Electromagnetic Fields and Breast Cancer on Long Island: A Case-Control Study

Transcription

1 American Journal of Epidemiology Copyright 2003 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 158, No. 1 Printed in U.S.A. DOI: /aje/kwg106 Electromagnetic Fields and Breast Cancer on Long Island: A Case-Control Study E. R. Schoenfeld 1, E. S. O Leary 1, K. Henderson 1, R. Grimson 1, G. C. Kabat 1, S. Ahnn 1, W. T. Kaune 2, M. D. Gammon 3, and M. C. Leske 1 for the EBCLIS Group 1 Department of Preventive Medicine, School of Medicine, Stony Brook University, Stony Brook, NY. 2 EM Factors, Richland, WA. 3 Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC. Received for publication December 20, 2002; accepted for publication April 9, The EMF and Breast Cancer on Long Island Study (EBCLIS) was a case-control study designed to evaluate the possible association between exposure to electromagnetic fields (EMFs) and breast cancer. Eligible women were participants in the population-based Long Island Breast Cancer Study Project, were under 75 years of age at enrollment, were residentially stable, and were identified between August 1, 1996, and June 20, Of those eligible, 576 cases and 585 controls participated in EBCLIS (87% and 83%, respectively). In-home data collection included various spot and 24-hour EMF measurements, ground-current magnetic field measurements, wire mapping of overhead power lines servicing the home, and an interview. Odds ratios and 95% confidence intervals were based on multivariate logistic regression analyses. All odds ratios were close to 1 and nonsignificant. For the highest quartile of 24-hour EMF measurements, the odds ratio was 0.97 (95% confidence interval (CI): 0.69, 1.37) in the bedroom and 1.09 (95% CI: 0.78, 1.51) in the most lived-in room. For the highest exposure category of ground-current measurements, the odds ratio was 1.13 (95% CI: 0.88, 1.44) in the bedroom and 1.08 (95% CI: 0.85, 1.38) in the most lived-in room. These and other EBCLIS results agree with other recent reports of no association between breast cancer and residential EMF exposures. breast neoplasms; carcinogens; electromagnetic fields; environmental exposure Abbreviations: EBCLIS, EMF and Breast Cancer on Long Island Study; EMF(s), electromagnetic field(s); LIBCSP, Long Island Breast Cancer Study Project. High breast cancer rates have persisted in the northeastern United States for over four decades (1, 2). Rates of female breast cancer on Long Island are among the highest in New York State, with an average annual incidence rate of per 100,000 women in Nassau County and per 100,000 in Suffolk County, as compared with a statewide average of per 100,000 between 1994 and 1998 (3, 4). While these high rates could at least partially be explained by an increased frequency of known risk factors among Long Island women (5 10), there is widespread community concern that these high rates are also due to undetermined environmental factors related to residence on Long Island. As a result of efforts by local breast cancer coalitions, a Congressional mandate was issued for a comprehensive study of environmental factors and breast cancer on Long Island, one additional New York State county, and one Connecticut county (11). Exposures of concern included contaminated drinking water; sources of indoor and ambient air pollution; electromagnetic fields (EMFs); pesticides and other toxic chemicals; and hazardous and municipal waste. The Long Island Breast Cancer Study Project (LIBCSP) originated from this mandate (12) and included several multidisciplinary studies by multiple investigators. The largest of these studies was the LIBCSP, a population-based case-control study designed to determine whether organochlorine pesticides and polycyclic aromatic hydrocarbons were associated with breast cancer among Long Island women (13). The potential role of EMFs was evaluated by means of a companion case-control study, the EMF and Breast Cancer on Long Island Study (EBCLIS). The EBCLIS is the subject of this report. Reprint requests to Dr. Elinor Schoenfeld, Department of Preventive Medicine, School of Medicine, Stony Brook University, Stony Brook, NY ( eschoenfeld@notes.cc.sunysb.edu). 47

2 48 Schoenfeld et al. The possible relation between EMF exposure and risk of breast cancer in men (14 20) and women (10, 21 52) has received considerable attention. One hypothesis regarding the association is based on experimental evidence that light and extremely low frequency EMFs affect melatonin production by the pineal gland, thus influencing mammary carcinogenesis; this has been observed in laboratory studies (53). The epidemiologic literature, as summarized previously, has yielded inconclusive results (54). An underlying limitation relates to difficulties in quantifying EMF exposure, as studies have often relied on proxy measures such as occupational categories, wire codes, and distance from power lines rather than direct measurements of current fields as markers of past exposure. To overcome these problems, EBCLIS investigators performed comprehensive in-home assessments of magnetic field exposure, including spot, 24- hour, and ground-current magnetic field measurements, information on wire coding, and self-reported information. The study s focus on in-home measurements was based on findings from a pilot study which determined that a large proportion of the participants incurred more than 75 percent of their cumulative EMF exposures at home (55). As an additional feature, the study population was limited to residentially stable women who had lived in the same Long Island home for at least 15 years. In this paper, we describe the study protocol and present results pertaining to the residential EMF measurements. MATERIALS AND METHODS Selection of cases and controls Women eligible for EBCLIS were a subset of participants in the LIBCSP case-control study who had lived in their current residence for at least 15 years and were under 75 years of age at the time of study recruitment (13). The residential criterion was selected to allow for EMF measurements that might best reflect long-term residential exposure. The upper age limit for eligibility was used because of the low number of persons aged 75 years recruited in the LIBCSP (13). The target sample size for EBCLIS was set at 600 cases and 600 controls so the study would have 90 percent power to detect odds ratios of 1.6 or greater with exposure prevalences of 20 percent or greater. To meet these sample-size goals, the investigators based recruitment on eligible case women (women with in-situ or invasive breast cancer) identified for LIBCSP between August 1, 1996, and June 20, 1997, and a random sample of controls