Maternal Metals Exposure and Infant Weight Trajectory: The Japan Environment and Children’s Study (JECS)

Abstract

Background:

To our knowledge, the association of maternal exposure to metallic elements with weight trajectory pattern from the neonatal period has not been investigated.

Objectives:

The goals of this study were to identify infant growth trajectories in weight in the first 3 y of life and to determine the associations of maternal blood levels of lead, cadmium, mercury, selenium, and manganese with growth trajectory.

Methods:

This longitudinal study, part of the Japan Environment and Children Study, enrolled 103,099 pregnant women at 15 Regional Centres across Japan between 2011 and 2014. Lead, cadmium, mercury, selenium, and manganese levels were measured in blood samples collected in the second (14–27 wk gestational age) or third trimester ($\ge 28\mathrm{wk}$). Growth trajectory of 99,014 children was followed until age 3 y. Raw weight values were transformed to age- and sex-specific weight standard deviation (SD) scores, and latent-class group-based trajectory models were estimated to determine weight trajectories. Associations between maternal metallic element levels and weight trajectory were examined using multinomial logistic regression models after confounder adjustment.

Results:

We identified 5 trajectory patterns based on weight SD score: 4.74% of infants were classified in Group I, very small to small; 31.26% in Group II, moderately small; 21.91% in Group III, moderately small to moderately large; 28.06% in Group IV, moderately large to normal; and 14.03% in Group V, moderately large to large. On multinomial logistic regression, higher maternal lead and selenium levels tended to be associated with increased odds ratios (ORs) of poor weight SD score trajectories (Groups I and II), in comparison with Group III. Higher levels of mercury were associated with decreased ORs, whereas higher levels of manganese were associated with increased ORs of “moderately large” trajectories (Groups IV and V).

Discussion:

Maternal lead, mercury, selenium, and manganese blood levels affect infant growth trajectory pattern in the first 3 y of life. https://doi.org/10.1289/EHP10321

Introduction

The World Health Organization (WHO) estimates that 15%–20% of all births have low birth weight (LBW)¹; in particular, 6% of infants born in East Asia and the Pacific, 13% in sub-Saharan Africa, and up to 28% in South Asia have LBW.² LBW due to intrauterine growth restriction affects immune function,³ infectious disease hospitalization,⁴ illness, and mortality during the neonatal period.⁵ Negative effects of LBW can continue over a long period and cause adverse health outcomes, including cognitive deficits and altered brain structure in school-age children,⁶ mortality in adolescence,⁷ and disease and mortality in later life.⁸ Reducing LBW thus is a serious public health concern worldwide.

Previous studies have reported an association between maternal exposure to several metallic elements, including iron,⁹ lead,¹⁰ cadmium,¹¹ mercury,¹² selenium,¹³ and manganese,¹⁴ and infant birth weight. In fact, a systematic review and meta-analysis identified an association of prenatal anemia and its most common cause—iron deficiency—with risk of LBW.⁹ As a consequence, two guidelines (WHO and Nordic Nutrition Recommendations) now recommend routine iron supplementation during pregnancy. However, other studies have reported that levels of several metallic elements, such as selenium and manganese, during pregnancy show only a marginal or no significant association with birth weight^15,16; thus, dietary intake recommendations for these metallic elements during pregnancy require further consideration.

According to the Developmental Origins of Health and Disease hypothesis,¹⁷ postnatal growth patterns, in addition to infant birth weight, also play a key role in neonatal prognosis. Previous studies examining infant growth trajectory patterns have done so by targeting specific groups based on, for example, adiponectin concentration,¹⁸ LBW status,¹⁹ birth weight quartile,²⁰ and prepregnancy body mass index (BMI).²¹ A few studies, however, have examined entire populations to identify growth trajectory patterns. For example, Carling et al. examined weight-for-length trajectory using prenatal cohort data obtained from 595 subjects,²² Wang et al. investigated BMI growth trajectory from age 2 to 6 y using large cohort data ($n=\mathrm{71,892}$),²³ and Marinac et al. studied body shape trajectory from age 5 to 60 y using large cohort data (4,280,712 person-years of follow-up).²⁴ Although weight standard deviation (SD) score is widely used in clinical settings to evaluate postnatal growth patterns, use of the weight SD score in longitudinal studies has to date been limited.²⁰ To our knowledge, no study has used multiple-repeated-measures data obtained from a large birth cohort study to investigate trajectory patterns in weight SD scores calculated for the neonatal period and examined their associations with maternal levels of metallic elements.

The ultimate goal of the nationwide birth cohort Japan Environment and Children’s Study (JECS) is to identify environmental factors that affect children’s health and development.^25–27 Lead, cadmium, mercury, selenium, and manganese are considered to be important coexposures when examining the effect of other chemical substances on child development, and the JECS has provided basic data on the levels of these five metallic elements.²⁵ In this prospective study, we examined data on maternal levels of blood metallic elements during pregnancy and repeated-measures data on infants’ weight SD scores from the neonatal period up to age 3 y. This study had two objectives: a) to identify weight SD score trajectories in the first 3 y of life and b) to examine the association of maternal levels of blood metallic elements (lead, cadmium, mercury, selenium, and manganese) with infant growth trajectories in weight.

Methods

Data Source and Study Population

The data sets were sourced from a nationwide government-funded birth cohort study in Japan, the JECS. The design of the JECS has been published elsewhere.^26,27 Briefly, a total of 103,099 women in early pregnancy were recruited and registered at 15 Regional Centres across Japan via co-operating health care providers and/or local government offices between 2011 and 2014. The profile paper reported that the children in the JECS covered approximately 45% of total live births within the Study Area,²⁷ and it is assumed that the JECS participants are a representative sample of the Japanese population. Maternal blood samples for metallic element testing were collected during mid–late pregnancy.²⁵ The present study used the jecs-ta-20190930 data set, released in October 2019, which contains follow-up data until the children were age 3 y (8 waves of follow-up). To be eligible for the study, individuals had to have repeated measurement data required for calculating weight SD score at least twice during the 8 waves of follow-up. Among 104,062 fetuses, the study excluded women who had miscarriages, stillbirths, or complications for unknown reasons at birth ($n=758$) and those who withdrew consent during the follow-up period ($n=\mathrm{3,623}$). We further excluded those with missing data or single-measurement data for weight and height during the 8 waves of follow-up ($n=663$), those with logical contradiction in monthly age data at the time of measurement without 3 months of data, or those with missing data for sex ($n=4$). The data source for the present study was the 99,014 children who had their weight SD score. The average number (SD) of follow-up assessments was 6.1 (1.7), and the total number of observations during follow-up was 599,192 (Figure 1). Among 99,014 children, 97,164 were singletons, 1,810 were twins, and 40 were triplets; in addition, we used 99,014 samples to identify the infant growth trajectories in weight in all participants in the JECS. We additionally conducted preliminary analysis without twins and triplets ($n=\mathrm{97,164}$).

Figure 1.

Flowchart of data processing to identify growth trajectory patterns in weight SD score in the first 3 y of life in the JECS study.

The JECS protocol was reviewed and approved by the Ministry of the Environment’s institutional review board on epidemiological studies and the Ethics Committees of all participating institutions. The JECS was conducted in accordance with the Helsinki Declaration and other nationally valid regulations and guidelines. Written informed consent was obtained from all participants.

