a benchmark concentration for carbon disulfide: analysis of the niosh carbon disulfide exposure...

REGULATORY TOXICOLOGY AND PHARMACOLOGY 24, 171–176 (1996)ARTICLE NO. 0122

A Benchmark Concentration for Carbon Disulfide: Analysis of theNIOSH Carbon Disulfide Exposure Database

BERTRAM PRICE,* TED BERNER,† RICHARD T. HENRICH,‡ JOHN M. STEWART,§ AND ELIZABETH J. MORANØ

*Price Associates, Inc., Washington DC; †ENVIRON Corporation, Arlington, Virginia; ‡Akzo Nobel Chemicals, Inc., Dobbs Ferry,New York; §Courtaulds Fibers, Inc., Axis, Alabama; and ØChemical Manufacturers Association, Arlington, Virginia

Received June 17, 1996

ral nerves. For the ulnar and peroneal nerves, the anal-ysis was based on measurements of maximum motorA statistical analysis of the NIOSH (National Insti-nerve conduction velocity (MCV), distal latency, andtute for Occupational Safety and Health) carbon disul-

fide (CS2) exposure database was conducted for pur- amplitude ratio. For the sural nerve, measurements ofposes of establishing a benchmark concentration sensory conduction velocity (SCV), distal latency, and(BMC) for CS2. The analysis addressed the effects of discrete amplitude ratio were analyzed. ConcerningCS2 exposure on the peripheral nervous system and on CS2 exposure effects on IHD risk factors, the responseischemic heart disease risk factors. The BMC is based variables analyzed were total serum cholesterol, low-on models relating response to exposure determined and high-density lipoprotein cholesterol (LDL andfrom statistical analysis of the continuous exposure HDL), triglyceride, fasting glucose concentration, anddata for individuals recorded in the NIOSH database. systolic and diastolic blood pressure.The results demonstrate that changes in the responses The NIOSH CS2 exposure data were analyzed pre-associated with increases in CS2 exposure at levels rep- viously by Johnson (1983) for PNS effects and by Ege-resented in the NIOSH database are relatively small land (1992) for effects on IHD risk factors. In these twoafter adjustment for confounders. The only response analyses, workers were classified into four groups—avariables that had statistically significant relation- control group (unexposed), and low-, medium-, andships with CS2 were the peroneal nerve MCV (motor

high-exposure groups. The current analysis addressesconduction velocity) and the peroneal nerve ampli-PNS endpoints and IHD risk factors, all potential con-tude ratio. Based on these results, BMCs of 16.2 andfounders identified by Johnson (1983) and Egeland18.5 ppm were derived for MCV and amplitude ratio,(1992), and exposure represented by the continuous ex-respectively. q 1996 Academic Press, Inc.posure data reported for individual workers in theNIOSH database. The exposure data consist of meanconcentrations for 25 job categories. These mean jobINTRODUCTIONcategory exposure concentrations are 0 parts per mil-lion (ppm) for workers in the control group and rangeA statistical analysis of the NIOSH carbon disulfide from 0.2 ppm to 15.4 ppm for workers exposed to CS2.(CS2) exposure database was conducted for purposes of

establishing a benchmark concentration (BMC) for METHODSCS2. A BMC is the exposure concentration of a chemi-cal, CS2 in the present analysis, corresponding to a The process leading to a BMC is inherently statisti-

cal (USEPA, 1995a; Crump, 1995; Kodell, 1995). Forprescribed increase in a health effect endpoint relativeto a background level. The BMC methodology has been the NIOSH CS2 exposure data, the process involved

the following three steps.applied by the United States Environmental ProtectionAgency (USEPA) for establishing reference concentra- 1. Develop a statistical relationship between re-

sponse, exposure, and confounders. Proceed if the rela-tions and is being considered by the OccupationalSafety and Health Administration for application in tionship between response and exposure is statistically

significant (P õ 0.05).developing permissible exposure limits (USEPA,1995a; OSHA, 1996). 2. Specify a benchmark response (BMR).

3. Invert the statistical relationship between re-The current analysis addresses the effects of CS2 ex-posure on the peripheral nervous system (PNS) and sponse and exposure to calculate a nominal BMC. The

BMC, then, is the 95% lower confidence limit of theon risk factors for ischemic heart disease (IHD). PNSeffects were evaluated for the ulnar, peroneal, and su- nominal BMC.

171 0273-2300/96 $18.00Copyright q 1996 by Academic Press, Inc.

All rights of reproduction in any form reserved.

AID RTP 1048 / 6e0b$$$141 10-14-96 21:17:38 rtpa AP: Reg Tox Pharm

172 PRICE ET AL.

Statistical Relationship Between Response and CS2 lated using a BMR equal to the response standard devi-ation multiplied by 1.1.Exposure

Regression analysis methods (Chatterjee and Price, Compute BMC1991; Neter et al., 1990) were used to identify and fit

The BMC was computed by inverting the statisticalresponse–exposure relationships to the CS2 data. Sim-relationship between the response variable (adjustedple plots of response versus exposure, partial regres-for confounders) and exposure to solve for the CS2 expo-sion plots, and plots of regression residuals were usedsure level corresponding to the BMR. The resultingto evaluate the impact of confounders that could maskexposure is referred to as the nominal BMC. The 95%the effects of exposure, identify outliers, and developlower confidence limit for the nominal BMC is thean estimate of the relationship between response andBMC. (Calculation of exact 95% confidence limits isexposure adequate for computing a BMC.described in Scheffe (1963).)

Benchmark ResponseDATA

The BMR is a specified change in mean responsefrom the response level in the control group (exposure PNS effects were evaluated for the ulnar, peroneal,equal to 0) that is used to determine the BMC. The and sural nerves. For the ulnar and peroneal nerves,BMR is quantified in the current analysis by setting the analysis was based on measurements of maximuman upper bound on additional risk. Additional risk is motor nerve conduction velocity (MCV), distal latency,the difference between the probability of an abnormal and amplitude ratio. For the sural nerve, the analysisresponse for workers exposed at the BMC and the prob- was based on measurements of sensory conduction ve-ability of an abnormal response for workers in the con- locity (SCV), distal latency, and discrete amplitude ra-trol group (Crump, 1995; Kodell, 1995). The upper tio. MCV for the ulnar and peroneal nerve and SCV forbound on additional risk in the current analysis was the sural nerve were adjusted for temperature to reflectset at 10% (0.10). Stated another way, if P(C ) repre- standard experimental reference conditions (Johnson,sents the probability of an abnormal response at con- 1983).centration C after adjustment for confounders, then The BMC analysis was designed to use the same dataP(0), the probability of an abnormal response in the used by Johnson. In selecting these data from the fullcontrol population after adjustment for confounders, NIOSH database, Johnson’s exclusion criteria were ap-was set to 0.01, and the difference or additional risk, plied. Therefore, workers with high blood lead levels,P(BMC) 0 P(0), was set to 0.10. high glucose levels, diabetes, and high alcohol con-

Abnormal was defined as follows. For each response sumption were excluded from the analysis. Other vari-variable exhibiting a statistically significant positive ables (confounders) that could mask exposure effects ifcorrelation with exposure, the 99th percentile of the not taken into account were included in the regressionresponse distribution in the control group (after adjust- analysis. These are age, height, weight, and race. Anment for confounders) was used to differentiate abnor- additional variable was formed to capture potential in-mal from normal responses. Responses larger than the teraction effects between exposure and age.99th percentile were considered to be abnormal. Where The analysis of IHD risk factors was based on mea-the response variable exhibits a negative correlation surements of total serum cholesterol, low- and high-with increasing exposure, abnormal was differentiated density lipoprotein cholesterol, triglyceride, fastingfrom normal by the 1st percentile of the response distri- glucose concentration, and systolic and diastolic bloodbution (after adjustment for confounders) in the control pressure. Confounders for the serum lipids were age,group. weight, height, education, smoking, alcohol consump-

For response data that follow a normal distribution tion, and race. For the blood pressure analyses, the listand are linearly related to exposure, the BMR corre- of confounders was expanded to include blood lead levelsponding to the additional risk characterization de- and pulse rate. The additional variable introduced inscribed above is a multiple of the response standard the PNS analysis that captures potential interactiondeviation (Crump, 1995). When the probability of an effects between age and exposure was included.abnormal response is 0.01 and additional risk is set at0.10, the multiple is 1.1 (Crump, 1995). The response– RESULTSexposure relationships that were estimated from theNIOSH CS2 exposure data are linear, the responses Statistical Analysis: PNS Endpointsfollow the normal distribution, and the response stan-dard deviation is constant across exposure levels. Table 1, a summary of the regression results for PNS

endpoints, displays multiple correlation coefficientsTherefore, the BMC for each response having a statisti-cally significant relationship with exposure was calcu- (R2s) and statistically significant confounders for each


173A BENCHMARK CONCENTRATION FOR CARBON DISULFIDE

TABLE 1 cally significant. The database includes complete dataon MCV and its predictors for 325 white and 19 non-Regression Results: PNS Endpointswhite workers. The average MCV after adjustment for

Statistically confounders was higher for nonwhites than for whites,significant Adjusted which explains the statistical significance of the raceEndpoint predictors R2

indicator. The MCV–exposure relationship was inves-Ulnar tigated further for nonwhites and whites separately.

MCV Age 0.05 The regression result for whites is virtually identicalDistal latency Age 0.05 to the result for all workers. Neither exposure nor age

Heightare statistically significant in the regression for non-Amplitude ratio Age 0.08white workers (Table 2).Height

Peroneal A BMC was computed from the MCV regression re-MCV Age 0.21 sults for all workers and for whites only. The BMR for

Height all workers is 4.875 m/s (1.1 times 4.432, the responseExposurestandard deviation estimated from the regression fit)Distal latency Height 0.04and the BMR for whites is 4.827 m/s. For the group ofWeight

Amplitude ratio Age 0.09 all workers, the nominal BMC is 22.6 ppm correspond-Height ing to an MCV equal to 40.1 m/s, which is the mean ofExposure MCV measurements in the control group, adjusted forSural

age and height, minus the BMR. The BMC is 16.2 ppm.SCV Height 0.22Weight For white workers, the nominal BMC is 23.2 ppm, cor-Exposure/Age responding to an MCV equal to 40.2 m/s. The BMC for

Distal latency Age 0.23 white workers also is 16.1 ppm.Height Regression analysis conducted for the peronealAmplitude Age 0.07

nerve amplitude ratio also indicated a statisticallyHeightWeight significant relationship with CS2 exposure. Age is the

only statistically significant confounder (Table 2). TheNote. MCV, Motor Conduction Velocity; SCV, Sensory Conduction corresponding BMR is 0.216 m/s and the associated

Velocity; R2, multiple correlation coefficient.BMC is 18.5 ppm.

The sural nerve. Regression analysis results for thesural nerve did not show a statistically significant rela-endpoint. The results for each endpoint are discussed tionship with the CS2 exposure variable for any of theindividually below. sural response measures. However, an indirect expo-sure relationship was indicated for the SCV response.The peroneal nerve. Regression analysis results for

the peroneal nerve showed a statistically significant SCV regression results indicated that height, weight,and the exposure–age interaction variable were statis-relationship to exposure for response variables MCV

and amplitude ratio. The regression equation for MCV tically significant (Table 3). These results suggest thatany reduction in SCV due to CS2 exposure is indirect,indicated a statistically significant relationship with

CS2 exposure as well as confounders age and height occurring through the age–exposure interaction. Theinteraction means that the rate of change in SCV with(Table 2). The race indicator variable also was statisti-

TABLE 2Peroneal Regression Summary for Carbon Disulfide Exposure

Regression coefficients (linear)

n Age Height Exposure Race Adj. R2 BMR BMC

MCVAll data 344 00.173* 00.613* 00.215* 3.368* 0.21 4.875 16.2Whites 325 00.178* 00.580* 00.207* NA 0.20 4.827 16.1Nonwhites 19 00.119 01.353* 00.344 NA 0.21 NA NA

Amplitude ratioAll data 344 00.003* NS 00.008* NS 0.05 0.216 18.5

Note. MCV, Maximum Motor Nerve Conduction Velocity; BMR, Benchmark Response; BMC, Benchmark Concentration; NA, Not Applica-ble; NS, Not Statistically Significant, therefore excluded from model; R2, multiple correlation coefficient.

* P value õ0.05 for t test to determine whether coefficient is significantly different from 0.


174 PRICE ET AL.

TABLE 3Sural Sensory Conduction Velocity (SCV) Regression Summary for Carbon Disulfide Exposure

Regression coefficients (linear)

Exposure Age exposuren Height Weight Age Exposure indicator interaction Adj. R2 BMC

All data 334 00.398* 00.018* 00.028 0.022 3.351* 00.135* 0.20 NAExposed 131 00.420* 0.011 00.164* 00.001 NA NA 0.26 NAControl 203 00.384* 0.022* 00.030 NA NA NA 0.06 NAControl (excluding 6 points) 197 00.409* 0.020* 00.061* NA NA NA 0.11 NA

Note. BMC, Benchmark Concentration; NA, Not Applicable; R2, Multiple Correlation Coefficient.* P value õ0.05 for t test to determine whether coefficients are significantly different from 0.

increasing age is different for exposed workers than for founders are listed in Table 4 along with the regressionR2 for each risk factor. Age, height, and weight areunexposed workers. A higher rate for exposed workers

would be interpreted as an indirect CS2 exposure effect. important confounders for most of the risk factors.The regression analysis for triglycerides indicated aThe regression coefficient measuring the reduction

in SCV with age was larger for the exposed group data statistically significant age–exposure interaction. Re-gression analysis applied to data for the exposed groupthan for the control group data. To better understand

this result, the relationship between SCV and age was of workers showed height, weight, and smoking as sta-tistically significant confounders, but age was not ainvestigated separately for the control group and the

exposure group. Initial regression results showed a sta- statistically significant confounder. Age was, however,tistically significant relationship between SCV and agein the exposed group, but no statistically significant

TABLE 4relationship in the control group. Partial regressionStatistically Significant Explanatory Variablesplots revealed seven (of 203) data points in the control

for Ischemic Heart Disease Risk Factorsgroup that could potentially distort the relationshipbetween SCV and age. After removing these points, Statistically significant Adjustedage, as would be expected, was a statistically signifi- Risk factor predictors R2

cant predictor for SCV in the control group (Table 2).Total serum cholesterol Height 0.15The rate of SCV reduction with age, as measured by

Weightthe regression coefficient for age, was greater in the Ageexposed group than in the control group (P õ 0.01). Race/exposure interaction

Low-density lipoprotein Weight 0.12Workers in the exposed group were older than those incholesterol Agethe control group, which explains part, but not all, of

Smokingthe difference in the age regression coefficients. Be-Race/exposure interactioncause of the absence of a statistically significant direct High-density lipoprotein Weight 0.17

relationship between CS2 exposure and SCV, a BMC cholesterol AgeEducationwas not calculated from the sural SCV data.Alcohol consumptionSmokingThe ulnar nerve. CS2 exposure was not statisticallyRace/exposure interactionsignificant in any of the analyses involving the ulnar

Triglyceride Height 0.06nerve. Age was a statistically significant predictor in Weightthe analysis of ulnar MCV. Age and height were statis- Agetically significant predictors of ulnar amplitude ratio, Age/exposure interaction

Fasting glucose Smoking 0.02and of ulnar latency (Table 1). Because of the absenceconcentrationsof a significant statistical relationship between CS2 ex-

Systolic blood pressure Weight 0.16posure and the ulnar response measures, a BMC was Pulse ratenot calculated from these data. Blood lead level

AgeAlcohol consumptionStatistical Analysis: IHD Risk Factors

Diastolic blood pressure Weight 0.15AgeNone of the IHD risk factors showed a statistically Smoking

significant relationship to CS2 exposure after adjust-Note. R2, multiple correlation coefficient.ment for confounders. The statistically significant con-


175A BENCHMARK CONCENTRATION FOR CARBON DISULFIDE

TABLE 5Summary of Benchmark Concentrations Derived from NIOSH CS2 Database

BMCSource Descriptiona Confounders BMRb (ppm)c

EPA (1995) MCV; grouped data; polynomial No Relative change Å 10% 11.8EPA (1995) MCV; grouped data; Weibull No Relative change Å 10% 11.8Price and Berner (1995) MCV; interval data; linear regression Yes Relative change Å 10% 20.0EPA (1995) (Setzer) MCV; interval data; cumulative exposure; linear regression Yes Relative change Å 10% 17.9Current analysis MCV; interval data; linear regression Yes Additional risk Å 10% 16.2Current analysis Amplitude ratio; interval data; linear regression Yes Additional risk Å 10% 18.5

a MCV, Motor Conduction Velocity.b BMR, Benchmark Response—fifth percentile of responses in the control group.c BMC, Benchmark Concentration.

statistically significant in the analysis of the unexposed nificant relationship between LDL and exposure. Theregression for nonwhite workers would show an expo-worker data. The age–exposure interaction is not,

therefore, an indication of an indirect exposure effect sure relationship, but only because the LDL average inthe unexposed group is uncharacteristically low. Whenfor triglycerides.