identified for LIBCSP through random digit dialing (for women aged <65 years) or from Health Care Financing Administration files (for women aged 65 years). EBCLIS controls were selected from this pool through frequency matching to the expected age distribution of EBCLIS cases. A previous article (13) described the design and methods of the LIBCSP (for women under age 75 years, independent of length of residence, there were 1,354 cases (86 percent participation) and 1,426 controls (69 percent participation)); it also discussed the impact of using random digit dialing on LIBCSP control selection. The protocol for EBCLIS recruitment was as follows. At the time of the LIBCSP in-home visit, participants received a brochure describing EBCLIS. A field supervisor identified 1,431 potentially eligible women from the LIBCSP (698 cases and 733 controls). EBCLIS staff determined that 1,365 of these women (663 cases and 702 controls) were eligible for EBCLIS and contacted them to invite participation. When participation was agreed upon, an EBCLIS in-home visit was scheduled as soon as possible. Data collection The EBCLIS data collection protocol had several components: interviews, measurement of magnetic fields and ground-current magnetic fields, and collection of data on the wire coding of overhead power lines. Interview data. Interview data were obtained in two parts. A set of questions regarding residential history and EMF exposures was administered to all LIBCSP study participants (n = 3,064) at the time of their LIBCSP comprehensive in-home visit. During the subsequent EBCLIS in-home visit, more detailed information on the participant s residential and housing construction histories (e.g., wiring, structural renovations) and EMF exposure was collected during a halfhour interview. Data gathered included self-reported information (e.g., questions on occupational exposures; questions on shift work or night work and other questions assessing exposure to light at night; additional residential questions) and more detailed information on the use of electrical appliances at home and at work. We plan to present separately the results of evaluations of EBCLIS data with regard to electric blanket use (56) and other EMF exposures. EMF residential measurements. EMF residential measurements, at the time of the EBCLIS interview, included spot measurements in three locations (the front door, the bedroom, and the most lived-in room (the room where the participant spent the most time other than the bedroom or kitchen)) (55) and 24-hour measurements in two locations (the bedroom and the most lived-in room) obtained with EMDEX II meters (Enertech Consultants, Campbell, California), as well as ground-current magnetic field measurements. Selection of these rooms was based on the pilot study, completed prior to the start of EBCLIS, which determined that exposures in those locations contributed the most to the total EMF exposure of an individual (55). The meters were programmed to record broadband ( Hz) and harmonic ( Hz) magnetic fields at 3-second intervals for the spot measurements and 15-second intervals for the 24-hour measurements. Typically, since the harmonic fields are considerably smaller than broadband fields, the broadband fields provide a good approximation of 60-Hz magnetic fields. The protocol is described below. Spot measurements. Front-door measurements were taken outside the participant s front door, within 1 m of the entrance to the home. In-home spot measurements were taken in the center of the bedroom and the center of the most lived-in room. Ground-current magnetic field measurements. A groundcurrent test load located at least 6 feet (1.8 m) away from the interviewer was used while spot measurements were

3 EMFs and Breast Cancer on Long Island 49 FIGURE 1. Participant recruitment for the EMF and Breast Cancer on Long Island Study (EBCLIS), Participants were defined as women who completed the in-home interview and had at least one electromagnetic field measurement. taken in the bedroom and the most lived-in room and three known levels of current (0 A, 2.2 A, and 4.3 A) were introduced into the home s wiring system. The measured changes in magnetic-field strength between these currents were used to calculate broadband and harmonic test-load coefficients, defined as the magnetic field produced by 1 A of test-load current. Details on this procedure have been published elsewhere (57). 24-hour measurements. The meter for the 24-hour measurement in the bedroom was placed under the bed most closely aligned with the location where the participant slept. The meter for the 24-hour measurement in the most lived-in room was placed as close as possible to the location where the participant spent most of her time in that room. These measurements were included to detect variations between daytime and nighttime values, as well as to obtain data for all women over the same time interval. These data allowed for case-control comparisons of exposures incurred during specific time periods (e.g., from 10:00 p.m. to 8:00 a.m.) and during specific activities (e.g., sleeping), thus permitting evaluation of sleep-time habits and the EMF patterns of shift workers. The meters were programmed to acquire magnetic field data until their memories were filled (~30 hours), and the meters were either retrieved by the interviewer or sent via express mail to the interviewer, who then downloaded the information. Data on the 24-hour measurements were extracted from these data files for analysis. Wiring maps. Wiring maps were obtained by trained personnel for all EBCLIS participants and nonparticipants. These data were used to classify residences according to both the five-category Wertheimer-Leeper method (modified to accommodate underground wiring and the types of wiring configurations on Long Island (55, 58)) and its threecategory version, the Kaune-Savitz method (30, 59, 60). Wire mapping of the residences of EBCLIS nonparticipants was used to evaluate potential differences in EMF exposure between participants and nonparticipants. Wire mapping was assigned in batches in order to mask the mappers as to the case/control status of participants. Quality control The study protocol included a number of steps designed to ensure quality control of all data collection procedures. The EMDEX II measurements and ground-current magnetic field data were evaluated using range and logic checks, as well as evaluation of timing and quantity of measurements. Logic checks and cross-checks were performed for all interview data acquired in the study, and a 10 percent random sample of entered data items from interviews, wire code maps, and measurements were verified against the original hard copy.

4 50 Schoenfeld et al. TABLE 1. Selected characteristics of cases and controls in the EMF and Breast Cancer on Long Island Study, Cases Controls Variable p value No. % No. % Age (years) NS* Race White Black Other NS Latina ethnicity Latina/Hispanic NS Missing data 1 1 Religion Catholic Protestant Jewish None Other NS Cigarette smoking Never smoker Current smoker (within last 12 months) Past/former smoker NS Education Less than high school graduation High school graduation Some college College graduation Postcollege education Missing data 1 Annual income <$15, $15,000 $19, $20,000 $24, $25,000 $34, $35,000 $49, $50,000 $69, $70,000 $89, $90, NS Missing data Alcohol use Ever drinking NS Table continues

5 EMFs and Breast Cancer on Long Island 51 TABLE 1. Continued Variable Cases Controls No. % No. % p value Age (years) at menarche < NS Missing data 5 3 Age (years) at first birth (for parous women) < NS Parity Parous NS No. of livebirths > Derived menopausal status Postmenopausal NS Missing data Benign breast disease Positive history Missing data 2 0 Hormone replacement therapy Ever use NS Oral contraceptives Ever use NS Missing data 2 1 Family history of breast cancer (mother, sister, or daughter) Positive history Missing data Fertility problems Yes Missing data 5 3 Lactation Ever lactated NS Quartile of body mass index at reference date < NS Missing data 3 5 * NS, not significant (p > 0.05). Derived according to the method of Gammon et al. (13). Weight (kg)/height (m) 2.

6 52 Schoenfeld et al. Repeat measures for wire coding were obtained by having each of the four mappers replicate the wire mapping of at least 45 of the other mappers homes. A total of 235 homes were thus remapped, with discrepancies being resolved by a third mapper. A complete description of the wire mapping protocol can be found in an article by O Leary et al. (58). To control for possible seasonal fluctuations in power usage and thus in measured EMF levels, we enrolled cases and controls and completed their in-home visits at the same rate within each season throughout the study. We closely monitored interview dates for cases and controls to evaluate our achievement of this goal. The concern was that magnetic field levels measured in many Long Island homes would exhibit a seasonal dependence, with the largest fields being measured during hot days in the summer, when demand for electrical power would be highest. In addition, study personnel recorded the daily maximum and minimum temperatures during the course of the data collection period. These temperatures were compared for cases and controls. No significant temperature differences were found, indicating satisfactory control of season-related differences in measurements. Statistical methods Summary statistics for all continuous variables, for EMF broadband, harmonic spot, and 24-hour measurements, and for test-load coefficients (in µt) were calculated; cutpoints for the distributions (quartiles) were based on measurements obtained for controls. The continuous variables included a computed estimate of personal exposure derived from the pilot study. In that study, estimated personal exposure predicted actual personal exposures, as determined from the arithmetic means of measurements taken by a personal EMF exposure meter (EMDEX Lite; Enertech) worn by each individual (55). Estimated personal exposure is derived from a regression equation that includes the means of the broadband 24-hour measurements in µt (in the bedroom and the most lived-in room) and the ground-current test-load coefficient in the most lived-in room (µt/a) (55). The Mann-Whitney U statistic was used to compare mean values between cases and controls for the above exposure variables, since the EMF measurement data were not normally distributed. Spearman correlation coefficients were calculated for the mean 24- hour broadband and harmonic measurements taken in the bedroom and the most lived-in room and for the composite estimated personal exposure measure. For all analyses, the referent categories were defined as those with the lowest in-home EMF exposures. Referent categories were defined as values below the lowest quartile for the spot and 24-hour exposure variables and the composite estimated personal exposure measure; 0.01 µt/ A for the ground-current magnetic field measurements; the combined grouping of underground and very low current configuration for the Wertheimer-Leeper wire code measurements (59); and low for the Kaune-Savitz wire code measurements (60). Data for the ground-current magnetic field measurements taken in the bedroom and the most lived-in room were also dichotomized as 0.01 µt/a (referent group) and >0.01 µt/a, with the latter category representing the presence of ground-current magnetic fields. In univariate analyses, associations between breast cancer status and other variables were tested using the chi-squared and Fisher s exact tests. Odds ratios and 95 percent confidence intervals were estimated for the risk of breast cancer associated with exposure to EMF, as estimated by the inhome measurements. Independent variables for multivariate analyses were selected from variables associated with breast cancer (p < 0.25) in the univariate analysis. Logistic regression analyses were used to identify significant predictors of breast cancer and to estimate adjusted odds ratios and their 95 percent confidence intervals. Logistic regression analyses (61, 62) identified the following variables as significant predictors of breast cancer in EBCLIS: age, parity, family history of breast cancer, education, and history of benign breast disease. These variables were included in all multivariate analyses. Body mass index (weight (kg)/height (m) 2 ) at the reference date (for both premenopausal and postmenopausal women) and body mass index at age 20 years (for postmenopausal women) were also found to be significant predictors and were included in analyses of the menopause subgroups. RESULTS Descriptive characteristics Figure 1 summarizes the overall recruitment results for EBCLIS. Of those eligible, 