7
Synthesis of Risk Considerations for Unencapsulated Electric Arc Furnace Slag Uses

The U.S. Environmental Protection Agency (EPA) considers risk to be the chance of harmful effects to human health or to ecological systems resulting from exposure to an environmental stressor.¹ This chapter synthesizes some of the key observations of this report concerning risks to human health associated with the use of unencapsulated electric arc furnace (EAF) slag, primarily for residential applications. As part of this synthesis, the chapter also explores previous risk assessments of EAF slag and provides an analysis of slag composition data for the purpose of ranking components by degree of hazard. In addition, the chapter identifies factors that may lead to the highest risks. While screening-level analyses of specific residential use scenarios of unencapsulated EAF slag indicated an exceedance of established risk thresholds and analyses of other scenarios reported risks below those thresholds, the data available to the committee were not sufficient to develop a general conclusion about human health risks associated with unencapsulated EAF slag use in the United States. The chapter identifies research needed to address key information gaps and reduce uncertainty in risk assessments associated with human exposure to EAF slag.

PREVIOUS RISK ASSESSMENTS ON ELECTRIC ARC FURNACE SLAG

To inform its deliberations, the committee examined five previous health risk assessments of EAF slag that included residential applications: Proctor et al. (2002), Exponent (2007), ToxStrategies (2011), Streiffer and Thiboldeaux (2015), and Mittal et al. (submitted).² The Streiffer and Thiboldeaux assessment was conducted by the State of Wisconsin Department of Health Services and focused on risks associated with exposure to slag from a specific EAF steel plant. The other four assessments comprise a series of initiatives funded either by the Steel Slag Coalition (a group of 63 companies that produce steel and/or process slag) or the National Slag Association (comprised of more than 80 member companies associated with the production and processing of iron and steel slag products) that were intended to cover multiple slag sources. The 2007, 2011, and submitted assessment are each intended to update earlier assessments. The comments presented subsequently summarize the observations provided in Appendix F stemming from the committee’s review. The committee did not provide a separate, detailed review of each assessment.

In addition, EPA Region 8 conducted a screening-level evaluation in 2020 to assess risks associated with residential applications of EAF slag (Folland and Simmons, 2020). EPA observed that EAF slag had been applied at approximately 100 properties within the Colorado Smelter Superfund study area. The assessment was conducted using intentionally conservative exposure assumptions in order to identify dominant risk drivers and eliminate relatively minor sources of risk from subsequent focused risk evaluation. The assessment concluded that risks are elevated above standard risk thresholds by several orders of magnitude based on concentrations of manganese (Mn) and chromium (Cr) measured in EAF slag samples collected at 21 properties in a community that has potential environmental justice concerns. The

___________________

¹ See https://www.epa.gov/risk/about-risk-assessment.

² While the committee was conducting its study, Deborah Proctor (ToxStrategies) and colleagues were carrying out a probabilistic risk assessment of residential EAF slag applications for covering driveways and unpaved roads. Preliminary materials provided to the committee regarding that assessment comprise the following: Proctor and Antonijevic (2022); Proctor (2022); Crystal Ball input data submitted by Deborah Proctor on February 21, 2023; and Manuscript submitted to Risk Analysis for peer review entitled “Probabilistic Risk Assessment of Residential Exposure to Electric Arc Furnace (EAF) Steel Slag Using Bayesian Model of Relative Bioavailability and PBPK Modeling of Manganese.”

screening-level assessment examined a range of exposure pathways and receptor scenarios and identified key sources of uncertainty that could be explored in a baseline risk assessment.

Overall, the committee found that the five risk assessments vary in terms of the specific aspects considered. Regarding slag sources and chemicals of potential concern (COPCs), the ranges of mean and maximum values (milligrams per kilogram; mg/kg) for the most commonly presented slag constituents in the risk assessments are shown in Table 7-1. Note that these values were reported for different kinds of sampling locations (e.g., slag pots versus open slag piles [ready for sale]), and some samples were from slag that had been separated by particle size. It was not clear how well the data accurately reflect the variability in slag composition across the population of EAF plants in the United States, as well as the variability from heat to heat within an EAF plant.

The receptor populations evaluated included the current and hypothetical future child and adult residents. However, a full range of susceptible groups that may be exposed and health outcomes likely associated with each hazardous constituent in slag had not been fully identified or characterized. Variations in toxicity assessment included choice of chronic oral RfD and use of toxicokinetic models to estimate internal dose of slag chemicals.

The committee identified concepts and approaches applied in some of the risk assessments that represent rather narrowly defined conditions, which would likely not be reasonable to extrapolate to residential populations throughout the United States. For example, one industry-funded risk assessment considered exposure scenarios in which slag covers only the driveway of a residence rather than a larger portion of the property. However, documentation of EPA’s slag sampling initiative at 21 residences in Pueblo, Colorado, indicates slag is used to cover a substantial portion of a property’s yard (PWT, 2020). Overall, the committee questioned the approaches to exposure assessment (e.g., routes and pathways of exposure considered and assumed values of exposure factors), the COPC concentrations associated with EAF slag, the effect of environmental conditions on fate and transport of weathered slag, the relevance of assumed activity patterns over time, potential confounding factors for each of the health outcomes of interest, and the sources of uncertainty and variability that may bias the assessment results. Because of the range of conditions represented by the risk assessments, the risk conclusions cannot be directly compared. Furthermore, the applicability of the overall conclusions of the assessments on a national scale is unclear without additional research to address key data gaps.

As discussed later in this chapter, the committee evaluated the chemical profile presented in Table 7-1 along with two additional data sets of EAF slag profiles to assess the variability of the set of COPCs in EAF slag from different locations and production processes. Also, the committee identified key assumptions and uncertainties that might distinguish an assessment for unencapsulated EAF slag from assessments of more conventional sources of contaminants in outdoor soil and indoor dust, as discussed subsequently.

TABLE 7-1 Examples of Ranges of Concentrations of Inorganic COPCs in EAF Slag as Reported in Five Risk Assessments

HAZARD RANKING OF CONSTITUENTS IN ELECTRIC ARC FURNACE SLAG

To identify COPCs of slag that warrant further consideration as potential contributors to human health risk, the committee applied a hazard ranking approach in which example slag composition profiles were compared with EPA’s regional screening levels (RSLs) for residential exposures to soil for a range of inorganic chemicals.³ The current RSLs (as of May 2023) provide risk-based screening levels calculated using the currently recommended toxicity values, default exposure assumptions, and physical and chemical properties. RSLs are discrete values set at a level below which exposures are unlikely to present an unacceptable human health risk.

It is important to note that the results are intended to convey a relative risk ranking rather than absolute risk estimate. Therefore, an exceedance of an RSL at this step does not necessarily mean that slag use will result in unacceptable levels of exposure and risk. Also, the soil RSLs for residents may introduce uncertainty when used to screen slag COPCs. It may be prudent to adjust one or more exposure factors to derive slag-related RSLs that are more applicable to slag exposure scenarios. Table 7-2 identifies sources of uncertainty, along with potential consequences in terms of bias introduced in a screening-level assessment by applying soil RSLs. When these sources of uncertainty collectively bias the application of the current EPA soil RSLs to be less stringent, the approach may not identify all COPCs in a hazard identification step of an EAF slag risk assessment. For example, ingestion of home-produced foods (e.g., fruits and vegetables, chicken meat and eggs) was not included in the hazard ranking. Key sources of uncertainty include the matrix-to-plant uptake factors and intake slope factors (e.g., change in concentration in meat or eggs per unit of average daily intake) for chicken meat and eggs. Although application of exposure factors from published studies with outdoor soil may under- or overestimate exposures associated with EAF slag for ingestion of home-produced foods, the net effect of not including that exposure pathway would likely bias the RSL toward being less stringent.

TABLE 7-2 Sources of Uncertainty in Applying EPA Soil RSLs to Identify and Rank Order COPCs in EAF Slag

Comparing Slag Composition Data Sets

Figure 7-1 illustrates the range of concentrations of EAF slag COPCs from the compositional profiles presented in Table 7-1 along with the EAF slag profiles from Piatak et al. (2021) (see Tables 2-4a,

___________________

³ See https://www.epa.gov/risk/regional-screening-levels-rsls.

2-4b, and 3-1) and analytical data of aged EAF slag obtained from residential yards within the Colorado Smelter Superfund study area in Pueblo, Colorado in 2019.⁴

Figure 7-1 presents the data on a log scale, which may exaggerate the correspondence in overall COPC profiles based on the average and range of concentrations reported by each data set. Also, without additional details regarding the sample locations represented in the risk assessment and Piatak et al. (2021) data sets, the committee assumed that some correspondence would be expected if there were at least some overlapping sources of data. Compared to the Piatak et al. (2021) data set, the average concentrations in samples from Pueblo were generally lower by a factor 2–10 for all inorganics except arsenic (6× higher), zinc (3×), and cobalt (1.2×). Compared to the data from the five risk assessments, the concentrations in samples from Pueblo were generally lower with the exception of mercury (10×), lead (8×), thallium (7×), zinc (7×), arsenic (6×), cadmium (4.3×), and cobalt (1.6×). Because concentrations of organics were not reported in most of the example chemical profiles available to the committee, they are not included in the hazard ranking evaluation.

The Pueblo data set is particularly useful for examining the extent to which the concentrations of a particular metal were correlated with the concentrations of other metals across samples collected from 21 residential properties. Given that the reported estimated age of the slag ranged from 1 to 15 years, the relatively high bivariate Spearman rank correlations indicate that patterns in co-occurring inorganics may persist over time. As indicated in Table 7-3, there is a strong correlation (Spearman rho ≥ 0.9) between Mn, Cr (total), and vanadium, among other constituents of potential toxicological concern. The correlation suggests that, at least for the weathered and sieved EAF slag material present in residential yards in Pueblo, it is highly likely that higher concentrations of several COPCs occur together in EAF slag and that exposures to elevated levels of metals could occur as co-exposures. This is particularly relevant for screening-level assessments when selecting a target hazard index (HI), which is the ratio of the potential exposure concentration of a substance to the level of that substance at which no adverse effects are expected. EPA provides risk assessors with the option to screen using HI =1 or HI = 0.1 (to account for potential mixtures). For EAF slag material, it appears that HI = 0.1 is warranted given the potentially high correlation among COPCs.⁵

The strong correlations between elements shown by multiple analyses of Pueblo slags are most easily understood if the minor and trace quantities of the correlated COPCs are in some mineralogical carrier present in variable abundance among slag samples. For instance, the correlations between Cr, Mn, V, and Al might be understood if spinel carries the signal. However, it is very difficult to understand why Ca and Ba should also be correlated with Cr, Mn, and V, as Ca and Ba are not present at any significant level in spinels. The spinel structure has only sites as large as octahedral on which Cr, Mn, V, and Fe are accommodated. To accommodate large cations like Ca and Ba, more capacious cubic sites are required by radius ratio considerations. However, some slags contain Ca-Al ferrite (brownmillerite), which has Ca as an essential constituent as well as cubic or more highly coordinated sites to accommodate Ba as well. Mombelli et al. (2016, 2014) report that brownmillerite does concentrate V and Ba, making plausible the conjecture that brownmillerite may be the carrier responsible for the correlation between these normally uncorrelated elements. Indeed, the normally unexpected correlation of Ca and Ba with the Cr-Mn-V suite may be a signal of the presence of brownmillerite. It follows that less basic, more siliceous EAF slags that support melilite instead of brownmillerite may not show the covariation of Ca and Ba with Cr, V, and Mn. Trace and minor element correlations are reflections of the major element carrier phase stabilities that may

___________________

⁴ C. Hodes, EPA, personal communication, February 25, 2022.

⁵ See https://www.epa.gov/risk/regional-screening-levels-rsls. Note that the term “hazard quotient” (HQ) is used to refer to a single constituent in which average daily dose is summed across pathways. Some risk assessors elect to switch to the term “hazard index” (HI) when any summation is performed (e.g., dose from ingestion and dermal contact for a single COPC). In its Microsoft Excel® tables that summarize RSLs released in May 2023, EPA uses the term “Target Hazard Quotient (THQ)” when referring to RSLs associated with one exposure pathway, and “Target Hazard Index (THI)” when referring to RSLs that integrate multiple exposure pathways for a single chemical.

change with quite small changes in bulk composition, as may be seen on the equilibrium liquidus diagrams in Appendix E.

The Piatak et al. (2021) data set is discussed in Chapter 3. The sample size varies by analyte, ranging from 7 samples for arsenic and cadmium to more than 30 samples for iron, aluminum, and Mn. The “risk assessment” category is based on Kaplan Meier arithmetic mean concentrations from 22 EAF slag samples reported by Proctor (2022).

Slag elements reported on the basis of compounds (e.g., Mn as MnO) were converted to elemental concentrations using the mass fraction of elements in the compound.

For the conversion of weight percent to concentration (mg/kg), 100% = 10⁶ mg per kg; but since results are reported on scale of 0 to 100, multiplier is conc (mg/kg) = Wt (%) × 10⁴.

The committee applied the following processing steps to the Pueblo data:

• We excluded three field duplicates (denoted by the suffixes “-02A” and “-02B” in the Sample IDs). While averaging of primary and field duplicate results reduces measurement error for results associated with the specific sample locations, this practice may introduce bias in estimates of variance (EPA, 2009).

• We used the subset of data described as sieved (< 250 µm) (denoted by the last letter “B” in the Sample ID).
• We excluded data qualified as “R” (rejected) and included nondetects by applying Kaplan Meier parameter estimation methods rather than substituting proxy values for the reported analytical detection limits.

Two sets of results were reported for hexavalent chromium (Cr⁶⁺), one using analytical method 7199 and a second using “SOP BAL-4300.” The latter method consistently yields results that are an order of magnitude higher for the same sample. Results presented in this chapter are based on total Cr, unless otherwise specified.

The final data set consists of results for 21 samples of EAF slag, with the exception of antimony (n = 18) and thallium (n = 19). The error bars represent the minimum and maximum concentrations.

TABLE 7-3 Correlation Matrix (Spearman rho) for Concentrations of Selected Metals in Slag (< 250 µmsieved fraction) from Different Residential Properties in Pueblo, Colorado

NOTE: The slag was collected from residential areas: n = 21; sieved faction < 250 µm.
The estimated ages of the sampled slag ranged from 1 to 15 years. Cr is total Cr.

The Pueblo data set also provides insights into the potential enrichment of some COPCs that may occur among the smaller particles that are more likely to adhere to skin and become a source of exposure through hand-to-mouth activity. Each slag sample was sieved, and data were reported for a particle size fraction < 250 µm and a particle size fraction >250 µm. Three COPCs exhibit moderate enrichment in the smaller size fraction (based on the ratio of paired concentrations [arithmetic mean ± standard deviation]): lead (ratio = 3.4 ± 1.3), zinc (ratio = 1.8 ± 1.0), and beryllium (ratio = 1.4 ± 0.3). For the remaining inorganics, including barium, Cr, Mn, and vanadium, there is no evidence of enrichment in the smaller size fractions, and, in fact, the mean Mn concentration in the <250 µm fraction is approximately 60 percent lower (ratio = 0.4 ± 0.2) than the mean in the >250 µm fraction. Generalizations to all EAF slag should be applied with care. These results reflect slag from one EAF steel plant and one environmental setting. As noted in Chapters 2 and 3, there are many different factors that can contribute to variability in constituents and concentrations present, the physical nature of the material, and the environmental and use conditions; each factor can play a role in determining if enrichment occurs in smaller particle size fractions.

Hazard-Ranking Results

A common source of screening levels applied to soil data is the EPA RSL.⁶ For chemicals that have both cancer and noncancer endpoints, the lower of the two RSLs is used. The same approach is used in this chapter (noting that for metals, cancer is the more critical effect for both Cr⁶⁺ and arsenic). Table 7-4 uses the exposure variables identified in Table 4-1 in Chapter 4 to generate adjusted RSLs that may apply to slag, and it presents a range of plausible alternative values that may apply based on a review of available risk assessments, EPA guidance, and professional judgment. Table 7-4 is not intended to be comprehensive

___________________

⁶ See https://www.epa.gov/risk/regional-screening-levels-rsls.

in the sense that it applies to all COPCs and slag materials; it is an approximation that informs the potential range of RSLs that may apply to slag in general. The overall shift from the default soil RSL to an adjusted slag RSL is likely to be modest, within a factor of 5 for the ingestion pathway and a factor of 3 for the inhalation pathway. The particulate emission factor is a potentially large source of uncertainty in estimates of the inhalation exposure concentration.

Using the average concentration profiles of slag COPCs from the risk assessments reviewed by the committee (see Figure 7-1, green bars), and increasing the EPA soil RSL by a factor of 4, by way of example to reflect a possible shift in exposure variables discussed previously, COPCs can be rank ordered based on the ratio of the COPC concentration divided by the adjusted screening level (herein referred to as exposure ratio). Figure 7-2 shows the results using this approach. The COPCs with the highest hazard ranking (exposure ratio > 1) are Mn, iron, Cr⁶⁺, vanadium, thallium, and antimony. Arsenic, cadmium, nickel, and cobalt have ratios between 0.1 and 1. Barium (total), beryllium, lead, silver, zinc, and mercury rank lowest with ratios of less than 0.1.

TABLE 7-4 Key Variables in Assessing Possible Exposures to EAF Slag Chemicals in Soil with Plausible Alternative Values

NOTE: IR = average daily soil ingestion rate; EV = event frequency of contact; AF = skin surface adherence factor; PEF = particulate emission factor; RBA = relative bioavailability; ABS = absorption factor adjustment for dermal; BAF = soil-to-plant bioaccumulation factor; RME = reasonable maximum exposure.

^a RME point estimate for young children (<6 years), assuming critical effect is noncancer.

^b Factor range is the ratio of the RME used for the current RSL divided by the plausible range of alternative exposure factors for IR, EV, AF, RBA, ABS, and the inverse for PEF.

^c Composite soil and dust ingestion rate = (IR outdoor soil × fraction of ingestion that occurs outdoors × fraction of outdoor time in contact with slag) + (IR indoor dust × mass fraction of outdoor slag in indoor dust) = (200 × 0.45 × 0.6) + (200 × 0.55) = 54 + 77 = 131 mg/day. Fraction of ingestion of outdoor soil and indoor dust is based on EPA’s assumptions applied for lead in the IEUBK Model. Fraction of outdoor time in contact with slag is professional judgment, shown here by way of example.

^d Ranges for PEF associated with (1) 1.36E+09 is commercial/industrial scenario assuming 0.5 acre source, 50 percent vegetative cover, 4.69 m/s mean annual wind speed; (2) 4.4E+08 is construction/off-site resident scenario over 30 years (EPA, 2002).

^e Ratio rounded to one significant digit.

RISK FACTORS AND DATA NEEDS

EAF slag is produced and available for use in a large portion of the United States. The committee identified 117 operating EAF steel plants and 91 slag processing facilities in 33 states. Many of those states allow unencapsulated EAF slag to be readily sold and distributed in commerce. The total amount of slag that could be potentially produced at full steel production capacity would be between 8 and 12 million tons per year. Actual annual slag production will follow the steel production quantities in a given year. Unencapsulated EAF slag is used mainly as an aggregate substitute for such applications as construction project entrances, residential driveways, road bases, parking lots, culverts, railroad beds, landscaping, unpaved trails, septic fields, soil remediation, agricultural soil conditioning, and remediation of acid mine drainage.

The chemical composition of slag varies widely according to the grade of steel produced, source of the scrap used as a feedstock, and EAF operational practices. The majority of the slag constituents are formed from the oxidation and addition of CaO and MgO₂ during melting to form slag rich in FeO, MnO, lime, silica, and alumina. Stainless steel slags contain significant quantities of chromium oxide due to oxidation of the high Cr content of stainless steels, in addition to FeO, MnO, lime, silica, and alumina.

A number of chemical elements can occur within the slag mineral matrix. According to a ranking approach applied by the committee, the slag components with potentially the highest hazard ranking (HI > 1) are Mn, Fe, Cr⁶⁺, V, Tl, and Sb. Components with HI values between 0.1 and 1 are Al, As, Cd, Ni, Co, and Cu. Components with HI values less than 0.1 are Ba, Be, Pb, Ag, Zn, and Hg. Standardized laboratory testing generally has found that minimal leaching of hazardous elements from EAF slag usually occurs in

concentrations that are not detectable or below regulatory limits. In general, studies to date have not examined the long-term fate of ferrous slag materials and trace element release under variable environmental conditions.

The detection of several persistent organic pollutants (POPs) in EAF slag (albeit at relatively low concentrations compared to EPA’s soil screening levels used for site remediation planning) indicates a need for further study. In addition, because the potential for slag contamination by POPs depends upon the composition of the steel scrap fed into an EAF, it underscores the need for EAF plants to meet the existing regulatory requirements under the U.S. Clean Air Act to prepare and implement a pollution prevention plan for metallic scrap selection and inspection to minimize the amount of halogenated plastics, lead, and free organic liquids that is charged to the furnace.

Potential exposure pathways and routes include incidental ingestion of slag particles mixed in outdoor soil or indoor dust, inhalation, and dermal contact. In addition, ingestion may occur from consumption of home-produced foods grown on properties with EAF slag. While an exposure assessment of EAF slag is likely to share many of the same approaches as those for contaminants in soil and dust, there may be some notable differences that play a role in the relative contributions of exposure pathways to estimates of long-term average daily dose. Also, when mixed with outdoor soil and indoor dust, EAF slag may alter the overall bioaccessibility and bioavailability of inorganic COPCs.

The committee focused on Cr and Mn as examples of COPCs that would be included in a more comprehensive, site-specific evaluation, which would include a broader list of COPCs. Due to data limitations in the scientific literature, the committee was unable to examine quantitative relationships between human health outcomes directly from the levels of Cr and Mn exposures expected from unencapsulated slag used for residential applications. Therefore, the committee relied on studies of health effects from exposures to Cr and Mn in general.

In previous assessments considered by the committee, screening-level analyses of specific residential use scenarios of unencapsulated EAF slag indicated an exceedance of established risk thresholds, and analyses of other scenarios reported risks below those thresholds. However, due to uncertainties in the current evidence stream, the committee was unable to make an overall characterization of risk related to unencapsulated EAF slag use in the United States.

Therefore, until more studies have characterized a wider range of weathered EAF slag materials and environmental conditions, the committee cautions against making generalizations from conclusions from published risk assessments.

Based upon its review of the available evidence and its judgment, the committee identified factors considered to have the potential to contribute to the highest risks from the use of unencapsulated EAF slag. Those factors also comprise key data needs. See Box 7-1. A greater understanding of these factors will help ensure that calculated slag-related risks are not overestimated or underestimated. The relative importance of these factors is expected to differ on a case-by-case basis.

PRIORITY RESEARCH NEEDS

EPA should coordinate with state agencies, the National Slag Association, and other organizations to address these specific research needs: