Term Project: Risk Assessment
of the American Creosote Site

Table of Contents:

1. Introduction
2. Site Background
3. Data Management
4. Data Reduction
5. Exposure Point Concentrations
   • Organics
   • Metals
6. Exposure Assessment
   • Incidental Soil Ingestion
   • Dermal Contact
   • Particulate Inhalation
7. Toxicity Assessment
8. Risk Characterization
9. Conclusions
10. References

Introduction

Site Background

The site is currently abandoned, but is bordered by residential properties and an apartment complex. Area residents have reported the presence of black, weathered, creosote-like material in their yards and nearby drainage ditches. The most recent site investigation focused on characterizing the extent and magnitude of offsite contamination due to stormwater flow and historical chemical spills. Numerous soil and drainage ditch sediment samples were collected down to a depth of 2 feet from residences and drainage pathways adjacent to the former facility. The samples consisted of approximately 400 samples from suspected areas of contamination (characterization samples) and approximately 70 samples from areas believed to be unaffected by historical site operations (background samples). The primary contaminants of concern were found to be polycyclic aromatic hydrocarbons (PAHs), PCP, copper, chromium, arsenic, polychlorinated dibenzodioxins, and polychlorinated dibenzofurans. The goal of this report is to perform a human health risk assessment of potential receptors in the offsite residential areas bordering the site.
 

Data Management

The next step was to set up a series of database queries to retrieve only the needed information about the samples. Based on the results of the site investigation, an intial list of chemicals of concern (COCs) was created. COCs are chemicals identified at the site which may be present at concentrations of potential human health concern. The intial COC list consisted of PAHs, PCP, and metals, and is shown as follows:

Queries for each of the 41 initial COCs were subsequently set up. The requested information included the sample identification number, station type (background or characterization), analysis, concentration, sample qualifier (detected or undetected), and sample matrix (soil). A query for the background and characterization samples for each COC was constructed, resulting in a total of 82 tables. A typical query window looked like this:

Data Reduction

• Frequency of Detection. If a constituent was detected in fewer than 5% of the characterization samples, the constituent was eliminated from further consideration.

• Comparison to Background. A student's t-test was performed on the data of each constituent to determine if the mean concentration of the characterization samples was different from that of the background samples. If the two data sets are described by the variables x and y, which have n and m data values and standard deviations of S_x and S_y, then the t-statistic can be calculated as follows:
  
  For calculation of the mean background and characterization sample concentrations, all data with no qualifiers or with "J" qualifiers (estimated concentration) were assumed to be positive identifications. If a constituent was reported as a non-detect ("U" qualifier), it was assumed to be present at one-half of the sample quantitation limit. If |t|<2 for a given constituent, the means of the two data sets are not considered different (the two data sets cannot be statistically distinguished from one another) and the chemical was eliminated from further consideration.

• Essential Nutrients. Naturally occurring inorganic chemicals that are considered as essential nutrients (calcium, iron, potassium, magnesium, and sodium) were eliminated as COCs.

Exposure Point Concentrations

Exposure Assessment

• Dermal Contact
• Incidental Ingestion
• Particulate Inhalation

Each receptor was evaluated using the reasonable maxium exposure (RME) assumption. The RME case is designed to be a measure of high-end exposure representing an estimate of upperbound risk (approximately the 95th percentile of the dose distribution). The purpose of the RME is to obtain a conservative estimate of exposure that is still within the range of possible exposures. Exposure was estimated for the the noncarcinogenic case, in which exposure is averaged over a given exposure duration. Three routes of exposure were investigated: incidental soil ingestion, dermal contact with soil, and soil particulate inhalation. The doses were expressed as intakes in units of milligrams contaminant per kilogram body mass per day (mg/kg-day). The primary parameters used in calculating intake are the chemical concentration, intake rate, and frequency and duration of exposure. The parameters and equations used to calculate the intakes are detailed in the following sections.

Incidental Soil Ingestion

Dermal Contact

Particulate Inhalation

Toxicity Assessment

• Whereas oral reference doses are typically based on administered doses, dermal exposures are usually expressed in terms of absorbed doses (amount that enters the bloodstream). IRIS contains a wealth of data on oral reference doses and these were retrieved for the COCs. However, EPA has yet published dermal toxicity values. As such, dermal reference doses were established by extrapolation from the oral values multiplication by gastrointestinal absorption factors: 1.0 for organics and 0.3 for metals (EPA, 1992).

• For inhalation toxicity data, values are not available for a majority of constituents. There are two generally practiced methods to deal with this. The first is simply to not evaluate toxicity and risk for inhalation because the uncertainty with the toxicity assessment is too high. The second is a process called route-to-route extrapolation, in which the inhalation reference doses are assumed to be equal to the oral doses. The latter method was employed for this assessment.