1,161 women completed EBCLIS (576 cases and 585 controls; participation rates of 87 percent and 83 percent, respectively). Comparisons of demographic data showed that participants were younger, had a higher income and educational level, and were more likely to be Caucasian than nonparticipants. There were no differences between participants and nonparticipants for categories of Wertheimer-Leeper wire codes; however, for Kaune-Savitz wire codes, participants had a higher percentage of low category homes and nonparticipants had a higher percentage of medium category homes. Percentages of homes in the highest exposure categories, that is, categories where a disease-exposure association would be most likely to be observed, were similar for participants and refusers, both with the Wertheimer-Leeper method ( very high current configuration : 7.9 percent and 7.5 percent, respectively) and with the Kaune-Savitz method ( high : 10.6 percent and 8.9 percent, respectively). The distribution of values in these wire code categories among nonparticipants did not differ between cases and controls. Since we had no data on the length of residence of LIBCSP nonparticipants, their eligibility status for EBCLIS could not be determined. Table 1 presents select demographic characteristics of study participants by case/control status. Overall, participants ranged in age from 20 years to 74 years (mean = 58.8 (standard deviation, 8.9); median, 58.2); 96 percent were Caucasian. With regard to religion, 57 percent were Catholic, 23 percent were Protestant, 19 percent were Jewish, and 1 percent were of other religions. Seven percent had less than a high school education. Twenty-five percent of participants reported an annual household income below $35,000,

7 EMFs and Breast Cancer on Long Island 53 TABLE 2. Summary statistics for residential electromagnetic field measurements and a composite estimated personal exposure measure among women aged <75 years, EMF and Breast Cancer on Long Island Study, * Measurement Mean 24-hour measurements (µt) Broadband * All p values from Mann-Whitney U testing were greater than 0.05 (not significant). SD, standard deviation. Includes the mean 24-hour broadband measurements taken in the bedroom and the most lived-in room and the test-load coefficient for the most lived-in room (55). although 13 percent refused to respond to this question. These demographic figures reflect the distribution by ethnicity, religion, and income observed for residents of Long Island in US Census data (data not shown). Eight percent of participants were nulliparous, 74 percent were postmenopausal, 17 percent reported a history of benign breast disease, and 16 percent reported a family history of breast cancer. There were significant differences between cases and controls for level of education (p = 0.03), number of livebirths (p = 0.05), history of benign breast disease (p = 0.001), family history of breast cancer (p = 0.001), and fertility problems (p = 0.004). Magnetic field measurements No. of participants Table 2 presents the distribution of in-home EMF measurements (broadband, harmonics, spot, 24-hour, and Cases Controls Mean SD Median Mean SD Median Bedroom 1, Most lived-in room 1, Harmonics Bedroom 1, Most lived-in room 1, Spot measurements (µt) Broadband Bedroom 1, Most lived-in room 1, Front door 1, Harmonics Bedroom 1, Most lived-in room 1, Front door 1, Ground-current test-load coefficients (µt/a) Broadband Bedroom 1, Most lived-in room 1, Harmonics Bedroom 1, < <0.001 Most lived-in room 1, < <0.001 Composite estimated personal exposure measure (µt) 1, ground-current magnetic field test-load coefficients) for the front door, the bedroom, and the most lived-in room, as well as the composite estimated personal exposure measure. Cases and controls had very similar mean values and almost identical median values for all measurements; no significant differences between the two groups were observed. All p values, except those for the harmonic spot measurements in the most lived-in room (p = 0.21) and the broadband spot measurements in the most lived-in room (p = 0.24), were greater than 0.5. Table 3 presents unadjusted and adjusted odds ratios for strata based on quartiles of the distributions of spot and 24- hour measurements, ground-current magnetic field test-load coefficients, sleep time, and the composite estimated personal exposure measure. Odds ratios were adjusted for age, parity, family history of breast cancer, education, and history of benign breast disease. No differences between

8 54 Schoenfeld et al. TABLE 3. Unadjusted and adjusted odds ratios for breast cancer according to quartiles of residential electromagnetic field measurements (mean spot and 24-hour broadband measurements), ground-current magnetic field test-load coefficients for the bedroom and the most lived-in room, an estimated personal exposure measure, and sleep time, EMF and Breast Cancer on Long Island Study, Cases Controls Unadjusted odds ratio Adjusted odds ratio* Measurement and magnetic field strata No. % No. % OR 95% CI OR 95% CI Spot measurements (µt) Bedroom < , , , , , , 1.60 Most lived-in room < , , , , , , 1.60 Front door < , , , , , , hour measurements (µt) Bedroom < , , , , , , 1.37 Most lived-in room < , , , , , , 1.51 Ground-current test-load coefficients (µt/a) Bedroom > , , 1.44 Most lived-in room > , , 1.38 Estimated personal exposure measure (µt) < , , , , , , 1.51 Sleep time (µt) < , , , , , , 1.39 * Adjusted for age, parity (number of livebirths), family history of breast cancer (defined as breast cancer in a mother, sister, or daughter), education (defined as high school graduation or less (referent), some college, college graduation, or postcollege education), and history of benign breast disease. OR, odds ratio; CI, confidence interval. Includes the mean 24-hour broadband measurements taken in the bedroom and the most lived-in room and the test-load coefficient for the most lived-in room (55). Measurements taken in the bedroom during the participant s usual sleeping time, independent of time of day.

9 EMFs and Breast Cancer on Long Island 55 TABLE 4. Unadjusted and adjusted odds ratios for breast cancer according to residential electromagnetic field exposure (Wertheimer-Leeper and Kaune-Savitz wire code classifications) among women aged <75 years with complete data (n = 1,161), EMF and Breast Cancer on Long Island Study, Cases Controls Unadjusted odds ratio Adjusted odds ratio* Wire code No. % No. % OR 95% CI OR 95% CI Wertheimer-Leeper classification Underground/very low current configuration Ordinary low current configuration , , 1.17 Ordinary high current configuration , , 1.11 Very high current configuration , , 1.48 Kaune-Savitz classification Low Medium , , 1.13 High , , 1.61 * Adjusted for age, parity (number of livebirths), family history of breast cancer (defined as breast cancer in a mother, sister, or daughter), education (defined as high school graduation or less (referent), some college, college graduation, or postcollege education), and history of benign breast disease. Data on covariates were missing for 29 women. OR, odds ratio; CI, confidence interval. Referent category. cases and controls were found; all of the 95 percent confidence intervals contained 1. All odds ratios were less than 1.30, with no evident dose-response relation with increasing levels of exposure. Additional analyses were performed in which persons at or above the 90th percentile of exposure were compared with those in the lowest exposure category; these analyses yielded similar results. Wire codes Table 4 presents the wire coding data for cases and controls using both the Wertheimer-Leeper classification scheme (30, 59) and the Kaune-Savitz classification scheme (60). No significant differences were observed for any wire coding level by case/control status, and no dose-response relation was observed. Similar results were obtained using the Kaune-Savitz scheme. All of the analyses were also conducted with data stratified according to menopausal status, county of residence, stage of disease (in-situ vs. invasive), and age. These separate analyses yielded similar results, with no indication of effect modification. Other possible covariates (e.g., age, seasonality of recruitment) were included in both the crude and adjusted analyses, with similar results being found in these analyses as well. DISCUSSION Overview This study was designed to evaluate the role of EMF exposure as a potential environmental risk factor for breast cancer, using a comprehensive set of measurements to directly assess exposures. The hypothesis under evaluation originated from basic research and from previous clinical and epidemiologic studies, which provided a basis for further exploration in EBCLIS (14 55). The results do not support the hypothesized association between residential EMF exposure and breast cancer, whether assessed by inhome measurements of magnetic fields, by ground-current magnetic fields, or by wire codes. Cases and controls had very similar spot and 24-hour broadband or harmonic measurements in the bedroom and the most lived-in room, as well as similar composite measures of exposure, with no trends or significant differences being detected. Analyses in which data were stratified by menopausal status, county of residence, and stage of disease (in-situ vs. invasive) did not alter the pattern. Furthermore, no additional covariates or confounders were identified from LIBCSP analyses that needed to be included in the EBCLIS analyses (13, 63, 64). Strengths and weaknesses When interpreting EBCLIS results, one should critically evaluate the limitations inherent in the case-control design and the overall study methodology. The study used the best available measures of exposure by incorporating several methods, including ground-current magnetic field measurements, thus being the first epidemiologic study to use this measure. By limiting the study to long-term residents and collecting information about housing construction over time, seasonal effects, and other factors that might have affected exposure, EBCLIS investigators have attempted to be as comprehensive as possible in assessing long-term EMF exposure. The study protocol incorporated many quality control procedures, as well as masking of interviewers, to minimize biases. Interviews and all data collection protocols followed a standardized, structured format that was used for both cases and controls. EBCLIS had a large sample size and good participation rates for cases and controls (87 percent and 83 percent,

10 56 Schoenfeld et al. respectively). Since EBCLIS participants were required to have participated in the LIBCSP first, the EBCLIS participation rates were dependent on participation in the LIBCSP. As a result, the good LIBCSP participation rates for case women under 75 years of age (86 percent) and the relatively lower participation rates reported for controls under 75 years of age (69 percent) carried over to EBCLIS (13). Given the recruitment methods used in the LIBCSP, it was not possible to determine how many persons who refused to participate in that study were potentially eligible for EBCLIS. Thus, it is difficult to determine the true impact of the lower control participation rates on the EBCLIS results, especially among older women, who would also have been more likely to have lived in their homes for 15 years or longer. Nonparticipation in the study could have led to bias if the frequency of exposure in the nonparticipating cases differed from that in the nonparticipating controls. However, both studies (LIBCSP and EBCLIS) were presented to potential subjects in the same way, misperceptions were clarified, and reasons for nonparticipation were collected and evaluated. For evaluation of nonparticipation in EBCLIS, the study protocol included wire mapping of all EBCLIS nonparticipants so a single surrogate measure of their in-home exposures could be obtained. There were no regional differences in participation, nor were the wire codes significantly different for participants versus nonparticipants using the Wertheimer-Leeper method. Using the Kaune-Savitz method, there were more participants who had low category homes than nonparticipants who had more medium exposure homes. To some extent, this difference may reflect age differences in participation, with fewer older women agreeing to participate in EBCLIS. Inadequate control for confounding factors could have resulted in biased estimation of risk. Cases and controls in EBCLIS were comparable with regard to age, and in all multivariate analyses we adjusted for age and other variables by systematically testing the effects of demographic, socioeconomic, and regional factors and known breast cancer risk factors. Given the large sample size, the extensive quality control procedures, and the extensive analytical control for potential confounders, the lack of association found is unlikely to have been the result of errors. Comparison with previous studies A number of studies have evaluated the relation of residential EMF exposures to breast cancer, but the results have been inconsistent, as has been reviewed in detail previously (54) and assessed through meta-analysis (65). Davis et al. (43) were the first to report on the results of comprehensive in-home measurements of EMFs as a risk factor for breast cancer. EBCLIS is only the second such study, and to our knowledge it is the first to have been conducted in the eastern United States. A third, similar study is under way in California, with results pending (66). Our findings are consistent with the population-based casecontrol study conducted by Davis et al. (43) in Seattle, Washington, which pioneered this method of measuring residential EMF exposure directly. Although the studies differed in their methods of case ascertainment and data collection, both studies reached the same conclusion: that there was no association between breast cancer and in-home EMF measurements and wire coding. Whereas the Seattle study did not include length of residence in the current home in its eligibility criteria, EBCLIS was based only on residentially stable women, thus restricting the study population to those with EMF measurements that might better reflect long-term exposure. Since the EMF measurements of short-term residents may not reflect their true exposure status, cumulative assessment of exposure requires multiple measurements at previous residences. To address this issue, Davis et al. obtained wire maps of all homes participants had lived in during the 10 years prior to study participation. Such information was available for 72 percent of participants (43). As the authors recognized, some uncertainty remained, since those measurements might not have reflected the actual past exposures of participants while they were residing in those homes. Thus, the long-term residency requirement for EBCLIS was an advantage for exposure assessment. Selection of long-term residents in EBCLIS might account for the larger proportion of older participants than was found in the Seattle study (45 percent vs. 40 percent, respectively, were over age 60 years), as well as the greater number of postmenopausal women in EBCLIS (75 percent vs. 66 percent). The odds ratios reported in both studies were of similar magnitude, regardless of whether bedroom exposure in EBCLIS was defined using a 24-hour measurement, a spot measurement, the 11:00 p.m. 7:00 a.m. period, or actual sleep time. Davis et al. (43) raised concerns that the lack of association between EMF exposure and breast cancer in their study might have been related to uncertainties in the timing of exposure and thus inadequacies in exposure assessment at the most etiologically relevant time. Since EBCLIS enrolled more residentially stable participants than the Seattle study did, the issue of exposure misclassification may have been minimized in EBCLIS if current measurements truly reflected measurements from the distant past. However, use of this proxy for long-term EMF exposure still might not be sufficiently powerful to measure the impact of residential EMF exposure on breast cancer risk. Researchers may need to explore newer ideas for measuring exposure (66) or consider EMFs as a surrogate for another exposure, such as the proposed association between light at night and melatonin production (67, 68). Conclusion In EBCLIS, we found no association between EMF levels (wire coding or measurements) and breast cancer risk. This study thus provides no empirical evidence suggesting that residential EMF exposures contribute to the risk of breast cancer on Long Island. ACKNOWLEDGMENTS This work was supported by grants CA/ES and CA/ ES from the National Cancer Institute and the National Institute of Environmental Health Sciences.

11 EMFs and Breast Cancer on Long Island 57 The EMF and Breast Cancer on Long Island Study (EBCLIS) Group Stony Brook, New York: Dr. M. Cristina Leske (Principal Investigator), Dr. Sang Ahnn, Judith M. Greene, Dr. Roger Grimson, Kevin Henderson, Dr. Geoffrey C. Kabat, Dr. Erin S. O Leary, and Dr. Elinor Randi Schoenfeld; Westat, Rockville, Maryland: Carol Haines and Jacqueline Slattery-Telonidis; EM Factors, Richland, Washington: Dr. William T. Kaune; Long Island Breast Cancer Study Project: Dr. Marilie D. Gammon (Principal Investigator), Dr. Julie A. Britton, Dr. Alfred I. Neugut, and Dr. Susan L. Teitelbaum. Advisory committee: Drs. David A. Savitz (chair), Louise Brinton, Richard Stevens, and Sholom Wacholder; Fred Dietrich; and Norma Logan. Local advisory committee: Mary Dowden, Miriam Goodman, and Mary Joan Shea. REFERENCES 1. Devesa SS, Grauman DG, Blot WJ, et al. Atlas of cancer mortality in the United States, Washington, DC: US GPO, (NIH publication no. (NIH) ). 2. Kulldorff M, Feuer EJ, Miller BA, et al. Breast cancer clusters in the northeast United States: a geographic analysis. Am J Epidemiol 1997;146: New York State Department of Health. New York State Cancer Registry data. Vol 3. Trends in cancer incidence and mortality in county, New York State, Nassau, and Suffolk counties. Albany, NY: New York State Department of Health, New York State Department of Health. New York State Cancer Registry data. Vol 1. Cancer incidence and mortality by county, New York State, Nassau, and Suffolk counties. Albany, NY: New York State Department of Health, (World Wide Web URL: cancer/1998/pdf/longisl.pdf). 5. New York State Department of Health, Bureau of Cancer Epidemiology. Long Island Breast Cancer Study Consortium. Long Island Breast Cancer Study report #1: overview and descriptive variables: breast cancer screening practices. Albany, NY: New York State Department of Health, State University of New York at Stony Brook, School of Medicine, Department of Preventive Medicine. Long Island Breast Cancer Study Consortium. Long Island Breast Cancer Study report #2: risk factors, regional distribution, pathology appraisal, evaluation of selection bias, and water sources and landfills. Stony Brook, NY: State University of New York at Stony Brook, New York State Department of Health, Bureau of Cancer Epidemiology and Division of Environmental Epidemiology and Occupational Health. Small area analysis of breast cancer incidence rates in Nassau and Suffolk counties, New York, Albany, NY: New York State Department of Health, New York State Department of Health, Bureau of Cancer Epidemiology. Long Island Breast Cancer Study Consortium. Long Island Breast Cancer Study report #3: the relationship of lifestyle factors to breast cancer risk. Albany, NY: New York State Department of Health, New York State Department of Health, Bureau of Cancer Epidemiology. Long Island Breast Cancer Study Consortium. Long Island Breast Cancer Study report #4: termiticide use and breast cancer risk. Albany, NY: New York State Department of Health, New York State Department of Health, Bureau of Environmental and Occupational Epidemiology. Location of 138 kv electric transmission lines in Nassau and Suffolk counties, New York, and breast cancer incidence, Albany, NY: New York State Department of Health, National Cancer Institute, Cancer Control and Population Sciences. Study of elevated breast cancer rates in Long Island. Public Law , June 10, Sec Potential environmental and other risks contributing to incidence of breast cancer. Bethesda, MD: National Cancer Institute. (World Wide Web URL: (Accessed December 2002). 12. National Cancer Institute, Cancer Control and Population Sciences. The Long Island Breast Cancer Study Project. Bethesda, MD: National Cancer Institute. (World Wide Web URL: epi.grants.cancer.gov/libcsp/). (Accessed December 2002). 13. Gammon MD, Neugut AI, Santella RM, et al. The Long Island Breast Cancer Study Project: description of a multi-institutional collaboration to identify environmental risk factors for breast cancer. Breast Cancer Res Treat 2002;74: Tynes T, Andersen A. Electromagnetic field exposure and male breast cancer. (Letter). Lancet 1990;336: Demers PA, Thomas DB, Rosenblatt KA, et al. Occupational exposure to electromagnetic fields and breast cancer in men. Am J Epidemiol 1991;134: Matanoski GM, Breysse PN, Elliott EA. Electromagnetic field exposure and male breast cancer. (Letter). Lancet 1991;337: Loomis DP. Cancer of breast among men in electrical occupations. (Letter). Lancet 1992;339: Tynes T. Electromagnetic fields and male breast cancer. Biomed Pharmacother 1993;47: Floderus B, Tornquist S, Stenlund C. Incidence of selected cancer in Swedish railway workers, Cancer Causes Control 1994;5: Rosenbaum PF, Vena JE, Zielezny MA, et al. Occupational exposures associated with male breast cancer. Am J Epidemiol 1994;140: Morton WE. Further investigation of housewife cancer mortality risk. Women Health 1982;7: Vagero D, Olin R. Incidence of cancer in the electronics industry: using the new Swedish cancer environment registry as a screening instrument. Br J Ind Med 1983;40: Vena JE, Graham S, Hellmann R, et al. Use of electric blankets and risk of postmenopausal breast cancer. Am J Epidemiol 1991;134: Vena JE, Freudenheim JL, Marshall JR, et al. Risk of premenopausal breast cancer and use of electric blankets. Am J Epidemiol 1994;140: Stevens RG. Re: Risk of premenopausal breast cancer and use of electric blankets. (Letter). Am J Epidemiol 1995;142: Vena JE, Marshall JR, Freudenheim JL, et al. Re: Risk of premenopausal breast cancer and use of electric blankets. The authors reply. (Letter). Am J Epidemiol 1995;142: Wertheimer N, Leeper E. Re: Risk of premenopausal breast cancer and use of electric blankets and use of electric blankets and risk of postmenopausal breast cancer. (Letter). Am J Epidemiol 1995;142: Vena JE, Freudenheim JL, Marshall JR, et al. Re: Risk of premenopausal breast cancer and use of electric blankets and use of electric blankets and risk of postmenopausal breast cancer. The authors reply. (Letter). Am J Epidemiol 1995;142: Theriault G. Electromagnetic fields and cancer risks. Rev Epidemiol Sante Publique 1992;40(suppl 1):

12 58 Schoenfeld et al. 30. Wertheimer N, Leeper E. Adult cancer related to electrical wires near the home. Int J Epidemiol 1982;11: Stevens RG, Davis S, Thomas DB, et al. Electrical power, pineal function, and the risk of breast cancer. FASEB J 1992;6: Guenel P, Raskmark P, Andersen J, et al. Incidence of cancer in persons with occupational exposure to electromagnetic fields in Denmark. Br J Ind Med 1993;50: Loomis DP, Savitz DA, Ananth CV. Breast cancer mortality among female electrical workers in the United States. JNCI 1994;86: Cantor KP, Dosemici M, Brinton LA, et al. Re: Breast cancer mortality among female electrical workers in the United States. (Letter). JNCI 1995;87: Coogan PF, Clapp RW, Newcomb PA. Occupational exposure to 60-Hertz magnetic fields and risk of breast cancer in women. Epidemiology 1996;7: Preston-Martin S. Breast cancer and magnetic fields. (Editorial). Epidemiology 1996;7: Tynes T, Hannevik M, Andersen A, et al. Incidence of breast cancer in Norwegian female radio and telegraph operators. Cancer Causes Control 1996;7: Fear NT, Roman E, Carpenter LM, et al. Cancer in electrical workers: an analysis of cancer registrations in England, Br J Cancer 1996;73: Kelsh MA, Sahl JD. Mortality among a cohort of electric utility workers, Am J Ind Med 1997;31: Coogan PF, Aschengrau A. Exposure to power frequency magnetic fields and risk of breast cancer in the Upper Cape Cod Incidence Study. Arch Environ Health 1998;53: Johansen C, Olsen JH. Risk of cancer among Danish utility workers a nationwide cohort study. Am J Epidemiol 1998; 147: Pollan M, Gustavsson P. High-risk occupations for breast cancer in the Swedish female working population. Am J Public Health 1999;89: Davis S, Mirick DK, Stevens RG. Residential magnetic fields and the risk of breast cancer. Am J Epidemiol 2002;155: McDowall ME. Mortality of persons resident in the vicinity of electricity transmission facilities. Br J Cancer 1986;53: Schreiber G, Swaen G, Maijers J, et al. Cancer mortality and residence near electricity transmission equipment: a retrospective cohort study. Int J Epidemiol 1993;22: Li C-Y, Theriault G, Lin RS. Residential exposure to 60-Hz magnetic fields and adult cancers in Taiwan. Epidemiology 1997;8: Feychting M, Forssen U, Rutquist LE, et al. Magnetic fields and breast cancer in Swedish adults residing near high-voltage power lines. Epidemiology 1998;9: Verkasalo PK, Pukkala E, Kaprio J, et al. Magnetic fields of high voltage power lines and risk of cancer in Finnish adults: nationwide cohort study. BMJ 1996;313: Gammon MD, Schoenberg JB, Britton JA, et al. Electric blanket use and breast cancer risk among younger women. Am J Epidemiol 1998;148: Zheng T, Holford TR, Mayne ST, et al. Exposure to electromagnetic fields from use of electric blankets and other in-home electrical appliances and breast cancer risk. Am J Epidemiol 2000;151: Laden F, Neas LM, Tolbert PE, et al. Electric blanket use and breast cancer in the Nurses Health Study. Am J Epidemiol 2000;152: McElroy JA, Newcomb PA, Remington PL, et al. Electric blanket or mattress cover use and breast cancer incidence in women years of age. Epidemiology 2001;12: Stevens RG. Electric power use and breast cancer: a hypothesis. Am J Epidemiol 1987;125: Caplan L, Schoenfeld ER, O Leary ES, et al. Breast cancer and electromagnetic fields a review. Ann Epidemiol 2000;10: Schoenfeld ER, Henderson K, O Leary E, et al. Magnetic field exposure assessment: a comparison of various methods. Bioelectromagnetics 1999;20: Kabat GC, O Leary ES, Schoenfeld ER, et al. Electric blanket use and breast cancer on Long Island. Epidemiology (in press). 57. Kaune WT, Zaffanella LE. Assessing historical exposures of children to power-frequency magnetic fields. J Exposure Anal Environ Epidemiol 1994;4: O Leary ES, Schoenfeld ER, Henderson K, et al. Wire coding in the EMF and Breast Cancer on Long Island Study relationship to magnetic fields. J Exposure Anal Environ Epidemiol (in press). 59. Wertheimer N, Leeper E. Electrical wiring configurations and childhood cancer. Am J Epidemiol 1979;109: Kaune WT, Savitz DA. Simplification of the Wertheimer- Leeper wire code. Bioelectromagnetics 1994;15: Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol 1. The analysis of case-control studies. (IARC Scientific Publication no. 32). Lyon, France: International Agency for Research on Cancer, Hosmer PW, Lemeshow S. Applied logistic regression. 2nd ed. New York, NY: John Wiley and Sons, Inc, Gammon MD, Santella RM, Neugut AI, et al. Environmental toxins and breast cancer on Long Island. I. Polycyclic aromatic hydrocarbon DNA adducts. Cancer Epidemiol Biomarkers Prev 2002;11: Gammon MD, Wolff MS, Neugut AI, et al. Environmental toxins and breast cancer on Long Island. II. Organochlorine compound levels in blood. Cancer Epidemiol Biomarkers Prev 2002;11: Erren TC. A meta-analysis of epidemiologic studies of electric and magnetic fields and breast cancer in women and men. Bioelectromagnetics 2001;5(suppl):S Slesin L, ed. Female breast cancer linked to EMFs for the third time. Associations still viewed cautiously. Microwave News 1996;16:1, Neutra RR, Del Pizzo V. A richer conceptualization of exposure for epidemiologic studies of the EMF mixture. Bioelectromagnetics 2001;5(suppl):S Davis S, Kaune WT, Mirick DK, et al. Residential magnetic fields, light-at-night, and nocturnal urinary 6-sulfatoxymelatonin concentration in women. Am J Epidemiol 2001;154: