Perchloroethylene-contaminated drinking water and the risk of breast cancer: additional results from Cape Cod, Massachusetts, USA. (Research).
Perchloroethylene (PCE; also called tetrachloroethylene) is a chlorinated solvent commonly used in industrial operations such as dry cleaning, textile processing, and metal degreasing [Agency for Toxic and Substances and Disease Registry (ATSDR) 1995]. Because PCE is mainly used in small, geographically scattered, and poorly controlled workplace settings, it has become a common contaminant of drinking water and Superfund sites.
Although industrial disposal is the typical source of drinking water contamination, PCE leached into the drinking water supplies of Cape Cod, Massachusetts, from an inner vinyl liner that was present in certain asbestos cement pipes (Larsen et al. 1983). The vinyl liner was introduced in the late 1960s in response to taste and odor complaints. A slurry of a vinyl plastic and PCE was used to coat the inside of the pipes just before shipping. The manufacturers assumed that the volatile PCE would disappear during the drying process; however, substantial quantities remained and slowly leached into the drinking water supplies.
More than a decade lapsed before the Massachusetts Department of Environmental Protection (DEP) learned that the vinyl liner was a source of PCE contamination. A survey in 1979 revealed that 660 miles of the vinyl-lined asbestos cement (VL/AC) pipe had been installed in Massachusetts, predominantly in the Cape Cod region (Massachusetts Department of Environmental Quality Engineering 1982). Typical levels in affected towns ranged from 1,600 to 7,750 [micro]g/L in low-flow locations, and from 1.5 to 80 [micro]g/L in medium- and high-flow locations (Demond 1982). DEP began a regular program of flushing and bleeding to correct the problem in 1980.
In response to public concern about elevated cancer rates and environmental pollution in the Cape Cod area, we conducted a population-based case-control study to evaluate the relationship between nine types of cancer and air and water pollution, including PCE-contaminated drinking water (Aschengrau et al. 1998; Paulu et al. 1999). Although our study suggested that women with high relative delivered doses (RDDs) of PCE-contaminated drinking water have an increased risk of breast cancer, firm conclusions were limited by the small proportion of exposed subjects, particularly when long latent periods were considered. We undertook the present study with a larger number of more recently diagnosed cases in order to evaluate further the hypothesis that PCE exposure increases the risk of breast cancer. The biologic rationale for our study stems from a hypothesis recently described by Labreche and Goldberg (1997) that organic solvents such as PCE may act either directly as genotoxic agents or indirectly through their metabolites to increase the risk of breast cancer.
Materials and Methods
Selection and enrollment of study population.
The case group was composed of women diagnosed with breast cancer from 1987 through 1993 who were permanent residents of eight Cape Cod towns (Barnstable, Bourne, Brewster, Chatham, Falmouth, Mashpee, Provincetown, and Sandwich) and whose diagnosis was reported to the Massachusetts Cancer Registry. Comparison of the Massachusetts Cancer Registry with other state cancer registries indicates nearly complete ascertainment for breast cancer (Massachusetts Department of Public Health 1995). For the vast majority of cases (94.6%) this was the first occurrence of breast cancer, but for a small percentage (5.4%) this was the second occurrence of a primary tumor.
The control group was composed of similarly aged women who were also permanent residents of the eight Cape Cod towns during 1987 through 1993. Because most cases were elderly and many were deceased at the start of data collection, three sources were used to identify controls in an efficient manner: a) living controls [less than or equal to] 64 years of age were selected by random-digit dialing; b) living controls [greater than or equal to] 65 years of age were randomly selected from a roster of Medicare beneficiaries provided by the Health Care Financing Administration (HCFA); and c) deceased controls were randomly selected from a roster of deceased residents of the eight Cape Cod towns provided by the Massachusetts Bureau of Health Statistics, Research, and Evaluation. The number of controls selected from each source was weighted to reflect the age and vital status distribution among cases.
Random-digit dialing identified a random sample of female telephone subscribers [less than or equal to] 64 years of age who lived in the study towns during 1987-1993. According to the 1990 Census, 97% of housing units in Massachusetts have telephone service (Bureau of the Census 1990). A total of 3,402 residences were identified using this method (Table 1). Approximately 68% did not have any residents who met the eligibility criteria, and another 9% would not respond to the eligibility-determining questions. About 15% never answered the telephone after many calls. Ultimately, 248 households were found with an eligible respondent.
Living controls [greater than or equal to] 65 years of age were identified using a roster of female Medicare beneficiaries provided by HCFA. Hatten (1980) has estimated that Medicare beneficiaries constitute 95% of the U.S. population [greater than or equal to] 65 years of age. HCFA controls (n = 468) were randomly selected from residents of the eight Cape Cod towns using a sampling scheme that frequency matched on age.
Controls who died from January 1987 through December 1996 were randomly selected from deaths occurring to female residents of the study towns. Individuals with any cause of death were eligible. Deceased controls (n = 234) were selected using a sampling scheme that was frequency matched on age and year of death.
All subject selection and enrollment procedures were approved by the Institutional Review Boards of Boston University Medical Center, the Massachusetts Cancer Registry, and the HCFA.
Up-to-date addresses and telephone numbers of subjects and, if needed, next-of-kin were determined using cancer registry and HCFA records, physicians and tumor registrars, birth and death records, voter registration lists, driver's license records, telephone books, directory assistance, Internet resources, and obituary notices in local newspapers. Following Massachusetts Cancer Registry guidelines, permission was obtained from physicians of living cancer cases before conducting the interview.
After obtaining verbal consent, trained interviewers conducted structured interviews to obtain information on demographic characteristics, confounding variables including family history of breast cancer, occupational exposure to PCE, and a 40-year residential history. Information was also obtained on tap and bottled water consumption, bathing habits, drinking water source, and use of drinking-water treatment devices.
Index years were randomly assigned to the controls according to the distribution of diagnosis years among the cases. Subjects who reported that they moved to the study area after their diagnosis or index year were excluded. Exposures occurring after the diagnosis or index year were not counted.
Overall, 80.7% of selected and eligible cases, 76.4% of the HCFA controls, 79.0% of the deceased controls, and 82.6% of eligible and contacted random-digit dial controls were included in the analysis (Table 1). The reasons for exclusion were failure to be located, to meet the eligibility criteria, and to obtain permission for an interview from a treating physician (cases only); subject refusal; reaching the target number of interviews (random digit dial controls only); and unknown PCE exposure status. This produced 672 cases and 616 controls for the analysis.
