Gestational weight gain mediates the effects of energy intake on birth weight among singleton pregnancies in the Japan Environment and Children’s Study

The JECS aims to explore environmental factors that affect the health and development of children and is being by the Environment, Japan. The detailed methodology and baseline profile of the study participants have been reported previously [16, 17]. Briefly, approximately 100,000 early pregnant women who live in 15 designated Study Areas across Japan were recruited between January 2011 and March 2014. They will be followed until the participating children reach 13 years of age. Participants were considered eligible if they met the following criteria: (1) they resided in a Study Area at the time of recruitment and were expected to reside continually in Japan, (2) they were expected to give birth after August 1, 2011, and (3) they were capable of understanding the Japanese language and completing a self-administered questionnaire.

This study uses the jecs-ag-20,160,424 dataset (released in June 2016) from the JECS, which comprises information from 104,102 fetal records. Therefore, formal sample size calculation was not performed in the present study. We restricted this study to 97,182 pregnancies resulting in live singleton births with no major birth defects. Then, we excluded participants based on predetermined exclusion criteria related to our exposure and outcome variables (Fig. 1): births before 28 weeks or after 42 weeks; missing or implausible daily energy intake (we used commonly applied values of < 500 or > 3500 kcal/day to define implausible daily energy intake [18, 19]); missing birth weight; implausible data on GWG (i.e. body weight recorded instead of GWG); and missing parity or newborn sex.

Nutritional information during pregnancy was obtained via a validated food frequency questionnaire (FFQ) [20, 21]. The FFQ was administered twice during pregnancy. The first time was in the first trimester to investigate the pre-pregnancy food frequency. The second FFQ conducted in mid-pregnancy (after week 22) was used in this study. The participants completed the FFQ by recalling their food consumption at the beginning of pregnancy (i.e., the point at which they realised that they were pregnant) and recording the quantity of each food item and the frequency with which it was consumed. Frequency categories ranged from ‘almost never’ to ‘7 or more times per day’ (or ‘10 glasses per day’ for beverages). Daily total energy intake was estimated using a food composition table that was developed based on the Standardised Tables of Food Composition in Japan (2010 edition) [22].

Anthropometric measurements of newborns, including birth weight (in grams) and gestational age at birth (in weeks), were extracted from medical records transcripts. Gestational age was calculated from the first day of the last menstrual period or by using ultrasound examinations of the fetus during the first trimester. Birth size was defined as small-for-gestational-age (SGA), appropriate-for-gestational-age (AGA), and large-for-gestational-age (LGA) for birth weights below the 10th percentile, 10th to 90th percentile, or above the 90th percentile, respectively, according to the Japanese parity and sex-specific neonatal birth weight chart for gestational age at birth [23].

Demographic, medical condition, and pregnancy-related data were collected through self-administered standard questionnaires or from medical records. GWG (mediator variable) was calculated by subtracting the self-reported pre-pregnancy body weight from the last measured weight closest to childbirth. Pre-pregnancy BMI was calculated by dividing the self-reported pre-pregnancy body weight (in kilograms) by the square of the height (in metres). The following information was also obtained: socioeconomic and lifestyle information, such as education, alcohol consumption, and smoking; occupation during pregnancy; physical activity levels (metabolic equivalent of a task measured as minutes per day) [24]; medical history of chronic diseases (diabetes, hypertension, heart disease, and kidney disease); nausea and vomiting during pregnancy; and receipt of health guidance.

Daily energy intake during pregnancy was stratified into three levels (low, moderate, and high) comprising approximately equal numbers of participants. The tertiles were the 33.4 and 67.9 percentiles, corresponding to the energy intake values of 1412 kcal and 1856 kcal, respectively. GWG was analysed as a continuous variable. Birth weight was studied as a continuous variable and birth size in categories (SGA, AGA, and LGA). Pre-pregnancy BMI was stratified as ‘underweight’ (< 18.5 kg/m²), ‘normal weight’ (18.5–24.9 kg/m²), and ‘overweight and obese’ (≥25 kg/m²). Because the Japanese physique is small and underweight by global standards, the Ministry of Health, Labour and Welfare (MHLW) has provided a guideline for pregnancy weight gain, calculated from the Japanese average and by pre-pregnancy BMI category. The participants were grouped into ‘low’, ‘appropriate’, and ‘high’ GWG categories (for descriptive statistics) within each pre-pregnancy BMI stratum. GWG was stratified by applying the appropriate GWG ranges of 9–12 kg and 7–12 kg for underweight and normal-weight pregnant women, respectively, according to the recommendations of the Ministry of Health, Labour, and Welfare, Japan [7]. Since there were no published recommendations for the overweight group, we used a GWG range of 5–7 kg, adopted from the Japanese Society for the Study of Obesity guidelines [25]. GWGs of ≤7 kg and ≤ 5 kg were applied for overweight (25–29.9 kg/m²) and obese (≥30 kg/m²) pregnant women, respectively. The participants were categorised into three age groups: ≤25, 26–35, and ≥ 36 years. Maternal educational attainment was stratified into three groups: ‘high school or less’, ‘vocational school/college’, and ‘university or higher education’. Considering that activity level may vary across job types [26], occupation during pregnancy was categorised as ‘unemployed’ (including students and women not classifiable by occupation), ‘low physical job’ (e.g. managers, engineers, and clerks), ‘moderate physical job’ (e.g. sales, services, and security workers), and ‘high physical job’ (e.g. manufacturing process workers, farmers, construction workers, carrying workers, and cleaners). Physical activity during pregnancy was grouped into three levels (low, moderate, and high) with approximately equal numbers of participants in each group. Other variables, such as alcohol consumption and smoking during pregnancy, chronic diseases, nausea and vomiting during pregnancy, and receipt of health guidance, were dichotomised as present (yes) or absent (no). Non-smoking and non-drinking women also included those who quit before or after they realised that they were pregnant.

Descriptive statistics of the participants were calculated for the entire study population and according to the levels of energy intake. The mean and standard deviation for continuous variables and numbers and percentages for categorical variables are reported. Since our primary aim was to examine how energy intake during pregnancy and GWG are associated with birth weight, we performed a mediation analysis. Figure 2 illustrates the conceptual framework of our study. We assumed energy intake preceded GWG in our analyses. The following points were the reasons for this assumption. First, the participants completed the FFQ by recalling their food consumption from conception (although FFQ was administered during mid-pregnancy). Second, the rate of weight gain is slower in the first trimester compared to the second or third [27] (although we used the total GWG). We compared the coefficients of the exposure (energy intake during pregnancy) derived from models with and without the mediator (GWG) (Fig. 2a) by employing the Karlson–Holm–Breen method using a linear regression model [28]. The Karlson–Holm–Breen method allowed, in a single command, the separation of the total effect of energy intake (c) into direct (ć) and indirect effects mediated by GWG (c – ć) while controlling for potential confounding variables. Based on the literature [1‐5, 29, 30] and considering the causal relationship with exposure and outcome variables (Fig. 2b), the following factors were considered potential confounders and controlled for in the mediation analysis: maternal age (years), parity, educational level, pre-pregnancy BMI (kg/m²), type of occupation and physical activity level during pregnancy, alcohol consumption and smoking during pregnancy, the presence of chronic diseases, receipt of health guidance, and nausea and vomiting during pregnancy. To increase the precision of the estimates, newborn sex and gestational age (weeks) were also adjusted. Next, we evaluated the associations between levels of energy intake (moderate energy intake as the referent) and birth size using a multinomial logistic regression model. The crude and adjusted relative risk ratios (RRR), adjusted risk differences, and corresponding 95% confidence intervals (CIs) for SGA and LGA births were reported. The same factors included in the mediation analysis, except newborn sex and gestational age, were included in the adjusted model. Additionally, we examined the characteristics of the women according to their inclusion status in our analysis. All statistical analyses were performed using Stata/MP version 15.1 (StataCorp., College Station, TX, USA). Statistical significance was defined as a two-sided P-value < 0.05.

This study was conducted according to the principles expressed in the Declaration of Helsinki. All procedures involving human participants were approved by the Japan Ministry of the Environment’s Institutional Review Board on Epidemiological Studies (No. 100910001) and the Ethics Committees of all participating institutions. Written informed consent was obtained from all participants.