Probabilistic risk assessment of residential exposure to electric arc furnace steel slag using Bayesian model of relative bioavailability and PBPK modeling of manganese

1 INTRODUCTION

Iron and steel slags are generated as coproducts of steel production and are classified according to the type of furnace in which the slag is generated (O'Connor et al., 2021). Electric arc furnace (EAF) slag is generated during carbon steel production in an EAF. EAF slag has many applications primarily related to construction, including road base, riprap, landscape aggregate, soil stabilization, and cover of unpaved rural roads (Al-Amoudi et al., 2017; Nguyen et al., 2022; US Geological Survey [USGS], 2021), for the treatment of phosphates and metals in wastewater (Ahmad et al., 2020; Elez et al., 2008; Lee et al., 2021), as well as new uses in cement production (Loureiro et al., 2022). The production of EAF slag in the US is determined by EAF carbon steel production and has been relatively consistent at approximately 5–9 million metric tons annually since 2016, of which approximately 85% is used for road base and land cover (National Academies of Sciences, Engineering & Medicine [NASEM], 2023; O'Connor et al., 2021; USGS, 2021, 2023).

Environmental applications of steel slag result in increased potential for human contact and the need for assessing potential health risks (O'Connor et al., 2021). Residential exposure has been of recent concern, prompting a National Academies of Science, Engineering and Medicine (NASEM) review of unencapsulated uses of steel slag (NASEM, 2023). The NASEM study identified uncertainties in the risk assessment of EAF steel slag and a number of research recommendations, some of which are addressed here to the extent of available data.

Many metals in EAF slag occur at concentrations that are higher than those typically found in US soils; however, the mobility of metals in slag is limited because the metals are bound to the slag mineral matrix, which is alkaline (O'Connor et al., 2021; Singh et al., 2021). Water leachate data have been used to calculate the water-solid partitioning coefficient (K_d) for metals in slag (Proctor et al., 2000). K_d values exceed 1000 L per kilogram (L/kg) for all metals in slag that occur at concentrations exceeding background levels of metals in soil, indicating low potential for leaching (Proctor et al., 2000). For most metals, K_d values increase with pH; thus, it is not surprising that metals in highly alkaline slag are not readily solubilized. Consistently, under the strongly acidic conditions of the US Environmental Protection Agency's (USEPA) Toxicity Characteristic Leaching Procedure (TCLP; US Environmental Protection Agency [USEPA], 1991), metals are not leached from iron and steel slag at levels that exceed TCLP standards (Garrabrants et al., 2022; Proctor et al., 2000; Singh et al., 2021). Preliminary findings from recent Leaching Environmental Assessment Framework extraction tests for EAF slag further support this finding (Garrabrants et al., 2022), and site-specific evaluations focused on leaching potential have concluded that leaching is unlikely to affect drinking water sources due to low solubility of metals in EAF slag and considering natural dilution and attenuation (Loncnar et al., 2022; Singh et al., 2021).

The current risk assessment updates previous evaluations (Proctor et al., 2000; Proctor et al., 2002; ToxStrategies, 2011). Similar to the approaches used here, these earlier risk assessments evaluated the potential for human health risk using probabilistic risk assessment (PRA) methods and measures of oral bioaccessibility, which is the percent of an element in a solid matrix that is soluble in in vitro simulated gastric conditions and available for absorption upon ingestion (USEPA, 2007), and are specific to EAF slag applications as a product, but not specific to any location. The Wisconsin Department of Health Services (Streiffer & Thiboldeaux, 2015) also conducted a screening level risk assessment for incidental soil ingestion and inhalation of EAF slag used at residential locations and on farm roads, assessing six child exposure scenarios in Wisconsin. The department concluded that manganese (Mn), iron (Fe), antimony (Sb), vanadium (V), and thallium (Tl) had hazard quotients exceeding unity, suggesting the potential for a health hazard. All these metals are evaluated here and in earlier assessments.

The metal considered to pose the most significant potential concern for human health in these assessments, as well as in the NASEM review (NASEM, 2023), is Mn which exists in EAF slag at approximately 3%. However, it has been shown that Mn is more concentrated in commercially available processed slag, which is typically 0–1 in. in diameter, as compared to small slag particles, those sieved to <300 or <75 μm (Figure S1). Importantly, Mn is an essential element and both the USEPA oral reference dose (RfD) and the Agency for Toxic Substances and Disease Registry (ATSDR) Tolerable Intake for Mn are based on the upper-bound of Mn in the human diet, rather than a measure of toxicity. This introduces uncertainty into the traditional risk assessment approach as most RfDs are based on an adverse effect level, or no effect level, divided by an appropriate uncertainty factor. This uncertainty was noted in the NASEM review, calling on USEPA to update its Mn toxicity criteria (NASEM, 2023).

