Science and Judgment in Risk Assessment (1994)

Page 479

Appendix
G
Improvement in Human Health Risk Assessment Utilizing Site- and Chemical- Specific Information: A Case Study

Del Pup, J.,1

Kmiecik, J.,2

Smith, S.,3

Reitman, F.1

1.0 Introduction

The U.S. Environmental Protection Agency (EPA) has classified 1,3-butadiene (butadiene) as a B2 (''probable") human carcinogen.4 Conservative screening level cancer risk estimates reported by EPA to rank sources and prioritize regulatory action associated emissions of butadiene from the Texaco Chemical Company, Port Neches, Texas facility with a maximum individual risk of 1 in 10. Although the agency emphasized that these screening level estimates should be viewed only as rough estimates of the relative risks posed by the facility under evaluation, and should not be interpreted to represent an absolute risk of developing cancer, the risk estimate generated a high level of concern. In this paper we provide a discussion of results of an effort to use site-specific data, species differences in the metabolism of butadiene, the Monte Carlo procedure, and other factors to estimate risk to the community. The effect of some of these factors is profound. For example, using this information, the range of risks at the closest residence is estimated to be 1 in 10,000,000 to 3 in 10,000. This range of

1Texaco, Inc.

2Texaco Chemical Company

3Radian Corporation

4EPA classifies chemicals for which there is sufficient evidence for carcinogenicity in experimental animals and inadequate or no evidence for carcinogenicity in humans as Group B2, "probable human carcinogens."

Page 480

uncertainty is driven largely by species differences in butadiene uptake and metabolism used in the slope factor.

The purpose of this study is twofold:

It is neither the intent of this paper to evaluate any cause and effect relationship between 1,3 butadiene exposure and cancer in humans, nor to provide the most scientifically defensible cancer potency estimates for 1,3-butadiene. Risks referred to in this paper are hypothetical estimates useful for regulatory purposes. These estimates assume as a matter of regulatory policy that a low-dose linear carcinogenic response to butadiene occurs in humans. Actual risks would be zero if butadiene is not carcinogenic to humans at these exposure levels.

Texaco initiated this evaluation in 1990 (Radian Corporation, 1990). That assessment focused on increasing the precision of the EPA screening level risk estimates based on more realistic representation of emissions, dispersion and exposure after completion of the Butadiene Modernization Project. This project centered around changing the extraction solvent used in the distillation process and in changing the "once-through" cooling water system to a recirculating cooling tower system in order to reduce butadiene emissions. Although based on site-specific information wherever possible, the risk assessment noted several sources of uncertainty that impacted interpretation of the risk estimates. Primary sources of uncertainty were identified as estimated emissions rates, assumptions and algorithms associated with dispersion modeling analysis, assumptions used to calculate inhalation exposure, and the theoretical estimate of the carcinogenic potency of butadiene, if any, in humans.

The Butadiene Modernization Project, now largely completed, has resulted in a process that is cleaner from both a product purity and environmental perspective. Butadiene emissions have been reduced more than 90 percent. Repeating the prior EPA screening level analysis predicts a maximum individual cancer risk after completion of this project in the range of 5-10 in 1000 based on a 70 year exposure to the maximum predicted annual-average ground level concentration 200 meters from the center of the plant. The current study was initiated to reexamine some of the sources of uncertainty in the risk estimates and to update the risk estimates, using the most site-specific and chemical-specific information available (Radian 1992a) The resulting risk estimates range from 3 in

Page 481

10,000 to 1 in 10,000,000 at the nearest residence, which are much lower than EPA's original 1 in 10 risk estimate. In addition, we also provide estimates of risk to the nearest residence and school using the Monte Carlo analyses. These provide central tendencies which result in even lower estimates of risk.