Outcomes of Interest

In this study, we collected children’s weight data from medical records transcribed at delivery and age 1 month. From 1 month after birth, self-administered questionnaires were continually mailed to participants every 6 months based on the children’s birthday.²⁶ From age 0.5 to 3 y, we used children’s weight data provided on self-administered questionnaires by their mothers and/or guardians, and corresponding weight values were selected if they were measured within 3 months ($\text{highest\hspace{0.17em}frequency}\pm 1\text{\hspace{0.17em}month}$) at the time of each survey as follows: “0.5 years” included 4 to 6 months, “1 year” included 10–12 months, “1.5 years” included 16–18 months, “2 years” included 22–24 months, “2.5 years” included 28–30 months, and “3 years” included 34–36 months. We accepted weight data from the self-administered questionnaires that were within $\text{mean}\pm 5\text{\hspace{0.17em}SD}$. We transformed the raw weight values to age- and sex-specific weight SD scores²⁸ and referred to the mean and SD of the reference population at each age in months provided in data from The Japanese Society for Pediatric Endocrinology.²⁹ Finally, we obtained repeated-measures data on weight SD scores across the 8 waves of follow-up for the following age groups: 0 months (at delivery), 1 month, 0.5 y, 1 y, 1.5 y, 2 y, 2.5 y, and 3 y.

Exposure Assessment

In the JECS study, lead, cadmium, mercury, selenium, and manganese were measured as important coexposures when seeking to evaluate the effect of other chemical contaminants on child health.²⁵ Blood samples ($33\mathrm{mL}$) were collected by medical staff when the participants visited cooperating health care providers in the second (14–27 wk of gestational age) or third trimester (28 wk or more of gestational age). Approximately half of the blood samples used in this study were collected at the second trimester, and about half were collected at the third trimester. Whole blood samples for chemical analysis were collected into $3-\text{mL}$ tubes with sodium ethylenediaminetetraacetic acid (EDTA), transferred to a central laboratory within 48 h, divided into cryo-biobanking tubes, and stored at $-80°\mathrm{C}$ until analysis. Detailed information on measurement of metallic elements, including details on chemicals, reagents, sample preparation, instrument analysis, calculations, quality control, and comparison of each element with previous studies has been published elsewhere.²⁵ Briefly, levels of metallic elements in this study were measured using inductively coupled plasma–mass spectrometry (ICP-MS) on an Agilent 7700 device (Agilent Technologies). Detection rate for measurement of metallic elements was 100%, and the method detection limits for each element were calculated based on Currie’s method using the following equation: $t_{(n-\mathrm{1,0.05})}\times 2\times s$, that t(n−1, 0.05) represents the Student’s t value under an α level of 0.05 with n−1 degrees of freedom, and s represents the standard deviation of blank measurements in n replicates ($n\ge 7$).²⁵

The analysis was conducted in three different laboratories and quality control data indicated that all three laboratories showed good agreement, resulting in very high precision with differences of $<10\%$ relative SD between each measurement.²⁵ Seven-point calibration curves had coefficients of determination higher than 0.9999. The method detection limits for lead, cadmium, mercury, selenium, and manganese were 0.129, 0.0234, 0.0490, 0.837, and $0.522\text{\hspace{0.17em}ng}/\mathrm{g}$, respectively.²⁵ Of the 99,014 children included in this study, maternal metallic element measurements were obtained for 95,010 (Figure 1).

Covariates

We used medical records transcribed during pregnancy and after delivery, and self-administered questionnaires conducted at the time of study registration, during mid- to late pregnancy and at child age 1 month to assess the following information: maternal age at delivery,³⁰ infant sex, gestational duration,³¹ mode of delivery,³² primipara, medical and obstetric history,^33,34 BMI (weight and height) before pregnancy, weight before pregnancy, weight at delivery, weight gain during pregnancy,³⁵ smoking habit,³⁶ drinking habits,³⁷ household income,³⁸ education,³⁹ physical activity levels using The International Physical Activity Questionnaire short form (IPAQ),^40,41 health-related quality of life using The Short Form-8 (SF-8),⁴² psychological distress using The Kessler Psychological Distress Scale (K6),⁴³ long-term dietary intake of foods and nutrients using the Food Frequency Questionnaire (FFQ),⁴⁴ and breastfeeding during the first month of life. Additionally, breastfeeding period and infant formula period were assessed in the self-administered questionnaires administered when the child was 1 y old.

Medical and obstetric history in the present study included clinically relevant medical conditions: anemia, hypertension, hyperlipidemia, congenital heart disease, Kawasaki disease, pregnancy hypertension, bronchial asthma, atopic dermatitis, drug allergy, connective tissue disease, other immunological disease, type 2 diabetes mellitus, gestational diabetes, hyperthyroidism/Basedow disease, hypothyroidism/Hashimoto’s disease, depression, chronic nephritis (IgA nephropathy/glomerulonephritis), other kidney disease, paramenia/menoxenia, endometriosis, hysteromyoma, adenomyosis uteri, uterine deformity, ovarian tumor/ovarian cyst, polycystic ovary syndrome, other gynopathy, breast cancer, cervical cancer, uterine cancer, placental abruption, and other abnormal pregnancy/abnormal delivery. For each condition, participants were asked whether a physician had diagnosed the specific condition (yes or no).

We used the food frequency questionnaire (FFQ) that formed part of the Japan Public Health Center-based Prospective Study for the Next Generation (JPHC-NEXT).⁴⁴ In this study, an FFQ was used to assess estimated daily intake of energy (kcal), protein (grams per day), fat (grams per day), carbohydrate (grams per day), vitamin D (micrograms per day), vitamin K (micrograms per day), vitamin B1 (milligrams per day), vitamin B2 (milligrams per day), niacin (milligrams per day), vitamin B6 (milligrams per day), vitamin B12 (micrograms per day), folic acid (micrograms per day), pantothenic acid (milligrams per day), vitamin C (milligrams per day), saturated fatty acids (grams per day), monounsaturated fatty acid (grams per day), polyunsaturated fatty acid (grams per day), and cholesterol (milligrams per day).

Categories of covariates in this study were mode of delivery (vaginal, induction of labor, vacuum extraction delivery, forceps delivery, cesarean); smoking habit (never, ex-smoker who quit before pregnancy, ex-smoker who quit from pregnancy, current); drinking habit (never, ex-drinker who quit before pregnancy, ex-drinker who quit from pregnancy, current); household income ($<2\text{\hspace{0.17em}million}$, $2–4\text{\hspace{0.17em}million}$, $4–6\text{\hspace{0.17em}million}$, $6–8\text{\hspace{0.17em}million}$, $8–10\text{\hspace{0.17em}million}$, $10–12\text{\hspace{0.17em}million}$, $12–15\text{\hspace{0.17em}million}$, $15–20\text{\hspace{0.17em}million}$, $\ge 20\text{\hspace{0.17em}million}$ Japanese yen); education (junior high school, high school, technical school, vocational school, junior college, university, graduate school); health-related quality of life, using the SF-8, which covers general health, physical function, role physical, bodily pain, vitality, social functioning, mental health, and role emotional, and physical component summary and mental component summary; and breastfeeding during the first month of life (mother’s milk only, mother’s milk and infant formula, infant formula only).

The adjusted logistics regression model used to examine the associations of maternal intake of metallic elements with weight SD score trajectory pattern included covariates by accumulated evidence and comparison with maternal metallic element levels; however, correlation values above 0.4 were excluded to avoid multicollinearity, and medical and obstetric history with a prevalence below 1.00% was also subsequently excluded. Selected confounders were maternal age at delivery, infant sex, gestational duration, mode of delivery, primipara, medical and obstetric history at early pregnancy (anemia, pregnancy hypertension, bronchial asthma, atopic dermatitis, drug allergy, hyperthyroidism/Basedow disease, hypothyroidism/Hashimoto’s disease, depression, other kidney disease, paramenia/menoxenia, endometriosis, hysteromyoma, ovarian tumor/ovarian cyst, polycystic ovary syndrome, other gynopathy, other abnormal pregnancy/abnormal delivery), BMI and height before pregnancy, weight gain during pregnancy, smoking at early pregnancy, drinking at early pregnancy, household income at midpregnancy, education at midpregnancy, IPAQ at early pregnancy, SF-8 at early pregnancy (general health, bodily pain), K6 at midpregnancy, FFQ at early pregnancy (energy), breastfeeding during the first month of life, and breastfeeding period until 12 months; these selected confounders were also included as individual covariates.

Statistical Analyses

First, we identified trajectory patterns in weight SD scores calculated for the first 3 y of life using latent-class group-based trajectory models in SAS PROC TRAJ (SAS Institute, Inc.),^45,46 which assumes that a study population comprises a mixture of finite latent groups within which people follow an approximately homogeneous growth trajectory. Cubic trajectory models were fitted for a fixed number of latent groups, and posterior probabilities for group membership assignment were calculated for each individual. Model selection for the number of trajectory groups and functional form were based on following criteria (Bayesian information criterion for the models and the precision of the group proportion, clinical trajectory interpretability, and the average posterior probability of assignment of the participants to their groups). Next, we compared maternal levels of metallic elements and covariates among the weight SD score trajectory groups using the ${\mathrm{\chi }}^2$ test or analysis of variance (ANOVA). Finally, multinomial logistic regression models were fitted to examine the associations between maternal levels of metallic elements and weight SD score trajectory in the adjusted model. Based on several previous studies’ reported nonlinear associations of maternal metallic elements with birth weight, we categorized exposure variables into four levels. We confirmed the overall trend for the associations of maternal metallic elements with infant growth trajectories in weight by the test for linear trend. Bonferroni correction was used to adjust for multiple comparisons. All statistical analyses were conducted with SAS (version 9.4; SAS Institute, Inc.). The level of significance was set at $p<0.05$.

Results

Mothers included in the study had a mean (SD) age at delivery of 31.17 (5.05) y, gestational duration of 38.77 (1.65) wk, weight before pregnancy of 53.22 (8.98) kg, and weight at delivery of 63.49 (9.41) kg. Among the infants, 51.25% were male, 56.71% were born vaginally, and 20.05% were delivered by cesarean. The mean birth weight was 3,010.09 (430.54) g, mean birth height was 48.85 (2.37) cm, and mean gestational duration was 38.77 (1.65) wk (Table 1; Table S1). Maternal levels of blood metallic elements were shown in gravimetric units and volumetric concentrations.²⁵ Median maternal blood levels were $5.85\text{\hspace{0.17em}ng}/\mathrm{g}$ and 0.62 $\mathrm{\mu }\mathrm{g}/\text{dL}$ for lead, $0.66\text{\hspace{0.17em}ng}/\mathrm{g}$ and $0.69\mathrm{\mu }\mathrm{g}/\mathrm{L}$ for cadmium, $3.64\text{\hspace{0.17em}ng}/\mathrm{g}$ and $3.82\mathrm{\mu }\mathrm{g}/\mathrm{L}$ for mercury, $168.0\text{\hspace{0.17em}ng}/\mathrm{g}$ and $176.5\mathrm{\mu }\mathrm{g}/\mathrm{L}$ for selenium, and $15.4\text{\hspace{0.17em}ng}/\mathrm{g}$ and $16.2\mathrm{\mu }\mathrm{g}/\mathrm{L}$ for manganese, respectively (Table 2).

Table 1.

Characteristics of this study subjects in the Japan Environment and Children’s study (JECS) 2011–2014, stratified by weight SD trajectory group.

Variableb	Weight SD trajectory groupa					Total	$p$-Valuec
	Group I: Very small to small (4.74%)	Group II: Moderately small (31.26%)	Group III: Moderately small to moderately large (21.91%)	Group IV: Moderately large to normal (28.06%)	Group V: Moderately large to large (14.03%)
Lead level ($\text{ng}/\mathrm{g}$) ($n=\mathrm{95,010}$)	6.59 (2.63)	6.37 (2.99)	6.46 (2.94)	6.21 (2.75)	6.32 (2.65)	6.35 (2.85)	$<.001$
Cadmium level ($\text{ng}/\mathrm{g}$) ($n=\mathrm{95,010}$)	0.80 (0.43)	0.74 (0.38)	0.77 (0.39)	0.73 (0.37)	0.76 (0.38)	0.75 (0.38)	$<.001$
Mercury level ($\text{ng}/\mathrm{g}$) ($n=\mathrm{95,010}$)	4.24 (2.49)	4.23 (2.48)	4.23 (2.54)	4.15 (2.45)	4.17 (2.49)	4.20 (2.49)	$<.001$
Selenium level ($\text{ng}/\mathrm{g}$) ($n=\mathrm{95,010}$)	169.8 (21.1)	170.5 (20.2)	169.7 (20.2)	169.9 (20.2)	169.5 (21.0)	169.99 (20.33)	$<.001$
Manganese level ($\text{ng}/\mathrm{g}$) ($n=\mathrm{95,010}$)	15.84 (4.86)	15.86 (4.67)	15.84 (4.63)	16.12 (4.67)	16.13 (4.67)	15.96 (4.67)	$<.001$
Infant sex (% male) ($n=\mathrm{99,014}$)	48.19	51.74	46.26	55.12	51.21	51.25	$<.001$
Infant birth height (cm) ($n=\mathrm{98,541}$)	43.73 (3.98)	48.36 (1.75)	48.23 (1.79)	50.05 (1.59)	50.22 (1.76)	48.85 (2.37)	$<.001$
Maternal age at delivery (y) ($n=\mathrm{99,014}$)	31.76 (5.15)	31.15 (5.04)	31.50 (5.05)	30.89 (5.03)	31.06 (5.03)	31.17 (5.05)	$<.001$
Gestational duration (wk) ($n=\mathrm{98,830}$)	35.56 (3.55)	38.78 (1.27)	38.39 (1.45)	39.37 (1.07)	39.21 (1.19)	38.77 (1.65)	$<.001$
Mode of delivery ($n=\mathrm{98,588}$)	—	—	—	—	—	—	$<.001$
Vaginal	34.55	58.97	54.12	60.49	55.61	56.71	—
Induction of labor	11.02	16.24	15.67	19.97	20.79	17.55	—
Vacuum extraction delivery	2.94	4.89	5.21	6.15	6.69	5.47	—
Forceps delivery	0.09	0.20	0.20	0.22	0.30	0.21	—
Cesarean	51.41	19.69	24.80	13.17	16.62	20.05	—
Primipara (% yes) ($n=\mathrm{99,014}$)	48.62	41.20	49.17	38.24	44.00	42.86	$<.001$
Medical and obstetric history at early pregnancy ($n=\mathrm{97,872}$)	—	—	—	—	—	—	—
Anemia (% yes)	18.20	18.44	17.64	19.36	18.29	18.49	$<.001$
Pregnancy hypertension (% yes)	3.33	1.36	1.57	1.13	1.50	1.45	$<.001$
Bronchial asthma (% yes)	10.82	11.04	10.20	11.23	11.10	10.91	.005
Atopic dermatitis (% yes)	16.27	16.07	15.20	16.13	14.76	15.72	$<.001$
Drug allergy (% yes)	2.94	2.67	2.75	2.43	2.34	2.59	.023
Hyperthyroidism/Basedow disease (% yes)	1.21	1.13	1.18	0.93	0.96	1.07	.029
Hypothyroidism/Hashimoto’s disease (% yes)	1.23	0.96	1.22	0.90	0.87	1.00	.001
Depression (% yes)	3.85	3.06	3.00	2.92	2.88	3.69	.012
Other kidney disease (% yes)	1.49	1.60	1.58	1.43	1.23	1.49	.030
Paramenia/menoxenia (% yes)	14.30	11.35	11.60	10.93	11.22	11.41	$<.001$
Endometriosis (% yes)	4.35	3.66	4.14	3.23	3.52	3.66	$<.001$
Hysteromyoma (% yes)	8.20	5.73	7.27	5.28	6.07	6.10	$<.001$
Ovarian tumor/ovarian cyst (% yes)	3.48	3.47	3.75	3.20	3.48	3.46	.025
Polycystic ovary syndrome (% yes)	2.99	2.34	2.17	2.34	2.14	2.31	.010
Other gynopathy (% yes)	4.95	3.92	4.32	3.80	3.88	4.02	$<.001$
Other abnormal pregnancy, abnormal delivery (% yes)	2.90	2.30	2.22	1.69	1.72	2.06	$<.001$
BMI before pregnancy (${\mathrm{kg}/\mathrm{m}}^2$) ($n=\mathrm{98,899}$)	20.89 (3.48)	20.78 (3.06)	21.09 (3.21)	21.49 (3.32)	22.08 (3.65)	21.23 (3.30)	$<.001$
Weight before pregnancy (kg) ($n=\mathrm{98,945}$)	51.20 (8.96)	51.19 (8.02)	52.88 (8.54)	54.09 (8.89)	56.57 (9.88)	53.13 (8.89)	$<.001$
Height before pregnancy (cm) ($n=\mathrm{98,952}$)	156.52 (5.38)	156.93 (5.21)	158.31 (5.28)	158.61 (5.17)	160.03 (5.34)	158.12 (5.35)	$<.001$
Weight at delivery (kg) ($n=\mathrm{96,747}$)	59.47 (9.13)	61.18 (8.61)	62.71 (8.52)	65.19 (9.71)	67.77 (9.65)	63.49 (9.41)	$<.001$
Weight gain during pregnancy (kg) ($n=\mathrm{96,685}$)	8.26 (4.21)	9.98 (4.85)	9.82 (4.08)	11.07 (5.71)	11.17 (4.25)	10.34 (4.91)	$<.001$
Smoking at early pregnancy (% yes) ($n=\mathrm{97,161}$)	—	—	—	—	—	—	$<.001$
Never	59.55	60.10	58.35	57.44	55.53	58.30	—
Ex-smoker who quit before pregnancy	21.75	22.90	23.06	25.03	24.47	23.70	—
Ex-smoker who quit from pregnancy	12.19	12.21	13.10	13.59	15.29	13.22	—
Current	6.50	4.79	5.49	3.95	4.71	4.78	—
Drinking at early pregnancy (% yes) ($n=\mathrm{97,439}$)	—	—	—	—	—	—	$<.001$
Never	35.93	35.83	33.85	34.60	32.48	34.58	—
Past	54.15	54.37	56.11	55.56	57.25	55.48	—
Current	9.91	9.80	10.03	9.85	10.27	9.94	—
Household income at midpregnancy (% yes) ($n=\mathrm{90,611}$)	—	—	—	—	—	—	$<.001$
$\mathrm{<}2\text{\hspace{0.17em}million}$ yen	6.64	5.57	5.46	5.69	5.63	5.64	—
$2–4\text{\hspace{0.17em}million}$ yen	35.15	34.83	33.32	35.18	34.09	34.50	—
$4–6\text{\hspace{0.17em}million}$ yen	31.93	33.39	32.93	32.93	33.13	33.06	—
$6–8\text{\hspace{0.17em}million}$ yen	15.18	15.95	16.64	15.64	15.78	15.96	—
$8–10\text{\hspace{0.17em}million}$ yen	6.67	6.20	7.01	6.54	6.68	6.56	—
$10–12\text{\hspace{0.17em}million}$ yen	2.60	2.27	2.80	2.17	2.65	2.43	—
$12–15\text{\hspace{0.17em}million}$ yen	0.84	0.94	1.05	0.91	0.95	0.95	—
$15–20\text{\hspace{0.17em}million}$ yen	0.58	0.56	0.52	0.53	0.68	0.56	—
$\ge 20\text{\hspace{0.17em}million\hspace{0.17em}yen}$	0.41	0.29	0.26	0.41	0.41	0.34	—
Education at mid-pregnancy (% yes) ($n=\mathrm{96,943}$)	—	—	—	—	—	—	$<.001$
Junior high school	5.08	4.75	4.17	5.08	5.12	4.78	—
High school	31.07	30.99	30.75	31.88	32.32	31.38	—
Technical school	1.59	1.59	1.62	1.68	1.71	1.64	—
Vocational school	23.48	22.65	22.90	22.73	23.64	22.90	—
Junior college	17.55	18.34	17.59	17.33	16.13	17.54	—
University	19.96	20.26	21.46	19.85	19.45	20.28	—
Graduate school	1.28	1.41	1.50	1.44	1.63	1.47	—
IPAQ at early pregnancy	—	—	—	—	—	—	—
$\text{MET}-\mathrm{minutes}$ ($n=\mathrm{96,020}$)	425.26 (762.09)	401.63 (722.96)	391.90 (699.26)	408.31 (716.09)	431.43 (758.51)	406.67 (723.04)	$<.001$
kcal ($n=\mathrm{96,015}$)	380.76 (708.13)	361.20 (693.38)	362.17 (662.69)	386.88 (764.90)	427.21 (773.43)	378.82 (720.27)	$<.001$
SF-8 at early pregnancy	—	—	—	—	—	—	—
General health  ($n=\mathrm{97,509}$)	46.25 (7.52)	46.42 (7.69)	46.36 (7.69)	46.52 (7.67)	46.71 (7.68)	46.47 (7.67)	$<.001$
Physical function  ($n=\mathrm{97,154}$)	45.69 (8.32)	46.25 (7.64)	46.10 (7.73)	46.43 (7.41)	46.41 (7.52)	46.26 (7.62)	$<.001$
Role physical  ($n=\mathrm{97,230}$)	42.55 (9.87)	43.29 (9.21)	43.10 (9.34)	43.44 (9.13)	43.56 (9.19)	43.29 (9.25)	$<.001$
Bodily pain  ($n=\mathrm{97,430}$)	49.88 (8.32)	50.15 (8.24)	50.24 (8.25)	50.22 (8.31)	50.27 (8.23)	50.19 (8.26)	.041
Vitality  ($n=\mathrm{97,521}$)	46.72 (7.76)	46.97 (7.68)	46.95 (7.73)	47.11 (7.57)	47.17 (7.66)	47.02 (7.66)	$<.001$
Social functioning  ($n=\mathrm{97,166}$)	43.51 (9.77)	43.90 (9.48)	43.66 (9.59)	44.09 (9.41)	44.09 (9.44)	43.91 (9.50)	$<.001$
Mental health  ($n=\mathrm{97,401}$)	46.89 (7.06)	47.18 (7.11)	47.17 (7.02)	47.21 (6.99)	47.31 (7.15)	47.19 (7.06)	.013
Role emotional  ($n=\mathrm{97,083}$)	46.60 (8.61)	47.03 (8.03)	46.84 (8.24)	47.19 (7.82)	47.11 (7.96)	47.03 (8.04)	$<.001$
Physical component summary  ($n=\mathrm{96,024}$)	44.60 (7.62)	45.10 (7.33)	45.01 (7.39)	45.25 (7.24)	45.34 (7.24)	45.13 (7.32)	$<.001$
Mental component summary  ($n=\mathrm{96,024}$)	45.96 (7.46)	46.16 (7.28)	46.06 (7.34)	46.26 (7.14)	46.25 (7.25)	46.17 (7.26)	$<.001$
K6 at midpregnancy ($n=\mathrm{97,220}$)	3.87 (4.07)	3.45 (3.75)	3.47 (3.73)	3.43 (3.72)	3.44 (3.73)	3.47 (3.75)	$<.001$
FFQ at early pregnancy ($n=\mathrm{97,859}$)	—	—	—	—	—	—	—
Energy (kcal/d)	1795.29 (842.24)	1808.6 (805.89)	1821.09 (796.46)	1835.81 (823.57)	1856.81 (940.89)	1825.13 (830.96)	$<.001$
Protein (g/d)	61.92 (34.28)	62.18 (33.02)	62.86 (33.00)	62.90 (33.19)	63.94 (39.20)	62.77 (34.06)	$<.001$
Fat (g/d)	60.75 (37.22)	61.25 (36.55)	61.77 (36.67)	61.83 (37.25)	62.15 (41.55)	61.63 (37.55)	.052
Carbohydrate (g/d)	242.61 (105.70)	244.69 (102.49)	245.86 (101.21)	249.51 (105.06)	252.90 (115.76)	247.35 (105.09)	$<.001$
Vitamin D ($\mathrm{\mu }\mathrm{g}/\mathrm{d}$)	5.31 (4.88)	5.30 (5.21)	5.28 (5.48)	5.28 (4.43)	5.37 (5.24)	5.30 (5.06)	.463
Vitamin K ($\mathrm{\mu }\mathrm{g}/\mathrm{d}$)	197.57 (185.76)	196.64 (169.38)	202.04 (175.72)	201.91 (182.58)	207.93 (197.34)	200.94 (179.45)	$<.001$
Vitamin B1 (mg/d)	0.85 (0.49)	0.86 (0.46)	0.86 (0.450)	0.87 (0.46)	0.88 (0.56)	0.87 (0.48)	$<.001$
Vitamin B2 (mg/d)	1.19 (0.84)	1.19 (0.85)	1.21 (0.84)	1.21 (0.86)	1.24 (0.97)	1.21 (0.87)	$<.001$
Niacin (mg/d)	14.96 (8.78)	14.81 (7.95)	15.06 (8.11)	14.88 (7.80)	15.15 (9.73)	14.94 (8.26)	$<.001$
Vitamin B6 (mg/d)	1.07 (0.61)	1.07 (0.57)	1.08 (0.56)	1.08 (0.57)	1.10 (0.66)	1.08 (0.59)	$<.001$
Vitamin B12 ($\mathrm{\mu }\mathrm{g}/\mathrm{d}$)	4.62 (3.98)	4.57 (3.99)	4.62 (4.31)	4.61 (3.85)	4.72 (4.63)	4.62 (4.12)	.009
Folic acid ($\mathrm{\mu }\mathrm{g}/\mathrm{d}$)	278.98 (187.66)	276.37 (175.78)	278.64 (166.76)	279.99 (180.14)	286.29 (202.42)	279.40 (179.68)	$<.001$
Pantothenic acid (mg/d)	6.12 (3.60)	6.15 (3.53)	6.23 (3.48)	6.26 (3.56)	6.39 (4.19)	6.23 (3.63)	$<.001$
Vitamin C (mg/d)	95.45 (83.47)	95.35 (76.82)	95.66 (72.96)	97.41 (79.16)	99.30 (79.71)	96.56 (77.40)	$<.001$
Saturated fatty acids (g/d)	19.25 (13.21)	19.53 (13.44)	19.76 (13.44)	19.79 (13.70)	19.89 (14.71)	19.69 (13.69)	.007
Monounsaturated fatty acid (g/d)	22.51 (14.20)	22.70 (13.70)	22.85 (13.82)	22.87 (13.91)	22.90 (15.30)	22.80 (14.04)	.262
Polyunsaturated fatty acid (g/d)	12.20 (7.13)	12.20 (6.78)	12.26 (6.79)	12.28 (6.93)	12.38 (8.31)	12.26 (7.07)	.128
Cholesterol (mg/d)	273.86 (243.73)	273.42 (238.45)	276.12 (241.19)	277.81 (241.65)	284.93 (267.19)	276.88 (244.45)	$<.001$
Breastfeeding during the first month of life ($n=\mathrm{96,852}$)	—	—	—	—	—	—	$<.001$
Mother’s milk only	29.70	44.28	35.96	46.51	40.28	41.85	—
Mother’s milk and infant formula	67.87	54.34	62.46	52.16	58.16	56.67	—
Infant formula only	2.43	1.38	1.58	1.33	1.57	1.48	—
Breastfeeding period until 12 months ($n=\mathrm{87,570}$)	8.79 (3.95)	9.84 (3.46)	9.34 (3.73)	9.88 (3.45)	9.43 (3.73)	9.63 (3.59)	$<.001$

Note: —, no data; BMI, body mass index; FFQ, Food Frequency Questionnaire; IPAQ, The International Physical Activity Questionnaire short form; K6, Kessler Psychological Distress Scale; SD, standard deviation; SF-8, The Short Form-8.

^a Proportion, mean (SD).

^b Other variables are shown in Table S1.

^c $p$-Values were calculated using ${\mathrm{\chi }}^2$ test or analysis of variance.

Table 2.

Maternal metallic element levels of this study subjects in the Japan Environment and Children’s Study (JECS).

	n	Unit	Mean	SD	25th percentile	Median	75th percentile
Lead	95,010	$\text{ng}/\mathrm{g}$	6.35	2.85	4.71	5.85	7.33
Cadmium	95,010	$\text{ng}/\mathrm{g}$	0.75	0.38	0.50	0.66	0.90
Mercury	95,010	$\text{ng}/\mathrm{g}$	4.20	2.49	2.54	3.64	5.19
Selenium	95,010	$\text{ng}/\mathrm{g}$	169.99	20.33	156.00	168.00	182.00
Manganese	95,010	$\text{ng}/\mathrm{g}$	15.96	4.67	12.60	15.40	18.70

Note: Metallic element levels were measured using whole blood samples. SD, standard deviation.

Trajectory Patterns in Weight SD Score

We identified five trajectories in weight SD score in the first 3 y of life: 4.74% of infants were classified in Group I, very small to small; 31.26% in Group II, moderately small; 21.91% in Group III, moderately small to moderately large; 28.06% in Group IV, moderately large to normal; and 14.03% in Group V, moderately large to large (Figure 2; Table S2). Children in Group I were born with a weight SD score of $-2.28$ ($\mathrm{2,088}\mathrm{g}$) and exhibited consistently light body weight throughout the follow-up period. Although children in Group II had weight SD scores that were similar to those in Group III [Group II, $-0.34$ ($\mathrm{2,864}\mathrm{g}$); Group III, $-0.47$ ($\mathrm{2,813}\mathrm{g}$)] at birth, their postnatal trajectory patterns differed. Children in Group II showed gradual body weight decline after birth such that they approached values observed in Group I at around age 2 y [Group II, $-0.78$ ($10.6\mathrm{kg}$ for boys and $10.1\mathrm{kg}$ for girls); Group I, $-1.03$ ($10.3\mathrm{kg}$ for boys and $10.0\mathrm{kg}$ for girls) at age 2 y]. Children in Group III reached the average weight until age 1 y and 0.48 weight SD score ($14.4\mathrm{kg}$ for boys and $13.9\mathrm{kg}$ for girls) at age 3 y. Group III showed estimated weight SD scores that were comparable to standard values both at birth and age 3 y. In contrast, children in Group IV and Group V had the same weight SD score of 0.80 ($\mathrm{3,300}\mathrm{g}$) at birth and remained above the average weight throughout the follow-up period. We confirmed that preliminary analysis without twins and triplets showed similar trajectory patterns in weight SD score (Table S3).

Figure 2.

Growth trajectories in estimated weight SD score in the first 3 y of life: estimating latent-class group-based trajectory models. Note: Group I, very small to small (black, $n=\mathrm{4,696}$); Group II, moderately small (gray dashed line, $n=\mathrm{30,954}$); Group III, moderately small to moderately large (gray, $n=\mathrm{21,696}$); Group IV, moderately large to normal (light gray dashed line, $n=\mathrm{27,781}$); and Group V, moderately large to large (light gray, $n=\mathrm{13,887}$).

Associations of Maternal Levels of Metallic Elements with Infant Growth Trajectory Patterns

Multinomial logistic regression models showed significant linear trends in quartiles of maternal levels of metallic elements with weight SD trajectory groups (Table 3). Higher quartiles of lead levels tended to be associated with larger ORs of being in Group I, and smaller ORs of being in Group IV and Group V, in comparison with Group III (Table 3; Figure 3). For cadmium, the highest level Q4 showed the low OR of being Group IV, whereas no clear trend was observed with the other trajectories (Table 3; Figure 4). Higher quartiles of mercury levels tended to be associated with decreased ORs of being in Group IV and Group V (Table 3; Figure 5). Higher quartiles of selenium levels tended to be associated with higher ORs of being in Group I and Group II (Table 3; Figure 6), whereas ORs of being Groups IV and V did not show clear relationships. Meanwhile, higher quartiles of manganese levels tended to be associated with increased ORs of being in Group IV and Group V (Table 3; Figure 7).

Table 3.

Associations between maternal metallic element levels and weight SD trajectory groups.

		Weight SD trajectory group
		Group I vs. III	Group II vs. III	Group IV vs. III	Group V vs. III
		Adjusted OR (95% CI)	Adjusted OR (95% CI)	Adjusted OR (95% CI)	Adjusted OR (95% CI)
Leada	Q1	Ref	Ref	Ref	Ref
	Q2	1.18 (1.04, 1.34)	0.95 (0.89, 1.00)	0.93 (0.87, 0.98)	0.90 (0.84, 0.97)
	Q3	1.15 (1.01, 1.30)	0.95 (0.90, 1.01)	0.86 (0.80, 0.91)	0.94 (0.87, 1.01)
	Q4	1.33 (1.18, 1.51)	0.93 (0.88, 0.98)	0.75 (0.71, 0.80)	0.80 (0.74, 0.86)
	Test for linear trend	$<.001$	.118	$<.001$	$<.001$
Cadmiumb	Q1	Ref	Ref	Ref	Ref
	Q2	0.91 (0.81, 1.03)	1.03 (0.97, 1.09)	0.98 (0.92, 1.04)	1.01 (0.94, 1.08)
	Q3	0.96 (0.84, 1.08)	0.98 (0.92, 1.04)	0.99 (0.93, 1.06)	1.02 (0.95, 1.10)
	Q4	0.99 (0.87, 1.13)	0.96 (0.91, 1.02)	0.88 (0.83, 0.94)	0.99 (0.92, 1.07)
	Test for linear trend	1.000	.248	.005	1.000
Mercuryc	Q1	Ref	Ref	Ref	Ref
	Q2	1.08 (0.95, 1.22)	1.06 (1.00, 1.12)	0.95 (0.89, 1.01)	0.99 (0.93, 1.07)
	Q3	1.14 (1.00, 1.29)	1.07 (1.01, 1.13)	0.92 (0.87, 0.98)	0.93 (0.86, 1.00)
	Q4	1.04 (0.92, 1.18)	1.02 (0.97, 1.09)	0.88 (0.83, 0.94)	0.90 (0.83, 0.96)
	Test for linear trend	1.000	1.000	$<.001$	.002
Seleniumd	Q1	Ref	Ref	Ref	Ref
	Q2	1.00 (0.89, 1.13)	1.06 (1.00, 1.12)	0.96 (0.90, 1.02)	0.91 (0.85, 0.98)
	Q3	1.10 (0.98, 1.25)	1.10 (1.04, 1.16)	0.95 (0.89, 1.01)	0.89 (0.83, 0.96)
	Q4	1.19 (1.06, 1.35)	1.11 (1.04, 1.17)	0.98 (0.92, 1.04)	0.93 (0.86, 1.00)
	Test for linear trend	.005	$<.001$	1.000	.154
Manganesee	Q1	Ref	Ref	Ref	Ref
	Q2	0.95 (0.84, 1.07)	1.10 (0.94, 1.05)	1.04 (0.98, 1.10)	1.05 (0.98, 1.13)
	Q3	0.90 (0.80, 1.02)	0.99 (0.93, 1.05)	1.04 (0.98, 1.10)	1.07 (0.99, 1.15)
	Q4	0.88 (0.77, 0.99)	1.04 (0.99, 1.11)	1.13 (1.06, 1.20)	1.16 (1.08, 1.25)
	Test for linear trend	.112	.943	.002	$<.001$

Note: Multinomial logistic regression model ($n=\mathrm{73,704}$). Dependent variable was weight SD trajectory group (vs. Group III). Independent variable was maternal metallic element. Adjusted for maternal age at delivery, infant sex, gestational duration, mode of delivery, primipara, medical and obstetric history (anemia, pregnancy hypertension, bronchial asthma, atopic dermatitis, drug allergy, hyperthyroidism/Basedow disease, hypothyroidism/Hashimoto’s disease, depression, other kidney disease, paramenia/menoxenia, endometriosis, hysteromyoma, ovarian tumor/ovarian cyst, polycystic ovary syndrome, other gynopathy, other abnormal pregnancy/abnormal delivery), BMI and height before pregnancy, weight gain during pregnancy, smoking at early pregnancy, drinking at early pregnancy, household income at midpregnancy, education at midpregnancy, IPAQ at early pregnancy, SF-8 at early pregnancy (general health, bodily pain), K6 at midpregnancy, FFQ at early pregnancy (energy), breastfeeding during the first month of life, breastfeeding period until 12 months. BMI, body mass index; CI, confidence interval; FFQ, Food Frequency Questionnaire; IPAQ, The International Physical Activity Questionnaire short form; K6, Kessler Psychological Distress Scale; OR, odds ratio; Q1, lowest quartile; Q2, second quartile; Q3, third quartile; Q4, highest quartile; Ref, reference; SD, standard deviation; SF-8, The Short Form-8.

^a Each quartile for lead represents the following (nanograms per gram): $\le 4.71$, 4.72–5.85, 5.86–7.33, $\ge 7.34$.

^b Each quartile for cadmium represents the following (nanograms per gram): $\le 0.495$, 0.496–0.662, 0.663–0.903, $\ge 0.904$.

^c Each quartile for mercury represents the following (nanograms per gram): $\le 2.54$, 2.55–3.64, 3.65–5.19, $\ge 5.20$.

^d Each quartile for selenium represents the following (nanograms per gram): $\le 156.0$, 157.0–168.0, 169.0–182.0, $\ge 183.0$.

^e Each quartile for manganese represents the following (nanograms per gram): $\le 12.6$, 12.7–15.4, 15.5–18.7, $\ge 18.8$.

Figure 3.

Associations of maternal lead levels with weight SD trajectory groups. Note: Multinomial logistic regression model. Independent variables of lead level are on the y-axis, and the odds ratios and 95% confidence intervals for weight SD trajectory group (vs. Group III) are on the x-axis. Each quartile for lead (nanograms per gram) represents the following: $\le 4.71$, 4.72–5.85, 5.86–7.33, $\ge 7.34$. Corresponding numerical results presented in Table 3. Q1, lowest quartile; Q2, second quartile; Q3, third quartile; Q4, highest quartile.

Figure 4.

Associations of maternal cadmium levels with weight SD trajectory groups. Note: Multinomial logistic regression model. Independent variables of cadmium level are on the y-axis, and the odds ratios and 95% confidence intervals for weight SD trajectory group (vs. Group III) are on the x-axis. Each quartile for cadmium (nanograms per gram) represents the following: $\le 0.495$, 0.496–0.662, 0.663–0.903, $\ge 0.904$. Corresponding numerical results presented in Table 3. Q1, lowest quartile; Q2, second quartile; Q3, third quartile; Q4, highest quartile.

Figure 5.

Associations of maternal mercury levels with weight SD trajectory groups. Note: Multinomial logistic regression model. Independent variables of mercury level are on the y-axis, and the odds ratios and 95% confidence intervals for weight SD trajectory group (vs Group III) are on the x-axis. Each quartile for mercury (nanograms per gram) represents the following: $\le 2.54$, 2.55–3.64, 3.65–5.19, $\ge 5.20$. Corresponding numerical results presented in Table 3. Q1, lowest quartile; Q2, second quartile; Q3, third quartile; Q4, highest quartile.

Figure 6.

Associations of maternal selenium levels with weight SD trajectory groups. Note: Multinomial logistic regression model. Independent variables of selenium level are on the y-axis, and the odds ratios and 95% confidence intervals for weight SD trajectory group (vs. Group III) are on the x-axis. Each quartile for selenium (nanograms per gram) represents the following: $\le 156.0$, 157.0–168.0, 169.0–182.0, $\ge 183.0$. Corresponding numerical results presented in Table 3. Q1, lowest quartile; Q2, second quartile; Q3, third quartile; Q4, highest quartile.

Figure 7.

Associations of maternal manganese levels with weight SD trajectory groups. Note: Multinomial logistic regression model. Independent variables of manganese level are on the y-axis, and the odds ratios and 95% confidence intervals for weight SD trajectory group (vs. Group III) are on the x-axis. Each quartile for manganese (nanograms per gram) represents the following: $\le 12.6$, 12.7–15.4, 15.5–18.7, $\ge 18.8$. Corresponding numerical results presented in Table 3. Q1, lowest quartile; Q2, second quartile; Q3, third quartile; Q4, highest quartile.

Discussion

Although previous studies have used birth cohort data to report growth trajectory patterns,^18,19,22 there are no data on weight SD score trajectory patterns that are comparable to those presented in this study. LBW is widely defined as a birth weight below $\mathrm{2,500}\mathrm{g}$, with 6% of births in East Asia and the Pacific categorized as LBW.² LBW prevalence in Japan has approximately doubled in the last three decades and was estimated to reach 5.3% in 2010.⁴⁷ In the present study of a nationwide birth cohort in Japan, 4.74% of infants were classified in Group I (very small to small), representing those with LBW (weight SD score below $-1.25$), a finding that largely overlaps with those reported in earlier studies. An interesting finding was that estimating latent-class group-based trajectory models also led to the classification of 31.26% of infants in Group II (moderately small), who fell within the normal weight range at birth but subsequently showed a gradual decline in weight such that they approached values observed in Group I at around age 2 y. Group I and Group II showed weight SD score below 0 consistently in the first 3 y of life, and we considered the infants belonging to Group I or Group II as showing poor growth. These findings indicate that even infants with normal weight at birth may exhibit poor postnatal growth, as indicated by a negative SD score, and may in turn have increased risk of adverse health outcomes, such as postrenal morbidity, neurological impairment, and chronic disease, in later life. Our findings emphasize that a large proportion—approximately one-third of children in Japan—corresponded to such a group.

After controlling for a comprehensive range of important confounders known to be risk factors for LBW, maternal levels of several metallic elements showed linear trend associations with infant growth trajectory patterns, even though live birth bias might lead to underestimation of the strength of exposure–outcome association.⁴⁸ First, higher maternal lead and selenium levels were associated with higher ORs of being in Group I. Previous studies have reported that maternal lead exposure during pregnancy is inversely associated with fetal growth. Irgens et al.⁴⁹ showed that the offspring of mothers who were occupationally exposed to lead were at higher risk of LBW than infants of women who were not. González-Cossío et al.⁵⁰ demonstrated that maternal bone-lead burden is inversely related to birth weight. A recent study from the JECS reported that even at a maternal blood lead level below $1.0\mathrm{\mu }\mathrm{g}/\text{dL}$, prenatal lead exposure was associated with decreased birth weight.⁵¹ The present results likewise showed that higher maternal lead levels were associated with increased risk of LBW and decreased risk of “moderately large” trajectories (Group IV and Group V) in offspring. Regarding selenium, earlier studies have reported inconsistent evidence. Makhoul et al.¹⁵ reported that higher umbilical cord selenium concentrations were significantly correlated with higher birth weight, whereas Nazemi et al.⁵² showed umbilical cord selenium concentrations that positively correlated with blood selenium concentrations were not different in low and adequate birth weight infants. Tsuzuki et al.⁵³ examined the association of maternal serum iron, zinc, copper, and selenium with birth weight and showed that only maternal selenium concentrations were positively correlated with neonatal birth weight among 44 infants. Results from the present study suggest that maternal exposure to selenium may not only be associated with LBW (Group I) but also with delayed negative effects on postnatal growth among infants with close-to-standard weight values at birth (Group II). Because the window between optimal and toxic concentrations of selenium for most living organisms is narrow,⁵⁴ cautious discussion is needed. The previous study reported that the median lead level in the JECS participants was similar to that of pregnant women in the United States and Canada, and the selenium level in the JECS was similar to that of pregnant Chinese women,²⁵ whereas the median selenium level was slightly higher than pregnant women in the United Kingdom and Australia.²⁵ Although the precise biological mechanisms of the effects of selenium are currently unclear, lead and selenium levels have strong positive associations with dietary intake,^15,55,56 which highlights the importance of control measures for preventing excessive lead and selenium intake in pregnant women.

Second, higher maternal mercury levels were negatively associated with the “moderately large” trajectories (Group IV and Group V) but had no clear association with LBW (Group I) in this study. Mercury is a well-known neurotoxicant, and exposure can cause adverse health effects. A previous studies reported that higher maternal mercury levels were associated with increased risk of small for gestational age (SGA),^57,58 and LBW^59,60; however, several studies reported nonsignificant association with birth weight.^12,61–63 Our results showed a partial negative effect of maternal mercury exposure on moderately large trajectories groups in the first 3 y of life. The fetus is highly sensitive to the toxic effects of mercury. After occupational exposure, dietary intake is the most important source of mercury exposure. In Japan, nonoccupational mercury exposure tends to be higher with high levels of fish and seafood consumption.⁶⁴ However, methylmercury was not measured in this study, and the source of mercury exposure cannot be identified. Fish and seafood are rich in protein, minerals, and omega-3 unsaturated fatty acids. Consuming seafood may be beneficial, despite potential contamination with methylmercury. Although the Ministry of Health, Labor and Welfare in Japan has created a pamphlet to call attention to excessive mercury intake among pregnant women,⁶⁵ it will be necessary to continue outreach activities to promote appropriate intake of seafood.

Third, the present study found that higher maternal manganese levels tended to be associated with increased ORs of the “moderately large” growth trajectory groups (Group IV and Group V). Although earlier studies have examined the association between maternal manganese levels and LBW, the results have been inconsistent, with some showing a linear association^66,67 and others showing a nonlinear association.^68–70 An earlier study from the JECS reported that nonlinear relationships between blood manganese level and birth weight only in male infants¹⁴ and maternal manganese levels were not related to risk of preterm birth.⁷¹ This study added evidence that higher maternal manganese levels contributed to the moderately large trajectory pattern in the first 3 y of life. Accumulating evidence indicates that manganese passes through the placenta via active transport and is necessary for fetal growth and development.^67,72 We suggest two potential mechanisms by which maternal manganese levels may be associated with the “moderately large” trajectory groups. First, manganese is a component of manganese superoxide dismutase (Mn-SOD),¹⁴ which is the body’s primary reactive oxygen species (ROS) scavenger.⁷³ Sufficient Mn-SOD may be needed to prevent ROS from affecting placental mitochondria, thereby enabling escape from placental cell damage and facilitating the transport of nutrients and oxygen to the fetus.^14,74 Second, manganese deficiency reduces the circulating insulin-like growth factor I (IGF-1) levels.¹⁴ Conversely, higher manganese levels may increase circulating IGF-1 levels. Increased maternal IGF-1 levels might augment placental mitogenesis and/or transplacental transfer of glucose, leading to fetal growth.^14,75 There is currently no domestic or international standard for blood manganese level. We suggest that defining a standard value, especially a lower limit of blood manganese and estimated average requirement for manganese during the gestation period, is a pressing task.

Fourth, an earlier systematic review and meta-analysis demonstrated that elevated maternal cadmium levels are associated with decreased birth weight and higher LBW risk⁷⁶; however, in the present study, we found that cadmium level had no clear linear trend association with weight SD trajectories. Previous study identified a negative correlation between maternal blood cadmium levels and birth weight in smoking pregnant women and highlights that cadmium may be a relevant biomarker for smoking toxicity on fetal development.⁷⁷ Another study showed that the association between higher maternal urinary cadmium levels and risk of preterm LBW was more pronounced among female infants than male infants.⁷⁸ Previous evidence suggests that maternal cadmium levels may affect infant growth among mothers with specific lifestyles and in infants with a specific gender. Future studies should examine the association between maternal cadmium levels and infant growth and trajectory patterns, stratified by characteristics such as mother’s lifestyle and infant’s gender.

The present study has some potential limitations. First, maternal blood samples were collected during the second or third trimester, at a mean (SD) gestation period of 27.3 (3.1) wk. Future studies should examine the critical period for determining postnatal growth patterns during the gestational period. Second, potential mechanisms underlying the association of maternal blood metallic elements, including lead, mercury, selenium, and manganese, with weight SD score trajectory remain to be examined in detail. Future studies should conduct genome analysis with a focus on transport protein families and the direct effect of metallic element levels in umbilical cord blood to clarify the potential mechanisms. Third, previous studies have typically measured selenium in serum and plasma, and cadmium was measured in urine. Because we only examined the association of postnatal growth patterns with selenium and cadmium in whole blood to illustrate the current exposure status, further research is needed to determine whether the association holds true in serum, plasma, and urine. Fourth, the present study did not identify the source of the metallic elements. Further study should examine important diet source that affect excessive metallic elements such as lead and selenium intake in pregnant women. Finally, this study assessed children’s weight with medical record transcripts at delivery to 1 month and self-administered questionnaires from age 0.5 to 3 y. We selected children’s weight data if they were measured within 3 months ($\text{highest\hspace{0.17em}frequency}\pm 1\text{\hspace{0.17em}month}$) at the time of each survey and accepted the data from the self-administered questionnaires that were within five SDs. The weight SD score trajectories in this study were based on two different data resources.

In conclusion, this nationwide government-funded birth cohort study identified five trajectories in weight SD score in the first 3 y of life, with 4.74% of infants classified in Group I, very small to small group; 31.26% in Group II, moderately small group; 21.91% in Group III, moderately small to moderately large group; 28.06% in Group IV, moderately large to normal group; and 14.03% in Group V, moderately large to large group. Multinomial logistic regression models showed that higher maternal lead and selenium levels tended to be associated with increased ORs of poor weight SD score trajectories, such as those observed in Group I and/or Group II, in comparison with Group III, highlighting the importance of control measures for preventing excessive lead and selenium dietary intake in pregnant women. Moreover, higher levels of mercury were associated with decreased ORs, and higher levels of manganese were associated with increased ORs of the “moderately large” trajectory groups (Group IV and Group V). We suggest that higher mercury levels may have negative effects on infant growth and that defining a standard value for blood manganese during the gestation period is a pressing task.