Risk of Female Breast Cancer Associated with Serum Polychlorinated Biphenyls and 1,1-Dichloro-2,29-bis(p-chlorophenyl)ethylene

This case-control study was designed to investigate the relationship between polychlorinated biphenyls (PCBs) and 1,1-dichloro-2,2*-bis(p-chlorophenyl)ethylene (DDE) and breast cancer risk in Connecticut. Cases were incident breast cancer patients who were either residents of Tolland County or who had a breast-related surgery at the Yale-New Haven Hospital in New Haven County. Controls were randomly selected from Tolland County residents or from patients who had newly diagnosed benign breast diseases or normal tissue at Yale-New Haven Hospital. A total of 475 cases and 502 controls had their serum samples analyzed for PCBs and DDE in 1995–1997. The ageand lipid-adjusted geometric mean serum level of DDE was comparable between the cases (460.1 ppb) and controls (456.2 ppb). The geometric mean serum level of PCBs was also comparable between cases (733.1 ppb) and controls (747.6 ppb). After adjustment for confounding factors, odds ratios of 0.96 (95% confidence interval, 0.67–1.36) for DDE and 0.95 (95% confidence interval, 0.68–1.32) for PCBs were observed when the third tertile was compared with the lowest. Further stratification by parity, lactation, and menopausal and estrogen receptor status also showed no significant association with serum levels of DDE or PCBs. The results by PCB congener groups also showed no major increased risk associated with any of the congener groups. Our study does not support the hypothesis that DDE and PCBs, as encountered through environmental exposure, increase the risk of female breast cancer. Introduction Environmental exposure to organochlorine compounds, particularly PCBs, DDT, and its most stable metabolite, DDE, recently have been suggested as risk factors for female breast cancer (1–4). It is conceivable that exposure to these environmental contaminants may increase breast cancer risk because some of the organochlorine compounds are animal carcinogens, estrogenically active, and inducers of cytochrome P-450 mixedfunction oxidase enzymes, which are involved in steroid hormone metabolism (5–11). Epidemiological studies linking PCB and DDE exposure to breast cancer risk, however, have produced inconclusive results. Among six follow-up studies that examined the relationship between PCBs and DDE and the risk of breast cancer, using a nested case-control study design, one found a doseresponse relationship between breast cancer risk and serum DDE levels, and a possible threshold effect with serum PCB levels (12). Another study by Krieger et al. (13) suggested an increased risk of breast cancer associated with higher serum levels of DDE among Caucasian and African-American women. Four other recent studies, however, did not find an increased risk associated with serum levels of either DDE or PCBs (14–17). Several pilot studies have assessed the relationship between levels of PCBs or DDE in adipose tissue and breast cancer risk (18–22). Four of the studies found higher adipose tissue levels of PCBs among breast cancer cases than noncancer controls (18–21), and one suggested an increased risk of breast cancer associated with body levels of DDE in women with ER-positive breast cancer (22). Two recent larger casecontrol studies, one using breast adipose tissue (23) and the other using buttock adipose tissue (24), did not find a positive association between adipose tissue levels of DDE and DDT and breast cancer risk. Three case-control studies that used blood drawn after the diagnosis of breast cancer assessed the risk of breast cancer associated with serum levels of PCBs or DDE (25–27). Although the studies conducted in North Vietnam (26) and Mexico (27) did not find an association between serum DDT and DDE levels and breast cancer risk, a study from upstate New York by Moysich et al. (25) did suggest an increased risk of breast cancer associated with serum levels of total PCBs among postmenopausal parous women who had never breast-fed an infant (OR 5 2.9; 95% CI, 1.0–7.3). Although recent epidemiological studies have not supported an overall association between PCBs and DDE exposure and breast cancer risk, several studies that have information on parity, lactation, and hormone receptor status seem to suggest Received 4/26/99; revised 11/10/99; accepted 11/30/99. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 Supported by Grant CA-62986 from the National Cancer Institute/National Institute of Environmental Health Sciences. 2 To whom requests for reprints should be addressed, at Department of Epidemiology, Yale University School of Medicine, 129 Church Street, Suite 700-704, New Haven, CT 06510. Phone: (203) 785-2882; Fax: (203) 764-9782; E-mail: [email protected]. 3 The abbreviations used are: PCB, polychlorinated biphenyl; DDT, 2,29-bis(pchlorophenyl)-1,1,1-trichloroethane; DDE, 1,1-dichloro-2,29-bis(p-chlorophenyl)ethylene; ER, estrogen receptor; OR, odds ratio; CI, confidence interval; YNHH, Yale-New Haven Hospital; BMI, body mass index. 167 Vol. 9, 167–174, February 2000 Cancer Epidemiology, Biomarkers & Prevention on August 27, 2017. © 2000 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from that reproductive factors and hormone receptor status may have an impact on the relationship between PCB and DDE exposures and subsequent development of breast cancer. To further address this issue, we report here the results from a case-control study that examined the relationship between serum levels of DDE and PCBs and breast cancer risk by menopausal status, parity and lactation, and by cases’ hormone receptor status. Materials and Methods Study Subjects. Cases, recruited from 1995 to 1997, were histologically confirmed, incident breast cancer patients (International Classification of Diseases for Oncology, 174.0–174.9) who were either residents of Tolland County, or who had a breast related surgery at YNHH, in New Haven County, Connecticut. Cases and controls were 30–80 years of age, had no previous diagnosis of cancer, with the exception of non-melanoma skin cancer, were alive at the time of interview, and were willing to donate at least 10 ml of blood for organochlorine compound analyses. Potentially eligible cases and controls from YNHH were identified using computerized patient information from the YNHH Surgical Pathology Department, where records of all newly completed breast-related surgeries are kept. We consecutively selected all breast cancer patients who met the study eligibility requirements as described above. A total of 326 incident breast cancer cases were recruited from YNHH. From the computerized files, we also randomly selected 347 controls who had had breast-related surgery and were histologically diagnosed with benign breast diseases. Efforts were made to frequency match the cases and controls by age within 5-year intervals (e.g., 30–34, 35–39, 40–44) with a 1:1 ratio by adjusting the number of controls randomly selected in each age stratum every few months. Of the 347 YNHH controls, 37 subjects were diagnosed with normal tissue, 45 with fibroadenoma, 107 with other nonproliferative benign breast diseases, and 158 with proliferative benign breast diseases without atypia. Diagnoses of atypical hyperplasia were excluded. The participation rates were 71% for controls and 77% for cases among the YNHH patients. In addition to the YNHH cases and controls drawn largely from New Haven County, we recruited cases and controls from Tolland County, Connecticut. The two counties have similar breast cancer incidence rates, and in recent years, also have had similar breast cancer mortality rates. Newly diagnosed cases with Tolland County addresses were identified from area hospital records by the Rapid Case Ascertainment Shared Resource of the Yale Cancer Center. A total of 149 cases were recruited. Population-based controls with Tolland County addresses were recruited using either random digit dialing methods for those below age 65 or from Health Care Finance Administration files for those age 65 and above. A total of 155 controls were recruited. Efforts were also made to frequency match the cases and controls by age within 5-year intervals with a 1:1 ratio by adjusting the number of controls randomly selected in each age stratum. The participation rates were 61% for controls and 74% for cases in Tolland County. The study pathologist (D. C.) reviewed all of the pathological diagnoses for breast cancer patients and benign breast disease controls diagnosed at YNHH and also reviewed the pathology reports for the 149 cases ascertained from Tolland County. Carcinomas were classified as in situ, invasive ductal, or invasive lobular, and were staged according to the TNM system (28). For patients diagnosed at YNHH, we also collected information on ER levels, which were measured immunohistochemically at the Pathology Department of YNHH. ER status was considered positive when the H-score was higher than 75, as described by McCarty et al. (29). Treatment information for breast cancer patients was also collected from the Yale New Haven Hospital Tumor Registry, where computerized files contain information (including treatment data) for all cancer patients seen at YNHH. Interviews. After approval by each subject’s hospital and physician, potential participants were approached by letter and then by phone, and those who consented were interviewed by a trained interviewer, either in their homes or at locations convenient for the subjects. A standardized, structured questionnaire was used to obtain information on major known or suspected confounding factors, including menstrual and reproductive history, lactation history, past medical history, family cancer history, occupation, diet, and demographic factors. Dietary information was collected using a scannable semiquantitative food frequency questionnaire developed by the Fred Hutchinson Cancer Research Center, designed to optimize estimation of fat intake. Each subject was asked to characterize her usual diet in the year prior to being interviewed for our study. Following the interview, the participant provided a blood sample, collected by venipuncture by our study staff. Blood Collection and Chemical Analysis. Blood samples were held in a cooler until serum was separated, usually within 1–3 h. The samples were then coded and stored in our study freezer at 284°C until they were sent in batches to the study laboratory at Colorado State University. All samples were kept frozen until analysis. Serum samples were analyzed in batches of 12, with each batch having ;5 cases, 5 controls, and 2 quality control samples. Laboratory personnel in Colorado were blind to the case-control status of samples being analyzed. The analytical methods for determining PCB and DDE levels in 1.0 ml of serum have been described elsewhere (30). Briefly, the method involved denaturation of protein by methanol, extraction of the compounds of interest in ethyl ether: hexane (1:1, v/v), gravimetric lipid determination, purification of the sample using Florisil chromatography, and identification and quantification of the compounds by gas chromatography. Serum residue results are reported as ppb on a lipid adjusted basis. To obtain a lipid adjusted residue value, the wet weight value was divided by the serum’s lipid content, and reported as nanograms of compound per gram of lipid. The quantitation limit (the smallest amount of a compound that can be quantified consistently) by this method was 1.5 ppb for both PCBs and DDE. The detection limits of the method for PCBs and DDE were half of their quantitation limits. For DDE, 95% of the samples were above the detection limit. For PCBs, .70% of the samples were above the detection limit. No consistent way of dealing with values below the detection limit is given among the studies in the literature. We used two approaches: no correction for values below the detection limit, and correction for these values (substituting a hypothetical value equal to half of the detection limit for the assay for nondetectable subjects). Both of the analyses reached the same conclusion. Therefore, we present only the results with corrections. In this study, total PCBs was defined as the sum of the following measured PCB congeners: 74, 118, 138, 153, 156, 170, 180, 183, and 187. Total DDE was defined as the serum level of p,p9-DDE. Strict quality control/quality assessment procedures were followed throughout sample analyses, including method spikes, reagent blanks, and quality control windows. Estimated recovery of the various analytes (including p,p9-DDE, and nine PCB 168 PCBs, DDE, and Breast Cancer Risk on August 27, 2017. © 2000 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from congeners) exceeded 95%, and the coefficients of variation for the various analytes were 9–15%. Data Analysis. The primary analyses involved comparisons of serum levels of DDE and PCBs between all cases and controls. Because the distribution of PCBs and DDE was skewed, a log transformation was used to better approximate the normality assumption, and thus its antilog provided the geometric mean. The age-adjusted geometric means were found by analysis of covariance on the log exposure, and the antilog of the least squares means provided summary statistics. The statistical significance for the adjusted geometric means of serum levels of DDE and PCBs was determined using analysis of covariance. Because earlier studies have suggested that environmental estrogens may affect only the incidence of hormone-responsive breast cancer (22, 31), serum levels of PCBs and DDE were also compared based on the cases’ ER status. It has been suggested that serum levels of PCBs and DDE may be artifactually increased in late-stage patients because of mobilization of energy from fat stores (32); therefore, we divided the cases into early (stages 0, I, and II) and later stages (stages III and IV), and each case category was compared with the control group. An earlier study suggested that chemotherapy might increase the serum level of PCBs, but not DDE (33). Therefore, we compared controls with breast cancer patients based on type of treatment and the elapsed time between start of treatment and the time that blood samples were drawn. For PCBs, we also examined the association based on PCB structural and biological-activity groups as proposed by Wolff et al. (34). Nine PCBs were grouped into three groups: (a) potentially estrogenic and weak phenobarbital inducer (congener 187); (b) potentially antiestrogenic and dioxin-like (congeners 74, 118, 138, 156, and 170); and (c) phenobarbital, CYP1A, and CYP2B inducers (congeners 153, 180, and 183). A linear logistic regression model was used to estimate the exposure and disease association and to adjust for potential confounders. We divided the serum levels of total PCBs and DDE into tertiles or quartiles based on the frequency distribution of the controls. Variables included in the final model were age (,47, 47–52, 53–63, $64 years), BMI (,21, 21–24.9, $25 kg/m), age at menarche (,13, 13–14, $15 years), lifetime months of lactation (0, 1–10, $11), age at first full-term pregnancy (nulliparous, ,20, 20–25, $26 years), number of live births (0, 1–3, $4), lifetime months of hormone replacement therapy (0, 1–71, $72), dietary fat intake in g/day (,46, 46–71, $72, unknown), family breast cancer history (including mother, sisters, and daughters), income 10 years before disease diagnosis or interview (,$20,000, $20,000–24,999, $$25,000, or unknown), and race (white, black, and other). The study site (Tolland County or New Haven County) was also adjusted when the risk was assessed for the entire study population. Maximum likelihood estimates of the parameters were obtained using SAS (35). Tests for trend were conducted using a likelihood ratio statistic in a logistic regression model. Breast cancer risk was also assessed based on menopausal, parity, and lactation status and by study site.