Leukemia and cumulative exposure to butadiene, styrene and benzene among workers in the synthetic rubber industry

Toxicology 113 (1996) 190-202

Leukemia and cumulative exposure to butadiene, styrene and benzene among workers in the synthetic rubber industry Maurizio Macaluso*a, Rodney Larson b, Elizabeth Delzell”, Nalini Sathiakumar”, Mary Hovingaa, Jim Julian”, David Muir”, Philip Cole” “Department OfEpiakmio~ogy, School of Public Health, University of Alabama at Birmingham, Birmingham, AL, USA bDepartment ofEnvironmental Engineering, Texas A & M University KingsviNe, Kingsville, TX, USA ‘Occupational Health Program, MeMaster University, Hamilton, Ontario, Canada

Abstract

Retrospective, quantitative estimates of exposure to 1,3-butadiene, styrene and benzene were developed for a follow-up study of leukemia mortality among 16610 subjects employed at six North American styrene-butadiene rubber manufacturing plants (418 846 person-years, 58 leukemia deaths). The estimation procedure entailed identifying work areas within each manufacturing process, historical changes in exposure potential and specific tasks involving exposure, and using mathematical models to calculate job- and time-period-specific average exposures. The resulting estimates were linked with the subjects’ work histories to obtain cumulative exposure estimates, which were employed in stratified and Poisson regression analyses of mortality rates. Mantel-Haenszel rate ratios adjusted by race, age, and cumulative styrene exposure increase with cumulative butadiene exposure from 1 in the nonexposed category to 4.5 in the category of 80 ppm-years or more (P = 0.01). The risk pattern is less clear and statistically nonsignificant for styrene exposure. A trend of increasing risk with butadiene exposure is still present after exclusion of the nonexposed category (P = 0.03). A parsimonious interpretation of the findings presented here, in light of previous epidemiologic studies, is that exposure to butadiene in the synthetic rubber industry produces a dose-related

increase

Keywords:

in the occurrence

Leukemia; Occupational Follow-up studies

of leukemia.

exposure; 1,3-Butadiene; Styrene; Synthetic rubber; Epidemiologic

1. Introduction

processes.

1,3-Butadiene (BD) and styrene (STY) are widely used in the manufacture of styrene-butadiene rubber (SBR) and in other industrial *Corresponding author, 206 BBRB, UAB Station, Birmingham, AL 35294-2170, USA. Tel: + 1 205 934 7159; fax: + 1 205 933 5671; e-mail: mmacaluso@,epi.soph.uab.edu. 0300-483X/96/315.00

0

PZZ SO300-483X(96)03444-0

According

to the International

methods;

Agency

for Research on Cancer (IARC), there is sufficient evidence that BD causes hematopoietic and other forms of cancer in animals, and limited evidence of carcinogenicity in humans (IARC, 1992). For STY, there is limited evidence of carcinogenicity in animals and inadequate evidence of carcinogenicity in humans (IARC, 1987a). Benzene (BZ) causes acute myelogenous leukemia in human

1996 Elsevier Ireland Ltd. All rights reserved

M. Macalusoet al.I ToxicologyI1 3 (19%) 190-202

beings (IARC, 1987b), and has been used as a solvent in laboratories and in synthetic rubber manufacturing. The present study expanded and updated previous studies of mortality among men employed at eight North American plants which manufacture SBR and related products (Matanoski and Schwartz, 1987; Matanoski et al., 1990, 1993; Meinhardt et al., 1982). Retrospective, quantitative estimates of exposure to BD, STY and BZ were developed for each worker to increase the power of the study to detect causal associations between these exposures and mortality from leukemia or other neoplasms. This report describes the exposure estimation procedure and presents preliminary results of the analysis of mortality from leukemia by cumulative exposure level.

2. Methods The methods employed to gather information on cohort membership, employment history and vital status of SBR workers are described in detail elsewhere (Delzell et al., this issue). Briefly, the investigation included 17964 men who worked at any one of the eight plants for at least 1 year during the period 1943-1991. Identifying and work history information on subjects was obtained from plant records and from data collected in previous follow-up studies. Vital status as of 1 January 1992, was determined for 96% of the cohort. The cause of death was ascertained for 97% of the decedents. Complete work histories were available for about 97% of the cohort. Work histories were complete but were not sufficiently specific for subjects from two plants (n = 1354), who were excluded from the exposure estimation procedure. Thus, exposure estimation was carried out for 16610 workers. Trained staff abstracted work history data from personnel records. Department and job title combinations (n = 8281) listed in the work histories were grouped into 308 homogeneous “work area groups”. The research team reviewed plant records relevant to exposure estimation, including results of industrial hygiene (IH) monitoring surveys,

