Association of personal exposure to power-frequency magnetic fields with pregnancy outcomes among women seeking fertility treatment in a longitudinal cohort study

Abstract

Objective:

To assess for the first time the potential relationships of personal exposure to magnetic fields (MF) with pregnancy outcomes among a cohort of women from a fertility clinic, addressing, through study design, some of the primary limitations of previous studies on this topic.

Design:

Longitudinal preconception prospective cohort.

Setting:

Fertility center.

Patient(s):

Our analysis included 119 women recruited from 2012 to 2018, who underwent in vitro fertilization (IVF) (n = 163 cycles) and/or intrauterine insemination (IUI) (n = 123 cycles).

Intervention(s):

Women wore personal exposure monitors continuously for up to three consecutive 24-hour time periods separated by several weeks.

Main Outcome Measure(s):

Implantation, clinical pregnancy, live birth, and pregnancy loss.

Result(s):

The median and maximum of the overall daily mean (daily peak) MF exposure levels were 1.10 mG (2.14 mG) and 15.54 mG (58.73 mG), respectively. MF exposure metrics were highest among women who changed environments four or more times per day. Overall, no statistically significant associations between MF exposure metrics and fertility treatment or pregnancy outcomes were observed in crude or adjusted models. Effect estimates, both positive and negative, varied by outcome and the exposure metric, including the way in which exposure was modeled.

Conclusion(s):

Personal MF exposures were not associated with fertility treatment outcomes or pregnancy outcomes. Despite its limited size, strengths of the study include a longitudinal repeated-measures design, the collection of personal MF exposure data across multiple days, and carefully documented outcome and covariate information among a potentially susceptible study population.

The majority of all conceptions end before 20 weeks of gestation, mostly due to failed uterine implantation (1, 2). In addition, approximately one in every three conceptions that do implant successfully still fail, often before a woman knows she is pregnant (1). In the United States, the rate of self-reported pregnancy loss continues to rise with ~20% of medically recognized pregnancies failing to reach full term. The risk of early pregnancy loss (<12 weeks of gestation) has also increased 2% annually from 1990 to 2011 (3). Although the determinants of pregnancy failures remain largely unknown, this adverse reproductive health condition is likely multifactorial, arising as a complex interplay of environmental, genetic, and lifestyle factors evident at the population level (4).

Since the late 1970s, exposure to power-frequency (50 or 60 Hz) magnetic fields (MF), a form of nonionizing radiation encountered wherever alternating current electricity is produced, distributed, and used, has been studied as a potential risk factor for miscarriage (pregnancy loss at <20 weeks of gestation) and other adverse reproductive health outcomes (5). This research priority arose from reports of miscarriage and birth defect clusters among video display terminal (VDT) operators in the U.S. and Canada (6), motivating a 20-year effort of epidemiology studies focusing on exposure to MF from VDT in addition to other sources in and around the home, such as electric blankets, heated water beds, and power lines. However, the overall evidence from those studies was inadequate to conclude that exposure to MF was a risk to female reproductive health, because they produced conflicting results and many were characterized by critical study design limitations, such as questionable accuracy of outcome measures, use of surrogate measures of personal exposure, and/or temporal ambiguity (5).

The debate on this topic was revived in the early 2000s following the publication of two epidemiology studies conducted in pregnant women enrolled in the California Kaiser Permanente Medical Care Program (7, 8). Those studies were particularly noteworthy because they were the first of their kind to characterize MF exposures with the use of personal exposure monitors, specifically over a single 24-hour period per woman. This was an improvement on historical approaches that primarily relied on exposure surrogates, such as residential wire code classification, self-reported use of electric devices, and spot measurements in residences and workplaces (5). Nevertheless, it should be noted that the collection of personal exposure data over a single 24-hour period is still likely prone to measurement error, which can be appreciable, especially for peak exposure metrics given that the period over which exposure data is collected is very narrow relative to the time window of miscarriage risk (5). Overall, these two epidemiology studies reported an increased risk of miscarriage associated with the maximum MF exposure level compared with null associations with average exposure level, suggesting a threshold of ~16 mG and a greater risk among “susceptible” women with subfertility and those who experienced earlier miscarriages (<10 weeks of gestation).

