When only one soil sample is submitted for RBA testing (usually a composite, or a sample from a small decision unit with homogeneous soils), only one RBA value will be available for use in the risk and cleanup goal calculations. Using only one soil sample for RBA testing, however, may not be appropriate, because soils are usually heterogeneous. Uncertainty may also exist about whether one RBA value adequately represents the RBA for an entire exposure area and depth range. In these cases, multiple soil samples should be submitted for RBA testing. When more than one soil sample is submitted for RBA analysis, multiple RBA values (one per sample) will be provided by the laboratory. Laboratory results for in vivo and in vitro methods are discussed in the Methodology section.

With multiple RBA analyses for the same exposure unit, several estimates of central tendency may be representative, including the mean, median, or some UCL on the mean. Currently, there is no clear guidance or precedent for which value to use. In choosing among these values, however, note that the RBA is usually applied to an estimate of the mean soil concentration and integrated into a final exposure estimate that contains the conservative RME exposure assumptions. Additionally, some USEPA guidance recommends using the straight arithmetic mean RBA value rather than a value that accounts for potential uncertainty (for example, the UCL on the mean). In the case of lead, for example, USEPA recommends using the arithmetic mean soil concentration as an exposure concentration in evaluating potential exposure to lead (USEPA 2015a). In doing so, it would be reasonable to also use the straight arithmetic mean lead RBA value in the exposure calculation.

If more conservatism is warranted due to concerns regarding the uncertainty in the RBA, a UCL on the mean RBA could be calculated. For example, a 90% or 95% UCL on the arithmetic mean could be used as the statistical measure for the RBA. In doing so, the most recent version of USEPA’s ProUCL software (USEPA 2016d) could be used to generate the UCL value. Because the UCL on the mean RBA may be applied to the 95% UCL on the mean soil concentration, the use of multiple UCLs should be evaluated for consistency with the overall approach used for adequate site characterization and with USEPA guidance (USEPA 1992b). The RBA estimate chosen should adequately characterize the RBA but, when integrated into the other parameters (for example, UCL on the mean soil concentration), not cause the overall risk estimates to become overly conservative.

When ISM is used for site characterization, the soil samples submitted for RBA testing should also be collected using ISM for consistency. ISM yields an estimate of the average concentration across a decision unit (exposure area), which may include one or more sampling units. It is a site-specific decision whether to collect soil samples for RBA analysis from one incremental sample or from replicate incremental samples within the sampling unit. If the site characterization data are available prior to submittal of samples for RBA testing, the degree of consistency between the triplicate results may be used to make this decision. Submitting IS samples for RBA analysis yields an average RBA across the sampling unit.

As indicated in Footnote 1 of Section 3.3.4 of the ITRC’s ISM guidance (ITRC 2012), ISM may be used for RBA testing. The guidance also notes, however, that “… to be representative of exposure, bioavailability studies must be performed on ISM samples that have not been processed by grinding. If other data quality objectives (DQOs) being fulfilled by ISM samples in the exposure DU require grinding as part of sample processing, the comparability of ground vs. unground samples should be evaluated as part of the study.”

Regardless of the estimate of central tendency, the variability in the individual RBA values should be discussed in the uncertainty analysis section of the HHRA. This discussion could entail a qualitative or quantitative sensitivity analysis (see Bioavailability in Risk Assessment).

Probabilistic risk assessment (PRA) can help to capture uncertainty and variability (USEPA 2014c) in the risk characterization. This approach could be a useful tool when considering RBA. A PRA uses distributions rather than discrete variables for some or each parameter used to calculate risks. In doing so, a PRA provides a distribution or range of risks as its output rather than a single point estimate of risk. In a probabilistic framework, soil RBA would be represented by a probability distribution rather than a discrete value. PRA may be worth considering, however, the increased cost of doing so and likelihood of acceptability by the oversight regulatory agency should be weighed against the potential benefits. As USEPA explains (2014c), a “PRA generally requires more resources than standard Agency default-based deterministic approaches.” As a result, this approach may be costlier to perform in comparison to a deterministic HHRA. The increased costs, however, may be offset by a reduction in the degree of risk management necessary to achieve acceptable risks.

Portions of the site are known to have been used as farmland. The site was developed for use as a commercial center in 1980 and significant filling and grading occurred. Concentrations of arsenic have been detected in soil samples. The higher concentrations of arsenic are colocated with the area that was used for farming in the past. At adjacent sites, however, agriculture may not be the source because the arsenic concentrations do not decrease with depth. The source may be naturally occurring arsenic in the form of a locally prevalent arsenopyrite.

• Background soil concentrations: No background level for arsenic was established.
• Maximum and average concentrations in soils, size of the data set: A single site with 20 samples; total arsenic concentrations ranged from 30 to 980 mg/kg. The 95% UCL on the mean = 278 mg/kg.
• Soil type: Soil characteristics (pH, clay, organic content, dominant minerology) are not known. Geographical history indicates some localized deposits of the mineral arsenopyrite, which is naturally high in arsenic.
• Source of arsenic: Analysis to determine arsenic species indicate that arsenic at the site is likely from historical agricultural land use.
• Present and future land use scenarios: The present and potential land use scenarios are residential or commercial/industrial.

No in vivo bioavailability value was established for the site. An in vitro method was used to establish bioavailability. Arsenic was extracted using a fluid that has properties which resemble gastrointestinal fluid (pH 1.5). This extract represented the fraction of soluble arsenic and is referred to as the in vitro bioaccessibility (IVBA) fraction.

The IVBA was used to predict the RBA using a correlation model by Brattin et al. (2013). Also, see Table 7-2 (in section 7.3.4.1) for other models in the literature for arsenic. This model is based on arsenic RBA measurements from in vivo juvenile swine studies. The RBA values estimated using the (Brattin and Casteel) regression equation (shown below) ranged from 20% to 28%. For samples where arsenic was nondetect (the arsenic concentration in the extracted fluid was below the method detection limit), the equation intercept of 19.7 (or 20%) was used as the default RBA.

The noncancer hazard estimates and cleanup goals for oral exposure to soil arsenic were calculated for a residential child, residential aggregate adult/child, and worker using 95% UCL on the mean and the maximum detected arsenic concentrations. Tables 9-1, 9-2, and 9-3 summarize these calculations.

Table 9‑1. Arsenic: Calculated hazard using 95% UCL on the mean (278 mg/kg)

Table 9‑2. Arsenic: Calculated hazard using maximum detected concentration (980 mg/kg)

Using backward calculation and assuming THQ = 1, the soil cleanup goals for arsenic calculated using different RBA values are shown in Table 9-3. The USEPA RSLs for arsenic are 39 and 650 mg/kg for residential (child) and outdoor worker, respectively.

Table 9‑3. Arsenic: Calculated soil cleanup goal (CGs) for various RBA values (THQ = 1)

The use of site-specific 28% RBA generates more refined risk estimates and cleanup goals for the soil ingestion exposure pathway. The tables above show that if the property is maintained for industrial or nonresidential land use, using the 95% UCL on the mean arsenic concentration with site-specific RBA resulted in an acceptable HQ value for workers compared to the use of the conservative maximum detected concentration of arsenic.

The noncancer hazard estimates and cleanup goals for oral exposure to soil arsenic were calculated using the equations and assumptions shown below.

Residential child noncancer hazard equation:

Residential aggregate noncancer hazard equation:

Where:

Worker noncancer hazard equation:

Where:

Example of residential, noncancer soil cleanup goal equation and assumptions:

Where:

An industrial plant has soil contaminated with PAHs, including BaP, at a limited area. The property was previously used as a manufactured gas plant. Based on initial risk estimates, BaP was the primary risk driver, so an in vivo bioavailability study was conducted for BaP.

• Background soil concentrations: No background level for BaP was established.
• Maximum and average concentrations in soils, size of the data set: The contaminated site has soil BaP concentrations ranging from 1 to 180 mg BaP/kg soil; 95% UCL on the mean = 11 mg/kg and mean = 1.1 mg/kg.
• Soil type: Soil characteristics are not known.
• Source of BaP: The potential sources of the PAHs are the manufactured gas process and waste materials including coal tar.
• Future land use: The potential future land use scenarios are residential or commercial/industrial.

An in vivo method using mice was used to evaluate bioavailability.

The range of measured RBA values (25% – 75%) was used.

BaP is a carcinogen with an MMOA; therefore, the cancer toxicity value (CSF) is adjusted using ADAFs, for residential receptors only.

Tables 9-4 and 9-5 present the calculated ELCR estimates and the soil cleanup goals for various RBA values.

Table 9‑4. BaP: Calculated ELCR using 95% UCL on the mean (11 mg/kg)

Using backwards calculations and assuming a target cancer risk (TCR) = 1×10^-6, the soil cleanup goals for BaP using the range of RBA are shown below.

Table 9‑5. BaP: Soil cleanup goals (CGs) for various RBA values (assumes TR = 1×10^-6)

The use of site-specific RBA refines the HHRA by lowering the uncertainty contributed by possible influence of the soil matrix on BaP bioavailability. As demonstrated above, the site-specific RBA decreases the ELCR estimate and increases the cleanup goal values. If the regulatory target ELCR is 1×10^-6, the use of site-specific bioavailability (and the 95% UCL on the mean concentration) does not change the outcome of the HHRA due to the high soil concentrations at the site. Where the regulatory target ELCR is 1×10^-5, however, site-specific bioavailability may affect the risk management decision for the site (using the 95% UCL on the mean concentration and a commercial/industrial land use scenario). Because of the mutagenic mode of action and high CSF of BaP, a bioavailability assessment may be useful if the soil concentrations are slightly above the cleanup goal or when the RBA value is expected to be less than 50%.

The ELCR estimates and cleanup goals for oral exposure to soil BaP were calculated using the equations and assumptions shown below.

Residential Child and Adult Aggregate Risk for a Carcinogen with an MMOA:

Where:

Worker Risk for a Carcinogen (MMOA adjustment does not apply to adults):

Where:

Residential Soil Cleanup Goal for a Carcinogen with MMOA adjustment:

Where:

Worker Soil Cleanup Goal for a Carcinogen (MMOA adjustment does not apply to adults):

Where:

The site is currently a residential property. Fill material containing waste residues from a smelter were imported to the property and used to fill low-lying areas. Elevated concentrations of lead were detected in surface soil from zero to one foot below ground surface in the filled areas.

• Background soil concentrations: No background level for lead was identified.
• Maximum and average concentrations in soils, size of the data set: A single site with 10 soil samples analyzed for lead bioavailability. Each soil sample was collected in the field as a five-point composite. Each sample was sieved by the laboratory and the material less than 150 microns was used for IVBA testing. The measured concentrations of lead in site soil ranged from 637 mg/kg to 677 mg/kg (average = 649 mg/kg).
• Soil type: Soil characteristics are not known.
• Source of lead: Fill material on site contains metal slag from a smelter.
• Future land use: Residential or commercial use is anticipated.

Bioavailability was evaluated using the IVBA method presented in USEPA guidance (USEPA 2007b).

The IVBA values reported by the laboratory ranged from 81% to 89%, with an average value of 85%.

Step 1—Estimation of In Vivo Relative Bioavailability

The following empirical linear regression model was previously established by USEPA (2007b) between RBAs derived using in vitro and in vivo (juvenile swine) procedures:

Using this equation, the in vivo RBA of lead in site soil was estimated based on the site-specific IVBA values measured in soil samples collected from the site. A range of site-specific RBAs corresponding to the minimum, average, and maximum was calculated. The sample-specific RBAs range from 68% to 75%, with an average RBA of 72%.

Step 2—Absolute Bioavailability

An estimate of absolute bioavailability is needed for the IEUBK model. Based on available information in literature on lead absorption in humans, USEPA estimates that the absolute bioavailability of lead from water and the diet is usually about 50% in children (USEPA 2010d). When a reliable site-specific RBA value for soil is available, it may be used to estimate a site-specific absolute bioavailability in that soil, as follows:

Where:
ABA_soil = ABA of lead in site soil ingested by a child
RBA_soil = site-specific RBA of lead

The calculated ABA_soil for the site ranges from 34% to 38%, with an average ABA_soil of 36%, which is higher than USEPA’s default ABA of 30% used in the IEUBK model.

The percentage of the exposed population with a predicted blood lead level (BLL) exceeding 5 µg/dL from oral exposure to lead in soil was calculated. The target BLL used for this example was 5 µg/dL; however, other states and agencies may use a different target BLL. The site-specific RBA value was used in the lead models in place of the default RBA values as follows:

• IEUBK Model: The default soil and dust GI bioavailability values were modified in the IEUBK model. The average estimated site-specific ABA_soil value (36%) was used in place of the default value (30%) for soil and dust. The remaining default input values were left unchanged in the IEUBK model.
• ALM: The AF_S,D input parameter value was modified in the ALM. The AF_S,D is the product of 0.20 (the absorption factor for soluble lead) and the RBA. The average estimated site-specific RBA (0.72) was used in place of the default value (0.6). The remaining default input values were left unchanged in the ALM.

For comparison purposes, the models were also run using USEPA’s default ABA_soil value (in the IEUBK Model) or RBA (in the ALM). The input assumptions used in the IEUBK Model and ALM for this lead example are presented in Table 9-6.

Table 9‑6. Lead: Input Assumptions Used in the IEUBK Model and ALM

The predicted results for a resident child and commercial worker using the average soil lead concentration and average site-specific RBA (in addition to the default RBA) are presented in Table 9-7.

Table 9‑7. Lead: Probability that calculated BLL > 5 µg/dL for soil concentration of 649 mg/kg

Assuming the target is 5% probability that BLL exceeds 5 µg/dL, the soil cleanup goals for lead were calculated using the IEUBK Model and ALM based on the default and average site-specific RBAs, as shown in Table 9-8.

Table 9‑8. Soil cleanup goal (CG) for two RBA values (assumes target is 5% probability that BLL > 5 µg/dL)

The site-specific RBA data for lead in soil provide more refined site-specific estimates of BLLs for potential site receptors. The site-specific RBA data, however, did not change the conclusions of the risk evaluation: the soil lead concentrations result in BLLs that are above USEPA-acceptable levels for a residential scenario and BLLs within USEPA-acceptable levels for a commercial scenario. The average site-specific RBA is higher than the default RBA, resulting in lower site-specific soil cleanup goals than the cleanup goals calculated using the default RBA.