Document jgwRZOo2RLRDwkpE0Vv1BjQdN

ABDOO134489 R Review of Current Quantitative Cancer Risk Assessment Evaluations By EPA, Review of the Air Dispersion Model employed by the Vista Chemical company Aberdeen, Mississippi, and Estimation of Cancer Risks at Different Distances from Emission Sources from the Vista Chemical Company Facility Prepared by: Clement International Corporation 9300 Lee Highway Fairfax, Virginia 22031 Prepare for: Vista Chemical Company P.O. Box 91 Aberdeen, Mississippi 39730 February 4, 1991 ABDOO134490 Introduction This report reviews and compares the use of two inhalation unit risk values that are reported in two different EPA documents for vinyl chloride* Comments are made as to the scientific validity of the unit risks values for estimating acceptable ambient air concentrations to protect human health, and a review was conducted on the air dispersion model used by Vista Chemical Company in their Aberdeen, Mississippi plant to characterize emissions levels. Exposures estimated from the model were then used to determine human cancer risks at different points from the emission sources under different production conditions. ABDOO134491 Review and Comparison of Determining Acceptable Ambient Air Concentrations for Vinyl Chloride Based on Carcinogenicity in two EPA Reports There are two recent EPA documents in which estimates were derived for ql* values for vinyl chloride. In the first document (Health Effects Assessment for Vinyl Chloride, EPA, 1984, office of Health and Environmental Assessment, EPA/540/1-86-036) the animal ql* was given as 4.23xl0~3 (mg/kg/day) and the human ql* was given as 2.5xl0"2 (mg/kg/day)"1 (see Appendix 1 for determination of animal ql* value using GL0BAL86). Inhalation studies in rats were used and total tumors were used as the toxicological endpoint. This ql* value can be converted to units of ppm"1 or (ug/m3)"1 as follows by employing the same conversion factors used by EPA in converting animal exposures in ppm to human equivalent exposures in mg/kg/day: 1. Correction forL_surface area between humans and rats [(1/(70 kg human/.35 kg rat))1/3] * 2.5xl0"2 4.27X10"3 (mg/kg/day)"1 2. Adjustment for body jweiaht of rat 4 27xl0"3/. 35 kg rat = 1.22X10"2 (mg/day)"1 3. Adjustment for, breathing rates of rats .223 m3/d * 1.22X10"2 = 2.72X10"3 (mg/m3)"1 or 2.72X10"6 (ug/m3)"1 or 2.56X103 (ug/m3*ppm)*2.72xl0*6 (ug/m3)"1 = 6.96X10"3 ppm"1 (EPA84 calculation was 6.80X10"3 ppm'1) 4. It was assumed that all the vinyl chloride inhaled was absorbed into systemic circulation. This is a conservative estimate. 5. The previous ql* value that the state of Mississippi used to determine cancer risk was 4.1xl0'6 (ug/m3)*1. This value is greater than the calculated value of 2.72X10"6 (ug/m3)-1 in step 3 above by 1.5. In a more recent update (Health and Environmental Effects Profile for Chloroethene, EPA, 1985, Environmental Criteria and Assessment Office ECAO-Cinn-P155) the rat ql* was determined to be 5.04xl0"2 (mg/kg/day) (see Appendix 2 for determination of the rat ql* value using GLOBAL86) and the human ql* was determined to ABDOO134492 be 2.95X10"1 (mg/kg/day)"1. Animal studies from the same investigator were used as in the EPA 1984 report; instead of using total tumor incidence, however, angiosarcomas of the liver were used as the only toxicological end point. Corrections to the dose were made since the observation period was extended after exposure stopped and an assumption was made that only 50% of the inhaled vinyl chloride was being absorbed. By using the same conversion factors above and correcting for % uptake the following inhalation ql* was determined. 1. Correction for surface area between humans andrats [(1/(70/.35))1/3]*2.95xl0'1 (mg/kg/d)"1 = 5.05X10"2 (mg/kg/d)"1 2. Adjustment for body weight, and breathing rate of rats [5.05xl0~2 (mg/kg/d)-1/.35kg]*.223 m3/d =* 3.22xlO"2 (mg/rn3)"1 or 3.22xl0's (ug/m3)"1 3. In this assessment EPA assumed that only 50% of the vinyl chloride was absorbed through the lungs. To correct for this the ql* value has to be multiplied by .5. (. 5) *3 22xl0~5 (ug/m3)"1 = l.filxlO"5 (ug/m3)"1 or 4 12xl0~2 ppm"1 The state of Mississippi is recommending a ql* value of 4.2xl0~5 (ug/m3)"1. This value is derived from the more recent EPA report just described from the ql* value of 2.95X10"1 (mg/kg/d)"1. This number is 2.6-fold greater than the number calculated above (1.61 xlO"5). Mississippi derived their number as follows. [2.95X10"1 (mg/kg/d)"1]*(.5)*(20 m3/d)*(l0-3mg/ug)/70 kg = 4.21x10" (ug/m3)"1 This calculation is incorrect, since the same conversion factors used by EPA to convert experimental levels (ppm) to human exposure equivalents (mg/kg/day) were not used by the state of Mississippi to back-calculate from (mg/kg/day) to ppm or ug/m3. Analysis _of the data used_bv EPA in determining: a human _ol* of 2.95xl0~3,_lmgykg/dav) 1 in their 1985 report Figure 1 illustrates the relationship between liver angiosarcoma and vinyl chloride exposure in rats. No dose-response relationship is obvious. Furthermore, EPA did not use all the data in determining the ql* value. It only selected data points 0, .172, .344 and .86 mg/kg/day. There was no rationale provided why other data points were omitted. In addition, the data points are not even taken from the same experiment. For example, data ft ABDOO134493 points 0.344 and 0.86 were from Maltoni et al. experiment BT15 and data point 1.719 was taken from Maltoni et al. experiment BT1. Therefore, the manner in which the ql* value was calculated is scientifically incorrect and meaningless. In the 1985 report the length of the experiment was reported to be up to 1029 days, whereas the exposure period to vinyl chloride was for 365 days. In the transformation of experimental doses to human equivalent doses an adjustment factor of 365/1029 was used. From a biological sense this dose-averaging approach may not be appropriate. The actual time to tumor data needs to be obtained in order to make the appropriate adjustments to dose. The first ql* value established by EPA in 1984 was based on total tumors rather than liver angiosarcomas. Although in some instances total tumors may not provide the most sensitive assessment of cancer risk, the data as shown in figure 2 do show a linear dose response relationship up to 239.1 mg/kg/day. Therefore, these data provide a much better data base to determine a human ql* value, than the data base given in the EPA 1985 report. Discussions with EPAIs Human Health Assessment Group EPA's Office of Health and Environmental Assessment often has contractors provide health profiles and reports on various chemicals. These documents are used primarily as guidance to provide available information oh the toxicity of various environmental agents and to provide the agency scientists and regulators with initial data to determine what chemicals potentially should take greater priority in being regulated. Although these documents have undergone some peer review within EPA they have not undergone the extensive scientific peer review that is usually necessary to support program office regulations. The more recent 1985 EPA report entitled "Health and Environmental Effects Profile for Chloroethene," which is being used by the State of Mississippi, is a first draft report and has the disclaimer that it is "for review purposes only and does not constitute Agency policy." Discussions with EPA scientists in the Human Health Assessment Group revealed that they are aware of the deficiencies in the approach taken in the 1985 EPA report for calculating a ql* value for vinyl chloride and that additional evaluations of the data base are necessary to derive a more scientifically sound quantitative risk assessment. Discussion The state of Mississippi elected to use a unit risk factor of 2.95X10"1 (mg/kg/day) "r taken from an EPA report "Health and Environmental Effects Profile for Chloroethene" September 1985. From this value they converted the units to obtain a ql* value of 4.21x10-5 ug/m3. This value is not appropriate to use to set ABDOO134494 acceptable ambient exposure levels to vinyl chloride for the following reasons: 1. The conversion from mg/kg/day to ug/m3 was done incorrectly (see above). The correct conversion would reduce the unit risk by a factor of 2.6. 2. The calculation was performed using a ql* value that EPA has not endorsed. This is stated in the disclaimer in the 1985 document. 3. The data used in calculating the ql* value in the 1985 EPA report are not of sufficient quality to determine a ql* value. a. There is no dose-response relationship. b. Results from different experiments were pooled in order to get a dose response relationship without any scientific rationale provided. c. Only certain data points were selected for determining the unit risk without any scientific rationale provided. d. Two of the data points available in the original Maltoni et al. paper were not included in the 1985 EPA report. No scientific rationale was provided. 4. Vinyl chloride is known to require metabolic activation to the proximate carcinogen. These pharmacokinetic considerations were not incorporated into the dose-response data used in determining the unit risk determination in the 1985 EPA report. 5. Many other studies are available than Maltoni et al. to estimate the carcinogenic risk of vinyl chloride. These other studies should be assessed, and pharmacokinetic/biologically based models should be considered to estimate more accurately the human carcinogenic risk of exposure to vinyl chloride gas. conclusions The use of the unit risk presented in the 1985 EPA report entitled "Health and Environmental Effects Profile for Chloroethene11 to establish acceptable ambient levels of vinyl chloride in the air is not scientifically justifiable because of the misuse of the data, lack of critical peer review of the methods employed in the report, and insufficient use of critical data pertinent in considering the carcinogenic risk of vinyl chloride. ABDOO134495 Until a more scientifically sound risk assessment is performed the former ql* value of 2.72xl0"6 (ug/m3)"1 from the 1984 EPA report entitled "Health Effects Assessment for Vinyl Chloride" should be used to estimate cancer risk for vinyl chloride in the air. Although there are also deficiencies in the data on which this ql* value is based, a solid dose-response relationship for total tumors (figure 2) was used to estimate risk. R APPENDIX 1 ABDOO134496 GLOBAL 86 (MAY 1986) BY RICHARD B. HOWE AND CYNTHIA VAN LANDINGHAM CLEMENT ASSOCIATES, 1201 GAINES STREET RUSTON, LA 71270 (318) 255-4800 INC R vinyl chloride: data taken from EPA 1984 report (total tumors) POLYNOMIAL DEGREE SELECTED BY PROGRAM, (POLY-DEGREE=0) CHI-SQUARE TEST USED IN SELECTION GROUP DOSE #RESPONSES OBSERVED/#ANIMALS #RESPONSES PREDICTED 1 .000000 2 4.90000 3 23.9000 4 47.8000 5 239.100 6/ 58 10/ 59 16/ 59 22/ 59 32/ 59 9.81 10.69 13.35 16.50 34.98 CHI-SQUARE GOODNESS OF FIT STATISTIC IS 5.6876 P-VALUE FOR THE CHI-SQ TEST WITH 3 DEGREES OF FREEDOM IS .1278379359 FORM OF PROBABILITY FUNCTION: P(DOSE) = 1 - exp( -Q0 - Q1 * D - Q2 * D*2 ) MAXIMUM LIKELIHOOD ESTIMATES OF DOSE COEFFICIENTS Q( 0) = Q( 1) Q( 2) = .185324445807 2983355597998E-03 .000000000000 MAXIMUM VALUE OF THE LOG-LIKELIHOOD IS -163.144925015 CALCULATIONS ARE BASED UPON EXTRA RISK G**LO**B*A*L**86**U**PP*E*R**C**O*N*F*ID*E*N*C*E***L*IM**IT*S**O*N**R**IS*K**F*O*R**F*I*X*E*D***D*O*SE** DOSE MLE RISK UPPER BOUND ON RISK CONFIDENCE COEFFICIENTS FOR LIMIT SIZE CONFIDENCE LIMIT 1.0000 297891E-03 4.22255E-03 95.0% Q( 0) .15535 Q( 1) = 4.23149E-03 Q( 2) .00000 NORMAL COMPLETION! APPENDIX 2 ABDOO134497 GLOBAL 86 (MAY 1986) BY RICHARD B. HOWE AND CYNTHIA VAN LANDINGHAM CLEMENT ASSOCIATES, 1201 GAINES STREET RUSTON, LA 71270 (318) 255-4800 INC vinyl chloride: data taken from EPA 1985 report (liver angiosarcomas) POLYNOMIAL DEGREE SELECTED BY PROGRAM, (POLY-DEGREE=0) CHI-SQUARE TEST USED IN SELECTION GROUP DOSE #RESPONSES OBSERVED//ANIMALS /RESPONSES PREDICTED 1 .000000 2 .344000 3 .860000 4 1.71900 0/363 1/119 5/120 1/ 60 CHI-SQUARE GOODNESS OF FIT STATISTIC IS .00 1.17 2.92 2.89 2.8312 P-VALUE FOR THE CHI-SQ TEST WITH 2 DEGREES OF FREEDOM IS .2427793524 FORM OF PROBABILITY FUNCTION: P(DOSE) * 1 - exp( -Q0 - Q1 * D - Q2 * DA2 ) MAXIMUM LIKELIHOOD ESTIMATES OF DOSE COEFFICIENTS Q( 0) a Q( 1) ** Q( 2) = .000000000000 2 867931930907E-02 .000000000000 MAXIMUM VALUE OF THE LOG-LIKELIHOOD IS -33.1407703774 CALCULATIONS ARE BASED UPON EXTRA RISK G**LO**B*A*L**86**U**P*P*E*R**C*O*N*F*ID*E*N*C*E**L**IM**IT*S***O*N**R*IS**K**F*O*R**F*IX*E*D***D*O*S*E* DOSE MLE RISK UPPER BOUND ON RISK CONFIDENCE LIMIT SIZE COEFFICIENTS FOR CONFIDENCE LIMIT 1.0000 2.82720E-02 491334E-02 95.0% Q( 0) = .00000 Q( 1) = 503816E-02 Q( 2) = .00000 NORMAL COMPLETION! ABDOO134498 Figure 1 mg/kg/day tumor incidence ABDOO134499 Figure 2 tumor incidence ABDOO134500 Inhalation Risk Assessment An inhalation risk assessment was conducted for off-site exposures to vinyl chloride emissions from the Vista Chemical Company in Aberdeen, Mississippi. The potential inhalation risks were calculated by combining a unit risk factor for vinyl chloride with dispersion modeling results of emissions from the Vista Chemical Company plant. The dispersion modeling was conducted by personnel at the Vista Chemical Company plant and results were provided to Clement International, Inc. for use in the risk assessment. Ambient air concentrations associated with vinyl chloride emissions were determined using the Industrial source Complex Short-Term (ISCST) air dispersion model. ISCST is an EPA recommended model for use in industrial settings such as the Vista Chemical Company plant. The ISCST model was run in the regulatory default mode which automatically selects model options recommended for use by the EPA. Meteorological data from the Montgomery Alabama National Weather Service station, for the years 1966 to 1970 were used in the ISCST model. A separate model run was conducted for each of the five years of meteorological data using a polar grid with 36 radials (10 degrees of arc per radial) and receptors placed at 100, 150, 200, 300, 400, 500, 750 and 1000 meters from the plant along each radial. Emissions from the plant were characterized in the ISCST model by 34 separate emission sources. These sources included both controlled stack emissions modeled as point sources, and fugitive emissions modeled as both volume and area sources. Analysis of the ISCST model results for each of the five years of meteorological data indicated that the 1970 data resulted in the highest annual average concentration for off-site receptors. A second ISCST model run was then conducted using the same emissions inputs and model options with a cartesian grid of receptor points at 200, 400, 600, 800, 1000, 1250, 1500, 1750, 2000, 3000, 3500, 4000, and 5000 meters from the plant in all directions. This resulted in model predictions of annual average concentrations at 676 receptors around the plant. The results from this model run were used by Clement International, Inc. in the inhalation risk assessment. The first step in the inhalation risk assessment was to identify the receptor locations to be evaluated. Two receptor grids were constructed for evaluation. In the first receptor grid model receptors located within the fence line boundaries of the plant were eliminated from the evaluation. In addition, model concentrations for receptors beyond 3500 meters from the plant were also eliminated from, the evaluation because the concentrations would result in negligible inhalation risks ABDOO134501 relative to the closer receptor locations evaluated and in the majority of cases they were beyond the populated areas around the plant. The second grid used in the inhalation risk assessment used the same receptor locations as the first with the exception that receptors located within the plant boundaries but outside of the fence line were excluded. For both of the grids evaluated in the inhalation risk assessment the maximum and average model predicted annual average concentrations were used. The average concentration used in the evaluation was the arithmetic average of all model predicted annual average concentrations for the receptors within each grid evaluated. Note that the use of an arithmetic average assumes an equal weighting of all model predicted concentrations. However, the spacing of the receptor grid increased with distance from the plant, while model predicted concentrations decreased with distance from the plant. Thus the average concentration used in this assessment" is conservative since a regular spacing of the grid out to 3500 meters from the site would have included many more concentrations with lower values than those predicted for receptors located within 1000 meters of the plant, resulting in a lower arithmetic average concentration. The ISCST model results used in the inhalation risk assessment were based on a 200 parts per million (ppm) vinyl chloride content in the slurry. The risk assessment however was conducted for a 150 ppm and 400 ppm vinyl chloride slurry concentration. According to Vista Chemical Company personnel the model results could be scaled for the changes in vinyl chloride slurry content. Vista Chemical Company personnel indicated that emissions from only a portion of the 34 sources evaluated in the ISCST model would be affected by changes in the slurry content. They had configured the ISCST model to calculate receptor concentrations separately for two source groups. One source group represented emissions from the 19 sources (source numbers 1-19) that were affected by slurry content. The second source group represented the 15 remaining sources (source numbers 20-34) whose emissions were unaffected by slurry content. The concentration at a receptor for all modeled vinyl chloride emissions from the plant is then the sum of the concentration at the receptor for each source group. Vista Chemical company personnel provided the following algorithm to convert the 200 ppm slurry results to the desired slurry content: *Xi(x,y) (x*y) *-1? where: yt concentration at a point due to all sources combined, (ug/nr) , Xx s concentration at a point due to source numbers 20 to 34, (ug/m3), ABDOO134502 Xo - concentration at a point due to source numbers 1 to 19. (ug/m3), DSC " the desired concentration of vinyl chloride in the slurry, (ppm). Using the algorithm shown above for a 150 and 400 ppm concentration of vinyl chloride in the slurry with the ISCST model output and unit risk factor for vinyl chloride of 2.72xl0~6 per ug/m3 yielded the following results: For a slurry concentration of 150 ppm vinyl chloride the average annual concentration for receptors located at or beyond the fence line was 0.66 ug/m3 for an inhalation risk of 2xl0"6, For a slurry concentration of 150 ppm vinyl chloride the maximum annual concentration for receptors located at or beyond the fence line was 13.62 ug/m3 for an inhalation risk of 4xl0~5, For a slurry concentration of 400 ppm vinyl chloride the average annual concentration for receptors located at or beyond the fence line was 1.65 ug/m3 for an inhalation risk of 4xl0"6, For a slurry concentration of 400 ppm vinyl chloride the maximum annual concentration for receptors located at or beyond the fence line was 32.41 ug/m3 for an inhalation risk of 9xl0~5, For a slurry concentration of 150 ppm vinyl chloride the average annual concentration for receptors located at or beyond the property line was 0.60 ug/m3 for an inhalation risk of 2xl0~6, For a slurry concentration of 150 ppm vinyl chloride the maximum annual concentration for receptors located at or beyond the property line was 7.34 ug/m3 for an inhalation risk of 2xl0"s, For a slurry concentration of 400 ppm vinyl chloride the average annual concentration for receptors located at or beyond the property line was 1.50 ug/m3 for an inhalation risk of 4xl0~*, For a slurry concentration of 400 ppm vinyl chloride the maximum annual concentration for receptors located at or beyond the property line was 17.73 ug/m3 for an inhalation risk of 5xl0"5. Hodel predictions for average and maximum air concentrations and the associated Cancer risks are summarized in Table 1. ABDOO134503 Table 1 Model Predictions for Average/Maximum Annual Air Concentrations of Vinyl Chloride and Associated Cancer Risks Slurry Concentrations (ppm) 150 150 400 400 Slurry Concentrations (ppm) 150 150 400 400 Average Concentration Fenceline Property line (ug/m3) (ug/m3) 0.66 0.60 1.65 1.50 Risk 2xl0` 2X10' 4x10' 4X10 Maximum Concentration Fenceline Property line (ug/m3) (ug/m3) Risk 13.62 4X10' 7.34 2X10 32.41 9X10' 17.73 5x10 R