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.
To the Editor:
In considering alternative explanations for our findings of an association between the exposure to maximum magnetic fields (MFs) during pregnancy and the risk of miscarriage, David Savitz in his commentary 1 hypothesizes that factors other than maximum MFs might explain the associations. With regard to the prospective study of Li et al2 and the prospective substudy of Lee et al, 3 he wonders whether the observed association might have arisen from a premonitory symptom of impending miscarriage or the lack of such a symptom that in turn could influence exposure behavior. He gives the example of nausea that is less common among those destined to have a miscarriage. Under this hypothesis, nausea slows down the women destined to be healthy while those destined to miscarry pass by sources of magnetic fields more frequently.
The authors of Li et al2 can evaluate this hypothesis directly, as we collected information on nausea and related symptoms such as vomiting in the original in-person interview. As expected, nausea and vomiting themselves were associated with a reduced risk of miscarriage [hazard ratio (HR) = 0.3 (95% confidence interval = 0.2–0.4) and HR = 0.3 (95% confidence interval = 0.2–0.4), respectively]. However, the frequency of nausea or vomiting was very similar for women exposed to maximum MF ≥16 mG (exposed) and to MF <16 mG (unexposed), regardless of whether these symptoms were measured since the last menstrual period or during the 7 days before interview/measurement (probably the more relevant period for influencing the exposure measurements) (Table 1). After adding the nausea and vomiting variables (coded as in Table 1) to the Cox model, the HR for the association of maximum MF with risk of miscarriage remained essentially the same, if not strengthened, compared with the HR without the adjustment for nausea or vomiting (Table 2). The results were similar when the analysis was restricted to women whose measurement was conducted during a typical day of pregnancy. Adjustment for nausea or vomiting also did not change the association between the time-weighted average MF matrix and the risk of miscarriage (Table 2).
These results do not support the suggestion by Savitz 1 that nausea or vomiting could influence maximum MF exposure or that adjusting for them might modify the association between maximum MF exposure and the risk of miscarriage.
Along the same line of thinking, Savitz 1 also wondered whether the measurements of exposure in late pregnancy in the nested case-control study of Lee et al3 might have produced a bias because control women, who were now 30 weeks pregnant and perhaps less mobile, were less likely to encounter high MF exposure than cases whose miscarriage ended in earlier gestations. The prospective substudy embedded in the case-control study was supposed to evaluate this kind of problem; 148 control women were measured at both around 12 weeks and around 30 weeks of gestation. The mean values of time-weighted average, maximum MF, and rate-of-change metric measured in early gestation (1.1 mG, 34 mG, and 0.9 mG per 10 seconds, respectively) were similar to those in late gestation (1.2 mG, 28 mG, and 0.7 mG per 10 seconds). In addition, Lee et al3 have noted that the results of the large nested case-control study had associations that were in the same direction as the small prospective substudy, which did not have the late measurement “problem.”
In summary, our data do not support the argument that factors such as nausea or vomiting explain our findings. Of course, we cannot address the general hypothesis that some premonitory symptoms other than nausea or vomiting could falsely produce an association. Unless this generic hypothesis can be specified in a way that could be tested, it could always be raised to question the results of our studies or any conceivable future studies. We believe that our findings need to be replicated, and we welcome ideas for testable alternative hypotheses that can be built into future studies.
Finally, a factual correction: Savitz states in his commentary that “The results of the case-control study by Lee et al. motivated the development of a prospective cohort study by Li et al. to determine whether the findings of a positive association with indices of peak magnetic fields and variability could be replicated, and they were.” Li et al.2 were the first to reveal an association between peak magnetic field exposure and miscarriage. In an earlier preliminary analysis of the prospective arm of their study, Lee, Neutra et al. (unpublished at the time and now published as Lee et al.3) had suggested a possible effect of TWAs above 2 mG. It was this association that motivated the funding of the Li et al. 2 study to specficially test the hypothesis of an association between TWA exposure and the risk of miscarriage. Lee, Neutra et al, 3 had looked at the effect of the 95th percentile of peak fields and not seen an association. After looking at TWA Li et al.2 innovated by also looking at dose-response of the full range of peak magnetic field. This revealed an association between peak magnetic field exposure and miscarriage. In fact, it was the finding by Li et al.2 of the effect of peak magnetic fields that prompted Lee et al. 3 to reanalyze their data on peak magnetic fields and subsequently confirmed the findings by Li et al.2
Raymond Richard Neutra
1. Savitz DA. Magnetic fields and miscarriage. Epidemiology 2002; 13: 238.
2. Li D-K, Odouli R, Wi S, et al
. A population-based prospective cohort study of personal exposure to magnetic fields during pregnancy and the risk of miscarriage. Epidemiology 2002; 13: 9–20.
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.