The health risk analysis undertaken by the author improves upon the EPA-generated health risk assessment by reevaluating assumptions pertinent to determining the maximum exposed individual risk and risks at various locations in the community. Risks were characterized for the conventional "worst case" 70-year exposure, the 30-year upper bound exposure, the 9-year average residential exposure, and the 95th percentile fraction of life exposed (FLE) based on national human activity pattern distributions. Assumptions used in the development of the EPA-sanctioned unit risk factor for butadiene and impact on the magnitude of risk using alternate unit risk factor assumptions were also evaluated. The assessment also evaluated differences between ground-level concentrations predictions by the Industrial Source complex Long-Term (ISCLT) and the Industrial Source Complex Short-Term (ISCST) atmospheric dispersion models. In addition, results using two meteorological data sets for the area and various decay coefficients for butadiene were evaluated.

This study addresses many of the issues, assumptions, and uncertainties inherent to inhalation pathway risk assessments. However, it should be noted that the analyses, conducted for the current study are site-specific and, therefore, the results may not be applicable to other source configurations, meteorological data sets, or other receptor populations. The study is intended to illustrate a process by which human risk assessments can be improved by using available site-and chemical-specific information.

2.0 Emission Statements

The facility produces butadiene by solvent extraction from a crude C4 stream. The process involves distilling the extracted butadiene to remove heavy ends and final polishing to obtain a butadiene product with purity of 99.7%. Potential sources of butadiene emissions included equipment components in the process units, tank farms, and on the product loading racks; cooling towers; process flares; the dock flare; steam boilers; wastewater treatment plant; the cracking unit; and the butadiene sphere. The butadiene emission estimates were based primarily on actual process data and source-specific information, and on Air Control Board and/or EPA approved emission factors.

It is recognized there are other butadiene sources in the Port Neches area (e.g. butadiene emissions from other area facilities). These other sources of butadiene emissions were not included in the analysis.

Page 482

3.0 Environmental Fate And Transport Modeling

Atmospheric fate and transport is usually assessed using a mathematical atmospheric dispersion model. Industrial Source Complex (ISC) Models are classified as "preferred" models in the EPA's "Guidelines on Air QUality Models (Revised), 1987 (EPA, 1987). Two versions of the ISC model are available. Both the Industrial Source Complex Short Term (ISCST) and the Industrial Source Complex Long-Term (ISCLT) model are steady state Gaussian plume models preferred for use with industrial complexes with flat terrain such as that found in the area of the facility.

3.1 Industrial Source Complex Model Comparisons

The ISCST model is designed for use in predicting concentrations using averaging periods from one hour to one year. This model utilizes discrete hourly meteorological data. The ISCLT model is designed for use in predicting annual-average concentrations. This model utilized meteorological data in the format of a STAR summary. The STAR summary is a joint-frequency distribution of wind speed, wind direction, and stability classification, processed from discrete hourly observations. The use of this meteorological data summary enables the ISCLT dispersion model program to calculate ambient concentrations much faster than ISCST because dispersion calculations are performed for a small number of meteorological categories rather than for every hour of the year. The ISCLT and ISCST use identical equations for calculating ambient concentrations, with the exception of several changes necessary for the incorporation of the STAR summary.

A model comparison using site-specific inputs revealed fairly good agreement between long-term and short-term results (Radian, 1992b). A 12.5% higher maximum off-property concentration was predicted using the long-term model, but the average concentration of all receptor locations predicted by both models were identical. Given the good agreement between the models, the requirement of evaluating butadiene for long-term or chronic effects, and the faster model execution time, the ISCLT was chosen for this analysis.

3.2 Effects of the use of Atmospheric Decay Coefficients in ISCLT

The ISCLT model provides a mechanism to account for pollutant removal by physical or chemical processes. There are three main chemical reactions which were considered important to evaluating atmospheric concentrations of butadiene, including: 1) reaction with hydroxyl radical (·OH);2) reaction with ozone (O3); and reaction with nitrogen trioxide radical (·NO3) (EPA, 1983). The reaction with ·OH is dominant during the day while reaction with ·NO3 is domi-

Page 483

nant at night. Ozone reactivity occurs during the day and night. All reactions are temperature dependent, with butadiene residence times being greater during the winter months and dependent on the chemical species available for reaction in the particular airshed of interest.