The regression analyses for HDL and LDL each indi- limited to workers in the exposed group, the relation-ship between LDL and CS2 exposure for nonwhitecate a statistically significant difference between racial

groups. The NIOSH database contains complete data workers is not statistically significant (i.e., LDL doesnot increase with increasing exposure).on HDL, LDL, and confounders for 17 nonwhite male

workers and 334 white male workers. An analysis byracial group, which is summarized in the following

DISCUSSIONparagraphs, demonstrates that the suggested race–ex-posure interaction in the initial regressions is not real,but an artifact of the data due to an imbalance in the BMCs previously were developed for CS2 based on

statistical analysis of the PNS measurements con-distribution of HDL and LDL measurements betweenwhite and nonwhite worker groups and the small num- tained in the NIOSH CS2 exposure database (USEPA,

1995b; Price and Berner, 1995; Setzer, 1995), the sameber of observations in the nonwhite worker groups.For HDL, the average for the nonwhite group is database employed in the current analysis. USEPA ini-

tially used the NIOSH data summarized into four expo-larger than the average for the white group (P Å 0.02).This difference is the basis for the statistically signifi- sure groups—unexposed, and low-, medium-, and high-

exposure groups. USEPA applied Polynomial and Wei-cant race variable in the initial regression. Also, for thenonwhite group, average HDL for those exposed to CS2 bull models (USEPA, 1995a). Both models yielded

BMCs equal to 11.8 ppm (See Table 5). Price andis lower than the average HDL for those not exposed.For the white group, average HDL levels for those ex- Berner (1995) and Setzer (1995) both used continuous

exposure data and adjusted for confounders in theirposed and those not exposed are statistically equal.Neither the regression for the white group nor the re- analyses. Setzer’s analysis included a measure of cu-

mulative CS2 exposure and length of employment.gression for the nonwhite group indicates a statisticallysignificant CS2 exposure relationship. Setzer found cumulative exposure to be statistically

significant for the peroneal nerve MCV response.For LDL, the CS2 exposure variable is statisticallysignificant, but the race variable is also statistically Setzer used a BMR equal to a 10% reduction in the

average peroneal MCV for the unexposed group andsignificant, suggesting a difference in the relationshipbetween LDL and CS2 exposure in the two racial calculated a BMC of 17.9 ppm based on cumulative

exposure divided by average length of employment.groups. The data, however, display an aberrant patternof LDL levels across racial and exposure groups. The Price and Berner (1995) used contemporaneous expo-

sure measurements and limited the confounders to age,average of LDL measurements for unexposed nonwhiteworkers is substantially less than the average LDL the only confounder found to be statistically significant

in the statistical analysis of PNS responses reportedfor unexposed white workers, 94.4 versus 119.1 mmol/liter, respectively. A difference of this magnitude would by Johnson (1983). Price and Berner (1995) computed

a BMC for the peroneal nerve MCV, the only responsenot be expected in an unexposed group. For exposedworkers, average LDL is 147.5 mmol/liter for nonwhite with a statistically significant relationship to expo-

sure, equal to 20 ppm. This BMC value was based onworkers and 125.7 mmol/liter for white workers. Theregression for white workers shows no statistically sig- a BMR equal to a 10% reduction in the peroneal MCV


176 PRICE ET AL.

average in the unexposed group after adjustment for cal relationships with CS2 exposure at levels below 15.4confounders. ppm, the maximum exposure level recorded in the

In the current analysis of PNS endpoints, all poten- NIOSH CS2 exposure database. Two PNS endpoints,tial confounders recorded in the NIOSH database were MCV and amplitude ratio, showed a statistically sig-included. BMCs were calculated for the peroneal nerve nificant trend with CS2 exposure. Both trends changeMCV and amplitude ratio, the only PNS endpoints ex- slowly with increasing exposure. The BMCs for thesehibiting statistically significant relationships with CS2 endpoints, 16.2 and 18.5 ppm, respectively, are largerexposure. The ‘‘additional risk’’ BMR was employed, than the maximum exposure represented in thewhich defines the nominal BMC as the CS2 concentra- NIOSH database. Health impairment associated withtion corresponding to a 10% additional risk of an abnor- CS2 exposure, therefore, is unlikely at exposure levelsmal response (Crump, 1995). Abnormal was defined as below 15.4 ppm.the first percentile point of the distribution of responsesin the unexposed group after adjustment for confound-

ACKNOWLEDGMENTSers. The BMCs for MCV and amplitude ratio are 16.2and 18.5 ppm, respectively.

