for partitioning and reactivity, and various models for toxicity (e.g., ref 7). Risk assessment also requires estimates of emission rates (e.g., t/year) for defined regions. Such information may only be available in semiquantitative form as production ranges (e.g., 10-100 t/year) with additional uncertainty associated with emission factors, i.e., the fraction of production that is emitted. For pesticides this can be 100%, whereas for fuels it can be 0.1% or less. In this paper we describe a screening level Risk Assessment, IDentification, And Ranking (RAIDAR) model that can be applied to substances for which little or no empirical property data are available and emission rates are known only approximately. Although the uncertainties in output may be high, the results may be adequate to “bin” substances into groups of similar risk and thus compare lower and higher risk potential.

Conceptual Approach Most models proceed in a “forward” direction using data on emissions and chemical and environmental properties to calculate concentrations in abiotic and biotic media. These concentrations are compared with endpoints judged to be acceptable or unacceptable. For example, comparison of an estimated or measured concentration to a selected endpoint concentration (e.g., PEC/PNEC) provides a risk quotient (RQ). The RAIDAR model proceeds in a “reverse” direction from the most sensitive endpoint, i.e., highest RQ, to calculate the corresponding critical emission rate (EC). EC is then compared, possibly semiquantitatively, with an estimate of the actual emission rate (EA). The ratio of EA to EC is then a metric of risk and EA/EC can be viewed as a screening level risk assessment factor (RAF). The principle underlying the “backtracking” approach has been outlined by Mackay et al. (8) as part of a general discussion on hazard and risk assessment using QSARs. Figure 1 provides a conceptual overview of the RAIDAR model. The linearity of the model equations is exploited by first running the model for an arbitrary unit emission rate (EU). Risk quotients of predicted unit emission concentrations (CU) to selected endpoint concentrations (CE) are calculated in representative media, i.e., RQ ) CU/CE. There may be several RQs but the largest, corresponding to the most sensitive endpoint, is then used to scale EC from EU, i.e., EC ) EU/RQ. The critical emission rate is compared to an estimate of actual emissions to derive the risk assessment 10.1021/es0514085 CCC: $33.50

 2006 American Chemical Society Published on Web 03/01/2006

FIGURE 1. Conceptual overview of the RAIDAR model. factor, i.e., RAF ) EA/EC, namely:

RAF )

CU EA × CE E U

(1)

Risk Endpoints. Risk endpoints fall into two general categories which we term “objective” and “effect” based. For existing contaminants there may be standards, guidelines, or target concentrations that are suggested as posing an acceptably low level of risk. Such objective concentrations have been defined for air, water, soils, and foodstuffs for a limited number of commercial chemicals on the basis that resulting exposure (usually to humans) is tolerable. The second category comprises endpoints at which a defined effect is expected. Examples are acute or chronic lethal concentrations, doses, and critical body residues. All organic chemicals are believed to exert at least a baseline or narcotic toxicity at a critical body residue (CBR) of approximately 5 mmol/kg wet weight with more selective, potent, or intrinsically toxic substances exerting effects at lower concentrations (e.g., refs 9 and 10). It has been estimated that 32% of 973 high-production-volume chemicals and 60% of all commercial chemicals are narcotic (11, 12). The toxic ratio (TR) of baseline CBR to actual or selective CBR reflects this potency with a value exceeding 10 indicating a mode of action significantly more potent than baseline (13, 14). Potent substances such as dioxins and pesticides may have very large TRs, at least for certain species. For carcinogens the usual practice is to define a reference dose or a slope of the incidence/dose curve rather than an absolute effect concentration. If a probability of one in one million is used, an objective concentration corresponding to this incidence can be estimated. To establish the critical emission rate either an “objective” or “effect” endpoint can be applied, but for coherent evaluations it is imperative that the selected endpoint be consistent for all chemicals. Model Selection. Rather than use a real region, it is more convenient to define an evaluative or typical region in which all chemicals can be consistently assessed. The preferred scale we suggest is approximately 105 km2, which has been used in the EQC model (15). For the purpose of estimating per capita emission rates or population intake fractions, a

human population of 30 million is reasonable for an industrialized region of this size, i.e., 300 persons per km2. This is approximately the population density of Belgium or Japan; it is lower than The Netherlands or the state of New Jersey which both exceed 400 persons per km2. This is believed to represent a conservative value of human activity and industrial development (16). Population does not enter the model calculations unless a per capita rate of emission is used, or a population intake fraction is calculated (17). Figure 2 illustrates abiotic and biotic components of the RAIDAR model environment. Key abiotic compartments in RAIDAR are the same as the EQC standard environment including a 10% area of water as lakes and rivers. Unlike EQC, the RAIDAR environment incorporates biota. The mass balance of chemical in the abiotic environment is established on a steady-state Level II or Level III basis from a defined unit emission rate including losses only by degradation and advection. Bioaccumulation is subsequently calculated and includes representative food webs, selected to include key trophic guilds and indicator species as well as agricultural species responsible for the primary human diet. Mass balance bioaccumulation calculations include exposures from the diet as well as from the surrounding environment and are in agreement with empirical observations of bioaccumulation potential in aquatic, terrestrial, and agricultural food webs as described in the Supporting Information (SI). Reliable estimates of metabolic transformation can be included. Details of the compartments, compositions, and parameters, as well as Level II and III equations describing fate and bioaccumulation, are in the SI. The primary application of the RAIDAR model is for risk assessment, however, the model also provides “secondary” output that may be useful for hazard assessment or for comparison with other hazard evaluations (Figure 1). The overall, reaction, and advection chemical residence times are calculated; with reaction residence time being the overall persistence (POV). The model provides bioaccumulation information, i.e., BAFs, accounting for dietary exposure in a range of receptors such as birds and mammals, including humans. These values are independent of emission rate but are dependent on the model and mode of entry. A characteristic travel distance for long range transport in air is VOL. 40, NO. 7, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2317

FIGURE 2. Representative ecological and human food webs in RAIDAR. calculated from an assumed air velocity and fraction of chemical mass in air as described in the SI (18). The environmental media and organisms in which the highest concentrations are likely are identified, thereby providing data that may be useful for monitoring purposes. Mode of entry (MOE) identifies the abiotic media to which the chemical is released (e.g., air, water, or soil). For Level III calculations MOE is an important determinant of fate and separate assessments are needed for each MOE. MOE is irrelevant for Level II fate calculations because the chemical instantaneously achieves equilibrium among all abiotic media (15). Here we examine and compare both Level II and III fate calculations in the RAIDAR framework.

Illustration Three chemicals with diverse physical-chemical properties were selected for illustrative purposes, namely benzo[a]pyrene (B(a)P), 1,2-dichloroethane (DCE), and hexachlorobenzene (HCB). Table 1 lists selected properties for these substances and objective and effect risk endpoints. An arbitrary unit emission rate of 1000 kg/h (1 t/h) was selected for Level II and Level III model simulations. Level III MOE scenarios included unit emissions exclusively to air, water, and soil, and a fourth with equal emissions of 3331/3 kg/h to all three media. Model output data for unit concentrations in all abiotic and biotic concentrations for five different model scenarios are summarized in Table SI-8. The calculations for critical emission rates and RAFs are described below for objective then effect endpoints and summarized in Tables 2 and 3, respectively. Objective Endpoints. Table 2 provides a summary of objective endpoint concentrations, unit concentrations, critical emission rates, estimated actual emission rates, RAFs, and relative screening level risk rankings for each of the three chemicals. B(a)P is a hydrophobic, relatively nonvolatile and environmentally immobile chemical. For B(a)P Level III unit emissions to air the concentration in air is the most sensitive objective endpoint. The unit emission concentration in air is 65 ng/m3 and the objective endpoint is 1.14 ng/m3 (Table 1) resulting in a RQ of 57.2. The critical emission rate is thus 1000/57.2 or 17.5 kg/h or 153.1 t/year. An estimated emission rate for B(a)P in 2003 in Canada is 24.1 t/year (19), thus the RAF is 24.1/153.1 or 0.157. For B(a)P emissions to water, 2318

9


TABLE 1. Physical-Chemical Properties (22), Selected Media-Specific Half-Lives, and Endpoints for Three Illustrative Chemicals (dw ) dry weight; ww ) wet weight) B(a)P

DCE

physical-chemical properties molar mass, g/mol 252.32 98.96 water solubility, g/m3 3.8 × 10-4 8.6 × 103 vapor pressure, Pa 7.0 × 10-7 1.1 × 104 log KOW 6.04 1.48 in air in water in soil in sediment

half-lives, days 0.2 200 600 1800

40.2 40 120 360

HCB 284.78 5.0 × 10-3 2.3 × 10-3 5.50 387.5 900 1800 5400

in airb, ww in waterc, ww in soild, dw in sedimente, dw

objective endpoint,a g/m3 1.14 × 10-9 3.85 × 10-8 2.17 × 10-9 1.0 × 10-5 5.0 × 10-3 5.2 × 10-4 2.4 × 10-1 2.4 × 10-1 1.2 × 10-1 7.66 × 10-2 1.2 × 10-1 4.8 × 10-2

in biota, ww

effect endpoint,f mmol/kg 9.7 × 10-3 5.0

5.0

a

Environmental quality guidelines from Oak Ridge National Laboratory (23) and Canadian Council of Ministers of the Environment (CCME) (24). b Set equal to the Reference Concentration for a threshold, i.e., noncarcinogenic substance, or calculated from the Inhalation Unit Risk using a risk level of one-in-a-million for a nonthreshold, i.e., carcinogenic substance. c Set equal to the lower of the CCME guideline for community water supplies and the CCME guideline for the protection of freshwater aquatic life. d Set equal to the CCME soil quality guideline for agricultural use. e Set equal to the CCME sediment quality guideline. f Critical body residue (CBR) calculated from selected “inherent Toxicity” value and “5% lipid BCF” model if empirical BCF unavailable (21).

sediment proves to be the most sensitive objective endpoint with a unit emission value of 48.7 g/m3 wet weight (244 g/m3 dry weight). The critical emission rate to water to achieve the objective of 0.0766 g/m3 is 1000/3180 or 0.315 kg/h (2.76 t/year) and the RAF is 8.74. For B(a)P unit emissions to soil a maximum objective risk quotient of 9.59 is calculated from the most sensitive objective endpoint, i.e., 0.24 g/m3 dry weight (soil), and the unit concentration in soil 2.3 g/m3 dry weight (1.15 g/m3 wet weight). The annual tonnage critical emission rate of 913 corresponds to a RAF of 0.0264. If unit emissions are equal to all three media then sediment is the

TABLE 2. Summary of Different Screening Level Risk Assessments for the Illustrative Chemicals Using Objective Endpoints chemical

model

MOEa

media of concern

CEb g/m3

CUb g/m3

EC t/year

EAc t/year

RAF EA/EC

relative rank

B(a)P

LII LIII LIII LIII LIII

N/A air water soil AWS

sediment air sediment soil sediment

7.66 × 10-2 1.14 × 10-9 7.66 × 10-2 2.40 × 10-1 7.66 × 10-2

4.38 6.50 × 10-8 2.44 × 102 2.3 82.1

1.53 × 102 1.53 × 102 2.76 9.13 × 102 8.18

24.1 24.1 24.1 24.1 24.1

1.57 × 10-1 1.57 × 10-1 8.74 2.64 × 10-2 2.95

1 1 1 1 1

DCE



air air air air air

3.85 × 10-8 3.85 × 10-8 3.85 × 10-8 3.85 × 10-8 3.85 × 10-8

9.30 × 10-7 9.31 × 10-7 6.88 × 10-7 8.96 × 10-7 8.38 × 10-7

3.63 × 102 3.62 × 102 4.90 × 102 3.77 × 102 4.02 × 102

9.52 9.52 9.52 9.52 9.52

2.63 × 10-2 2.63 × 10-2 1.94 × 10-2 2.53 × 10-2 2.37 × 10-2

2 2 2 2 2

HCB



air air sediment air sediment

2.17 × 10-9 2.17 × 10-9 4.80 × 10-2 2.17 × 10-9 4.80 × 10-2

9.78 × 10-7 9.89 × 10-7 69.6 1.17 × 10-7 23.6

19.5 19.2 6.04 1.63 × 102 17.8

2.95 × 10-2 2.95 × 10-2 2.95 × 10-2 2.95 × 10-2 2.95 × 10-2

1.51 × 10-3 1.53 × 10-3 4.88 × 10-3 1.81 × 10-4 1.66 × 10-3

3 3 3 3 3

a

MOE ) mode of entry; AWS ) air, water, soil.

b

Soil and sediment are on a dry weight basis. c Estimated actual emission rates (19).

TABLE 3. Summary of Different Screening Level Risk Assessments for the Illustrative Chemicals Using Effect Endpoints CE mmol/kg

CU mmol/kg

EC t/yr

EAb t/yr

RAF EA/EC

relative rank

9.71 × 10-3 9.71 × 10-3 9.71 × 10-3 9.71 × 10-3 9.71 × 10-3

18.2 2.22 × 102 6.72 × 102 2.26 2.27 × 102

4.67 3.83 × 10-1 1.27 × 10-1 37.7 3.75 × 10-1

24.1 24.1 24.1 24.1 24.1

5.15 62.8 1.90 × 102 6.39 × 10-1 64.1

1 1 1 1 1

aquatic mammal aquatic mammal piscivorous fish terrestrial invertebrate terrestrial invertebrate

5.0 5.0 5.0 5.0 5.0

1.96 × 10-6 1.96 × 10-6 1.13 × 10-4 1.60 × 10-4 5.37 × 10-5

2.23 × 1010 2.23 × 1010 3.87 × 108 2.73 × 108 8.16 × 108

9.52 9.52 9.52 9.52 9.52

4.27 × 10-10 4.26 × 10-10 2.46 × 10-8 3.49 × 10-8 1.17 × 10-8

3 3 3 3 3

aquatic mammal aquatic mammal aquatic mammal terrestrial carnivore aquatic mammal

5.0 5.0 5.0 5.0 5.0

3.08 × 10-1 4.95 × 10-1 62.7 2.80 21.3

1.42 × 105 8.85 × 104 6.99 × 102 1.56 × 104 2.06 × 103

2.95× 10-2 2.95× 10-2 2.95 × 10-2 2.95 × 10-2 2.95 × 10-2

2.08 × 10-7 3.33 × 10-7 4.22 × 10-5 1.88 × 10-6 1.43 × 10-5

2 2 2 2 2

chemical

model

MOEa

B(a)P



aquatic mammal terrestrial carnivore aquatic mammal terrestrial carnivore aquatic mammal

DCE



HCB



a

media of concern

MOE ) mode of entry; AWS ) air, water, soil.

b

Estimated actual emission rates (19).

most sensitive objective endpoint producing the largest RQ of 1071 and a RAF of 2.95. These RAFs indicate that, depending on MOE, B(a)P has the potential to exceed objectives. For the Level II model, which does not require MOE information, the sediment is the most sensitive objective endpoint with a RAF of 0.157. For DCE, a highly volatile and mobile chemical with a high Henry’s Law constant and low hydrophobicity, air is the most sensitive endpoint for all scenarios with low RAFs between 0.019 and 0.026. The Level II results are similar because of the similar overall environmental partitioning calculated by these models, i.e., percent in air. This suggests that the selection of model and MOE for Level III may be less important for certain mobile and volatile chemicals. HCB is a highly persistent, hydrophobic, and semivolatile chemical. The most sensitive objective endpoint is sediment when Level III emissions are to water or equal to air, water, and soil. The RAF is greatest (approximately 0.005) when HCB emissions are directly to water, reflecting the partitioning from water to organic matter in the water column and transport to sediment. Air is the most sensitive objective endpoint when HCB is emitted to air or soil. The RAF is lowest (approximately 0.0002) for emissions to soil. Level II results indicate that air is the most sensitive objective endpoint with a RAF of approximately 0.0015. Effect Endpoints. Acute lethality is selected as a welldefined and consistent evaluative “effect” endpoint. For polar

and nonpolar narcotic chemicals this value is approximately 5 mmol/kg wet weight. For substances that exert a more selective mode of toxic action (acute), CBR is derived from the toxic ratio (TR) of these chemicals to that of narcotics as suggested by Maeder et al. (20). A chemical with a TR greater than or equal to 10 is considered a selective toxicant while others are considered to be narcotic. For this case study CBRs and TRs were calculated by selecting data from Environment Canada’s Domestic Substances List (DSL) “inherent Toxicity” (iT) and bioconcentration factor database (21) as summarized in the SI. Acute toxicity was selected rather than extrapolation to chronic levels since this may introduce additional uncertainty for chemicals which exhibit toxicity greater than baseline. Table 3 lists effect endpoint concentrations, unit concentrations, critical emission rates, estimated actual emission rates (19), RAFs, and relative screening level risk rankings for the three chemicals. For B(a)P, the most sensitive species depend on MOE; however, assuming negligible metabolic transformation they are consistently upper trophic level organisms. For Level III calculations terrestrial organisms are the most sensitive species for emissions to air and soil, whereas aquatic mammals are the most sensitive for emissions to water and for equal emissions to air, water, and soil. The screening level RAFs are near unity or greater for all model scenarios. Of the three chemicals B(a)P is identified as being the most acutely toxic. VOL. 40, NO. 7, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2319

TABLE 4. Seventy Chemicals Binned According to Their Risk Assessment Factora (RAF) and Relatively Ranked for Priority Assessment from Level II and Level IIIb Models Based on “Effect” Endpoints ID no.

b

CAS RN

1 2 3 4 5 6 7 8 9 10 11 12 13

50-00-0 50-32-8 51-28-5 67-56-1 67-63-0 67-66-3 67-72-1 68-12-2 70-30-4 71-36-3 71-43-2 74-85-1 76-60-8

14

77-54-3

15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

78-87-5 78-93-3 79-01-6 79-34-5 79-92-5 79-94-7 84-74-2 85-01-8 85-68-7 87-86-5 91-20-3 95-50-1 95-63-6 101-61-1 104-40-5 106-46-7 107-06-2 107-21-1 108-10-1 108-88-3 108-90-7 110-54-3 110-82-7 111-42-2 115-07-1 117-81-7 118-74-1 120-12-7 120-83-2 124-18-5 127-18-4 129-00-0 603-35-0 608-93-5 1163-19-5 1222-05-5

51 52 53 54 55 56 57 58 59 60

1330-78-5 1582-09-8 1897-45-6 1912-24-9 3380-34-5 3846-71-7 4979-32-2 5285-60-9 6386-38-5 13560-89-9

61 62 63 64 65 66 67

15646-96-5 23593-75-1 25973-55-1 26140-60-3 32534-81-9 32536-52-0 54079-53-7

68 69 70

54827-17-7 63936-56-1 89347-09-1

chemical name

L II bin

L III bin

L II rank

L III rank

formaldehyde benzo[a]pyrene (BaP) phenol, 2,4-dinitromethanol 2-propanol (isopropyl alcohol) methane, trichloro- (chloroform) ethane, hexachloroformamide, N,N-dimethylphenol, 2,2′-methylenebis[3,4,6-trichloro- (hexachlorophene) 1-butanol (n-butyl alcohol) benzene ethene (ethylene) phenol, 4,4 -(3H-2,1-benzoxathiol-3-ylidene)bis [2,6-dibromo-3-methyl-, S,S-dioxide (bromocresol green) 1H-3a,7-methanoazulen-6-ol, octahydro-3,6,8,8-tetramethyl-,acetate, [3R-(3a′,3a,6a′,7,8aa′)]propane, 1,2-dichloro 2-butanone (methyl ethyl ketone) ethene, trichloro- (TCE) ethane, 1,1,2,2-tetrachlorobicyclo[2.2.1]heptane, 2,2-dimethyl-3-methylene (camphene) phenol, 4,4 -(1-methylethylidene)bis[2,6-dibromo- (tetrabromobisphenol A) 1,2-benzenedicarboxylic acid, dibutyl ester (DBP) phenanthrene 1,2-benzenedicarboxylic acid, butyl phenylmethyl ester (BBP) phenol, pentachloronaphthalene benzene, 1,2-dichloro benzene, 1,2,4-trimethylbenzenamine, 4,4 -methylenebis[N,N-dimethyl- (Michler’s hydride) phenol, 4-nonylbenzene, 1,4-dichloro ethane, 1,2-dichloro (DCE) 1,2-ethanediol 2-pentanone, 4-methyl (methyl isobutyl ketone) benzene, methylbenzene, chlorohexane cyclohexane ethanol, 2,2′-imino-bis (diethanolamine) 1-propene (propylene) 1,2-benzenedicarboxylic acid, bis(2-ethylhexyl) ester (DEHP) hexachlorobenzene (HCB) anthracene phenol, 2,4-dichloro- (DCP) decane ethene, tetrachloro- (PCE) pyrene phosphine, triphenylbenzene, pentachlorobenzene, 1,1-oxybis[2,3,4,5,6-pentabromocyclopenta g -2-benzopyran, 1,3,4,6,7,8-hexahydro-4,6,6,7,8,8-hexamethyl(musk 50) phosphoric acid, tris(methylphenyl) ester benzenamine, 2,6-dinitro-N,N-dipropyl-4-(trifluoromethyl)1,3-benzenedicarbonitrile, 2,4,5,6-tetrachloro- (chlorothalonil) 1,3,5-triazine-2,4-diamine, 6-chloro-N-ethyl-N′-(1-methylethyl)- (atrazine) phenol, 5-chloro-2-(2,4-dichlorophenoxy)- (triclosan) phenol, 2-(2H-benzotriazol-2-yl)-4,6-bis(1,1-dimethylethyl)2-benzothiazolesulfenamide, N,N-dicyclohexylbenzenamine, 4,4 -methylenebis[N-(1-methylpropyl)benzenepropanoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy-, methyl ester 1,4:7,10-dimethanodibenzo[a,e]cyclooctene, 1,2,3,4,7,8,9,10,13,13,14,14-dodecachloro-1,4,4a,5,6,6a,7,10,10a,11,12, 12a-dodecahydrohexane, 1,6-diisocyanato-2,4,4-trimethyl1H-imidazole, 1-[(2-chlorophenyl)diphenylmethyl]phenol, 2-(2H-benzotriazol-2-yl)-4,6-bis(1,1-dimethylpropyl)terphenyl benzene, 1,1 -oxybis-, pentabromo deriv. benzene, 1,1 -oxybis-, octabromo deriv. propanedinitrile, [[4-[[2-(4-cyclohexylphenoxy)ethyl]ethylamino]-2-methylphenyl]methylene][1,1 -biphenyl]-4,4 -diamine, 3,3,5,5 -tetramethylbenzene, pentabromo(tetrabromophenoxy)1,3,4-thiadiazole, 2,5-bis(tert-nonyldithio)-

E B C E F G G F B F F G D

E B A E F G G C A F F G C

40 6 17 38 45 67 69 58 4 47 53 62 23

45 10 4 42 47 67 69 19 3 48 55 61 20

D

D

34

40

F F F G G B E C E B D F F E C F G E F F G F F E G D F C D G G B D E D D

F F F G G A D C D A D F F E C F G D F F G F F C G D F B D G G B C E C D

52 51 50 70 65 5 36 15 35 3 31 59 48 41 13 56 68 43 55 46 64 57 60 42 66 25 49 8 27 61 63 1 28 44 30 26

54 53 52 70 65 2 32 30 34 9 38 59 50 44 24 57 68 36 56 49 64 58 60 28 66 33 51 16 35 62 63 11 29 46 23 37

C D B D B D D C D D

B D A A A B C B C C

14 32 7 21 2 20 33 12 29 18

14 39 6 1 5 13 27 18 26 22

E E D D C C D

E D B C C A B

39 37 19 22 16 9 24

43 31 15 21 25 7 17

F C C

E A B

54 10 11

41 8 12

a A: RAF g 102; B: 100 e RAF < 102; C: 10-2 e RAF < 100; D: 10-4 e RAF < 10-2; E: 10-6 e RAF < 10-4; F: 10-8 e RAF < 10-6; G: RAF < 10-8. Level III model results presented based on emissions to air.

2320

9


FIGURE 3. “Chemical space” occupied by 70 case study substances. The long-dashed line represents log KOA ) 16 and the short-dashed line represents log KOA ) 0. For DCE, the most sensitive organisms again depend on MOE. Level III emissions to air and to water identify an aquatic or marine species as having the greatest risk quotient, albeit with very low RAFs (10-8 to 10-10). Level III emissions to water result in a higher-lipid-content fish being the most sensitive effect endpoint. Initially this seems counterintuitive based on the high volatility of the chemical; however, a minimal quantity in the water results in modest bioconcentration in fish. The chemical does not biomagnify but combined exposure from the air and the diet render the aquatic mammal the most sensitive effect endpoint for Level II and Level III emissions to air. DCE emissions directly to soil and equal emissions to air, water, and soil have the greatest impact on the terrestrial invertebrate. For HCB, the most sensitive effect endpoint is the aquatic mammal for all scenarios except Level III emissions to soil (terrestrial carnivore). HCB has the potential to biomagnify in each of the representative food webs resulting in higher internal concentrations at higher trophic levels (see Table SI-8). Because HCB is regulated to low emission rates and it has low baseline toxicity the RAF is low. The relative screening level risk ranks B(a)P first, followed by DCE, and HCB for each scenario according to objective endpoints (Table 2). The effect endpoint ranks the chemicals as B(a)P, HCB, and DCE from highest to lowest RAF for each scenario (Table 3). The most sensitive effect endpoint or target organism changes between scenarios; however, the relative ranking remains consistent within the Level II and the four Level III MOE model selections. Application to 70 Chemicals. To illustrate the concept more broadly, 70 chemicals were selected from the DSL as summarized in Table 4. Property data, effect endpoints, estimated emission rates, and summary model output are given in the SI. Figure 3 depicts some properties of these substances in a “chemical space” diagram and shows the wide variation in partitioning properties. For ionizing chemicals, the model requires water solubility and octanolwater partition coefficient (KOW) data for the neutral species as well as pKa. Objective endpoints were only available for few chemicals, therefore acute lethality was chosen as a consistent effect endpoint for all 70 chemicals. Figure 4 illustrates the results from Level II and Level III calculations as a plot of logarithms of the RAF for the 70 substances. This figure suggests that chemicals can be “binned” according to the RAF. Table 4 lists the bin results and ranks the 70 chemicals according to the RAF for Level II and Level III emissions to air. Figure 4 and Table 4 indicate that RAF bin categories for effect endpoint may vary somewhat between Level II and Level III models; however, the relative ranking of chemicals is similar (Figure SI-3).

FIGURE 4. Comparison of Level II and Level III risk assessment factors (RAF) and risk identification “bins” for 70 chemicals. Level II (0) and Level III emissions to air (b), water (2), soil ([), and equal emissions to air, water, and soil (×). Results including the most sensitive endpoint, critical emission rate, RAF, risk identification “bin”, and relative rank are summarized for the Level II and for each of the four Level III model scenarios in the SI. A relatively high RAF clearly indicates a need for further assessment but some substances with low RAFs may require more detailed considerations, especially if emissions are localized. This illustration consistently ranks certain phenolic compounds, i.e., tetrabromobisphenol A, hexachlorophene, triclosan, and pentachlorophenol, as well as some polycyclic aromatic hydrocarbons (PAHs), i.e., B(a)P and pyrene, and certain PBDEs as having relatively greater risk potential than many other chemicals from classes such as hydrocarbons and chlorinated hydrocarbons. Many of these higher ranking phenols and PAHs are identified as possessing a mode of action more potent than narcosis reflecting the fact that empirical “iT” values selected for this case study are from the most sensitive species. For example, photoinduced toxicity values are included for many of the PAHs. Substances identified as being more selectively toxic generally ranked higher than narcotics according to the Level II and Level III model scenarios. However, certain acutely narcotic chemicals that are persistent, such as certain PBDE derivatives, consistently rank in the top 10. This evaluation illustrates that risk ranking is dependent on emission quantity, persistence, routes of exposure, and the intrinsic toxicity of the chemical. Figure 4 should not be considered indicative of the universe of commercial chemicals. Many of these 70 chemicals are included in programs previously developed to assess the risk of chemicals identified as potential concern to humans and the environment. Thus they represent a skewed subset of commercial chemicals and a more “random sample” of chemicals may rank lower by comparison. Although model VOL. 40, NO. 7, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2321

uncertainty may be significant the discrepancy between high and low RAF rankings may be very apparent for most chemicals. Ranking is relative to the selected endpoint and the chemicals included in the evaluation, i.e., a RAF less than unity may still represent a potential risk.

Discussion Consideration of these results leads to certain key issues being identified as critical to the successful implementation and improvement of the model. RAIDAR is structured to bring together information on chemical partitioning, reactivity, environmental fate and transport, bioaccumulation, exposure, endpoints, and emission rates in a coherent mass balance framework to yield an overall estimate of risk at a screening level. This evaluation demonstrates that when ranking chemicals on the basis of risk or binning them into groups of similar risk it is essential to consider both intrinsic toxicity and exposure as influenced by quantity released to the environment and subsequent fate and transport. A substance with relatively low toxicity that is persistent and bioaccumulative can pose greater risk compared to more potent substances. It is hoped that the approach may be useful for identifying those chemicals of commerce of greatest potential concern in order that they can be more fully and accurately evaluated using monitoring data in conjunction with models describing chemical fate and effects in more detail in actual regional environments and food webs. Emission Data and Level II and Level III Simulations. The uncertainties of quantity released and MOE are significant for many chemicals that require preliminary assessment. The model demonstrates the importance of having available reliable emission rate data. An advantage of the RAIDAR modeling approach is that highly uncertain emission rates are not used as direct input to the model where they could reduce the credibility of the results. They are only used at the end of the assessment when evaluating the magnitude of the RAF, i.e., EA/EC. Changes in import, production, or the use of the chemical for significantly different purposes can be assessed in advance by inspecting the RAF. Emission rate data may be subject to restrictions due to commercial confidentiality but often an approximate industry-wide estimate may be sufficient. In many cases a low risk ranking demonstrates that there is no need for better than orderof-magnitude emission estimates. Thus, the model can help focus efforts to establish emissions or release inventories as well as other key empirical data where they are most needed (e.g., reaction rates, monitoring). We suggest that “binning” the substances into ranges of RAF provides a useful screening level risk assessment for prioritizing large numbers of chemicals for further evaluations. The assessment can be run for Level II or Level III fate conditions and with experience guidance can be obtained on a preferred approach. A tiered strategy may be desirable. In many cases the Level II and III results are similar suggesting that the simpler Level II model may suffice. For example, it is unlikely that the more demanding Level III simulation will increase the screening level risk identification bin from a Level II simulation that identifies chemicals binned as “F” or “G” to “A”, “B”, or “C”. The results show that the Level II and III distributions are similar when Level III MOE is to a compartment into which the chemical tends to partition. Results tend to be most different when emission is into a compartment to which the chemical tends not to partition and the most sensitive endpoint is in that compartment. If MOE data are available for the chemical they should be applied and emission factors dependent on physicalchemical properties such as volatility can also be used as in EUSES (2). In the absence of information the “one-third to each” MOE approach is a reasonable alternative. 2322

9


Model Limitations. Models such as RAIDAR necessarily present a simplification of the complex environmental behavior of chemicals. The resulting limitations must therefore be appreciated. In light of experience, model structure and parameters are expected to require revision. For example, polyparameter Linear Free Energy Relationships (ppLFERs) could be included to improve assessments for chemicals whose partitioning may be poorly described by a single parameter. Currently, certain chemical groups such as pigments, and dyes, and perfluorinated and ionizing substances may not be well simulated. The model is designed to assess the “farfield” impact of chemical exposure and does not presently include “nearfield” sources such as indoor or product use exposure for humans in domestic or industrial settings. The model’s credibility can be established by comparing its intermedia partitioning predictions to empirical and monitoring data (e.g., air-water concentration ratios, BAFs). All models, including RAIDAR, require accurate data on transport rate coefficients and physical-chemical properties, especially biodegradability. Metabolic transformation rate data can be included in the food webs, but in absence of accurate values negligible transformation is assumed. It may be desirable to allow a degree of metabolic conversion for hydrophobic substances known to be subject to metabolism (e.g., PAHs). Endpoints. If regulatory environmental objectives are available there is no need to consider the food web results, thus simplifying the assessment. Effect or toxicity endpoints require that there be consistency for all chemicals, for example the selection of either an acute or chronic toxicity CBR or estimated no-effect concentration. Calculated RAFs and the relative ranking of chemicals may vary according to the selected endpoint. When chemicals have widely different adverse effects such as short-term acute toxicity versus reproductive effects there is no substitute for inspection of the results for toxicological significance and ranking or binning may be impossible. Definitive toxicity data for each representative species is ideal. The role of the model is to highlight the likely concentrations for detailed consideration. Endpoint selection and interpretation of the results are then critically important.

Acknowledgments We dedicate this paper to Dr. Tom Feijtel of Procter and Gamble (Brussels) who died in a tragic accident in September 2005. Inspiration for this paper arose from a conversation with Dr. Feijtel at the 2002 SETAC Europe meeting in Vienna in which he advocated a strategy for the early comparison of likely release quantities of commercial chemical with quantities back-calculated using multimedia models. This paper has been a direct result of sharing his experience and insights into hazard and risk assessment of chemicals. The authors thank Environment Canada, Health Canada, the Natural Sciences and Engineering Research Council of Canada (NSERC), and the consortium of chemical companies that support research at the Canadian Environmental Modelling Centre.

Supporting Information Available More detailed description of the RAIDAR model and illustrative model output. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Vermeire, T. G.; Jager, D. T.; Bussian, B.; Devillers, J.; den Haan, K.; Hansen, B.; Lundberg, I.; Niessen, H.; Robertson, S.; Tyle, H.; van der Zandt, P. T. J. European Union System for the Evaluation of Substances (EUSES). Principles and structure. Chemosphere 1997, 34, 1823-1836.