Annualized pollutant decay factors were developed by Radian Corporation for use with the ISCLT model based on site-specific temperatures and airshed data estimates. The decays were annualized to address the long-term or chronic exposure aspects of the study. Due to the low solubility of butadiene, physical removal processes such as pollutant incorporation into clouds and rain were not considered to be important pollutant degradation processes and were not considered in this analysis.

Figures 3-1, 3-2, 3-3, and 3-4 illustrate concentration isopleths for no decay, low decay, median decay, and high decay of butadiene, respectively. These results indicate that the inclusion of pollution decay in the transport and fate analysis of butadiene has only minimal effects on predicted ground-level concentrations near the facility. However, as distance from the facility increases, inclusion of butadiene decay in the fate and transport analysis significantly decreases predicted ground-level concentrations.

3.3 Alternative Meteorological Data Set Comparison for the ISCLT Model

Two quality-assured sets of meteorological data were evaluated for use in this analysis: 1) a 14-year composite annual joint frequency distribution of wind speed, wind direction, and stability class (STAR) data processed from the National Weather Service (NWS) hourly surface observations at the County Airport, located approximately four miles from the plant boundary; and 2) a two-year composite STAR data set processed from 1990 and 1991 Regional Planning Commission (RPC) continuous observations at another County Airport location, approximately three miles form the plant boundary. The RPC data were selected for use in the majority of the analyses due to the continuous nature of the observations and the use of measured mixing heights. However, to examine the sensitivity of the risk estimates to changes in the meteorological data set, the ISCLT dispersion model was run with identical inputs, varying only the meteorological data. At nearby locations, predicted concentrations using RPC data were 25 to 100% higher than predicted concentrators using the NWS data. Using the RPC data, concentrator isopleths would extend farther to the east and are more rounded. using the NWS data, the isopleths would show more of a northsouth bias (Radian, 1992a).

Page 484

FIGURE 3-1
Concentration Isopleths (µg/m3). No Butadiene Decay

Page 485

FIGURE 3-2
Concentration Isopleths (µg/m3). Low Butadiene Decay

Page 486

FIGURE 3-3
Concentration Isopleths (µg/m3). Median Butadiene Decay

Page 487

Page 488

4.0 Human Health Assessment

Risk characterization involves integrating exposure and toxicity information into quantitative and qualitative expressions of potential health risk. For potential carcinogens such as butadiene, risk can be characterized by estimating the potential for carcinogenic effects or by estimated ambient air concentration with health-based ambient guidelines or standards.

To characterize potential carcinogenic effects, estimated risks that an individual will develop cancer over a lifetime of exposure to butadiene were calculated from projected intakes and the cancer slope factor. The cancer slope factor converts estimated daily intakes directly to an estimate of incremental risk as follows:

The slope is often an upper 95th percentile confidence limit of the probability of response based upon experimental animal data and an assumption of linearity in the low-dose portion of dose-response curve. Therefore, the carcinogenic risk estimate will generally be an upper-bound estimate, indicating that the "true-risk", if any, will probably not exceed the risk estimates based on the slope factor and is likely to be less than that predicted.

Individuals may be exposed to chemical in air by inhalation of chemicals in the vapor phase or adsorbed to particulate. Dermal absorption of vapor phase chemicals such as butadiene is considered to be lower than inhalation intakes and, therefore, was not quantified in this risk assessment (EPA, 1989). Inhalation of airborne vapor-phase chemicals can be quantified using the following formula:

Lifetime exposure must be evaluated to determine cancer risk. To provide a conservative analysis of lifetime community exposure, the exposed population (represented by an average 70 kg adult) has been assumed to inhale (at an average rate of 20 m3/day) predicted ground-level concentrations continuously, 24

Page 489

hours/day, 365 days/year, for a 70 year exposure duration. More recently, EPA has employed ''reasonable maximum" assumptions of 24 hours/day, 350 days/year for 30 years.

