From the 1University of Wisconsin Comprehensive Cancer Center, Madison, WI;
2Fred Hutchinson Cancer Research Center, Seattle, WA;
3Departments of Epidemiology and
6Nutrition, Harvard School of Public Health, Boston, MA;
4Norris Cotton Cancer Center, Dartmouth-Hitchcock Medical Center, Lebanon, NH; and
5Channing Laboratory, Department of Medicine, Harvard Medical School and Brigham and Women’s Hospital, Boston, MA.
Exposure to electromagnetic fields has been hypothesized to increase breast cancer risk, possibly because of reductions in the pineal gland production of melatonin, a hormone that may inhibit mammary carcinogenesis.1–4 Other possible mechanisms include increased proliferation of breast cancer cells (for example, MCF-7 cells), disruption of signal transduction pathways, or inhibition of cell differentiation processes.3,5,6 Electromagnetic exposures may occur at work, through residential proximity to electromagnetic field sources, or within homes.5,7–9
Electric blanket use is among the greatest contributor to electromagnetic exposure from residential appliances.10–12 A few studies have assessed the association between electric blanket use and the incidence of breast cancer.13–18 One case-control study19 found an increased overall risk of 1.45 [95% confidence interval (95% CI) = 1.08–1.94] associated with continuous use throughout the night. In contrast, four other studies did not observe an important association between electric blanket use and breast cancer.15–18 We present data from our large population-based case-control study of breast cancer risk and exposure to electromagnetic fields through the use of electric blankets among women 50–79 years of age.
Subjects and Methods
Identification of Breast Cancer Cases
All female residents of Wisconsin, Massachusetts (excluding metropolitan Boston), and New Hampshire 50–79 years of age with a new diagnosis of breast cancer reported between January 1992 and December 1994 were eligible for this study. Information on cancer site, histology, extent of disease, demographic factors, and follow-up physician was available from each state’s cancer registry. According to an institutionally approved protocol in each state, the physician of record for each potentially eligible case was contacted by mail for permission to approach the subject. Only cases with listed telephone numbers, a valid driver’s license (50–64 years of age, as verified by self-report), and without previous breast cancer were eligible. Of the 6,839 eligible cases, physicians refused contact for 158 (2.3%), 293 (4.3%) were deceased, 83 (1.2%) could not be located, and 620 (9.1%) refused to participate. Thus, data for 5,685 women were available for analysis, for an overall response rate of 83.1%. Response rates varied somewhat between the three states, 78% (N = 1,624) for Massachusetts, 79% (N = 595) for New Hampshire, and 87% (N = 3,466) for Wisconsin. According to cancer registry reports, 98% of these cases had confirmation of invasive breast carcinoma through histologic or cytologic analysis or other means.
Identification of Population Controls
In each state, community controls were randomly selected from two sampling frames. Women 50–64 years of age were selected from lists of licensed drivers, and women 65–79 years of age were selected from a roster of Medicare beneficiaries supplied by the Health Care Financing Administration. Controls were selected at random within age strata to yield an age distribution similar to that of the cases within each state, and had to meet the eligibility criterion of having a listed telephone number. Of the 7,655 potential controls without a personal history of breast cancer, 183 (2.4%) had died, 124 (1.6%) could not be located, 1,397 (18.2%) refused to participate, and 5,951 completed the study interview. Thus, the overall response rate was 78%: 70% (N = 1,783) for Massachusetts, 69% (N = 604) for New Hampshire, and 84% (N = 3,564) for Wisconsin.
Cases and controls were sent letters briefly describing the study before they were contacted by telephone. The 45-minute telephone interview, conducted between July 1992 and July 1995, elicited information on known or suspected risk factors for breast cancer, including reproductive and menstrual history, history of alcohol use, mammogram history, height and weight, medical history, and demographic factors. Exposures before diagnosis (for cases) were ascertained. For controls a comparable reference date was defined that corresponded to the average dates of diagnosis of cases within 5-year age strata (on average, this was 1 year before interview and varied slightly between study states). To maintain blinding, information on the woman’s screening practices as well as on her personal and family history of breast cancer was not obtained until the end of the interview. For 87% of the cases and 96% of the controls, the interviewer remained unaware of the case/control status of the subject until the end of the interview.
Electric blanket or mattress cover use history was included in the questionnaire from June 1994 through July 1995. The women were queried, “Have you ever slept with an electric blanket or mattress cover when it was turned on?” Women who reported a positive history were further asked, “How long did you use it?” Subjects were also asked whether they were using an electric blanket or mattress cover during the reference year. (“Were you using it in [the reference year]?”)
Analysis was limited to women interviewed between June 1994 and July 1995 (1,956 cases and 2,499 controls). We excluded 7 case women with breast cancer and 1 control woman who could not provide complete electric blanket usage information. Thus, 1,949 cases and 2,498 controls remained for this analysis.
To assess the reliability of questionnaire responses, a sequential sample of case and control subjects from Wisconsin and Massachusetts were reinterviewed. After an average of 3.4 months (range = 2–6 months), 181 cases (75% of those who gave permission to be reinterviewed) and 186 controls (83%) were successfully recontacted and reinterviewed. We used Cohen’s kappa with 95% lower confidence limits (LCLs) to measure the reliability of the subjects’ responses regarding never/ever and never/former/current electric blanket use,20 and intraclass correlation coefficients to assess the reproducibility of reported duration of electric blanket use. The kappa for never-use/ever-use was 0.75 (LCL = 0.66) among case women and 0.82 (LCL = 0.74) among control women. The kappa for never-/former/current use was 0.70 (LCL = 0.61) among case women and 0.83 (LCL = 0.75) among control women. The intraclass correlation coefficient for duration of electric blanket use was 0.62 (LCL = 0.54) among case women and 0.77 (LCL = 0.72) among control women.
We used odds ratios and 95% CIs from logistic regression models to estimate relative risks.21 We used conditional models stratified according to age and study site to accommodate the slightly different age distribution of case subjects and controls in each state. Ever-use vs never-use of electric blankets or mattress covers was represented in the logistic regression models as a dichotomous variable. A categorical variable described never/former/current categories of electric blanket users. Among ever-users, we categorized the duration of electric blanket or mattress cover use (in months) as approximate quartiles based on the distribution among controls, in separate models, and as a continuous linear term (months of use). Covariates in the models included known or suspected risk factors for breast cancer, including menopausal status (premenopausal or postmenopausal, defined as no menstruation for
6 months); age at menopause (five categories); age at first full-term pregnancy (five categories); education (four categories); body mass index [quartiles, self-reported weight (kg) divided by height (m2)]; usual alcohol consumption (continuous variable); and family history of breast cancer in mother, daughter, or sister (no/yes).
Results
Compared with controls, women with breast cancer were more likely to have a heavier body mass index, higher educational attainment, family history of breast cancer, later age at first full-term pregnancy (or be nulliparous), and later age of menopause (Table 1). All these associations were in the expected direction. Cases and controls were similar in age (mean age for cases was 65.0 years and mean age for controls was 66.3 years at reference date).
A similar proportion of women with breast cancer reported electric blanket or mattress cover use (43%) as age-matched controls (44%) (Table 2). Compared with women who had never used an electric blanket or mattress cover, the relative risk of breast cancer in women who had ever used an electric blanket or mattress cover was 0.93 (95% CI = 0.82–1.06). The relative risk for former users was 1.01 (95% CI = 0.87–1.17), whereas the relative risk of breast cancer for current users was 0.79 (95% CI = 0.66–0.95). The nonparametric regression curve indicates that there is no meaningful association between duration of electric blanket or mattress cover use and breast cancer risk (Figure 1).
Discussion
In this large, population-based case-control study of breast cancer, ever-use of electric blankets or mattress covers was not associated with any increase in breast cancer risk. Our confidence in a lack of positive association for ever-use is strengthened by the large study size and the control of many potentially confounding factors. Breast cancer risk was lower among current users, although we found no association according to duration of use.
Of the six previous studies of the relation between electric blanket use and breast cancer risk, most have not demonstrated an association.13–18 These studies have examined various characteristics of exposure, including duration of use13–18 and use throughout the night.13–15,17 Because electromagnetic field exposure is posited to increase hormone levels, hormonal characteristics including hormone receptor status22 may modify the association with breast cancer. Stratification by receptor status has yielded similar null odds ratios between electric blanket use and breast cancer, however.15,17,18 In addition, no interaction has been observed with menopausal status.13–15,17,18 Although some of these studies were limited by small size13,14,16,17 and low response rates,13,14 in the large Nurses’ Health Study cohort the relative risk was 1.08 (95% CI = 0.95- 1.24).18 Studies that investigated residential geographic proximity to electromagnetic field sources have generally not shown an association with female breast cancer risk.23–25 Studies of occupational electromagnetic field exposure and risk of breast cancer in men and women have reported inconsistent results, with some suggesting a positive association26–34 and others showing no association.35–43
Associations between breast cancer and electric blanket use may be confounded by other exposures. Lawrence44 has suggested that women who use electric blankets might use them for socioeconomic or comfort reasons and that these women might differ from other women. For example, some women might use an electric blanket to help save money on home heating costs. Similarly, lean women might choose to use electric blankets more than heavier women to stay warm at night. In our study we were able to control for educational attainment (a surrogate of socioeconomic status), body mass index, and other risk factors, although such adjustment did not meaningfully alter the relative risk estimates.
With any case-control study, selective recall of past exposures is of some concern. Recall of current electric blanket use is probably more accurate than recall of former use, or of duration of previous use.45 In our study population, the reliability of reported electric blanket or mattress cover use was reassuring for both ever-use/never-use and currency of use, although concordance in reporting between the initial and second interviews was slightly better for controls than cases.
Accurate and relevant assessment of electromagnetic field exposure is difficult in a telephone interview type of survey. The exposure indicators we used are only imprecise proxies for electromagnetic exposure.4,9,46 Information on patterns of use (continuous use throughout the night or bed-warming use only, seasonal usage patterns, etc) and the products (purchase date, type of appliance, etc) was not obtained. For example, electric blanket wiring configuration changed in the late 1980s, which significantly reduced electromagnetic field exposure potential. In addition, we did not obtain information regarding exposure to lights at night or electromagnetic field exposures proximal to the home and at work,47 or regarding other electrical home appliances or wiring that may influence total exposure.10 Exposure misclassification and weakened estimates of effect might be produced in analysis of ever-use or a category specifying use of long duration; however, examination of current use should not result in such misclassification.
It has been suggested that electromagnetic fields might affect melatonin homeostasis and thus positively affect hormone homeostasis and cancer.48 Putative melatonin receptors have been cloned,48 and the membrane receptors (Mel1a and Mel1b), found primarily in the brain, may indirectly influence hormone-dependent carcinogenesis, such as breast cancer. Recent reviews have noted the difficulty in demonstrating an association between exposure to electromagnetic fields in the range of 50–60 Hz, that is, common household exposures, and depressed melatonin levels in humans.5,49 Because exposures in clinical studies were short term and few electromagnetic parameters were assessed, these studies probably do not simulate the kinds of exposure that women experience in and around their home and work environments.49–52 Some have suggested more direct mechanisms by which electric blanket use may affect breast cancer,13,15 such as an increased cell proliferation, disruption of signal transduction pathways, or inhibition of differentiation.5,6 Nevertheless, limited epidemiologic or clinical evidence exist to support these possible mechanisms.52,53
To our knowledge an inverse association, such as we observed, has not been hypothesized for electromagnetic field exposure. It is unlikely that using electric blankets or mattress covers reduces the risk of breast cancer, since there is no hypothesized mechanism of action, no dose-response relation, and insufficient latency between exposure and the development of breast cancer. These data may represent confounding that was not measured in the study. In conclusion, our findings, taken together with previous studies, do not support a positive association between electric blanket or mattress cover use and breast cancer risk among women 50–79 years of age.
We thank Barry E. Storer, Henry Anderson, and Han Kim for assistance at various stages of this project; the staff of the Wisconsin Cancer Reporting System; the Massachusetts Tumor Registry; the staff of the New Hampshire Cancer Registry; and the study managers, interviewers, and programmers.