(2) Vermeire, T. G.; Rikken, M.; Attias, L.; Boccardi, P.; Boeije, G.; Brooke, D.; de Bruijn, J.; Comber, M.; Dolan, B.; Fischer, S.; Heinemeyer, G.; Koch, V.; Lijzen, J.; Muller, B.; Murray-Smith, R.; Tadeo, J. European Union System for the Evaluation of Substances (EUSES): the second version. Chemosphere 2005, 59, 473-485. (3) Webster, E.; Mackay, D.; Di Guardo, A.; Kane, D.; Woodfine, D. Regional differences in chemical fate model outcome. Chemosphere 2004, 55, 1361-1376. (4) McKone, T. E. CALTOX, A multimedia total exposure model for hazardous-waste sites; UCRL-CR-111456PtI-IV; U.S. Department of Energy, Lawrence Livermore National Laboratory; U.S. Government Printing Office: Washington, DC, 1993. (5) U.S. EPA PBT Profiler. http://www.pbtprofiler.net/default.asp. (6) U.S. EPA EPI (Estimation Programs Interface) Suite, version 3.12; November 30, 2004. http://www.epa.gov/oppt/exposure/ docs/episuitedl.htm. (7) Moore, D. R. J.; Breton, R. L.; MacDonald, D. B. A comparison of model performance for six Quantitative Structure-Activity Relationship packages that predict acute toxicity to fish. Environ. Toxicol. Chem. 2003, 22, 1799-1809. (8) Mackay, D.; Hubbarde, J.; Webster, E. The role of QSARs and fate models in chemical hazard and risk assessment. QSAR Comb. Sci. 2003, 22, 106-112. (9) McCarty, L. S.; Mackay, D. Enhancing ecotoxicological modeling and assessment. Environ. Sci. Technol. 1993, 27, 1719-1728. (10) Escher, B. I.; Hermens, J. L. M. Modes of action in ecotoxicology: Their role in body burdens, species sensitivity, QSARs, and mixture effects. Environ. Sci. Technol. 2002, 36, 4201-4217. (11) 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, 743-748. (12) 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, 39-58. (13) Lipnick, R. L.; Watson, K. R.; Strausz, A. K. A QSAR study of the acute toxicity of some industrial organic-chemicals to Goldfish: Narcosis, electrophile and proelectrophile mechanisms. Xenobiotica 1987, 17, 1011-1025. (14) Verhaar, H. J. M.; Van Leeuwen, C. J.; Hermens, J. L. M. Classifying environmental pollutants. 1. Structure-Activity Relationships

(15)

(16)

(17)

(18)

(19) (20)

(21) (22)

(23)

(24)

for prediction of aquatic toxicity. Chemosphere 1992, 25, 471491. Mackay, D.; Di Guardo, A.; Paterson, S.; Cowan, C. Evaluating the environmental fate of a variety of types of chemicals using the EQC model. Environ. Toxicol. Chem. 1996, 15, 1627-1637. Cousins, I. T.; Mackay, D. Multimedia mass balance modelling of two phthalate esters by the Regional Population-based Model (RPM). In The Handbook of Environmental Chemistry 3Q; Staples, C. A., Ed.; Springer-Verlag: Berlin, 2003; pp 179-200. MacLeod, M.; McKone, T. E. Multimedia persistence as an indicator of potential for population-level intake of environmental contaminants. Environ. Toxicol. Chem. 2004, 23, 24652472. Beyer, A.; Mackay, D.; Matthies, M.; Wania, F.; Webster, E. Assessing long-range transport potential of persistent organic pollutants. Environ. Sci. Technol. 2000, 34, 699-703. Environment Canada. National Pollutant Release Inventory 2003; April 13, 2005. http://www.ec.gc.ca/pdb/npri/npri_home_e.cfm. 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, 3659-3666. Environment Canada. Domestic Substances List - Preliminary Ecological Categorization Decisions; April 2005. (CD-ROM). Mackay, D.; Shiu, W. Y.; Ma, K. C. Physical-Chemical Properties and Environmental Fate and Degradation Handbook; CRCnetBASE 2000, Chapman & Hall CRCnetBASE; CRC Press LLC: Boca Raton, FL. (CD-ROM). ORNL. Risk Assessment Information System Ecological Benchmarks Search Tool; Oak Ridge National Laboratory, Environmental Sciences Division: Oak Ridge, TN, September 2003. http://risk.lsd.ornl.gov/homepage/benchmark.shtml. CCME. Canadian Environmental Quality Guidelines; The Canadian Council of Ministers of the Environment, December 2003. http://www.ccme.ca/.

Received for review July 19, 2005. Revised manuscript received January 26, 2006. Accepted January 27, 2006. ES0514085

VOL. 40, NO. 7, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2323

Screening Level Risk Assessment Model for Chemical Fate and

Recommend Documents