4.1 Characterization of Risk

To characterize the risks, both the health variables and the exposure variables were combined under three scenarios, the Base Case, Worst Case and Best Case (Table 4.1). For example, the Worst Case includes inputs that reflect a highly conservative approach whereas the Base Case and Best Case make use of different levels of sophistication in the utilization of site-specific data, exposure assumptions, and recent biological data on the uptake and metabolism of butadiene.

The ISCLT model calculates an ambient concentration at each point (or receptor) provided in the model input. Receptor placement was designed to identify the location of the maximum off-property concentration. Additional receptors were also placed at the nearest residences and the nearest school complexes in several directions. Therefore, concentrations at several locations of special interest were determined. Table 4-2 summarizes the Base Case maximum individual risk calculations for each of the nearby receptor locations. Risk estimates at the closest residences were 1 in 10,000. Risk estimates at the location of maximum off-property concentration were about 5 times higher. Estimated risks at the school locations were lower, ranging from 7 in 100,000 to 4 in 1,000,000. This can be compared with the approximate 1 in 4 background risk of developing fatal cancer in the U.S. population (Harvard School of Public Health, 1992). Refinements to this assessment were made by evaluating additional variables impacting on the risk estimates. Some of these, particularly the slope factor, have a high level of uncertainty.

4.1.1 Effect of Exposure Assumptions

Realistically, very few people remain in the same location for a lifetime. To account for exposure durations less than a lifetime, the following formula can be used to quantify the Lifetime Average Daily Exposure (LADE) (Price et al.1991):

Page 490

National statistics are available on the upper-bound (30 years) and average (9 years) number of years spent by individuals at one residence (EPA 1989, 1991). The "upper-bound" value was used as the exposure duration when calculating the reasonable maximum residential exposures. An exposure frequency of 350 days/year was used, assuming 15 days/year are spent away from home. Assuming a 70 year lifetime, the FLE is an average of 0.12 and a reasonable maximum of 0.41.

A Point Source Exposure Model (PSEM) was used to characterize the distri-

Page 491

bution of exposures received over a long-term period based on information on mobility, mortality, and daily activity patterns (Price et al 1991). PSEM models the time of residence in the zone of impact and the amount of time spent at home as variables that yield a probability density function for the FLE. The model predicted that, on average, individuals live in their current house for 16.5 years and spend 18 hours per day at home. The average value of the FLE calculated by PSEM using national statistics was 0.16. The median value was 0.12, and the 95th percentile of the distribution was 0.42. Inhalation exposures for all receptor locations were calculated based on residential exposure assumptions, using these FLE values. It is assumed for this report that 1,3-butadiene concentrations are the same inside and outside the home. No attempt has been made here to validate this assumption.

Several exposure scenarios were examined in this assessment including: 1) "worst case" (24 hours/day, 365 days/year for 70 years); 2) "reasonable maximum" (24 hours/day, 350 days/year for 30 years); 3) 95th percentile FLE based on national human activity pattern distributions, and 4) average FLE based on national human activity pattern distributions. Figure 4-1, 4-2, and 4-3 illustrate the areal extent encompassed by several risk levels using traditional ''worst case", reasonably maximum, and average exposure assumptions. As indicated in the figures, the areal extent encompassed by specific risk levels is very sensitive to changes in the time, frequency, and duration of exposure.

Page 492

FIGURE 4-1
Area Encompassed by Specific Risk Levels-Traditional Worst Case Exposure Assumptions

Page 493

FIGURE 4-2
Area Encompassed by Specific Risk Levels-Reasonable Maximum Exposure Assumptions (Base Case)

Page 494

FIGURE 4-3.
Area Encompassed by Specific Risk Levels-Average Exposure Assumptions

Page 495

4.1.2 Effect of Cancer Slope Factor Assumptions

The EPA-sanctioned slope factor for butadiene of 1.8 (mg/kg-day)-1 was used in all previous analyses (IRIS 1992). In the current analysis, however, risk estimates were also generated using alternative slope factors based on research that the EPA slope factor may overpredict risks to the human population.5 Cancer slope factors can be converted to unit risk estimates to determine the risk per unit air or water concentration. The inhalation unit risk can be calculated by dividing the slope factor by 70 kg (average body weight for an adult) and multiplying by 20 m3/day (adult average inhalation rate) assuming a 70 year exposure period (EPA 1989, 1991).