191

historical aerial photographs and plot plans, or‘ganizational charts, bargaining agreements, information on protective equipment and safety policies, on employee safety training programs and on engineering projects, and the reports of NIOSH surveys conducted at six plants. The team also conducted in-depth walk-through surveys of the plants and held meetings with key plant management officials, such as the managers of manufacturing operations, quality control laboratories, research and development, pilot plant(s), safety/industrial hygiene and engineering and maintenance. Meetings with management were followed by interviews with workers who had a history of long-term employment in specific work areas and work area groups. These interviews comprised a large part of the historical information collected, and addressed area layout, equipment and material flow, process operations, job titles of workers employed in routine operations or maintenance/cleanup, potential exposure sources and exposure control systems. Efforts were made to interview at least two employees with specific work experience in the same area and time period, to minimize the likelihood that an unreliable interview would have a large influence on exposure estimation. The exposure estimation procedure consisted of four major steps: (1) process analysis described individual manufacturing processes at each plant, work areas within each process and specific operations performed in each area, and identified job titles assigned to carry out operations with exposure potential as well as historical changes that may have affected exposure; (2) job analysis specified component tasks associated with exposure potential, task-specific determinants of exposure (equipment used, duration and frequency of the task, work practices, presence of exposure reduction mechanisms) and historical changes in the job with respect to exposure determinants, and resulted in the compilation of historical job profiles that identify all sources of exposure; (3) exposure estimation entailed the specification of exposure models, the estimation of exposure intensities for specific tasks in different time periods, the computation of job- and time periodspecific summary indices, the estimation of expo-

192

M. Macaluso et al. I Toxicology 113 (19%)

sure intensities for generic work area groups (e.g. “production worker”) in different time periods and resulted in the compilation of job-exposure matrices (JEMs) for BD, STY and BZ; (4) linkage of the JEMs with subjects’ work histories allowed the calculation of cumulative exposure indices for individual workers. The information derived from the interviews with plant management and other long-term employees was summarized in documents that were submitted to the participating plants for review, and reviewers’ comments were used to revise the process and job descriptions. The exposure associated with each job is assumed to derive from an area “background” level, corresponding to the average air concentration of BD, STY and BZ in the work area where the job was performed, and of “task-specific” exposure levels which replaced the background level during the performance of specific duties. For work in buildings, the background level was computed using a dilution ventilation formula (ACGIH, 1992). The determinants of air concentration in this model are the emission rate, which expresses the amount of monomer evaporating in the unit time, and the ventilation rate, which expresses the amount of dilution due to air flowing through the work area. Potential sources of monomer release in the area, including equipment leaks, loading/unloading tasks, sampling tasks, routine maintenance and cleaning tasks, were summed to compute the average emission rate during one work shift. Ventilation rates were estimated from direct observation of the work area, from descriptions obtained through interview, and from available engineering or safety data. Usually, data on building size, surface of openings, and air flow through fans, hoods, and equipment exhaust systems were available. Taskspecific exposure levels were estimated using the same dilution ventilation formula when the exposure did not originate from a point source (e.g. vessel cleaning, inspection of a compressor house). For work in open areas such as the tank farm, a subjective estimate of the background level was obtained to recognize that exposure was present throughout the work shift at levels higher than in offices or outside the plant. Sub-

190-202

jective background estimates were generally very low (often less than 1 ppm) and had a minor influence on the average exposure level of workers who also carried out tasks entailing exposure potential. Exposure originating from a point source, such as minor maintenance of a pump or latex sampling from a reactor, was calculated using a nearfield air dispersion model derived from the Pasquill-Gifford equation (Lipton and Lynch, 1987). In this model, the determinants of concentrations in air were the emission rate, the distance from the source of exposure and the wind velocity across the area. For each task we compiled data on the variables to be used in the appropriate model. Emission rate estimates were based on information gathered in interviews (e.g. the approximate amount of unstripped latex purged plus the sample size for a reactor sample during batch polymerization) and on chemical parameters such as vapor pressure of the chemical and the evaporation area. Distance also was estimated from interviews or from direct observation of the task and the work environment (e.g. 0.5 m for arm length, 2 m for an operator standing on a grid or platform, with emissions occurring from a drainage ditch). Three air velocities were used: 4.4 m/s for work in open air, on elevated structures; 2.2 m/s for work in open air at the ground level; and 1.1 m/s for work in semi-enclosed areas. To compute 8-h time-weighted average (TWA) exposures, each task-specific intensity estimate (in ppm) was multiplied by the total number of minutes of exposure during a work shift to obtain the task-specific number of ppm-minutes of exposure. Next, the remaining part of the 8 h (or less, if appropriate) was multiplied by the estimated area background level to obtain the number of ppm-minutes of background exposure. Finally, the sum of ppm-minutes of exposure was divided by 480 to obtain the 8-h TWA. Estimates of the annual number of exposure peaks (i.e. short time periods in which intensity exceeds a specified level) also were computed but were not included in the present analysis. The exposure levels of certain work area groups including poorly specified jobs, surpervisory jobs, and plant-wide main-

M. Macaluso et 01. I Toxicology 113 (I 996) 190202

tenance jobs were not estimated directly, but were derived from other categories. For example, for polymerization workers not otherwise specified, exposure estimates reflected an average of the estimates computed for the tank farm, pigment, reactor and recovery operators, weighted by the relative frequency of workers employed in each well-specified area. The final plant-specific JEMs consist of summary exposure estimates for specific work area groups and for each year between 1943 and 1992. Each record contains a plant code; a work area group code; a calendar year; the estimated BD, STY and BZ concentrations (in ppm) corresponding to the work area/job code and the year; and the estimated number of annual BD, STY, BZ exposure peaks. The work history files were merged with the JEMs to link each subject’s employment periods with the appropriate work area group- and year-specific exposure estimates. Finally, for each subject exposure indices were multiplied by the length of corresponding employment period, expressed in fractions of a year, and were summed across the subject’s work history to obtain cumulative exposure estimates. The overall and cause-specific mortality experience of the study group was assessed by comparing their rates with general population mortality rates, using the standardized mortality ratio (SMR) as the measure of association. Most analyses compared the cohort’s mortality rates with USA male general population or with Ontario male rates (Delzell et al., this issue). In the analysis of mortality rates by quantitative estimates of exposure to BD, STY and BZ, we computed cumulative ppm-years and peak-years for each person-year of follow-up in the study and for each decedent as of his death date. The dynamic allocation of cumulative exposure to person-years was done using an approximate method, assigning the entire follow up in one calendar year (i.e. one person-year or less) to the value of cumulative exposure and other covariates computed as of the mid-point of the interval. Person-years of follow up and deaths were then grouped according to categories of cumulative exposure (ppm-years) and of important covariates, including age, calendar period, years since

193

hire and race. The person-year data grouped by BD, STY and BZ ppm-years categories were used in standard SMR analyses and in stratified internal comparisons. To ensure comparability between cohort mortality rates and reference (general population) mortality rates, only decedents whose death certificate reported leukemia as the underlying cause of death (n = 51) were included in SMR analyses. Leukemia was reported as contributory cause of death on the death certificate of seven additional decedents. These decedents were included as leukemia cases in analyses using internal comparisons. In the latter, Mantel-Haenszel rate ratios (RR) were computed as measures of association. Ninety-five percent confidence intervals for the RR were computed using the variance estimator proposed by Breslow, and the statistical significance of the association between categorized cumulative exposure and leukemia RR was evaluated using a linear trend chi-square test (Breslow and Day, 1987). We also used Poisson regression models to examine mortality rates within the cohort and to estimate the effect of one exposure variable adjusting for multiple possible confounders (Breslow and Day, 1987).

3. Results A total of 52 in-depth interviews were conducted with plant management, and 164 with longterm employees. Analysis of the interview data resulted in the identification of 446 plant-specific tasks or work area backgrounds entailing exposure to BD, STY or BZ (Table 1). The number of specific tasks varied by plant from a minimum of 50 to a maximum of 164. Task-specific exposure estimates were computed for an average of two historical periods per task, to reflect significant changes in manufacturing process or in safety procedures. Area background and task-specific exposure estimates were employed in the computation of work shift time-weighted exposure averages for a total of 664 plant-specific work area groups (Table 2). The number of work area groups varied by plant from a minimum of 64 to a maximum of 239. Estimates of 8-h TWA expo-

M. Macaluso et al. I ToxicologyII3 (19%) NO-202

194

Table 1 Numbers of specific tasks (including area backgrounds), number of task-time period combinations, exposure estimates and numbers of task-time period combinations with nonzero exposure (by plant) Plant

1 2 4 5 7 8

Tasks

68 50 63 48 53 164

Time periods

129 108 124 96 II6 322

BD exposure (ppm)

STY exposure (ppm)

median and maximum

BZ exposure (ppm)

Median

Max

ppm r 0

Median

Max

ppm > 0

Median

Max

ppm>O

29 10 23 I4 16 9

9000 3292 3924 3670 3292 10000

81 73 83 59 64 169

3.0 2.6 2.7 2.7 3.0 3.0

253 253 253 253 253 253

73 77 82 70 90 I58

2.0 0.5 0 2.3 0 10

188 0.5 0 2.3 0 840

II I 0 2 0 95

Table 2 Work area groups, median and maximum time-weighted average exposure estimates and number of work area group-year combinations with nonzero exposure in the job-exposure matrix (by plant) Plant

1 2 4 5 7 8

Work area groups

84 93 100 83 64 239

BD (8-h TWA, ppm)

STY (8-h TWA, ppm)

Median

Max.