The current risk assessment uses current and scientifically advanced approaches with newly collected EAF slag characterization data, including oral bioaccessibility, particle size, and moisture content data, to assess the potential for residential exposures to EAF slag to pose a human health hazard. Specifically, we use PRA methods with EAF slag-specific exposure parameters to the extent available. To account for the limited solubility and expected limited bioavailability of most metals in EAF slag, new measures of oral bioaccessibility and relative bioavailability (RBA), which is defined as the percent of an element absorbed from EAF slag relative to the percent absorbed of the same element in the studies used to set toxicity criteria (USEPA, 2007), were developed for this assessment. RBA estimates are only available for arsenic and Mn, and for the latter, RBA data are based on a recent in vivo rat model specific to EAF slag (Proctor et al., 2023). A novel Bayesian model to describe variability for this parameter was used in the PRA. Finally, for Mn, a previously published physiologically based pharmacokinetic (PBPK) model was used to assess the potential for Mn accumulation in the target tissue of the brain and cause neurotoxicity (Campbell et al., 2023).

2 MATERIALS AND METHODS

2.5 Probabilistic risk assessment

PRAs allow for the characterization of variability and uncertainty in exposure estimates and increase the scientific rigor of risk assessment as compared to determinist methods (Burmaster & Anderson, 1994). It is more meaningful than screening level approaches, such as RBCs, that rely on the upper bound of all exposure variables multiplied together. Variability and uncertainty in the risk assessment models and exposure parameters are quantified using probability distributions, but toxicity criteria are not typically represented as a probability distribution in USEPA regulatory risk assessments (USEPA, 2001, 2014) and were assessed as point estimates here. The distributions mathematically represent the range of possible values for a parameter value and the probability associated with each value. In a PRA, distributions of exposure parameters are multiplied together generating probabilistically based risks and hazards. Thus, PRAs avoid the overestimation that results from multiplying upper bound estimates together, and the upper percentiles of the resulting risk and hazard distributions are more representative of the true upper bound of potential exposure.

Probabilistic methods were used to quantify the distribution of inhalation and oral exposures and used to calculate increased cancer risk and noncancer hazard from residential exposure to specific environmental slag applications. The following exposure scenarios were evaluated:

• Residential exposure (including children and adults) to slag cover for an unpaved driveway or other exposed surfaces at a residential property (residential driveway scenario);
• Residential exposure (including children and adults) to dust generated by vehicle traffic on an unpaved roadway with slag cover adjacent to the residence (residential roadside scenario).

Figure 1 depicts the conceptual site model for the PRA. The current evaluation focuses on inhalation and oral exposures. Dermal exposures were not quantitatively evaluated because metals in EAF slag are not expected to be soluble on the skin surface and available for absorption; further, the dermal pathway for contact with slag was found to be insignificant in a previous risk assessment (Proctor et al., 2002). Similarly, Streiffer and Thiboldeaux (2015) did not include exposure by dermal contact in their risk assessment.

For the residential driveway scenario, exposures via inhalation and incidental ingestion were estimated. For the resident living next to unpaved roads with slag cover (residential roadside scenario), only inhalation exposure was included, as it was assumed that incidental ingestion was not relevant for this scenario (Figure 1). For both exposure scenarios, henceforth referred to as the “driveway” and “roadside” scenarios, exposure estimates were made for children and adults at the 50th and 95th percentiles of exposure. The equations used to estimate cancer risk and noncancer hazards for the driveway and roadside scenarios are included in Supplemental Material B.

The PRA modeling was conducted in Oracle Crystal Ball, version 11.1.3. Monte Carlo sampling was utilized, and 200,000 trials were run to ensure numeric stability. In conducting the PRA, distributions of exposure parameters (e.g., metals concentration in slag, soil ingestion rate, and exposure duration) and modeling parameters (e.g., particulate emission factors [PEF]) were developed from slag-specific data or published sources. Some were taken from the published literature and USEPA guidance, whereas others were empirically determined. All distribution assumptions and their bases are shown in Table 1. Select distributions are described in greater detail herein.

TABLE 1. Exposure distributions used in the probabilistic risk assessment (PRA).

Variable	Applicable scenario	Description	Distribution definitions	Units	Basis
BWadult	Driveway, roadside	Body weight, adult	Custom continuous ranges (BW from NHANES 2011 18+ years)	kg	NHANES 2011-12 (CDC,2015)
BW)child	Driveway, roadside	Body weight, child	Custom continuous ranges (BW from NHANES 2011 1–6 years)	kg	NHANES 2011-12 (CDC,2015)
Cslag	Driveway, roadside	Concentration of COIs in slag	Empirical distribution	mg/kg	EAF slag characterization program; 38 samples
ED	Driveway, roadside	Exposure duration for child and adult. For child the maximum duration is 6 years	Custom continuous ranges (residential tenure data)	Year	USEPA Exposure Factors Handbook (EFH) Ch 16. (2011) Table 16-108 Residency Occupancy Period (both sexes)
ETdust, ETslag	Driveway	Exposure time for slag ingestion (through outdoor slag or indoor dust) at residence	Triangular (min—0, likely—0.56, max—1)	Unitless	EFH Table 16-1: the adjustment factor of 0.56 selected as the most likely value for child and adults was based on the ratio between time spent by 3 to <6-year-old playing in sand/gravel (60 min) to time spent outdoors (107 min); selected factor based on child since higher exposure
ETinh	Driveway, roadside	Exposure time for slag inhalation at residence	Lognormal (mean—0.09, 95th—0.3, truncate max = 1)	Unitless	Adjustment factors for child and adults based on mean time spent outdoors; selected factor based on adult since higher mean exposure time. EFH Table 16-22 Mean time spent outdoors at residence for 16 to 64 (mean 136.4 min, 95% = 435 min)
IRadult	Driveway	Slag ingestion rate for adult	Lognormal (mean—30, 95th—100)	mg/d	EFH Ch 5. (2017) Table 5-1 Soil and dust ingestion rate for 12 years through adult
IRchild	Driveway	Slag ingestion rate for child	Lognormal (mean—80, 95th—200)	mg/d	EFH Ch 5. (2017) Table 5-1 Soil and dust ingestion rate for 1 to <6-year-old
PEF	Driveway	Wind driven particulate emission factor, resident, driveway	Custom continuous ranges with PEFs ranging from 2.7E08 to 3.7E12	m3/kg	PEF values from USEPA Soil Screening Guidance 1996 for a 0.5-acre source with 50% vegetative cover
	Roadside	Unpaved roads particulate emission factor, using Fresno Met data to represent arid conditions	Triangular (min—3.8E7, likely—1.1E8, max—2.3E08)	m3/kg	AERMOD-determined using Met data for Fresno and emission factors from AP-42 Ch 13.2.2. PEF based on maximum annual average of 5 years of Met data for distances up to 50 m from the road
RBA	Driveway	Antimony bioavailability via oral exposure	Uniform (min—67.00, max—75.00)	%	EAF slag characterization program bioaccessibility data
		Arsenic bioavailability via oral exposure	Uniform (min—35.39, max—54.35)	%	EAF slag characterization program bioaccessibility data, converted to RBA using USEPA (2017) equation
		Hexavalent Chromium bioavailability via oral exposure	Triangular (min—4.80, max—31.00, likely—16.70)	%	EAF slag characterization program bioaccessibility data for total chromium
		Iron bioavailability via oral exposure	Triangular (min—8.40, max—27.00, likely—16.48)	%	EAF slag characterization program bioaccessibility data
		Manganese bioavailability via oral exposure	RBA data for the lung Lognormal (logmean—−2.085, LogSD—1.198, upper limit—63)	%	In vivo bioavailability Bayesian linear regression analysis for Mn
		Vanadium bioavailability via oral exposure	Triangular (min—61.00, max—79.00, avg—73.40)	%	EAF slag characterization program bioaccessibility data

• Abbreviations: BW, body weight; d, day; d/yr, days per year; EAF, electric arc furnace; ED, exposure duration; EF, exposure frequency; ET, exposure time; m³/d, cubic meter per day; m³/kg, cubic meter per kilogram; mg/d, milligram per day; PEF, particulate emission factor; RBA, relative bioavailability; SD, standard deviation.

2.5.1 Particulate emission factors

PEFs relate the concentration of a metal in slag to the concentration of particulates arising from various emission and transport processes. USEPA developed standard PEFs in its Soil Screening Guidance (USEPA, 1994, 1996). For the driveway scenario, the USEPA residential PEF, based on a 0.5-acre source, with 50% vegetative cover, was used to model fugitive dust generation due to wind erosion. To define a distribution of PEFs, values calculated by USEPA for locations across the US were used (USEPA, 1996).

For the roadside scenario, dispersion modeling was employed to determine the distribution of PEF values. Unit-emission air concentrations were modeled with four orientations of the receptor to roadway (north–south, east–west, northeast–southwest, northwest–southeast) at various indicated distances of a receptor (e.g., a home) from the roadway (10, 20, 30, 40, and 50 m). Five-year meteorological data from Fresno, CA (i.e., 2004–2008) were employed in the AERMOD dispersion model to represent the potential for exposure in arid conditions. PEFs were derived using the inverse of the estimated 5-year maximum air concentrations from area-normalized, unit air emissions across a roadway segment, that were scaled using slag-specific PM10 emissions from unpaved road traffic based on USEPA AP-42 for an unpaved road assuming dimensions of 6 m by 1045 m (USEPA, 2006).