The racial and vital status and geographic distributions of included and excluded eligible cases and non-random digit dial controls were quite similar. However, excluded subjects tended to be older than included subjects (e.g., 55% and 58% of excluded cases and non-random digit dial controls, respectively, were [greater than or equal to] 70 years of age).
PCE exposure estimation. Relative exposure to PCE-contaminated water was estimated using an algorithm developed by Webler and Brown (1993) that took into account specific information about the water pipe that supplied each subject's Cape Cod home. An ordinal estimate of exposure to PCE-contaminated water, the relative delivered dose (RDD), was defined as the estimated mass of PCE that entered the home through the drinking water during a specified period. The algorithm for estimating the RDD is based on a model for PCE leaching from vinyl-lined pipe that was developed by Demond (1982). The model assumes that the initial amount of PCE in a pipe is proportional to its inner surface area. The rate at which the initial stock leaches depends on the physical characteristics of the pipe and the water flow. Thus, the Webler-Brown algorithm estimates the pipe's initial stock from its diameter and length, and the leaching rate from the age of the pipe and water flow, including the flow rate (Webler and Brown 1993).
The water flow rate is influenced by the geometry and load on the water distribution system. The Webler-Brown algorithm simplified the effect of geometry by considering four generic cases: dead ends, circles, circles with taps, and in-line (Webler and Brown 1993). A specific subject was considered as one or a combination of these geometries.
The pipe load depends on the number of connected houses, the date of connection, and the water consumption of each house. Computerized parcel and street maps obtained from local authorities were used to determine the location and spacing of the house connections. The time of a household's connection to public drinking water was determined from subject interviews. Subjects who stated that their household was served by a private well for their entire Cape Cod residency (n = 14) were considered unexposed. Water flow was assumed to be unidirectional, and all houses were assumed to draw the same amount of water during a subject's residency.
To implement the Webler-Brown model, the locations, installation dates, and diameters of the VL/AC pipes in all public water supply systems in the study area were determined from maps provided by the local water suppliers or the Massachusetts DEP. Next, the public water distribution system data were digitized using geographic information system (GIS) mapping software (ArcView GIS, Environmental Systems Research Institute, Inc., Redlands, CA). GIS software was also used to geocode all subjects' Cape Cod residences to the parcel level. These data were linked to produce maps with spatial information on the parcel addresses and street locations of the subjects together with information on the water pipe configurations and attributes, including length, diameter, installation year, and composition (VL/AC or not). All residences located on a street with VL/AC pipe or on a street downstream from a VL/AC pipe were set aside for a detailed exposure assessment.
The exposure assessments involved making a judgment about the water flow direction and the load distribution because this information was not provided by the water companies. The water flow direction was determined by examining the water source locations and pipe sizes. The load distribution was determined by judging the point where residences connected to water mains. A protocol was designed so that all judgments were made in a consistent manner. All exposure assessments were performed by a single individual who was unaware of the subject's disease status. Intraobserver agreement was examined by performing 25 assessments twice several months apart, and interobserver agreement was tested by having another individual reevaluate 45% of the assessments. The levels of both intra-and interobserver agreement were excellent (Spearman correlation coefficients = 0.98 and 0.96, respectively; p = 0.0001).
Data analysis. First, ever-exposed versus never-exposed women were compared. Next, women with low and high cumulative RDDs were compared with never-exposed women. A low RDD was defined as an exposure level up to and including median cumulative RDD among exposed controls. Three successive categories of high exposure were also defined: above the 50th percentile, above the 75th percentile, and above the 90th percentile.
The empirical latent period, which is composed of both the induction and latent period, is the interval between the causal action of PCE and the diagnosis date or index year (Rothman 1981). Because it is not known if PCE might act as an early, intermediate, or late causal component, the analysis considered a range of empirical latent periods: 0, 5, 7, 9, 11, 13, 15, 17, and 19 years.
Women were considered exposed if they had at least one exposed residence during the relevant time period. RDDs were summed over all exposed addresses. The cumulative exposure was determined for each empirical latent period assumption. For example, only cumulative exposure that occurred more than 15 years before the diagnosis or index year was counted when a 15-year empirical latent period was assumed.
The exposure odds ratio (OR) was used to estimate the strength of the relationship between PCE exposure and the occurrence of breast cancer. The potential modifying effects of bathing habits, bottled water and water filter use were examined in stratified analyses. Ninety-five percent profile likelihood confidence intervals were used to indicate the precision of the crude associations (SAS 1994).
Multiple logistic regression was used to control simultaneously for potential confounding variables (Rosner 2000). The antilog of the beta coefficient of the exposure variable was used to estimate the OR. Adjusted analyses were conducted only for contrasts with at least three exposed cases and three exposed controls. A group of core confounders was controlled in all regression analyses: age at diagnosis or index year, vital status at interview, family history of breast cancer, personal history of prior breast cancer (before the current diagnosis or index year), age at first live birth or stillbirth, and occupational exposure to PCE. Additional potential confounders were added to the logistic regression models along with the core confounders, including history of benign breast disease; past use of diethylstilbestrol, oral contraceptives, and menopausal hormones; cigarette smoking history; alcohol drinking history; history of ionizing radiation treatment; quetelet index (measure of obesity); race; marital status; religion; educational level; and physical activity level. None of these additional variables changed the adjusted estimates by more than 10%, and so the final model included only the core confounders. We calculated 95% confidence intervals (95% CIs) for the adjusted ORs using maximum likelihood estimates of the standard errors (Rosner 2000).
Analyses were conducted among the present study population and among the present and prior study populations combined. Subject selection and data collection methods for the two studies were alike (Aschengrau et al. 1998). Although subjects in the prior study were permanent residents of fewer Cape Cod towns (Barnstable, Bourne, Falmouth, Mashpee, and Sandwich) during an earlier period (1983-1986), the demographic characteristics and history of breast cancer risk factors were similar in the two populations.