Following the publication of these studies, there was debate in the scientific literature because it was hypothesized that the findings may be confounded by differences in mobility patterns in women with healthy pregnancies (i.e., more likely to experience nausea) compared with those who miscarried (i.e., less likely to experience nausea) (9, 10); the results of several subsequent exposure science analyses supported the hypothesis of residual confounding due to unmeasured physical activity (11–13).

We set out to further the state of the science by conducting an epidemiologic analysis of personal MF exposure and pregnancy outcomes using data from a well established prospective cohort of women undergoing assisted reproductive technologies. The present study design addressed some of the primary shortcomings of previous investigations through the accurate collection of outcome data, measurement of personal exposure data across multiple days, consideration of potential confounding due to differential physical activity, and being conducted among a potentially susceptible population.

METHODS

Study Population

Study participants were a subset of women from the Environment and Reproductive Health (EARTH) study, an established longitudinal prospective preconception cohort study of the environmental, dietary, and lifestyle impacts on reproductive health (14). Women (18–46 years of age) were recruited from Massachusetts General Hospital (MGH) Fertility Center. Approximately 60% of women who were approached enrolled in the parent study (15). Questionnaires were administered to collect demographic information (i.e., age, race/ethnicity, education level). The present analysis includes a subset of 119 women who were recruited from 2012 to 2018, underwent in vitro fertilization (IVF) (n = 163 cycles) and/or intrauterine insemination (IUI) (n = 123 cycles) and consented to wearing a personal MF exposure monitor. Research protocols were approved by the Ethics and Research Committees of MGH, Harvard T.H. Chan School of Public Health, and University of Michigan. The study was described in detail to all participants and informed consent was obtained from each of them.

Personal Magnetic Field Exposure Monitors

Personal MF exposure monitor specifications and details have been previously described (13, 16). Briefly, women were asked to wear an Emdex Lite (Enertech), a small battery-powered personal MF exposure monitor worn at hip level that is light-weight (170 g) and approximately the size of a mobile phone (12 × 6 × 2.5 cm). Monitors were calibrated to measure MF exposure levels in milligauss (1 mG = 0.1 μT) from 40 to 1,000 Hz (including the power frequency of 60 Hz). The Emdex Lite measures the MF level in the x (horizontal), y (vertical), and z (lateral) planes. The resultant combined effect of all three planes was used to quantify MF exposure for the present analysis. Women wore the monitors continuously while undergoing fertility treatment for up to three consecutive 24-hour time periods separated by several weeks, except for times when they were sleeping, bathing, showering, or swimming.

Time-Activity Diary

Women were asked to complete a time-activity diary on each of the days on which they wore a personal MF exposure monitor (13). This diary, which was modeled after the one used in the U.S. National Children’s Study, documented daily activities by location (inside or outside: home, work or school, somewhere else, in transit) and nausea at 30-minute intervals.

Clinical Protocols and Pregnancy Measures

During each cycle, clinical data were abstracted from the participant’s electronic health record. Clinical protocols and cycle measures have been previously detailed (15, 17). Infertility diagnosis was established by an MGH physician in accordance with the Society of Assisted Reproductive Technology (SART) (18). On infertility evaluation including infertility diagnosis and other clinical factors, one of three ovarian treatment protocols was selected: 1) luteal-phase GnRH agonist; 2) follicular-phase GnRH agonist or “flare” stimulation; or 3) GnRH antagonist (19). Throughout gonadotropin stimulation and up to 2 days before oocyte retrieval, serum E₂, follicle size and count, and endometrial thickness were monitored for each participant (15). Fertilization occurred via IVF or intracytoplasmic sperm injection (ICSI) and was confirmed 17–20 hours later. Fertilization was determined by the presence of a cytoplasmic halo and two pronuclei (20, 21). Clinical outcomes were assessed for participants who continued with embryo transfer. Successful implantation for a given transfer was confirmed when serum β-hCG levels were >6 mIU/mL, ~17 days after oocyte retrieval (15). Clinical pregnancy was defined as the presence of an intrauterine pregnancy via ultrasound (≥6 weeks of gestation) along with elevated levels of β-hCG. Live birth was defined as the birth of a neonate at or after 24 weeks of gestation. Total pregnancy loss was defined as any loss of a pregnancy before 20 weeks of gestation, including biochemical losses.

