The survey instrument was derived from previously developed questionnaires [28-30] with special attention to reproductive developmental stages. The questionnaire captured reproductive risk factors such as: parity, duration of lactation, menstrual and menopausal history, use of hormone replacement therapy and oral contraceptives, and family history. Demographic and lifestyle risk factors included: income, education, physical activity, weight and body mass index (BMI), alcohol use, smoking history, and residential history. Up to 12 jobs were recorded for each participant, i.e. periods of employment. These included start and end dates for each job as well as free-text descriptions of job activities which were used to inform coding of occupation, industry and exposure. Work history was available in time units of one year.

Cases were recruited over a six year period from mid 2002 through mid 2008 with the following criteria: new diagnosis of histologically confirmed breast cancers (ICD-9 Code – 174) [31], excluding recurrences; current residence in Essex or Kent Counties; willingness and ability to participate in a one to two hour interview with adequate language facility. Upon receipt of informed consent, names, addresses and telephone numbers of cases were provided by WRCC staff. Information outlining the study was mailed to each of the referrals and follow-up telephone calls were made by research personnel to schedule interviews. To minimize selective recruitment bias, the information disclosed the goal of understanding the causes of cancer but did not identify a focus on occupation or environment. After informed consent was obtained, the patient’s date of diagnosis and tumor pathology regarding estrogen receptor (ER) or progesterone receptor (PR) status were accessed.

Community controls from the same geographic study area were recruited from 2003 through 2007. Randomly selected households were obtained through computer-generated telephone numbers and linkage to mailing addresses. The same study information which was sent to cases, which made no reference to occupational or environmental factors, was mailed to potential controls and followed-up with telephone calls. Eligibility requirements were the same as the cases, with the exception that only one person per household was allowed and could have no prior history of breast or ovarian cancer. Interviewers followed a scripted recruitment and interview plan. Cases and controls were compensated $20 for their time.

Each job was coded using the North American Industry Classification System (NAICS) [32] and National Occupational Classification (NOC) [33]. Within each job, multiple NAICS and NOC classifications were allowed. In order to characterize exposures in the subjects’ work activities, all unique NAICS and NOC combinations that occurred in the study’s collective work history were compiled and classified in one of 32 sectors, which we identified as “minor sectors” and in 8 sectors that we identified as “major” sectors of primary interest which were based on prior hypotheses and consideration of potential exposures to mammary carcinogens [14] or EDCs [34] (Table 1). Several minor sectors potentially of interest for breast cancer investigation, such as textiles, footwear, printing, ceramics, furniture, jewelry and electronics, were combined as light manufacturing due to small numbers of cases.

Each unique NAICS/NOC combination was further assigned an exposure classification code signifying the likely presence and intensity of carcinogen and/or EDC exposures in the manner of expert panel assignment [28,35]. Investigators, who were blind to case/control status, assigned exposure categories as “low, moderate or high” based on general process characteristics and prior professional knowledge of chemical hazards. For example, Table 2 displays the NAICS/NOC combinations in the Plastics major sector and the assigned exposure codes. This assignment was implemented by investigators with extensive experience 1) in exposure assessment within the occupational health clinic network associated with the Ontario Workers Compensation system and 2) in a wide range of industrial hygiene evaluations including automobile and parts manufacturing, health care, casinos, food production and agriculture. The assigned exposure strata were randomly checked by team members to ensure consistency and validity. NAICS/NOC codes also determined a social class variable (white collar, blue collar, unknown) based on NOC text.

In preliminary analyses minor sectors were examined as: a) categorical variables (minor sector of longest [lagged] duration, a mutually exclusive classification); and b) continuous variables (lagged durations of employment in all minor sectors) [36]. Next, cumulative exposure metrics were calculated as the sum over time of the assigned exposure levels from each NAICS/NOC activity using two weighting schemes. The first assigned to the categories “low, moderate and high” the weights 0, 1 and 2, respectively, and summed these over time; the second assigned the weights 0, 1 and 10. The two weightings permitted a choice to be made concerning the ratio of average exposure levels in the “moderate” vs. “high” categories, which would be expected to vary widely across processes, workplaces and sectors. When a job comprised multiple NAICS/NOC categories (because of mixed activities, holding more than one position at a time, or sequential employment within one year) equal weight was assigned to each element of the job in assessing exposure. Duration and cumulative exposure metrics were lagged 5 years, i.e., summed up until 5 years prior to a subject’s diagnosis (cases) or participation in study (controls), accounting for delay between primary carcinogenic events and clinical diagnosis. Cumulative exposures were calculated: a) generically combining all sectors, b) within the eight major sectors, and c) for some additional groupings of special interest, some of which were derived from preliminary observations such as in food manufacturing or automotive plastics.

Because food canning was a major activity in this region, related exposures were examined. Polymer lining of cans was approved in the 1960’s by the US Food and Drug Administration [37] and became widespread, internationally, in the 1970’s. To test for a role of chemicals in canning, we defined canning exposure as work in the canning industry NAICS codes after 1973 by which time epoxy coatings were being widely introduced [38]. Food and beverage can coatings have been found to contain bisphenol A (BPA), which is a recognized EDC [39].

Prior research suggested that early employment in agriculture may predispose individuals to higher risk from subsequent occupational exposures [27]. A biological interaction term for cumulative exposure and prior agricultural work was constructed for several of the major-sectors of concern (e.g., automotive plastics, canning) as a weighted sum over time of the sector exposures, where the weight was the then-current cumulative exposure in (prior) farm work. The exposure contribution from a given year was the exposure level rating of a job multiplied by the person’s cumulative, previous, agricultural exposure. Models were then fit with the usual cumulative exposure terms for major sectors together with these interaction terms.

Cumulative exposures for some analyses were partitioned into time (age) windows representing distinct stages of breast development that could affect risk [12,40]. Cumulative exposure accruing in each window was calculated, with windows defined on age as follows: a) before menarche, b) menarche to first full term pregnancy, c) first full term pregnancy to menopause, d) after menopause. In the absence of a first completed pregnancy or premenopausal status, subjects would have no observation time in windows c or d, respectively.

Results were based on frequency-matched case–control analyses using a loglinear specification in multiple conditional logistic regression [41]. Matching was achieved by stratifying the cases and controls in three-year age intervals such that, on average, ages of controls were within about 1.5 years of the cases. Due to sparse data, all subjects below age 30 were assigned the same matching stratum. Odds ratios (OR) from logistic regression models are interpreted as estimates of relative “risk” throughout this report. In addition to reproductive risk factors, demographic risk factors were included in all models, including: smoking (pack-years and pack-years squared) calculated up to the age of diagnosis/participation, education in three levels (less than high school, high school and some college, college degree), and family income (<$40,000, >$40,000 blue collar, >$40,000 white collar). Employment duration terms (linear and squared) were statistically significant and included in all matched analyses (except the initial descriptive analysis by minor sector of longest duration). For investigation of breast cancer restricted to specific receptor classifications, breast cancer cases of other types were excluded. Only three estrogen/progesterone receptor categories were examined due to small numbers of cases in other types: ER+/PR+; ER+/PR-; ER-. For examination of menopausal status, subjects were classified on whether age was greater than age at menopause when augmented with a five-year lag. Additive relative rate model specifications were also evaluated using conditional logistic regression [42], which permitted testing for effect measure modification, or interaction, in an additive model context.

The results display both p-values, showing the probabilities of chance associations, and confidence intervals, showing the range of true parameter values that would produce the observed estimates with probability > 2.5 percent (two-tailed).

Using the generic cumulative exposure metric (across all minor sectors) with the 0, 1, 2 exposure weighting scheme produced a statistically significant excess risk of breast cancer; 10 years in a high-exposed job had an associated 29% increase (OR = 1.29; 95% CI, 1.10-1.51) (Table 6, model 1). With the (0,1,10) weighting scheme, a stronger association resulted (OR = 1.42; 95% CI, 1.18-1.73), with a 42% increase in risk after 10 years in jobs assessed as likely high-exposure (model 2). Applying the 0, 1, 10 weighting scheme within major sectors identified excess breast cancer risk: in agriculture (OR = 1.34; 95% CI, 1.03-1.74; for 10 years in high-exposure jobs), plastics (OR = 2.43; 95% CI, 1.39-4.22), metal work (OR = 1.73; 95% CI, 1.02-2.92) and in bars-gambling work (OR = 2.20; 95% CI, 0.91-5.29) (model 3). There was no additional risk, beyond that found in farming in general, for specific farming activities involving corn cultivation since 1978 when atrazine use became common or greenhouse work. The excess in chemicals/petrochemicals was based on only 6 cases. Including additional terms for categories of special interest slightly strengthened the major category associations (Table 6, model 4).

The analysis revealed excess risk with work in high exposure food canning jobs (OR = 2.35; 95% CI, 1.00-5.53, for 10 years work) (Table 6, model 4). This metric was motivated by the endocrine disruptor hypothesis and by preliminary findings of an excess in those for whom food manufacturing was the sector of longest duration (Table 4). There was a possible excess in a group that includes toll booth operators (with potentially high vehicle emission exposures) (OR = 1.17; 95% CI, 0.44-3.14) but this group was limited by small numbers. The strongest association was with automotive plastics manufacturing (OR = 2.68; 95% CI, 1.47-4.88, p=0.0013). Within the auto industry in general, excess breast cancer appeared to be limited to small automotive parts suppliers, which would include some plastics operations (OR = 2.48; 95% CI, 1.00-6.10).

There was no evidence of risk modification related to prior work in agriculture for subsequent work in metals or canning (Table 6, model 5). For bars-gambling work the estimate for the interaction term was stronger (OR=2.38, 95%CI=0.58-9.79; for 1 year of farm work prior to 10 years of bars-gambling exposure) than for the main effect, although both were not statistically significant (Table 6, model 5). For automotive plastics the estimate of a doubling of risk for one year of prior farm work was not statistically significant (OR=2.31, 95%CI=0.53-9.98).

Partitioning the generic Cumulative Exposure Metrics I and II into time-windows suggests that the most important exposures affecting breast cancer risk occur in the third time window – from first full term pregnancy to menopause; the elevation was smaller for the first, second and fourth time-windows although there was limited power to distinguish them (Table 7). For Metric II, the point estimates for the second and third windows were close. Exposures in farming and bars-gambling work exhibited the same pattern whereas for the metal-related, plastics, and canning metrics the most important period appeared to be the second time-window – menarche to first full term pregnancy – before breast tissue is fully differentiated.

Examination of specific estrogen receptor (ER) or progesterone receptor (PR) types in the major sectors showing excess breast cancer produced distinct associations across receptor types (Table 8). The farming, metals, bars-gambling and particularly automotive plastics (OR = 3.63; 95% CI, 1.90-6.94, p=10^-4) sectors all exhibited excesses for the ER+/PR+ receptor type, but farming had a stronger excess in the ER- category (OR = 1.71; 95% CI, 1.12-2.62, p=0.014). The canning excess appeared to be entirely in the ER+ /PR- and ER- groups. Including the interaction terms for prior farm work identified possible effect modification for metals (ER+/PR-), bars-gambling (ER+ /PR+), and plastics (ER-), and a statistically significant interaction for prior farming and canning for the ER+ /PR- receptor status (OR = 1.81; 95% CI, 1.08-3.04, p=0.025) but not for ER+/PR+ or ER- receptor status.

Models fit with an additive relative rate specification generally fit less well than with the loglinear form. For example, the automotive plastics estimate with the loglinear model was OR=2.68 (1.47-4.88), p=0.0013 whereas the linear relative rate model produced OR=4.03 (1.43-6.64), p=0.023. With the interaction terms, the same pattern was observed as with the loglinear form, but confidence intervals were wider.

Restricting the analysis to premenopausal women resulted in many fewer cases (373 out of 1006) and considerably higher estimates of relative risk (Table 9) as in high exposed jobs in automotive plastics (OR=5.10, 95% CI=1.68-15.5) or canning (OR=5.20, 95% CI=0.95-28.4). Thus 10 yrs in that work was associated with a five-fold excess in breast cancer incidence. Adding a term for body mass index (BMI, centered at 25) produced a reduced odds of breast cancer with BMI (for 10 unit increase, OR = 0.78; 95% CI, 0.61-0.99), a slightly weaker association for automotive plastics, and a stronger association for canning (OR = 5.70; 95% CI, 1.03-31.5). In the analysis of postmenopausal breast cancer (633 cases), estimated risks associated with specific sectors were lower, particularly for automotive plastics and canning sectors. Terms for total employment duration, which were not statistically significant for premenopausal breast cancer, were statistically significant for postmenopausal cancer, with an estimated 6% decline in risk for each additional year of employment. BMI was a strong risk factor for postmenopausal breast cancer (for 10 unit increase in BMI, OR = 1.37; 95% CI, 1.12-1.68) but with small changes in major-sector risk estimates on addition of the BMI term.

Many of the women in this study had a background in farming or in the automotive plastics sector and this provided sufficient statistical power to show consistency with our prior studies [26,27]. Similarly, statistical power was sufficient to reliably identify elevated risks associated with food canning, bars-gambling and metalwork. In other sectors, such as construction, petrochemical, printing, and textile manufacturing sectors, there was a lack of statistical power.

This study found elevated breast cancer risk among women who had farmed. Agriculture in southwestern Ontario is diverse with tomato, corn and peach production being major activities. No additional risk, beyond what was found for farming in general, was observed for corn cultivation when atrazine was used but the labor-intensive activities there (detasseling) may have had low exposures. Several pesticides act as mammary carcinogens in animal bioassays [14]; many are EDCs [40]. In several cohort studies no elevated risk was observed among farming women [4] but some of these studies did not examine specific exposures or their timing. The Agricultural Health Study [48], while inconclusive, found risk was elevated among postmenopausal women whose husbands used specific pesticides [49]. A recent study found that young women exposed to DDT before the age of fourteen had an excess breast cancer risk before age fifty [50]. Band et al. [21] found in pre- and postmenopausal cases (combined) elevated breast cancer risk in fruit and other vegetable farming (OR = 3.11, 90% CI 1.24-7.81). There was an even greater breast cancer risk among women ever employed in other vegetable farming (OR = 7.33, 90% CI 1.16-46.2). One important aspect of farming in terms of endocrine disruption is that employment tends to begin earlier than other occupations. This may impart particular risks for those in pre-pubescent or pre-parity windows of vulnerability [51].

The plastics manufacturing jobs held by the women who participated in this study involved primarily injection molding. Injection molding and related processes take molten mixtures of resins, monomer, multiple additives, and sometimes lamination films, and form them into plastic pieces of defined dimensions and configuration. Emissions of vapors or mists from these hot processes can include plasticizers, ultraviolet-protectors, pigments, dyes, flame-retardants, un-reacted resin components and decomposition products. Further exposure comes from skin contact in handling and performing finishing tasks [52].

Many plastics have been found to release estrogenic chemicals [53]. Furthermore such additives as phthalates, and polybrominated diphenyl ethers (PBDE) have been identified as EDCs [40]. Cumulative exposure to mixtures of various estrogenic chemicals may compound the effect [54]. Some of the monomers present in the manufacturing of polymers (such as BPA, butadiene, and vinyl chloride) have been identified as mutagenic and/or carcinogenic [55]. Several monomers, additives, and related solvents, such as vinyl chloride, styrene, and acrylonitrile have been identified as mammary carcinogens in animal studies [14].

A near doubling of the risk for female breast cancer was found among plastics and rubber industry workers (SIR = 1.8; 95% CI, 1.4-2.3) [19]. Two other studies report increased breast cancer risk among rubber and plastics workers. Gardner et al. observed an OR of 1.4; 95% CI, 0.69-2.84, p=.26 after 10 years employment [56]. Ji et al. observed an OR of 2.0; 95% CI, 0.9-4.3 for those who were ever employed as plastics process machine operators [57]. Adding weight to this is a more than quadrupling of breast cancer risk found among male workers in the rubber and plastics industries (OR = 4.5; 95% CI, 0.7-28) [58]. Villeneuve et al. [22] found an increased breast cancer risk among French plastics and rubber product makers (OR = 1.8; 95% CI, 0.9-3.5). Labrèche et al. recently found an excess risk of breast cancer for occupational exposure to acrylic fibers (OR = 7.69; 95% CI, 1.5-40) and for nylon fibers (OR =1.99; 95% CI, 1.0-3.9) when exposures occurred before age thirty-six [23]. It was also reported that exposure to acrylic and rayon fibers and monoaromatic hydrocarbons doubled the risk of estrogen/progesterone positive tumours. The observation in this study of a robust association with automotive plastics manufacturing suggests that the risk factors are widespread and common in this sector. In the geographical study area, plastics production was primarily automotive. As a result, the non-automotive group was much smaller with less statistical power to detect associations. The absence of excess breast cancer incidence among non-automotive plastics workers could also be related to the types of polymers, additives and processes used in the manufacturing of automotive versus non-automotive products.

Food canning industry exposures could include pesticide residues and exposures specific to canning processes involving lead (historically) and coating emissions. Canning processing has been found to significantly reduce levels of residual pesticide [59] through washing, boiling, and peeling, which conceivably expose food processing workers. Some operations in this industry produce epoxy-coated cans at the food processing facility. In others, coated cans come from a supplier and are then hot washed. In either case, it is plausible that coating constituents are released into the plant atmosphere. Unlike typical consumer exposures that occur through ingestion of food packaged in epoxy-lined cans, the exposures to BPA from heated can liners experienced by canning workers occurs primarily through inhalation. The bioavailabilty of BPA that has been inhaled or absorbed dermally has been found to be eliminated at a slower rate than BPA ingested through food or drink [60]. The associations observed for canning show higher specificity with respect to receptor type and menopausal status than was observed for automotive plastics. For etiologic effects, this would be expected because the relevant exposure in canning, if polymer-related, would be more homogeneous than that across diverse plastics manufacturing activities. There is little epidemiological research on this subject. However, Ji et al. found elevated breast cancer risk for work in food canning (OR = 3.5; 95% CI, 1.2-10.1) after adjusting for reproductive history and SES [57]. Band et al. found elevated premenopausal breast cancer risk for work in food and beverage processing (OR = 3.45; 90% CI, 1.22-9.78) [21].

There were also important findings for those employed in bars or such gambling establishments as casinos and racetracks. The strong (OR = 2.28 after 10 years) (Table 6, model 4) but statistically significant (p=0.04 with one-tailed test) excess in breast cancer among bars-gambling workers implicates second-hand smoking [6]. Environmental tobacco smoke has been identified as a major occupational health concern in the bars-gambling industries [61,62]. The time period of lagged exposures studied here largely occurred prior to restrictive smoking regulations. Increase may also be related to disruption in circadian rhythms and decreased melatonin production resulting from work at night [63].

The findings for metal work, which includes foundries, metal stamping, fabrication and metalworking, have important implications for a broad range of blue collar industrial operations. Although these industries expose workers to metallic fume, metalworking fluids, PAHs, solvents, and other hazards [64,65], there has been little epidemiological research on associated breast cancer risk. A weak association was found for breast cancer risk and soluble metalworking fluids [18]. Several studies have found associations between PAH exposure and breast cancer risk [4,66]. Petralia et al. [66] found elevated breast cancer risk among premenopausal women exposed to PAHs and benzene. Risk was found to be increased among young women exposed to solvents in a variety of industrial settings [67].

Despite the likely presence of diverse carcinogen or EDC exposures within industrial sectors, some distinct specificity of receptor-type associations was observed. Of sectors showing elevated breast cancer, automotive plastics and the metals-related sectors would be expected to have the most diverse mix of carcinogen and EDC exposures; the canning and bars-gambling sectors would be expected to have the least diverse. Our study found statistically significant associations with canning in two receptor types (ER+/PR- and ER-), one exhibiting a statistically significant interaction with prior farm work. If the association is etiologic, this suggests that more than one mechanism may be involved. There was also a statistically significant increase in ER- tumor status among women employed in farming. Although there has been little research in this area, Danish researchers noted an association between ER- tumor status and exposure to dieldrin [68]. Whether or not the observed differences in hormone receptor status found in this study can be explained by the current level of understanding of the impact of EDC exposures on receptor status, they are indicative of the benefit of occupational investigations with more rigorous retrospective exposure assessments for investigating endocrine and other mechanistic aspects of breast carcinogenesis [69].

Finding distinct patterns for pre- versus postmenopausal breast cancer, such as increased premenopausal breast cancer among women employed in automotive plastics, adds confidence that exposure associations may be etiologic even though the exposure specificity is limited. Some of the events leading to a cancer diagnosis could occur throughout a woman’s life, i.e., both pre- and postmenopausal exposures could be involved in postmenopausal cases. Nevertheless, observing differences on menopausal status allows examination of some mechanistic hypotheses. Other studies of premenopausal breast cancer identified smoking [46] and such EDCs as benzene and PAHs [66] as risk factors.

Selection bias arising from participation that jointly depends on exposure history and case-status is unlikely to have played a significant role because study candidates did not know the intent of the study, and the participation rate among cases was relatively high. Among controls there would be even less likelihood of an exposure-driven decision. Uncontrolled confounding was undoubtedly present which is one reason why income terms were included in the model. Many lifestyle and health-related risk factors are associated with income. It would be quite unlikely for minor uncontrolled confounding to produce the strong specific associations observed.

Shift work was examined but did not produce statistically significant findings. In a study aggregating diverse workplaces, it is likely that exposures themselves depend on shift, making any shift-effect difficult to interpret. For example, not all employers operate a midnight shift and that shift can have maintenance, support and custodial activities that are absent in day-shift work and that could influence types or levels of exposures.

Exposure assessment based on survey instrument-derived work histories coded in NAICS and NOC categories has inevitable misclassification that dilutes or occludes observable associations [70]. Changing trends in technology and manufacturing are a further source of misclassification, possibly playing a role in some sectors like food manufacturing and dry cleaning. In the case of the food sector, focus on the specific subsector for canning produced a stronger association.

Models using the additive relative rate specification fit less well than with the loglinear choice, which assumes an exponential rather than linear dependence on the exposure metric. This suggests that the weighting scheme assuming a 10-fold higher exposure with “high” vs. “moderate” may have understated this ratio.

There was an under-representation of potentially highly exposed migrant farm and greenhouse laborers because they were not treated at the regional cancer center. This exclusion may have underestimated the risk estimate. Furthermore the role of carcinogens and EDCs that are ubiquitous in society will be underestimated in this study against the inflated background.

While this study was unable to identify exposure to specific chemicals, associations were observed between breast cancer carcinogens and EDCs. EDCs and biological windows of vulnerability are not currently considered in the establishing of occupational exposure limits (OELs). These findings, along with mounting evidence from other recent studies of harm from low level EDC exposure, point to the need to re-evaluate OELs and their relevancy in regulatory protection.