|
|
| |||||||||||||||||||||||||||||||||||||||||
Background Previous studies found associations between childhood leukemia and surrogate indicators of exposure to magnetic fields (the power-line classification scheme known as "wire coding"), but not between childhood leukemia and measurements of 60-Hz residential magnetic fields.
Methods We enrolled 638 children with acute lymphoblastic leukemia (ALL) who were under 15 years of age and were registered with the Children's Cancer Group and 620 controls in a study of residential exposure to magnetic fields generated by nearby power lines. In the subjects' current and former homes, data collectors blinded to the subjects' health status measured magnetic fields for 24 hours in each child's bedroom and for 30 seconds in three or four other rooms and outside the front door. A computer algorithm assigned wire-code categories, based on the distance and configuration of nearby power lines, to the subjects' main residences (for 416 case patients and 416 controls) and to those where the family had lived during the mother's pregnancy with the subject (for 230 case patients and 230 controls).
Results The risk of childhood ALL was not linked to summary time-weighted average residential magnetic-field levels, categorized according to a priori criteria. The odds ratio for ALL was 1.24 (95 percent confidence interval, 0.86 to 1.79) at exposures of 0.200 µT or greater as compared with less than 0.065 µT. The risk of ALL was not increased among children whose main residences were in the highest wire-code category (odds ratio as compared with the lowest category, 0.88; 95 percent confidence interval, 0.48 to 1.63). Furthermore, the risk was not significantly associated with either residential magnetic-field levels or the wire codes of the homes mothers resided in when pregnant with the subjects.
Conclusions Our results provide little evidence that living in homes characterized by high measured time-weighted average magnetic-field levels or by the highest wire-code category increases the risk of ALL in children.
The shortcomings of earlier epidemiologic studies have been extensively reviewed.10,12,13,14,15 Inconsistent findings, discrepancies between results based on proxy estimates and those based on direct magnetic-field measurements, and the absence of supportive laboratory evidence or a plausible biologic mechanism of disease causation10,16 have resulted in uncertainties about the relation, if any, between childhood leukemia and exposure to magnetic fields. Widespread concern and the limitations of previous studies led us to evaluate residential exposure to magnetic fields in a comprehensive case–control study of acute lymphoblastic leukemia (ALL) in childhood, conducted by the Children's Cancer Group.
Methods
Subjects
The methods of this study are described in detail elsewhere.17 Briefly, a group of the 1914 children with ALL and the 1987 controls participating in a nationwide telephone-interview study conducted by the Children's Cancer Group was eligible for the assessment of residential exposure to magnetic fields. Eligible case children received a diagnosis of ALL before the age of 15 years, between 1989 and 1994, and were registered with the Children's Cancer Group. Eligible controls were selected by random-digit telephone dialing18 and were individually matched to the children with ALL according to the first eight digits of the telephone number, age, and race. Eligibility for the assessment of magnetic-field exposure was restricted to the 851 case patients and the 825 controls who participated in the initial telephone interview (representing 96 percent and 75 percent, respectively, of those who were eligible) and who resided in one of nine states (Illinois, Indiana, Iowa, Michigan, Minnesota, New Jersey, Ohio, Pennsylvania, and Wisconsin) on the reference date, defined as the date of diagnosis of ALL for each case patient. The same date was assigned to the case patient's matched control for the purpose of determining which children's residences would have magnetic-field assessments. Mothers of 98 percent of the children with ALL (832 case patients) and 97 percent of the controls (n = 801) who responded to the telephone interview also provided lifetime residential histories for the subjects. Because we did not evaluate 65 of these case patients and 76 of the controls further, once the sample-size goals had been achieved, 767 case patients and 725 controls were eligible for measurements of residential magnetic fields.17
For each child under the age of five years, we attempted to measure magnetic fields in all the homes the subject had lived in for at least six months and required that at least 70 percent of the child's life have been spent in the measured homes. For each child over the age of five, we measured one or two homes, provided that the child had lived in them for at least 70 percent of the five years immediately preceding the reference date. We chose the five-year reference period closest to the date of diagnosis because of hypothesized cancer-promoter effects, since no evidence exists that the low strength of residential magnetic fields can induce genotoxic effects.10,16
Overall, 78 percent of the eligible patients participated (83 percent participation among the 767 case patients eligible for residential measurements times 98 percent participation in the lifetime residential history times 96 percent participation in the initial telephone interview), as did 63 percent of the controls (86 percent participation among the 725 eligible controls times 97 percent and 75 percent, respectively), resulting in a final study population of 638 case patients and 620 controls. Reasons for nonparticipation included refusal by the child's parents, inability to locate the child or too many changes of residence, lack of approval by the hospital institutional review board for the magnetic-field measurements, and refusal by the child's physician (this was a factor only for the children with cancer). Some subjects could not be included because the current occupants of subjects' former homes denied permission for the magnetic-field measurements.
We ascertained the residential wire-code category for a subgroup of the pairs of children with ALL and their controls who were eligible for magnetic-field measurements. We restricted assessment of wire codes to pairs in which both the case patient and the matched control had "residential stability" — that is, both paired members had lived in one home for at least 70 percent of the reference period (this residence is hereafter designated the "main residence"). Among the 428 such residentially stable pairs identified, 12 pairs were excluded because the technician could not locate the home or accurately diagram nearby power lines at one of the residences. Technicians assessed most homes eligible for wire-code classification even if they could not obtain permission to measure magnetic fields, since access to the residence or the surrounding property was not necessary for wire coding. To evaluate the risk of ALL associated with the subject's residential wire code during the mother's pregnancy with the subject, technicians evaluated residences in which the subject's family had resided for at least five months during the index pregnancy ("residence during pregnancy") for all subjects under the age of three years (151 matched case–control pairs) and for those whose homes were assessed as part of the wire coding of the main residence, for a total of 230 case–control pairs.
Measurement Protocol
Magnetic-Field Measurements
Technicians blinded to the subjects' case or control status used an Emdex-C meter (Electric Field Measurements, West Stockbridge, Mass.), which measures extremely-low-frequency magnetic fields (40 to 300 Hz, a range that includes 50-Hz and 60-Hz levels, frequencies evaluated in prior epidemiologic studies) with a three-axis induction-coil sensor.17 Derived from two personal-exposure studies,19,20 the standardized measurement protocol included a 24-hour measurement in the child's bedroom (with the meter placed under or adjacent to the bed); 30-second measurements in the center of the child's bedroom, the family room, the kitchen, and the room in which the mother slept during the index pregnancy; and a 30-second outdoor measurement made within 0.9 m (3 ft) of the front door.17
Wire Coding
Technicians (who were unaware of whether a case patient or a control currently or formerly lived in each residence evaluated) drew diagrams and recorded systematically the distance from the home of any overhead power lines within 46 m (150 ft) of the residence, including transmission lines, thick and thin three-phase primary-distribution lines (which carry electric power from substations to surrounding neighborhoods), any open (with separated wires) or spun (with wires bound together) secondary distribution lines, and first-span secondary distribution lines.17 On the basis of the diagrams, a computer algorithm assigned a wire code to each residence according to the five-category Wertheimer–Leeper classification1,3,11 and the modified three-category Kaune–Savitz scheme.21 As in earlier studies,3,5 we found that measured magnetic-field levels (i.e., the arithmetic means of 24-hour measurements from 858 residences) rose with increasing Wertheimer–Leeper1,3,11 and Kaune–Savitz21 wire-code categories (unpublished data).
Statistical Analysis
Magnetic-Field Measurements
For each eligible residence, a summary magnetic-field level was calculated from a weighted average of the room measurements. The weights were derived from the personal-exposure study and based on the estimated time spent by children according to age.17,19,20 If measurements were not obtained in all rooms, then the weighted average was based on a standardized hierarchy of measurements.17 The primary measure of exposure for each subject was an average of the summary level for all the eligible measured homes, weighted according to the duration of residence. We used odds ratios and 95 percent confidence intervals to estimate the risk of ALL.22 Before undertaking any case–control comparisons, we identified four exposure categories for residential magnetic-field levels (<0.065 µT [the reference group], 0.065 to 0.099 µT, 0.100 to 0.199 µT, and >0.200 µT), based on the distribution of measurements in the control homes. These categories were similar to those used in earlier investigations.3,5,6 We calculated results using unmatched analysis as well as analysis of matched case–control pairs.22
We used stratified and logistic-regression analyses to explore the effects of age at the reference date, sex, race (though the very small number of nonwhites limited this evaluation), socioeconomic status (indicated by family income, the mother's and father's educational level and occupation, home ownership, and family size), temporal factors (year, season, and time of day when the measurements were made), demographic characteristics (degree of urbanization and type of residence), and dose–response relations using continuous measurements.22 We also evaluated birth order, birth weight, the mother's age at the child's delivery, and medical x-ray studies during pregnancy as potential confounding factors. We excluded nine case patients and one control who had Down's syndrome, since this disorder has been linked to 10-to-40-fold increases in the risk of acute leukemia.23 We included 629 case patients and 619 controls in the final unmatched analysis, and 463 case–control pairs in the matched analysis.
Wire Coding
Because the relation between power-line configurations and magnetic-field strength may vary geographically,5,24 we retained the matched design of the initial nationwide phase of the study for the wire coding of the main residence. The Wertheimer–Leeper wire-code categories used in the analysis include underground (buried) power lines plus very-low-current configuration (the reference group), ordinary low-current configuration, ordinary high-current configuration, and very-high-current configuration.1,3,11 The modified Kaune–Savitz wire-code categories were as follows: low (the reference group), medium, and high.21 We used matched-pairs analysis to evaluate the risk of ALL in relation to the wire-code category of the main residence (including 408 case–control pairs, after the exclusion of subjects with Down's syndrome) and the residence during pregnancy (a total of 225 pairs, including 149 pairs of subjects under the age of three, after the exclusion of subjects with Down's syndrome); conditional logistic regression was used to control for socioeconomic and demographic factors and other potential confounders.22
Results
Characteristics of the Subjects
The controls were similar to the case patients (Table 1), except for their higher total family income (P<0.001). ALL was not associated with the mother's age at delivery of the subject, the number of children in the family, the birth order of the subject (data not shown), the type of residence, the degree of urbanization, home ownership, or the interval between the reference date and the date of the measurements (data not shown). All estimates of risk have been adjusted for the age of the subject at the reference date, the subject's sex, the mother's educational level, and family income.
|
Risk estimates based on the summary residential magnetic-field exposures for a priori measurement categories did not differ significantly from unity either for all the subjects (629 case patients and 619 controls) or for the 463 matched pairs (Table 2), nor did risk increase significantly with increasing summary magnetic-field levels (P for trend = 0.22 for the unmatched analyses and 0.12 for the matched analyses). Risk was higher with estimated summary exposures of 0.300 µT or more (odds ratio, 1.72; 95 percent confidence interval, 1.03 to 2.86; 45 case patients and 28 controls); however, risk did not increase significantly with increasing exposure when exposure was evaluated as a continuous variable (P for trend = 0.15 for the unmatched analysis and 0.09 for the matched analysis).
|
Main-Residence Wire-Code Patterns
For the main residence, we found no association between the risk of ALL and residence in a home classified in the highest wire-code category according to either wire-code classification (Table 3). There were no positive or statistically significant dose–response trends, and results were not materially changed when adjusted for potentially confounding variables.
|
As regards the homes resided in during pregnancy by the mothers of 257 case patients and 239 controls, the odds ratio for ALL was 0.75 (95 percent confidence interval, 0.45 to 1.24) for a magnetic-field level of 0.065 to 0.099 µT, as compared with the reference category (<0.065 µT); 1.32 (95 percent confidence interval, 0.81 to 2.15) for a level of 0.100 to 0.199 µT; and 1.24 (95 percent confidence interval, 0.69 to 2.23) for a level of 0.200 µT or higher (P for trend = 0.25). Among the 225 matched pairs whose mothers' residences during pregnancy were wire-coded, the odds ratios for ALL were 1.20 (95 percent confidence interval, 0.74 to 1.95) for the Wertheimer–Leeper code-configuration category of "ordinary low"; 1.07 (95 percent confidence interval, 0.61 to 1.86) for "ordinary high"; and 1.49 (95 percent confidence interval, 0.66 to 3.37) for "very high," as compared with the reference category of "underground plus very low" (P for trend =0.07). For children under the age of three whose mothers' homes during pregnancy were wire-coded (149 matched pairs), the odds ratios were not significantly elevated and the risks did not increase significantly with higher wire-code categories (P for trend =0.19).
Discussion
We found no significant excess risk of childhood ALL associated with time-weighted average summary residential magnetic-field levels of 0.200 µT or greater, nor did we observe any significant dose–response trends. There was a tendency for the risk to be higher among subjects with summary exposure levels of 0.300 µT or more, but the number of children with such high levels was small. The risk of childhood ALL was not associated with high wire codes for either the subject's main residence or the mother's residence during pregnancy. Adjustment for socioeconomic, demographic, or other potentially confounding variables had little effect on the risk.
In contrast to three earlier U.S. studies,1,3,5 we found no association between the highest wire-code category and an elevated risk of childhood ALL. Our data demonstrated a significant correlation between measured magnetic fields and wire codes (unpublished data), as was found in previous studies.24,25,26 The lack of association between childhood ALL and wire-code categories is particularly noteworthy since public concern10 has been driven primarily by the excess risks linked with surrogate or historical estimates of residential magnetic-field exposure.1,3,5,6
The results of our measurements of magnetic-field levels, like those of four earlier investigations,3,4,5,6 also show no significant increase in the risk of ALL among children whose residences had measured magnetic-field levels of 0.200 µT or higher, based on a priori categories. The small increase in risk at estimated exposures of 0.300 µT or more derived from a significant excess incidence of ALL at the intermediate level of 0.400 to 0.499 µT, but the odds ratios were close to unity for estimated exposure levels of 0.500 µT or greater, and the P value for trend was not significant. We cannot exclude the possibility of a small increase in risk among children in homes with very high magnetic-field levels, as suggested in studies using historical estimates of residential magnetic-field exposure.4,6,7
We designed our investigation to address the limitations of earlier studies, particularly the lengthy intervals (typically years or decades) between the diagnosis of ALL and measurements of magnetic fields. In our study magnetic fields were usually measured within 24 months after the date of diagnosis in the children with ALL.17 Previous studies also included fewer cases of childhood leukemia, measured fields during a smaller proportion of the reference period or lacked a standardized reference interval for the evaluation of magnetic fields, and evaluated fewer potential confounding variables. Some of the earlier studies selected controls who moved less frequently than the case patients or failed to blind data collectors to the case or control status of the subjects living in each residence evaluated.10,12,13,14,15 We measured residential magnetic-field levels for nearly four times the numbers of case patients and controls in the largest previous investigation.5 An important strength of our study was that magnetic-field measurements covered more than 95 percent of the reference period for 77 percent of subjects and more than 90 percent of the reference period for 83 percent of subjects.17
We made a major effort to achieve a high rate of participation in the study, despite the substantial burden for families (an average of three hours for interviews and measurements). Overall, 78 percent of eligible case patients and 63 percent of eligible controls participated. Many of the reasons for not participating were unrelated to refusal by the subjects or their parents; they included refusals of permission for testing by current occupants of former residences or the failure of subjects to meet eligibility requirements (such as residential stability).
To address concern about possible response bias,27,28,29 we instructed the technicians to diagram the homes of 119 children who were identified during random-digit dialing as potential controls but whose parents declined permission for participation; we found that the proportion of these homes assigned by the computer algorithm to the highest wire-code category was similar to that among the subjects in our study.17 Moreover, the technicians diagrammed virtually all eligible residences of subjects whose families refused permission for magnetic-field measurements, since neither residential nor property access was necessary for assigning wire codes to residences. Residential mobility was similar for case patients and controls in this study, in contrast to an earlier investigation,3 which has been criticized because the case patients changed residences considerably more often than the controls.10,12,13,14,15 Additional strengths of our investigation included the collection of the exposure data on a blinded basis; the personal-exposure studies to develop19 and evaluate20 the measurement protocol; the routine calibration of all magnetic-field (Emdex) meters; the lengthy initial training, retraining, and site visits of measurement staff; the independent rediagramming of a substantial proportion of residences, which showed good concordance of assigned wire codes (unpublished data); and the regular review of all measurements, with detailed investigation of potential errors.17
A limitation of our investigation and all previous studies is the absence of measurements for individual residences in the years preceding the diagnosis of cancer. It is not known how well a single 24-hour measurement characterizes contemporary exposure, much less magnetic-field exposure years earlier. Very limited data suggest a moderate correlation between repeated spot measurements taken in the same residential location several years after the initial measurements.30 To examine the reproducibility and seasonal variation of magnetic-field measurements, we initiated a detailed longitudinal study of 50 homes in Detroit and Minneapolis. The preliminary results suggest good reproducibility and relatively little seasonal variation over a one-year period (Banks R, et al.: unpublished data). Repeated measurements in a large sample of homes over a longer period would help to resolve this issue. The selection of controls by random-digit dialing has known weaknesses,27 but the use of alternative control groups was not feasible.17 The only major difference between the case patients and the controls — a higher family income among controls — was probably due to the use of controls obtained by random-digit dialing,27 but this difference did not confound the relation between magnetic-field exposure and childhood ALL.
In summary, our comprehensive case–control investigation did not find significantly increased risks of ALL associated with time-weighted average summary residential magnetic-field measurements or with residence in homes characterized by a high wire-code category during the five years immediately preceding the diagnosis of ALL or during the index pregnancy. The finding of a tendency for risk to be higher at measured magnetic-field levels of 0.300 µT or greater was based on small numbers and was not characterized by a consistent pattern or a significant trend. Our results provide little support for the hypothesis that living in homes with high time-weighted average magnetic-field levels or in homes close to electrical transmission or distribution lines is related to the risk of childhood ALL.
Supported in part by a grant from the National Cancer Institute (RO1 CA48051) and by the University of Minnesota Children's Cancer Research Fund.
We are indebted to the members of the Advisory Committee (including Dr. Lawrence Fischer, director of the Institute for Experimental Toxicology, Michigan State University [chairperson]; Dr. Ron Brookmeyer, Department of Biostatistics, Johns Hopkins University School of Hygiene and Public Health; Dr. Raymond Greenberg, vice-president for academic affairs and provost, Medical University of South Carolina; Dr. Martin Misakian, National Institute of Standards and Technology; and former member Dr. Howard Wachtel, Department of Electrical Engineering, University of Colorado) for their guidance and their numerous constructive suggestions on all aspects of the study. We are also indebted to Dr. John Boice, Jr., former chief of the Radiation Epidemiology Branch, National Cancer Institute, for his support in all phases of the study; to the staff of Westat, Inc., Rockville, Md. (including Beth Bridgeman, Kathy Deutchman, Susan Englehart, Susan Gardner, Vickie Griffis, Teferra Hailu, Barbara Hood, Nancy LaVerda, Pat Leonard, Judy Light, Bob McConnell, Pat Mueller, Arbarna Nathan, Margaret Pacious, Michelle Tanenbaum, Shirley Tipton, Susan Ditty Van-Till, and Freda Wentz), for data collection and data-management support; to the employees of Enertech, Campbell, Calif., for wire coding (Bob Workley and Esther Workley) and for programming meters (Richard Iriye); and to Mr. Jan Erik Deadman, School of Occupational Health, McGill University, Montreal, for advice on the assessment of magnetic-field exposure.
Source Information
From the Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, Md. (M.S.L., E.E.H., R.A.K., D.R.F., S.W., R.E.T.); the Children's Cancer Group, Arcadia, Calif. (L.L.R., R.K.S.); the Division of Pediatric Epidemiology and Clinical Research, University of Minnesota School of Medicine, Minneapolis (L.L.R., R.K.S.); EM Factors, Richland, Wash. (W.T.K.); Westat, Inc., Rockville, Md. (C.M.H., S.N.); and Information Management Services, Rockville, Md. (C.T.H.). Investigators and institutions participating in the Children's Cancer Group are listed in the Appendix.
Address reprint requests to Dr. Linet at the Division of Cancer Epidemiology and Genetics, National Cancer Institute, Executive Plaza North, Suite 408, Bethesda, MD 20892-7362.
References
The principal investigators and participating institutions in the Children's Cancer Group (with grants from the National Cancer Institute in parentheses) were as follows: W.A. Bleyer, A. Khayat, H. Sather, M. Krailo, J. Buckley, D. Stram, and R. Sposto, Group Operations Center, Arcadia, Calif. (CA13539); R. Hutchinson, University of Michigan Medical Center, Ann Arbor (CA02971); S. Shurin, Rainbow Babies and Children's Hospital, Cleveland (CA20320); E. Baum, Children's Memorial Hospital, Chicago (CA07431); F.L. Johnson, Wyler Children's Hospital, Chicago; F. Ruymann, Children's Hospital of Columbus, Columbus, Ohio (CA03750); J. Mirro, Children's Hospital, Pittsburgh (CA36015); W. Woods, University of Minnesota, Minneapolis (CA07306); A. Meadows, Children's Hospital, Philadelphia (CA11796); P. Brietfield, Riley Hospital for Children, Indianapolis (CA13809); R. Wells, Children's Hospital Medical Center, Cincinnati (CA26126); R. Tannous, University of Iowa Hospitals and Clinics, Iowa City (CA29314); G. Gilchrist, Mayo Clinic, Rochester, Minn. (CA28882); and M. Donaldson, University of Medicine and Dentistry of New Jersey, Camden.
| |||||||||||||||||||||||||||||||||||||||||
Related Letters:
Leukemia and Exposure to Magnetic Fields
Wartenberg D., Stevens R. G., Levallois P., Gauvin D., Gochfeld M., Funk H., Neutra R. R., Kabat G. C., Linet M. S., Tarone R. E., Robison L. L., Campion E. W.
Extract |
Full Text
N Engl J Med 1997;
337:1471-1474, Nov 13, 1997.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