Main Outcomes

Pregnancy outcomes in this analysis included implantation, clinical pregnancy, pregnancy loss, and live birth. Each outcome was dichotomized as either successful (yes) or not (no).

Statistical Analysis

Demographic and reproductive characteristics and personal MF exposure monitor daily wear times for women were calculated with the use of medians, interquartile ranges (IQRs), frequencies, and percentages as appropriate. Geometric means (GMs) and associated 95% confidence intervals (CIs) and selected percentiles of mean (arithmetic) and peak (measurements in the 90th percentile) MF measurements were calculated for total daily monitor wear time data and stratified by environment (inside or outside: home, work or school, somewhere else, in transit) per participant. GMs and associated 95% CIs were also calculated for mean, median, and peak exposure metrics stratified by number of changes in environment (≤3, 4–9, and ≥10) per day.

MF exposure metrics (mean, median, and peak) were calculated per woman (i.e., overall across all sampling days; n = 119) and per day (n = 538). Spearman correlation coefficients were calculated for all exposures to evaluate correlations among the various MF metrics. Mean, median, and peak metrics were right skewed and transformed by the natural logarithm (ln) for use in subsequent analyses. Confounding was evaluated with the use of prior knowledge and descriptive statistics and included age, body mass index (BMI; kg/m²), race (white/other), education (high school/some college, college graduate, graduate degree), nausea on sampling days (yes/no), and daily number of changes in environment (i.e., an indicator of overall physical activity). Reproductive characteristics also considered were previous pregnancy, day 3 FSH levels (IUI/L), initial infertility diagnosis (female factor, male factor, or unexplained), previous intrauterine insemination (IUI) (yes/no), previous IVF (yes/no), IVF cycle (yes/no), treatment protocol (antagonist, flare, or luteal-phase agonist), E₂ trigger levels (pmol/L), endometrial thickness mm, ICSI (yes/no), and measurement-day nausea (yes/no). Final covariates were included if they were associated with exposure in our cohort, associated with exposure based on previous studies, and known to be a predictor of IVF outcomes (11, 13, 22). Final models were adjusted for age, BMI, race (white/other), endometrial thickness, IVF (yes/no), and number of changes in environment per day. Missing values (n = 11) for endometrial thickness were replaced with the median (9). Associations of MF exposure metrics (per cycle and per day) with successful (yes/no) clinical outcome (implantation, clinical pregnancy, pregnancy loss, and live birth) were evaluated with the use of cluster weighted generalized estimating equation (CWGEE) log-binomial models, where the weight was calculated as the inverse of the cluster size (total no. of cycles) (23, 24). CWGEEs have some advantages when accounting for multiple cycles per women (i.e., nonignorable cluster size), because they may provide more precise point estimates compared with other common statistical approaches for dichotomous IVF outcomes (24). Metrics were first treated as ln-transformed continuous variables, and results represent an increase or decrease in the probability of a successful clinical outcomes with a 1-mG increase in MF exposure. We also categorized MF exposure into tertiles where P values for trend (P-trend) were calculated with the use of the median ln-transformed MF exposure metric for each tertile.

We conducted several sensitivity analyses. We first considered stratifying the models by number of changes in environment (<4 and ≥4) owing to higher metric readings among women who changed environments more often. Previous analyses have also evaluated associations with certain thresholds (time-weighted average [TWA] ≥3 mG and maximum ≥16 mG) of exposure (8), but the limited number of observations at these thresholds (≥3 mG: 33 cycles, 90 days; ≥16 mG: 1 cycle, 2 days) was too low to be considered in the present analysis. Instead, we reran our models with women who had peak MF ≥2 mG, which has also been previously used as a cutoff for the TWA (7). We also explored the possibility of nausea confounding the relationship between MF exposure and miscarriage by modeling the associations of reported measurement-day nausea (yes/no) with pregnancy outcomes (10). Analyses were performed with the use of SAS 9.4 (SAS Institute) and R version 3.3.5. P values of <.05 were considered to be statistically significant.

RESULTS

Demographic and reproductive characteristics of this cohort (n = 119 women) are detailed in Table 1 and are similar to previous subsets from the EARTH cohort. Most women were white (75%), in their mid-thirties (median 34 years), and had a normal BMI (median 23 kg/m²). Few women (21%) reported ever smoking and more than half (63%) held a graduate degree. Half of the initial infertility diagnoses were unexplained (54%), followed by male factor (27%) and female factor (19%). More of the cycles were IVF cycles than IUI cycles (57% vs. 43%). Median day 3 FSH level was 7.0 IU/L (IQR 6.0–8.0). A substantial number of women (69%) underwent luteal-phase agonist protocol compared with cryopreservation or using donor eggs (13%), flare (9%), or antagonist (8%) protocols. The median peak E₂ level was 1,472 pmol/L and endometrial thickness was 9 mm. Few treatment cycles were ICSI (26%).

TABLE 1.

Demographic and reproductive characteristics of 119 women (163 in vitro fertilization cycles and 123 intrauterine insemination (IUI) cycles) from the EARTH cohort.

Characteristic	Median or n	IQR or %
Demographic
Age, y	34	31–37
Race/ethnicityds
Black/Asian/other	29	25
White/Caucasian	90	75
Body mass index, kg/m2	23	21–26
Ever smoker	25	21
Education
High school/some college	7	5
College graduate	37	32
Graduate degree	75	63
Reproductive characteristics
Day 3 FSH levels, IU/L	7	6–8
Initial infertility diagnosis
Female factor	23	19
Male factor	32	27
Unexplained	64	54
Treatment protocol
Antagonist	24	8
Flare	26	9
Luteal phase agonist	196	69
Cryo/donor	40	13
IUI	123	43
E2 trigger levels, pmol/La	1472	731–1,851
Endometrial thickness, mmb	9	8–11
ICSI cycles	73	26

Note: Cryo = cryopreservation; ICSI = intracytoplasmic sperm injection; IQR = interquartile range.

^an = 29 missing

^bn = 11 missing.

The majority of women wore their personal exposure monitors for ~10 hours (median) each day (IQR 2.4–14.0; Table 2). Most time was spent either inside the home (median 7.6 h) or inside work or school (median 6.3 hours), with similar mean MF levels (GM 0.81 mG and 0.79 mG, respectively). Less time was spent outside at home (median 0.5 h), work or school (median 1.0 h), or somewhere else (median 2.0 h). Peak exposures were similar for outside the home (GM 0.93 mG) and work or school (GM 0.88 mG), yet considerably higher for times categorized as somewhere else (GM 1.40 mG). Most women spent ~3.6 hours (median) in transit each day, with peak levels considerably higher (GM 3.28 mG) than peak exposures inside the home (GM 1.32 mG) and work or school (GM 1.35 mG). All MF exposure metrics (mean, median, and peak) per woman and per day were significantly and positively correlated with each other (Supplemental Fig. 1 [available online at www.fertstert.org]). Median MF levels were strongly correlated (r = 0.72), followed by means (r = 0.60) and peaks (r = 0.45). Women who had fewer changes in environment per day (≤3) had lower average personal exposure monitor readings (GM 0.70 mG) compared with those with more than three changes (–−9: GM 1.02 mG; ≥10: GM 1.03 mG; Fig. 1). However, peaks were higher for women with ≥10 changes (GM 1.98 mG) compared with those with 4–9 changes (GM 1.83 mG).

TABLE 2.

Distribution of personal magnetic field exposure metrics (mG) and daily magnetic field exposure monitor wear time (h) by environment of women from the EARTH cohort.

Environment	Daily wear time, median (IQR)	Metric	GM	95% CI	Percentile
					10th	25th	50th	90th	Maximum
Totala	10.4 (2.4–14.0)	Mean	1.07	0.95–1.19	0.51	0.70	1.10	2.05	15.54
		Peak	1.99	1.76–2.24	0.87	1.34	2.14	4.48	58.73
Inside at homea	7.6 (1.4–13.4)	Mean	0.81	0.70–0.93	0.30	0.46	0.80	2.19	7.83
		Peak	1.32	1.12–1.56	0.45	0.68	1.24	4.18	58.73
Inside at work or schoolb	6.3 (0.3–8.5)	Mean	0.79	0.66–0.94	0.32	0.46	0.67	2.07	15.54
		Peak	1.35	1.11–1.65	0.57	0.72	1.23	3.81	54.01
Inside somewhere elsec	2.0 (1.0–5.7)	Mean	1.09	0.95–1.25	0.43	0.79	1.14	2.31	4.75
		Peak	2.10	1.79–2.46	0.75	1.54	2.18	5.04	9.40
Outside at homed	0.5 (0.5–2.5)	Mean	0.93	0.75–1.15	0.31	0.55	0.91	2.4	3.58
		Peak	1.63	1.28–2.08	0.53	0.84	1.81	4.71	8.18
Outside at work or schoole	1.0 (0.5–3.5)	Mean	0.88	0.71–1.09	0.41	0.60	0.82	2.07	2.70
		Peak	1.69	1.34–2.12	0.79	1.19	1.54	4.19	4.82
Outside somewhere elsef	2.0 (0.5–6.5)	Mean	1.40	1.20–1.63	0.63	0.92	1.42	2.98	11.75
		Peak	3.11	2.66–3.63	1.36	2.06	3.06	7.51	32.13
In transitg	3.6 (1.0–9.0)	Mean	1.55	1.40–1.73	0.79	1.12	1.49	2.81	7.71
		Peak	3.28	2.92–3.70	1.71	2.20	3.16	6.63	21.00

CI = confidence interval; GM = geometric mean; Mean = arithmetic mean; Peak = 90th percentile.

^a600 sampling days from 119 women, sleeping and non-wear readings omitted.

^b187 sampling days from 76 women.

^c200 sampling days from 92 women.

^d83 sampling days from 50 women.

^e44 sampling days from 29 women.

^f192 sampling days from 29 women.

^f300 sampling days from 102 women.

FIGURE 1.

Geometric means (95% confidence intervals [CIs]) for magnetic field exposure metrics (mG; mean, median, and peak) by daily number of changes in environment among 119 women from the EARTH cohort. Mean = arithmetic; peak = 90th percentile. ^a440 sampling days from 119 women. ^b102 sampling days from 97 women. ^c57 sampling days from 43 women.

Associations with ln-transformed continuous MF exposure metrics per cycle with clinical outcomes are presented in Figure 2 and Supplemental Table 1 (Supplemental Tables 1–7 are available online at www.fertstert.org). In crude models, none of the effect estimates were statistically significant, but largely indicated an increase in the probability of successful implantation, clinical pregnancy, and live birth in relation to an increase in MF exposure metrics per cycle. In adjusted models, an increase in peak MF exposure per cycle was associated with decreased estimated probabilities of successful implantation (relative risk [RR] 0.81, 95% CI 0.60–1.10), clinical pregnancy (RR 0.80, 95% CI 0.56–1.14), and live birth (RR 0.87, 95% CI 0.59–1.30); similar findings were observed when relationships were explored in relation to the mean and median MF levels per cycle. Daily peak MF exposure was positively associated with successful implantation (RR 1.06, 95% CI 0.86–1.31), clinical pregnancy (RR 1.06, 95% CI 0.86–1.30), and live birth (RR 1.09, 95% CI 0.89–1.33; Supplemental Table 2); similar findings were observed for the daily mean and median MF levels. No associations were observed when MF exposure metrics were modeled as tertiles (Supplemental Table 3). Inverse associations with daily MF exposure metrics were observed for total pregnancy loss in crude and adjusted models (Supplemental Table 4).

FIGURE 2.

Relative risks (RR) (95% confidence intervals [CIs]) for the association of implantation, clinical pregnancy, and live birth with a 1-mG increase in magnetic field exposure metrics (median, mean, and peak) per cycle among 119 women (n = 286 cycles) from the EARTH cohort. Models adjusted for age, body mass index (BMI), race (white/other), endometrial thickness, in vitro fertilization (yes/no), and number of changes in environment per day. Results are presented on a logarithmic scale.

When stratified by number of environment changes per day, we found overall no statistically significant associations between daily MF exposure metrics with implantation, clinical pregnancy, and live birth in both crude and adjusted models (Supplemental Table 5). In particular, among women with <4 environment changes per day, inverse associations were consistently observed and the opposite for women with ≥4 environment changes per day. In models including only those women with peak MF levels ≥2 mG per day (n = 186 days) or per cycle (n = 69 IVF/IUI cycles), an increase in MF exposure was largely associated with a decrease in the estimated probability of implantation, clinical pregnancy, and live birth (Supplemental Table 6). The magnitude of the effect estimates for analyses stratified on physical activity and among this subset of women with peak MF levels ≥2 mG, were in the range of those that were observed in our analyses described above.

DISCUSSION

In this study, repeated measures of personal exposure to MF and pregnancy outcomes (implantation, clinical pregnancy, and live birth) were evaluated in 119 women who underwent IVF and/or IUI at an academic fertility clinic in Boston, Massachusetts. Effect estimates were heterogeneous, varying by outcome and exposure metric, including the manner in which exposure was modeled, and none approached statistical significance. Overall, there was no evidence for a relationship between personal exposure to MF and the fertility treatment or pregnancy outcomes.

The present analysis was, in part, motivated by the findings of two relevant miscarriage epidemiology studies among pregnant women living in northern California, which suggested that peak MF exposure is the biologically relevant exposure metric for this outcome and is perhaps threshold dependent (7, 8). These findings generated important commentary resulting in a hypothesis that physical activity level explains the observed associations, because increased physical activity level may result in greater chance of encountering an elevated peak MF exposure as well as be a sign of low nausea and perhaps a less healthy pregnancy (10). We inquired if participants were experiencing nausea on the days for which we collected activity diaries and MF exposure was measured. Physical activity, measured as number of changes in environment, was not associated with pregnancy outcomes (data not presented). Also, when we modeled measurement-day nausea (yes/no) with pregnancy outcomes, we did not observe any notable changes in associations, and thus differential physical activity did not appear to be confounding the relationship between MF exposure and pregnancy outcomes in our sample (Supplemental Table 7). While peak MF exposure levels were higher among women who changed environments ≥4 times per day, the effect estimates of the associations for all pregnancy outcomes were negative only among women with fewer (<4) daily environment changes. These results are contrary to the Savitz hypothesis (10), but the relationships observed in our analyses stratified on physical activity should be interpreted with caution because none of the effect estimates approached statistical significance.

A more recent analysis (n = 913) from the Kaiser Permanente research team reported an almost threefold (hazard ratio [HR] 2.72, 95% CI 1.42–5.19) increase in the risk of miscarriage with a 24-hour 99th-percentile MF exposure ≥2.5 mG (the minimum threshold for the highest three exposure quartiles) on a “typical day” during pregnancy after adjusting for maternal age, race, education, smoking during pregnancy, and previous miscarriage, compared with the lowest quartiles (22). However, there was a clear inverse dose-response relationship when exposure was separately modeled as quartiles in relation to miscarriage risk (Q2 HR 3.29, 95% CI 1.59–6.79; Q3 HR 3.01, 95% CI 1.48–6.12; Q4 HR 2.02, 95% CI 0.95–4.28), suggesting that miscarriage risk reported at the 2.5 mG threshold may be driven by factors other than MF exposure. Similar to what has been demonstrated for the maximum, the 99th percentile also has poor reproducibility over time, which likely results in appreciable exposure misclassification (13, 25). It is also noteworthy that the authors indicated that their results did not change with additionally adjusting their models for self-reported nausea and vomiting at any time during pregnancy. However, as previously pointed out by Savitz et al., analysis of pregnancy-related morning sickness symptoms on the exposure measurement day (as was performed in our study) would be required to accurately account for this potential confounder (11).

Our findings are consistent with several literature reviews published by authoritative scientific bodies, which have concluded that the evidence does not support maternal exposure to MF as a risk factor for adverse female reproductive outcomes (26–29). Although there is some experimental evidence in mammalian and nonmammalian species that exposure to MFs may disrupt early development and fertility (27, 30), no apparent biophysical mechanism is known that would trigger biological effects from low-level MF exposure, such as is encountered in the general population, and as a result lead to impacts on female reproduction. Consequently, it may be worthwhile to explore whether some of the positive findings reported in the literature may be due to residual confounding from co-exposure to an environmental agent that, similarly to MF, results in ubiquitous exposure in the general population but whose potential as a female reproductive toxicant is much more biologically plausible and supported by the experimental and epidemiologic literature. For example, we have previously reported increased odds of early IVF failure with exposure to ambient air pollution (31) and associations with exposure to organophosphate esters and reproductive end points later in gestation (21).

The present study has several limitations and strengths. The relatively small sample size had reduced power to detect statistically significant associations. However, the women in the study wore personal monitors for up to three separate 24-hour periods, thus reducing exposure measurement error associated with the use of repeated within-woman exposure measurements. The original study design also included personal activity monitors that, unfortunately, did not provide reliable data to more quantitatively evaluate confounding by physical activity. However, we attempted to account for this deficiency by using time-activity diary information, which previously has been shown to be associated with peak MF exposure in an earlier subset of the women (13) and by controlling for exposure measurement-day nausea in our models. Our cohort of sub-fertile women may not be generalizable to the general population, but such cohorts are a potentially “susceptible” population, which, if true, increases the probability of detecting an MF exposure-related association relative to the general population. Our study design expands on other preconception cohorts by capturing early developmental and clinical end points otherwise not included in studies of the general population. To the best of our knowledge, this is the first study to evaluate the relationship between personal MF exposure and early pregnancy outcomes such as implantation, biochemical pregnancy, and preclinical pregnancy loss. The prospective design of this established cohort allowed measurement of exposure to precede the outcome, reducing bias from temporal ambiguity that may arise from retrospective studies. Exposure was also characterized with the use of personal exposure monitors, which, relative to surrogate measures, provide more valid data because personal exposure monitors can capture variability in exposure over space and time (10, 25).

CONCLUSION

The present study found that within the range of MF exposures experienced by the women in our fertility clinic cohort, there was no evidence for a relationship with fertility treatment or pregnancy outcomes in unadjusted and adjusted models. Given the continued interest in this topic, future studies with larger cohorts and more detailed exposure metrics are still warranted to explore the potential relationship between MF exposure and reproductive health. Ideally, to minimize important sources of potential bias, these studies should be longitudinal and collect repeated measures of personal exposure and carefully document outcome and covariate data, including those relevant to physical activity. An important gap that should be considered is the male partner’s exposure, which might play an important role in the etiology of the reproductive health outcome of interest.