EPA calculated in inhalation risk estimate for butadiene of 2.8-04 (µg/m3)-1, based on an absorption factor of 54%, which was derived from preliminary results of an absorption study conducted in mice and sponsored by the National Toxicology Program (NTP). The procedure for determining animal-to-equivalent human dose was adjusted to account for the fact that at high concentrations, the internal dose (mg/kg) is not directly proportional to external concentrations. A final report of the NTP study has been published and differs significantly from the preliminary results (Bond et al. 1986). Results from the final report suggested that butadiene retention by mice in the initial study may have been overestimated by a factor of five. Based on these data, risk estimates derived using EPA-sanctioned values for butadiene should be adjusted downward by approximately a factor of five. Based on the discussion published in EPA's Integrated risk Information System (IRIS, 1992), EPA used an absorption factor of 54% in calculating a slope factor. IRIS states that differences between the retention of butadiene reported in the initial and final study have been accounted for in EPA's calculations. Assuming this is correct, there is no need to make adjustments in risk estimates based on the EPA value. However, the animal upper-limit slope factors are identical to those published by the EPA in 1985, suggesting that this correction has not been made (EPA, 1985). If the correction was not made, the downward adjustment by a factor of five is appropriate.

The respiratory systems of humans differ from experimental animals in many ways. These differences result in variations in air flow, deposition of inhaled agents, as well as the retention of that agent. The dose of partially soluble vapors, such as butadiene, is proportional to oxygen consumption. Oxygen consumption is, in turn, proportional to (body weight) and is also proportional to the

5A cohort epidemiologic study of workers employed at this facility between 1943 and 1979 showed a statistically significant deficit for all causes of death and all cancers. There was, however, a statistically significant excess of deaths from lymphosarcoma. This was concentrated in workers employed less than 10 years and first employed prior to 1946 (Divine, 1990).

Page 496

solubility of the gas in body fluids, which is expressed as an absorption coefficient for the gas. In the absence of experimental evidence to the contrary, the absorption coefficient is assumed to be the same for all species. Therefore, butadiene exposure concentrations (in ppm) used in animal studies were assumed to be equivalent to the same concentration in humans. However, smaller animals have higher minute respiratory volumes per unit of body weight to supply their relatively larger requirements for oxygen. Since the dose of butadiene (by inhalation) is proportional to oxygen consumption, species with higher minute respiratory volumes would be expected to have larger body burden of the chemical.

Studies have been conducted which indicate that nonhuman primates absorb considerably less butadiene than mice (Dahl et al. 1990). At 10 ppm, mice retain approximately 6.6-fold more butadiene than monkeys. The human species is much more closely related to the monkey than the mouse, both physically and anatomically. Therefore, primate retention data should be used as a basis for estimating retention by humans. On this basis, risk estimates derived from EPA sanctioned toxicity values should be adjusted downward by a factor of six.

In quantitative risk modeling, internal concentrations of butadiene were used as a measure of dose. However, in doing so, species differences in metabolism of butadiene were ignored. In studies sponsored by NTP (Dahl et al. 1990), mice were shown to attain approximately 590-fold higher blood levels of the monoepoxide (a DNA-reactive and mutagenic metabolite of butadiene, assumed to be a toxic metabolite) than did primates.6 Based on the assumption that humans metabolize butadiene in a manner that is more closely related to nonhuman primates, humans should be approximately 590-fold less sensitive to butadiene's carcinogenic effects than mice. Therefore, estimates of risk should be adjusted by a factor of 590 to account for species differences in metabolism of butadiene. Use of the internal concentration of the monoepoxide would obviate the need to adjust for difference in retention of inhaled butadiene.

The available comparative studies suggest that the equivalent potency of butadiene in humans could be substantially less than that used as the basis for EPA's calculated cancer slope factor. Based on the available data, the slope factor could be adjusted downward (i.e., to indicate lower potency for humans) by a factor of 30 (5 × 6 based on current retention data for the mouse and mouse/primate differences in retention) to 590 (based on mouse/primate differences in blood levels of the monoepoxide). Since risks change proportionally to changes in the butadiene slope factor, the risks using the alternative slope factors are lowered by a factor of 30 to 590. Figures 4-4. 4-5 and 4-6 illustrate the way in

6Metabolites were tentatively identified, based on co-distillation with standards.

Page 497

FIGURE 4-4.
Area Encompassed by Specific Risk Levels-EPA Slope Factor (Base Case)

Page 498

FIGURE 4-5.
Area Encompassed by Specific Risk Levels-EPA Slope Factor Adjusted by a Factor of 30

Page 499

FIGURE 4-6.
Area Encompassed by Specific Risk Levels-EPA Slope Factor Adjusted by a Factor of 590

Page 500

which changes in the butadiene slope factor affect the area encompassed by specific risk levels.

5.0 Probabilistic Monte Carlo Simulation

Risk estimates resulting from a series of "worst case" assumptions can be expected to overestimate actual risk. However, there is no way for the regulator, industry representatives, the potentially exposed population, or other interested parties to interpret the degree of conservatism. EPA risk assessments are expected to address the range of risk including the central tendency and high end portion of the risk distribution (EPA, 1992). In addition, they are expected to include a statement of confidence in the risk assessment itself. Stochastic analysis of risk provides a distribution of estimated risks based on the use of probability density functions for input parameters instead of single point estimates.

Monte Carlo simulation calculates risk through numerous iterations using randomly generated values from the defines probability functions. The resulting distribution of risk estimates makes greater use of the scientific evidence and data related to exposure and theoretical risk without sacrificing conservatism. Monte Carlo avoids compounding of "worst case" assumptions and uncertainty, and provides quantitative information on the uncertainty in the risk values.

The shape of the distribution and the range between low and high end estimates portray the uncertainties incorporated in the assessment and can be used to interpret the level of confidence in the assessment. A narrow range between 5th and 95th percentile of the distribution implies a low level of overall uncertainty and, consequently, a high level of confidence in the assessment. A broad range implies a high level of uncertainty.

In this assessment, the range in the risk estimates from the 5th and 95th percentile at the closest residence was 4 in 100,000,000 to 2 in 10,000 (Radian, 1992b). This range spans almost four orders of magnitude, indicating a very high level of uncertainty. The range in estimated risk from the 5th to the 95th percentile at the closest schools was 5 in 10,000,000,000 to 6 in 1,000,000. This range spans more than four orders of magnitude. The slightly greater span in the risk range at this location results form the greater potential influence of butadiene decay in the atmosphere as the distance from the facility increases. Therefore, the level of confidence in the estimates of risk associated with butadiene at the facility can only be described as low.

Page 501

6.0 Conclusions

A number of variables examined in this risk assessment significantly impacted the final theoretical risk estimates. These variables included: 1) the meteorological data used in transport and fate modeling; 2) butadiene decay factors; 3) exposure time, frequency, and duration; and 4) the slope factor for butadiene.

Base Case estimates were developed including inputs for key variables that are relatively conservative. The sensitivity of Base Case estimates to varying inputs for these key variables was evaluated. The Base Case predicted risks in the range of 1 in 10,000 at the nearest residences, and 4 in 1,000,000 to 7 in 100,000 at the nearest schools. Worst Case estimates were only two to three times higher than Base Case estimates.

Best Case estimates, which provide an additional measure of the level of uncertainty associated with the estimates, ranged from more than three to four orders of magnitude lower than Worst Case estimates. The butadiene slope factor contributes almost three orders of magnitude to the theoretical risk estimates separating the Worst Case and Best Case scenarios. While the butadiene decay factor did not significantly affect the risk estimates at nearby locations, this effect was location dependent. The Base Case risk estimates (1 in 10,000 at the nearest residences) represents an upper-bound to the risk associated with the butadiene emissions from the facility. The ''true risk" is unlikely to be higher, and is most likely lower. An examination of some of the key variables that influence estimates of theoretical risk indicates that the maximum individual risk at the nearest residences may be as low as 1 in 10,000,000. Risk estimates in this report should be considered in comparison to the approximate 1 in 4 background fatal cancer risk in the U.S. population. In all cases the risk would be zero if butadiene is not carcinogenic in humans at prevailing exposure levels.

References

1. Bond, J.A., Dahl, A.R., Henderson, R.F., Dutcher, J.S., Mauderly, J.L., Birnbaum, L.S., 1986. Species differences in the disposition of inhaled butadiene. Toxicol. Appl. Pharmacol. 84: 617-627, 1986.

2. Dahl, A.R., Bechtold, W.E., Bond. J.A., Henderson, R.J., Muggenburg, B.A., Sun, J.D., and Birnbaum, L.S., 1990. Species Differences in the Metabolism and Disposition of Inhaled 1,3-Butadiene and Isoprene. Environ. Health Perspect. 86: 65-69, 1990.

3. Divine, B.J. 1990 An Update on Mortality Among Workers at a 1,3-Butadiene Facility - Preliminary Results. Environmental Health Perspectives 86: 119-128.

4. Harvard School of Public Health, The Center for Risk Analysis. Annual Report. 1992.

Page 502

5. Integrated Risk Information System (IRIS), 1992. U.S. EPA Information Network On-Line Database. Data retrieved January, 1992.

6. Price, P.S., Sample, J., Streiter, R., 1991. PSEM. A model of Long-Term Exposures to Emissions from Point Sources. Presented at the 84th annual meeting of the Air and Waste Management Association, Vancouver, British Columbia, June 16-21, 1991. 91-172.3.

7. Radian Corporation, 1990. Site-Specific Evaluation of Potential Cancer Risk Associated with 1,3-Butadiene Emissions from the Texaco, Port Neches Facility. Prepared for Texaco Chemical Company, May 14, 1990.

8. Radian Corporation, 1992a. Site-Specific Evaluation of Potential Cancer Risk Associated with 1,3-Butadiene Emissions from the Texaco, Port Neches Facility. Prepared for Texaco Chemical Company, April 16, 1992.

9. Radian Corporation, 1992b. Technical Memorandum from Randy Parmley, Radian Corporation to Jim Kmiecik, Texaco Chemical Company on "Summary of ISCST and ISCLT Model Comparison," dated 31 January 1992.

10. Radian Corporation 1992c. Site-Specific Evaluation of Potential Cancer Risk Associated with 1,3-Butadiene Emissions from the Texaco, Port Neches Facility. Addendum I-Probabilistic Monte Carlo Simulation. August 10, 1992.

11. U.S. Environmental Protection Agency (EPA), 1983. Health and Environmental Effects Profile for 1,3-Butadiene. EPA/60/0-84/120. May, 1983.

12. U.S. Environmental Protection Agency (EPA), 1985. Mutagenicity and Carcinogenic Assessment of 1,3-Butadiene. EPA/600/8/85-004f.

13. U.S. Environmental Protection Agency (EPA), 1987. Industrial Source Complex (ISC) Dispersion Model. Addendum to User's Guide. 1987.

14. U.S. Environmental Protection Agency (EPA), 1989. Risk Assessment Guidance for Superfund. Volume 1 Human Health Education Manual (Part A). Interim Final. EPA 540/1-89-002. December, 1989.

15. U.S. Environmental Protection Agency (EPA), 1991. Risk Assessment Guidance for Superfund. Volume 1 Human Health Evaluation Manual. Supplemental Guidance. Standard Default Exposure Factors. March 25, 1991.