In utero exposure to mercury and childhood overweight or obesity: counteracting effect of maternal folate status

The study protocol was approved by the Institutional Review Boards of Boston Medical Center and Johns Hopkins Bloomberg School of Public Health. Written informed consent was obtained from each study child’s biological mother or the study child. Participants of this study came from the Boston Birth Cohort (BBC), who were enrolled at birth and followed prospectively thereafter at the Boston Medical Center. Detailed information on participant enrollment has been described previously [21]. Briefly, the mother-infant pairs were enrolled 24 to 72 h after delivery. After obtaining written informed consent, trained research staff interviewed mothers using a standardized questionnaire. Maternal blood was drawn for measurement of Hg in red blood cell (RBC) and plasma folate concentrations. Pertinent clinical information was obtained by a review of maternal and infant medical records, including prenatal ultrasonographic reports, laboratory reports, and information on pregnancy complications, and birth outcomes. As of July 2018, 3163 children who enrolled in the BBC and received primary care at the Boston Medical Center were followed. Of these children, 1551 mothers had measurements taken for Hg concentration in the RBCs. There were 109 children excluded due to a lack of body mass index (BMI) data between ages 2–15 years. Finally, this study included 1442 mother-child pairs who had complete data (Additional file 1: Figure S1). The characteristics were similar between the included and excluded samples, except that more Black women and less preterm births were in the included samples (Additional file 1: Table S1).

Maternal RBC-Hg concentrations were measured using inductively coupled plasma mass spectrometry (ICP-MS) on an Agilent 8900 QQQ (Agilent Technologies Inc., Santa Clara, CA) at the New Jersey Department of Health, Environmental and Chemical Laboratory Services, Metals Laboratory, Trenton, New Jersey. This laboratory is certified by the US Centers for Disease Control and Prevention, Clinical Laboratory Improvement Amendment, and Clinical Laboratory Improvement Services for analysis of clinical samples. The analytical procedures followed standard protocols for quality control and assurance. The quality control includes a calibration curve comprised of a minimum of five points, continuous calibration verification, continuous calibration blank verification, duplicate samples, laboratory fortified matrix samples, blind duplicates, and a minimum of four second source quality control samples for every run. Calibration stock solution (catalog# SM-2107-042-50, High-Purity Standards, SC) and second source quality control samples (Blood Metals Quality Control Specimens, Lot 018, Wisconsin State Laboratory of Hygiene, WI) were used as National Institute of Standards and Technology (NIST)-traceable certified reference materials to calibrate and monitor the accuracy and interday and intraday repeatability of the analysis. The second source samples must recover within the established recovery range set by the manufacturer or within a tighter range established by the laboratory. The coefficient of variation (CV) was < 5.0%. The limit of detection (LOD) was 0.28 μg/L. If the concentration was below the LOD, the value was assigned to be the value of the LOD divided by the square root of two.

Maternal plasma folate concentrations were measured by a commercial laboratory via chemiluminescent immunoassay using a MAGLUMI 2000 Analyzer (Snibe Co., Ltd., Shenzhen, China). Each run included a calibration curve with five points and three internal quality controls. If the value of folate was above the maximum LOD, plasma samples were diluted at a ratio of 1:2 and then re-run. The inter-assay CV was < 4% [22]. In our previous study [11], we found that maternal plasma folate level ≥ 20.4 nmol/L was associated with a lower risk of child OWO and mitigated the inter-generational link of obesity. As such, we defined adequate maternal folate as maternal plasma folate ≥ 20.4 nmol/L.

Maternal variables, including age at time of delivery, race/ethnicity, parity, education, pre-pregnancy height and weight, cigarette smoking, and fish consumption, were based on maternal questionnaire interview. Maternal pre-pregnancy BMI was calculated as pre-pregnancy weight (kg) divided by squared height (m), and further dichotomized into non-OWO (BMI < 25 kg/m²) and OWO (BMI ≥ 25 kg/m²). Fish consumption was grouped into none, 1–2 servings/week, and ≥ 3 servings/week. Maternal pregnancy complications, including diabetes mellitus (either gestational diabetes or pre-existing diabetes), hypertensive disorders (pre-eclampsia, eclampsia, gestational hypertension, chronic hypertension, and HELLP [hemolysis, elevated liver enzymes, and low platelets syndrome]), were abstracted from medical records. Gestational age was estimated based on the first day of the last menstrual period, as recorded in the maternal medical record, or early (< 20 weeks) prenatal ultrasonographic results, as detailed in a previous publication [21].

Child’s birthweight and gender were abstracted from the medical records. Birthweight for gestational age was calculated according to an established local reference population, and controlled for infant gender, gestational age, and ethnicity [21]. Fetal growth pattern was defined by birthweight for gestational age and grouped into small for gestational age (SGA) (gestational age specific birthweight < 10th percentile), large for gestational age (LGA) (birthweight > 90th percentile), and appropriate for gestational age (AGA) (birthweight in the 10th to 90th percentile for gestational age). Information regarding infant breastfeeding history was primarily assessed within the first 2 years of follow-up visits, and grouped into exclusively breastfeeding, exclusively formula feeding, or mixed breast and formula feeding [23].

Child weight and height were extracted from the medical records, which were measured by trained medical staff using the same clinical protocol and equipment during well-child visits. Before data analyses, careful data checking and cleaning of weight and height data was performed. At first, we removed extreme or biologically implausible values. Then, we identified outliners or erroneous height and weight values based on growth curve. When possible, erroneous height and weight values were corrected, otherwise the points were deleted. Age- and sex-specific BMI z-scores and percentiles were calculated using US national reference data [24], derived from a US representative sample. OWO was defined as BMI ≥ 85th percentile for age and sex [25]. The BBC has used rolling enrollment, so the length of postnatal follow-up and number of well-child visits varied greatly across study participants. It is known that OWO at an older age is more likely to persist into adulthood. Thus, we chose the last visit in each time period as the end point for OWO.

Demographic and clinical data are presented as either mean ± standard deviation (SD) or n(%) stratified by maternal Hg categories. Unadjusted trend p values across maternal Hg categories were calculated by Mantel-Haenszel χ² for categorical variables and linear regression for continuous variables. The relationship between maternal Hg and child OWO is displayed using locally weighted regression smoothing plots (implemented using PROC LOESS in SAS, a nonparametric regression method).

As shown in Fig. 1, maternal Hg levels were positively associated with an increased risk of child OWO (panel a). When stratified by maternal pre-pregnancy OWO and/or diabetes, the association was strongest among OWO and diabetic mothers (panel b). Furthermore, the risk of child OWO among OWO and diabetic mothers was the highest across the whole Hg spectrum. Multivariate regression models (Table 2) showed that the increased risk of child OWO was mainly concentrated in the top Hg quartile (Q4) with an RR of 1.24 (95% CI 1.05–1.47), while second (Q2) and third (Q3) Hg quartiles were not significantly associated with the risk of child OWO, as compared with the lowest quartile (Q1). When we combined Q1–Q3 as the reference, the top Hg quartile was associated with increased risk of child OWO (RR = 1·19; 95% CI 1.05–1.35; p = 0·007). These associations persisted from preschool age (2–5 years), to school age (6–9 years) and into early adolescence (10–15 years).

We found a significant combined effect of maternal Hg levels and either or both pre-pregnancy OWO and diabetes on child OWO risk. Children of OWO mothers in the top quartile of Hg had an increased risk of OWO (RR = 1.87; 95% CI 1.55–2.25) compared to those with non-OWO mothers in lower quartiles of Hg (Fig. 2, a). Similarly, the risk of OWO was significantly higher in children of diabetic mothers in the top quartile of Hg than those with non-diabetic mothers in lower quartiles of Hg (RR = 1.45; 95% CI 1·15–1.83). When the presence of maternal OWO and diabetes was categorized into three groups: none (without OWO and diabetes), any (either OWO or diabetes), and both (OWO and diabetes), the risk of OWO (RR = 2.06, 95% CI 1.56–2.71) was highest in children born to mothers with both OWO and diabetes and in the top quartile of Hg, compared with children of mothers without OWO and diabetes and in low quartiles of Hg (Fig. 2a). However, there was no evidence of an interaction between maternal RBC-Hg and pre-pregnancy OWO or between maternal Hg and diabetes with respect to the risk of child OWO (p > 0.05). These association patterns were similar for preschool age children (Fig. 2b), school age children (c), and adolescents (d).

Maternal plasma folate was inversely related with Hg concentration (Additional file 1: Figure S3). As shown in Table 3, the rate of child OWO was significantly lower in children whose mothers in the highest Hg quartile and with adequate maternal folate than children whose mothers with a low folate levels (43.6% vs. 60.9%; RR = 0·66; 95% CI 0.51–0·85; p = 0.001). Adequate maternal folate status was not associated with reduced child OWO risk (RR = 0.96; 95% CI 0.80–1.15; p = 0.658) among children of mothers with Hg levels in the lower three quartiles. The interaction between maternal RBC-Hg and plasma folate on child OWO risk was marginal (p = 0.086 for interaction).

To control for other potential confounders, we further adjusted for frequency of maternal fish consumption (Additional file 1: Table S2), maternal RBC-selenium levels (Additional file 1: Table S3), and maternal RBC-lead levels (Additional file 1: Table S4). The results were similar to those reported above. A series of subgroup analyses were performed, and the associations were not changed materially among fish consumers (Additional file 1: Table S5), term births (Additional file 1: Table S6), Blacks (Additional file 1: Table S7), boys (Additional file 1: Table S8), girls (Additional file 1: Table S9), and breastfed children (Additional file 1: Table S10). To further evaluate the associations, we analyzed child BMI z-score as a continuous outcome (Additional file 1: Table S11), and again the results were consistent.