Document b57K1wp0okwpnBynpO14mwM6o
American Journal of Industrial Medicine 16:617-630 (1989)
Assessment of Exposure to Chemicals in a Complex Work Environment
M. Gerald Ott, PhD, M. Jane Teta, Dr ph, and Howard L. Greenberg, PhD
Exposure information was evaluated for two large chemical manufacturing facilities and a research and development center in support of occupational health studies of employ ees assigned to these facilities. Methodology and rationale underlying the exposure categorization are provided, and descriptive exposure statistics are presented for a sam ple of 774 employees. Analysis of work patterns and exposure profiles revealed that 1) employee transfers among various production work areas did not follow a predictable pattern, 2) over 41% of the chemicals identified were present in multiple work areas, and 3) individuals exposed to one chemical of toxicologic interest were also likely to be exposed to other similarly toxic materials. The use of both work area and chemicalspecific exposure measures is recommended, as each may be helpful in addressing etiologic questions regarding complex work environments.
Key words: carcinogenesis, case-control methods, occupational epidemiology, chemical manufac ture
INTRODUCTION One of the more pressing needs in occupational cancer research is for the
development of agent-based classifications of exposure [Cole and Merletti. 1980], Several distinct approaches for obtaining such exposure information have arisen. Investigators who conduct broad population- or disease-based research have devel oped systems to classify chemical agent exposures using either in-depth interviews coupled with technical review of various industrial workplaces or publicly available environmental data bases [Gerin et al., 1985; Hoar et al., 1985]. Recently, the use of biological markers has been advocated as a means of selectively improving exposure assessments [Perera et al., 1982].
Researchers whose primary focus is the study of specific industries or work environments have most often relied on linkage of company personnel and other records to describe and then categorize employee exposures [Lloyd and Ciocco, 1969; Gamble and Spirtas, 1976; Ott et al., 1975; Com and Esmen, 1979; Bond et al..
Arthur D. Little, Inc., Cambridge, MA (M.G.O.). Union Carbide Corporation, Danbury, CT (M.J.T.). ARCO, Inc., Los Angeles (H.L.G.). Address reprint requests to Dr. M. Jane Teta, Director of Epidemiology, Union Carbide Corporation, 39 Old Ridgebury Rd., Danbury, CT 06817-0001. Accepted for publication June 5, 1989.
1989 Alan R. Liss, Inc.
UCC 099471
*18 Ott et al.
1986], The usual strategy has been to define work areas that are internally homoge neous in regard to exposure profiles and to link specific agent exposures to employees through knowledge of past work assignments and environmental data relevant to each work area.
The present effort concerns a particular chemical manufacturing and research environment that dates back to the 1920s. Our major goal was to develop an exposure classification system that would be applicable to a variety of epidemiologic research projects. Thus, we sought to retain die maximum of detail regarding the specific occupational exposures that had occurred in this workplace over the 50 year period of interest. Using a lowest common denominator approach, we relied primarily on data contained in historical employment and chemical production records. Linkage of exposure information to individual employees was based on determination of the work area assignments of employees and the nature of chemical processing associated with each work area.
In this report we: 1) describe the methodology supporting our categorization of work exposures; 2) present descriptive exposure statistics for 774 male employees included in a nested case-control study of multiple myeloma. non-Hodgkin's lym phoma, and leukemia [Ott et al., 1989]; and 3) discuss the implications of the methodology and the observed exposure profiles on the conduct of etiologic research related to cancer.
METHODS AND MATERIALS Research Setting
In 1979 a long-term mortality follow-up study of Union Carbide employees in the Kanawha Valley of West Virginia was initiated [Rinsky et al., 1988], The final study cohort consisted of 29,139 men who had ever worked at one of three selected company locations between January 1, 1940, and December 31, 1978. The facilities at these locations included two large chemical manufacturing complexes (one in South Charleston, West Virginia, and a second in Institute, West Virginia) and a research and development center. There were also 4,579 women employees who had been assigned to these company locations. However, the women worked primarily in salaried classifications (78%) and for short time periods (71% were employed for less than 5 years by the company). Since follow-up was judged to be incomplete, women were excluded from both the prior cohort analyses and the present study.
The complicated nature of these work environments was recognized at the outset of the study. Consequently, efforts were begun in the late 1970s to reconstruct a history of chemical processing at the production facilities. Two retired senior manufacturing personnel were contracted to review available records and prepare chronological listings of substances present in each chemical processing area as well as to describe briefly the processes. The listings were compiled from resources that included production reports, manuals of industrial processes, and engineering records. The reconstruction of process history was approximate because of the size of the task, the necessity of relying on personal recall and judgment in many instances, and because certain operations dated as far back as 1925. Oversight guidance was provided by site industrial hygienists. Over 1,000 chemicals were specifically iden tified as having been used or produced at some time between 1925 and 1978 within these facilities. The collected materials provided the core information for developing
UCC 099472
Assessment of Exposure to Chemicals
619
an exposure classification system to reflect the many agent exposures within these work environments.
Although a mortality analysis of the study cohort through 1978 showed fewer observed than expected cancer deaths, there was an increase in lymphosarcoma and reticulosarcoma deaths (48 observed vs. 34.0 expected deaths based on comparison to U.S. white male death rates) and marginal increases in deaths attributed to other lymphatic and hematopoietic tissue cancer (18 observed vs. 16.3 expected deaths) and leukemia (54 observed vs. 49.1 expected deaths) [Rinsky et al., 1988]. Addi tional analyses based on a regrouping of diagnosis categories to identify separately non-Hodgkin's lymphoma, multiple myeloma, and leukemia indicated a relative in crease in standardized mortality ratios with longer employment for each of these categories and supported the need for case-control studies of these diseases. In turn, planning for the case-control studies gave added impetus to completing the exposure classification effort.
Overall Strategy
The recognition of underlying etiologic relationships as occupational in origin is highly dependent on: 1) the degree of heterogeneity of exposure within the examined population; 2) the ability to define indices that reflect the extent of exposure to each causative factor; and 3) sufficiently large numbers of exposed individuals to detect the relationships. Of the above elements, the only one directly under investigator control is the selection of exposure indices. Three measures of exposure were developed for this purpose, one based on identifiable work areas, a second based on specific ex posure agents, and a third based on chemical structure groups.
In many respects, these exposure measures complement one another. The clas sification based on assignment to work areas subdivides all production and nonpro duction areas and work activities into mutually exclusive and exhaustive categories, such that every job assignment maps into a single work area. Work areas describe unique exposure settings. However, the chemical exposure profiles within work areas may vary widely in complexity. Single process chemical units can give rise to a rather limited list of agents and exposure conditions, whereas multiprocess and specialty chemical units give rise to variable and difficult to quantify opportunities for expo sure. Nevertheless, using this approach, evidence can be developed of work-related disease occurrences that are localized to specific work areas. Clearly, it may not be possible to isolate the offending agent or agents.
Classification based on the presence of specific environmental agents and chem ical structure groups allows the grouping of individuals across work areas based on a shared exposure. This increases statistical power with respect to the specific agent at the expense of increased heterogeneity of the exposure milieu. Although it is possible to estimate the extent of exposure to a given agent for individuals assigned to pro duction units, this is not generally the case for nonproduction work assignments (e.g., maintenance, distribution, warehouse, construction, and so forth). In fact, no attempt was made to associate specific substances with employment in these work assign ments, other than exposures that are peculiar to a given trade, such as asbestos with the insulation trade. Thus, there is more overall uncertainty although greater speci ficity for a specific exposure agent versus a work area categorization approach. With the agent classification approach, the critical decision point is the selection of par-
620 Ott et al.
TABLE I. Subdivision of Production and Nonproduction Areas and Activities Into Defined Work Areas in Three UCC Facilities, 1940-1978
Functional activity
No. of distinct work areas
Administrative and business services Construction and maintenance Production support Research and development Production
Acid/ester/anhydride units Acrylate units Alcohol/ketone/aldehyde units Amines units Ethylene/propylene-denvative units Hydrocarbons units Miscellaneous chemicals units Oxide-derivative units Specialty/developmental chemicals units Vinylic monomers and polymers units
Total
4 11 7 6 83 5 2 7 2 4 18 10 14
5 16
111
ticular chemicals to be evaluated from among the 1,000+ candidates. With the work area approach, the concern is how narrowly or broadly to define work areas.
Definition of Work Areas
There are six major activities within the facilities included in this study: 1) administrative and business services; 2) construction; 3) plant-wide maintenance; 4) production support services; 5) research and development; and 6) production. Employees engaged in administrative and business activities generally worked in office settings separate from the production operations. Employees assigned to construction included tradesmen and laborers who constructed new facilities and expansions of existing facilities. Plant-wide maintenance work was carried out by maintenance tradesmen and laborers who could be assigned duties in central shop areas or throughout the production facilities. Production support encompasses: material transfer and distribution activities; janitorial and groundskeeping work; utilities generation (power plant operation) and distribution; and industrial hygiene, health, security, and safety support. Employees assigned to research and development activities worked in laboratory settings or were active in new plant start-ups throughout the production facilities. Also included in this group were engineering design personnel. Finally, the employees assigned to chemical production included operators, foremen, and some maintenance personnel who were assigned to specific production units.
