Chemosphere 120 (2015) 697–705

Contents lists available at ScienceDirect

Chemosphere journal homepage: www.elsevier.com/locate/chemosphere

Review

Cumulative risk assessment lessons learned: A review of case studies and issue papers Sarah S. Gallagher a,⇑, Glenn E. Rice b, Louis J. Scarano c, Linda K. Teuschler b, George Bollweg d, Lawrence Martin e a American Association for the Advancement of Science (AAAS) Science and Technology Policy Fellow, U.S. Environmental Protection Agency (EPA) Office of the Science Advisor, 1200 Pennsylvania Ave. NW., Washington, DC 20460, USA b U.S. EPA National Center for Environmental Assessment, 26 W. Martin Luther King Dr., Cincinnati, OH 45268, USA c U.S. EPA Office of Pollution Prevention and Toxics, 1200 Pennsylvania Ave. NW., Washington, DC 20460, USA d U.S. EPA Region 5 Air and Radiation Division, 77 W. Jackson Blvd., Chicago, IL 60604, USA e U.S. EPA Office of the Science Advisor, 1200 Pennsylvania Ave. NW., Washington, DC 20460, USA

h i g h l i g h t s  Authors identify six lessons learned from ten US EPA cumulative risk assessments.  Due to a population focus cumulative risk assessments need engaged stakeholders.  Tiering can focus the scope of cumulative risk assessments and prioritize stressors.  An iterative approach for cumulative assessments reduces complications of multiple stressors.  Quantifying risks in vulnerable populations is important, but data gaps remain.

a r t i c l e

i n f o

Article history: Received 24 March 2014 Received in revised form 9 October 2014 Accepted 11 October 2014

Handling Editor: Tamara S. Galloway Keywords: Cumulative risk assessment Vulnerability Stakeholder involvement Tiered approach Chemical mixture Multiroute exposure

a b s t r a c t Cumulative risk assessments (CRAs) examine potential risks posed by exposure to multiple and sometimes disparate environmental stressors. CRAs are more resource intensive than single chemical assessments, and pose additional challenges and sources of uncertainty. CRAs may examine the impact of several factors on risk, including exposure magnitude and timing, chemical mixture composition, as well as physical, biological, or psychosocial stressors. CRAs are meant to increase the relevance of risk assessments, providing decision makers with information based on real world exposure scenarios that improve the characterization of actual risks and hazards. The U.S. Environmental Protection Agency has evaluated a number of CRAs, performed by or commissioned for the Agency, to seek insight into CRA concepts, methods, and lessons learned. In this article, ten case studies and five issue papers on key CRA topics are examined and a set of lessons learned are identified for CRA implementation. The lessons address the iterative nature of CRAs, importance of considering vulnerability, need for stakeholder engagement, value of a tiered approach, new methods to assess multiroute exposures to chemical mixtures, and the impact of geographical scale on approach and purpose. Ó 2014 Published by Elsevier Ltd.

Abbreviations: BMD10, benchmark dose estimate for 10%; CAG, Common Assessment Group; CEP, Community Environmental Partnership; CRA, cumulative risk assessment; CRPF, cumulative relative potency factor; DBP, disinfection byproduct; EPA, Environmental Protection Agency; ERA, ecological risk assessment; FQPA, Food Quality Protection Act; HAP, hazardous air pollutant; HI, hazard index; HQ, hazard quotient; ICED, index chemical equivalent dose; MOA, mode of action; NATA, National-Scale Air Toxics Assessment; OP, organophosphorus pesticides; OPP, Office of Pesticide Programs; RAIMI, Regional Air Impact Modeling Initiative; ReVA, Regional Vulnerability Assessment; RPF, relative potency factor. ⇑ Corresponding author at: Oak Ridge Institute for Science and Education Fellow, U.S. EPA Office of Solid Waste and Emergency Response, 1200 Pennsylvania Ave. NW., Washington, DC 20460, USA. Tel.: +1 202 566 1394. E-mail address: [email protected] (S.S. Gallagher). http://dx.doi.org/10.1016/j.chemosphere.2014.10.030 0045-6535/Ó 2014 Published by Elsevier Ltd.

698

S.S. Gallagher et al. / Chemosphere 120 (2015) 697–705

Contents 1. 2. 3.

4. 5.

Background and purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Scope of the paper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lessons learned from the case studies and issue papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Lesson 1: iterative nature of CRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Lesson 2: importance of vulnerability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Lesson 3: importance of stakeholder engagement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Lesson 4: the value of a tiered approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5. Lesson 5: the need for new methods for CRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6. Lesson 6: implications of spatial scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Background and purpose Evaluations of environmental hazards and human health risks are expanding from single chemical or simple chemical mixture approaches to more comprehensive approaches that examine risks posed by exposures to multiple stressors, including chemical, physical, biological, and psychosocial stressors. Traditional environmental risk assessment approaches focus on chemical or microbial hazards, independently of other hazards or stressors. Attempting to analyze ‘‘real world’’ exposures and improve the accuracy of the characterization of risks, cumulative risk assessments (CRAs) examine human health and environmental risks from the perspective that populations are exposed simultaneously to multiple stressors via multiple exposure routes and pathways (Callahan and Sexton, 2007). The U.S. Environmental Protection Agency (EPA) has long recognized the potential importance of expanding the focus of risk assessment activities beyond single chemicals (Browner, 1995). In 2003, the EPA published the Framework for Cumulative Risk Assessment (herein called the Framework) (US EPA, 2003b). The Framework defines cumulative risk as ‘‘the combined risks from aggregate exposures to multiple agents or stressors’’ and emphasizes considering population vulnerabilities. After its publication, the EPA initiated two efforts to amass information for developing CRA Guidelines. First, the EPA collected Agency assessments that addressed one or more aspects of CRA and examined the utility of the methods used in these CRA ‘‘case studies’’. Second, the EPA commissioned five issue papers that investigated key CRA topics deemed critical to understand and improve the accuracy of risks predicted by CRA methods. 2. Scope of the paper This article describes six key ‘‘lessons learned’’ from the CRA case studies (Table 1) and issue papers (Table 2). While evaluating these individually to determine their contributions to critical aspects of CRA, we note that several of these lessons apply generally to risk assessment practices. The case studies highlight approaches that EPA program and regional offices and research programs developed to address risks posed by exposures to multiple stressors. They reflect varied geographic scales, including CRAs conducted at both national and community levels, and varied scope, with three evaluating ecological endpoints, six evaluating human health endpoints, and one evaluating both human health and ecological endpoints separately (rather than in an integrated manner). The five issue papers (Table 2) investigated specific topics that include: articulating the challenges to conducting CRAs, evaluating

698 698 698 698 699 700 701 702 702 703 704 704 704

combined health effects from multiple stressors, incorporating vulnerability into CRA, assessing environmental mixtures, and using biomarkers to inform CRA (Callahan and Sexton, 2007; DeFur et al., 2007; Menzie et al., 2007; Ryan et al., 2007; Sexton and Hattis, 2007).

3. Lessons learned from the case studies and issue papers 3.1. Lesson 1: iterative nature of CRA EPA’s Framework recognized that iteration would be essential in the conduct of CRAs. The three phases of a CRA -planning, scoping, and problem formulation; analysis; and risk characterizationwould not always be conducted unidirectionally (US EPA, 2003b). As different types of stressors and population vulnerabilities are identified and associated risks are characterized, the need for additional data on interactions among stressors of interest, other stressors that could be related (e.g., cause same health effect), or population vulnerabilities may be recognized and require the collection of additional data or further analysis. Similarly, during the conduct of a CRA, unanticipated risk management options may become apparent that could entail additional analysis or reconsideration of the risk assessment approach. These are simple examples of the role that an iterative approach may have during a CRA. In its recommendations for improving the utility of risk assessment, the National Research Council (2009) also identified this as being an important risk assessment practice. The Clinch and Powell Valley Ecological Risk Assessment exemplifies the iterative conduct of a CRA (Diamond et al., 2002; US EPA, 2002a). The assessment goal was to determine whether mining, urbanization, and agricultural activities in the watershed were adversely impacting fish and mussel species. The conceptual models for this assessment provided the following: (1) identify the exposed populations (i.e., fish, mussels); (2) identify possible sources of stressors (e.g., runoff from mining activities); (3) specify adverse effects (e.g., unacceptable losses of native fish); and (4) conceptualize the pathways by which the stressors impact assessment endpoints (e.g., runoff enters water bodies where fish reside). The relationships depicted in these conceptual models provided the initial inputs to the analysis plan. As the assessment progressed, unanticipated gaps were identified for specific stressor and effects data (e.g., a lack of water chemistry data prevented the analysis of relationships between water quality stressors and land uses, including urbanization). To overcome this challenge, the analytical approach was modified; nearby land-use activities along with habitat quality information served as surrogates for water quality stressor levels (US EPA, 2002a).

699

S.S. Gallagher et al. / Chemosphere 120 (2015) 697–705 Table 1 Summary of the ten case studies. Title

Goal

Results

Lessons learned

Identified key stressors (e.g., proximity to mining activities) and management approaches (e.g., restoration of abandoned mines)

Iteration enhanced development of an improved analysis plan. An initial conceptual model, depicting the relationships between land uses and stressors, was revised based on data availability and the preliminary analysis

Individual data integration methods can address different aspects of vulnerability and can be sensitive to different data problems. Used combinations of integration methods to address data limitations Importance of identifying vulnerable populations, including groups with preexisting diseases and elevated stressor exposures, who could experience adverse health outcomes from additional stressor exposures or are more likely to develop more severe outcomes than the general population

Iterative nature of cumulative risk assessment (CRA) Clinch and Powell Valley Watershed Ecological Risk Assessment (ERA) US EPA (2002a)

Establish a baseline for future watershed assessments and determine if multiple stressors were adversely impacting the fish and mussel species

Importance of considering vulnerability Regional Vulnerability Assessment for the MidAtlantic Region US EPA (2002e)

Develop a tool to identify ecosystems most vulnerable to loss or permanent harm and stressors posing the greatest risks

Developed a decision aid that identifies vulnerable ecosystems and stressors driving ecological change

Air Screening Assessment for Cook County, Illinois and Lake County, Indiana (Argonne National Laboratory, 2004)

Evaluate the cumulative cancer and noncancer inhalation hazards from air pollutants in the study area

Provided alternative tools for evaluating air quality. Cumulative hazard indicators (e.g., toxicity weights for emissionsa). Characterized sources/contributions of stressors, spatial distribution of stressors, and potentially vulnerable populations

Importance of stakeholder engagement Identify how stressors were impacting water quality in Waquoit Bay, and which were the most important to manage to improve water quality Evaluate impact of multiple emissions sources on local air quality. Identify effective prevention efforts to improve community health

Utilizing stakeholder input on environmental concerns and important local resources, identified stressor of greatest concern

Providing opportunities for stakeholder involvement is essential throughout the CRA

Involved stakeholders in initial project stages. Changes in scope due to resource and data constraints were communicated ineffectively

Consistent stakeholder engagement facilitates community understanding of goals, limits, and questions that a CRA can answer

Industrial Surface Impoundment Sites US EPA (2000c)

Estimate human health and ecological risks posed by managing wastes in surface impoundments, and ‘‘protectiveness’’ of regulations

Baltimore CEP Study US EPA (2000a)

Evaluate impact of multiple sources on local air quality; identify effective prevention efforts to improve community health

For human health, single chemical hazards/ risks aggregated across exposure pathways were generally below levels of concern. For ecological effects, the assessment suggested possibility of adverse effects Established an emission sources inventory for air pollutants. Using a three-tiered, screening approach, estimated aggregate exposure levels and identified chemicals of concern

Tiered approaches allow facilities that are unlikely to pose a risk to human health or the environment to be screened out, reducing the number of facilities requiring more resource-intensive assessment A risk-based screening approach was used to focus the assessment. Refined exposure estimates as tiers progressed

Examine health risks posed by multiple pathways of exposures to mixtures of OPs acting by a common mode of action Evaluate cumulative, multiroute exposures to DBPs and assess potential health risks

Compared the risk posed by OP exposures via different pathways, identified major sources of exposure, and evaluated if risk mitigation efforts effectively reduced exposure Developed method to evaluate groups of DBPs whose components caused toxicity via different modes of action. Estimated internal doses of DBPs based on multiroute exposure modeling

The Relative Potency Factors Approach is useful for evaluating mixture risk

National-Scale Air Toxics Assessment US EPA (2006a)

Assess potential health risks associated with inhalation exposure to 180 chemicals (individually and combined)

Regional Air Impact Modeling Initiative US EPA (2003c)

Identify a standard approach for permitting authorities to assess aggregate risks from exposure to multiple pollutants, sources, and pathways

Calculated cumulative cancer risks and noncancer hazards. Identified national and regional risk drivers. Developed model-tomonitor comparisons, and limitations of estimates Provided detailed contaminant-, source-, and receptor-specific information supporting the identification and prioritization of risk management opportunities

This national-scale assessment enabled EPA to identify specific hazardous air pollutants (HAPs) and emission sources that contributed the most to national population risk The finer resolution spatial level enables risk assessors to consider cumulative impacts on neighborhoods when evaluating facility siting proposals

Waquoit Bay ERA US EPA (2002f)

Baltimore Community Environmental Partnership (CEP) Study US EPA (2000a) Value of a tiered approach

New methods for chemical mixtures and multiroute exposure Organophosphorus Pesticide (OP) CRA US EPA (2002d, 2006a) Drinking Water Disinfection By-Product (DBP) Mixtures US EPA (2003a); Teuschler et al. (2004)

A probabilistic approach can be used to estimate DBP multiple route exposures. The Cumulative Relative Potency Factor approach can be used to estimate mixture risks

Implications of Spatial Scale

a

Emission mass (pounds per year) was multiplied by Risk Screening Environmental Indicators toxicity weights to estimate the relative hazard for each pollutant.

3.2. Lesson 2: importance of vulnerability Risk assessors have long recognized the importance of considering vulnerable populations in assessments of environmental chemicals; current approaches address some aspects of vulnerability, e.g., an uncertainty factor for sensitive human populations when deriving reference doses (US EPA, 2002c). However, gaps remain

in identifying and accurately quantifying risks in vulnerable populations; both for conventional, single chemical risk assessments and CRAs (Sexton and Hattis, 2007). Vulnerability includes predisposition to risk of harm from exposure to stressors due to biological susceptibility, or differential exposure, preparedness, or ability to recover (cited in US EPA, 2003b; Kasperson, Personal communication). Vulnerability implies

700

S.S. Gallagher et al. / Chemosphere 120 (2015) 697–705

Table 2 Summary of the five issue papers. Source

Main points

Callahan and Sexton (2007)

 Provides historical regulatory and scientific basis for transitioning from single chemical assessments to CRA  Identifies major challenges making CRAs more complex than single stressor risk assessments, including time-related exposure aspects, vulnerability, at-risk populations, psychosocial stress, and approaches for combining exposures and hazards

Menzie et al. (2007)

 Reviews importance of conceptual models for CRA  Describes screening approaches and analytical tools for assessing combined effects  Utilizes a phased approach for stressor-based and effects-based assessments

DeFur et al. (2007)

 Identifies and presents examples of human and ecological vulnerabilities that could affect CRAs  Proposes conceptual model for incorporating vulnerability into CRA

Sexton and Hattis (2007)

 Presents criteria for identifying high-priority mixtures  Identifies cumulative exposure assessment challenges, including background and differential exposures  Describes types of interactive effects that can affect response

Ryan et al. (2007)

 Illustrates how biomarkers can help elucidate linkages between environmental exposures and health outcomes  Presents a framework for using an array of biomarkers as a surrogate for an ideal biomarker

an altered relationship between stressors and receptors or changes in the shape of the dose-response curve for a given effect (DeFur et al., 2007). Differential exposure magnitudes, timing of exposures, and co-occurrence of stressors alter stressor-receptor relationships and can potentially render populations vulnerable (Sexton and Hattis, 2007). Population groups may be more vulnerable to adverse effects from environmental chemical exposures (1) during certain life stages (e.g., puberty), (2) when they have experienced previous chemical exposures, (3) due to genetic predisposition, or (4) due to specific preexisting diseases. Following an exposure, vulnerability can result in an increased likelihood of experiencing an adverse health outcome or developing more severe outcomes than the general population. The Air Screening Assessment for Cook County, Illinois, and Lake County, Indiana utilized geographic information system mapping to analyze spatial distributions of emissions data and estimate cumulative hazard levels. These levels were used to rank the relative hazard contribution of specific pollutants, sources, and locations. This assessment also included spatial analysis of vulnerable populations, as represented by disease rates/severity indices related to potentially increased exposure to air toxics (Argonne National Laboratory, 2004). The selected health conditions, diseases, and biomarkers included elevated blood lead concentration as well as leukemia, asthma, and acute respiratory infection incidence. By combining the spatial analysis, pollutant source, and disease rates by location the study identified geographic areas with potentially vulnerable populations. Vulnerability also is important in ecological CRAs. Proximity to urban or industrial areas can increase the number of stressors impacting an ecosystem. The Regional Vulnerability Assessment (ReVA) for the Mid-Atlantic Region developed a decision tool that identified vulnerable ecological areas to help prioritize risk management resources (US EPA, 2002e, 2003d). The ReVA developed spatial data sets, characterizing joint distributions of valued resources (e.g., water quality and wildlife habitats) and stressors (e.g., land-use change and pollution) using eleven data integration methods (e.g., Highest/Lowest Quintiles, Principle Component Analysis). Because the individual methods varied in their sensitivities and ability to answer different assessment questions (e.g., ranking vs. vulnerability), a suite of integration methods was recommended to identify vulnerable ecosystems. 3.3. Lesson 3: importance of stakeholder engagement While important for any risk assessment, it is likely that more stakeholders will exist for a CRA and that their perspectives will be more diverse because CRA has a population focus and involves multiple stressors and multiple exposures. Communicating with

stakeholders throughout the CRA is essential to identify stakeholder concerns, to obtain information relevant to the assessment and to provide insight into the assessment process, assessment results, decisions and consequences (US EPA, 2003b). Early engagement also allows risk assessors to identify and clarify stakeholder concerns to ensure that the CRA addresses these issues or to communicate that specific concerns cannot be addressed. The EPA’s Superfund program requires specific public involvement, to ensure that community concerns are considered during decision-making. Site teams prepare formal Community Involvement Plans that specify outreach activities, including personal interviews and public meetings. Using this or a similar process for communicating with stakeholders during a CRA could increase the acceptability of the analysis. Stakeholder engagement also can provide critical information for the assessment, including identification of the following: relevant stressors, exposure pathways, potentially vulnerable groups, and other stressor exposures potentially affecting population health. The Waquoit Bay Watershed Assessment identified stressors impacting Waquoit Bay along coastal Massachusetts and presented decision alternatives (sometimes called ‘‘risk management options’’) to improve water and habitat quality (US EPA, 2002f; Serveiss et al., 2004). The collaborative CRA planning process involved varied stakeholders. Initially, EPA held a public meeting of concerned citizens and organizations to receive input on environmental concerns and valued resources; participants identified over 30 concerns. Following this, EPA established an interdisciplinary workgroup of scientists and resource managers. Considering the stakeholder input along with the goals of local, regional, and national resource management organizations and nongovernmental organizations, the workgroup developed the following management goal: ‘‘reestablish and maintain water quality and habitat conditions in Waquoit Bay and associated wetlands, freshwater rivers, and ponds’’ (US EPA, 2002f; Serveiss et al., 2004). Although such early planning meetings can increase the duration of the planning phase and delay the start of the assessment phase, early efforts can build consensus on risk assessment objectives and decision alternatives, reducing delays throughout the remaining CRA process. Stakeholder involvement also facilitates understanding of the goals and limitations of a CRA. In the Baltimore Community Environmental Partnership (CEP) Air Committee Report (Baltimore CEP Study), a CRA was initiated because of community concerns about the health impacts of air pollution (US EPA, 2000a). Stakeholder involvement was initially effective. A committee comprised of local residents, industry managers, university scientists, and officials from the EPA, the state, the city, and the county designed a project to evaluate 175 chemicals emitted into the air from more than 125

701

S.S. Gallagher et al. / Chemosphere 120 (2015) 697–705

facilities. Some community members expected the study to confirm their presupposition that local air quality posed unacceptable health risks (US EPA, 2000a). Ultimately, resource limitations and data gaps precluded the researchers’ ability to link the air quality measures to disease incidence. Because these limitations were ineffectively communicated, some stakeholders were disappointed when the study results could not fully address their concerns, highlighting the importance of properly communicating the CRA scope and limits throughout the assessment process. Stakeholders can help risk assessors understand how cultural practices and activities may affect CRA. For example, the cultural practices of some Native American tribes can lead to exposures that differ markedly from the general population (e.g., diets including significant quantities of locally caught fish). Understanding these practices can increase the accuracy of exposure and risk estimates (Callahan and Sexton, 2007). 3.4. Lesson 4: the value of a tiered approach While ‘‘tiering’’ approaches are useful in many risk assessments, they are particularly valuable in CRAs, as they can focus the CRA’s scope and prioritize stressors that may require further data collection or analysis (Menzie et al., 2007). In CRAs, tiering facilitates the evaluation of multiple stressor exposures, by allowing stressors of limited public health concern to be screened out, focusing the resources available on stressors of most concern. Decisions to include or exclude specific stressors must be presented logically and transparently. A concern for screening approaches is that a stressor, group of stressors, or sources may be inappropriately screened out (i.e., false negative results). When tiered screening approaches are employed, the criteria must be sufficiently

conservative to reduce the likelihood of such an event. To prevent stressors from being screened out inappropriately, the initial tiers should utilize conservative benchmarks that are below the thresholds at which individual health effects might occur (Menzie et al., 2007). This strategy will enable the risk assessor to retain stressors that may have screened out on their own, but may need to be evaluated for potential joint toxicity in a mixture. Additional tiers perform in depth analyses on a small set of stressors. A strength of using tiered approaches is that the level of effort can be tailored to the purpose of the specific tier in the assessment with increasing focus on specific stressors, groups of stressors, or other concerns (e.g., vulnerable population or lifestage) (Meek et al., 2011). The Baltimore CEP Study focused available resources by conducting a three-tiered, risk-based screening approach (Fig. 1) to evaluate the chemicals released to the atmosphere from multiple sources (US EPA, 2000a). At each tier, the cancer risk and noncancer Hazard Quotient (HQ) were estimated for each individual chemical. The accuracy of the exposure estimates increased at each tier. Chemicals with either cancer risk or HQ estimates above screening values were evaluated further in the next tier. Tier 1 used conservative exposure modeling assumptions and identified 17% of the original chemicals for further evaluation. Tier 2 utilized atmospheric dispersion modeling to improve the accuracy of exposure estimates; the results indicated that 24% of the remaining chemicals (4% of the original chemicals) required further analysis. Tier 3 utilized monitoring data from each facility; 2% of the chemicals initially considered were identified as potential chemicals of concern. The Industrial Surface Impoundment Risk Assessment employed a tiered approach to assess whether management of wastewater treatment plant discharges and sludges in surface impoundments

175 Chemicals Released to Outdoor Air

Tier 1: Calculate cancer risk and noncancer Hazard Quotient (HQ) for each individual chemical using conservative exposure modeling.

Is the cancer risk < 10-6 And the HQ1 (for multiple chemicals) were screened out (see Table 9 of Johnson et al., 2003). The EPA conducted more resource intensive, site-specific assessments on those that remained. 3.5. Lesson 5: the need for new methods for CRA Several case studies highlighted the development of methods to improve CRA for multiple exposure routes, pathways, and chemicals. The Organophosphorus Pesticide (OP) CRA was the first assessment conducted by EPA’s Office of Pesticide Programs (OPP) in response to the 1996 Food Quality Protection Act (FQPA) requirement that the EPA consider the human health risks posed by cumulative exposures to pesticides that act through a ‘‘common mechanism of toxicity’’ when establishing allowable pesticide residue levels in foods (US EPA, 2002d). The FQPA did not specifically define the phrase ‘‘common mechanism of toxicity’’. After much debate, including a manuscript published by an expert panel (Mileson et al., 1998), OPP published guidance documents (US EPA, 1999, 2002a) with the following definition: ‘‘Common Mechanism of Toxicity pertains to two or more pesticide chemicals or other substances that cause a common toxic effect(s) by the same, or essentially the same, sequence of major biochemical events (i.e., interpreted as mode of action).’’ This concept of mode of action (MOA), defined as a series of key events leading to a toxic outcome, distinguished this terminology from ‘‘mechanism of action’’, which implies a greater degree of detailed knowledge at the cellular or subcellular level. Subsequently, this definition has been applied generally by EPA to group chemicals for combined analysis in chemical mixture risk assessments (US EPA, 2000b, 2002a). For the hazard assessment, OPs causing a common neurotoxic effect (i.e., cholinesterase inhibition) were identified as members of a Common Assessment Group (CAG). To assess the CAG members as a mixture, OPP used a relative potency factor (RPF) approach, based on dose addition. The RPF approach assumes doses of component chemicals that act in a toxicologically similar manner can be added together after scaling the doses by their potencies, relative to a selected index chemical, to generate an index chemical equivalent dose (ICED). The sum of the component ICEDs produces an ICED for the mixture, which is compared with the index chemical’s dose response information to evaluate potential health hazards from exposure to the mixture. The OP with the most robust toxicity data, methamidophos, was chosen as the P P HI = [HQj] = [Ej/RfDj], where Ej and RfDj are the daily exposure and reference dose of chemical j. 1

index chemical, and an RPF was estimated for each OP and exposure route by comparing its benchmark dose estimate for 10% inhibition of brain cholinesterase (BMD10) with that of methamidophos. The RPFs were used to convert multiple pesticide exposure estimates to ICEDs for comparison with dose response information on methamidophos. EPA developed a probabilistic exposure assessment approach to calculate the cumulative OP exposure distributions from all sources, incorporating behavioral and environmental factors for the relevant exposure routes (US EPA, 2002d, 2006b). Cumulative exposures were estimated for the following pathways: food, drinking water, and residential exposures. The exposure distributions were compared with a point of departure (e.g., BMD10) for methamidophos to estimate the hazard posed by the mixture. Using the results, EPA compared the hazard posed by the different pathways and identified specific crops and groups of OPs posing the greatest hazard. A Feasibility Study of Cumulative Risk Assessment Methods for Drinking Water Disinfection By Product (DBP) Mixtures proposed a novel risk assessment method to evaluate multiroute mixture exposures, the cumulative relative potency factors (CRPF) method (US EPA, 2003a; Teuschler et al., 2004). This case study also used an innovative probabilistic approach to combine DBP water concentrations with human activity and water use patterns, and kinetic modeling to generate multiroute internal doses of DBPs for use in the CRPF calculations. Using the CRPF method (Fig. 2), DBPs potentially causing a common effect were grouped for assessment as a mixture. However, a MOA analysis of the mixture components may indicate that different MOAs may be known or hypothesized that lead to the same adverse outcome. Thus, the components can be assigned to common MOA subgroups and an index chemical with high-quality dose-response data can be selected for each MOA subgroup. The RPF approach is then applied and an ICED is developed for each MOA subgroup, which is used with its index chemical’s dose-response information to estimate the MOA subgroup risk. Because the MOA subgroups are assumed to cause toxicity via different MOAs independently of each other, the estimated MOA subgroup risks can then be summed assuming response addition to estimate risk from exposure to the entire mixture. Thus, the CRPF method allows the estimation of cumulative risks for DBPs that have a common adverse outcome, but act by different MOAs. 3.6. Lesson 6: implications of spatial scale The spatial scale of a CRA affects the methods used, data needs, and types of risk management questions that can be addressed. While data for a single chemical assessment can be directly measured on a site specific basis, CRAs often require additional types of data for stressors and population factors that are not usually collected in the risk assessment process, but are needed to improve estimates of risk. To include these stressors in a non-resource intensive way, they may need to be obtained from existing sources that may have collected the data in different spatial scales than the site data. The National-Scale Air Toxics Assessment (NATA) is an ongoing, nationwide analysis of potential exposures to and health effects from hazardous air pollutants (HAPs) in the US. The 1999 NATA2 used atmospheric emissions data to estimate cancer risks and noncancer hazards associated with inhalation exposures to 177 HAPs (US EPA, 2002b). The frequency, duration and magnitude of exposures to these HAPs were estimated using atmospheric dispersion 2 In 2011 U.S. EPA released the fourth version of NATA based on calendar year 2005 emissions information that were the most complete and up-to-date at the time (http://www.epa.gov/nata2005).

703

S.S. Gallagher et al. / Chemosphere 120 (2015) 697–705

MOA Analysis of Chemical Mixture Group Finds Mixed MOAs for an Adverse Effect

Adverse Effect

Subgroup A Risk of Adverse Effect Using RPFs (RA)

Dose Addion Based on Common MOA Within Subgroup B

Dose Addion Based on Common MOA Within Subgroup A

Division of Group into MOA Subgroups A,B

Adverse Effect

Event: Mechanism of Acon Key Event: Mode of Acon

Subgroup B Risk of Adverse Effect Using RPFs (RB)

Sum Subgroup Risks (RA + RB)

Response Addion Based on Independent MOAs Across Subgroups

Chemical Mixture Group Total Risk of Adverse Effect Using CRPF Approach Fig. 2. Schematic of cumulative relative potency factor (CRPF) approach for chemical mixture group with mixed modes of action (MOAs).

models and human activity pattern data. Cancer risks from exposures to mixtures of HAPs were estimated assuming response addition, and the noncancer hazards for respiratory effects and for neurological effects were estimated using the hazard index method (US EPA, 2000b). The results enabled EPA to identify specific HAPs posing the greatest risk and determine the relative impact of different emission sources affecting health, including locations contributing significantly to population health risks. While this national-scale assessment evaluated contaminants of potential concern across large geographical areas (e.g., county, state), the exposure and risk estimates at finer spatial scales (e.g., neighborhood level) were less certain, potentially requiring different models and emissions data consistent with the spatial scale of interest. The Regional Air Impact Modeling Initiative (RAIMI)—Initial Phase (herein called RAIMI Pilot) was implemented to develop a tool for assessing cumulative health risks from exposures to chemicals emitted by multiple atmospheric emission sources (US EPA, 2003c). The study goal was to develop a standardized approach for assessing and prioritizing inhalation exposures based on air emissions data, using facility-specific data obtained from federal and state regulatory emissions databases, regulatory files, and permit applications. Compared to NATA, the methods developed for the RAIMI Pilot evaluate more refined spatial levels, allowing the estimation of risks at the neighborhood level. This finer resolution enables risk assessors to consider cumulative risks or impacts on local communities when evaluating permitting decisions or facility siting proposals. In addition, RAIMI links estimated risks to specific sources, pathways, and contaminants, allowing assessors to

identify sources potentially causing elevated health risks (i.e., ‘‘hot spots’’). Because of its local spatial scale focus, RAIMI can prioritize interventions reducing community exposures and risks.

4. Discussion Each of these six lessons learned are not necessarily new considerations for risk assessment. Clearly, evaluating multiple stressors and vulnerabilities are complicating factors that set CRA apart from conventional single chemical risk assessment. The following key contributions from the efforts reviewed here are important for the continued development of CRA:  Using an iterative process (e.g., continued reexamination of assessment goals) promotes efficient use of resources and ensures that the CRA results address the primary concerns.  Consideration of population vulnerabilities could improve estimation of population risks and ensures that a selected risk management option is adequately protective.  Early and regular communication with stakeholders enhances process transparency and informs stakeholders of changes in the analysis as well as the logic underlying the selection of a risk management option.  A tiered approach focuses resources on the most important factors contributing to risk. The criteria used in developing tiers need to be balanced, yet sufficiently conservative so that important factors are not inappropriately screened out.

704

S.S. Gallagher et al. / Chemosphere 120 (2015) 697–705

 Continued development of novel risk assessment methods is needed for evaluating multiple stressors encountered in CRAs, including approaches for assessing combined chemical and nonchemical stressors and evaluating dose-response relationships in vulnerable populations.  The spatial scale of a CRA affects data needs, model choice, and the utility of the information for comparing risk management options. Challenges for CRAs include combining data from disparate sources, such as separate national databases, and using analyses of national-scale data to evaluate risks on a community scale. While EPA utilizes aspects of CRA on a limited basis, additional research and case studies could greatly increase the application rates of CRA. CRA could benefit from the development of additional methods for evaluating risks or hazards posed by combinations of chemicals that act via different pathways as well as methods to predict risks from combined exposures to chemical and nonchemical stressors. The ‘‘omics’’ technologies, including genomics, proteomics, and metabolomics, could potentially elucidate the molecular mechanisms of such toxicities. Patterns of changes in the concentrations of related biomolecules can form molecular signatures of toxicity, which can be used to identify and group stressors eliciting the same response and affecting a common health outcome (Pleil et al., 2012). Recently, Thomas et al. used transcriptional changes following chemical exposure to predict pathwaybased points of departure, correlating these pathways with apical endpoints previously identified through ‘‘traditional’’ animal testing (Thomas et al., 2012). One challenge is identifying cellular changes that are indicative of a toxic effect because not all cellular changes lead to an adverse effect. There also may be adaptive cellular responses that will require further elucidation, such as the up-regulation of genes for xenobiotic metabolism and the role of such responses in altering the risks posed by toxic chemicals. Once these technologies have been developed, it will be important to evaluate the new data streams in case studies to understand how they can be used in combination with traditional toxicity information to support the grouping of chemicals for CRA. In the future, they also may be useful in dose-response assessment as suggested by Thomas et al. Research into biomarkers, including xenobiotics, metabolites, or DNA- and protein-adducts may fill some gaps regarding exposure and risk in vulnerable populations. Exposure biomarkers have identified populations experiencing differential chemical exposures (Schulte and Hauser, 2012). Some genetic factors and diseases associated with vulnerable populations also exhibit biomarkers (Ryan et al., 2007). Multiple researchers have used genetic epidemiology to quantify population vulnerability, linking adverse health outcomes to measures of genetic variation, such as single-nucleotide polymorphisms (Cantor et al., 2010; McHale et al., 2010). Another major challenge is that CRA requires integration across a number of different scientific disciplines, each having their own terminology. In some cases, different terms are used for the same concept. One important concept for CRA that is often represented interchangeably by multiple words is vulnerability (e.g., susceptibility and sensitivity). Alternatively, the same word can represent different concepts. At the EPA, there is the added complexity that different programs may use the same term with slightly different meanings depending on the regulatory context. This communication disconnect hinders cross-fertilization between disciplines and EPA programs. Communication issues can also affect the ability of the field to share and integrate data. A similar challenge had to be addressed the way the biomedical fields represented genes. In 1998, the Gene Ontology Consortium was formed to develop a common terminology for annotating genes, gene products, and

sequences. A similar set of standardized rules and requirements for data collection and reporting is essential for CRA; enabling scientists to develop databases, tools, and models capable of integrating data collected from disparate resources. While this article focuses on the work conducted by EPA, important research on the science underlying and application of CRA is being conducted outside of the Agency. Much of this work has focused on new methods for developing quantitative estimates of risk for complex assessments. The European Food Safety Authority (EFSA) assessed the utility of using a ‘‘common adverse outcome’’ approach to accommodate chemicals with dissimilar modes of action (EFSA, 2013). In addition, workshops have been held to evaluate how emerging data streams can be used to characterize human variability to address some aspects of population vulnerability (Zeise et al., 2013) as well as on new approaches for incorporating nonchemical stressors into CRA, including a systems-based approach for predicting the interactions between physical and chemical stressors (Rider et al., 2014). 5. Conclusion Because CRAs evaluate risks posed by exposures to multiple stressors, possibly including different types of stressors, and consider vulnerability, this emerging area of risk analysis is more complex than traditional single chemical environmental risk assessments. Our examination shows that EPA programs and regions are pursuing an increased understanding of CRA through their unique regulatory mandates. The lessons learned from the case studies and issue papers highlight important considerations as EPA and others attempt to make scientific advances in this field. Acknowledgements Sarah Gallagher was supported by a Science and Technology Policy Fellowship from the American Association for the Advancement of Science – United States. References Argonne National Laboratory, 2004. Air Screening Assessment for Cook County, Illinois, and Lake County, Indiana, Chicago, IL. ftp://public-ftp.agl.faa.gov/ OMP%20PFC%2006-19-C–00-ORD/ EIS%20and%20ROD%20Administrative%20Record/Disk04/03_March%202004/ 11_99_22253.pdf. Browner, C., 1995. Memorandum on the EPA Risk Characterization Program and Attached Policy for Risk Characterization, 21 March 1995. U.S. Environmental Protection Agency, Washington, DC. Callahan, M.A., Sexton, K., 2007. If cumulative risk assessment is the answer, what is the question? Environ. Health Perspect. 115, 799–806. Cantor, K.P., Villanueva, C.M., Silverman, D.T., Figueroa, J.D., Real, F.X., Garcia-Closas, M., Malats, N., Chanock, S., Yeager, M., Tardon, A., Garcia-Closas, R., Serra, C., Carrato, A., Castano-Vinyals, G., Samanic, C., Rothman, N., Kogevinas, M., 2010. Polymorphisms in GSTT1, GSTZ1, and CYP2E1, disinfection by-products, and risk of bladder cancer in Spain. Environ. Health Perspect. 118, 1545–1550. DeFur, P.L., Evans, G.W., Cohen Hubal, E.A., Kyle, A.D., Kyle, A.D., Morello-Frosch, R.A, Williams, D.R., 2007. Vulnerability as a function of individual and group resources in cumulative risk assessment. Environ. Health Perspect. 115, 817– 824. Diamond, J.M., Bressler, D.W., Serveiss, V.B., 2002. Assessing relationships between human land uses and the decline of native mussels, fish, and macroinvertebrates in the Clinch and Powell river watershed. USA Environ. Toxicol. Chem. 21, 1147–1155. EFSA (European Food Safety Authority), 2013. Scientific opinion on the relevance of dissimilar mode of action and its appropriate application for cumulative risk assessment of pesticides residues in food. EFSA J. 11, 3472. Johnson, B., Balserak, P., Beaulieu, S., Cuthbertson, B., Stewart, R., Truesdale, R., Whitmore, R., Young, J., 2003. Industrial surface impoundments: environmental settings, release and exposure potential and risk characterization. Sci. Total Environ. 317, 1–22. McHale, C.M., Zhang, L., Hubbard, A.E., Smith, M.T., 2010. Toxicogenomic profiling of chemically exposed humans in risk assessment. Mutat. Res. 705, 172–183. Meek, M.E., Boobis, A.R., Crofton, K.M., Heinemeyer, G., Raaij, M.V., Vickers, C., 2011. Risk assessment of combined exposure to multiple chemicals: a WHO/IPCS framework. Regul. Toxicol. Pharmacol. 60, S1–S14.

S.S. Gallagher et al. / Chemosphere 120 (2015) 697–705 Menzie, C.A., MacDonell, M.M., Mumtaz, M., 2007. A phased approach for assessing combined effects from multiple stressors. Environ. Health Perspect. 115, 807– 816. Mileson, B.E., Chambers, J.E., Chen, W.L., Dettbarn, W., Ehrich, M., Eldefrawi, A.T., Gaylor, D.W., Hamernik, K., Hodgson, E., Karczmar, A.G., Padila, S., Pope, C.N., Richardson, R.J., Saunders, D.R., Sheets, L.P., Sultatos, L.G., Wallace, K.B., 1998. Common mechanism of toxicity: a case study of organophosphorus pesticides. Toxicol. Sci. 41, 8–20. National Research Council, 2009. Science and Decisions: Advancing Risk Assessment. The National Academies Press, Washington, DC. Pleil, J.D., Williams, M.A., Sobus, J.R., 2012. Chemical Safety for Sustainability (CSS): human in vivo biomonitoring data for complementing results from in vitro toxicology-A commentary. Toxicol. Lett. 215, 201–207. Rider, C.V., Boekelheide, K., Catlin, N., Gordon, C.J., Morata, T., Selgrade, M.K., Sexton, K., Simmons, J.E., 2014. Cumulative risk: toxicity and interactions of physical and chemical stressors. Toxicol. Sci. 137, 3–11. Ryan, P.B., Burke, T.A., Cohen Hubal, E.A., Cura, J.J., McKone, T.E., 2007. Using biomarkers to inform cumulative risk assessment. Environ. Health Perspect. 115, 833–840. Schulte, P.A., Hauser, J.E., 2012. The use of biomarkers in occupational health research, practice, and policy. Toxicol. Lett. 213, 91–99. Serveiss, V.B., Bowen, J.L., Dow, D., Valiela, I., 2004. Using ecological risk assessment to identify the major anthropogenic stressor in the Waquoit Bay Watershed, Cape Cod. Massachusetts. Environ. Manage. 33, 730–740. Sexton, K., Hattis, D., 2007. Assessing cumulative health risks from exposure to environmental mixtures-three fundamental questions. Environ. Health Perspect. 115, 825–832. Teuschler, L.K., Rice, G.E., Wilkes, C.R., Lipscomb, J.C., Power, F.W., 2004. A feasibility study of cumulative risk assessment methods for drinking water disinfection by-product mixtures. J. Toxicol. Environ. Health A 67, 755–777. Thomas, R.S., Clewell 3rd, H.J., Allen, B.C., Yang, L., Healy, E., Andersen, M.E., 2012. Integrating pathway-based transcriptomic data into quantitative chemical risk assessment: a five chemical case study. Mutat. Res. 746, 135–143. US EPA, 1999. Proposed Guidance on Cumulative Risk Assessment of Pesticide Chemicals that Have a Common Mechanism of Toxicity. Office of Pesticide Programs, Washington, DC. US EPA, 2000a. Baltimore Community Environmental Partnership Air Committee Technical Report. Office of Pollution Prevention and Toxics, Washington, DC. EPA/744/R-00/005. . US EPA, 2000b. Supplementary Guidance for Conducting Health Risk Assessment of Chemical Mixtures. Risk Assessment Forum, Washington, DC. . US EPA, 2000c. Surface Impoundment Study. Technical Plan for Human Health and Ecological Risk Assessment. Office of Water, Washington, DC. .

705

US EPA, 2001. Industrial Surface Impoundments in the United States. Office of Solid Waste and Emergency Response, Washington, DC. EPA/530/R-1/005. . US EPA, 2002a. Clinch and Powell Valley Watershed Ecological Risk Assessment. National Center for Environmental Assessment, Washington, DC. . US EPA, 2002b. National-Scale Air Toxics Assessment for 1996: Estimated Emissions, Concentrations and Risk. Office of Air Quality Planning and Standards, Washington, DC. . US EPA, 2002c. A Review of the Reference Dose and Reference Concentration Processes. Risk Assessment Forum, Washington, DC. . US EPA, 2002d. Revised Organophosphorus Pesticide Cumulative Risk Assessment. Office of Pesticides Programs, Washington, DC. . US EPA, 2002e. The US EPA’s Regional Vulnerability Assessment Program: A Research Strategy for 2001 2006. Office of Research and Development, Washington, DC. . US EPA, 2002f. Waquoit Bay Watershed Ecological Risk Assessment: The effect of land-derived nitrogen loads on estuarine eutrophication. National Center for Environmental Assessment, Washington, DC. . US EPA, 2003a. The Feasibility of Performing Cumulative Risk Assessments for Mixtures of Disinfection By-products in Drinking Water. National Center for Environmental Assessment, Cincinnati, OH. EPA/600/R-03/051. US EPA, 2003b. Framework for Cumulative Risk Assessment. Risk Assessment Forum, Washington, DC. . US EPA, 2003c. Regional Air Impact Modeling Initiative: Study Pilot—Initial Phase. Office of Solid Waste and Emergency Response, Washington, DC. EPA/906/R-01/ 001. . US EPA, 2003d. Regional Vulnerability Assessment for the Mid-Atlantic Region: Evaluation of Integration Methods and Assessments Results. Office of Research and Development, Washington, D.C. EPA/600/R-03/082. http://www.epa.gov/ reva/docs/vulnerable.pdf>. US EPA, 2006a. National-Scale Air Toxics Assessment for 1999: Estimated Emissions, Concentrations and Risk. Office of Air Quality Planning and Standards, Washington, DC. . US EPA, 2006b. Organophosphorus Cumulative Risk Assessment-2006 Update. Office of Pesticides Programs, Washington, DC. . Zeise, L., Bois, F.Y., Chiu, W.A., Hattis, D., Rusyn, I., Guyton, K.Z., 2013. Addressing human variability in next-generation human health risk assessments of environmental chemicals. Environ. Health Perspect. 121, 23–31.