Watkins, Deborah K. MS; Chiazze, Leonard Jr ScD; Fryar, Cheryl MSPH
The inclusion of women and minorities in occupational cancer epidemiologic research, for the most part, has not been commonplace. Zahm et al reviewed eight epidemiologic journals over a 20-year period and found that 46% of the published articles "clearly indicated that the study subjects were limited to white men."1 Of the remaining articles, 54% either "clearly indicated that the study populations included other race-gender groups or made no mention of race or gender but seemed highly likely to have included others because of where the studies were conducted." Although 35% of the journal articles indicated that the study population included white women, only 14% presented any race-gender specific analyses in the paper. Two percent of the studies reported inclusion of nonwhite women (1% with detailed analyses). The inclusion of nonwhite men was mentioned in 7% of the articles, with only 3% reporting detailed analyses.
Epidemiologic investigations of the potential health effects of man-made vitreous fibers (MMVF) have focused primarily on white men. Although the European historical cohort study of MMVF production workers in 13 MMVF plants enrolled women, most of the published analyses have been for both sexes combined.2,3 The US MMVF cohort is being updated to include approximately 3800 women from 11 fibrous-glass plants.4 A recently published update of a study of three Swedish MMVF production plants enumerated women in the cohort, and even though gender-specific analyses were conducted, results are reported for men and women combined.5 Shannon et al included women in their study of a Canadian glass-filament plant but focused only on men after finding no lung cancers among the women.6
The Division of Occupational Health Studies, Georgetown University Medical Center, undertook an historical cohort mortality study of Owens Corning's Anderson, South Carolina, continuous filament plant. The decision was made to collect information on anyone who worked at the Anderson plant for a minimum of one year between the opening of the plant in 1951 and December 31, 1991. This article reports the results of women and minorities working at Anderson during that time period. The results for white men have been presented separately.7
Records for all persons hired at the Anderson plant were retrieved from company records both at the plant and from corporate headquarters. Although all records were collected, the study cohort was limited to those employees who had worked for at least one year at the Anderson plant between the opening of the plant in 1951 and December 31, 1991. The Social Security Employee Quarterly report (IRS form 941) was used for independent cohort verification. Of the total number of employees, 1074 white women (23.2%), 130 black women (2.8%), 494 black men (11%), and 2933 (63%) white men met the cohort entrance criterion.
Tracing for vital status was carried out through the Social Security Administration Death Master File, Equifax Mortality Data Files, the Health Care Financing Administration, credit bureaus, telephone directory assistance, PhoneDisc, voter registration records, state motor vehicle bureau records, tracing by a local contact in South Carolina, and the National Death Index. The distribution of the cohort by race, gender, and vital status is given in Table 1.
State vital statistics offices could not locate four death certificates for white women and three for black men. In cause-of-death analysis, deaths listed as "cause unknown" are included in the number of total deaths. Death certificates were coded by a qualified nosologist according to the International Classification of Disease revision in effect at the time of death.
Race was unknown for 580 persons (12.2% of the total cohort). The vast majority of those of known race (86%) were white, although that proportion varies with year of hire. Of the 580 cohort members with race listed as unknown, race was assigned randomly by year of hire according to the proportion of those with race known by gender in the hire year. On that basis, 539 cohort members were assigned to the white race category and 41 to the black race category. We verified the race assignment of subsequent deaths among the persons listed as "race unknown" when we compared the race recorded on the death certificate with that assigned.
An historical environmental reconstruction (HER) of the Anderson plant, patterned after that reported previously for Owens Corning's Newark, Ohio, plant,8 was undertaken to characterize the working environment at the Anderson plant from its beginning in 1951 through 1991.
The employee work history provided the linkage between the individual employee and the HER. The work history contains job titles, department, start dates for each job, and status (hire, termination, leaves of absence, rate changes, etc). Each line of the work history was entered into a computerized work history database. Each line of the work history was then assigned a process code corresponding to a list of processes developed from the HER.
A list of exposures for each process was developed from information contained in the HER. Because few records for quantitative exposure exist prior to 1970, an exposure assessment committee, consisting of current and former Owens Corning employees knowledgeable in industrial hygiene and the current and historical plant processes, was established to develop quantitative estimates of potential exposure to each substance for each process from the date of the plant's opening. Exposure assessments were developed for respirable glass fibers, total particulate, asbestos, refractory ceramic fibers (RCF), respirable silica, formaldehyde, total chrome and arsenic.
A cohort member's working experience at the Anderson plant can be categorized into three broad categories-production, nonproduction, and a combination of the two. The number and percentage of cohort members by race and gender for whom work histories are available is provided in Table 2. Of the white women in the Anderson plant cohort, 87% held only production-related positions during their employment, 5.5% had both production and nonproduction positions, and only 7.5% worked only in nonproduction-related activities. For white men, the percentages were 89.2% for positions only in production, 6.2% for both production and nonproduction positions, and 4.6% for nonproduction positions only.
The percentage of black men who worked only in a production-related capacity was 93.4%; 3.5% worked in both production and nonproduction positions, and 3.1% held only nonproduction positions. The percentage of black women employed only in production activities was 90.6%; 5.5% held both production and nonproduction-related positions, and 3.9% held only nonproduction posts.
Results for the cohort follow-up are presented as standardized mortality ratios (SMR). Mortality at the Anderson plant is compared with (1) mortality (specific for race and gender) for the entire United States (national standard), and (2) mortality specific for race and gender for Anderson County, South Carolina (local standard).
Standardized mortality ratios are presented using both national and local standards for white women (Tables 3 and 4), black men (Tables 5 and 6), and black women (Tables 7 and 8). The results presented below are for SMRs with at least four deaths, because SMRs based on fewer than four deaths are likely to be unstable.
Among white women, there are 107 total deaths and relatively few decedents for most specific causes of death. The only statistically significant SMR is for motor-vehicle accidents on the national standard, with eight observed deaths compared with 3.16 expected (SMR = 253; 95% confidence interval [CI], 109.2 to 498.6). Although this SMR is elevated on the county standard (SMR = 174.0), it is not statistically significant. Four of the auto accidents occurred while the women were actively employed at the Anderson plant and of these four, one accident occurred per decade. The remaining four auto accidents occurred after termination of employment (between five and 16 years postemployment).
On the national standard, all cancers combined is below unity and only slightly elevated on the county standard. The SMR for total digestive cancer is based on six deaths and is below unity on the national standard (SMR = 79.5) and only slightly above unity on the county standard (SMR = 105.9). The lung cancer SMR based on four deaths is less than unity on both the national (SMR = 49.9) and county (72.1) standards.
The SMR for breast cancer on the national standard was 89.6 (nine observed, 10.04 expected; 95% CI, 41.0 to 170.1). On the county standard, the SMR was slightly above one (SMR = 108) with 8.33 expected (95% CI, 49.4 to 205.1). Other female genital organ cancers (including three ovarian, one vaginal, and one site unspecified) was also elevated both on the national and county standards.
The SMR for cancers of lymphatic and hematopoietic tissue combined was slightly elevated on both the national and the county standards (SMR = 119.4, national standard; SMR = 147, county standard).
For the non-cancer causes of death, slight elevations on both the national and county standards are seen for cerebrovascular disease and cirrhosis. In addition to the motor-vehicle accidents mentioned above, suicides are also elevated on both the national and county standards (four observed; 2.71 expected, SMR = 147.7, 95% CI, 40.2 to 378.2; and 2.53 expected, SMR = 158.1, 95% CI, 43.1 to 404.9).
The SMR for nonmalignant respiratory disease, minus influenza and pneumonia, is slightly elevated but is not statistically significant on either the national or county standard. The SMRs for heart disease are in deficit on both the national and county standard.
Black Men and Women
No significant excesses among black men were found, based on the national standard. The SMR for all cancers combined is below one on both the national (SMR = 84.2) and county (SMR = 82.3) standards. The only cancer cause of death with four or more deaths is total digestive cancer, and the SMR is below one on both the national (SMR = 86.2) and county (SMR = 94.2) standards. No other cancer cause has more than two deaths, therefore SMRs must be interpreted with extreme caution. A statistically significant elevation for-leukemia on the county standard was found among black men but is based on only two deaths. SMRs for lung cancer among black men are below unity for both the national (SMR = 33.6) and county (SMR = 29.6) standards but are based on only two deaths.
Significant deficits were observed for heart disease among black men on both the national (SMR = 54.2) and county (SMR = 53.7) standards. The SMR for deaths from homicides was also significantly below one on both the national and county standards but is based on only two deaths.
There were only four deaths among black women (two cerebro-vascular, one heart disease, and one other cause of death). As a result, any analysis by cause of death is uninformative.
Summary and Discussion
Studies of women and minorities in the workplace have not been of the same magnitude as those completed for white men. There are many justifiable reasons why this has been the case-too few women and/or minority employees for a cohort to be defined, not enough deaths for meaningful analysis, and the unique difficulties in tracing women.
Characteristics of race/gender groups in the Anderson cohort prove to have many similarities. The age at hire for the four race/gender groups is strikingly similar: 26.8 for white women, 26.2 for white men, 27.1 for black men, and 26.7 for black women. The youngest age at hire for any of the race/gender groups was 15.5 for white men, whereas the youngest white woman was only slightly older, at 15.9 years. The minimum age for black men was 16.6 and for black women, 18.2. What is most interesting is the maximum age at hire-for white men, 60.8; for black men, 55.9. Women were slightly younger at the maximum age at hire and very similar-for white females, 55.9, and for black women, 55.4.
The average length of employment is also similar. For white men, the mean length of employment was 11.8 years vs 10.6 for black men; for women, 8.5 years for white women vs 8.0 for black women. The minimum length of employment was one year, to meet the cohort entrance criterion. The maximum length of employment did vary, with white men having the longest length of service at 41.4 years. Black men had 39.8 years, white women 37.3 years, and black women 23.1 years of service.
A majority of the cohort held either only production jobs or a combination of production and nonproduction jobs. Of this cohort majority, black men held the highest percentage (96.9%), followed closely by black women (96.1%); white men were third with 95.4%, and white women held 92.5%.
SMRs were recalculated for white women, black men, and black women, excluding any cohort member who had only nonproduction work experience. The results are essentially the same as those presented in Tables 3 through 8. For white women, the number of deaths dropped from 107 to 98 (four from malignant neoplasms, two from cerebrovascular disease, one external cause of death, and two from all other causes). Motor-vehicle accidents are still elevated but are no longer statistically significant. Ischemic heart disease is significantly below one. For black men, the total number of deaths is reduced by one (hypertension without mention of heart disease). The deficit for heart disease on both the national and county standards does become statistically significant. The number of deaths for black women is cut in half (from four to two) when the nonproduction-only employees are removed. The two remaining deaths are from cerebrovascular disease.
The difficulty in tracing women for vital status follow-up is not an insurmountable obstacle, as evidenced by this study. Only 5% of the white women and 3% of the black women are coded as "vital status unknown," compared with 3% of the white men and 5% of the black men in the Anderson cohort. Understanding that potential problems could exist in tracing women, investigators should include in their protocols the necessity of collecting additional data related to women, including any name changes that have been recorded on the employment application and/or work history as well as inclusion of father's name for use in matching with the National Death Index.
With the exception of a statistically significant increase for motor-vehicle accidents among white women, there are no causes of death in any of the race/gender groups with statistically significant excesses. Although the statistical power of this study of women and minorities is low and may not be adequate to detect risks of the magnitude typically of interest in studies of men,3 it nevertheless represents the largest single cohort of white women assembled to date in either a wool or continuous filament fiberglass manufacturing facility. It also represents the first study of a cohort of black men and women in the MMVF industry.
This study was supported by a grant to Georgetown University from Owens Corning. The authors gratefully acknowledge the contributions of Valerie Biggs, Michael Chiazze, Jason DeLozier, and Joseph Kozono, and the members of the exposure assessment committee, Dr. Jon Konzen, Mr. Robert Holler, and Mr. Charles Axten.
1. Zahm SH, Pottern LM, Lewis DR, et al. Inclusion of women and minorities in occupational cancer epidemiologic research. J Occup Med. 1994;36:842-847.
2. Saracci R, Simonato L, Acheson ED, et al. Mortality and incidence of cancer of workers in the man made vitreous fibres producing industry: an international investigation at 13 European plants. Br J Ind Med. 1984;41:425-436.
3. Simonato L, Fletcher AC, Cherrie JW, et al. The International Agency for Research on Cancer historical cohort study of MMMF production workers in seven European countries: extension of the follow-up. Ann Occup Hyg. 1987;31(48):603-623.
4. Stone RA, Marsh GM, Henderson VL, et al. Statistical power to detect occupationally related respiratory cancer risk in a cohort of female employees in the US man-made vitreous fiber industry. J Occup Med. 1994;36:899-901.
5. Plato N, Westerholm P, Gustavsson P, et al. Cancer incidence, mortality and exposure-response among Swedish man-made vitreous fiber production workers. Scand J Work Environ Health. 1995;21:353-361.
6. Shannon HS, Jamieson E, Julian JA, et al. Mortality of glass filament (textile) workers. Br J Ind Med. 1990;47:533-536.
7. Chiazze L, Watkins DK, Fryar C. Historical cohort mortality study of a continuous filament fiberglass manufacturing plant. I. White males. J Occup Environ Med. 1997;39:432-441.
8. Chiazze L, Watkins DK, Fryar C, Kozono J. A case-control study of malignant and non-malignant respiratory disease among employees of a fiberglass manufacturing facility. II. Exposure assessment. Br J Ind Med. 1993;50:717-725.
A Publishing Criterion?
"Someone asked me how I could publish a book by Dick Morris, and I said that if I stopped publishing books by people whose sex lives I disapproved of, I'd have no books at all." (Harold Evans, Random House CEO)
From Perspective. Newsweek, September 16, 1996, p35.
© Williams & Wilkins 1997. All Rights Reserved.