Increased standardized incidence ratio of breast cancer in female electronics workers

The discovery of uncontrolled hazardous waste sites has emerged, in many countries, as an issue of major environmental and public health concern [1]. In 1994, a hazardous waste site, polluted by the dumping of solvents from a former factory, was discovered in Taoyuan, Taiwan; the factory was built in 1970, and was in operation for the manufacturing of electronic appliances up until 1992. This discovery subsequently emerged as a serious case of soil and groundwater contamination accompanied by suspected occupational cancer.

The factory had been inspected eight times by the Taiwanese government's inspection agency, with multiple violations of the regulations having been recorded. There were, for example, six separate citations amongst these violations for inadequate wind speed and/or capture velocity of the factory's ventilation devices [2]. The presence of tetrachloroethylene (PCE), 1,1-dichloroethylene, 1,1-dichloroethane, methylene chloride, trichloroethane (TCH), 1,1,2-trichloroethane, 1,2-dichloroethane and cis-1,2-dichloroethylene was subsequently detected in the soil and groundwater [3, 4]. From a review of the literature at the International Agency for Research on Cancer (IARC), the US National Toxicology Program and the Integrated Risk Information System, we find an association has been reported between 1,1-dichloroethylene, 1,1-dichloroethane, methylene chloride, 1,2-dichloroethane and carbon tetrachloride, and the incidence of mammary gland adenocarcinomas in animals [5].

A prior study of risk assessment provided evidence on exposure to chlorinated alkanes and alkenes (CA) in community residents, accompanied by the increased risk of liver cancer [6, 7], whilst a proportionate cancer morbidity ratio (PCMR) study found a significant increase in breast cancer (1.4 times) amongst workers in this company, as compared to workers in the weaving industry [8]. However, a subsequent standardized mortality ratio (SMR) study found no significant increase in cancer between 1990 and 1997 [9]. A further retrospective morbidity study, based on the same cohort, found a significant increase in the crude standardized incidence ratio for female breast cancer (SIR: 1.2, 95% CI: 1.0–1.4) [10]; however, the authors could find no significant dose-response related to duration of employment, and therefore make no conclusive inferences.

Nonetheless, a formal hypothesis linking breast cancer to solvents has been proposed by Canadian researchers, who found the persistent presence of many kinds of solvents in the breast milk of women which could potentially lead to an increase in the opportunity for initiating, or promoting cancer [11]. The period of adolescence, during which breast tissue undergoes substantial cellular proliferation, may diminish the time available for DNA repair [12]; thus, exposure to potential carcinogenic factors prior to such cellular differentiation may potentially increase the risk of cancer in later life [13]. Since an increase in mammary cancer has been found in female mice after chronic exposure to a mixture of CA [14], this retrospective cohort study was carried out based upon standardized incidence ratios (SIRs), in order to test the hypothesis of increased breast cancer amongst these female workers, with a 23-year follow-up period.

A retrospective design was adopted for this study, incorporating secondary vital statistics data and employment records from the labor insurance database on a former electronics factory in Taiwan. The factory, which was located at latitude 24°N and longitude 121°E, covered an area of 36,000 km². At the end of 2001, the total population in Taiwan was 22,092,387, comprising of 11,312,728 males and 10,779,659 females. The original cohort comprised of 64,000 female employees who had worked at the factory between 1973 and 1992. Three other workers with cancer were also excluded, because their diagnoses were established prior to the time of their first employment at the factory.

Following the further exclusion of 15 workers, each of whom had less than one full day of employment, we were left with a total of 63,982 workers for subsequent analysis. We obtained vital statistics data, reported by geographical location in Taiwan, from the Ministry of the Interior (MoI), within which both local and national vital statistics are centralized, along with census information. The Taiwan household registration program is designed to collate and supply demographic information, with every birth and death being ascertained by a formal certificate written and attested by a physician.

The coding of the sites of cancer was based upon the International Classification of Diseases for Oncology issued by the Department of Health, and the diagnoses of cancer for this cohort were determined by linking employee identification numbers for the entire cohort, from 1 January 1979 to 31 December 2001, with the data obtained from the Taiwan National Cancer Registry. The cancer registry is a population-based registry containing information on newly-diagnosed cancer patients in all hospitals in Taiwan with 50 beds or more. The registry is financially supported by a government agency, with registration fees being paid to the reporting hospitals on the basis of each case number reported.

The raw employment data on this cohort, provided by the Institute of Occupational Safety and Health and covered the period from 21 May 1973 to 31 December 1997, were retrieved from the database of the Bureau of Labor Insurance (BLI) of Taiwan. The overall duration of employment in this study was therefore accumulated from the beginning of the coverage of labor insurance to the date of the termination of employment.

Based upon knowledge amassed from our prior qualitative studies, for those subjects with incomplete BLI records, we made the following assumptions for the estimation of duration of employment. Where the date of first employment (and labor insurance) was missing, this was assumed to be the earliest age, i.e., 14 years; where the date of termination of employment was missing, this was assumed to be the end of factory operations in 1992. The Labor Insurance Act, promulgated by the government in 1958, subsequently came into effect in 1960. Thereafter, all Taiwanese citizens between the ages of 15 and 60 years were required to join the employment insurance program, as insured workers, either through their employer or through the establishment to which they belonged.

The latent periods of cancer in different organ systems are generally quite different. Although assuming minimal latency based upon a literature review may inadvertently increase the counting, within the denominators, of 'person-years at risk', we still made such choices in order to obtain conservative estimates. The latent periods of different cancers were: five years for cancer of thyroid [15], leukemia [16], 15 years for cancer of breast [17, 18] and cervix uteri [19], and 10 years for the others [20].

Since workers in the early-1970s were exposed to high levels of solvents within a very short period of time, we stratified the durations of employment into less than, and more than, one year. 'Person-years at risk' for each individual worker were accrued after the assumed latent periods. The National Cancer Registry of Taiwan came into operation on 1 January 1979; thus, the collation of numbers of new cancer cases was based upon the earliest date of cancer occurrence, death, or the date when the person was last known to be alive and free of cancer on 31 December 2001. The number of cases was then stratified by age, calendar year, duration of employment, and the time of first employment.

Due to the lack of detailed information on exposure for each worker, we could only explore the dose-response relationship by stratifying the duration of employment into one month, and one, five, 10, 15 and 20 years. In order to account for the ambiguous nature of the latent periods, we conservatively apply the time since first employment as the time lag, thereby assuming that all workers became exposed once they were employed at the factory [21]. The short durations of exposure were decided by the extensive personal interviews from our qualitative study in 2003, which found short-term high exposure in female workers during the early-1970s [22].

No data on solvent exposure had been kept by the factory, and although we attempted to produce a reconstruction of such exposure, our dataset was too limited and crude to permit any possible linkage to individual workers. Following a serious outbreak of chemical hepatitis in another electronics factory in 1972 [23], the government issued strict regulations on the use of organic solvents within factories; these regulations were promulgated on 20 June 1974, followed by other amendments prior to 1990 (on 16 February 1976 and 21 November 1978) [24]. Under the current law in Taiwan, the use of CA is now strictly regulated in all workplaces.

The regulations in Taiwan cover trichloromethane (TCA), 1,1,2,2,-tetrachloroethane, tetra chloromethane, 1,2,-dichloroethylene, 1,2,-dichloroethane, carbon disulfide, TCE, PCE, methylene chloride, 1,1,1-trichloroethane and their mixtures. However, the factory inspection records on this company indicated that there were multiple violations of standards for ventilation devices, particularly in the early years [2]. We stratified the cohort into four groups to represent the cancer risks during different periods, according to the three promulgation and amendment dates noted above.

Although we found an increased risk of breast cancer amongst young female workers first employed prior to 1974 (as shown in Table 3), it does not necessarily follow that such an association is causal, since our data did not allow for the control of potential confounders, including exposure to ionizing radiation [26–28], smoking [29], alcohol consumption [30, 31], menopause after the age of 55 [32], low parity [33–38], late first pregnancy [39–41], and a positive family history [42–44]. Thus, careful scrutiny of the findings should be undertaken prior to drawing any conclusions on the association.

Following promulgation, on 16 April 1974, Article 12 of the Labor Safety and Health Act in Taiwan, required every worker to undergo a pre-employment physical examination, which included a chest roentgenography. Female workers first employed in Taiwan prior to 1975 were therefore less likely to be exposed to ionizing radiation than those whose employment began later; thus, it seemed unlikely that the hypothesis of increased ionizing radiation was a potential confounder within this association. Similarly, according to the regular survey data undertaken by the Tobacco and Wine Monopoly Bureau, the smoking and alcohol consumption rates of adult women were relatively low in the early-1970s; there has, however, been an increasing trend over the last three decades for alcohol but not for smoking, as illustrated in Figure 1[45, 46]. Therefore, it also seems unlikely that either smoking or alcohol consumption were potential confounders in this study.

Of the 286 cases of breast cancer in females, one was below 30 age group, 101 were in the 30–39 age group, 144 were in the 40–49 age group and the remaining 40 cases were in the above 50 age group. Thus, menopause after the age of 50 cannot explain the increased risk. Moreover, since the age of bearing a first child has tended to shift over the past three decades, from 20–24 years to 25–29 years, or even into the early-thirties (Figure 1), the likelihood of a late first pregnancy for women first employed prior to 1974 was also lower [47]. It therefore appears that family history was the only factor we were unable to rule out.

Historically, a higher socioeconomic status (SES) has been reported to be associated with increased breast cancer verified by a number of epidemiological studies, but few previous studies could adjust for all relevant potential confounding factors. Braaten et al. suggested that this positive gradient in risk for breast cancer can be fully explained by established breast cancer risk factors [48]. Although there was no information on level of education from the BLI database of original cohort, we still retrieved relevant educational data of 47,348 workers (74%) by linking to the Taiwan Birth Registration Database that began in 1978 [49]. The proportions of senior high and senior vocational schools for female workers were significantly higher than general female adults, as summarized in Figure 2. Based on the employees' earnings survey of Directorate General of Budget, Accounting and Statistics, the income of female workers in electronics manufacturing were above the average in the early 1970s [50], which was raised from 124 to 486 US dollars during the period between 1978 and 1988. As this factory is invested and owned by multinational enterprise, the wages were even higher than average electronics [22]. On the contrary, female workers in this factory were in a higher socioeconomic class, which might be explanatory to the decreased morbidity due to thyroid cancer, as was also found by Levi et al [51]. And there might still be possibility of residual confounding by SES for increased risk of breast cancer.

Since this study used the same dataset of cancer registry to link both the exposed cohort and the comparison cohort, and since it ensured that all cancer cases were ascertained, there was very good comparability in the collection of information; thus, any selection bias was minimal, or indeed negligible. Moreover, although we do not have the actual exposure data, random misclassification generally leads toward null [52]. The assumptions of the first and last dates of employment for missing values would also generally lead to an underestimation of SIR.

Throughout its period of operation, a wide range of solvents were used within this factory [53, 54], including isopropyl alcohol (IPA), acetone, methyl ethyl ketone (MEK), cleaner, cobehn spray (trichloromethane, TCA) and 1,1,1-trichloroethane, and trichloroethylene (TCE). Furthermore, TCE was once used extensively for degreasing and cleaning [22, 53], with our prior qualitative study indicating that many female workers had used the abovementioned CA on a daily basis prior to 1974, for cleaning dirt off clothes, wiping job utensils and tabletops, and even for washing their hands [22].

An extensive review suggests an association of TCE exposure with kidney cancer, liver cancer, and non-Hodgkin's lymphoma as well as for cervical cancer, Hodgkin's disease, and multiple myeloma [55]. All of above cancers were analyzed in our study but no increase of SIR has been found. The analysis is also limited by the lack of detailed exposure information.

After stratification by the year of first employment, the SIR for breast cancer increased from 1.09 to 1.38 (Table 3); if allowing 10 years of employment, this increased further to 1.62 (95% CI: 1.02–2.42). Since the cancer registry database was only available for the period up to 2001, any workers whose first employment was after 1986 were not eligible for inclusion within the cohort if a latent period of 15 years was taken into consideration. Thus, we tentatively conclude that an association does exist for the period prior to 1974, when CA was in use, and that there is a need for continued follow-up to further examine this association.

Due to the lack of detailed information on the early-1970s, we can make only limited inferences with regard to overall exposure. Based upon our in-depth review of formal inspection records [2], and the monthly magazines published by the company [53, 54], we find that the only organic solvent under strict regulation which no longer appeared after 1974, was TCE (Table 5). We therefore suspect that TCE, and/or its mixtures, may be the most likely agent responsible for our findings. However, our results should be interpreted with caution since it is clear that co-exposure to other solvents did occur amongst our subjects.

In summary, there was a significantly increased risk of breast cancer amongst female workers within this factory prior to the introduction of strict regulations on the use of CA in all workplaces. Such risks were more prominent amongst female workers entering the workforce prior to 1974. However, our inference is limited by the lack of individual information on both the intensity of occupational exposure and family history.