TWA>0

Median

0.7 0.7 1.3 1.7 1.3 1.1

39 64 46 44 59 50

2004 2375 2684 1853 1448 5314

0.6 0.8 0.5 1.1 0.9 0.5

sure for BD varied from zero to 64 ppm, but the median BD TWA among work area groups entailing any BD exposure was below 2 ppm in all plants, indicating that most work area groups were characterized by low average exposure levels. Estimates of TWA exposure to STY were lower, varying from zero to 7.7 ppm. The median STY TWA among work area groups entailing any STY exposure varied by plant from a minimum of 0.5 ppm to a maximum of 1.1 ppm. BZ exposure was present in three of the six plants, the median TWA among BZ-exposed work area groups being below 1 ppm (max: 38 ppm). The JEMs used in the linkage with work histories consisted of 33 178 plant, work area group and calendar year-specific sets of exposure estimates. Linkage of the JEM with the work histories of subjects employed at the six plants provided a historical description of the exposure profile of

Max 7.7 7.7 7.7 5.2 7.7 19

BZ (8-h TWA, ppm) TWA >O

Median

Max

TWA>0

1969 2864 2790 2106 1890 4885

0.09 0.2 0 0 0 0.6

6.6 0.5 0 0 0 38

629 28 0 0 0 2595

the cohort over five decades. Fig. 1 displays the average BD, STY and BZ exposure levels of active SBR workers, by decade. As was perceived by management and long term employees, and partially documented by limited IH data available for the last two decades, exposure levels show a downward trend. Average BD exposure estimates of active workers decreased from 4.2 ppm in the 1940s to 0.4 ppm in the early 1990s. During the same period, average STY exposure estimates decreased from 1.8 ppm to 0.1 ppm, and average BZ exposure estimates decreased from 0.2 ppm to 0.1 ppm. These figures reflect in part a reduction in the prevalence of exposure with time. Fig. 2 displays the historical trend of average exposure estimates among active workers with nonzero exposure. Average BD exposure among exposed workers were 5.9 ppm in the 1940s and decreased to 0.8 ppm in the

M. Macaluso et al. I Toxicology113 (19%) 190-202

195

6

5

n ED @pm) n SW (ppm)

l~~(ppm) I

1

0 1940s

1950s

1960s

1970s

1980s

1990+

Fig. 1. Estimates of average BD, STY and BZ exposure among all employeesat six plants, by decade.

early 1990s. During the same period, average STY exposure estimates among the exposed decreased from 2.2 ppm to 0.2 ppm. Average BZ exposure estimates do not show a clear trend, indicating that the decrease observed in Fig. 1 is mostly due to a reduction in BZ exposure prevalence. A large proportion of the cohort experienced exposure to BD or STY during employment in the SBR industry (Table 3). As of the end of follow up, 75% of the cohort members had been exposed to BD, the median cumulative exposure being 11.2 ppm-years. Eighty-three percent of the cohort had been exposed to STY (median cumulative exposure: 7.4 ppm-years). BZ exposure prevalence was 25% (median cumulative exposure: 2.9 ppm-years). Decedents from all causes combined showed a similar pattern of exposure prevalence, although their median BD and STY exposure levels were higher. Leukemia decedents

had both higher exposure prevalence rates and higher median cumulative exposure levels among the exposed. Eighty-six percent of the leukemia decedents had some BD exposure, and their median cumulative exposure was 36.4 ppm-years, about three times as high as other employees evaluated at the end of the follow up period and about twice as high as all decedents combined. Ninety percent of the leukemia decedents had some STY exposure, and their median cumulative exposure was 22.4 ppm-years, also about three times as high as all SBR workers and twice as high as all decedents combined. BZ exposure was less frequent among leukemia decedents than among other employees, although the median cumulative exposure was about twice as high (5.5 ppm-years). Further analyses of mortality rates by categories of cumulative exposure indicated the absence of an association between leukemia and BZ exposure. Increasing levels of cumulative

M. MacaiusoetaLIToxicolbgy113 (19%)1%-202

196

1940s

1950s

1960s

1970s

1960s

1990+

Fig. 2. Estimates of average BD, STY and BZ exposure among exposed employees at six plants, by decade.

Table 3 Numbers of subjects exposed, exposure prevalence (P) and median cumulative exposure to BD, STY and BZ (in ppm-years) among all study subjects, all decedents and leukemia decedents Group

All study subjects All decedents All leukemia decedents

BD exposure

STY exposure

BZ exposure

n

P (%)

Median (ppm-y)

n

P (%)

Median (ppm-y)

n

P (%)

Median (ppm-y)

12412 3271 50

75 77 86

11.2 19.0 36.4

13825 3526 52

83 83 90

7.4 9.7 22.4

4202 975 11

25 23 19

2.9 3.2 5.5

exposure to BZ were weakly associated with leukemia mortality rates, and the association was nil after controlling for exposure to BD and STY. Thus, BZ exposure was excluded from the additional analyses.

Leukemia SMRs increased with cumulative BD and STY exposure categories (Table 4). The leukemia SMR was 264 in the BD exposure category of 80 ppm-years or more, indicating a 164% excess mortality over the general popula-

hf. Macaluso et al. I Toxicology 113 (19%) NO-202

197

Table 4 Person-years, leukemia decedents, SMRs and mortality rate ratios by cumulative exposure to BD and STY (six plants) Cumulative BD exposure (ppm-years)

Person-years Decedents’ SMR” RR’*’

0


1-19

20-79

so+

x2

P

102900 8 76 1

100992 4 41 2.0

90 807 12 133 2.1

82885 16 166 2.4

41261 18 264 4.5

6.4

0.01

Cumulative STY exposure (ppm-years)

Person-years Decedents’ SMRb RR’,*

0

<5

5-9

10-39

40+

x2

P

67 453 6 89 1

139936 7 63 0.9

65094 11 161 5.4

101787 16 136 3.4

44576 18 235 2.7

2.2

0.14

‘Includes seven decedents for whom leukemia was listed only as a contributory cause of death. bExcludes seven decedents for whom leukemia was listed only as a contributory cause of death. ‘Mantel-Haenszel rate ratio adjusted by race, age and cumulative exposure to STY. dMantel-Haenszel rate ratio adjusted by race, age and cumulative exposure to BD.

tion rates. In contrast, the SMR was 76 for subjects with zero cumulative BD exposure, indicating a 24% deficit as compared with the general population rates. Mantel-Haenszel mortality RRs adjusted by race, age, and cumulative STY exposure showed a progression from 1 in the nonexposcd category to 4.5 in the category of 80 ppm-years or more. The linear trend test was statistically significant (P = 0.01). The pattern of risk progression was less clear and statistically nonsignificant for STY exposure. The SMRs increased with increasing cumulative STY exposure, but Mantel-Haenszel RR estimates adjusted by cumulative BD exposure showed an inconsistent pattern of association, with the highest risk being associated with the 5-9 ppm-years category, and decreasing with higher exposure levels. Further stratified analyses were conducted to assess whether the association with BD exposure was present after exclusion of the nonexposed category, and to assess the presence of interaction with cumulative STY exposure. Subjects with

zero cumulative exposure include salaried employees, and this category may be unsuitable for comparisons with blue collar workers, even after adjustment by age and race. Also, a large part of the association may be related to the difference between the nonexposed and all of the exposed categories combined, rather than to a progression of risk with exposure level. To clarify these issues, Mantel-Haenszel RRs for categories of increasing cumulative exposure were computed after exclusion of all subjects who had zero cumulative exposure to either BD or STY (Table 5). After adjustment by race, age and cumulative STY exposure, the cumulative BD exposure category of 20-79 ppm-years showed a 50% increase in leukemia mortality exposure over the category of O.l- 19 ppm-years, and the category of 80 ppm or more showed a 70% excess. The linear trend test was still significant after exclusion of the nonexposed (P = 0.03) confirming the presence of a progression in risk with increasing exposure level. The data suggested an effect of cumulative STY exposure even after adjustment

M. Macaluso et al. I Toxicology I13 (19%)

198

NO-202

Table 5 Leukemia rate ratios’ and 95% confidence intervals, by cumulative exposure to BD and STY among subjects exposed to both monomers (six plants) BD (ppm-years)

STY (ppm-years) 0.1-9

OveralP to-39

40+

0.1-19 20-79 80+

1 3.5 (1.1-11) 5.1 (1.3-21)

1.7 (0.5-6.0) 2.3 (0.7-7.3) 4.9 (1.6-15)

7.0 (2.2-22) 2.7 (0.7- 11) 5.7 (2.0- 16)

Overall’

1

0.9 (0.4-2.0)

1.6 (0.6-3.7)

1 1.5 (0.7-3.2) 1.7 (0.8-3.9)

‘Mantel-Haenszel rate ratios adjusted by race, age, BD or STY sx2 = 4.7, P = 0.03. ‘x2 = 2.8, P = 0.09.

by cumulative BD exposure, although the risk progression was less convincing and was statistically nonsignificant. Table 5 also displays Mantel-Haenszel RR estimates for combinations of cumulative BD and STY exposure categories. Under the additive risk model, RRi,j = 1 + (RRic - 1) + (RRoj - 1). Thus, for example, the RR corresponding to the group with 20-79 ppm-year of BD exposure and lo-39 ppm-years of STY exposure is expected to equal the sum of three terms: (a) the baseline of 1 represents the RR of the reference group (i.e. 0.1-19 ppm-years of BD exposure and 0.1-9 ppm-years of STY exposure); (b) the “pure” effect of BD exposure, as estimated by the excess RR in the group with 20-79 ppm-years of BD exposure and 0.1-9 ppm-years of STY exposure (i.e. 3.5 1 = 2.5); and (c) the “pure” effect of STY exposure, as estimated by the excess RR in the group with lo-39 ppm-years of STY exposure and 0.1-19 ppm-years of BD exposure (i.e. 1.71 = 0.7). Thus, the expected RR in the group was 4.2, whereas a Mantel-Haenszel RR of 2.3 was observed. In general, the pattern of association observed in Table 5 is compatible with a subadditive risk model, i.e. the combined effect of BD and STY exposure appears to be less than the expected. On the other hand, the RR estimates were based on small numbers of leukemia cases in each category, and the imprecision of the estimates was such that no firm conclusions could be reached about the presence of a negative

interaction (antagonism) between monomers. The pattern was further complicated by the considerable correlation between exposure to BD and STY. Spearman’s rank correlation coefficient was 0.5 for BD ppm-years and STY ppm-years among workers evaluated as of the end of follow up. It was not possible to examine the association between BD and leukemia among subjects who were not exposed to STY, as no leukemia death occurred among such subjects. Also, there were only two leukemias among men exposed to STY but not to BD. Additional analyses were conducted by fitting Poisson regression models with a variety of exposure categorization schemes, and with additional potential confounders such as years since hire. These analyses confirmed the presence of an association between BD exposure and mortality from leukemia, and further reduced the credibility of the association between STY exposure and leukemia.

4. Discussion Epidemiologic studies are essential in assessing the relation between occupational exposures and chronic diseases. Such studies are most informative if they incorporate a valid exposure assessment. Exposure measurements are rarely available for many study subjects. Thus, exposure

IU. Macaluso et al. I Toxicology I13 (19%) MO-202

must be estimated using information on subjects’ history of employment by production area, job title, task, duration of employment or combinations of these (Checkoway et al., 1989). Systems for linking exposure estimates with job titles, tasks or production areas can be referred to as JEMs (Rice, 1991). Several alternative strategies for developing a JEM for ED, STY and BZ exposure in the SBR industry were considered for the present study. Development of ordinal or semi-quantitative exposure scores would have been feasible. Plant records and interviews with long-term employees were used for retrospective exposure assessment, and summary information compiled from these sources could have been used by multiple raters to assign exposure scores with an acceptable degree of reproducibility (Macaluso et al., 1993a,b). A subjective, ordinal exposure score system was used in a case-control study of leukemia and other neoplasms among SBR workers (Santos-Burgoa et al., 1992). On the other hand, the process by which independent raters assign subjective exposure scores is difficult to document, and inter-rater agreement, although acceptable, is usually not very good (Hayes et al., 1986; Kromohut et al., 1987; Macaluso et al., 1993b). Finally, quantitative exposure estimates are highly desirable for risk assessment. Statistical analysis of IH data could have been used to generate exposure estimates for specific work area groups (see, for example, Hornung et al., 1994; Samuels et al., 1985). IH monitoring programs started in the late 1970s at most of the SBR plants. Data from the NIOSH industrywide survey also were limited to recent time periods. Francis et al. (this issue) discuss the limitations associated with using the sparse existing IH data to estimate BD exposure among SBR workers. Furthermore, evidence gathered during the study questioned the validity of historical IH data and suggested that technical problems in the sampling, storage and analytic procedures used in the past may have led to substantial underestimation of BD exposure levels. The exposure estimation procedure employed in this study consisted of the identification of the determinants of exposure in each job and time

199

period, and of the specification of a detailed exposure model whose components could individually be examined and validated. The procedure was based on established IH principles that are used for purposes such as ventilation system design (ACGIH, 1992). Similar procedures were used in a study of gasoline transportation workers to estimate work shift TWA exposure to total hydrocarbons from task-specific exposure levels. In that study, however, taskspecific exposure levels were estimated from actual IH data, rather than calculated (Smith et al., 1991). The present investigation is the first in which mechanistic models have been used systematically to obtain exposure estimates for a large number of individual workers. The procedure has several limitations. First, although exposure estimates are quantitative and are expressed in ppm, they are not actual measurements. The validity of the estimates is a function of the amount and accuracy of data collected at the plants, and of the assumptions used in the calculations. Thus, misclassification of exposure levels is likely. Even if the models are correct and the exposure estimates developed for well specified work area groups are valid, work histories abstracted from personnel records do not always characterize the job assignments with precision. The exposure estimates developed for such work area groups are imprecise assessments of the exposure levels experienced by any individual worker. This is a limitation of the work histories, rather than a limitation of the exposure estimation procedure, and could be avoided, for example, in a nested case-control study that allowed a detailed reconstruction of the job assignments of a selected group of employees. Only limited efforts were feasible to estimate a range of credibility for the exposure estimates. To the extent that was feasible, variability in the frequency and duration of a specific task was used in the calculation of exposure indices, and a range was computed for certain task-specific exposure estimates. The boundaries of these ranges were used to compute alternative cumulative exposure estimates for individual workers. The result of this effort led to only modest differences

200

M. Macaluso et at!/Toxicology 113 (19%) 190-202

in estimates of cumulative exposure for individual workers, and mortality analyses conducted using the alternative estimates yielded essentially the same results as the analyses presented here. Finally, the procedure lacked extensive validation because of budget constraints. Limited validation of the exposure estimates was conducted at one plant, where measurement of air concentrations of BD and STY were performed and compared to calculated exposure levels. The objectives of the investigation were to evaluate the validity of four methods for measuring the air concentration of BD and STY in the workplace and to carry out a limited field evaluation of the validity of the mathematical models. A detailed report on these studies will be published elsewhere. Briefly, the NIOSH method used in the extent-of-exposure industry-wide survey of 1986 (Fajen et al., 1990) appeared to be the least sensitive of a series of BD measurement methods tested, confirming the suspicion that NIOSH data and possibly other available historical industrial hygiene measurements may have underestimated exposure to BD and STY in SBR manufacturing. On the other hand, our calculated exposures appeared to be too high for BD, and perhaps too low for STY. Review of the assumptions used in developing the estimates revealed errors in the derivation of emission rates. The procedures for calculating emission rates were revised, and the recalculated exposure estimates were in close agreement with IH measurements obtained using valid sampling and testing procedures. The estimates employed in analyses conducted for the present report reflect the revised emission rate calculations. Despite its limitations, the exposure estimation procedure employed in this study has substantial strengths. A major advantage of this strategy is that determinants of exposure such as the frequency and duration of a specific task, the distance of an operator from an emission source and, to a lesser extent, the frequency and intensity of monomer emissions, can be estimated with precision from plant records and from the recollection of long term employees. The procedure also involved an evaluation by selected plant

experts who reviewed the completeness of our identification of determinants of exposure for the most important work area groups and time periods, and who reviewed the correctness of our assumptions. Another strength of the procedure is that exposure estimates developed a priori from the exposure models could be compared with existing IH data in a context that ensures independence of estimates and data, although the dubious accuracy of historical BD measurements limits the interpretability of such comparisons. Finally, the job and task descriptions and all assumptions employed to calculate the exposure estimates are fully documented, and it is feasible and inexpensive to audit and revise the exposure estimates as new information is accrued. The results of the analysis of leukemia mortality by estimated cumulative exposure to BD and STY lend credibility to the exposure estimation procedure. The findings are consistent with the observation of elevated leukemia mortality among long term workers, and among subjects in work area groups that are commonly believed to entail exposure to BD. Because of the correlation between BD and STY exposure, there is some uncertainty as to whether the leukemia excess is attributable entirely to one monomer. Leukemia risk was associated more strongly and consistently with BD than with STY in this study and in a previous case-control study (Santos-Burgoa et al., 1992). Moreover, studies of workers in the chemical and plastics industries in the USA (Okun et al., 1985; Wong, 1990; Bond et al., 1992) and in Europe (Frentzel-Beyme et al., 1978; Hodgson and Jones, 1985; Kogevinas et al., 1993), who were generally exposed to levels of STY much higher than levels experienced by SBR workers and had no BD exposure, have reported no or minimal associations with leukemia. Thus, a parsimonious interpretation of the findings presented here is that exposure to BD produces a dose-related increase in the occurrence of leukemia among exposed workers. Analyses addressing the form of the dose-response relation, the effect of peak exposure to monomers, and the relation of BD exposure to specific leukemia cell types will be reported elsewhere.

M. Macaluso et aL IToxico1ogy 113 (19%) 190-202

Acknowledgements This work was done under contract with the International Institute of Synthetic Rubber Producers. The authors gratefully acknowledge the cooperation of the management of the participating companies and of the. long-term employees who participated in over 200 interviews, the industrial hygiene consulting provided by Arthur Dunlop, and the technical assistance provided by Linda Gunnells and Ayumi Domori.

References ACGIH (1992) Industrial Ventilation, 2lst Edition. A Manual of Recommended Practice. American Conference of Governmental Industrial Hygienists, Cincinnati, OH, pp. 2-5. Bond, G.G., Bodner, K.M., Olsen, G.W. and Cook, R.R. (1992) Mortality among workers engaged in the develop ment or manufacture of styrene-based products-an update. Stand. J. Work Environ. Health 18, 145-154. Breslow, N.E. and Day, N.E. (1987) Statistical methods in cancer research. Volume II. The design and analysis of cohort studies. IARC Scientific Publication No. 82, Lyon, France. Checkoway, H., Pearce, N.E. and Crawford-Brown, D.J. (1989) Research Methods in Occupational Epidemiology. Oxford University Press, New York, pp. 18-45. Delzell, E., Sathiakumar, N., Hovinga, M., Macaluso, M., Julian, J., Larson, R., Cole, P. and Muir, C.F. (1996) A follow-up study of synthetic rubber workers. Toxicology (this issue). Fajen, J.M., Roberts, D.R., Ungers, L.J. and Radha Krishnan, E. (1990) Occupational exposure of workers to 1,3-butadiene. Environ. Health Perspect. 86, 11-18. Francis, M., Matanoski, G., Elliot, E. and Correa-Villaseiior. A. (1996) Butadiene exposure measures for the assessment of risk of gastrointestinal cancers in a case-control study of synthetic rubber producers. Toxicology (this issue). Frentzel-Beyme, R., Thiess, A.M. and Wieland, R. (1978) Survey of mortality among employees engaged in the manufacture of styrene and polystyrene at the BASF Ludwigshafen works. Stand. J. Work Environ. Health 4, 231-239. IARC (1987a) Styrene. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Suppl. 7. Overall Evaluations of Carcinogenicity: An Updating of IARC Monographs l-42. International Agency for Research on Cancer, Lyon, pp. 345-347. IARC (1987b) Benzene. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Suppl. 7. Overall Evaluations of Carcinogenicity: An Updating of IARC

201

Monographs l-42. International Agency for Research on Cancer, Lyon, pp. 120- 122. IARC (1992) 1,3-Butadiene. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Vol. 54. Occupational Exposures to Mists and Vapours from Strong Inorganic Acids; and Other Industrial Chemicals. International Agency for Research on Cancer, Lyon, pp. 237-285. Hayes, R.B., Raatgever, J.W., de-Bruyn, A. and Germ, M. (1986) Cancer of the nasal cavity and paranasal sinuses, and formaldehyde exposure. Int. J. Cancer 37,487-492. Hodgson, J.T. and Jones, R.D. (1985) Mortality of styrene production, polymerization and processing workers at a site in northwest England. Stand. J. Work Environ. Health 11, 347-352. Homung, R.W., Greife, A.L., Stayner, L.T., Steenland, N.K., Herrick, R.F., Elliott, L.J., Ringenburg, V.L. and Morawetz, J. (1994) Statistical model for prediction of retrospective exposure to ethylene oxide in an occupational mortality study. Am. J. Ind. Med. 25, 825-836. Kogevinas, M., Ferro, G., Saracci, R., Andersen, A., Biocca, M., Coggon, D., Gennaro, V., Hutchings, S., Kolstad, H., Lundberg, I., Lynge, E. and Partanen, T. (1993) Cancer mortality in an international cohort of workers exposed to styrene. In: M. Sorsa, K. Peltonen, H. Vainio and K. Hemminki (Eds), Butadiene and Styrene: Assessment of Health Hazards. IARC Scientific Publications No. 127. International Agency for Research on Cancer, Lyon, pp. 289-300. Kromhout, H., Oostendorp, Y., Heederick, D. and Boleij, J.S.M. (1987) Agreement between qualitative exposure estimates and quantitative exposure measurements. Am. J. Ind. Med. 12, 551-562. Lipton, S. and Lynch, J. (1987) Health Hazard Control in the Chemical Process Industry. Wiley, New York, p. 133. Macaluso, M., Delzell, E., Rose, V. and Perkins, J. (1993a) Use of organic solvents and potential worker exposure in the motor vehicle manufacturing industry. Am. J. Ind. Med. 23,449-460. Macaluso, M., Delzell, E., Rose, V., Perkins, J. and Oestenstad, K. (1993b) Inter-rater agreement in the assessment of solvent exposure at a car assembly plant. Am. Ind. Hyg. Assoc. J. 54, 351-359. Matanoski, G.M. and Schwartz, L. (1987) Mortality of workers in styrene-butadiene polymer production. J. Occup. Med. 29, 675-680. Matanoski, G.M., Santos-Burgoa, C. and Schwartz, L. (1990) Mortality of a cohort of workers in the styrene-butadiene polymer manufacturing industry (1943- 1982). Environ. Health Perspect. 86, 107- 117. Matanoski, G., Francis, M., Correa-Villasefior, A., Elliott, E., Santos-Burgoa, C. and Schwartz, L. (1993) Cancer epidemiology among styrene-butadiene rubber workers. In: M. Sorsa, K. Peltonen, H. Vainio and K. Hemminki (Eds), Butadiene and Styrene: Assessment of Health Hazards. IARC Scientific Publications No. 127. International Agency for Research on Cancer, Lyon, pp. 363-374.

202

M. Macaluso et al. I Toxicology 113 (19%) NO-202

Meinhardt, T.J., Lemen, R.A., Crandall, MS. and Young, R.J. (1982) Environmental epidemiologic investigation of the styrene-butadiene rubber industry. Mortality patterns with discussion of the hematopoietic and lymphatic malignancies. Stand. J. Work Environ. Health 8, 250-259. Okun, A.H., Beaumont, J.J., Meinhardt, T.J. and Crandall, MS. (1985) Mortality patterns among styrene-exposed boatbuilders. Am. J. Ind. Med. 8, 193-205. Rice, C.H. (1991) Retrospective exposure assessment: a review of approaches and directions for the future. In: S.M. Rappaport and T.J. Smith (Eds), Exposure Assessment for Epidemiology and Exposure Control, Lewis, Chelsea, pp. 185-197.

Samuels, S.J., Lemasters, G.K. and Carson, A. (1985) Statistical methods for describing occupational exposure measurements. Am. Ind. Hyg. Assoc. J. 46,427-433. Santos-Burgoa, C., Matanoski, GM., Zeger, S. and Schwartz, L. (1992) Lymphohematopoietic cancer in styrene-butadiene polymerization workers. Am. J. Epidemiol. 136, 843-854. Smith, T.J., Hammond, S.K., Hallock, M. and Woskie, S.R. (1991) Exposure assessment for epidemiology: characteristics of exposure. Appl. Occup. Environ. Hyg. 6,441~447. Wong, 0. (1990) A cohort mortality study and a case-control study of workers potentially exposed to styrene in the reinforced plastics and composites industry. Br. J. Ind. Med. 47, 743-462.