Study design, population, and settings
The data used in this study were obtained from the Japan Environment and Children’s Study (JECS), an ongoing cohort study that began in January 2011 to determine the effect of environmental factors on children’s health. The target number of enrolled pregnant women was 100,000. Partners were also recruited, although their participation was not mandatory. In the JECS, pregnant women were recruited between January 2011 and March 2014. The eligibility criteria for participants were as follows: (1) residing in the study area at the time of recruitment, (2) expected delivery after August 1, 2011, and (3) capable of comprehending the Japanese language and completing the self-administered structured questionnaire in Japanese. This study was registered in the UMIN Clinical Trials Registry (number UMIN000030786). Details of the JECS project have been described previously [16,17,18]. The JECS protocol was reviewed and approved by the Institutional Review Board on Epidemiological Studies of the Ministry of the Environment (ethical number 100910001) as well as by 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 each participant.
The present study was based on the “jecs-an-20180131” dataset released in March 2018 containing information on 98,255 mothers who had a singleton live birth, including 50,563 with the fathers’ registration. Specifically, we focused on questionnaire data regarding developmental screening as self-described by mothers when their child was 12 months old. The screening tool was the Ages and Stages Questionnaire, third edition (ASQ-3) . Maternal medical information, additional pregnancy details, and medical history were collected from subject medical record transcriptions for adoption as other covariates.
Information on socioeconomic status, smoking habit of the mother and partner, and maternal alcohol consumption during pregnancy was collected during the second/third trimester of pregnancy (T2) by means of self-reported questionnaires. Details on a parental history of neurodevelopmental disorders, epilepsy, and mental disease were also collected from T2 questionnaires as described by the mother and partner. Maternal anthropometric data before and during pregnancy, complications and medication during pregnancy related to HDP, diabetes mellitus/gestational diabetes mellitus (DM/GDM), and neonatal information was gathered from medical records. Pre-pregnancy BMI was calculated according to WHO standards as body weight (kg)/height (m)2 and categorized as underweight (BMI < 18.5), normal weight (BMI 18.5–24.9), overweight (BMI 25.0–29.9), and obese (BMI ≥ 30).
The main outcomes of interest were ASQ-3 domain scores at the age of 12 months. The ASQ-3 is a parent-reported comprehensive first-level developmental screening tool for children aged 1–66 months with 30 items in five domains: communication, gross motor, fine motor, problem-solving, and personal–social skills. Each item describes a skill, ability, or behavior to which the parent responds “yes” (10 points), “sometimes” (5 points), or “not yet” (0 points). Parents sometimes omit items when they are unsure of how to respond or because they have concerns about their child’s performance of the item. ASQ-3 scores were not calculated if there were three or more omitted items in a given domain. In the case of one or two omitted items, an adjusted total domain score was calculated by adding the averaged item score either once for one omission or twice for two omissions. The score calculated for each domain was categorized as normal development (above cutoff) or referral zone (below two standard deviations). The manual for the original ASQ recommends that a child be considered as screen positive if his/her score falls below the referral cutoff in any one of the five domains .
Participants with established risk factors of developmental delay, such as neonatal asphyxia, and physical abnormality at birth, including infection, respiratory distress, congenital abnormality, hearing disability, and chromosomal abnormalities, were excluded to investigate the effects of maternal GWG on neurodevelopment in infants without obvious underlying disease during the neonatal period (Fig. 1).
GWG in this study was subdivided as below, within, or above the reference values of the 2009 IOM guidelines widely used throughout the world. The IOM guideline ranges for total GWG based on pre-pregnancy BMI are as follows: 12.7–18.1 kg for underweight women, 11.3–15.9 kg for women of normal weight, 6.8–11.3 kg for overweight women, and 5.0–9.1 kg for obese women (Supplemental table S1).
The covariates in our models were selected a priori based on previous literature and biologic plausibility [20,21,22,23,24]. We estimated the effects of GWG after adjusting for demographic data including maternal age, pre-pregnancy BMI, parental smoking habit, maternal drinking habit, maternal highest level of education, annual household income, parental history of neurodevelopmental disorders, epilepsy, and mental disease, as well as obstetric and medical variables such as parity, means of pregnancy (including spontaneous pregnancy and assisted reproductive techniques, such as ovulation induction and artificial insemination or in vitro fertilization), use of folic acid supplements, complications during pregnancy (including DM/GDM, HDP, and intrauterine growth restriction), means of delivery, birth weight, gender, method of feeding, and neonatal jaundice in the newborn period requiring treatment such as phototherapy and exchange transfusion. Parental medical history of neurodevelopmental disorders included attention deficit hyperactivity disorder, learning disability, autism, Asperger’s syndrome, pervasive developmental disorder, and others. Parental history of mental disease included depression, schizophrenia, and anxiety disorder. Intrauterine growth restriction was defined as estimated fetal weight less than −1.5 standard deviations of standard weight based on gestational age in Japan.
Distribution normality was confirmed by the Kolmogorov–Smirnov test. Data are expressed as the mean ± standard deviation or the median (interquartile range) depending on whether they are normally distributed or not. Possible differences in maternal age, pre-pregnancy BMI, GWG, gestational age, and birth weight between subjects with normal development and developmental delay were assessed by the unpaired t-test or the Mann–Whitney U test based on the presence or absence of normal distribution, respectively. We also categorized continuous and ordinal variables, such as maternal age (< 35 or ≥ 35 years), pre-pregnancy BMI, GWG (below, within, or above), annual household income (< 4,000,000, 4,000,000–7,999,999, or ≥ 8,000,000 JPY), gestational age (< 37 or ≥ 37 weeks), and birth weight (< 1500, 1500–2499, or ≥ 2500 g). Fisher’s exact tests or chi-square tests were performed to compare covariates between groups stratified by category as well as by the presence of developmental delay. Additionally, differences in the scores of each domain among the three GWG groups were assessed by one-way repeated measures of analysis of variance (ANOVA) followed by post hoc (Bonferroni) testing. We employed multiple logistic regression models to investigate developmental delay at 1 year as the dependent variable in association with maternal GWG. Infants below and above the cutoff for each domain were categorized as “delayed” and “normal,” respectively. GWG was subdivided as below, within (reference), or above IOM guidelines. The models were adopted to calculate adjusted odds ratios (ORs) and their 95% confidence intervals (CIs) controlling covariates, as described above. Spearman’s rank correlation coefficient was used to check for multicollinearity of covariates. The variable of gestational age was excluded from the covariates because it was multicollinear with birth weight. Hosmer–Lemeshow testing was used to assess the goodness of fit of the models. We also analyzed the subjects without registered fathers to evaluate for possible selection bias.
All statistical analyses were performed using SPSS statistical software version 27 (SPSS Inc., Chicago, IL). All tests were two-tailed, and P-values of less than 0.05 were considered to indicate statistical significance.