Savitz, David A.
Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
Is there a link between exposure to magnetic fields and the risk of miscarriage? Past epidemiologic studies have been few and inconsistent, 1,2 but with the publication in this issue of results from two large and sophisticated studies in Northern California, 3,4 the question takes on a new level of interest. The results of the case-control study by Lee et al.3 motivated the development of a prospective cohort study by Li et al.4 to determine whether the findings of a positive association with indices of peak magnetic fields and variability could be replicated, and they were.
There is scant biologic justification for exploring a link between magnetic fields and miscarriage. 1,5 Previous epidemiologic research mainly addresses electric blankets and heated waterbeds (pertinent to nighttime exposure only). Two early reports found an association between use of electric blankets or ceiling cable radiant heat (in seasons with falling temperatures) and risk of miscarriage. 6,7 These were limited by use of vital record reports of miscarriage in a previous pregnancy. More rigorous studies have reported small positive 8 or null 9 associations for electric blanket use during pregnancy and miscarriage, and neither study found an association with use of waterbed heaters.
In occupational settings, the study of magnetic fields and miscarriage risk has focused on video display terminals (VDTs). There is a sizable literature that indicates no association between work with VDTs and miscarriage. 10 However, this is of limited relevance to extremely low-frequency magnetic fields since the equipment and work location of most VDT users does not produce elevated exposure. One attempt to isolate subsets of VDT workers with measurable increases in exposure found no association (adjusted OR = 0.9, 95% CI = 0.6–1.4) and no gradient in risk with duration of exposure. 11 In contrast, a Finnish occupational study 12 subdivided VDTs based on the measured magnetic fields close to the device and found no association with low exposures (<4.0 mG) but fairly strong associations (relative risks of 2–4) with the highest exposure.
The literature on general residential exposures (most similar to the studies from California) is limited to one report of a sizable but highly imprecise association between elevated magnetic field exposure and subclinical early pregnancy loss 13 and another small study of clinically recognized miscarriage with no indication of an association. 14 No studies have considered miscarriage risk in relation to multiple sources of magnetic fields through measurement or modeling.
A major strength and also a serious potential pitfall of the two new studies lies in the construction of exposure indices. Previously, time-weighted averages and area measurements (eg, average magnetic field in the bedroom) have been used to quantify exposure. These two summary measures suffer from well-recognized limitations.
Unlike the fixed meters used in epidemiologic studies, women are not stationed at the center of the room, one meter above the floor. Measurements obtained in this manner fail to incorporate differences in exposure that result from moving through different environments. If personal meters (recording exposures through a range of occupied environments) can capture behavioral sources of variability in exposure, then the personal measurements will be more valid than those of an area monitor.
The mechanism by which magnetic fields might cause health effects is not known, nor is the relevant exposure metric. Physical agents (even more than chemicals) offer an almost infinite menu of options for measurement, including averages, peaks, variability, and others. To rely on a time-weighted average solely because it is the most familiar summary index is unjustified.
Both studies reported in this issue incorporate the two refinements of behavioral determinants of exposure and multiple indices of exposure. But each of these refinements has pitfalls. The misclassification of fixed area measurements relative to individual exposure is often argued to be non-differential with regard to health outcome, ie, the accuracy, though less than perfect, is not modified by disease status. When individual behavior is incorporated into exposure measures, there is a distinct opportunity for behavioral differences associated with health outcome to introduce differential misclassification. This is particularly true when the measurement period overlaps the period of disease etiology and identification.
The use of alternative measures, namely peak exposures and indices of exposure variability, may well exacerbate the vulnerability to differential misclassification in the California miscarriage studies. Any home, workplace, or community environment has specific and distinct sources of high magnetic field exposure. The more mobile the study participant is within her home or workplace, the greater her opportunity to encounter at least one source of high exposure to magnetic fields and, similarly, to experience substantial variability in her exposure over time. The extent of a woman’s “micromobility” will not affect fixed area measurements at all and will affect time-weighted average personal measurements little, but will affect indices based on peaks and temporal variability greatly. In both of the California studies, the data suggest little or no association between time-weighted average magnetic fields and miscarriage. The support for an association is based principally on maximum magnetic fields 3,4 and rate of change metrics. 3
Perhaps the investigators have pinpointed biologically important indices, as they suggest. But it seems even more plausible that the results are based on behavioral differences between women with healthy pregnancies and women who either experienced a miscarriage or were destined to have one. Specifically, if the “micromobility” of women who had or go on to have miscarriages differ from that of women whose pregnancies continue, the measures of association between these indices of magnetic fields and miscarriage will be affected. This hypothesis of a specific form of bias is readily testable.
In the case-control study by Lee et al., 3 exposure was assessed at 30 weeks’ gestation for women whose pregnancies continued and at the equivalent point relative to the onset of pregnancy for women who had miscarried. Although the time from conception to measurement was fixed, the women with healthy pregnancies were late in their pregnancies at the time of measurement while the women with miscarriage were several months past their pregnancy loss. Women who are well into their third trimester of pregnancy are less mobile on average than non-pregnant women simply due to their size, awkwardness, and energy level. To produce an artifactual difference in exposure, independent of any causal effect of magnetic fields on miscarriage, non-pregnant women (ie, miscarriage cases) need only move around more. In so doing, they would encounter more unusual magnetic field sources in the home, workplace, and community and thus would have higher peak exposures than their pregnant counterparts. The question of who will have the greater peak exposure translates into who is more likely to happen to be in close contact with a photocopier machine or an industrial motor, to drive near a high-tension power line, or to pass near a sewing machine in operation. The investigators consider this possibility only in passing, but it may offer a parsimonious explanation for their constellation of findings, namely a gradient of increasing risk associated with greater maximum values and rate of change metrics, but little or no association for time-weighted averages.
A similar pathway may be operating in the prospective cohort study, although the behavioral differences in “micromobility” have a different origin. Lack of nausea in early pregnancy is associated with increased risk of miscarriage, with women who do not experience nausea 2–3 times more likely to miscarry than women who do experience nausea. 15–17 The nausea and related symptoms of aversion to smells and foods is thought to reflect elevated estrogens that are favorable predictors of fetal viability. The prospective study by Li et al.4 measures exposure in the etiologically relevant time period, concurrent with the period in early pregnancy over which losses are being identified. However, many pregnancy losses that become apparent at gestational ages after 8–10 weeks actually cease to develop some weeks earlier. 15,18 In fact, one clear reason a lack of nausea predicts subsequent clinically recognizable pregnancy loss is that it reflects fetal demise that has already occurred.
The illustration of caffeine may be useful. The determination of whether caffeine is causally related to miscarriage has proved almost intractable 15 because of the complicating effect of nausea. Nausea simultaneously discourages coffee consumption and indicates a healthy pregnancy. Lack of nausea, a sign of the loss yet to be manifest, permits higher levels of caffeine consumption, so that an association between higher levels of caffeine consumption and subsequent loss may merely reflect the impact of nausea on both factors. Clarification of when the fetal demise occurs, eg, by early ultrasound, would be needed to distinguish between exposure assessment that occurs before and after the loss, regardless of when it becomes clinically apparent.
How does this apply to magnetic fields? Nausea is likely to discourage “micromobility,” just as it discourages caffeine or alcohol consumption. All other things being equal, a woman experiencing nausea will be less likely to move around her home or workplace or community, and therefore less likely to experience the diverse magnetic field sources in those places. As a result, she is less likely to encounter high magnetic field peaks and less likely to have substantial magnetic field variability over time. At its extreme, nausea can keep a woman in bed or at least in her home for much of the day. Thus, nausea (a marker of low risk of miscarriage) will be associated with lower peaks and variability in magnetic fields. Women who have lost or will soon lose their pregnancies are less likely to be nauseated and more likely to be mobile and thus will tend to have higher peaks and more variability in exposure. Thus, nausea may explain the association between magnetic field peaks and miscarriage, just as it potentially explains the association between caffeine and miscarriage.
Li et al.4 take some constructive steps to address this hypothesis. They examined the representativeness of the measurement day and found that restriction to “normal” days tended to enhance rather than diminish the associations, counter to the alternative explanation I offer. However, to a woman who spends every day for some period of pregnancy suffering from nausea, a day in bed may be reported as “normal” so that restriction to such days may not address the behavioral sources of exposure variation attributable to nausea. The contribution from separate locations to the total exposure indices was examined, but more could be done to address “micromobility” in both of the studies. There is no easy solution to the problem of nonconcurrent exposure measurement in a retrospective study, ie, no easy way to measure the typical early pregnancy exposure of women who have already either gone on to late pregnancy or to have a pregnancy loss. In a prospective study, some explicit effort to examine the role of nausea would be feasible, analogous to that applied to the study of caffeine. 16 It seems that such data were collected, but Li et al.4 do not provide information on how nausea may have contributed to the observed associations with maximum magnetic field or lack of association with time-weighted average magnetic fields.
The two studies from Northern California bring new attention to the possibility that magnetic fields may be causally related to miscarriage. The issue deserves further scrutiny and is likely to get it, despite the inherent challenges to the study of both this particular exposure and outcome. Prior to this research, the evidence supporting an etiologic relation between magnetic fields and miscarriage could have been summarized as “extremely limited,” based on studies of workplace exposures and use of electric blankets. With publication of these reports, I believe the evidence in support of a causal association is raised only slightly. These two new studies provide fairly strong evidence against an association with time-weighted average magnetic fields and moderately strong evidence for an association with other indices; both of these findings may be due to an artifact resulting from a laudable effort to integrate behavior and environment.
1. Portier CJ, Wolfe MS (eds). Assessment of Health Effects from Exposure to Power-Line Frequency Electric and Magnetic Fields. NIH Publication No. 98–3981. Research Triangle Park, North Carolina: National Institute of Environmental Health Sciences, 1998.
2. Shaw GM. Adverse human reproductive outcomes and electromagnetic fields: a brief summary of the epidemiologic literature. Bioelectromagnetics 2001; 5: S5–S18.
3. Lee GM, Neutra RR, Hristova L, Yost M, Hiatt RA. A nested case-control study of residential and personal magnetic field measures and miscarriages. Epidemiology 2002; 13: 21–31.
4. Li D-K, Odouli R, Wi S, et al. Personal exposure to magnetic fields during pregnancy and the risk of spontaneous abortion. Epidemiology 2002; 13: 9–20.
5. National Research Council. Possible Health Effects of Exposure to Residential Electric and Magnetic Fields. Committee on the Possible Effects of Electromagnetic Fields on Biological Systems, Board on Radiation Effects Research, Commission on Life Sciences, National Research Council. Washington, DC: National Academy Press, 1997.
6. Wertheimer N, Leeper L. Possible effects of electric blankets and heated waterbeds on fetal development. Bioelectromagnetics 1986; 7: 13–22.
7. Wertheimer N, Leeper L. Fetal loss associated with two seasonal sources of electromagnetic field exposure. Am J Epidemiol 1989; 129: 220–224.
8. Belanger K, Leaderer B, Hellenbrand K, et al. Spontaneous abortion and exposure to electric blankets and heated water beds. Epidemiology 1998; 9: 36–42.
9. Lee GM, Neutra RR, Hristova L, Yost M, Hiatt RA. The use of electric bed heaters and the risk of clinically recognized spontaneous abortion. Epidemiology 2000; 11: 406–415.
10. Delpizzo V. Epidemiological studies of work with video display terminals and adverse pregnancy outcomes (1984–1992). Am J Industr Med 1994: 26: 465–480.
11. Schnorr TM, Grajewski BA, Hornung RW, et al. Video display terminals and the risk of spontaneous abortion. N Engl J Med 1991; 324: 727–733.
12. Lindbohm M-L, Hietanen M, Kyyrönen P, et al. Magnetic fields of video display terminals and spontaneous abortion. Am J Epidemiol 1992; 136: 1041–1051.
13. Juutilainen J, Matilainen P, Saarikoski S, Läärä E, Suonio S. Early pregnancy loss and exposure to 50-Hz magnetic fields. Bioelectromagnetics 1993; 14: 229–236.
14. Savitz DA, Ananth CV. Residential magnetic fields, wire codes, and pregnancy outcome. Bioelectromagnetics 1994; 15: 271–273.
15. Stein Z, Susser M. Miscarriage, caffeine, and the epiphenomena of pregnancy: the causal model. Epidemiology 1991; 2: 163–167.
16. Fenster L, Eskenazi B, Windham GC, Swan SH. Caffeine consumption during pregnancy and spontaneous abortion. Epidemiology 1991; 2: 168–164.
17. Fenster L, Hubbard AE, Swan SH, et al. Caffeinated beverages, decaffeinated coffee, and spontaneous abortion. Epidemiology 1997; 8: 515–523.
18. Cashner KA, Christopher CR, Dysert GA. Spontaneous fetal loss after demonstration of a live fetus in the first trimester. Obstet Gynecol 1987; 70: 827–830.
© 2002 Lippincott Williams & Wilkins, Inc.