For the residential roadside scenario, vehicle traffic data for unpaved roads were needed. Individual states’ Departments of Transportation (DOT) were contacted to obtain these data. Although state DOTs from across the United States were contacted, very few had data regarding this parameter, with the most robust data sets coming from the states of New York, Virginia, and Montana. We used the median traffic count of 60 vehicles per day (VPD) from the Virginia and New York data sets, and 162 VPD based on the median from Montana. The unpaved road emission factor incorporated EAF slag-specific average silt content and average moisture content data, 4.5% and 2.8%, respectively, collected as a part of the EAF slag characterization project. A triangular distribution of the roadside PEF was constructed using the minimum, average, and maximum PEFs across the two VPDs. The PEF distribution parameters are presented in Table 1.

2.5.2 Incidental slag ingestion rate

Because data on slag-specific incidental ingestion rates are not available, USEPA-recommended adult and child soil ingestion rates were used in the PRA, represented by lognormal distributions defined using the mean and 95th percentile values. Additionally, an adjustment factor to the soil ingestion rate was applied to better characterize the potential for contact with slag at a residence, relative to soil, because EAF slag is a hard, jagged cementitious aggregate, which is less likely to be contacted than soil. For this distribution, a mean ratio of 0.56 was developed from data in the USEPA Exposure Factors Handbook (USEPA, 2011: Chapter 16. Table 16-1). Specifically, the factor was based on the ratio between the assumed time in contact with slag (i.e., children ages 3 to less than 6 years old play in sand or gravel on average for 60 min per day) to the time spent outdoors (an average of 107 min; Table 1). This adjustment factor, referred to as exposure time (ET slag) in Table 1, was defined as a triangular distribution with a minimum value of 0 (no contact), a most likely value of 0.56, and a maximum value of 1 (contact equal to that for soil). The 95th percentile of this distribution was 0.85, and the distribution was also applied as the fraction of soil, or slag, ingestion as indoor dust (ET dust) in the PRA because contacted soil and slag are more likely to be tracked indoors.

3 RESULTS

3.5 Probabilistic risk assessment

The PRA-generated, probabilistically based exposure estimates and values at the 50th and 95th percentiles of the exposure distribution were chosen to represent typical and upper-bound exposure conditions, respectively, consistent with USEPA (2001) guidance.

3.5.1 Residential driveway scenario risk and hazards

Increased cancer risk and noncancer hazard associated with slag exposure in the driveway scenario to the various COIs for adult and child residents are presented in Table 5. For the driveway scenario, the 50th and 95th percentile estimates of increased cancer risk from exposure to arsenic in slag were 2E − 07 and 1E − 06, respectively, and for exposure to CrVI in slag were 2E − 08 and 7E − 07, respectively. The complete probability distributions for increased cancer risk for arsenic and CrVI are presented in Figure 2. Noncancer hazard quotients (HQs) for all COIs in the driveway scenario at the 50th and 95th percentiles ranged from 6E − 04 to 2 for the child resident, and from 4E − 05 to 0.2 for the adult resident. The maximum estimated hazard for the driveway scenario, with an HQ of 2, was estimated for child exposures to Mn in slag at the 95th percentile using the more conservative RfD of 0.024 mg/kg-day. Using the RfD without the threefold modifying factor of 0.071 mg/kg-day, the HQ was 0.7, and this RfD is considered more appropriate for exposures to Mn in EAF slag. The probability distributions for HQ for all COIs are presented in Figure 3. Additionally, the total hazard index (HI) was evaluated by aggregating HQs across target organs (i.e., cardiovascular, developmental, and digestive). All HI by target organ were at or less than 1 for the residential driveway scenario, except for the 95th percentile HI of 2 that was estimated for the nervous system based on oral exposure to Mn, as discussed above (Table S3).

TABLE 5. Probabilistic risk assessment (PRA) results from residential driveway scenario for children and adults exposed by incidental ingestion and inhalation of windblown particles.