The research for this article was sponsored by the Carbon DisulfideIt is notable that the three PNS analyses based on Panel of the Chemical Manufacturers Association. The authors thankcontinuous exposure data, although using different the following companies for their support: Akzo Nobel Chemicals,confounders, measures of exposure, and statistical Inc., Courtaulds Fibers, Inc., Flexel, Inc., Lenzing, North American

Rayon Corporation, PPG Industries, Inc., Teepak, Inc., and Viskasemodels, converged on a BMC around 20 ppm. It alsoCorporation.is notable that extrapolation of the statistical models

beyond 15.4 ppm, the largest exposure measurement inthe NIOSH database, was necessary to find the BMC. REFERENCESTherefore, even where the PNS response–exposure re-lationship is statistically significant, it is relatively flat Chatterjee, S., and Price, B. (1991). Regression Analysis by Example,

2nd ed. Wiley, New York.and the PNS effects of CS2 exposure are negligible forCrump, K. S. (1995). Calculation of benchmark doses from continu-the range of exposures analyzed.

ous data. Risk Anal. 15, 79–89.In the current analysis, none of the IHD risk factorsdeJesus, P. V., Hausmanowa-Petrusewicz, I., and Barchi, R. L.have a statistically significant relationship with CS2

(1973). The effect of cold on nerve conduction of human slow andexposure. A previous analysis of the NIOSH data (Ege-fast nerve fibers. Neurology 23, 1182–1187.land, 1992) reported statistically significant CS2 effects

Egeland, G. M., et al. (1992). Effects of exposure to carbon disulphidefor diastolic blood pressure and LDL. In the currenton low density lipoprotein cholesterol concentration and diastolicanalysis with all confounders considered, CS2 exposure blood pressure. Br. J. Ind. Med. 49, 287–293.

was not a statistically significant predictor of diastolic Fajen, J., Albright, B., and Leffingwell, S. S. (1981). A cross-sectionalblood pressure (P Å 0.08). In addition, Egeland’s statis- medical and industrial hygiene survey of workers exposed to car-tically significant result for LDL is a consequence of bon disulfide. Scand. J. Work Environ. Health 7, 20–27.the distribution of LDL measurements for 17 nonwhite Johnson, B. L., et al. (1983). Effects of the peripheral nervous system

of workers’ exposure to carbon disulfide. Neurotoxicology 4, 53–workers. LDL measurements for nonwhite workers in65.the unexposed group are lower than expected in com-

Kodell, R. L., et al. (1995). Neurotoxicity modeling for risk assess-parison to LDL measurements for unexposed whitement. Regul. Toxicol. Pharmacol. 22, 24–29.workers. There is no statistically significant relation-

Neter, J., Wasserman, W., and Kurner, M. H. (1990). Applied Linearship between LDL and exposure for white workers (nStatistical Models, 3rd ed., Irwin, Burr Ridge, IL.Å 334) and no increasing trend in LDL with exposure

OSHA (1996). 61 FR 1947 January 24, 1996.for nonwhite workers in the exposed group. The statis-Price, B., and Berner, T. (1995). A benchmark dose for carbon disul-tically significant LDL exposure relationship that

fide: Analysis of nerve conduction velocity measurements from theshows up when regression analysis is applied to dataNIOSH exposure database. Prepared for the Carbon Disulfide

for all workers should be viewed as an artifact resulting Panel of the Chemical Manufactures Association.from a small number of low LDL measurements in the Scheffe’, H. (1963). The Analysis of Variance, p. 53. Wiley, New York.unexposed nonwhite worker group (an insufficient

Setzer, R. W. (1995). Memo from R. Woodrow Setzer to Dan Guth,sample size). Carbon disulfide BD analysis. USEPA Health Effects Research

Laboratory, Research Triangle Park, NC.CONCLUSIONS USEPA (1995a). The Use of the Benchmark Dose Approach in Health

Risk Assessment. EPA/630/R-94/007, USEPA, Washington, DC.Neither the PNS endpoints nor the IHD risk factors USEPA (1995b). Integrated Risk Information System. Office of

Research and Development, USEPA, Washington, DC.evaluated in the current analysis have notable statisti-


a benchmark concentration for carbon disulfide: analysis of the niosh carbon disulfide exposure...

Documents