Work areas were defined within the activity groups, and separately for the Institute and South Charleston plant facilities. The research and development center was treated as pan of the South Charleston complex. A detailed subdivision of production work areas was possible and desirable since distinct work areas with relatively homogeneous exposure profiles could be reasonably defined for production units (see Table I). For each employee, duration of assignment in the work area was calculated as an index of exposure to that area.
AJCC
Assessment of Exposure to Chemicals
621
The potential for occupational exposure to plant materials existed for many maintenance, production support services, and research employees. However, on a historical basis, such exposures were not well defined. This is because either the individual employees have been assigned tasks throughout the plant complex on an ad hoc basis (as in maintenance work) or because the materials being worked with changed frequently (as in a research setting). Thus, for men in nonproduction jobs, fewer, broader work area groupings were defined for analysis purposes. Specific trades groups such as electricians and insulators were separately identified.
Definition of Environmental Agents
Well over 1,000 substances (raw materials, catalysts, intermediates, and prod ucts) were initially identified through the agent listings developed for each chemical processing unit. After editing to eliminate synonyms, biologically inert materials, and different formulations of essentially the same products and to replace trade names with generic chemical names when possible, there remained an edited chemical agent file of 1,020 unique substances linked to the various chemical processing units on a year-by-year basis.
With the assistance of chemists and toxicologists, all agents were subcatego rized by chemical structure using as a guide the functional classification described by Soderman [ 1982]. Several chemical groups were formed based on end-use rather than active structure group. For example, antioxidants and catalysts were separately iden tified. A summary list of the 52 major groups is provided in Table II. A further subdivision of chemical families by size of molecule was not utilized in this study. The present grouping was aimed at separating agents by structures likely to have importance relative to carcinogenesis. The same chemicals could easily be regrouped to reflect other dimensions of toxicity such as absorption, distribution, and excretion characteristics of the molecules, as well as known organ-specific toxicity.
Chemical groups of notable concern from a carcinogenic standpoint included epoxides, halogenated compounds, fused cyclics, nitriles, and vinylics [Soderman, 1982; Fishbein, 1979]. Twenty such groups of particular interest for the case-control study are designated in Table 11. An individual was considered positive for exposure to a chemical group based on the presence of any member of that group in a work area to which the employee was assigned. Often, however, multiple members of a chem ical group were found in areas in which any one was present. The only index of exposure to a chemical group was the duration of contact (measured in years) with at least one member of that group. Analysis of exposure-response relationships based on assessing exposure to groups of structurally similar chemicals has many limitations in terms of identifying possible adverse health effects. However, this assessment was only considered an adjunct to work-area and chemical-specific analyses.
There were also 21 individual chemicals selected for analysis based on their carcinogenic potential and their use in multiple work areas (see Table III). If a particular substance had been used in only one work area (this was true for 598 of the 1,020 identified environmental agents), essentially all information available relative to that substance was contained within the work area analyses. Therefore, these 598 agents were not considered as candidates for individual analysis. On a selective basis, exposure intensity scores have been developed for individual agents, but have not been used in the present case-control research.
UCC 099475
622 Ott et al.
TABLE II. Subdivision of Chemicals by Major Functional Groups at Three UCC
Facilities, 1940-1978
_______________________________________________
Chemical group
No. of chemicals in group
Chemical group
No. of chemicals in group
Acrylates*
15 Glycols
18
Alcohols
46 Haloalkancs*
- 16
Aldehydes saturated*
25 Haloaryls*
8
Aldehydes unsaturated* 11 Haloethers*
4
Alkanes
15 Halohydrins
3
Alkenes* Alkyl acids
18 Inorganic irritants 25 Isocyanates
27 6
Alkyl esters Alkyl ethers Alkyl sulfates*
30 Ketones 16 Lactones 6 Mercaptans
21 2
7
Alkylamines Alkylene amines*
31 Metal salts high tox*
22 Metal salts low tox
73
Amine alcohols Amides
20 Nitriles* 5 Peroxides*
11 9
Anhydrides*
12 Phthalates*
21
Anionic surfactants
6 Poly ethers
71
Antioxidants*
12 Pyrans
4
Aromatic amines
12 Pyridines*
8
Aromatic esters
2 Resins and latexes
45
Aryl ketones
3 Silicates
7
Aryls* Carbamates
27 Sulfones 3 Triols
3 10
Catalysts
46 Unclassified
90
Cationic surfactants
2 Vinylics*
26
Epoxides*
15
Epoxy esters
4
Fused cyclics*
20
Glycol ethers
39
''Chemical groups selected for emphasis in the case-control study. "Includes cadmium, chrome, cobalt, manganese, mercury, nickel, silver, tin, and vanadium compounds.
Selection of Subjects
The employees included in the case-control study consisted of 129 men with lymphatic or hematopoietic tissue malignancies except Hodgkin's disease and 645 controls selected from the total cohort according to an unmatched incidence density sampling design described by Greenland and Thomas [1982],
Linking Exposure Measures to Employees
Linkage of potential chemical exposures to individual employees required a history of assignments for each employee, an algorithm to map assignments (desig nated by payroll account numbers) into work areas, and a history of chemical usage associated with each work area. The steps required to complete the linkage are outlined in Figure 1. The record linkage system was dynamic, since both payroll account numbers and chemical usage changed over time. Dates, therefore, had to be part of the link between account numbers and work areas. During their working careers, employees could be assigned to multiple work areas and, consequently, experience exposure to many different chemicals. To determine whether employee
,UCC 099
Assessment of Exposure to Chemicals
TABLE 111. Individual Chemicals Appearing in More Than One Work Area and Selected for Analysis at Three UCC Facilities, 1940-1978
Chemical*
Acetaldehyde Acetonitrile Acrolein Acrylonitrile Allyl alcohol Ally] chloride Benzene Butadiene Butylene oxide Dichloroethyl ether Dioxane Epichlorohydrin Ethylene chlorohydrin Ethylene dichloride Ethylene oxide Formaldehyde Ketene Propylene oxide Styrene Vinyl acetate Vinyl chloride
No. of work areas
16 5 7 10 4 3 22 10 2 7 3 3 4 11 29 6 2
17 9 10 6
"Certain chemicals of interest such as dimethyl sulfate are not included here, since alkyl sulfates are being examined as a group and most members of this group tended to be present in the same work areas.
623
transfers between work areas followed predictable patterns and whether exposures to any given chemical were independent of exposures to other selected chemicals, correlation analyses were performed.
RESULTS
The combined case-control group of 774 men represents 2.7% of the total cohort of 29,139 employed men. In general, individuals selected for the case-control study had earlier hire dates compared with the overall distribution of employees in the study cohort. For example, 82.9% (642) of the case-control employees had been hired before 1950 compared with 53.8% (15,677) of the total employee cohort. Casecontrol employees also had longer employment durations than employees in the overall cohort. Thus, whereas 41.7% (12,151) of the men in the total cohort worked for 10+ years in the facilities, 57.8% (447) of the men in the case-control group had worked that long.
Using the algorithm described in Figure 1, the number of men assigned to each work area and to each chemical exposure category was determined based on cumu lative work history information. The percentage of the work force ever assigned to each major activity group was 10.1% for administrative and business services, 34.9% for construction, 25.8% for plant-wide maintenance, 38.8% for production support
free 099477
624 Ott et al.
Fig. 1. Outline of job history record processing to obtain exposure measures by employee. *Intensit> of exposure scores have been selectively developed for individual chemicals such as ethylene oxide, but were not used in this research.
services, 25.5% for research and development, and 25.8% for production. Assign ments in construction and production support services were of relatively short dura tion, with 47 and 52% of the men, respectively, having worked in these activities for less than 1 year. Maintenance and production assignments were of longer duration, with 45 and 51% of the men, respectively, having worked in these activities for at least 10 years.
In Table IV, the distribution of employees within work areas is given by pro duction status and by employment duration within work area. The nonproduction activity groups accounted for 28 of the 111 work areas. Since these work areas were defined rather broadly, there were far more employees assigned per nonproduction work area than per production work area.
There were 200 men in the case-control group who had ever worked in pro duction areas. Considering all chemicals and cumulative work histories, these pro duction workers came in contact with an average of 58 different chemicals during their assignments in production work areas. For the 52 selected chemical groups, the distribution of employee contact with one or more members of these groups among the 200 production employees in the sample is given in Table V. The percentage of production employees ever exposed ranged from 1.5% for carbamates, cationic sur factants, and epoxy resins to 90% for the unclassified group of chemicals. For more important chemical groups such as vinylics, 78.5% of the production employees had ever worked with one or more of these substances, and 45.5% had worked with these substances for at least 5 years. Thus, at some point during their careers, a high percentage of production employees worked in a production department where vi nylics such as vinyl chloride were present. Members of other chemical groups, such as alcohols, inorganic irritants, and metal salts of low toxicity, were also relatively ubiquitous in the production facilities.
UCC 099478
Assessment of Exposure to Chemicals
TABLE IV. Distribution of 111 Work Areas by Production/Nonproduction Status of Work Area and by Number of Male Employees Assigned to Area
No. of employees
0 1-4 5-9 10-24 25-74 75 +
No. of distinct work areas
Ever employed in area
Employed 5 + years in area
Production*
Nonproduction*
Production
Nonproduction
5 0 28 2
38 1 46 4
26 4
66
12 9
3 11
28
05
06
00
There were 83 production and 28 nonproduction work areas in the study.
625
Similar information is provided in Table VI for 21 individual chemicals. As would be expected, lower percentages of production employees were exposed to individual substances than to chemical groups. Exposure percentages among produc tion employees ranged from 3.5% for butylene oxide to 48.0% for benzene. With a 5 year minimum period of contact, the highest exposure percentage was 26.5% for benzene.
Correlation analyses were performed for men ever assigned to production areas to evaluate the patterns of employee contact with different members of chemical groups, individual chemicals, and work areas. Product moment correlation coeffi cients were calculated using each employee's total duration of contact with a partic ular chemical or work area. The distributional properties of the correlation coeffi cients calculated from exposure durations are unknown. However, the correlation coefficients were only required to serve as simple measures of association in this application. A summarization of all possible pairwise correlation coefficients among chemicals, chemical groups, and work areas is provided in Table VII.
Focusing on the 21 individual chemicals, the distribution of correlations was skewed to the right, with 18.1% of all correlations exceeding 0.5. There were notably high between-chemical correlations for several chemical subgroups. For example, within a subgroup of six chemicals (dichloroethyl ether, dioxane, epichlorohydrin, ethylene chlorohydrin, ethylene dichloride, and ethylene oxide), 13 of the 15 corre lation coefficients were greater than 0.5. Vinyl chloride, acrylonitrile, and vinyl acetate were also highly correlated, presumably because these three chemicals were often used concomitantly to produce copolymers. Finally, allyl alcohol, allyl chlo ride, acrolein, epichlorohydrin, and ketene shared relatively high correlation coeffi cients with one another.
A somewhat similar pattern of correlations emerged when chemical groups were substituted for individual chemicals in the analyses. However, 89% rather than about 64% of all correlations were positive, indicating more overlap in exposures to broad families of chemicals. Among all 52 chemical groups, pairwise correlation coefficients exceeded 0.5 in 12.8% of the comparisons.
Analyses among the various production work areas revealed a totally different pattern of interrelationships. Among the 52 production work areas with three or more employees having assignments in the area, most pairwise correlations were slightly negative and only five (0.3%) exceeded 0.5. The relative independence of work area
vcc 0994 79
626 Ott et al,
TABLE V. Percent of Employees Exposed to One or More Chemicals Within 52 Chemical Groups Among 200 Production Workers at Three UCC Facilities, 1940-1978
Chemical group
Ever exposed
Exposed 5 + years
Acrylates Alcohols Aldehydes saturated Aldehydes unsaturated Alkanes Alkenes Alkyl acids Alkyl esters Alkyl ethers Alkyl sulfates Alkylamines Alkylene amines Amine alcohols Amides Anhydrides Anionic surfactants Antioxidants Aromatic amines Aromatic esters Aryl ketones Aryls Carbamates Catalysts Cationic surfactants Epoxides Epoxy esters Fused cyclics Glycol ethers Glycols Haloalkanes Haloaryls Haloethers Halohydnns Inorganic irritants Isocyanates Ketones Lactones Mercaptans Metal salts, high toxicity Metal salts, low toxicity Nitriles Peroxides Phthalales Poly ethers Pyrans Pyridines Resins and latexes Silicates Sulfones Triols Unclassified Vinyltcs
22.0 (44)* 86.5 (173) 45.0 (90) 37.0 (74) 52.5 (105) 66.5 (133) 53.0 (106) 46.5 (93) 56.5 (113) 15.0 (30) 25.0 (50) 21.5 (43) 23.5 (47) 11.5 (23) 33.5 (67) 17.0 (34) 38.5 (77) 12.5 (25)
4.5 (9) 8.0 (16) 55.5 (111) 1.5 (3) 55.5 (111) 1.5 (3) 45.5 (91) 1.5 (3) 22.5 (45) 29.5 (59) 38.0 (76) 45.0 (90) 10.0 (20) 23.0 (46) 22.0 (44) 88.0 (176) 3.0 (6) 70.5 (141) 2.0 (4) 17.0 (34) 46.5 (93) 87.0(174) 21.5 (43) 32.0 (64) 10.5 (21) 31.5 (63) 3.5 (7) 14.5 (29) 43.0 (86) 7.0(14) 7.5 (15) 12.0 (24) 90.0 (180) 78.5 (157)
11.0 54.5 21.0 16.0 29.0 39.0 27.0 26.0 33.0
7.0 11.0 8.0 7.5 5.5 16.5 5.0 20.0 4,0
--
2.5 33.0
1.0 28.0 -- 23.5
-- 11.5 13.0 16.5 22.0
3.0 11.0 12.0 59.5 2.0 44.0
1.0 4.5 21.5 58.0 11.5 14.5 2.5 14.5 1.5 6.0 23.5 2.5 2.5 . 5.0 56.5 45.5
*Total number of employees in parentheses.
UCC 099480
Assessment of Exposure to Chemicals
TABLE VI. Percent of Employees Ever Exposed to 21 Specified Chemicals Among 200 Production Workers at Three UCC Facilities, 1940-1978
Chemical
Acetaldehyde Acetonitrile Acrolein Acrylonitrile AUyl alcohol Allyl chloride Benzene Butadiene Butviene oxide Dichloroethyl ether Dioxane Epichiorohydrin Ethylene chlorohydrin Ethylene dichlonde Ethylene oxide Formaldehyde Ketene Propylene oxide Styrene Vinyl acetate Vinyl chloride
Ever exposed
30.5 (61)* 5.0 (10) 12.5 (25) 25.0 (50) 9.0 (18) 10.0 (20)
48.0 (96) 30.5 (61)
3.5 (7) 22.5 (45) 10.5 (21) 10.5 (21) 18.5 (37) 22.0 (44) 35.5 (71) 15.0 (30)
9.0 (18) 28.5 (57) 13.5 (27) 40.5 (81) 28.5 (57)
Exposed 5 + years
11.5 2.5 3.0 9.0 3.0 4.0 26.5 13.5 2.0 11.0 4.0 4.5 11.0 9.5 20.0 4.5 3.0 15.0 4.5 20.5 12.0
Total number of employees in parentheses.
627
TABLE VII. Distribution of Pairwise Correlation Coefficients for 21 Selected Chemicals, 52 Chemical Groups, and 52 Work Areas Among 200 Male Employees Ever Assigned to Production Areas
Correlation coefficient*
<-0.100 -0.099 to -0.001 0.0 to 0.099 0 1 to 0.199 0.2 to 0.299 0.3 to 0 499 0.4 to 0.699 0.7 + Total
21 selected chemicals
No. Percent
12 5.7 64 30.5 25 11.9 22 10.5 12 5.7 37 17.6 20 9.5 18 8.6 210
52 chemical groups
No. Percent
2 136 246 254 228 290 125 45 1326
0.2 10.3 18.5 19.2 17.2 21.9 9.4 3.4
52 work areas with 3+ individuals assigned to area
No. Percent
0 1)67
102 20 13 18
3 2 1326
0.0 88.0
7.7 1.5 1.0 1.4 0.2 0.1
"For each employee, the duration of exposure (measured in years) was determined for all chemicals, chemical groups, and work areas. Then correlation coefficients were computed using the individual employee as the unit of observation.
assignments among employees reflects the mutually exclusive definition of work areas and, by inference, the lack of strong systematic patterns in the way individuals transferred between work areas. Conversely, many of the chemicals used within each
UCC 099481
628 Ott et al.
production work area were interrelated, thus giving rise to an increased likelihood that individuals exposed to one chemical of toxicologic importance would also be exposed to other similar chemicals.
DISCUSSION
The task of linking personnel and other historical records to conduct a retro spective exposure assessment can be quite onerous. Nevertheless, when the targeted health outcomes (e.g., site-specific cancers) become manifest only many years after initial exposure, this approach appears to make optimal use of admittedly imprecise exposure information, but information that is specific to the relevant exposure period. From an epidemiologic perspective, the lack of precision in the exposure estimates is of concern. Nondifferential misclassification may bias relative risk estimates toward the null, reduce power, and have unpredictable effects on tests of trend in subgroup analyses [Flegal et al., 1986). In particular, our ability to be specific in associating exposure agents only with production job assignments is a study limitation. In es sence, we have presumed that having daily assignments in specific production work areas will lead to greater average exposure opportunities to a given agent than having been assigned to production support or maintenance activities on a plant-wide basis. There is some evidence in other records to support this presumption. For example, review of medical visits related to acute ethylene oxide exposures revealed far higher visit rates among ethylene oxide production personnel than for maintenance person nel, who were only intermittently working in these units. The supplemental work area classification is a partial, but incomplete, solution to the problem of having different levels of knowledge regarding exposure potential within the various work activities. Better tracking of the exposures of nonproduction personnel is needed to support occupational health research. An additional limitation of the exposure assessment is our inability to account for differences in personal work practices that may lead to widely disparate exposures. There is no easy remedy for this limitation.
There are other biologic and statistical impediments to detecting etiologic re lationships in complex work environments. An important biologic impediment may be our lack of understanding of the mechanisms of cancer causation, particularly as they relate to low level or intermittent exposures and to the effects of direct-acting carcinogens versus agents that may promote or influence the progress of cancer. Toxicological studies of animals have generally been directed at determining the effects of high intensity exposure to specific agents and have not yet been particularly helpful in addressing the multiple exposure issue.
While genotoxic agents may be of concern at relatively low exposure intensi ties, the promotional activities of a much broader class of agents may occur only with high intensity exposures, which are less likely in work environments with active industrial hygiene programs. Thus, the mix of complete carcinogens or genotoxic agents and agents capable of promotion could be an important determinant of the overall impact of a work environment on its subject employee population, with genotoxicity playing a dominant role.
A number of large cohort mortality studies have examined cancer mortality rates among persons employed by U.S. chemical, petroleum, and rubber manufacturing companies [Andjelkovic et al., 1976; Delzell and Monson, 1981; Divine et al., 1985; Hanis et al., 1985; Kaplan, 1986; Wen et al., 1983; Wong et al., 1986; Austin and
UCC 099482
Assessment of Exposure to Chemicals
629
Schnatter, 1983; Bond et al., 1985; O'Berg et al., 1987; Ott et al., 1985; Ott, 1982; Pifer et al., 1986]. Overall, cancer mortality has generally been found to be less than that of the corresponding U.S. population. Although there are important methodologic concerns regarding these studies (see Savitz and Moure [1984] for a discussion of the petroleum industry studies), the findings argue against exposures to the many different chemicals used in these work settings having substantial independent or multiplicative carcinogenic effects. These findings support a practical epidemiologic strategy emphasizing detection and control of complete carcinogens that would be expected to represent a small fraction of the substances in use in these industries. Of the 21 likely carcinogenic agents noted in multiple work areas of the present work environment, 11 were recognized to be mutagenic, four were negative in mutagenic ity studies, and information was incomplete for the six remaining substances. Our exposure assessment strategy does provide for epidemiologic investigation of these substances, and the list of targeted substances can be readily adjusted to accommodate advances in toxicologic knowledge.
For reasons cited by Savitz and Moure [1984], the large cohort study findings do not preclude the possibility of small risk elevations for relatively large segments of the exposed workforce. Such a possibility is difficult to evaluate epidemiologically. Furthermore, if there are many indirect pathways through which low level exposures to a broad group of toxicants influence the risk of cancer, then all but the most direct relationships with individual substances would be hard to identify. The multiple exposure question is particularly difficult to evaluate because of all the possible combinations of exposure. We have attempted to simplify this issue by subcategorizing chemical substances according to structure, thereby reducing the combinations of exposure circumstances to be considered.
ACKNOWLEDGMENTS The technical assistance of Don Gosselin is gratefully acknowledged.
REFERENCES
Andjelkovic D. Taulbee J. Symons M (1976): Mortality experience of a cohort of rubber workers, 1964-1973. J Occup Med 18.387-394.
Austin SG. Schnatter AR (1983): A cohort mortality study of petrochemical workers. J Occup Med 25:304-320
Bond GG. Flores GH, Shellenberger RJ, Cartmill JB, Fishbeckand WA, Cook RR (1986): Nested case-control study of lung cancer among chemical workers. Am J Epidemiol 124:53-66.
Bond GG. Shellenberger RJ, Ftshbeck WA. Cartmill JB, Lasich BJ, Wymer KT. Cook RR (1985): Mortality among a large cohort of chemical manufacturing employees. JNCJ 75:859-869.
Cole P, Merletti F (1980) Chemical agents and occupational cancer. In Demopoulis HB. Mehiman MA teds): "Cancer and the Environment." Park Forest South, 1L: Patholox Publishers. Inc., pp 399-418
Corn M. Esmen N (1979). Workplace exposure zones for classification of employee exposures to physical and chemical agents. Am Ind Hyg Assoc J 40:47-57.
Delzell E. Monson RR (1981): Mortality among rubber workers 111. Cause-specific mortality, 19401978. J Occup Med 23:677-684.
Divine BJ. Barron V, Kaplan SD (1985): Texaco Mortality Study 1. Mortality among refinery, petro chemical, and research workers. J Occup Med 27:445 -447,
Fishbetn L (1979): "Potential Industrial Carcinogens and Mutagens." Amsterdam: Elvesier Scientific Publishing Company.
UCC 099483
630 Ott et al.
Flegal KM, Brownie C, Haas JD (1986): The effects of exposure misclassification on estimates of relative risk. Am J Epidemiol 123:736-731.
Gamble J, Spinas R (1976): Job classification and utilization of complete work histories in occupational epidemiology. J Occup Med 18:399-404.
Gdrin M, Siemiatycki J, Kemper J, Bigin D (1985): Obtaining occupational exposure histories in epidemiologic case-control studies. J Occup Med 27:420-426.
Greenland S, Thomas DC (1982): On tbe need for the rare disease assumption in case control studies. Am J Epidemiol 116:547-553.
Hams NM, Shellenberger LG, Donaleski DL, Sales EA (1985): A retrospective mortality study of workers in three major U.S. refineries and chemical plants. J Occup Med 27:283-292.
Hoar SK, Santodonato J, Cameron TP, Kelsey MI (1985): Monographs on human exposures to chemicals in the workplace. J Occup Med 27:585-586.
Kaplan SD (1986): Update of a mortality study of workers in petroleum refineries. J Occup Med 28:514-516.
Lloyd JW. Ciocco A (1969): Long-term mortality study of steelworkers. I. Methodology. J Occup Med 11:299-310.
O'Berg MT, Burke CA, Chen JL, Walrath J, Pell S, Gallie CR (1987): Cancer incidence and mortality in the DuPont company: An update. J Occup Med 29:245-261.
Ott MG (1982): "Effects of Selection and Confounding on Mortality in an Occupational Cohort." Ph.D. dissertation. University of Michigan, Ann Arbor, Ml.
Ott MG. Carlo GL, Steinberg S, Bond GG (1985): Mortality among employees engaged in chemical manufacturing and related activities. Am J Epidemiol 122:311-322.
On MG, Hoyle HR, Langner RR, Scharnweber HC (1975): Linking industrial hygiene and health records. Am Ind Hyg Assoc J 36:760-766.
On MG, Teta MJ, Greenberg HL (1989): Lymphatic and hematopoietic tissue cancer in a chemical manufacturing environment. Am J Ind Med (in press).
Perera FP, Poirier MC, Yupsa SH, Nakayama J, Javetzki A, Cumen MM, Knowles DM, Weinstein IB (1982): A pilot project in molecular cancer epidemiology: Determination of benzolajpyrene-- DNA adducts in animal and human tissues by immunoassays. Carcinogenesis 3:1405-1410.
Pifer JW. Hearne FT, Friedlander BR. McDonough JR (1986): Mortality study of men employed at a large chemical plant, 1972 through 1982. J Occup Med 28:438-444.
Rinsky RA, Ott MG. Ward E, Greenberg HL. Halperin W, Lest T (1988): Study of mortality among chemical workers in the Kanawha valley of West Virginia. Am J Ind Med 13:429-438.
Savitz DA. Moure R (1984): Cancer risk among oil refinery workers--A review of epidemiologic studies. J Occup Med 26:662-670.
Soderman JV (ed) (1982): "Handbook of Identified Carcinogens and Noncarcinogens: CarcinogenicityMulagenicity Database," Vol 1, Chemical Class File. Boca Raton, FT.: CRC Press.
Wen CP. Tsai SP. McClellan WA. Gibson RL (1983): Long-term mortality study of oil refinery workers. I. Mortality of hourly and salaried workers. Am J Epidemiol 118:526-542.
Wong O. Morgan RW, Bailey WJ, Swercicki RE. Claxton K, Kheifets LA (1986): An epidemiological study of petroleum refinery employees. Br J Ind Med 43:6-17.
UCC 099484