Correspondence
Leukemia and Exposure to Magnetic Fields
N Engl J Med 1997; 337:1471-1474November 13, 1997
- Article
To the Editor:
The study of the possible association between acute lymphoblastic leukemia (ALL) in children and residential exposure to magnetic fields by Linet and colleagues (July 3 issue)1 was well designed and conducted. The evidence does not support all the authors' conclusions or Campion's call, in his editorial,2 for an end to further research on exposure to magnetic fields. Some results of the study are positive, and many issues remain unresolved.
Although the authors note that there was no significant effect at 0.2 μT, an a priori and arbitrary exposure level,3 the results show an elevated albeit not statistically significant odds ratio, which is consistent with the results of other studies. With an exposure cutoff point of 0.3 μT, which was recommended on the basis of analyses of previous studies,4 the authors report a larger and statistically significant odds ratio in the unmatched analysis (the results of the matched analysis are not reported). However, Linet and colleagues summarily dismiss this result by asserting that the number of children with exposures of 0.3 μT or higher was small. In fact, there were 45 case patients and 28 controls with these exposures, numbers that represent a small proportion of the total but that are statistically robust and rarely considered small in epidemiology. Even when the negative results obtained with the use of wire codes to categorize exposure are included in the meta-analysis I originally conducted for the National Academy of Sciences,5 the combined result is still positive and statistically significant.
Finally, the interpretation of the results is limited by possible biases. Although we know ALL is associated with higher socioeconomic status, the case patients had markedly lower family income than the controls. There was also a substantial difference in the response rates among case patients and controls (78 percent and 63 percent, respectively), and many subjects' residences were not wire coded (over one third).
Given these limitations of the study by Linet et al. and these positive results, abandoning research on exposure to magnetic fields on the basis of this study is premature. Given that ALL is the most common childhood cancer and that we have little insight into its causes, it is imperative from a public health standpoint that we aggressively seek explanations.
5 ReferencesDaniel Wartenberg, Ph.D.
Environmental and Occupational Health Sciences Institute, Piscataway, NJ 08855-11791
Linet MS ,Hatch EE ,Kleinerman RA , et al. Residential exposure to magnetic fields and acute lymphoblastic leukemia in children. N Engl J Med 1997;337:1-7
Full Text | Web of Science | Medline2
Campion EW . Power lines, cancer, and fear. N Engl J Med 1997;337:44-46
Full Text | Web of Science | Medline3
Wartenberg D ,Northridge M . Defining exposure in case-control studies: a new approach. Am J Epidemiol 1991;133:1058-1071
Web of Science | Medline4
Wartenberg D ,Savitz DA . Evaluating exposure cutpoint bias in epidemiologic studies of electric and magnetic fields. Bioelectromagnetics 1993;14:237-245
CrossRef | Web of Science | Medline5
National Research Council. Possible health effects of exposure to residential electric and magnetic fields. Washington, D.C.: National Academy Press, 1997.
To the Editor:
As an examination of the association between actual residential magnetic-field measurements and the risk of leukemia in children, the study by Linet et al. is by far the largest performed to date, and the results are troubling. The abstract and the accompanying editorial depict the study as definitive and the results as completely negative, with the matched analysis yielding an odds ratio of only 1.24 for an exposure of 0.2 μT or greater. However, the data in the body of the report are not nearly so reassuring. First, each case patient was matched to a control in the same local community within the nine-state study area, and in the matched analysis, the odds ratio for an exposure of 0.2 μT or greater was 1.53. It is unclear why the authors highlighted the results of the unmatched analysis in the abstract when the results of the matched analysis were appropriate. The authors retained the matching in the wire-code analyses because of possible variation in the relation between the wire code and the magnetic field among communities. Retaining the matching is no less important in the magnetic-field analyses, because of possible variation in the mean magnetic field among communities, as well as for minimizing the possible effects of other local environmental factors.
In the unmatched analysis, the odds ratio for an exposure of 0.3 μT or greater was 1.72, with a 95 percent confidence interval that excluded 1.0; the odds ratio in the matched analysis may have been even higher, but it was not given. The authors discount this result because it was not their a priori hypothesis.
Richard G. Stevens, Ph.D.
Pacific Northwest National Laboratory, Richland, WA 99352To the Editor:
Although the data reported by Linet et al. show that children exposed to magnetic fields of low intensity are not at increased risk for leukemia, we do not believe that these results can be extrapolated to people living close to high-voltage power lines.
The results reported by Linet et al. do not indicate that exposure to a time-weighted average residential magnetic-field level >0.2 μT constitutes a significant risk for childhood ALL; however, a statistically significant association was observed for levels >0.3 μT. The authors were careful not to draw a conclusion about this possible risk because of the small numbers and the absence of a dose–response relation. Very few people are usually exposed to residential magnetic-field levels >0.3 μT. However, this level might be regularly exceeded among people living close to high-voltage power lines. For instance, McMahan et al.1 found that the mean residential exposure for people living close to two 220-kV lines and two 66-kV lines was 0.49 μT. Moreover, we reported a geometric mean of 0.71 μT for people living near a 735-kV line, with time-weighted values ranging from 0.46 to 1.14 μT.2
2 ReferencesPatrick Levallois, M.D.
Denis Gauvin, M.Sc.
Centre de Santé Publique de Québec, Beauport, QC G1E 7G9, Canada1
McMahan S ,Ericson J ,Meyer J . Depressive symptomatology in women and residential proximity to high-voltage transmission lines. Am J Epidemiol 1994;139:58-63
Web of Science | Medline2
Levallois P ,Gauvin D , St. -Laurent J, Gingras S, Deadman JE. Electric and magnetic field exposures for people living near a 735-kilovolt power line. Environ Health Perspect 1995;103:832-837
CrossRef | Web of Science | Medline
To the Editor:
Linet et al. conclude, “Our results provide little evidence that living in homes characterized by high . . . magnetic-field levels . . . increases the risk of ALL in children.” In their case–control study of 638 case patients with ALL, the odds ratio was 1.24 (95 percent confidence interval, 0.86 to 1.79) for magnetic-field levels of 0.2 μT or greater, as compared with levels of less than 0.065 μT. Epidemiologic studies frequently lack the power to demonstrate that a 20 percent increase in the relative risk or odds ratio is significant at the 95 percent level, and this study is no exception. But an odds ratio of 1.24 can hardly be considered a negative result, as Campion and the press reported. One must remember that the true odds ratio is just as likely to be 1.79 as it is to be 0.86.
In July 1997, Feychting et al.1 reported on a study of leukemia and brain cancer in relation to occupational and residential magnetic fields in Sweden. They did not measure the fields themselves. For a residential exposure >0.2 μT, the relative risk of leukemia was 1.3 (95 percent confidence interval, 0.8 to 2.2), which is very similar to the results reported by Linet et al. For an occupational exposure at a level higher than 0.2 μT, the relative risk of leukemia was 1.7 (95 percent confidence interval, 1.1 to 2.7). The subjects with the highest exposure levels had a relative risk of 3.7. Also in 1997, Kheifets et al.2 reported an odds ratio of approximately 1.2 (range, 0.4 to 1.88) for adult leukemia with occupational exposure to electric fields of 10 V per meter or greater.
Although I am not convinced that there is an association, the remarkable consistency among the risk estimates in these three very different studies suggests an alternative hypothesis — that exposure to electromagnetic fields results in approximately a 20 percent increase in the risk of leukemia. Although this is a value that Buffler3 considers “very weak,” many persons would take seriously a 20 percent increase in their own risk of leukemia. To be sure, three studies do not make a meta-analysis. But far from laying the issue to rest, these recent studies point to the need for additional high-quality studies.
3 ReferencesMichael Gochfeld, M.D., Ph.D.
University of Medicine and Dentistry of New Jersey, Piscataway, NJ 08855-11791
Feychting M ,Forssen U ,Floderus B . Occupational and residential magnetic field exposure and leukemia and central nervous system tumors. Epidemiology 1997;8:384-389
CrossRef | Web of Science | Medline2
Kheifets LI ,London SJ ,Peters JM . Leukemia risk and occupational electric field exposure in Los Angeles County, California. Am J Epidemiol 1997;146:87-90
Web of Science | Medline3
Buffler PA. Uses of epidemiology in environmental medicine. In: Brooks SM, Gochfeld M, Herzstein J, Jackson RJ, Schenker MB, eds. Environmental medicine. St. Louis: Mosby, 1995:46-62.
To the Editor:
In the study by Linet et al., I am confused by the use of tesla as opposed to gauss as the unit of measurement. Was the study measuring something other than electromagnetic radiation? All the work the Environmental Protection Agency did on the subject involved electromagnetic radiation, and the measurements used in those studies are expressed in gauss units.
However, in Table 2 in the report by Linet et al., it appears that there is a statistically elevated odds ratio at the level of 0.400 to 0.499 μT: 3.28 in the unmatched analysis and 6.41 in the matched analysis. I looked for an explanation in the text, but there is none. The authors state that the increased risk at exposure levels of 0.300 μT or greater “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.”
Harold Funk
Communications, Energy and Paperworkers Union of Canada, Local 226, Vancouver, BC V5V 4H5, CanadaTo the Editor:
I disagree with Dr. Campion's assessment that the study by Linet et al. was so superior in quality and size that it has laid to rest the hypothesis that electromagnetic fields cause childhood ALL.
Campion attributes a litany of biases to the studies preceding the study by Linet et al., implying that most of the biases apply to most of the studies. This is misleading. The reason the National Academy of Sciences1 concluded that the association between the wire code and the risk of leukemia (quite aside from the implications for electromagnetic fields) warranted further research was that no common pattern of confounding, recall bias, measurement error, or other factors provided an easy explanation of the association. The real qualitative advance of the study by Linet et al. was to obtain measurements within two years of the diagnosis. Perhaps this is why they came almost as close as the Swedish researchers2 in showing a direct effect of measurements of electromagnetic fields. The Swedish group reconstructed the dose just before the diagnosis.
The report by the National Academy of Sciences included a meta-analysis of 11 studies worldwide and concluded that the results were compatible with an odds ratio of 1.5 for a relation between wire codes and the risk of leukemia. Linet et al. would have had to perform wire coding for 1500 case patients and controls to obtain that odds ratio. Instead, they show the results for 408 case patients and controls, with an odds ratio of 0.88 and a 95 percent confidence interval that includes 1.5.
The problem here is that, as with many environmental effects, an odds ratio of 1.3 or 1.5, if real, would be of regulatory concern but is at the margin of what we epidemiologists can reliably show. The ubiquity of electromagnetic fields warrants continued careful consideration.
2 ReferencesRaymond Richard Neutra, M.D., Dr.P.H.
California Department of Health Services, Berkeley, CA 94704-10111
National Research Council. Possible health effects of exposure to residential electric and magnetic fields. Washington, D.C.: National Academy Press, 1997.
2
Feychting M ,Ahlbom A . Magnetic fields and cancer in children residing near Swedish high-voltage power lines. Am J Epidemiol 1993;138:467-481
Web of Science | Medline
To the Editor:
Campion's editorial raises important questions about the interplay between public concern about cancer and epidemiologic research on electromagnetic fields. However, I cannot agree with his statement, “It is sad that hundreds of millions of dollars have gone into studies that never had much promise of finding a way to prevent the tragedy of cancer in children.” The absence of an animal model does not constitute proof that a given agent has no carcinogenic effect in humans. Campion implies that we can know a priori that a potential hazard is harmless. Unfortunately, this is not possible. Furthermore, the fact that the early studies of electromagnetic fields and cancer, which indicated an association, were crude does not mean that they should have been discounted. More often than not, the first studies identifying a new risk factor for a disease are small and methodologically limited. Over the 18-year period since the first study, by Wertheimer and Leeper,1 was reported, there has been a progression toward more methodologically rigorous studies of electromagnetic fields. The study by Linet et al., which Campion appropriately praises, would never have been proposed and funded if it had not built on these earlier studies.
Epidemiologic research can proceed only by weighing the existing evidence and by constantly honing epidemiologic methods and improving the quality of the studies in a given area. Far from having been wasted, the money spent on this research has yielded important information regarding the apparent lack of a strong association between electromagnetic fields and childhood leukemia. This information should both help allay the fear and galvanize researchers to pursue new hypotheses. In addition, the methodologic improvements benefit the study of other outcomes, including breast cancer, for which a plausible biologic mechanism exists2 but convincing results are still lacking.
2 ReferencesGeoffrey C. Kabat, Ph.D.
University Medical Center, Stony Brook, NY 11794-80361
Wertheimer N ,Leeper E . Electrical wiring configurations and childhood cancer. Am J Epidemiol 1979;109:273-284
Web of Science | Medline2
Stevens RG ,Davis S ,Thomas DB ,Anderson LE ,Wilson BW . Electric power, pineal function, and the risk of breast cancer. FASEB J 1992;6:853-860
Web of Science | Medline
Author/Editor Response
The authors reply:
To the Editor: We concluded that our results “provide little evidence” that the risk of ALL is increased among children living in homes with high magnetic fields or high wire codes. We disagree that this conclusion is “completely negative” and believe it conveys our view that the results are reassuring rather than alarming.
Analysis of exposures at levels of 0.2 μT (2.0 mG) or higher was chosen in part because our study had inadequate power at higher levels to detect odds ratios of a magnitude considered likely given previous findings for measured fields.1 Nonetheless, we noted the significant odds ratio of 1.72 at exposure levels of 0.3 μT or greater in the unmatched analysis (odds ratio in the matched analysis, 1.82; 95 percent confidence interval, 0.92 to 3.60). We did not “summarily dismiss” this result but mentioned it three times in the Discussion section and noted the possibility of an increased risk at such high levels. We tended to discount the increase in our interpretation because of the post hoc nature of the analysis, the inconsistent dose–response pattern (a markedly increased risk at a level of 0.400 to 0.499 μT but no increased risk at a level of 0.5 μT or higher), the absence of a significant trend, and the small number of exposed children.
We presented the results of the unmatched analysis in addition to the results of the matched analysis because of the larger sample in the unmatched analysis. None of the matched analyses for exposure levels of 0.2 μT or higher, 0.3 μT or higher, 0.4 μT or higher, or 0.5 μT or higher showed a significantly increased risk.
Not all magnetic-field estimates in the Swedish study were for the year preceding the diagnosis. Magnetic fields were estimated for the last year before the diagnosis in which the case patients lived near a power line, which could have been several years before the diagnosis for those who moved.2 Parallels drawn between the Swedish estimates and our measurements are of uncertain relevance.
The relative risk of 1.5 reported by the National Academy of Sciences was from a meta-analysis comparing the combined two highest wire-code categories with all the lower wire-code categories.1 Using this categorization, we calculated an odds ratio of 0.94 (95 percent confidence interval, 0.68 to 1.29), which indicates no increased risk near high-voltage lines. Although the National Academy of Sciences committee could not explain previously reported associations between childhood leukemia and high wire codes, the two studies with the highest odds ratios for wire codes have been noted to have flaws that could have led to biased estimates. Future meta-analyses should take the quality of the studies into account.
A lower participation rate among controls than among case patients might produce a slight shift toward lower wire-code levels among controls,3 as would the selection of controls with a higher socioeconomic status.4 Thus, both potential sources of bias would tend to inflate the risk estimates. The analyses were adjusted for measures of socioeconomic status.
4 ReferencesMartha S. Linet, M.D.
Robert E. Tarone, Ph.D.
National Cancer Institute, Bethesda, MD 20892Leslie L. Robison, Ph.D.
Children's Cancer Group, Arcadia, CA 910061
National Research Council. Possible health effects of exposure to residential electric and magnetic fields. Washington, D.C.: National Academy Press, 1997.
2
Feychting M ,Ahlbom A . Magnetic fields and cancer in children residing near Swedish high-voltage power lines. Am J Epidemiol 1993;138:467-481
Web of Science | Medline3
Kleinerman RA ,Linet MS ,Hatch EE , et al. Magnetic field exposure assessment in a case-control study of childhood leukemia. Epidemiology 1997;8:575-583
CrossRef | Web of Science | Medline4
Jones TL ,Shih CH ,Thurston DH ,Ware BJ ,Cole P . Selection bias from differential residential mobility as an explanation for associations of wire codes with childhood cancer. J Clin Epidemiol 1993;46:545-548
CrossRef | Web of Science | Medline
Author/Editor Response
Dr. Campion replies:
If one is thoroughly convinced that there is an association between an exposure and a disease, then it may be tempting to approach any new study as an opportunity to find something that supports the association and ignore the rest of the data. The report by Linet et al. made it perfectly clear that in the subgroup analyses there was one odds ratio that was statistically significant — namely, the odds ratio for the group that had the second-highest exposure with the direct measurements (but not with the indirect, wire-code measurements). One cannot place much weight on an association found only in a post hoc analysis of a subgroup representing only 2.2 percent of the children with leukemia.
I have to disagree with Dr. Kabat's view that the research on electromagnetic fields has been an effective use of resources. Years ago, on the basis of inadequate data, a premature conclusion that residential exposure to electromagnetic fields was hazardous and caused ALL was widely publicized. That belief produced the pressures that made the larger studies necessary. We now find that there is little evidence to support such a conclusion, so we really have not learned anything at all about leukemia. It is unfortunate, even irrational, if large amounts of research dollars are spent because of fear and suspicion and an uncritical science that begins with the proposition that nearly anything might be hazardous until proved otherwise.
Edward W. Campion, M.D.
- Citing Articles (3)
Citing Articles
1
A. Maes, M. Collier, S. Vandoninck, P. Scarpa, L. Verschaeve. (2000) Cytogenetic effects of 50 Hz magnetic fields of different magnetic flux densities. Bioelectromagnetics 21:8, 589-596
CrossRef2
Richard K. Severson, Julie A. Ross. (1999) The causes of acute leukemia. Current Opinion in Oncology 11:1, 20
CrossRef3
D Wartenberg. (1998) Residential magnetic fields and childhood leukemia: a meta-analysis.. American Journal of Public Health 88:12, 1787-1794
CrossRef