The two studies also had similar exposure assessment methods (Aschengrau et al. 1998). The major differences were that the assessments for the present study used GIS software instead of tax assessor maps, had new information on some additional VL/AC pipe locations, and considered as exposed residences adjacent to and downstream from VL/AC pipe. (The prior study considered as exposed only residences adjacent to VL/AC pipe.) These differences contributed to the higher exposure prevalence in the present study. Adjusted analyses of the combined data took this difference into account by controlling for the study in which the subject was enrolled.
The study subjects were predominantly white, [greater than or equal to] 60 years of age, educated [greater than or equal to] 12 years, and postmenopausal at diagnosis or index year (Table 2). Consistent with the literature, more cases than controls had a first-degree family history of breast cancer and prior benign breast disease and were either nulliparous or had a late age at first birth. A similar proportion of cases and controls had occupational exposure to PCE and had ever lived near a dry cleaner. Water use patterns, including bathing habits, the use of bottled water, and the use of water treatment devices, were also similar.
Overall, 23.1% of cases (n = 155) and 22.1% of controls (n = 136) were considered ever exposed to PCE-contaminated drinking water when latency was ignored, and from 1.3% (19 years' latency) to 19.2% (5 years' latency) of cases and from 1.5% (19 years' latency) to 17.3% (5 years' latency) of controls were considered ever exposed when latency was considered (Table 3). When latency was ignored, the RDD estimates obtained from the Webler-Brown model ranged from 0.001 to 243.8 among exposed controls, and the estimates at the median, 75th, and 90th percentile were 2.5, 12.1, and 29.2, respectively (Table 4). The minimum RDD increased and the maximum RDD decreased as more years of latency were considered. In addition, the cutoffs for median and the 75th percentile increased, whereas the cutoffs for the 90th percentile remained relatively stable as more years of latency were assumed.
No increases or small increases in the crude OR for breast cancer were seen among ever-exposed women either when latency was considered or when it was ignored (ORs, 0.5-1.3; Table 3). The ORs changed only slightly when confounding variables were controlled using logistic regression models (ORs, 1.0-1.4).
When the RDD was dichotomized at the median, no increases or small increases were seen in the crude and adjusted ORs for exposure levels below and above the median when latency was taken into account and when it was ignored (Table 5). However, small to moderate elevations in the crude and adjusted ORS were seen among women whose exposure levels were above the 75th percentile when 0-15 years of latency were taken into account (crude ORs, 1.4-1.7; adjusted ORs, 1.5-1.9). A similar pattern was seen among women whose exposure levels were above the 90th percentile: there were small to moderate increases in the ORs when 0-15 years of latency were taken into account (crude ORs, 1.3-2.5; adjusted ORs, 1.3-2.8). The number of subjects whose exposure level was above the 75th and 90th percentiles was too small to provide meaningful results when more than 15 years of latency was taken into account.
Few changes in these findings were seen when the analyses were limited to postmenopausal women (adjusted ORs, 1.6-1.9 and 1.4-2.6 for exposure levels > 75th and > 90th percentiles, respectively). In addition, stratified analyses of ever-exposed women did not reveal any effect modification by bottled water use. However, ever-exposed subjects who took only baths at their exposed residence consistently had a reduced risk of breast cancer (ORS, 0.7-0.8 for 0-15 years of latency) compared with those who took only showers (ORs, 0.9-1.3 for 0-15 years of latency) and those who took a combination of baths and showers (ORs, 1.0-3.3). In addition, ever-exposed subjects who used a water treatment device at their exposed residence consistently had a reduced risk of breast cancer (ORs, 0.6-0.9 for 0-15 years of latency). The precision of the latter estimates was low because of the small proportion of cases and controls who used water treatment devices (12.9% and 13.7%, respectively; Table 2).
When the data from the present and previous studies were combined, small to moderate increases in the crude OR were seen among ever-exposed subjects when 0-19 years of latency were considered (ORs, 1.3-1.8; Table 6). However, the ORs were reduced when confounding variables were controlled (ORs, 1.0-1.3; Table 6).
In addition, no increases or modest increases in the crude and adjusted ORs were seen in the combined data when the exposure levels were dichotomized at the median (adjusted ORs, 0.9-1.5; Table 7). However, moderate increases in the ORs were seen among women whose exposure levels were above the 75th percentile when 0-15 years of latency were taken into account (adjusted ORs, 1.6-1.9). Several of these ORs were statistically significant.
Increases in the ORs were also observed among women whose exposure levels were above the 90th percentile when 0-15 years of latency were considered; these increases were generally smaller or equal to those observed at the 75th percentile (adjusted ORs, 1.3-1.9). The cutoffs for the 75th and 90th percentiles in the combined data were similar or higher than those in the present study (Table 4). Little or no change in these findings was seen when the combined analyses were limited to postmenopausal women (adjusted ORs, 1.4-2.0 and 1.3-2.2 for exposure levels > 75th and > 90th percentiles, respectively).
When the combined data were stratified according to bottled water use, the relative risks among ever-exposed women who did not consume bottled water were slightly higher (ORs, 1.2-2.1 for 0-17 years' latency) than among those who did consume bottled water (ORs, 0.8-1.4 for 0-17 years' latency). We were unable to examine the impact of bathing habits and water treatment devices in the combined data because of differences in the data collection instruments.
The results of the present study and those of the prior and present studies combined suggest that women with the highest RDDs of PCE-contaminated drinking water have small to moderate increases in the risk of breast cancer compared with unexposed women (adjusted ORs from combined data, 1.6-1.9 for > 75th percentile and 1.3-1.9 for > 90th percentile). The magnitudes of the findings for the two highest exposure levels are similar; however, the precision of the estimates for exposure levels > 75th percentile are greater than those for exposure levels > 90th percentile. No meaningful increases in risk were observed at lower RDDs. Stratified analyses did not reveal any effect modification by bottled water use or bathing habits but suggested that ever-exposed subjects who used a water treatment device had a reduced risk of breast cancer (ORs, 0.6-0.8 for 0-15 years of latency).
These results are likely affected by exposure misclassification because we used the Webler-Brown model to estimate the historical PCE exposures. Incorrect model assumptions or errors in determining the model's input variables would have led to errors in estimating the RDDs. We are currently conducting a validation study of the model by comparing the RDD estimates with actual PCE concentrations in water samples that were taken in 1980 when the contamination on Cape Cod was first discovered. However, because the PCE exposure assessments were conducted blindly, we think that these errors are likely nondifferential, and so associations with the dichotomous exposure measure are likely biased toward the null. Associations involving multiple exposure levels may be biased either toward or away from the null.
The study results are unlikely to be affected by biased selection of cases and controls. Breast cancer cases were obtained from all incident cases reported to the Massachusetts Cancer Registry. Comparison with other state cancer registries indicates nearly complete reporting for this cancer site (Massachusetts Department of Public Health 1995). Follow-up and interview rates were similar for cases and controls as were available demographic characteristics of participants and nonparticipants.
The proportion of nonparticipants who lived on a street with VL/AC pipe at the diagnosis or index year was 14.2% (14.4% for cases and 13.9% for non-RDD controls). This is lower than the exposure prevalence among participants because it is based on only a single address instead of a 40-year residential history. However, it is unlikely that it biased the results because the exposure prevalence is nearly identical for nonparticipating cases and controls.
The study results are also unlikely to be affected by observation bias. Although the interviewers were not masked to the disease status of the subjects, the interview was highly structured; the questions were closed-ended, carefully written, and tested; and the staff was well trained in appropriate interview techniques. In addition, comparable information quality was achieved by selecting deceased controls who had proxy interviews in the same manner as deceased cases. Lastly, the PCE exposure assessments were conducted blindly and independently of the interview.
With respect to confounding, the variables age at diagnosis or index year, vital status at interview, family history of breast cancer, personal history of breast cancer (before the current diagnosis), age at first live birth or stillbirth, and occupational exposure to PCE were controlled in all multivariate analyses. Many additional potential confounders were considered, but none changed the ORs by more than 10%. Residual confounding by unmeasured factors, including other environmental pollutants, is a possible but an improbable explanation for the findings because these factors would have to be relatively common, closely correlated with PCE exposure, and fairly strong risk factors for breast cancer in order to produce the observed increases in the ORs. However, we are currently conducting another population-based case-control study among female residents of all 15 Cape Cod towns in collaboration with Silent Spring Institute in order to investigate the risk of breast cancer associated with nonenvironmental and environmental exposures, and so it will be possible to assess directly the associations between PCE and other exposures in the future.
The International Agency for Cancer Research (IARC) has classified PCE as a probable human carcinogen on the basis of animal experiments and epidemiologic studies (IARC 1995). The National Toxicology Program (NTP) also classifies PCE as an agent reasonably anticipated to be carcinogenic to humans (NTP 2000). Oral administration has produced liver cancer in mice, and inhalation exposure has produced kidney cancer and leukemia in rats and liver cancer in mice [Mennear et al. 1986; National Cancer Institute (NCI) 1977]. Mammary tumors have not been observed in these species.
Most of the human evidence regarding the carcinogenicity of PCE comes from studies of dry cleaners and other occupationally exposed groups (e.g., Blair et al. 1990; Ruder et al. 2001). Although these studies suggest associations with several cancer sites, particularly esophageal, bladder, cervical, intestinal, and lung cancer and non-Hodgkin lymphoma, the findings for breast cancer, as summarized below, are mainly null.
Three proportional mortality studies of women employed as launderers and dry cleaners found a deficit of breast cancer deaths (Duh and Asal 1984; Katz and Jowett 1981; Walker et al. 1997). The largest of these studies (n = 8,163 deaths) found a 9% decreased proportion of breast cancer deaths among white women and a 37% decreased proportion in black women (Walker et al. 1997). Interestingly, there were several cases of male breast cancer, and although substantially more than expected, the numbers were too small to produce stable effect estimates. These employees were combined laundry and dry cleaning workers, and so many were not exposed to PCE.
Cohort studies of mortality among dry cleaning workers primarily exposed to PCE have also reported no increases in breast cancer deaths. A National Institute for Occupational Safety and Health cohort study (n = 1,111 women) found a deficit of breast cancer deaths among all women employed in dry cleaning shops [standardized mortality ratio (SMR), 0.91; 95% CI, 0.55-1.40] and among the subgroup of workers exposed only to PCE (SMR, 0.78; 95% CI, 0.28-1.69) (Ruder et al. 2001). A National Cancer Institute mortality study of dry cleaning union members (n = 4,046 women) also found no increase in the risk of breast cancer death (SMR, 1.0; 95% CI, 0.7-1.4) (Blair et al. 1990). Lastly, a study of mortality among aircraft manufacturing workers found a small but statistically unstable increase in the breast cancer mortality among female workers who were routinely exposed to PCE (SMR, 1.16; 95% CI, 0.32-2.97) (Boice et al. 1999).
The results of two cancer incidence studies are both null for breast cancer. A large cohort study from Denmark, Norway, Sweden, and Finland found an 11% decreased incidence of breast cancer among female launderers and dry cleaners [n = 23,714; standardized incidence ratio (SIR), 89; 95% CI, 83-97] (Andersen et al. 1999). An unknown proportion were exposed to PCE. In addition, a population-based study in the Portland, Oregon-Vancouver, British Columbia, Canada, area found a lower breast cancer incidence rate among women working in laundry and dry cleaning jobs compared with that among all women (age-standardized rates, 77.4 vs. 100.3 per 100,000 person-years, respectively) (Morton 1995).
In contrast, a Canadian case-control study found a 4.9-fold increased risk of breast cancer among postmenopausal women usually employed in laundry and dry cleaning (95% CI, 1.3-18.7) (Band et al. 2000). This study was unusual because it controlled for the well-established breast cancer risk factors. In addition, a proportional mortality ratio (PMR) study found a statistically significant excess of breast cancer in male dry cleaning and laundry workers [PMR = 1,275 in white men (n = 4) and PMR = 1,587 in black men (n = 2)] (Walker et al. 1997).
Studies of workers exposed to trichloroethylene (TCE) are also relevant to the interpretation of our findings because TCE and PCE are closely related chlorinated ethylenes and share a similar toxicology (IARC 1995). A retrospective cohort study of aircraft maintenance workers (n = 3,717 women) found a 1.8-fold increase in the breast cancer mortality rate (95% CI, 0.9-3.3) among women who were mainly exposed to TCE compared with those with no chemical exposure (Blair et al. 1998). Another study of aircraft maintenance workers found a 1.3-fold increase in the breast cancer mortality rate among women routinely exposed to TCE (n = 192; 95% CI, 0.5-2.7) (Boice et al. 1999). In contrast, a Finnish cohort study of workers who had been biologically monitored for TCE, PCE, and tricholoroethane found a decreased incidence of breast cancer among women (n = 1,924) exposed to any of these halogenated hydrocarbons (SIR, 0.84; 95% CI, 0.44-1.48) (Anttila et al. 1995). Two mortality studies of aircraft manufacturing workers exposed to TCE also found decreased breast cancer death rates [SMR, 0.91; 95% CI, 0.52-1.48 (Garabrant et al. 1998); SMR, 0.75; 95% CI, 0.55-1.00 (Morgan et al. 1998)].
Most of these occupational studies have a number of limitations that likely reduce their ability to detect an association with breast cancer. First, they suffer from exposure misclassification, stemming mainly from the use of broad job titles to classify individuals, and grouping together workers with a variety of chemical exposures. Second, most of the studies included a relatively small number of female subjects and so have low statistical power to detect small to moderate associations. Third, competing causes of death may account for the unelevated rates of breast cancer observed in the mortality studies.
Lastly, almost all of the occupational studies are missing information on key breast cancer confounders, such as age at first birth, family history of breast cancer, and socioeconomic status. In particular, because most women employed in dry cleaning and laundering jobs are from low socioeconomic classes, their baseline risk of breast cancer is lower than that of the general population (Kelsey 1993). Supporting evidence comes from a study of breast cancer incidence by major occupational groups that found a 33% lower incidence among blue-collar workers than among white-collar workers (Morton 1995). Thus, it is quite likely that the occupational studies with a general population comparison group are biased toward the null because of residual confounding by socioeconomic status.
Recently, Labreche and Goldberg (1997) described a hypothesis that lipophilic solvents like PCE can accumulate in breast tissue and have carcinogenic properties by acting either directly as genotoxic agents or indirectly through their metabolites. Although there are currently few epidemiologic studies on a relationship with breast cancer, either pro or con, this direction of research should be pursued in order to identify preventable causes of this common form of cancer.
Table 1. Selection and enrollment of breast cancer cases and controls. HCFA Deceased Random digit Cases controls controls dial controls Selected 858 468 234 3,402 Excluded Never located or contacted 54 16 20 515 Ineligible (a) 25 74 34 2,639 Physician not located or refused 41 -- -- -- Subject refusal 65 77 19 33 Target reached (b) -- -- -- 58 PCE exposure status unknown 1 -- 3 -- Included in analysis 672 301 158 157 (a) Includes cases and non-random digit dial controls who did not meet residence criteria (n = 98), cases without breast cancer (n = 2), HCFA controls who died (n = 33), and random digit dial controls who did not meet the age, sex, and residence criteria (n = 2,319) or who refused to answer the eligibility questions (n = 320). (b) These random digit dial controls were identified as eligible but not contacted for interview because the target number of random digit dial controls had already been interviewed. Table 2. Distribution of selected characteristics of breast cancer cases and controls (%). Cases Controls Characteristic (n = 672) (n = 616) White 98.4 95.8 Age at diagnosis or index year (years) 1-49 16.5 16.7 50-59 12.2 13.6 60-69 31.5 29.9 70-79 28.4 26.0 [greater than or equal to] 80 11.3 13.8 Educational level [greater than or equal to] 12 years 90.2 87.4 Alive at interview 71.9 74.4 Age at first live birth or stillbirth < 30 years 61.0 66.8 [greater than or equal to] 30 years 14.5 12.8 Nulliparous 24.5 20.4 Prior breast cancer (a) 5.4 4.8 Prior benign breast disease (b) 24.9 20.6 Family history of breast cancer (c) 25.6 15.5 Postmenopausal at diagnosis or 88.7 84.6 index year Definite or possible occupational 15.5 14.8 exposure to PCE (b,d) Residence near dry cleaner (b) 0.6 0.2 Bathing habits (b,e) Only baths 12.7 15.9 Only showers 40.6 35.6 Both baths and showers 46.7 48.5 Regularly drank bottled water (b,e) 28.2 27.7 Used a water treatment device (b,e) 12.9 13.7 (a) These women had a prior breast cancer diagnosis before current diagnosis year for cases and index year for controls. (b) Before the diagnosis or index year. (c) Mother or sister. (d) Exposed jobs included work in dry cleaning, chemical processing and manufacturing, metal degreasing and cleaning, and textile and electrical transformer manufacturing. (e) While living at their Cape Cod residence(s). Table 3. PCE exposure history of breast cancer cases and controls, crude (a) and adjusted (b) ORs, and 95% CIs. Latency PCE-exposed PCE-exposed period cases controls Crude OR Adjusted OR (years) (n= 672) (n = 616) (95% CI) (95% CI) 0 155 136 1.1 (0.8-1.4) 1.1 (0.8-1.4) 5 129 107 1.1 (0.8-1.5) 1.2 (0.9-1.6) 7 111 96 1.1 (0.8-1.4) 1.1 (0.8-1.5) 9 97 85 1.1 (0.8-1.5) 1.1 (0.8-1.5) 11 79 65 1.1 (0.8-1.6) 1.2 (0.8-1.7) 13 61 45 1.3 (0.8-1.9) 1.3 (0.9-2.0) 15 44 31 1.3 (0.8-2.1) 1.4 (0.9-2.3) 17 21 21 0.9 (0.5-1.7) 1.0 (0.6-2.0) 19 9 9 0.9 (0.4-2.4) 1.1 (0.4-2.9) (a) The OR was calculated relative to never-exposed cases (n = 517) and controls (n = 480). (b) Controlled for age at diagnosis or index year, vital status at interview, family history of breast cancer, personal history of breast cancer (before current diagnosis or index year), age at first live birth or stillbirth, and occupational exposure to PCE. Table 4. Distribution of cumulative RDDs among PCE-exposed controls according to latency period for the present study only and for prior and present studies combined. Latency period 75th 90th (years) Minimum Maximum Median percentile percentile Present study 0 0.001 243.8 2.5 12.1 29.2 5 0.02 243.2 4.0 14.2 38.0 7 0.05 242.1 4.6 14.4 37.9 9 0.03 239.4 5.9 14.4 35.0 11 0.1 233.0 6.3 13.9 37.3 13 0.1 217.5 10.2 16.9 36.8 15 0.6 200.6 10.3 18.3 33.9 17 1.3 191.6 8.2 21.5 40.6 19 2.6 169.6 13.6 19.8 169.6 Present and prior studies combined 0 0.001 243.8 3.6 15.5 41.8 5 0.02 243.2 6.9 17.6 41.7 7 0.05 242.1 6.9 18.2 40.9 9 0.03 239.4 6.4 16.5 38.4 11 0.1 233.0 6.8 18.5 37.3 13 0.1 217.5 10.3 18.9 36.8 15 0.6 200.6 10.3 18.3 49.1 17 1.3 191.6 8.2 21.5 40.6 19 2.6 169.6 13.6 19.8 169.9 Table 5. Crude (a) (COR) and adjusted (b) ORs (AOR) for breast cancer according to various PCE exposure levels in the present study. PCE exposure level Latency [less than or period (years) equal to] Median > Median 0 Case/control (n) 68/68 87/68 COR (95% CI) 0.9 (0.6-1.3) 1.2 (0.8-1.7) AOR (95% CI) 0.9 (0.7-1.4) 1.2 (0.9-1.7) 5 Case/control (n) 58/54 71/53 COR (95% CI) 1.0 (0.7-1.5) 1.2 (0.9-1.8) AOR (95% CI) 1.1 (0.7-1.6) 1.3 (0.9-1.9) 7 Case/control (n) 42/48 69/48 COR (95% CI) 0.8 (0.5-1.3) 1.3 (0.9-2.0) AOR (95% CI) 0.8 (0.5-1.3) 1.4 (0.9-2.1) 9 Case/control (n) 42/43 55/42 COR (95% CI) 0.9 (0.6-1.4) 1.2 (0.8-1.9) AOR (95% CI) 1.0 (0.6-1.5) 1.3 (0.8-2.0) 11 Case/control (n) 36/33 43/32 COR (95% CI) 1.0 (0.6-1.7) 1.2 (0.8-2.0) AOR (95% CI) 1.1 (0.6-1.7) 1.3 (0.8-2.2) 13 Case/control (n) 32/23 29/22 COR (95% CI) 1.3 (0.7-2.2) 1.2 (0.7-2.2) AOR (95% CI) 1.4 (0.8-2.4) 1.3 (0.7-2.4) 15 Case/control (n) 26/16 18/15 COR (95% CI) 1.5 (0.8-2.8) 1.1 (0.6-2.2) AOR (95% CI) 1.6 (0.9-3.1) 1.2 (0.6-2.4) 17 Case/control (n) 11/11 10/10 COR (95% CI) 0.9 (0.4-2.2) 0.9 (0.4-2.3) AOR (95% CI) 1.0 (0.4-2.5) 1.1 (0.4-2.6) 19 Case/control (n) 6/5 3/4 COR (95% CI) 1.1 (0.3-3.7) 0.7 (0.2-3.1) AOR (95% CI) 1.3 (0.4-4.4) 0.8 (0.2-3.9) PCE exposure level Latency period (years) > 75th Percentile > 90th Percentile 0 Case/control (n) 52/34 28/13 COR (95% CI) 1.4 (0.9-2.2) 2.0 (1.0-3.9) AOR (95% CI) 1.5 (1.0-2.4) 2.2 (1.1-4.3) 5 Case/control (n) 43/26 14/10 COR (95% CI) 1.5 (0.9-2.5) 1.3 (0.6-3.0) AOR (95% CI) 1.6 (1.0-2.7) 1.3 (0.5-3.0) 7 Case/control (n) 41/24 14/9 COR (95% CI) 1.6 (0.9-2.7) 1.4 (0.6-3.4) AOR (95% CI) 1.7 (1.0-2.9) 1.4 (0.6-3.4) 9 Case/control (n) 36/21 15/8 COR (95% CI) 1.6 (0.9-2.8) 1.7 (0.7-4.1) AOR (95% CI) 1.7 (1.0-3.0) 1.8 (0.8-4.4) 11 Case/control (n) 30/16 11/6 COR (95% CI) 1.7 (0.9-3.2) 1.7 (0.6-4.6) AOR (95% CI) 1.9 (1.0-3.5) 1.6 (0.6-4.5) 13 Case/control (n) 19/11 8/4 COR (95% CI) 1.6 (0.8-3.4) 1.9 (0.6-6.2) AOR (95% CI) 1.7 (0.8-3.7) 1.9 (0.6-6.6) 15 Case/control (n) 12/7 8/3 COR (95% CI) 1.6 (0.6-4.1) 2.5 (0.7-9.4) AOR (95% CI) 1.8 (0.7-4.8) 2.8 (0.7-11.0) 17 Case/control (n) 4/5 1/2 COR (95% CI) 0.7 (0.2-2.8) 0.5 (0.0-5.1) AOR (95% CI) 0.9 (0.2-3.5) -- (c) 19 Case/control (n) 2/2 0/0 COR (95% CI) 0.9 (0.1-6.6) 0.0 (--) AOR (95% CI) -- (c) -- (c) (a) The OR was calculated relative to never-exposed cases (n = 517) and controls (n = 480). (b) Controlled for age at diagnosis or index year, vital status at interview, family history of breast cancer, personal history of breast cancer (before current diagnosis or index year), age at first live birth or stillbirth, and occupational exposure to PCE. (c) Adjusted analyses were not performed if there were fewer than three exposed cases and three exposed controls. Table 6. PCE exposure history of breast cancer cases and controls, including crude (a) and adjusted (b) ORs (95% CIs) of combined data from previous and present studies. Latency PCE-exposed PCE-exposed period cases controls Crude Adjusted (years) (n = 930) (n = 1,302) OR (95% CI) OR (95% CI) 0 191 217 1.3 (1.0-1.6) 1.1 (0.9-1.4) 5 154 163 1.4 (1.1-1.8) 1.2 (0.9-1.5) 7 128 135 1.4 (1.1-1.8) 1.1 (0.8-1.5) 9 111 110 1.5 (1.1-2.0) 1.2 (0.9-1.6) 11 86 83 1.5 (1.1-2.1) 1.1 (0.8-1.6) 13 65 52 1.8 (1.3-2.7) 1.3 (0.9-2.0) 15 44 35 1.8 (1.2-2.9) 1.3 (0.8-2.1) 17 21 21 1.5 (0.8-2.7) 1.0 (0.5-1.9) 19 9 9 1.5 (0.6-3.7) 1.1 (0.4-2.9) (a) The OR was calculated relative to never-exposed cases (n = 739) and controls (n = 1,085). (b) Controlled for age at diagnosis or index year, vital status at interview, family history of breast cancer, personal history of breast cancer (before current diagnosis or index year), age at first live birth or stillbirth, occupational exposure to PCE, and study of origin (present or prior). Table 7. Crude (a) (COR) and adjustedb (AOR) ORs for breast cancer according to various PCE exposure levels: combined data from present and previous studies. PCE exposure level Latency [less than or period (years) equal to] Median > Median 0 Case/control (n) 91/109 100/108 COR (95% CI) 1.2 (0.9-1.6) 1.4 (1.0-1.8) AOR (95% CI) 1.0 (0.7-1.3) 1.2 (0.9-1.7) 5 Case/control (n) 79/82 75/81 COR (95% CI) 1.4 (1.0-2.0) 1.4 (1.0-1.9) AOR (95% CI) 1.1 (0.8-1.5) 1.3 (0.9-1.8) 7 Case/control (n) 59/68 69/67 COR (95% CI) 1.3 (0.9-1.8) 1.5 (1.1-2.1) AOR (95% CI) 0.9 (0.6-1.3) 1.3 (0.9-1.9) 9 Case/control (n) 48/55 63/55 COR (95% CI) 1.3 (0.9-1.9) 1.7 (1.2-2.4) AOR (95% CI) 0.9 (0.6-1.4) 1.4 (0.9-2.0) 11 Case/control (n) 39/42 47/41 COR (95% CI) 1.4 (0.9-2.1) 1.7 (1.1-2.6) AOR (95% CI) 1.0 (0.6-1.5) 1.4 (0.9-2.1) 13 Case/control (n) 35/26 30/26 COR (95% CI) 2.0 (1.2-3.3) 1.7 (1.0-2.9) AOR (95% CI) 1.4 (0.8-2.3) 1.3 (0.8-2.3) 15 Case/control (n) 26/18 18/17 COR (95% CI) 2.1 (1.2-3.9) 1.6 (0.8-3.0) AOR (95% CI) 1.5 (0.8-2.8) 1.1 (0.6-2.3) 17 Case/control (n) 11/11 10/10 COR (95% CI) 1.5 (0.6-3.4) 1.5 (0.6-3.5) AOR (95% CI) 1.0 (0.4-2.4) 1.0 (0.4-2.6) 19 Case/control (n) 6/5 3/4 COR (95% CI) 1.8 (0.5-5.8) 1.1 (0.2-4.9) AOR (95% CI) 1.3 (0.4-4.2) 0.9 (0.2-4.1) PCE exposure level Latency period (years) > 75th percentile > 90th percentile 0 Case/control (n) 59/54 18/21 COR (95% CI) 1.6 (1.1-2.3) 1.3 (0.7-2.4) AOR (95% CI) 1.6 (1.1-2.4) 1.3 (0.7-2.6) 5 Case/control (n) 50/40 17/16 COR (95% CI) 1.8 (1.2-2.8) 1.6 (0.8-3.1) AOR (95% CI) 1.6 (1.0-2.6) 1.5 (0.7-3.0) 7 Case/control (n) 46/33 17/13 COR (95% CI) 2.0 (1.3-3.2) 1.9 (0.9-4.0) AOR (95% CI) 1.8 (1.1-2.9) 1.7 (0.8-3.6) 9 Case/control (n) 40/27 16/11 COR (95% CI) 2.2 (1.3-3.6) 2.1 (1.0-4.6) AOR (95% CI) 1.9 (1.1-3.2) 1.9 (0.8-4.4) 11 Case/control (n) 29/20 12/8 COR (95% CI) 2.1 (1.2-3.8) 2.2 (0.9-5.4) AOR (95% CI) 1.8 (1.0-3.3) 1.8 (0.7-4.8) 13 Case/control (n) 17/13 8/5 COR (95% CI) 1.9 (0.9-4.0) 2.3 (0.8-7.2) AOR (95% CI) 1.6 (0.7-3.5) 1.7 (0.5-5.2) 15 Case/control (n) 12/8 2/3 COR (95% CI) 2.2 (0.9-5.4) 1.0 (0.2-5.9) AOR (95% CI) 1.7 (0.7-4.3) -- (c) 17 Case/control (n) 4/5 1/2 COR (95% CI) 1.2 (0.3-4.4) 0.7 (0.1-8.1) AOR (95% CI) 0.9 (0.2-3.4) -- (c) 19 Case/control (n) 2/2 0/0 COR (95% CI) 1.5 (0.2-10.4) 0.0 (--) AOR (95% CI) -- (c) -- (c) (a) The OR was calculated relative to never-exposed cases (n = 739) and controls (n = 1,085). (b) Controlled for age at diagnosis or index year, vital status at interview, family history of breast cancer, personal history of breast cancer (before current diagnosis or index year), age at first live birth or stillbirth, occupational exposure to PCE, and study of origin (present or prior). (c) Adjusted analyses were not performed if there were fewer than three exposed cases and three exposed controls.
Andersen A, Barlow L, Engeland A, Kjaerheim K, Lynge E, Pukkala E. 1999. Work-related cancer in the Nordic countries. Scand J Work Environ Health 25(suppl 2):1-116.
Anttila A, Pukkala E, Sallmen M, Hernberg S, Hemminki K. 1995. Cancer incidence among Finnish workers exposed to halogenated hydrocarbons. J Occup Environ Mod 37:797-806.
Aschengrau A, Paulu C, Ozonoff D. 1998. Tetrachloroethylene-contaminated drinking water and the risk of breast cancer. Environ Health Perspect 106(suppl 4): 947-953.
ATSDR. 1995. Toxicological Profile for Tetrachloroethylene. Atlanta, GA:Agency for Toxic Substances Disease Registry.
Band PR, Le ND, Fang R, Deschamps M, Gallagher RP, Yang P. 2000. Identification of occupational cancer risks in British Columbia. J Occup Environ Med 42:284-310.
Blair A, Hartge P, Stewart PA, McAdams M, Lubin J. 1998. Mortality and cancer incidence of aircraft maintenance workers exposed to trichloroethylene and other organic solvents and chemicals: extended follow up. J Occup Environ Med 55:161-171.
Blair A, Stewart A, Tolbert P, Grauman D, Moran FX, Vaught J, et al. 1990. Cancer and other causes of death among a cohort of dry cleaners. Br J Ind Med 47:162-168.
Boice JD, Marano DE, Fryzek JP, Sadler CJ. 1999. Mortality among aircraft manufacturing workers. Occup Environ Med 56:581-597.
Bureau of the Census. 1990. Equipment and Fuels, Massachusetts. Washington, DC:U.S. Department of Commerce.
Demond AH. 1982. A Source of Tetrachloroethylene in the Drinking Water of New England: An Evaluation of the Toxicity of Tetrachloroethylene and the Prediction of Its Leaching Rates from Vinyl-lined Asbestos-cement Pipe [MS Thesis]. Cambridge, MA:Massachusetts institute of Technology.
Duh R-W, Asal NR. 1984. Mortality among laundry and dry cleaning workers in Oklahoma. Am J Public Health 74:1278-1280.
Garabrant DH, Held J, Langholz B, Bernstein L. 1988. Mortality of aircraft manufacturing workers in Southern California. Am J Ind Med 13:683-693.
Hatten J. 1980. Medicare's common denominator: the covered population. Health Care Financ Rev 2:53-64.
IARC. 1995. Dry Cleaning, Some Chlorinated Solvents, and Other Industrial Chemicals. IARC Monogr Eval Carcinog Risks Hum 63.
Katz RM, Jowett D. 1981. Female laundry and dry cleaning workers in Wisconsin: a mortality analysis. Am J Public Health 71:305-307.
Kelsey JL. 1993. Breast cancer epidemiology. Epidemiol Rev 15:256-263.
Labreche FP, Goldberg MS. 1997. Exposure to organic solvents and breast cancer in women: a hypothesis, Am J Ind Mod 32:1-14.
Larsen CD, Love TO, Reynolds G, 1983. Tetrachloroethylene leached from lined asbestos-cement pipe into drinking water. J Am Water Works Assoc 75:184-188.
Massachusetts Department of Environmental Quality Engineering. 1982. Status Report on Tetrachloroethylene Contamination of Public Drinking Water Supplies by Vinyl-Lined Asbestos Cement Pipe. Boston, MA:Commonwealth of Massachusetts.
Massachusetts Department of Public Health, 1995, Cancer Incidence in Massachusetts 1982-1992: City/Town Supplement. Boston, MA:Massachusetts Department of Public Health.
Mennear J, Maronpot R, Doorman G, Eustis S, Huff J, Haseman J, et al. 1986. Toxicological and carcinogenic effects of inhaled tetrachloroethylene in rats and mice. In: New Concepts and Developments in Toxicology: Proceedings of the Fourth International Congress of Toxicology held in Tokyo, Japan, July 21-25, 1986 (Chambers P, Gehring P, Sakai F, eds). New York:Elsevier, 201-210.
Morgan RW, Kelsh MA, Zhao K, Heringer S, 1998. Mortality of aerospace workers exposed to trichloroethylene. Epidemiology 9:424-431.
Morton WE. 1995. Major differences in breast cancer risks among occupations. J Occup Environ Med 37:328-335.
NCI. 1977. Bioassay of Tetrachloroethylene for Possible Carcinogenicity (CAS No. 127-18-4). Technical Report 13. Bethesda, MD:National Cancer Institute. Available: http://ntp-sarvar.niehs.nih.gov/htdocs/LT-studies/ TR013.html [accessed 26 November 2002].
NTP. 2000. Report on Carcinogens, Ninth Edition. Research Triangle Park, NC:National Toxicology Program.
Paulu C, Aschengrau A, Ozonoff D, 1999, Tetrachloroethylene-contaminated drinking water in Massachusetts and the risk of colon-rectum, lung, and other cancers. Environ Health Perspect 107:265-271.
Rosner B. 2000. Fundamentals of Biostatistics. 5th ed. Pacific Grove, CA:Duxbury.
Rothman KJ. 1981. Induction and latent periods. Am J Epidemiol 114:253-259.
Ruder AM, Ward EM, Brown DP. 2001. Mortality in dry-cleaning workers: an update. Am J Ind Med 39:121-132.
SAS. 1994. JMP Statistics and Graphics Guide, Version 3 of JMP. Cary, NC:SAS Institute, Inc.
Walker JT, Burnett CA, Lalich NR, Sestito JP, Halperin WE, 1997. Cancer mortality among laundry and dry cleaning workers. Am J Ind Mod 32:614-619.
Webler T, Brown HS. 1993. Exposure to tetrachloroethylene via contaminated drinking water pipes in Massachusetts: a predictive model. Arch Environ Health 48:293-297.
Ann Aschengrau, (1) Sarah Rogers, (1) and David Ozonoff (2)
(1) Department of Epidemiology; and (2) Department of Environmental Health, Boston University School of Public Health, Boston, Massachusetts, USA
Address correspondence to A. Aschengrau, 715 Albany Street, Boston, MA 02118 USA. Telephone: (617) 638-5228. Fax: (617) 638-4458. E-mail: email@example.com
We thank the study participants, the local water companies, the Massachusetts Department of Public Health, and the Health Care Financing Administration.
The project was supported by grant 2P42 ES07381 from the National Institute of Environmental Health Sciences (NIEHS) with funds from the U.S. Environmental Protection Agency (EPA). Its contents are solely the authors' responsibility and do not necessarily represent the official views of the NIEHS or the EPA.
Received 15 May 2001; accepted 23 July 2002.
|Printer friendly Cite/link Email Feedback|
|Publication:||Environmental Health Perspectives|
|Date:||Feb 1, 2003|
|Previous Article:||Personal P[M.sub.2.5] exposure and markers of oxidative stress in blood. (Research).|
|Next Article:||Aggressive behavior and serum testosterone concentration during the maturation process of male mice: the effects of fetal exposure to bisphenol A:...|