	Cancer risk		Hazard index—child		Hazard index—adult
Constituent of interest	50th percentile	95th percentile	50th percentile	95th percentile	50th percentile	95th percentile
Antimony	-	-	3E − 02	2E − 01	1E − 03	2E − 02
Arsenic	2E − 07	1E − 06	5E − 03	3E − 02	3E − 04	3E − 03
Hexavalent chromium	2E − 08	7E − 07	6E − 04	2E − 02	4E − 05	2E − 03
Iron	-	-	7E − 02	3E − 01	4E − 03	3E − 02
Manganese (RfD = 0.024 mg/kg-day)	-	-	3E − 01	2E + 00	2E − 02	2E − 01
Manganese (RfD = 0.071 mg/kg-day)	-	-	1E − 01	7E − 01	8E − 03	7E − 02
Vanadium	-	-	1E − 01	7E − 01	8E − 03	6E − 02

3.5.2 Residential roadside scenario risk and hazards

PRA results for the residential roadside scenario are presented in Table 6. This scenario, which only assessed the inhalation of slag particulates resulting from transport from unpaved roads, resulted in increased cancer risk and noncancer hazard approximately two orders of magnitude less than the driveway scenario for most COIs. The 95th percentile cancer risk estimates for arsenic and CrVI were 6E − 09 and 9E − 08, respectively. Noncancer HQs for all COIs at the 50th and 95th percentiles were identical for child and adults, as hazards are based on RfCs and not adjusted for body weight or inhalation rate, consistent with current USEPA (2009) guidance for inhalation risk assessment, and ranged from 1E − 05 to 0.3. All HIs by target organ were less than 1 for the residential roadside scenario (Table S3).

TABLE 6. Probabilistic risk assessment (PRA) results from residential roadside scenario for children and adults exposed by inhalation of particles from traffic on unpaved road (arid conditions).¹

	Cancer risk		Hazard index—child		Hazard index—adult
Constituent of interest	50th percentile	95th percentile	50th percentile	95th percentile	50th percentile	95th percentile
Antimony	–	–	1E − 05	2E − 04	1E − 05	2E − 04
Arsenic	4E − 10	6E − 09	5E − 05	5E − 04	5E − 05	5E − 04
Hexavalent Chromium	2E − 09	9E − 08	8E − 05	3E − 03	8E − 05	3E − 03
Manganese	–	–	5E − 02	3E − 01	5E − 02	3E − 01
Vanadium	–	–	3E − 03	2E − 02	3E − 03	2E − 02

• ¹ No inhalation toxicity criteria exist for iron.

3.5.3 PRA sensitivity

Sensitivity data were extracted for the child driveway evaluation, specifically for noncancer hazard from exposures to Mn in slag. Figure 4 shows the contribution to variance observed across all distributions used in the PRA, with at least a 1% contribution to variance. The variance observed in this particular HQ is influenced the most by the Mn RBA distribution (46%), followed by the ingestion rate distribution (32%) and the slag ingestion adjustment ET factor distribution (14%).

3.6 PBPK modeling

PBPK model simulation results showed that for both the driveway and the roadside exposure scenarios at both the 50th or 95th percentiles of exposure, estimated peak levels of Mn in the globus pallidus, whole blood, and liver were relatively consistent with the levels estimated for background exposure (Table 7). Exposure by ingestion for the driveway scenario resulted in higher estimations of Mn in tissues than for the roadside scenario (Table 7). The predicted tissue concentrations specifically in the globus pallidus across the entire course of the simulation for the 95th percentile of exposure for both scenarios indicate that Mn did not significantly accumulate relative to the background scenario (Figure 5). Further, both the exposure scenarios and the background scenario globus pallidus levels show the same trend with concentrations peaking at ∼3.5 years of age, followed by a sharp decrease until ∼13 years, ending with a slower and steady decrease to a low plateau. Similar trends between background and exposure scenarios were also observed in liver and whole blood (Figure S2). Maximum predicted concentrations in the globus pallidus for the driveway exposure scenario for the 95th percentile exposed child were 0.616 μg/g of globus pallidus tissue and 0.578 μg/g in the roadside scenario, only slightly above and essentially equal to, respectively, the background scenario prediction of 0.575 μg/g (Table 7). Importantly, these maximum predicted concentrations in the globus pallidus fall below the range of tissue-based no-observed-adverse-effect levels (NOAELs) of 0.7 and 0.9 μg/g developed by Schroeter et al. (2012) and Gentry et al. (2017).

TABLE 7. Predicted manganese concentrations.

	Peak Manganese concentration (μg/g)
Exposure percentile	Globus pallidus		Whole blood		Liver
Driveway exposure scenario (oral + inhalation)
	Background	Exposure	Background	Exposure	Background	Exposure
50%	0.575	0.579	0.00932	0.00940	2.66	2.68
95%		0.616		0.01012		2.81

Roadside exposure scenario (inhalation only)
50%	0.575	0.575	0.00932	0.00932	2.66	2.66
95%		0.578		0.00938		2.67

4 DISCUSSION

EAF slag is commonly used as fill material and for rural road and driveway cover in residential areas even though the levels of several metals in EAF slag exceed residential screening levels for soil (Table 3). Recognizing that EAF slag is a cementitious aggregate, and that screening level risk assessments for soil are not specifically applicable to EAF slag, we sought to conduct a refined risk assessment for residential exposures to metals in EAF slag specifically. Approaches used to refine the risk assessment included the use of probabilistic methods, incorporation of bioaccessibility and RBA data, and PBPK modeling of Mn to assess the potential for Mn accumulation in the target tissue of the brain. For general consistency with USEPA's guidance and application of PRA methods, point estimate toxicity criteria were used in this risk assessment to focus on variability in exposure (USEPA, 2001, 2014). USEPA has determined that an HI or HQ exceeding 1 suggests that a hazard may exist, and that further evaluation is warranted (USEPA, 1989). USEPA considers total cancer risks in the range of 1 × 10⁻⁶ to 1 × 10⁻⁴ to be acceptable (USEPA, 1989, 1990). The HQs and cancer risk calculated for residential exposure to metals in EAF slag are within and below these acceptable risk targets, with the exception of an HQ for Mn of 2 calculated for the driveway scenario using the most conservative USEPA Mn RfD. Using the Mn RfD without the threefold modifying factor results in HQs <1.

CrVI and arsenic were the two carcinogenic metals measured in EAF slag at concentrations exceeding residential soil RBCs. Although the greater risk was estimated to be associated with arsenic (Tables 5 and 6), it is worth noting that arsenic concentrations in EAF slag do not exceed levels in natural soil (Dragun & Chekiri, 2005), and that the risk estimate, even at the upper bound, was within the acceptable risk range.

CrVI was measured more frequently, and at higher concentrations, in samples from this characterization program compared to previous studies (Proctor et al., 2000; ToxStrategies, 2011). Our expectation is that this observation is due to analytical and sample preparation variability, rather than a change in the CrVI content of EAF slag. Because EAF slag is a highly alkaline matrix enriched with total chromium (Table 3), the potential for chromium oxidation during sample preparation and analysis exists. CrVI concentrations in this study could also be overestimated due to crushing the samples prior to analysis. The NASEM (2023) report called for additional CrVI characterization in EAF slag, in particular regarding the effects, if any, of environmental weathering. In this study, newly produced slag was evaluated, including slag that had been stockpiled for up to 1 year, and the mean concentration was 9.3 mg/kg (Table 3), which is approximately equal to that (10 mg/kg) reported for EAF slag samples collected from residential properties in Pueblo, Colorado, that were sieved to <250 μm and had been in place for up to 15 years (NASEM, 2023).

The NASEM Committee reviewing the potential risks posed by unencapsulated uses of EAF slag identified CrVI bioaccessibility and bioavailability as data gaps for risk assessment. In this study, total chromium bioaccessibility data were used for CrVI dose calculations because past experience has demonstrated that CrVI is not detectable in the USEPA Method 1430 extraction test for solids containing low levels of CrVI, such as EAF slag (Table 3). This decision may over or underestimate CrVI bioaccessibility and RBA at low levels, because CrVI is more soluble than CrIII. However, because CrVI is reduced to CrIII in acidic conditions, including the stomach (Kirman et al., 2017), incidental ingestions of the relatively low levels of CrVI in EAF slag are not likely to overwhelm reduction in the stomach or pose a cancer hazard. As stated by the NASEM Committee, “Unless the reductive capacity is overwhelmed, there is minimal risk of hexavalent chromium-prompted disease burden” (NASEM, 2023).

The risk estimates for CrVI developed herein are conservatively based on the current USEPA draft IRIS toxicity criteria (USEPA, 2022), which assume a linear dose-response extrapolation from cancers observed in the animals exposed by drinking water to high concentrations of CrVI over the course of a lifetime and include ADAFs assuming a mutagenic mode of action. The CrVI dose by ingestion at the 95th percentile from the driveway scenario is calculated as 2.1E − 6 mg/kg-day, and if bioaccessibility is assumed to be 100%, the 95th percentile dose is 1.3E − 5 mg/kg-day. By comparison, the human-equivalent lower confidence limit on the benchmark dose for tumors in mice, which is the basis for the USEPA draft oral slope factor, is 0.31 mg/kg-day. Thus, at the 95th percentile dose, the margin of exposure (MOE) is 147,000, and assuming 100% bioaccessibility in the dose calculation, the MOE is 24,000. Given these large MOEs, the potential cancer risk associated with CrVI exposure is considered negligible despite the uncertainty regarding CrVI bioaccessibility in EAF slag.

Manganese has been identified in this and previous risk assessments for EAF slag as potentially causing a hazard because the levels of Mn in EAF slag exceed USEPA residential soil RSLs by more than 10-fold (Table 3). Assessing risk from Mn oral exposure is challenging because the available toxicity criteria (USEPA RfD and ATSDR MRLs) are based on dietary intake. As Mn is an essential nutrient, uptake and elimination of Mn are controlled through homeostasis (Aschner & Aschner, 2005). When uptake exceeds excretion, Mn may increase in the brain resulting in neurotoxicity, particularly affecting motor function. However, due to the lack of a clear NOAEL and lowest observed-adverse-effect level relationship, and because the application of traditional uncertainty factors may not be appropriate given homeostatic control of Mn (Gentry et al., 2017; Perry et al., 2023), a distribution of Mn RfDs could not be developed with confidence at this time. As an alternative, we assessed point estimate RfDs with and without the USEPA threefold modifying factor. We also used the ATSDR MRL as a point estimate for inhalation exposure, although equivalent values in published studies are 10- to 30-fold higher (Bailey et al., 2009; Gentry et al., 2017).

To further address potential uncertainty and use advanced PBPK modeling approaches for Mn risk assessment, the most current version of the published human models (Campbell et al., 2023) was used to evaluate both EAF slag exposure scenarios. The PBPK modeling results (Table 7) indicate that the Mn exposure scenarios evaluated here are not expected to significantly increase Mn in tissues for either EAF slag exposure scenario as compared to background Mn from normal dietary, drinking water, and ambient air exposures.

Gentry et al. (2017) and Schroeter et al. (2012) evaluated the dose response for neurological effects associated with Mn exposure using the model-predicted Mn concentrations in the globus pallidus for humans and nonhuman primates (NHPs), respectively. In Schroeter et al. (2012), the dose response was evaluated in 14 NHP studies involving exposure by different routes of administration, and a categorical regression model (EPA CatReg software) with severity scores for several neurological observations by dose. Clinical signs of neurotoxicity were scored 0 (no response), or 1, 2, or 3 (mild-to-severe), and summed across six categories of observations, with total scores ranging from 0 to 9 in the 14 studies that provided measures of neurotoxicity (Schroeter et al., 2012). These authors reported a steep dose-response curve starting at peak Mn concentrations of 7 μg/g in the globus pallidus for neurological effects. The CatReg-estimated extra risk dose for 10% response (ERD10) for a severity score of 1 was a peak globus pallidus concentration of 0.8 μg/g. The ERD10 for a severity score of 1 (0.8 μg/g) is considered the tissue-based NOAEL.

Gentry et al. (2017) conducted a similar study using the Schroeter et al. (2011) Mn PBPK model and measures of neurological effects among workers exposed primarily by inhalation. Gentry et al. (2017) used blood Mn data as a biomarker and predicted the target-tissue NOAEL from the available human studies. A no-statistical-significance-of-trend (NOSTASOT) test was used with Fisher's exact test to predict target-tissue Mn concentrations at the NOAELs for abnormal eye-hand coordination of 0.7 and 0.9 μg/g for the endpoint of abnormal hand steadiness. Importantly, Schroeter et al. (2012) and Gentry et al. (2017) provided highly consistent tissue NOAELs as thresholds for adverse effects for Mn in the globus pallidus, ranging from 0.7 to 0.9 μg/g.

Additionally, Ramoju et al. (2017) used PBPK modeling to predict a 10% extra risk concentration (ERC10) of 0.55 μg/g globus pallidus, which was modeled for adult workers, with background levels of Mn in the globus pallidus set at 0.4 μg/g. These authors calculated an occupational exposure associated with the ERC10 of 0.55 μg/g of 70 μg/m³ for a 5-year average exposure. The ERC10 from Ramoju et al. (2017) was not used as a point of comparison because it is based on occupational exposures among adults, and background Mn in the globus pallidus was set below model-predicted background for young children. However, it is worth noting that the ERC10 is 0.15 μg/g higher than background, which is an increase far higher than that predicted among children (0.041 μg/g) in our model for EAF slag exposures (Table 7). The tissue-based NOAELs of 0.7–0.9 μg/g from Gentry et al. (2017) and Schroeter et al. (2012) exceed the model predicted Mn doses from the EAF slag residential exposures scenarios, the highest of which is 0.616 μg/g for exposures among young children (Table 7).

Regarding Mn in EAF slag, it is worth noting that an in vivo rat RBA study (Proctor et al., 2023) was conducted to support and refine this risk assessment. Three tissues were studied: the lung, representative of systemic absorption and upon which the RBA used herein was derived; the liver, representative of uptake and excretion under hepatobiliary controls; and the striatum, which is the rat brain tissue representative of the globus pallidus in humans and NHPs. In this study, at doses of Mn as high as 39 mg/kg-day for 14 days from EAF slag ingestion, no increase in the absorption of Mn in the striatum was observed. Further, in the lung tissue, the absorption of Mn from chow showed a positive DT-C, whereas that from EAF slag ingestion was slightly negative. This is thought to be due to competition from iron, which occurs at levels six times higher in EAF slag than Mn. Both Mn and iron compete for the same divalent metal transport absorption sites, which limits the systemic uptake of Mn from EAF slag (Proctor et al., 2023). Thus, iron is thought to play a protective role, limiting the absorption of Mn upon ingestion of EAF slag. The RBA for Mn for this PRA was restricted to DT-C values which were positive to ensure conservatism in this assessment, but negative RBA values are possible based on the in vivo data and known competition between iron and Mn for absorption sites.

For the current study, we developed a new distribution of RBA values using Bayesian modeling. The 50th and 95th percentiles of the distribution were 16% and 52%, respectively, and the 50th percentile is similar to that reported in Proctor et al. (2023) for the lung of 14%. The 95th percentile was similar to the high measure of bioaccessibility for Mn measured in the current study (62%; Table 4). Importantly, in the interpretation of the results of the Bayesian modeling and the RBA, the numbers of samples that predicted the slag ingestion dose group D-TC slope was negative were much greater than those that predicted a positive slope (28% positive slope prediction vs. 72%), although those are the only samples that were used. So, although it is apparent that the RBA values derived from this modeling are indeed reflective of the relationship between the two Mn administration groups, the number of samples with a negative EAF slag group D-TC slope indicates that its slope is more than likely negative, as was observed in vivo (Proctor et al., 2023). Taken together, the Bayesian model demonstrates an ability to fit these relationships and derive an RBA distribution within a defined range (0–0.63), while still providing reliable estimations and an uncertainty distribution that can be used to better characterize the RBA than traditional methods have done. Importantly, this RBA is for Mn in EAF slag specifically, whereas the RBA of Mn from other sources may be higher. Among individuals with anemia or limited vitamin intake, the RBA may be higher although the very high iron content of EAF slag (18%; Table 3) likely compensates, such that the RBA of Mn from EAF slag in systemic and target tissues is arguably zero, if not negative.

The NASEM Committee commented on the lack of information available to assess the impacts of cumulative exposure to EAF slag in disadvantaged communities (NASEM, 2023). Although limited information currently exists to assess this, recent USEPA documents (USEPA, 2023c, 2023d) provide helpful information. First, EAF slag is typically used in areas in relatively close proximity to EAF steel mills, due to the high density of EAF slag, which is a disadvantage for long-distance transportation. USEPA (2023c) evaluated demographics by census blocks within 5 and 50 km of 88 carbon and specialty EAF steel mills in the United States considering race and ethnicity, educational attainment, poverty level, age, and linguistic isolation. The study found that most demographic factors were largely consistent with that of US averages, although the residential population within 5 km of an EAF mill was 17% African American as compared to 12% in the overall US population, and 14% Hispanic or Latino compared with 19% overall. The percentage of people below the poverty line within 5 km of the mills was 17% as compared to 13% in the United States. These metrics were consistent, that is, within a few percentage points, with the US population within 50 km of the mills, which is an area where EAF slag is most likely to be used.

Finally, the USEPA (2023d) reported research on the cumulative effects of mild repeated stress and drinking water exposure to Mn in rats from gestational day 13 to postnatal day 9, an important window in neurodevelopment corresponding to the second and third trimester in human pregnancy. Serum corticosterone was measured at several timepoints, with and without induction of a complex series of induced stress measures, and two Mn drinking water concentrations, the higher of which resulted in decreased water consumption and weight gain. Mn in drinking water increased Mn in blood and brain tissue, but there was no effect of stress on the Mn concentrations. Mn exposure affected behavioral function in offspring, but no effect of stress alone or in combination with Mn was observed, and the combination of stress seemed to attenuate the effects of Mn (USEPA, 2023d). Although consideration of site-specific factors at some EAF slag application locations may be warranted, the low RBA of Mn in EAF slag, homeostatic control of Mn accumulation in target tissues, and currently available data support that Mn in EAF slag should not contribute significantly to a cumulative hazard, including at locations where nonchemical stressors are elevated compared to the general population.

5 CONCLUSION

The findings of this PRA indicate that residential exposures to EAF slag are not likely to pose a significant cancer risk or noncancer hazard. Additionally, PBPK modeling of Mn showed that at expected (50th percentile) and upper bound (95th percentile) Mn doses from residential exposure scenarios, systemic absorption is controlled by homeostasis, and that the slag-specific exposure scenarios result in Mn tissue levels consistent with background and below published tissue no-effect levels for Mn in the brain. Future research efforts may consider an advanced study using two-dimensional PRA once greater confidence in the Mn toxicity criteria has been established using PBPK modeling.