This prospective cohort study with preconception enrollment of women upon discontinuation of contraception for purposes of becoming pregnant does not provide conclusive evidence to support the reproductive toxicity of metals at background concentrations. However, we detect a trend towards an increased probability for pregnancy among women with greater blood Mg concentrations, and a trend towards a decreased probability for pregnancy among women with greater blood Zn concentrations. Opposing patterns of effect are observed for various metals, underscoring the importance of considering a mixture of metals rather than relying on a single exposure in keeping with the nature in which humans are exposed. Metals were selected for inclusion in this study based on previously reported associations with female fertility, or on recognized essential roles in normal human physiology. Our study is strengthened by its prospective design, quantification of exposure prior to conception and the hierarchical data structure including the daily capture of information on potential confounding factors such as cigarette smoking and alcohol consumption reported to adversely affect fecundity [reviewed by [37
Of all metals, only Mg and Zn blood concentrations are suggestive of possible significant trends with TTP albeit in opposing directions. A 51.51% increase in the cycle specific conditional probability for pregnancy is suggested for each 3.60 μg/L increase in the covariate adjusted blood Mg concentration, and a 27.67% decrease in the conditional probability of pregnancy is suggested for each 0.54 μg/L increase in the covariate adjusted blood Zn concentration, though with borderline statistical significance (i.e., P=0.062 and P=0.114, respectively). For blood Mg, this trend is consistent with the unadjusted analysis in which concentrations for women with a positive pregnancy test are 1.05-fold (95% CI 1.01-1.10) that of women with no positive pregnancy test. Moreover, the proportion of women with Mg values below the 10th %tile of the sample distribution is more than four-fold higher among women who did not conceive than among women who did (P=0.026). The unadjusted analysis for Zn, in contrast, is not consistent with the multivariable results and indicates no association, suggesting the importance of the multivariable adjustment for additional metals and covariates.
The associations suggested by this study are difficult to interpret, given the dearth of prior investigation into the effects of Mg and Zn on human female fecundity. The results of two previously published preliminary studies are consistent with the suggested positive association between blood Mg concentrations and human pregnancy. In a study of 150 zygotes contributed by 25 IVF patients, investigators reported a statistically significant increase in embryonic cell cleavage (4.2 cells vs. 4.9 cells; P=0.006) following a doubling of the Mg2+
concentration in growth media [38
]. However, the results of this study may be artificially over-powered in that the authors erroneously assumed the zygotes to be independent outcomes despite multiple zygotes per woman in the analysis (i.e., embryonic cell cleavage scores are correlated within woman). Additional but somewhat more tangential evidence is provided by a small case-series of 12 women with a history of infertility or spontaneous pregnancy loss, all of whom conceived following four to six months of dietary Mg and Se supplementation leading to normalization of otherwise abnormal red blood cell Mg and glutathione-S-peroxidase concentrations [39
]. Changes in Mg concentrations as a function of sex-steroid hormones are well-recognized; inverse associations have been reported for serum progesterone and testosterone concentrations [40
]. The latter may be of consequence to our suggested results, though the temporality of the association is reversed and not likely to account for the findings. We identified no prior reports describing the results of studies considering Zn and female fecundity, in spite of the well-recognized association between Zn deficiency, pregnancy complications and adverse birth outcomes [reviewed by [42
]], and a positive association with male fecundity [reviewed by [44
While the exact mechanisms by which metals such as Mg and Zn might exert an effect on female fecundity are unknown, these trace elements are essential for the proper function of metalloenzymes which regulate cell division [45
] and apoptosis [reviewed by [48
]]. In addition, Mg binds to and has a stabilizing effect on cell membranes, proteins and nucleic acids [49
], and protects lipoproteins from reactive oxygen species [50
] which might compromise plasma membrane integrity [51
]. It is tempting to speculate that lower body burdens of Mg lead to a decrease in the activity of enzymes necessary for cell replication, concurrent to increased concentrations of reactive oxygen species which damage cell membranes and compromise a developing conceptus. It is possible that one or both of these mechanisms might underlie the association observed between Mg and fecundity. While speculative, inappropriate Zn concentrations might alter patterns of programmed cell death and, thereby, disrupt embryonic growth and implantation, leading to diminished fecundity.
In addition to the associations suggested for Mg, Zn and fecundity, our results corroborate previously reported associations between PCB exposure, lifestyle factors and TTP in this study population [29
,]. Furthermore, our study results underscore the importance of modeling environmental exposures in the context of lifestyle factors that adversely impact human reproduction, particularly cigarette smoking and alcohol consumption as recently reviewed [37
]. Moreover, the frequency of intercourse during the fertile window demonstrates a positive and statistically significant association with fecundity, as anticipated, underscoring the validity of our multivariable model. However, no association is indicated between age and fecundity during multivariable modeling, in spite of the well-recognized adverse impact of aging on the probability of pregnancy [52
]. We suspect this null result reflects the limited age range of participants who were 25 years of age at least and 35 years of age at most at the start of this study.
Although the toxic nature of As, Cd, Pb, and Ni are well characterized, we detect no suggested association with the probability for a positive pregnancy test in either the bivariate or in the multivariable context. Several prior reports, in contrast, described increased rates of spontaneous pregnancy loss (i.e., 1.5-2.5 fold) in association with chronic exposure to groundwater contaminated by inorganic As [53
], however, with one exception [56
], these studies were conducted in As endemic regions characterized by substantial geogenic groundwater contamination by inorganic species. In contrast, participants in the current study did not reside in a recognized As endemic region and thus measured concentrations are likely to reflect the fairly innocuous organic species [57
], although given the unspeciated nature of our As analysis we are unable to verify this presumption.
Two recent studies report a decreased probability for pregnancy in association with increasing blood Pb concentrations. In a case-control study of 64 infertile women and 83 controls with proven fertility, an increased adjusted odds for blood Pb concentration >25 μg/L was reported (OR 2.94, 95%CI 1.18-7.34) [15
]. However, the mean blood Pb concentration for the control group in that study (31.2 μg/L), was approximately double that for participants in the current study. In a prospective cohort study of couples undergoing IVF, Al Saleh et al. [14
] reported an inverse association for blood Pb and oocyte fertilization (OR 0.38, 95%CI 0.14-0.99), although this association did not extend to pregnancy as assessed by hCG testing (OR 0.55 95% CI 0.23-1.31). Again, these associations were reported for concentrations of blood Pb approximately twice those of participants in the current study (medians 30.1 μg/L vs. 15.0 μg/L, respectively). Moreover, the results of this study did not accommodate the correlated nature of IVF endpoints and may thus be spuriously over-powered. In contrast, no association was reported between unadjusted blood Pb concentration and TTP in a retrospective cohort study of 41 primigravid women with blood Pb concentrations similar to those measured in our study (median 12.0 μg/L) [20
]. An earlier study of 121 women with occupational exposure [17
] reported no association between covariate adjusted blood Pb concentrations ranging from to <103.6-497.3 μg/L and retrospectively reported time to pregnancy (relative risk 0.82, 95% CI 0.52-1.31). We are unaware of any prior studies evaluating the effects of Ni on human female fecundity and are, consequently, unable to interpret our null result in the context of the literature.
Previous investigators have indicated somewhat paradoxical indications of positive associations between adjusted Cd concentrations measured in follicular fluid [12
], and even urine [19
], and oocyte fertilization among women undergoing IVF; however, no such indication is provided by blood Cd concentrations in the current study, although the effect estimate is indeed positive. In the study by Al-Saleh et al. [14
] there was no association suggested for blood Cd concentrations and oocyte fertilization in vitro
(OR 0.45 95%CI 0.07-2.70), or pregnancy (OR 1.72 95%CI 0.74-4.04) at concentrations approximately 2.5-fold lower than those in the current study (mean 0.62 μg/L). We are unaware of any additional studies which assessed blood Cd in association with female fecundity. We detected two outlying Cd values, though we were unable to determine why. These values, equal to 22 μg/L and 478 μg/L blood, are 21 or 477 interquartile ranges above the median value for the Cd distribution, respectively. To assess the influence of these data points we repeated our statistical analysis following re-assignment of these values to the median for our sample distribution equal to 1.0 μg/L blood with no notable changes in results.
Our inability to observe an adverse effect from known toxic metals may stem from the low and relatively comparable metal concentrations across women irrespective of fecundity. With few exceptions, concentrations of metals measured in this study are within the population reference intervals reported for whole blood [58
]. All As (<23.0 μg/L) , Ni (1.0-28.0 μg/L) and Pb (<250 μg/L) concentrations in our cohort are within population reference intervals for whole blood, although concentrations of Cd exceed the population reference interval (3.9 μg/L) for eleven (13.8%) women, by 1.1 to 474.1 μg/L, and concentrations of Se exceed the population reference interval (>234.0 μg/L) for 36 (45.0%) women, by 6.0-96.0 μg/L. To our knowledge, there are no population reference intervals available for Mg and Zn measured in whole blood. The majority of Mg [45
] and Zn [59
] ions are sequestered within cells, with concentrations being tightly regulated by a variety of passive and active transport mechanisms. There is a substantial degree of inter-compartment variation in Mg and Zn concentrations, which precludes simple reference interval extrapolation from one matrix (i.e., plasma) to another (i.e., whole blood). Recent estimates do, however, suggest that a large proportion of American women are in fact Mg deficient, a consequence of modern food processing, preservation and packaging technologies [45
The results of the current study are limited by several factors, foremost of which is the limited sample size available for analysis. We conducted post hoc
power analyses to determine cohort size of sufficient statistical power to assess Mg and Zn relative to pregnancy at background concentrations for future research initiatives and to help interpret our results. To this end, we estimated that a cohort size comprising 91 and 155 women are sufficient for assessing the Mg and Zn related effects, respectively [60
]. Given our limited sample, we were unable to assess outcomes such as pregnancy loss that are woman level and not cycle level events. An additional and important limitation is our consideration only of female exposure and covariate data in this study. Ignoring male factors may have introduced exposure misclassification, as recent evidence indicates that male metals exposure may be even more relevant to conception than female exposure [19
], as well as unmeasured confounding. Our exposure assessment comprises the simultaneous consideration of several metals, an approach which is more likely consistent with the nature by which human exposure occurs than is represented by studies considering exposure to metals in isolation. Our approach of specifying a unique regression variable for each metal in the absence of observed colinearity allowed us to simultaneously assess these exposures in relation to TTP. However, our use of single preconception measure of blood metals concentrations may have introduced exposure misclassification for non-persistent metals, those other than Cd or Pb. Whereas the results of this study are only suggestive and useful for hypothesis generation, our detection of potential effects for Mg and Zn on human fecundity is of importance.
The results of this prospective cohort study with preconception enrollment of women and longitudinal capture of covariates relevant for TTP are suggestive of a possible relation with Mg and Zn, although in opposing directions with Mg shortening and Zn lengthening TTP. These results represent the first report of associations between Mg, Zn and human female fecundity in the scientific literature, to the best of our knowledge. The imprecision of regression coefficients as reflected by the wide confidence intervals underscores the importance of cautious interpretation. Still, we believe these results merit a more rigorous consideration of the potential impact of Mg and Zn exposures on female fecundity using a larger sample size with sufficient statistical power.