• Of all the chemicals in the IRIS database, perhaps the least is known about PAH toxicity. Even the data for oral toxicity is sparse. As such, PAHs for which there is no oral reference dose were assumed to have similar toxic effects as pyrene, for which there was a published value.

Due to the need to make assumptions such as these, the uncertainties related to the toxicity assessment can be substantial. Adding to the uncertainty question, are published reference dose data even valid? Reference doses are typically developed using laboratory studies in which animals are exposed to high doses of chemicals. The results are then extrapolated to low doses in humans. What is the uncertainty in this process? Modern science has been able to validate the procedures and results of the data reduction and exposure assessment processes fairly well. However, the toxicity assessment remains the weak link in the risk assessment process. As scientific advancements in the understanding of human toxicants are made, the accuracy of the overall risk assessment process will improve.

Risk Characterization

From the summary table, it can be seen that hazard indices of 1 and 0.4 were calculated for the child and adult receptors, respectively. Due to the many uncertainties involved up to this point in the risk assessment, the hazard indices were reported to only one significant digit. The highest hazard quotients occurred in the soil ingestion and dermal contact exposure scenarios. Among the chemicals, over 95% of the total risk was attributable to only 6 constituents: antimony, arsenic, barium manganese, nickel, and vanadium. Interestingly, PAHs, which are known creosote constituents, contributed very little to the overall hazard indices. In evaluating the hazard indices, a few things should be considered:

• The calculated values could be higher than 1 and 0.4:  For this risk assessment, many chemicals were eliminated without passing through the data reduction process. These include pesticides, PCBs, dioxins/furans, volatile organic chemicals, and semivolatile organic compounds other than PAHs. If these constituents were included, the hazard index would likely be higher.

• The calculated values could be lower than 1 and 0.4:  The summation of hazard quotients over all COCs assumes that noncancer effects are additive. In reality they are not. For example, barium is toxic to the cardiovascular system whereas manganese is known to adversely affect the central nervous system. As such, a preferable analysis method might be to calculate a different hazard index for each major organ/system. In addition, the RME exposure case is designed to be an estimate of upperbound risk; hazard indices based on an average exposure case would be lower.

Conclusions

• Automate the Data Reduction. Writing visual basic scripts to query the database and perform statistical calculations would dramatically simplify the data analysis and reduction process. It would have allowed the differentiation of soil samples based on depth and quality control type (normal or duplicate). In addition, automation would make possible the calculation of all chemicals (not just the initial COCs). In this manner, cumulative risk could be evaluated.

• Include a CTE Exposure Scenario. In addition to the RME case, exposure assessments often include a central tendency exposure (CTE) case. The CTE case is derived using average physiological parameters and activity patterns, and is intended to provide "average" exposure estimates. Hence, risk estimates based on CTE are more likely to apply than RME values. The calculation process of CTE dose is identical to the RME case except that some of the exposure parameters are changed.

• Include Ingestion of Homegrown Produce as an Exposure Route. For surface soil contamination, this exposure route is recommended by EPA when evaluating a residential populations as receptors. The procedure involves estimating the chemical concentrations in produce rooted in contaminated soil. As such, development of plant uptake factors based on chemical and soil properties would be necessary.

• Investigate Carcinogenic Risk. This would involve several steps. For the exposure assessment, exposure would be averaged over a 70-year lifetime. In the toxicity assessment, chemical-specific cancer slope factors would be retrieved from the IRIS database. Excess lifetime carcinogenic risk could then be calculated by dividing intake by the slope factor.

In closing, I would like to thank EPA Region VI and my former employer, Roy F. Weston, Inc. for allowing me to use the site data for this term project. At their request, I've changed some basic details about the site (name, description, etc.), but the data is real. In particular, I'd like to thank Andy Kallus and Steve Mitchell of Weston for taking the time to answer my numerous questions.

References

1. EPA (U.S.Environmental Protection Agency), 1989. Risk Assessment Guidance for Superfund. Volume I: Human Health Evaluation Manual Part A. Interim Final. Office of Solid Waste and Emergency Response (OSWER), Washington, D.C. OSWER Directive 9285.701A.

2. EPA, 1991. Human Health Evaluation Manual, Part B: "Development of Risk-Based Preliminary Remediation Goals". OSWER Directive 9285.701B.

3. EPA, 1992. Supplemental Region VI Risk Assessment Guidance. April 1992.

4. EPA, 1996. Integrated Risk Information System (IRIS). EPA Toxicological Database, Washington, D.C.

   Return to Eric Tate's home page