Breast Cancer Incidence and Exposure to Metalworking Fluid in a Cohort of Female Autoworkers

Breast Cancer Incidence and Exposure to Metalworking Fluid in a Cohort of Female Autoworkers Abstract Breast cancer is the leading cancer diagnosed among women, and environmental studies have produced few leads on modifiable risk factors for breast cancer. Following an Institute of Medicine recommendation for occupational studies of women highly exposed to potential breast cancer risk factors, we took advantage of an existing cohort of 4,503 female autoworkers in Michigan exposed to metalworking fluid (MWF), complex mixtures of oils and chemicals widely used in metal manufacturing worldwide. Cox proportional hazards models were fit to estimate hazard ratios for incident breast cancer (follow-up, 1985–2013) and cumulative exposure (20-year lag) to straight mineral oils (a known human carcinogen) and water-based soluble and synthetic MWF. Because the state cancer registry began decades after the cohort was defined, we restricted our analyses to subcohorts of women hired closer to the start of follow-up. Among those hired after 1969, the hazard ratio associated with a 1 interquartile-range increase in straight MWF exposure was 1.13 (95% confidence interval: 1.03, 1.23). In separate analyses of premenopausal breast cancer, defined by age at diagnosis, the hazard ratio was elevated for exposure to synthetic MWF (chemical lubricants with no oil content), possibly suggesting a different mechanism in the younger women with breast cancer. This study adds to the limited literature regarding quantitative chemical exposures and breast cancer risk. breast cancer, metalworking fluid, occupational exposure Breast cancer is the leading cancer diagnosed in women in the United States (1). It is estimated that, in 2016, there will be 246,660 new cases and 40,450 deaths attributable to breast cancer among women in the United States (2). Despite the high numbers of women affected by this disease, well-established risk factors are estimated to account for only 41% of breast cancer cases (3), and studies of environmental exposures associated with breast cancer risk have produced few additional leads on modifiable risk factors over the past 20 years (4). A 2012 Institute of Medicine review of environmental risk factors for breast cancer noted that the initial identifications of many known human carcinogens were based on studies of high exposures in occupational settings and recommended additional breast cancer studies of worker populations (4). Despite provocative evidence that metalworking fluid (MWF) contaminants cause mammary gland tumors in laboratory animals (5) and that occupational MWF-exposure levels are appreciable, few studies have been conducted on associated breast cancer risk. MWF are coolants and lubricants widely used in industrial machining and grinding operations, and are categorized into 3 classes on the basis of composition: straight, soluble, and synthetic. Straight MWF are complex mixtures of paraffinic, naphthenic, and aromatic compounds refined from mineral oil (6). The carcinogenic properties of mineral oil, classified as carcinogenic to humans (7), are thought to be due primarily to their polycyclic aromatic hydrocarbon (PAH) content (8). Soluble MWF are oils emulsified in water. Synthetic fluids are water-soluble chemical lubricants without oil. Ethanolamines and nitrites, added to synthetic MWF to inhibit corrosion and adjust pH, interact to form nitrosamines. Although, to our knowledge, there have been no toxicologic studies in which investigators specifically examined the association of MWF and breast cancer, MWF components including PAH and nitrosamines have been implicated as mammary gland carcinogens (5). Additionally, results of studies on ambient air PAH exposure are suggestive of increased breast cancer risk (9, 10), and nitrosamines are suspected endocrine disruptors (11). Women comprise a growing proportion of the 4.4 million US workers potentially exposed to MWF (12), as well as of the growing workforce employed in metal manufacturing worldwide. The global manufacturing workforce is estimated to be about 30% female (13). The global market volume of MWF in 2013 was 2.3 million tons; Asia had the largest share, 41.5%; North America’s share was 28.0% (14). Few epidemiologic studies on breast cancer risk and MWF exposure exist (15–22). Results of an early study indicated reduced incidence of death due to breast cancer among women working in automotive manufacturing, based on standardized mortality ratios (15, 16). More recently, researchers conducting case-control studies reported elevated breast cancer risk associated with occupation in automotive manufacturing and metal products/metal work (19–21). These studies, however, did not examine a specific compound; instead, occupation was used as a proxy for exposure. Three breast cancer studies have been conducted in the United Autoworker-General Motors (UAW-GM) cohort; 2 were limited in power and relied on either a combination of incident cases and deaths (17) or only deaths (22). In the 1 study in which investigators examined only incident breast cancer cases, Friesen et al. (23) reported a slightly elevated risk associated with exposure to straight MWF. We took advantage of this large cohort of female autoworkers exposed to MWF to examine the relationship between quantitative MWF exposure and breast cancer incidence. We have 9 additional years of follow-up and restricted our analysis to registry-identified incident cases. The aim of this study was to examine the exposure-response relationship between cumulative MWF exposure and breast cancer risk in a cohort of occupationally exposed female autoworkers. METHODS Study population The UAW-GM study was a joint labor-management funded study designed in 1984 to examine cancer mortality and its relation to MWF exposure. This cohort has been described elsewhere (24, 25). Briefly, the original study included 46,316 hourly workers from 1 of 3 automobile manufacturing plants in Michigan who were exposed to MWF primarily via inhalation (24). All hourly employees who had worked at least 3 years before January 1, 1985, were included in the cohort. Cohort participants who were alive on January 1, 1985, when the Michigan Cancer Registry began (n = 33,915), were included in the incidence cohort. Analyses for the present study were restricted to female autoworkers in the incidence cohort (n = 4,567). We excluded women who were missing more than 50% of their employment history (n = 59) or were hired before 1938 (n = 5). This latter restriction was imposed to address potential left truncation bias in the main cohort (26). The final study population comprised 4,503 female workers. Outcome assessment The UAW-GM incidence cohort was linked with the Michigan Cancer Registry to identify incident cancer cases diagnosed between January 1, 1985, and December 31, 2013. Data are collected by the Michigan Department of Community Health as part of the Michigan Cancer Surveillance Program (27), which participates in the National Program of Cancer Registries of the Centers for Disease Control and Prevention (28). We obtained data on first diagnosis of primary breast cancer (International Classification of Disease for Oncology, Third Edition, codes C50.0–C50.9), including in situ and invasive tumors, since 1985. We identified 221 incident cases of breast cancer in the study population, including 72 cases diagnosed when the woman was age 55 years or younger. Most of the women with breast cancer had invasive tumors (82.8%). Data on vital status were obtained from the National Death Index (National Center for Health Statistics, Hyattsville, Maryland) (29). Exposure assessment Cumulative exposure estimates for each MWF type were calculated for each woman in the UAW-GM cohort on the basis of detailed employment records available from hire through 1994 and of a time-varying job-exposure matrix. An extensive retrospective exposure assessment was conducted to develop the job-exposure matrix (30, 31). Size-fractionated MWF concentrations were estimated as an 8-hour time-weighted average (milligrams per cubic meter) based on several hundred personal and area airborne-exposure measurements collected by study industrial hygienists during the mid-1980s. A set of multipliers to adjust MWF concentration for temporal trends was developed on the basis of nearly 400 historical air-sampling measurements (collected from 1958 through 1987), review of historical records, and interviews with plant personnel, and were last updated in 1995. For this study, we used the respirable size fraction (<3.5 μm), which mostly deposits in the alveolar region, of the MWF exposure estimates. The job-exposure matrix was combined with employment records to estimate time-varying, annual average daily exposure to straight, soluble, and synthetic MWF (milligrams per cubic meter) throughout the women’s entire employment. Missing employment information was interpolated by averaging exposures from previous and subsequent jobs. We then calculated cumulative time-weighted exposure to the 3 fluid types (reported as milligrams per cubic meter–years) for each subject for the duration of follow-up. To account for breast cancer latency, cumulative exposures for each fluid type were lagged; because employment data ended December 31, 1994, 19 years before end of follow-up, we used a 20-year lag. This assumes exposure accumulated in the 20 years preceding a diagnosis did not affect cancer risk. Data analyses Cox proportional hazards models were fit to estimate hazard ratios and 95% confidence intervals associated with exposure to straight, soluble, and synthetic MWF (20-year lag) on breast cancer incidence. Exposure was defined as cumulative exposure to each fluid type for the respirable size fraction. MWF exposures were modeled as continuous and categorical. For continuous exposure, hazard ratios were scaled to a 1-interquartile-range increase among women with MWF exposure in the full study population; for the 3 fluid types, straight, soluble, and synthetic, these were 0.318, 0.979, and 0.270 mg/m3-years, respectively. For categorical exposure, the referent groups were women with no MWF exposure. For each fluid type, the median level among women with breast cancer who were exposed to the fluid was used as the cutpoints for the exposure categories. Tests for trend were conducted by setting the value to the median for exposure category, modeling exposure as a continuous variable, and testing for nonzero slope by using a likelihood ratio test. Age was the time metric for all models. All models included baseline covariates for race (white or black), an established breast cancer risk factor, manufacturing plant (to adjust for plant-specific characteristic not captured elsewhere), socioeconomic status, regional difference, and year of hire (B-spline with 3 degrees of freedom and equally spaced knots (32)) to account for secular trends in exposure, including PAH content, personal protective equipment, and plant ventilation. A time-varying covariate for calendar year (B-spline with 3 degrees of freedom and equally spaced knots) was included to account for secular trends in breast cancer diagnosis. To assess the linearity of exposure-response relationships, we evaluated additional Cox models with penalized splines (2 degrees of freedom) for MWF exposure. By necessity, only UAW-GM cohort members alive on January 1, 1985, were included in the incidence cohort, thereby creating a left-truncated cohort. Left truncation occurs when not all otherwise eligible persons are enrolled in the cohort. In this study, downward bias arises from left truncation because the proportion of women susceptible to the effect of exposure decreased over time (26). To reduce this potential bias, we restricted analyses to a series of subcohorts defined by year of hire. Narrowing the interval between hire and start of the cancer registry reduced the opportunity for susceptible women to die before start of follow-up in 1985. We defined 2 subcohorts: those hired in or after 1959 (1959–1981) and those hired in or after 1969 (1969–1981) (20 and 10 years, respectively, before the last year of hire among women with breast cancer). The greater the restriction, the less left truncation bias we expect. We also examined premenopausal breast cancer as an outcome using age at diagnosis as an indicator of menopausal status. The statistical methods used were the same as those described for the main analysis. For these analyses, however, a series of 4 age cutpoints, 55, 54, 53, and 52 years, was used to define premenopausal breast cancer cases, and follow-up ended upon reaching that age. A separate model was made for each age cutpoint. The younger age cutpoints improved specificity but reduced sensitivity in identifying premenopausal breast cancer cases. For these analyses, we did not examine subcohorts defined by year of hire, because these cases were hired later. Among women with breast cancer diagnosed when they were age 55 years or younger, the earliest year of hire was 1966. Thus, left truncation bias is less of a concern here. No apparent violation of the underlying assumption of proportional hazards was detected based on correlations between the Schoenfeld residuals for each MWF of interest and the ranked failure times. SAS software, version 9.4 (SAS Institute, Inc., Cary, North Carolina) was used for all analyses except Cox models with penalized splines, which were conducted in R, version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria). Use of human participants’ data in this study was reviewed and approved by the Office for the Protection of Human Subjects at the University of California, Berkeley. RESULTS Table 1 lists demographic and exposure characteristics of the 4,503 female autoworkers, including the 221 women with breast cancer, who composed the study population. The cohort was predominantly white, but more than 25% of the women were black. The total number of active female workers in the 3 plants and their MWF exposure by year are presented in Figure 1. Slightly more than half of the women, 50.4%, were hired in 1974 or later. As of December 31, 1994, the last date of known employment, 39.0% of the women remained actively employed in 1 of the 3 plants. Table 1. Demographic and Exposure Characteristics of Cohort Members Who Were Alive in 1985, Including Women With Breast Cancer, United Autoworkers-General Motors Incidence Cohort, Michigan, 1985–2013 Characteristics  Women With Breast Cancer  Incidence Cohort  No.  %  Mean (Range)  Mean (SD)  No.  %  Mean (Range)  Mean (SD)  No. of subjects  221        4,503        No. of person-years  3,519        108,595        Year of birth      1938 (1907–1959)        1940 (1893–1961)    Year of hire      1969 (1942–1979)        1969 (1940–1981)    Age at hire, years      30.2 (18.0–55.1)        29.3 (16.7–58.0)    Duration of employment, years      18.1 (3.2–47.8)        16.2 (2.9–47.8)    Race                   White  155  70.1      3,256  72.3       Black  66  29.9      1,247  27.7      Plant                   Plant 1  25  11.3      509  11.3       Plant 2  122  55.2      2,612  58.0       Plant 3  74  33.5      1,382  30.7      Vital status as of 2013                   Alive  132  59.7      2,947  65.5       Deceased  89  40.3      1,556  34.6      Year of diagnosis      1999 (1985–2013)            Age at diagnosis, years      62.1 (34.8–94.1)            Cumulative MWF exposure, mg/m3-yearsa                   Straight        0.29 (1.08)        0.20 (1.75)   Soluble        0.68 (1.28)        0.51 (1.42)   Synthetic        0.09 (0.43)        0.08 (0.44)  Characteristics  Women With Breast Cancer  Incidence Cohort  No.  %  Mean (Range)  Mean (SD)  No.  %  Mean (Range)  Mean (SD)  No. of subjects  221        4,503        No. of person-years  3,519        108,595        Year of birth      1938 (1907–1959)        1940 (1893–1961)    Year of hire      1969 (1942–1979)        1969 (1940–1981)    Age at hire, years      30.2 (18.0–55.1)        29.3 (16.7–58.0)    Duration of employment, years      18.1 (3.2–47.8)        16.2 (2.9–47.8)    Race                   White  155  70.1      3,256  72.3       Black  66  29.9      1,247  27.7      Plant                   Plant 1  25  11.3      509  11.3       Plant 2  122  55.2      2,612  58.0       Plant 3  74  33.5      1,382  30.7      Vital status as of 2013                   Alive  132  59.7      2,947  65.5       Deceased  89  40.3      1,556  34.6      Year of diagnosis      1999 (1985–2013)            Age at diagnosis, years      62.1 (34.8–94.1)            Cumulative MWF exposure, mg/m3-yearsa                   Straight        0.29 (1.08)        0.20 (1.75)   Soluble        0.68 (1.28)        0.51 (1.42)   Synthetic        0.09 (0.43)        0.08 (0.44)  Abbreviation: MWF, metalworking fluid. a Cumulative exposure lagged 20 years. Figure 1. View largeDownload slide Number of active workers (gray vertical bars) and annual exposure prevalence to straight (blue line), soluble (red line), and synthetic (green line) metalworking fluids (MWF) by year among female members of the United Autoworker-General Motor incidence cohort who were alive in 1985 (n = 4,503), Michigan, 1940–1994. Figure 1. View largeDownload slide Number of active workers (gray vertical bars) and annual exposure prevalence to straight (blue line), soluble (red line), and synthetic (green line) metalworking fluids (MWF) by year among female members of the United Autoworker-General Motor incidence cohort who were alive in 1985 (n = 4,503), Michigan, 1940–1994. The percentages of women ever exposed to straight, soluble, and synthetic MWF were 53.7, 84.1, and 38.0, respectively. We observed elevated rates of incident breast cancer among those with greater straight MWF exposure, after reducing bias due to left truncation. When we examined the full study population, using continuous exposure, results were null for all fluid types (Table 2). In the analyses using subcohorts restricted by year of hire, however, the hazard ratios for continuous straight-MWF exposure were increasingly farther from the null with more restrictive subcohorts (i.e., with decreased time between hire and the start of follow-up). Using categorized exposure to examine the exposure-response relation revealed elevated hazard ratios, but with wide confidence intervals, for straight MWF in the full study population (Table 2). Results were stronger and presented a more positive exposure-response in the subcohort analyses. No elevated hazard ratios were observed for either soluble or synthetic MWF. Models with penalized spline coding for straight MWF exposure showed an approximately linear exposure-response relationship with the hazard ratio on the log scale, up to the 99th percentile of exposure among women with breast cancer in the 2 subcohorts restricted by year of hire (1959–1981 or 1969–1981) (Web Figure 1, available at https://academic.oup.com/aje). Table 2. Adjusted Hazard Ratiosa for Breast Cancer Incidence Relative to Cumulative Exposure to Metalworking Fluid in Female Autoworkers (n = 4,503) in the Full Cohort and in Subcohorts Defined by Year of Hire, United Autoworkers-General Motors Incidence Cohort, Michigan, 1985–2013 Cumulative Exposure by MWF Type, mg/m3-years  All Women With Breast Cancer (n = 221)  Restricted by Year of Hire  Hired 1959–1981 (n = 172)  Hired 1969–1981 (n = 157)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                           Straight                            0  124  74,802  1.00  Referent  106  69,019  1.00  Referent  95  65,022  1.00  Referent    0.0001–0.1120  48  17,440  1.39  0.90, 2.14  36  12,760  1.64  0.99, 2.73  33  11,888  1.67  0.99, 2.83    ≥0.1121  49  16,353  1.32  0.86, 2.01  30  9,251  1.64  1.01, 2.67  29  8,406  1.74  1.05, 2.86     P for trend      0.43      0.13      0.08   Soluble                            0  80  52,727  1.00  Referent  74  50,819  1.00  Referent  65  48,070  1.00  Referent    0.0001–0.4999  70  29,660  0.90  0.61, 1.34  61  25,716  0.90  0.58, 1.39  59  24,371  1.00  0.63, 1.59    ≥0.5000  71  26,208  0.88  0.56, 1.40  37  14,496  0.73  0.42, 1.26  33  12,875  0.77  0.43, 1.39     P for trend      0.87      0.34      0.33   Synthetic                            0  168  86,772  1.00  Referent  133  75,619  1.00  Referent  121  71,094  1.00  Referent    0.0001–0.0699  26  10,213  0.89  0.53, 1.49  22  9,099  0.72  0.40, 1.31  19  8,555  0.62  0.33, 1.17    ≥0.0700  27  11,610  0.77  0.47, 1.26  17  6,312  0.80  0.43, 1.48  17  5,667  0.87  0.46, 1.64     P for trend      0.39      0.71      0.96  Continuousb                           Straight      1.00  0.98, 1.02      1.11  1.01, 1.22      1.13  1.03, 1.23   Soluble      1.00  0.91, 1.09      0.97  0.74, 1.28      0.98  0.73, 1.30   Synthetic      0.98  0.90, 1.07      0.82  0.53, 1.29      0.91  0.60, 1.38  Cumulative Exposure by MWF Type, mg/m3-years  All Women With Breast Cancer (n = 221)  Restricted by Year of Hire  Hired 1959–1981 (n = 172)  Hired 1969–1981 (n = 157)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                           Straight                            0  124  74,802  1.00  Referent  106  69,019  1.00  Referent  95  65,022  1.00  Referent    0.0001–0.1120  48  17,440  1.39  0.90, 2.14  36  12,760  1.64  0.99, 2.73  33  11,888  1.67  0.99, 2.83    ≥0.1121  49  16,353  1.32  0.86, 2.01  30  9,251  1.64  1.01, 2.67  29  8,406  1.74  1.05, 2.86     P for trend      0.43      0.13      0.08   Soluble                            0  80  52,727  1.00  Referent  74  50,819  1.00  Referent  65  48,070  1.00  Referent    0.0001–0.4999  70  29,660  0.90  0.61, 1.34  61  25,716  0.90  0.58, 1.39  59  24,371  1.00  0.63, 1.59    ≥0.5000  71  26,208  0.88  0.56, 1.40  37  14,496  0.73  0.42, 1.26  33  12,875  0.77  0.43, 1.39     P for trend      0.87      0.34      0.33   Synthetic                            0  168  86,772  1.00  Referent  133  75,619  1.00  Referent  121  71,094  1.00  Referent    0.0001–0.0699  26  10,213  0.89  0.53, 1.49  22  9,099  0.72  0.40, 1.31  19  8,555  0.62  0.33, 1.17    ≥0.0700  27  11,610  0.77  0.47, 1.26  17  6,312  0.80  0.43, 1.48  17  5,667  0.87  0.46, 1.64     P for trend      0.39      0.71      0.96  Continuousb                           Straight      1.00  0.98, 1.02      1.11  1.01, 1.22      1.13  1.03, 1.23   Soluble      1.00  0.91, 1.09      0.97  0.74, 1.28      0.98  0.73, 1.30   Synthetic      0.98  0.90, 1.07      0.82  0.53, 1.29      0.91  0.60, 1.38  Abbreviations: CI, confidence interval; HR, hazard ratio; MWF, metalworking fluid. a Each Cox regression model included cumulative exposure to the 3 fluid types (size <3.5 μm; 20-year lag), used age as time scale, and was adjusted for year of hire, calendar year, race, and manufacturing plant. b HR per each 1-interquartile-range increase among women exposed to MWF in the full study population; for straight, soluble, and synthetic fluids, these are 0.318, 0.979, and 0.270 mg/m3-years, respectively. Results from the premenopausal breast cancer incidence analyses suggest an increased hazard associated with synthetic MWF exposure but null associations for straight and soluble MWF. The hazard ratio for continuous synthetic MWF exposure was elevated when using age 54 years as the cutpoint to define premenopausal cases and increased in magnitude with lower age cutpoints (Table 3). When using a categorical exposure metric, a strong exposure-response was observed for synthetic MWF, particularly with younger age cutpoints (Table 3). Models with penalized spline coding for synthetic MWF exposure showed a positive exposure-response relationship regardless of age cutpoint used through the 99th percentile of exposure among women with breast cancer (Figure 2). As with the other models, hazard ratios were higher when the premenopausal case definition was based on younger age cutpoints. Table 3. Adjusted Hazard Ratiosa of Premenopausal Breast Cancer Incidence Relative to Cumulative Exposure to Metalworking Fluid in Female Autoworkers, United Autoworkers-General Motors Incidence Cohort, Michigan, 1985–2013 Cumulative Exposure by MWF Type (mg/m3-years)  Age at Diagnosis  ≤55 Years (n = 3,263; n = 72 Women With Breast Cancer)  ≤54 Years (n = 3,211; n = 65 Women With Breast Cancer)  ≤53 Years (n = 3,148; n = 60 Women With Breast Cancer)  ≤52 Years (n = 3,092; n = 49 Women With Breast Cancer)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                                   Straight                                    0  54  48,105  1.00  Referent  48  46,151  1.00  Referent  45  44,131  1.00  Referent  39  42,082  1.00  Referent    0.0001–0.0799  9  5,470  0.95  0.36, 2.52  8  4,968  1.08  0.37, 3.10  7  4,488  0.82  0.26, 2.52  5  4,022  0.88  0.23, 3.31    ≥0.0800  9  4,208  1.08  0.44, 2.66  9  3,692  1.47  0.56, 3.82  8  3,196  1.23  0.45, 3.37  5  2,728  1.15  0.34, 3.91     P for trend      0.89      0.41      0.54      0.80   Soluble                                    0  42  39,347  1.00  Referent  39  38,184  1.00  Referent  36  36,963  1.00  Referent  33  35,675  1.00  Referent    0.0001–0.3319  15  11,070  0.79  0.37, 1.71  14  10,189  0.81  0.36, 1.85  13  9,324  0.81  0.35, 1.89  9  8,488  0.60  0.22, 1.63    ≥0.3320  15  7,366  0.83  0.33, 2.09  12  6,437  0.72  0.26, 2.00  11  5,529  0.76  0.27, 2.19  7  4,669  0.55  0.16, 1.88     P for trend      0.81      0.61      0.72      0.47   Synthetic                                    0  56  50,721  1.00  Referent  50  48,481  1.00  Referent  46  46,207  1.00  Referent  39  43,914  1.00  Referent    0.0001–0.0899  8  5,059  1.04  0.37, 2.96  8  4,601  1.16  0.39, 3.48  8  4,155  1.42  0.46, 4.39  5  3,725  1.33  0.34, 5.28    ≥0.0900  8  2,003  2.40  0.88, 6.54  7  1,729  2.71  0.92, 7.98  6  1,453  2.78  0.88, 8.73  5  1,193  3.76  1.04, 13.51     P for trend      0.07      0.05      0.09      0.04  Continuousb                                   Straight      0.81  0.51, 1.27      0.87  0.57, 1.35      0.87  0.54, 1.39      0.82  0.42, 1.58   Soluble      0.90  0.64, 1.27      0.91  0.59, 1.39      0.84  0.51, 1.38      0.75  0.39, 1.44   Synthetic      1.06  0.95, 1.17      1.13  0.99, 1.30    1.15  1.00, 1.32      1.17  1.01, 1.35  Cumulative Exposure by MWF Type (mg/m3-years)  Age at Diagnosis  ≤55 Years (n = 3,263; n = 72 Women With Breast Cancer)  ≤54 Years (n = 3,211; n = 65 Women With Breast Cancer)  ≤53 Years (n = 3,148; n = 60 Women With Breast Cancer)  ≤52 Years (n = 3,092; n = 49 Women With Breast Cancer)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                                   Straight                                    0  54  48,105  1.00  Referent  48  46,151  1.00  Referent  45  44,131  1.00  Referent  39  42,082  1.00  Referent    0.0001–0.0799  9  5,470  0.95  0.36, 2.52  8  4,968  1.08  0.37, 3.10  7  4,488  0.82  0.26, 2.52  5  4,022  0.88  0.23, 3.31    ≥0.0800  9  4,208  1.08  0.44, 2.66  9  3,692  1.47  0.56, 3.82  8  3,196  1.23  0.45, 3.37  5  2,728  1.15  0.34, 3.91     P for trend      0.89      0.41      0.54      0.80   Soluble                                    0  42  39,347  1.00  Referent  39  38,184  1.00  Referent  36  36,963  1.00  Referent  33  35,675  1.00  Referent    0.0001–0.3319  15  11,070  0.79  0.37, 1.71  14  10,189  0.81  0.36, 1.85  13  9,324  0.81  0.35, 1.89  9  8,488  0.60  0.22, 1.63    ≥0.3320  15  7,366  0.83  0.33, 2.09  12  6,437  0.72  0.26, 2.00  11  5,529  0.76  0.27, 2.19  7  4,669  0.55  0.16, 1.88     P for trend      0.81      0.61      0.72      0.47   Synthetic                                    0  56  50,721  1.00  Referent  50  48,481  1.00  Referent  46  46,207  1.00  Referent  39  43,914  1.00  Referent    0.0001–0.0899  8  5,059  1.04  0.37, 2.96  8  4,601  1.16  0.39, 3.48  8  4,155  1.42  0.46, 4.39  5  3,725  1.33  0.34, 5.28    ≥0.0900  8  2,003  2.40  0.88, 6.54  7  1,729  2.71  0.92, 7.98  6  1,453  2.78  0.88, 8.73  5  1,193  3.76  1.04, 13.51     P for trend      0.07      0.05      0.09      0.04  Continuousb                                   Straight      0.81  0.51, 1.27      0.87  0.57, 1.35      0.87  0.54, 1.39      0.82  0.42, 1.58   Soluble      0.90  0.64, 1.27      0.91  0.59, 1.39      0.84  0.51, 1.38      0.75  0.39, 1.44   Synthetic      1.06  0.95, 1.17      1.13  0.99, 1.30    1.15  1.00, 1.32      1.17  1.01, 1.35  Abbreviations: CI, confidence interval; HR, hazard ratio; MWF, metalworking fluid. a Each Cox regression model included cumulative exposure to the 3 fluid types (size <3.5 μm; 20-year lag), used age as time scale, and was adjusted for year of hire, calendar year, race, and manufacturing plant. Multiple age cutpoints were used to define premenopausal women. b HR per 1-interquartile-range increase among women exposed to MWF who were 55 years of age or younger; for straight, soluble, and synthetic fluids, these are 0.157, 0.474, and 0.092 mg/m3-years, respectively. Figure 2. View largeDownload slide Adjusted hazard ratios for premenopausal breast cancer incidence as a smoothed function of 20-year lagged cumulative exposure to synthetic metalworking fluid (MWF) as estimated in a Cox regression model using penalized splines (2 degrees of freedom) based on a cohort of female workers in the United Autoworker-General Motor incidence cohort who were alive in 1985 in Michigan, 1985–2013. Premenopausal breast cancer was defined as diagnosis by 4 select age cutpoints. The 95% confidence interval (dotted lines) is shown for the model using age 55 years as the cutpoint. Models used age as the time scale and were adjusted for cumulative exposure to straight and soluble MWF, year of hire, calendar year, race, and manufacturing plant. The graph is truncated at the 99th percentile of synthetic fluid exposure among women with breast cancer (0.60 mg/m3-years). The rug plot indicates exposure of the women with breast cancer who were 55 years of age or younger. Figure 2. View largeDownload slide Adjusted hazard ratios for premenopausal breast cancer incidence as a smoothed function of 20-year lagged cumulative exposure to synthetic metalworking fluid (MWF) as estimated in a Cox regression model using penalized splines (2 degrees of freedom) based on a cohort of female workers in the United Autoworker-General Motor incidence cohort who were alive in 1985 in Michigan, 1985–2013. Premenopausal breast cancer was defined as diagnosis by 4 select age cutpoints. The 95% confidence interval (dotted lines) is shown for the model using age 55 years as the cutpoint. Models used age as the time scale and were adjusted for cumulative exposure to straight and soluble MWF, year of hire, calendar year, race, and manufacturing plant. The graph is truncated at the 99th percentile of synthetic fluid exposure among women with breast cancer (0.60 mg/m3-years). The rug plot indicates exposure of the women with breast cancer who were 55 years of age or younger. DISCUSSION In a prospective cohort of 4,503 female autoworkers from the UAW-GM study, we examined MWF exposure and its association with incident breast cancer. Exposure to straight MWF, but neither soluble nor synthetic MWF, was positively associated with breast cancer. This was evident in the analyses using subcohorts restricted by year of hire. When restricted to women hired in 1959 or later, the hazard ratio for straight MWF exposure was elevated in the continuous and categorical analyses. Results became stronger when using the subcohort restricted to women hired in 1969 or later. This more restrictive subcohort was designed to reduce left truncation bias and may explain the stronger results observed in these analyses. Though power was adequate for the main analysis, when we restricted analysis to women presumed to be premenopausal, the number of women in each exposure category was sparse. Results, however, were modestly suggestive of an increased risk associated with higher levels of synthetic MWF exposure. The associations were more pronounced when younger age cutpoints defined the women with cancer. These definitions were more specific for premenopausal breast cancer, which is more important than sensitivity for reducing bias when the outcome is rare (33). Our interpretation of these results, however, was constrained by the small number of premenopausal women with breast cancer. Moreover, these results were apparently inconsistent with the inverse association with synthetic MWF observed in the main analysis. One possible explanation for this inconsistency is that synthetic MWF were introduced later. Exposed older women, therefore, had to have remained employed longer, introducing survivor bias among the older cases. Three previous breast cancer studies relied on data from the UAW-GM study (17, 22, 23). The first study combined incident cases and deaths, and imputed date of diagnosis, if it was missing (17). The modest risk increase observed for cumulative soluble MWF exposure was inconsistent with the literature on latency for breast cancer, suggesting higher risk associated with exposure in the decade preceding diagnosis. The present analysis was restricted to incident cases and we imposed a 20-year lag on cumulative exposures based on the latency period for breast cancer (34–36). The second study examined the incidence of several cancers and reported an increased and borderline significant hazard ratio associated with exposure to straight MWF among younger women in the cohort (23). In the third study, based on 43 deaths due to breast cancer, a hazard ratio of 1.4 (95% confidence interval: 0.7, 2.5) was reported associated with higher straight-MWF exposure (22). Death, however, is a poor surrogate for breast cancer incidence, given a 5-year survival rate of 88% (37). After accounting for left truncation, we observed a similar but marginally larger hazard ratio associated with straight-MWF exposure when using a continuous metric and a categorical metric. The additional 9 years of follow-up compared with the second study and the use of the more appropriate breast cancer outcome measure of incidence rather than death, compared with the third study, may have led to the slightly more positive results. The contrasting results in the pooled analysis versus the subanalysis restricted to premenopausal breast cancer suggest a possibly distinct biologic mechanism for this type of breast cancer. However, the distribution of cumulative straight-MWF exposure was markedly lower among women with premenopausal breast cancer, with a mean of 0.02 (standard deviation, 0.05) mg/m3-years based on an age cutpoint of 55 years compared with all cases (mean, 0.29 (standard deviation, 1.08) mg/m3-years), and may also have contributed to a lack of association for straight MWF in the premenopausal analyses. The PAH content of straight MWF may be the causal factor in the elevated risk. Air pollution exposure, as a proxy for PAH, has been associated with postmenopausal breast cancer. Total suspended particulate exposure was associated with postmenopausal, but not premenopausal, breast cancer in a population-based, case-control study in western New York (9). A later study in the same population modeled total PAH exposure and similarly found exposure at first birth associated with postmenopausal, but not premenopausal, breast cancer (10). The only study, to our knowledge, to examine occupational PAH exposure found an increased odds of premenopausal breast cancer among those exposed (35). Because benzene was a coexposure in the study, the authors additionally evaluated women exposed exclusively to PAH and found null results. Study results of the Long Island Breast Cancer Study Project indicated a positive association of PAH-DNA adducts, which are short-term PAH biomarkers of exposure, with incident breast cancer (38, 39). There is less evidence for synthetic fluids and premenopausal breast cancer risk. Nitrosamines are a class of potentially hazardous contaminants found in synthetic fluids (40). N-Nitrosodiethanolamine, a type of nitrosamine, has been found in synthetic fluids (41, 42) and in postshift urine samples of workers exposed to water-based MWF (43). It has been demonstrated to induce DNA damage in animal and human cells (44), and metal workers exposed to higher levels of N-nitrosodiethanolamine had more DNA single-strand breaks in blood cells (45). Although N-nitrosodiethanolamine is classified as possibly carcinogenic to humans and is an animal carcinogen (46), there is no specific research on this compound and breast cancer risk. Bias due to the healthy-worker survivor effect is a concern in occupational epidemiology (47–49). We recently assessed the presence of the healthy-worker survivor effect in the UAW-GM breast cancer incidence cohort based on 3 necessary underlying conditions: 1) leaving work predicts future exposure, 2) leaving work is associated with disease outcome, and 3) prior exposure increases probability of leaving work (50). The first condition is a given, because subjects who leave work are no longer exposed. We found prior soluble and synthetic, but not straight, MWF exposure associated with leaving work among female autoworkers. This supports the third condition for soluble and synthetic, but not straight, MWF. Although breast cancer was not examined in that assessment, we used the same statistical method and found that having left work was not associated with breast cancer incidence, indicating a lack of evidence for the second condition. Overall, these results imply that healthy-worker survivor bias did not influence our estimates of the association between MWF exposure and breast cancer risk, particularly for straight MWF. Our main limitation is unmeasured potential confounders. Although we controlled for age and race, we did not account for several other breast cancer risk factors, including socioeconomic status, family history of breast cancer, age at menarche or menopause, age at first full-term pregnancy, parity, breastfeeding history, use of oral contraceptives or postmenopausal hormones, and alcohol consumption. Because the women in our study cohort were all blue-collar workers employed at the same 3 plants, we did not expect large differences in socioeconomic status. Additionally, we did not expect family history of breast cancer, age at menarche or menopause, or use of oral contraceptives or postmenopausal hormones to be related to MWF exposure. Last, prior studies of this cohort found no association between MWF exposure and cirrhosis death, a proxy for alcohol consumption (51). Therefore, the potential confounders of most concern are those related to childbearing: age at first full-term pregnancy, parity, and breastfeeding history. Women who were younger at first full-term pregnancy, with higher parity, and who breastfed longer had reduced breast cancer risk and were likely to have lower MWF exposure due to taking maternity leave, entering the workforce later, and/or leaving the workforce earlier. This would produce positive confounding and bias results upward. In the present study, however, we saw no evidence of leaving work being associated with reduced breast cancer risk. Furthermore, we would expect such a bias to affect the results for all 3 fluid types, not just the straight MWF in the pooled analyses or the synthetic fluids in the premenopausal analyses. Although unmeasured confounding may account for some portion of the results observed, this study makes a contribution to the literature. This study adds to the limited literature regarding quantitative chemical exposures and breast cancer risk in humans. This study used a well-characterized occupational cohort of women working in automotive plants who had long cancer incidence follow-up and provided evidence of an increased risk of breast cancer associated with cumulative straight MWF exposure. ACKNOWLEDGMENTS Author affiliations: Environmental Health Sciences Division, School of Public Health, University of California at Berkeley, Berkeley, California (Erika Garcia, Ellen A. Eisen); and Epidemiology Division, School of Public Health, University of California at Berkeley, Berkeley, California (Patrick T. Bradshaw). This work was financially supported by the National Institute of Occupational Safety and Health at the Centers for Disease Control and Prevention (grants T42OH008429 and R01OH010028). Conflict of interest: none declared. Abbreviations MWF metalworking fluid PAH polycyclic aromatic hydrocarbons UAW-GM United Autoworker-General Motors REFERENCES 1 US Cancer Statistics Working Group. United States Cancer Statistics: 1999–2011 Incidence and Mortality Web-based Report. Atlanta, GA: Department of Health and Human Services, Center for Disease Control and Prevention, and National Cancer Institute; 2014. 2 SEER. Cancer Statistics Factsheets: Female Breast Cancer. Bethesda, MD: National Cancer Institute. http://seer.cancer.gov/statfacts/html/breast.html. Accessed November 7, 2016. 3 Madigan MP, Ziegler RG, Benichou J, et al.  . Proportion of breast cancer cases in the United States explained by well-established risk factors. J Natl Cancer Inst . 1995; 87( 22): 1681– 1685. Google Scholar CrossRef Search ADS PubMed  4 Institute of Medicine. Chapter 7: Recommendations for future research. Breast Cancer and the Environment: A Life Course Approach . Washington, DC: The National Academies Press; 2012: 325– 345. PubMed PubMed  5 Rudel RA, Attfield KR, Schifano JN, et al.  . Chemicals causing mammary gland tumors in animals signal new directions for epidemiology, chemicals testing, and risk assessment for breast cancer prevention. Cancer . 2007; 109( 12 suppl): 2635– 2666. Google Scholar CrossRef Search ADS PubMed  6 Neale M. Tribiology Handbook . New York, NY: John Wiley and Sons; 1973. 7 International Agency for Research on Cancer. Mineral Oils, Untreated or Mildly Treated. Lyon, France: International Agency for Research on Cancer; 2012. 8 Catchpole WM, Macmillan E, Powell H. Specifications for cutting oils with special reference to carcinogenicity. Ann Occup Hyg . 1971; 14( 2): 171– 179. Google Scholar PubMed  9 Bonner MR, Han D, Nie J, et al.  . Breast cancer risk and exposure in early life to polycyclic aromatic hydrocarbons using total suspended particulates as a proxy measure. Cancer Epidemiol Biomarkers Prev . 2005; 14( 1): 53– 60. Google Scholar PubMed  10 Nie J, Beyea J, Bonner MR, et al.  . Exposure to traffic emissions throughout life and risk of breast cancer: the Western New York Exposures and Breast Cancer (WEB) study. Cancer Causes Control . 2007; 18( 9): 947– 955. Google Scholar CrossRef Search ADS PubMed  11 Office of Chemical Safety and Pollution Prevention, Office of Water, and Office of Research and Development, US Environmental Protection Agency. Endocrine Disruptor Screening Program: Universe of Chemicals for Potential Endocrine Disruptor Screening and Testing. Washington, DC: US Environmental Protection Agency; 2012. 12 Steenland K, Burnett C, Lalich N, et al.  . Dying for work: the magnitude of US mortality from selected causes of death associated with occupation. Am J Ind Med . 2003; 43( 5): 461– 482. Google Scholar CrossRef Search ADS PubMed  13 International Labour Office. Global Employment Trends for Women: March 2009 . Geneva, Switzerland: International Labour Office; 2009. 14 Frost and Sullivan. Analysis of the Global Metalworking Fluids Market. Mountain View, CA; 2014. 15 Delzell E, Beall C, Macaluso M. Cancer mortality among women employed in motor vehicle manufacturing. J Occup Med . 1994; 36( 11): 1251– 1259. Google Scholar CrossRef Search ADS PubMed  16 Beall C, Delzell E, Macaluso M. Mortality patterns among women in the motor vehicle manufacturing industry. Am J Ind Med . 1995; 28( 3): 325– 337. Google Scholar CrossRef Search ADS PubMed  17 Thompson D, Kriebel D, Quinn MM, et al.  . Occupational exposure to metalworking fluids and risk of breast cancer among female autoworkers. Am J Ind Med . 2005; 47( 2): 153– 160. Google Scholar CrossRef Search ADS PubMed  18 Brophy JT, Keith MM, Gorey KM, et al.  . Occupation and breast cancer: a Canadian case-control study. Ann NY Acad Sci . 2006; 1076: 765– 777. Google Scholar CrossRef Search ADS PubMed  19 Hansen J. Breast cancer risk among relatively young women employed in solvent-using industries. Am J Ind Med . 1999; 36( 1): 43– 47. Google Scholar CrossRef Search ADS PubMed  20 Villeneuve S, Févotte J, Anger A, et al.  . Breast cancer risk by occupation and industry: analysis of the CECILE study, a population-based case-control study in France. Am J Ind Med . 2011; 54( 7): 499– 509. Google Scholar CrossRef Search ADS PubMed  21 Brophy JT, Keith MM, Watterson A, et al.  . Breast cancer risk in relation to occupations with exposure to carcinogens and endocrine disruptors: a Canadian case-control study. Environ Health . 2012; 11: 87. Google Scholar CrossRef Search ADS PubMed  22 Friesen MC, Betenia N, Costello S, et al.  . Metalworking fluid exposure and cancer risk in a retrospective cohort of female autoworkers. Cancer Causes Control . 2012; 23( 7): 1075– 1082. Google Scholar CrossRef Search ADS PubMed  23 Friesen MC, Costello S, Thurston SW, et al.  . Distinguishing the common components of oil- and water-based metalworking fluids for assessment of cancer incidence risk in autoworkers. Am J Ind Med . 2011; 54( 6): 450– 460. Google Scholar CrossRef Search ADS PubMed  24 Eisen EA, Tolbert PE, Monson RR, et al.  . Mortality studies of machining fluid exposure in the automobile industry I: a standardized mortality ratio analysis. Am J Ind Med . 1992; 22( 6): 809– 824. Google Scholar CrossRef Search ADS PubMed  25 Eisen EA, Bardin J, Gore R, et al.  . Exposure-response models based on extended follow-up of a cohort mortality study in the automobile industry. Scand J Work Environ Health . 2001; 27( 4): 240– 249. Google Scholar CrossRef Search ADS PubMed  26 Applebaum KM, Malloy EJ, Eisen EA. Left truncation, susceptibility, and bias in occupational cohort studies. Epidemiology . 2011; 22( 4): 599– 606. Google Scholar CrossRef Search ADS PubMed  27 Michigan Department of Health and Human Services. Michigan Cancer Surveillance Program. http://www.michigan.gov/mdch/0,4612,7-132-2945_5221-16586--,00.html. Accessed May 5, 2015. 28 Division of Cancer Prevention and Control, National Center for Chronic Disease Prevention and Health Promotion. National Program of Cancer Registries. Centers for Disease Control and Prevention; 2014. http://www.cdc.gov/cancer/npcr/. Accessed November 5, 2014. 29 Cowper DC, Kubal JD, Maynard C, et al.  . A primer and comparative review of major US mortality databases. Ann Epidemiol . 2002; 12( 7): 462– 468. Google Scholar CrossRef Search ADS PubMed  30 Hallock MF, Smith TJ, Woskie SR, et al.  . Estimation of historical exposures to machining fluids in the automotive industry. Am J Ind Med . 1994; 26( 5): 621– 634. Google Scholar CrossRef Search ADS PubMed  31 Woskie SR, Smith TJ, Hallock MF, et al.  . Size-selective pulmonary dose indices for metal-working fluid aerosols in machining and grinding operations in the automobile manufacturing industry. Am Ind Hyg Assoc J . 1994; 55( 1): 20– 29. Google Scholar CrossRef Search ADS PubMed  32 de Boor C. A Practical Guide to Splines . New York, NY: Springer-Verlag; 1978. Google Scholar CrossRef Search ADS   33 Rothman KJ, Greenland S, Lash TL. Modern Epidemiology . 3rd ed. Philadelphia, PA: Wolters Kluwer Health/Lippincott Williams & Wilkins; 2008. 34 Aschengrau A, Paulu C, Ozonoff D. Tetrachloroethylene-contaminated drinking water and the risk of breast cancer. Environ Health Perspect . 1998; 106( suppl 4): 947– 953. Google Scholar CrossRef Search ADS PubMed  35 Petralia SA, Vena JE, Freudenheim JL, et al.  . Risk of premenopausal breast cancer in association with occupational exposure to polycyclic aromatic hydrocarbons and benzene. Scand J Work Environ Health . 1999; 25( 3): 215– 221. Google Scholar CrossRef Search ADS PubMed  36 Kelsey JL, Berkowitz GS. Breast cancer epidemiology. Cancer Res . 1988; 48( 20): 5615– 5623. Google Scholar PubMed  37 American Cancer Society. Cancer Facts & Figures 2014. Atlanta, GA: American Cancer Society; 2014. 38 Gammon MD, Sagiv SK, Eng SM, et al.  . Polycyclic aromatic hydrocarbon-DNA adducts and breast cancer: a pooled analysis. Arch Environ Health . 2004; 59( 12): 640– 649. Google Scholar CrossRef Search ADS PubMed  39 Gammon MD, Santella RM, Neugut AI, et al.  . Environmental toxins and breast cancer on Long Island. I. Polycyclic aromatic hydrocarbon DNA adducts. Cancer Epidemiol Biomarkers Prev . 2002; 11( 8): 677– 685. Google Scholar PubMed  40 National Institute for Occupational Safety and Health, US Centers for Disease Control and Prevention. Criteria for a Recommended Standard: Occupational Exposure to Metalworking Fluids. Cincinnati, OH: US Department of Health and Human Services; 1998. 41 Fan TY, Morrison J, Rounbehler DP, et al.  . N-Nitrosodiethanolamine in synthetic cutting fluids: a part-per-hundred impurity. Science . 1977; 196( 4285): 70– 71. Google Scholar CrossRef Search ADS PubMed  42 Keefer LK, Goff U, Stevens J, et al.  . Persistence of N-nitrosodiethanolamine contamination in American metal-working lubricants. Food Chem Toxicol . 1990; 28( 7): 531– 534. Google Scholar CrossRef Search ADS PubMed  43 Ducos P, Gaudin R. N-Nitrosodiethanolamine urinary excretion in workers exposed to aqueous metalworking fluids. Int Arch Occup Environ Health . 2003; 76( 8): 591– 597. Google Scholar CrossRef Search ADS PubMed  44 Holzer J, Voss B, Karroum S, et al.  . A comparative study of chemically induced DNA damage in isolated nasal mucosa cells of humans and rats assessed by the alkaline comet assay. J Toxicol Environ Health A . 2008; 71( 13– 14): 936– 946. Google Scholar CrossRef Search ADS PubMed  45 Fuchs J, Burg J, Hengstler JG, et al.  . DNA damage in mononuclear blood cells of metal workers exposed to N-nitrosodiethanolamine in synthetic cutting fluids. Mutat Res . 1995; 342( 1– 2): 95– 102. Google Scholar CrossRef Search ADS PubMed  46 International Agency for Research on Cancer. Some Industrial Chemicals. Lyon, France: International Agency for Research on Cancer; 2000. 47 Arrighi HM, Hertz-Picciotto I. The evolving concept of the healthy worker survivor effect. Epidemiology . 1994; 5( 2): 189– 196. Google Scholar CrossRef Search ADS PubMed  48 Buckley JP, Keil AP, McGrath LJ, et al.  . Evolving methods for inference in the presence of healthy worker survivor bias. Epidemiology . 2015; 26( 2): 204– 212. Google Scholar CrossRef Search ADS PubMed  49 Picciotto S, Hertz-Picciotto I. Commentary: healthy worker survivor bias: a still-evolving concept. Epidemiology . 2015; 26( 2): 213– 215. Google Scholar CrossRef Search ADS PubMed  50 Garcia E, Picciotto S, Costello S, et al.  . Assessment of the healthy worker survivor effect in cancer studies of the United Autoworkers-General Motors cohort. Occup Environ Med . 2017; 74( 4): 294– 300. Google Scholar CrossRef Search ADS PubMed  51 Eisen EA, Tolbert PE, Hallock MF, et al.  . Mortality studies of machining fluid exposure in the automobile industry. III: A case-control study of larynx cancer. Am J Ind Med . 1994; 26( 2): 185– 202. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Epidemiology Oxford University Press

Breast Cancer Incidence and Exposure to Metalworking Fluid in a Cohort of Female Autoworkers

Loading next page...
 
/lp/ou_press/breast-cancer-incidence-and-exposure-to-metalworking-fluid-in-a-cohort-Ge9hFEKzly
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0002-9262
eISSN
1476-6256
D.O.I.
10.1093/aje/kwx264
Publisher site
See Article on Publisher Site

Abstract

Abstract Breast cancer is the leading cancer diagnosed among women, and environmental studies have produced few leads on modifiable risk factors for breast cancer. Following an Institute of Medicine recommendation for occupational studies of women highly exposed to potential breast cancer risk factors, we took advantage of an existing cohort of 4,503 female autoworkers in Michigan exposed to metalworking fluid (MWF), complex mixtures of oils and chemicals widely used in metal manufacturing worldwide. Cox proportional hazards models were fit to estimate hazard ratios for incident breast cancer (follow-up, 1985–2013) and cumulative exposure (20-year lag) to straight mineral oils (a known human carcinogen) and water-based soluble and synthetic MWF. Because the state cancer registry began decades after the cohort was defined, we restricted our analyses to subcohorts of women hired closer to the start of follow-up. Among those hired after 1969, the hazard ratio associated with a 1 interquartile-range increase in straight MWF exposure was 1.13 (95% confidence interval: 1.03, 1.23). In separate analyses of premenopausal breast cancer, defined by age at diagnosis, the hazard ratio was elevated for exposure to synthetic MWF (chemical lubricants with no oil content), possibly suggesting a different mechanism in the younger women with breast cancer. This study adds to the limited literature regarding quantitative chemical exposures and breast cancer risk. breast cancer, metalworking fluid, occupational exposure Breast cancer is the leading cancer diagnosed in women in the United States (1). It is estimated that, in 2016, there will be 246,660 new cases and 40,450 deaths attributable to breast cancer among women in the United States (2). Despite the high numbers of women affected by this disease, well-established risk factors are estimated to account for only 41% of breast cancer cases (3), and studies of environmental exposures associated with breast cancer risk have produced few additional leads on modifiable risk factors over the past 20 years (4). A 2012 Institute of Medicine review of environmental risk factors for breast cancer noted that the initial identifications of many known human carcinogens were based on studies of high exposures in occupational settings and recommended additional breast cancer studies of worker populations (4). Despite provocative evidence that metalworking fluid (MWF) contaminants cause mammary gland tumors in laboratory animals (5) and that occupational MWF-exposure levels are appreciable, few studies have been conducted on associated breast cancer risk. MWF are coolants and lubricants widely used in industrial machining and grinding operations, and are categorized into 3 classes on the basis of composition: straight, soluble, and synthetic. Straight MWF are complex mixtures of paraffinic, naphthenic, and aromatic compounds refined from mineral oil (6). The carcinogenic properties of mineral oil, classified as carcinogenic to humans (7), are thought to be due primarily to their polycyclic aromatic hydrocarbon (PAH) content (8). Soluble MWF are oils emulsified in water. Synthetic fluids are water-soluble chemical lubricants without oil. Ethanolamines and nitrites, added to synthetic MWF to inhibit corrosion and adjust pH, interact to form nitrosamines. Although, to our knowledge, there have been no toxicologic studies in which investigators specifically examined the association of MWF and breast cancer, MWF components including PAH and nitrosamines have been implicated as mammary gland carcinogens (5). Additionally, results of studies on ambient air PAH exposure are suggestive of increased breast cancer risk (9, 10), and nitrosamines are suspected endocrine disruptors (11). Women comprise a growing proportion of the 4.4 million US workers potentially exposed to MWF (12), as well as of the growing workforce employed in metal manufacturing worldwide. The global manufacturing workforce is estimated to be about 30% female (13). The global market volume of MWF in 2013 was 2.3 million tons; Asia had the largest share, 41.5%; North America’s share was 28.0% (14). Few epidemiologic studies on breast cancer risk and MWF exposure exist (15–22). Results of an early study indicated reduced incidence of death due to breast cancer among women working in automotive manufacturing, based on standardized mortality ratios (15, 16). More recently, researchers conducting case-control studies reported elevated breast cancer risk associated with occupation in automotive manufacturing and metal products/metal work (19–21). These studies, however, did not examine a specific compound; instead, occupation was used as a proxy for exposure. Three breast cancer studies have been conducted in the United Autoworker-General Motors (UAW-GM) cohort; 2 were limited in power and relied on either a combination of incident cases and deaths (17) or only deaths (22). In the 1 study in which investigators examined only incident breast cancer cases, Friesen et al. (23) reported a slightly elevated risk associated with exposure to straight MWF. We took advantage of this large cohort of female autoworkers exposed to MWF to examine the relationship between quantitative MWF exposure and breast cancer incidence. We have 9 additional years of follow-up and restricted our analysis to registry-identified incident cases. The aim of this study was to examine the exposure-response relationship between cumulative MWF exposure and breast cancer risk in a cohort of occupationally exposed female autoworkers. METHODS Study population The UAW-GM study was a joint labor-management funded study designed in 1984 to examine cancer mortality and its relation to MWF exposure. This cohort has been described elsewhere (24, 25). Briefly, the original study included 46,316 hourly workers from 1 of 3 automobile manufacturing plants in Michigan who were exposed to MWF primarily via inhalation (24). All hourly employees who had worked at least 3 years before January 1, 1985, were included in the cohort. Cohort participants who were alive on January 1, 1985, when the Michigan Cancer Registry began (n = 33,915), were included in the incidence cohort. Analyses for the present study were restricted to female autoworkers in the incidence cohort (n = 4,567). We excluded women who were missing more than 50% of their employment history (n = 59) or were hired before 1938 (n = 5). This latter restriction was imposed to address potential left truncation bias in the main cohort (26). The final study population comprised 4,503 female workers. Outcome assessment The UAW-GM incidence cohort was linked with the Michigan Cancer Registry to identify incident cancer cases diagnosed between January 1, 1985, and December 31, 2013. Data are collected by the Michigan Department of Community Health as part of the Michigan Cancer Surveillance Program (27), which participates in the National Program of Cancer Registries of the Centers for Disease Control and Prevention (28). We obtained data on first diagnosis of primary breast cancer (International Classification of Disease for Oncology, Third Edition, codes C50.0–C50.9), including in situ and invasive tumors, since 1985. We identified 221 incident cases of breast cancer in the study population, including 72 cases diagnosed when the woman was age 55 years or younger. Most of the women with breast cancer had invasive tumors (82.8%). Data on vital status were obtained from the National Death Index (National Center for Health Statistics, Hyattsville, Maryland) (29). Exposure assessment Cumulative exposure estimates for each MWF type were calculated for each woman in the UAW-GM cohort on the basis of detailed employment records available from hire through 1994 and of a time-varying job-exposure matrix. An extensive retrospective exposure assessment was conducted to develop the job-exposure matrix (30, 31). Size-fractionated MWF concentrations were estimated as an 8-hour time-weighted average (milligrams per cubic meter) based on several hundred personal and area airborne-exposure measurements collected by study industrial hygienists during the mid-1980s. A set of multipliers to adjust MWF concentration for temporal trends was developed on the basis of nearly 400 historical air-sampling measurements (collected from 1958 through 1987), review of historical records, and interviews with plant personnel, and were last updated in 1995. For this study, we used the respirable size fraction (<3.5 μm), which mostly deposits in the alveolar region, of the MWF exposure estimates. The job-exposure matrix was combined with employment records to estimate time-varying, annual average daily exposure to straight, soluble, and synthetic MWF (milligrams per cubic meter) throughout the women’s entire employment. Missing employment information was interpolated by averaging exposures from previous and subsequent jobs. We then calculated cumulative time-weighted exposure to the 3 fluid types (reported as milligrams per cubic meter–years) for each subject for the duration of follow-up. To account for breast cancer latency, cumulative exposures for each fluid type were lagged; because employment data ended December 31, 1994, 19 years before end of follow-up, we used a 20-year lag. This assumes exposure accumulated in the 20 years preceding a diagnosis did not affect cancer risk. Data analyses Cox proportional hazards models were fit to estimate hazard ratios and 95% confidence intervals associated with exposure to straight, soluble, and synthetic MWF (20-year lag) on breast cancer incidence. Exposure was defined as cumulative exposure to each fluid type for the respirable size fraction. MWF exposures were modeled as continuous and categorical. For continuous exposure, hazard ratios were scaled to a 1-interquartile-range increase among women with MWF exposure in the full study population; for the 3 fluid types, straight, soluble, and synthetic, these were 0.318, 0.979, and 0.270 mg/m3-years, respectively. For categorical exposure, the referent groups were women with no MWF exposure. For each fluid type, the median level among women with breast cancer who were exposed to the fluid was used as the cutpoints for the exposure categories. Tests for trend were conducted by setting the value to the median for exposure category, modeling exposure as a continuous variable, and testing for nonzero slope by using a likelihood ratio test. Age was the time metric for all models. All models included baseline covariates for race (white or black), an established breast cancer risk factor, manufacturing plant (to adjust for plant-specific characteristic not captured elsewhere), socioeconomic status, regional difference, and year of hire (B-spline with 3 degrees of freedom and equally spaced knots (32)) to account for secular trends in exposure, including PAH content, personal protective equipment, and plant ventilation. A time-varying covariate for calendar year (B-spline with 3 degrees of freedom and equally spaced knots) was included to account for secular trends in breast cancer diagnosis. To assess the linearity of exposure-response relationships, we evaluated additional Cox models with penalized splines (2 degrees of freedom) for MWF exposure. By necessity, only UAW-GM cohort members alive on January 1, 1985, were included in the incidence cohort, thereby creating a left-truncated cohort. Left truncation occurs when not all otherwise eligible persons are enrolled in the cohort. In this study, downward bias arises from left truncation because the proportion of women susceptible to the effect of exposure decreased over time (26). To reduce this potential bias, we restricted analyses to a series of subcohorts defined by year of hire. Narrowing the interval between hire and start of the cancer registry reduced the opportunity for susceptible women to die before start of follow-up in 1985. We defined 2 subcohorts: those hired in or after 1959 (1959–1981) and those hired in or after 1969 (1969–1981) (20 and 10 years, respectively, before the last year of hire among women with breast cancer). The greater the restriction, the less left truncation bias we expect. We also examined premenopausal breast cancer as an outcome using age at diagnosis as an indicator of menopausal status. The statistical methods used were the same as those described for the main analysis. For these analyses, however, a series of 4 age cutpoints, 55, 54, 53, and 52 years, was used to define premenopausal breast cancer cases, and follow-up ended upon reaching that age. A separate model was made for each age cutpoint. The younger age cutpoints improved specificity but reduced sensitivity in identifying premenopausal breast cancer cases. For these analyses, we did not examine subcohorts defined by year of hire, because these cases were hired later. Among women with breast cancer diagnosed when they were age 55 years or younger, the earliest year of hire was 1966. Thus, left truncation bias is less of a concern here. No apparent violation of the underlying assumption of proportional hazards was detected based on correlations between the Schoenfeld residuals for each MWF of interest and the ranked failure times. SAS software, version 9.4 (SAS Institute, Inc., Cary, North Carolina) was used for all analyses except Cox models with penalized splines, which were conducted in R, version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria). Use of human participants’ data in this study was reviewed and approved by the Office for the Protection of Human Subjects at the University of California, Berkeley. RESULTS Table 1 lists demographic and exposure characteristics of the 4,503 female autoworkers, including the 221 women with breast cancer, who composed the study population. The cohort was predominantly white, but more than 25% of the women were black. The total number of active female workers in the 3 plants and their MWF exposure by year are presented in Figure 1. Slightly more than half of the women, 50.4%, were hired in 1974 or later. As of December 31, 1994, the last date of known employment, 39.0% of the women remained actively employed in 1 of the 3 plants. Table 1. Demographic and Exposure Characteristics of Cohort Members Who Were Alive in 1985, Including Women With Breast Cancer, United Autoworkers-General Motors Incidence Cohort, Michigan, 1985–2013 Characteristics  Women With Breast Cancer  Incidence Cohort  No.  %  Mean (Range)  Mean (SD)  No.  %  Mean (Range)  Mean (SD)  No. of subjects  221        4,503        No. of person-years  3,519        108,595        Year of birth      1938 (1907–1959)        1940 (1893–1961)    Year of hire      1969 (1942–1979)        1969 (1940–1981)    Age at hire, years      30.2 (18.0–55.1)        29.3 (16.7–58.0)    Duration of employment, years      18.1 (3.2–47.8)        16.2 (2.9–47.8)    Race                   White  155  70.1      3,256  72.3       Black  66  29.9      1,247  27.7      Plant                   Plant 1  25  11.3      509  11.3       Plant 2  122  55.2      2,612  58.0       Plant 3  74  33.5      1,382  30.7      Vital status as of 2013                   Alive  132  59.7      2,947  65.5       Deceased  89  40.3      1,556  34.6      Year of diagnosis      1999 (1985–2013)            Age at diagnosis, years      62.1 (34.8–94.1)            Cumulative MWF exposure, mg/m3-yearsa                   Straight        0.29 (1.08)        0.20 (1.75)   Soluble        0.68 (1.28)        0.51 (1.42)   Synthetic        0.09 (0.43)        0.08 (0.44)  Characteristics  Women With Breast Cancer  Incidence Cohort  No.  %  Mean (Range)  Mean (SD)  No.  %  Mean (Range)  Mean (SD)  No. of subjects  221        4,503        No. of person-years  3,519        108,595        Year of birth      1938 (1907–1959)        1940 (1893–1961)    Year of hire      1969 (1942–1979)        1969 (1940–1981)    Age at hire, years      30.2 (18.0–55.1)        29.3 (16.7–58.0)    Duration of employment, years      18.1 (3.2–47.8)        16.2 (2.9–47.8)    Race                   White  155  70.1      3,256  72.3       Black  66  29.9      1,247  27.7      Plant                   Plant 1  25  11.3      509  11.3       Plant 2  122  55.2      2,612  58.0       Plant 3  74  33.5      1,382  30.7      Vital status as of 2013                   Alive  132  59.7      2,947  65.5       Deceased  89  40.3      1,556  34.6      Year of diagnosis      1999 (1985–2013)            Age at diagnosis, years      62.1 (34.8–94.1)            Cumulative MWF exposure, mg/m3-yearsa                   Straight        0.29 (1.08)        0.20 (1.75)   Soluble        0.68 (1.28)        0.51 (1.42)   Synthetic        0.09 (0.43)        0.08 (0.44)  Abbreviation: MWF, metalworking fluid. a Cumulative exposure lagged 20 years. Figure 1. View largeDownload slide Number of active workers (gray vertical bars) and annual exposure prevalence to straight (blue line), soluble (red line), and synthetic (green line) metalworking fluids (MWF) by year among female members of the United Autoworker-General Motor incidence cohort who were alive in 1985 (n = 4,503), Michigan, 1940–1994. Figure 1. View largeDownload slide Number of active workers (gray vertical bars) and annual exposure prevalence to straight (blue line), soluble (red line), and synthetic (green line) metalworking fluids (MWF) by year among female members of the United Autoworker-General Motor incidence cohort who were alive in 1985 (n = 4,503), Michigan, 1940–1994. The percentages of women ever exposed to straight, soluble, and synthetic MWF were 53.7, 84.1, and 38.0, respectively. We observed elevated rates of incident breast cancer among those with greater straight MWF exposure, after reducing bias due to left truncation. When we examined the full study population, using continuous exposure, results were null for all fluid types (Table 2). In the analyses using subcohorts restricted by year of hire, however, the hazard ratios for continuous straight-MWF exposure were increasingly farther from the null with more restrictive subcohorts (i.e., with decreased time between hire and the start of follow-up). Using categorized exposure to examine the exposure-response relation revealed elevated hazard ratios, but with wide confidence intervals, for straight MWF in the full study population (Table 2). Results were stronger and presented a more positive exposure-response in the subcohort analyses. No elevated hazard ratios were observed for either soluble or synthetic MWF. Models with penalized spline coding for straight MWF exposure showed an approximately linear exposure-response relationship with the hazard ratio on the log scale, up to the 99th percentile of exposure among women with breast cancer in the 2 subcohorts restricted by year of hire (1959–1981 or 1969–1981) (Web Figure 1, available at https://academic.oup.com/aje). Table 2. Adjusted Hazard Ratiosa for Breast Cancer Incidence Relative to Cumulative Exposure to Metalworking Fluid in Female Autoworkers (n = 4,503) in the Full Cohort and in Subcohorts Defined by Year of Hire, United Autoworkers-General Motors Incidence Cohort, Michigan, 1985–2013 Cumulative Exposure by MWF Type, mg/m3-years  All Women With Breast Cancer (n = 221)  Restricted by Year of Hire  Hired 1959–1981 (n = 172)  Hired 1969–1981 (n = 157)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                           Straight                            0  124  74,802  1.00  Referent  106  69,019  1.00  Referent  95  65,022  1.00  Referent    0.0001–0.1120  48  17,440  1.39  0.90, 2.14  36  12,760  1.64  0.99, 2.73  33  11,888  1.67  0.99, 2.83    ≥0.1121  49  16,353  1.32  0.86, 2.01  30  9,251  1.64  1.01, 2.67  29  8,406  1.74  1.05, 2.86     P for trend      0.43      0.13      0.08   Soluble                            0  80  52,727  1.00  Referent  74  50,819  1.00  Referent  65  48,070  1.00  Referent    0.0001–0.4999  70  29,660  0.90  0.61, 1.34  61  25,716  0.90  0.58, 1.39  59  24,371  1.00  0.63, 1.59    ≥0.5000  71  26,208  0.88  0.56, 1.40  37  14,496  0.73  0.42, 1.26  33  12,875  0.77  0.43, 1.39     P for trend      0.87      0.34      0.33   Synthetic                            0  168  86,772  1.00  Referent  133  75,619  1.00  Referent  121  71,094  1.00  Referent    0.0001–0.0699  26  10,213  0.89  0.53, 1.49  22  9,099  0.72  0.40, 1.31  19  8,555  0.62  0.33, 1.17    ≥0.0700  27  11,610  0.77  0.47, 1.26  17  6,312  0.80  0.43, 1.48  17  5,667  0.87  0.46, 1.64     P for trend      0.39      0.71      0.96  Continuousb                           Straight      1.00  0.98, 1.02      1.11  1.01, 1.22      1.13  1.03, 1.23   Soluble      1.00  0.91, 1.09      0.97  0.74, 1.28      0.98  0.73, 1.30   Synthetic      0.98  0.90, 1.07      0.82  0.53, 1.29      0.91  0.60, 1.38  Cumulative Exposure by MWF Type, mg/m3-years  All Women With Breast Cancer (n = 221)  Restricted by Year of Hire  Hired 1959–1981 (n = 172)  Hired 1969–1981 (n = 157)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                           Straight                            0  124  74,802  1.00  Referent  106  69,019  1.00  Referent  95  65,022  1.00  Referent    0.0001–0.1120  48  17,440  1.39  0.90, 2.14  36  12,760  1.64  0.99, 2.73  33  11,888  1.67  0.99, 2.83    ≥0.1121  49  16,353  1.32  0.86, 2.01  30  9,251  1.64  1.01, 2.67  29  8,406  1.74  1.05, 2.86     P for trend      0.43      0.13      0.08   Soluble                            0  80  52,727  1.00  Referent  74  50,819  1.00  Referent  65  48,070  1.00  Referent    0.0001–0.4999  70  29,660  0.90  0.61, 1.34  61  25,716  0.90  0.58, 1.39  59  24,371  1.00  0.63, 1.59    ≥0.5000  71  26,208  0.88  0.56, 1.40  37  14,496  0.73  0.42, 1.26  33  12,875  0.77  0.43, 1.39     P for trend      0.87      0.34      0.33   Synthetic                            0  168  86,772  1.00  Referent  133  75,619  1.00  Referent  121  71,094  1.00  Referent    0.0001–0.0699  26  10,213  0.89  0.53, 1.49  22  9,099  0.72  0.40, 1.31  19  8,555  0.62  0.33, 1.17    ≥0.0700  27  11,610  0.77  0.47, 1.26  17  6,312  0.80  0.43, 1.48  17  5,667  0.87  0.46, 1.64     P for trend      0.39      0.71      0.96  Continuousb                           Straight      1.00  0.98, 1.02      1.11  1.01, 1.22      1.13  1.03, 1.23   Soluble      1.00  0.91, 1.09      0.97  0.74, 1.28      0.98  0.73, 1.30   Synthetic      0.98  0.90, 1.07      0.82  0.53, 1.29      0.91  0.60, 1.38  Abbreviations: CI, confidence interval; HR, hazard ratio; MWF, metalworking fluid. a Each Cox regression model included cumulative exposure to the 3 fluid types (size <3.5 μm; 20-year lag), used age as time scale, and was adjusted for year of hire, calendar year, race, and manufacturing plant. b HR per each 1-interquartile-range increase among women exposed to MWF in the full study population; for straight, soluble, and synthetic fluids, these are 0.318, 0.979, and 0.270 mg/m3-years, respectively. Results from the premenopausal breast cancer incidence analyses suggest an increased hazard associated with synthetic MWF exposure but null associations for straight and soluble MWF. The hazard ratio for continuous synthetic MWF exposure was elevated when using age 54 years as the cutpoint to define premenopausal cases and increased in magnitude with lower age cutpoints (Table 3). When using a categorical exposure metric, a strong exposure-response was observed for synthetic MWF, particularly with younger age cutpoints (Table 3). Models with penalized spline coding for synthetic MWF exposure showed a positive exposure-response relationship regardless of age cutpoint used through the 99th percentile of exposure among women with breast cancer (Figure 2). As with the other models, hazard ratios were higher when the premenopausal case definition was based on younger age cutpoints. Table 3. Adjusted Hazard Ratiosa of Premenopausal Breast Cancer Incidence Relative to Cumulative Exposure to Metalworking Fluid in Female Autoworkers, United Autoworkers-General Motors Incidence Cohort, Michigan, 1985–2013 Cumulative Exposure by MWF Type (mg/m3-years)  Age at Diagnosis  ≤55 Years (n = 3,263; n = 72 Women With Breast Cancer)  ≤54 Years (n = 3,211; n = 65 Women With Breast Cancer)  ≤53 Years (n = 3,148; n = 60 Women With Breast Cancer)  ≤52 Years (n = 3,092; n = 49 Women With Breast Cancer)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                                   Straight                                    0  54  48,105  1.00  Referent  48  46,151  1.00  Referent  45  44,131  1.00  Referent  39  42,082  1.00  Referent    0.0001–0.0799  9  5,470  0.95  0.36, 2.52  8  4,968  1.08  0.37, 3.10  7  4,488  0.82  0.26, 2.52  5  4,022  0.88  0.23, 3.31    ≥0.0800  9  4,208  1.08  0.44, 2.66  9  3,692  1.47  0.56, 3.82  8  3,196  1.23  0.45, 3.37  5  2,728  1.15  0.34, 3.91     P for trend      0.89      0.41      0.54      0.80   Soluble                                    0  42  39,347  1.00  Referent  39  38,184  1.00  Referent  36  36,963  1.00  Referent  33  35,675  1.00  Referent    0.0001–0.3319  15  11,070  0.79  0.37, 1.71  14  10,189  0.81  0.36, 1.85  13  9,324  0.81  0.35, 1.89  9  8,488  0.60  0.22, 1.63    ≥0.3320  15  7,366  0.83  0.33, 2.09  12  6,437  0.72  0.26, 2.00  11  5,529  0.76  0.27, 2.19  7  4,669  0.55  0.16, 1.88     P for trend      0.81      0.61      0.72      0.47   Synthetic                                    0  56  50,721  1.00  Referent  50  48,481  1.00  Referent  46  46,207  1.00  Referent  39  43,914  1.00  Referent    0.0001–0.0899  8  5,059  1.04  0.37, 2.96  8  4,601  1.16  0.39, 3.48  8  4,155  1.42  0.46, 4.39  5  3,725  1.33  0.34, 5.28    ≥0.0900  8  2,003  2.40  0.88, 6.54  7  1,729  2.71  0.92, 7.98  6  1,453  2.78  0.88, 8.73  5  1,193  3.76  1.04, 13.51     P for trend      0.07      0.05      0.09      0.04  Continuousb                                   Straight      0.81  0.51, 1.27      0.87  0.57, 1.35      0.87  0.54, 1.39      0.82  0.42, 1.58   Soluble      0.90  0.64, 1.27      0.91  0.59, 1.39      0.84  0.51, 1.38      0.75  0.39, 1.44   Synthetic      1.06  0.95, 1.17      1.13  0.99, 1.30    1.15  1.00, 1.32      1.17  1.01, 1.35  Cumulative Exposure by MWF Type (mg/m3-years)  Age at Diagnosis  ≤55 Years (n = 3,263; n = 72 Women With Breast Cancer)  ≤54 Years (n = 3,211; n = 65 Women With Breast Cancer)  ≤53 Years (n = 3,148; n = 60 Women With Breast Cancer)  ≤52 Years (n = 3,092; n = 49 Women With Breast Cancer)  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  No.  Person-Years  HR  95% CI  Categorical                                   Straight                                    0  54  48,105  1.00  Referent  48  46,151  1.00  Referent  45  44,131  1.00  Referent  39  42,082  1.00  Referent    0.0001–0.0799  9  5,470  0.95  0.36, 2.52  8  4,968  1.08  0.37, 3.10  7  4,488  0.82  0.26, 2.52  5  4,022  0.88  0.23, 3.31    ≥0.0800  9  4,208  1.08  0.44, 2.66  9  3,692  1.47  0.56, 3.82  8  3,196  1.23  0.45, 3.37  5  2,728  1.15  0.34, 3.91     P for trend      0.89      0.41      0.54      0.80   Soluble                                    0  42  39,347  1.00  Referent  39  38,184  1.00  Referent  36  36,963  1.00  Referent  33  35,675  1.00  Referent    0.0001–0.3319  15  11,070  0.79  0.37, 1.71  14  10,189  0.81  0.36, 1.85  13  9,324  0.81  0.35, 1.89  9  8,488  0.60  0.22, 1.63    ≥0.3320  15  7,366  0.83  0.33, 2.09  12  6,437  0.72  0.26, 2.00  11  5,529  0.76  0.27, 2.19  7  4,669  0.55  0.16, 1.88     P for trend      0.81      0.61      0.72      0.47   Synthetic                                    0  56  50,721  1.00  Referent  50  48,481  1.00  Referent  46  46,207  1.00  Referent  39  43,914  1.00  Referent    0.0001–0.0899  8  5,059  1.04  0.37, 2.96  8  4,601  1.16  0.39, 3.48  8  4,155  1.42  0.46, 4.39  5  3,725  1.33  0.34, 5.28    ≥0.0900  8  2,003  2.40  0.88, 6.54  7  1,729  2.71  0.92, 7.98  6  1,453  2.78  0.88, 8.73  5  1,193  3.76  1.04, 13.51     P for trend      0.07      0.05      0.09      0.04  Continuousb                                   Straight      0.81  0.51, 1.27      0.87  0.57, 1.35      0.87  0.54, 1.39      0.82  0.42, 1.58   Soluble      0.90  0.64, 1.27      0.91  0.59, 1.39      0.84  0.51, 1.38      0.75  0.39, 1.44   Synthetic      1.06  0.95, 1.17      1.13  0.99, 1.30    1.15  1.00, 1.32      1.17  1.01, 1.35  Abbreviations: CI, confidence interval; HR, hazard ratio; MWF, metalworking fluid. a Each Cox regression model included cumulative exposure to the 3 fluid types (size <3.5 μm; 20-year lag), used age as time scale, and was adjusted for year of hire, calendar year, race, and manufacturing plant. Multiple age cutpoints were used to define premenopausal women. b HR per 1-interquartile-range increase among women exposed to MWF who were 55 years of age or younger; for straight, soluble, and synthetic fluids, these are 0.157, 0.474, and 0.092 mg/m3-years, respectively. Figure 2. View largeDownload slide Adjusted hazard ratios for premenopausal breast cancer incidence as a smoothed function of 20-year lagged cumulative exposure to synthetic metalworking fluid (MWF) as estimated in a Cox regression model using penalized splines (2 degrees of freedom) based on a cohort of female workers in the United Autoworker-General Motor incidence cohort who were alive in 1985 in Michigan, 1985–2013. Premenopausal breast cancer was defined as diagnosis by 4 select age cutpoints. The 95% confidence interval (dotted lines) is shown for the model using age 55 years as the cutpoint. Models used age as the time scale and were adjusted for cumulative exposure to straight and soluble MWF, year of hire, calendar year, race, and manufacturing plant. The graph is truncated at the 99th percentile of synthetic fluid exposure among women with breast cancer (0.60 mg/m3-years). The rug plot indicates exposure of the women with breast cancer who were 55 years of age or younger. Figure 2. View largeDownload slide Adjusted hazard ratios for premenopausal breast cancer incidence as a smoothed function of 20-year lagged cumulative exposure to synthetic metalworking fluid (MWF) as estimated in a Cox regression model using penalized splines (2 degrees of freedom) based on a cohort of female workers in the United Autoworker-General Motor incidence cohort who were alive in 1985 in Michigan, 1985–2013. Premenopausal breast cancer was defined as diagnosis by 4 select age cutpoints. The 95% confidence interval (dotted lines) is shown for the model using age 55 years as the cutpoint. Models used age as the time scale and were adjusted for cumulative exposure to straight and soluble MWF, year of hire, calendar year, race, and manufacturing plant. The graph is truncated at the 99th percentile of synthetic fluid exposure among women with breast cancer (0.60 mg/m3-years). The rug plot indicates exposure of the women with breast cancer who were 55 years of age or younger. DISCUSSION In a prospective cohort of 4,503 female autoworkers from the UAW-GM study, we examined MWF exposure and its association with incident breast cancer. Exposure to straight MWF, but neither soluble nor synthetic MWF, was positively associated with breast cancer. This was evident in the analyses using subcohorts restricted by year of hire. When restricted to women hired in 1959 or later, the hazard ratio for straight MWF exposure was elevated in the continuous and categorical analyses. Results became stronger when using the subcohort restricted to women hired in 1969 or later. This more restrictive subcohort was designed to reduce left truncation bias and may explain the stronger results observed in these analyses. Though power was adequate for the main analysis, when we restricted analysis to women presumed to be premenopausal, the number of women in each exposure category was sparse. Results, however, were modestly suggestive of an increased risk associated with higher levels of synthetic MWF exposure. The associations were more pronounced when younger age cutpoints defined the women with cancer. These definitions were more specific for premenopausal breast cancer, which is more important than sensitivity for reducing bias when the outcome is rare (33). Our interpretation of these results, however, was constrained by the small number of premenopausal women with breast cancer. Moreover, these results were apparently inconsistent with the inverse association with synthetic MWF observed in the main analysis. One possible explanation for this inconsistency is that synthetic MWF were introduced later. Exposed older women, therefore, had to have remained employed longer, introducing survivor bias among the older cases. Three previous breast cancer studies relied on data from the UAW-GM study (17, 22, 23). The first study combined incident cases and deaths, and imputed date of diagnosis, if it was missing (17). The modest risk increase observed for cumulative soluble MWF exposure was inconsistent with the literature on latency for breast cancer, suggesting higher risk associated with exposure in the decade preceding diagnosis. The present analysis was restricted to incident cases and we imposed a 20-year lag on cumulative exposures based on the latency period for breast cancer (34–36). The second study examined the incidence of several cancers and reported an increased and borderline significant hazard ratio associated with exposure to straight MWF among younger women in the cohort (23). In the third study, based on 43 deaths due to breast cancer, a hazard ratio of 1.4 (95% confidence interval: 0.7, 2.5) was reported associated with higher straight-MWF exposure (22). Death, however, is a poor surrogate for breast cancer incidence, given a 5-year survival rate of 88% (37). After accounting for left truncation, we observed a similar but marginally larger hazard ratio associated with straight-MWF exposure when using a continuous metric and a categorical metric. The additional 9 years of follow-up compared with the second study and the use of the more appropriate breast cancer outcome measure of incidence rather than death, compared with the third study, may have led to the slightly more positive results. The contrasting results in the pooled analysis versus the subanalysis restricted to premenopausal breast cancer suggest a possibly distinct biologic mechanism for this type of breast cancer. However, the distribution of cumulative straight-MWF exposure was markedly lower among women with premenopausal breast cancer, with a mean of 0.02 (standard deviation, 0.05) mg/m3-years based on an age cutpoint of 55 years compared with all cases (mean, 0.29 (standard deviation, 1.08) mg/m3-years), and may also have contributed to a lack of association for straight MWF in the premenopausal analyses. The PAH content of straight MWF may be the causal factor in the elevated risk. Air pollution exposure, as a proxy for PAH, has been associated with postmenopausal breast cancer. Total suspended particulate exposure was associated with postmenopausal, but not premenopausal, breast cancer in a population-based, case-control study in western New York (9). A later study in the same population modeled total PAH exposure and similarly found exposure at first birth associated with postmenopausal, but not premenopausal, breast cancer (10). The only study, to our knowledge, to examine occupational PAH exposure found an increased odds of premenopausal breast cancer among those exposed (35). Because benzene was a coexposure in the study, the authors additionally evaluated women exposed exclusively to PAH and found null results. Study results of the Long Island Breast Cancer Study Project indicated a positive association of PAH-DNA adducts, which are short-term PAH biomarkers of exposure, with incident breast cancer (38, 39). There is less evidence for synthetic fluids and premenopausal breast cancer risk. Nitrosamines are a class of potentially hazardous contaminants found in synthetic fluids (40). N-Nitrosodiethanolamine, a type of nitrosamine, has been found in synthetic fluids (41, 42) and in postshift urine samples of workers exposed to water-based MWF (43). It has been demonstrated to induce DNA damage in animal and human cells (44), and metal workers exposed to higher levels of N-nitrosodiethanolamine had more DNA single-strand breaks in blood cells (45). Although N-nitrosodiethanolamine is classified as possibly carcinogenic to humans and is an animal carcinogen (46), there is no specific research on this compound and breast cancer risk. Bias due to the healthy-worker survivor effect is a concern in occupational epidemiology (47–49). We recently assessed the presence of the healthy-worker survivor effect in the UAW-GM breast cancer incidence cohort based on 3 necessary underlying conditions: 1) leaving work predicts future exposure, 2) leaving work is associated with disease outcome, and 3) prior exposure increases probability of leaving work (50). The first condition is a given, because subjects who leave work are no longer exposed. We found prior soluble and synthetic, but not straight, MWF exposure associated with leaving work among female autoworkers. This supports the third condition for soluble and synthetic, but not straight, MWF. Although breast cancer was not examined in that assessment, we used the same statistical method and found that having left work was not associated with breast cancer incidence, indicating a lack of evidence for the second condition. Overall, these results imply that healthy-worker survivor bias did not influence our estimates of the association between MWF exposure and breast cancer risk, particularly for straight MWF. Our main limitation is unmeasured potential confounders. Although we controlled for age and race, we did not account for several other breast cancer risk factors, including socioeconomic status, family history of breast cancer, age at menarche or menopause, age at first full-term pregnancy, parity, breastfeeding history, use of oral contraceptives or postmenopausal hormones, and alcohol consumption. Because the women in our study cohort were all blue-collar workers employed at the same 3 plants, we did not expect large differences in socioeconomic status. Additionally, we did not expect family history of breast cancer, age at menarche or menopause, or use of oral contraceptives or postmenopausal hormones to be related to MWF exposure. Last, prior studies of this cohort found no association between MWF exposure and cirrhosis death, a proxy for alcohol consumption (51). Therefore, the potential confounders of most concern are those related to childbearing: age at first full-term pregnancy, parity, and breastfeeding history. Women who were younger at first full-term pregnancy, with higher parity, and who breastfed longer had reduced breast cancer risk and were likely to have lower MWF exposure due to taking maternity leave, entering the workforce later, and/or leaving the workforce earlier. This would produce positive confounding and bias results upward. In the present study, however, we saw no evidence of leaving work being associated with reduced breast cancer risk. Furthermore, we would expect such a bias to affect the results for all 3 fluid types, not just the straight MWF in the pooled analyses or the synthetic fluids in the premenopausal analyses. Although unmeasured confounding may account for some portion of the results observed, this study makes a contribution to the literature. This study adds to the limited literature regarding quantitative chemical exposures and breast cancer risk in humans. This study used a well-characterized occupational cohort of women working in automotive plants who had long cancer incidence follow-up and provided evidence of an increased risk of breast cancer associated with cumulative straight MWF exposure. ACKNOWLEDGMENTS Author affiliations: Environmental Health Sciences Division, School of Public Health, University of California at Berkeley, Berkeley, California (Erika Garcia, Ellen A. Eisen); and Epidemiology Division, School of Public Health, University of California at Berkeley, Berkeley, California (Patrick T. Bradshaw). This work was financially supported by the National Institute of Occupational Safety and Health at the Centers for Disease Control and Prevention (grants T42OH008429 and R01OH010028). Conflict of interest: none declared. Abbreviations MWF metalworking fluid PAH polycyclic aromatic hydrocarbons UAW-GM United Autoworker-General Motors REFERENCES 1 US Cancer Statistics Working Group. United States Cancer Statistics: 1999–2011 Incidence and Mortality Web-based Report. Atlanta, GA: Department of Health and Human Services, Center for Disease Control and Prevention, and National Cancer Institute; 2014. 2 SEER. Cancer Statistics Factsheets: Female Breast Cancer. Bethesda, MD: National Cancer Institute. http://seer.cancer.gov/statfacts/html/breast.html. Accessed November 7, 2016. 3 Madigan MP, Ziegler RG, Benichou J, et al.  . Proportion of breast cancer cases in the United States explained by well-established risk factors. J Natl Cancer Inst . 1995; 87( 22): 1681– 1685. Google Scholar CrossRef Search ADS PubMed  4 Institute of Medicine. Chapter 7: Recommendations for future research. Breast Cancer and the Environment: A Life Course Approach . Washington, DC: The National Academies Press; 2012: 325– 345. PubMed PubMed  5 Rudel RA, Attfield KR, Schifano JN, et al.  . Chemicals causing mammary gland tumors in animals signal new directions for epidemiology, chemicals testing, and risk assessment for breast cancer prevention. Cancer . 2007; 109( 12 suppl): 2635– 2666. Google Scholar CrossRef Search ADS PubMed  6 Neale M. Tribiology Handbook . New York, NY: John Wiley and Sons; 1973. 7 International Agency for Research on Cancer. Mineral Oils, Untreated or Mildly Treated. Lyon, France: International Agency for Research on Cancer; 2012. 8 Catchpole WM, Macmillan E, Powell H. Specifications for cutting oils with special reference to carcinogenicity. Ann Occup Hyg . 1971; 14( 2): 171– 179. Google Scholar PubMed  9 Bonner MR, Han D, Nie J, et al.  . Breast cancer risk and exposure in early life to polycyclic aromatic hydrocarbons using total suspended particulates as a proxy measure. Cancer Epidemiol Biomarkers Prev . 2005; 14( 1): 53– 60. Google Scholar PubMed  10 Nie J, Beyea J, Bonner MR, et al.  . Exposure to traffic emissions throughout life and risk of breast cancer: the Western New York Exposures and Breast Cancer (WEB) study. Cancer Causes Control . 2007; 18( 9): 947– 955. Google Scholar CrossRef Search ADS PubMed  11 Office of Chemical Safety and Pollution Prevention, Office of Water, and Office of Research and Development, US Environmental Protection Agency. Endocrine Disruptor Screening Program: Universe of Chemicals for Potential Endocrine Disruptor Screening and Testing. Washington, DC: US Environmental Protection Agency; 2012. 12 Steenland K, Burnett C, Lalich N, et al.  . Dying for work: the magnitude of US mortality from selected causes of death associated with occupation. Am J Ind Med . 2003; 43( 5): 461– 482. Google Scholar CrossRef Search ADS PubMed  13 International Labour Office. Global Employment Trends for Women: March 2009 . Geneva, Switzerland: International Labour Office; 2009. 14 Frost and Sullivan. Analysis of the Global Metalworking Fluids Market. Mountain View, CA; 2014. 15 Delzell E, Beall C, Macaluso M. Cancer mortality among women employed in motor vehicle manufacturing. J Occup Med . 1994; 36( 11): 1251– 1259. Google Scholar CrossRef Search ADS PubMed  16 Beall C, Delzell E, Macaluso M. Mortality patterns among women in the motor vehicle manufacturing industry. Am J Ind Med . 1995; 28( 3): 325– 337. Google Scholar CrossRef Search ADS PubMed  17 Thompson D, Kriebel D, Quinn MM, et al.  . Occupational exposure to metalworking fluids and risk of breast cancer among female autoworkers. Am J Ind Med . 2005; 47( 2): 153– 160. Google Scholar CrossRef Search ADS PubMed  18 Brophy JT, Keith MM, Gorey KM, et al.  . Occupation and breast cancer: a Canadian case-control study. Ann NY Acad Sci . 2006; 1076: 765– 777. Google Scholar CrossRef Search ADS PubMed  19 Hansen J. Breast cancer risk among relatively young women employed in solvent-using industries. Am J Ind Med . 1999; 36( 1): 43– 47. Google Scholar CrossRef Search ADS PubMed  20 Villeneuve S, Févotte J, Anger A, et al.  . Breast cancer risk by occupation and industry: analysis of the CECILE study, a population-based case-control study in France. Am J Ind Med . 2011; 54( 7): 499– 509. Google Scholar CrossRef Search ADS PubMed  21 Brophy JT, Keith MM, Watterson A, et al.  . Breast cancer risk in relation to occupations with exposure to carcinogens and endocrine disruptors: a Canadian case-control study. Environ Health . 2012; 11: 87. Google Scholar CrossRef Search ADS PubMed  22 Friesen MC, Betenia N, Costello S, et al.  . Metalworking fluid exposure and cancer risk in a retrospective cohort of female autoworkers. Cancer Causes Control . 2012; 23( 7): 1075– 1082. Google Scholar CrossRef Search ADS PubMed  23 Friesen MC, Costello S, Thurston SW, et al.  . Distinguishing the common components of oil- and water-based metalworking fluids for assessment of cancer incidence risk in autoworkers. Am J Ind Med . 2011; 54( 6): 450– 460. Google Scholar CrossRef Search ADS PubMed  24 Eisen EA, Tolbert PE, Monson RR, et al.  . Mortality studies of machining fluid exposure in the automobile industry I: a standardized mortality ratio analysis. Am J Ind Med . 1992; 22( 6): 809– 824. Google Scholar CrossRef Search ADS PubMed  25 Eisen EA, Bardin J, Gore R, et al.  . Exposure-response models based on extended follow-up of a cohort mortality study in the automobile industry. Scand J Work Environ Health . 2001; 27( 4): 240– 249. Google Scholar CrossRef Search ADS PubMed  26 Applebaum KM, Malloy EJ, Eisen EA. Left truncation, susceptibility, and bias in occupational cohort studies. Epidemiology . 2011; 22( 4): 599– 606. Google Scholar CrossRef Search ADS PubMed  27 Michigan Department of Health and Human Services. Michigan Cancer Surveillance Program. http://www.michigan.gov/mdch/0,4612,7-132-2945_5221-16586--,00.html. Accessed May 5, 2015. 28 Division of Cancer Prevention and Control, National Center for Chronic Disease Prevention and Health Promotion. National Program of Cancer Registries. Centers for Disease Control and Prevention; 2014. http://www.cdc.gov/cancer/npcr/. Accessed November 5, 2014. 29 Cowper DC, Kubal JD, Maynard C, et al.  . A primer and comparative review of major US mortality databases. Ann Epidemiol . 2002; 12( 7): 462– 468. Google Scholar CrossRef Search ADS PubMed  30 Hallock MF, Smith TJ, Woskie SR, et al.  . Estimation of historical exposures to machining fluids in the automotive industry. Am J Ind Med . 1994; 26( 5): 621– 634. Google Scholar CrossRef Search ADS PubMed  31 Woskie SR, Smith TJ, Hallock MF, et al.  . Size-selective pulmonary dose indices for metal-working fluid aerosols in machining and grinding operations in the automobile manufacturing industry. Am Ind Hyg Assoc J . 1994; 55( 1): 20– 29. Google Scholar CrossRef Search ADS PubMed  32 de Boor C. A Practical Guide to Splines . New York, NY: Springer-Verlag; 1978. Google Scholar CrossRef Search ADS   33 Rothman KJ, Greenland S, Lash TL. Modern Epidemiology . 3rd ed. Philadelphia, PA: Wolters Kluwer Health/Lippincott Williams & Wilkins; 2008. 34 Aschengrau A, Paulu C, Ozonoff D. Tetrachloroethylene-contaminated drinking water and the risk of breast cancer. Environ Health Perspect . 1998; 106( suppl 4): 947– 953. Google Scholar CrossRef Search ADS PubMed  35 Petralia SA, Vena JE, Freudenheim JL, et al.  . Risk of premenopausal breast cancer in association with occupational exposure to polycyclic aromatic hydrocarbons and benzene. Scand J Work Environ Health . 1999; 25( 3): 215– 221. Google Scholar CrossRef Search ADS PubMed  36 Kelsey JL, Berkowitz GS. Breast cancer epidemiology. Cancer Res . 1988; 48( 20): 5615– 5623. Google Scholar PubMed  37 American Cancer Society. Cancer Facts & Figures 2014. Atlanta, GA: American Cancer Society; 2014. 38 Gammon MD, Sagiv SK, Eng SM, et al.  . Polycyclic aromatic hydrocarbon-DNA adducts and breast cancer: a pooled analysis. Arch Environ Health . 2004; 59( 12): 640– 649. Google Scholar CrossRef Search ADS PubMed  39 Gammon MD, Santella RM, Neugut AI, et al.  . Environmental toxins and breast cancer on Long Island. I. Polycyclic aromatic hydrocarbon DNA adducts. Cancer Epidemiol Biomarkers Prev . 2002; 11( 8): 677– 685. Google Scholar PubMed  40 National Institute for Occupational Safety and Health, US Centers for Disease Control and Prevention. Criteria for a Recommended Standard: Occupational Exposure to Metalworking Fluids. Cincinnati, OH: US Department of Health and Human Services; 1998. 41 Fan TY, Morrison J, Rounbehler DP, et al.  . N-Nitrosodiethanolamine in synthetic cutting fluids: a part-per-hundred impurity. Science . 1977; 196( 4285): 70– 71. Google Scholar CrossRef Search ADS PubMed  42 Keefer LK, Goff U, Stevens J, et al.  . Persistence of N-nitrosodiethanolamine contamination in American metal-working lubricants. Food Chem Toxicol . 1990; 28( 7): 531– 534. Google Scholar CrossRef Search ADS PubMed  43 Ducos P, Gaudin R. N-Nitrosodiethanolamine urinary excretion in workers exposed to aqueous metalworking fluids. Int Arch Occup Environ Health . 2003; 76( 8): 591– 597. Google Scholar CrossRef Search ADS PubMed  44 Holzer J, Voss B, Karroum S, et al.  . A comparative study of chemically induced DNA damage in isolated nasal mucosa cells of humans and rats assessed by the alkaline comet assay. J Toxicol Environ Health A . 2008; 71( 13– 14): 936– 946. Google Scholar CrossRef Search ADS PubMed  45 Fuchs J, Burg J, Hengstler JG, et al.  . DNA damage in mononuclear blood cells of metal workers exposed to N-nitrosodiethanolamine in synthetic cutting fluids. Mutat Res . 1995; 342( 1– 2): 95– 102. Google Scholar CrossRef Search ADS PubMed  46 International Agency for Research on Cancer. Some Industrial Chemicals. Lyon, France: International Agency for Research on Cancer; 2000. 47 Arrighi HM, Hertz-Picciotto I. The evolving concept of the healthy worker survivor effect. Epidemiology . 1994; 5( 2): 189– 196. Google Scholar CrossRef Search ADS PubMed  48 Buckley JP, Keil AP, McGrath LJ, et al.  . Evolving methods for inference in the presence of healthy worker survivor bias. Epidemiology . 2015; 26( 2): 204– 212. Google Scholar CrossRef Search ADS PubMed  49 Picciotto S, Hertz-Picciotto I. Commentary: healthy worker survivor bias: a still-evolving concept. Epidemiology . 2015; 26( 2): 213– 215. Google Scholar CrossRef Search ADS PubMed  50 Garcia E, Picciotto S, Costello S, et al.  . Assessment of the healthy worker survivor effect in cancer studies of the United Autoworkers-General Motors cohort. Occup Environ Med . 2017; 74( 4): 294– 300. Google Scholar CrossRef Search ADS PubMed  51 Eisen EA, Tolbert PE, Hallock MF, et al.  . Mortality studies of machining fluid exposure in the automobile industry. III: A case-control study of larynx cancer. Am J Ind Med . 1994; 26( 2): 185– 202. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Journal

American Journal of EpidemiologyOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial