American Journal of Industrial Medicine 2269-83 (1992)

General Mortality and Respiratory Cancer Among a Cohort of Male Chemical Workers in California Cecil M. Burchflel, PhD, MPH, Janice B. Cartmill, RN, Fredric D. Axe, BA, CIH, and Gregory G. Bond, PhD, MPH

Cohort mortality and nested case-control studies were conducted involving 2,901 men employed 1 year or more between 1940 and 1986 at any of four California facilities of a major chemical company. Employees experienced fewer deaths from each of the major causes than were expected based on US.,California, and local county mortality rates. Respiratory cancer was significantly elevated in one socioeconomic category comprised of operators (SMR = 157, 95% CI = 109-220). The 34 cases who died from respiratory cancer and 136 matched controls, all of whom were operators, were included in a nested case-control study. Departments in which subjects had worked were grouped into 13 work assignment or product categories by an industrial hygienist without knowledge of case-control status. Smoking habits and other occupational exposures were ascertained by telephone interview from subjects or surrogate-responders. As expected, current cigarette smoking was strongly related to respiratory cancer. After adjustment for smoking, cases were significantly more likely than controls to have ever worked in one of the 13 work areas (supervision, services, and business support). However, no dose-response relationship was evident with duration of employment in this work area and the departments involved were associated with plant security and not chemical production. Results were similar when a 15-year latency period was assumed. These findings suggest that the excess of respiratory cancer mortality among operators was most likely due to differences in cigarette smoking or other factors not ascertained, rather than to a specific occupational exposure. 6 1992 Wiey-Liss, Inc. Key words: cohort studies, chemical industry, epidemiology, lung neoplasms, occupational mortality, retrospective studies, smoking


Surveillance is recognized as an important step in the prevention of occupational disease [Landrigan, 19891. While a variety of data sources are available for occupational surveillance, the use of death certificates (with their recognized limitations) is still an effective means of monitoring the health of a workforce [Melius et al., 19891. Cohort mortality studies of occupational groups can provide a valuable form of surveillance, especially when combined with a nested case-control study. The nested Department of Epidemiology, The Dow Chemical Company, Midland, MI (C.M.B., J.B .C., G.G.B.). Western Division Human Resources, Dow Chemical U.S.A., Pittsburg, CA (F.D.A.). Address reprint requests to Dr. Gregory G . Bond, The Dow Chemical Company, Department of Epidemiology, 1803 Building, Midland, MI 48674. Accepted for publication December 13, 1991. 0 1992 Wiley-Liss, Inc.


Burchfiel et al.

case-control design provides an efficient method which avoids the necessity of collecting exposure data from all current and former employees, yet allows the use of all cases of a specific outcome identified in a general cohort mortality study. A previous cohort mortality study of male employees who had been employed for one or more years between 1940 and 1969 at the Western Division of The Dow Chemical Company revealed no significantly elevated mortality risks [Ott et al., 19851. This previous study was limited by relatively few deaths in some causespecific categories, the use of one national comparison population, and its exploratory nature. The present study expands the cohort to include 982 additional employees hired after 1969, extends follow-up an additional seven years, utilizes an internal comparison method as employed previously, and uses state and local mortality rates, in addition to national rates, for external comparison. It also uses a nested casecontrol design to investigate an excess of respiratory cancer which was identified in one occupational group (operators) during the cohort mortality phase of this investigation. This study examines whether cases of respiratory cancer were more likely than controls to have worked in specific departments or work areas and uses interview data to adjust for the role of cigarette smoking and other occupational exposures. MATERIALS AND METHODS Cohort Mortality Study The study population included 2,901 men who were employed for one or more years between January 1, 1940 and December 31, 1986 at one of four California locations (Pittsburg, Walnut Creek, Torrance, and Fresno) of The Dow Chemical Company. Women were not included in the study cohort because, historically, few were employed in manufacturing areas. An additional 982 eligible men who were hired between 1970 and 1986 were identified from computerized personnel records and supplemented the 1,919 men previously studied by Ott et al. [1985]. The majority of employees were white and were hired at age 20-29 years (Table I). Hiring was fairly uniform across time and nearly one third were hired before 1950. Many different job functions were represented in the workforce, most of which would have involved the potential for chemical and physical agent exposures. The three most common job titles of operators, professionalltechnical, and laborers accounted for over 75% of the positions at entry. Seventy percent of employees were assigned to the Pittsburg production facilities, with the remainder working at three other locations (Walnut Creek 118, Torrance 15%, and Fresno 4%). Vital status follow-up was ascertained from 1940 through 1986 primarily via linkage to the records of the National Death Index and the Social Security Administration. These sources were supplemented with efforts through the California Department of Motor Vehicles, credit agencies, and personal contact. Death certificates were obtained for 96.7% of the decedents and underlying cause of death was coded by an experienced nosologist according to the revision of the International Classification of Diseases in effect at the time of death. As of January 1, 1987, 2,249 (77.5%) of the cohort members were alive, 583 (20.1%) were deceased, and only 69 (2.4%) could not be traced. Mean duration of employment was 11.9 years (median = 8.6 years). A sizeable proportion (21.2%) of the workers were employed more than 20 years. Persons for whom vital status could not be established tended to be shorter-term employees (74% of the 69 men worked

Respiratory Cancer in Chemical Workers


TABLE I. Distribution of Chemical Workers by Selected Characteristics, California 194&1986 Characteristics Race White Black Hispanic Asian American Indian Age at Hire 15-19 20-29 30-39 40-49 50




2,545 105 189 57 5

87.7 3.6 6.5 2.0 0.2

228 1,540 719 294 120

7.9 53.1 24.8 10. I 4.1

870 1,027 1.004

30.0 35.4 34.6

147 767 25 1 87 856 218 575

5.1 26.4 8.7 3.0 29.5 7.5 19.8

Year of Hire


Socio-economic category I



Job title

Managerdadministrators Professionahechnicd Craftsmen Technicians Operators Office/clerical and service Laborers and other unskilled

between 1 and 5 years). Although not shown, they were comparable in age and year of hire to the workers who were established as alive. Standardized mortality ratios (SMRs) and their 95% confidence intervals (CI) [Miettinen, 1976; Fisher, 19561 were calculated using a modified life-table program developed by Monson [1974] that incorporated comparisons with national, state, and local population groups. Person-years for cohort members were accumulated across five-year age- and calendar-year-specific intervals beginning with the first anniversary of their hiring or January 1, 1940, whichever was most recent. Subjects continued to contribute person-years at risk until the earliest of the following events: 1) the employee was lost-to-follow-up, in which case he was withdrawn from the life table on the date of last contact; 2) the employee died; or 3) December 31, 1986. To serve as an indicator of socioeconomic status, five job title categories at entry were defined. The five levels included managerdadministrators and professionalkechnical (I), craftsmen and technicians (11), operators (111), office/clerical and service (IV), and laborers and other unskilled (V). In addition to the indirect standardization approach of the SMR, internal comparisons were also utilized. The Mantel-Haenszel method adapted for cohort studies [Hakulinen, 1981; Gilbert, 19821 was used to determine the influence of age at hire, period of hire, and duration of employment on mortality and to provide a means of direct adjustment for differences in age and calendar year in comparing subgroups internally.


Burchfiel et al.

Nested Case-Control Study

All cases of respiratory cancer among socioeconomic category 111 identified in the cohort mortality study were matched to four controls on socioeconomic level, gender, race, and year of birth within 5 years. Controls were selected randomly from all present and former employees who worked as operators and met the matching criteria. Incidence-density sampling of controls was used which involved selection of controls from the group of persons who are at risk of developing respiratory cancer [Pearce et al., 19891. The study population consisted of 34 cases and 136 matched controls: a total of 163 individuals were involved since seven of the cases also served as controls for other cases. Telephone interviews were conducted with the subjects or their next of kin to ascertain history of smoking and occupational exposures that may have occurred elsewhere (e.g., farming, shipbuilding, insulation, refineries, rubber, printing, wmdworking, steel, asbestos, arsenic, wood dust, sulfur dioxide, and coke ovens). Of 163 individuals in this phase of the study, 140 (86%) were located. A total of 133 individuals were interviewed; a knowledgeable respondent was not found for one individual and six refused. While 55 subjects were alive and interviewed directly, surrogate respondents were interviewed for 73 deceased subjects and 5 who were alive, yet unable to participate. Surrogate responders included spouses (32%), sons or daughters (49%), and other relatives or friends (19%). As expected, subjects for whom interviews were obtained were born later, entered the company more recently, stopped working sooner, and tended to work somewhat longer compared with subjects not interviewed. In addition, cases and controls who were interviewed did not differ significantly with respect to year of birth, race, age at hire, year of entry, and duration of employment. Work histories at the facility under study which included job classification, department, and date were extracted from computerized records. Departments that were relatively homogeneous with respect to possible occupational exposures were grouped into work assignment or product categories by an industrial hygienist (F.D.A.) without knowledge of case-control status. Work assignment categories were similar to those used in a previous mortality study of Western Division employees [Ott et al., 19851, with several additions. A total of 13 categories were created from 113 departments. Analyses were performed using the Statistical Analysis System [SAS Institute, Inc., 19851. The chi-square test was used to compare proportions for categorical variables and the unpaired t test was used to compare means for continuous variables. Workers were classified by whether or not they had ever worked in each of the 13 product categories. Only exposures of controls which occurred prior to the death of their matched case were counted in the analyses. Both conditional and unconditional logistic regression models were used to calculate odds ratios and 95% CISadjusted for cigarette smoking [Prentice, 19761. Of the various measures of smoking, current smoking yielded the most stable estimates and was selected for adjustment purposes. Since the results using conditional logistic regression were virtually identical, only the unconditional results are presented. A chi-square test for trend was used to examine whether odds ratios increased linearly with duration of employment (0 years,

General mortality and respiratory cancer among a cohort of male chemical workers in California.

Cohort mortality and nested case-control studies were conducted involving 2,901 men employed 1 year or more between 1940 and 1986 at any of four Calif...
826KB Sizes 0 Downloads 0 Views