Temporal relationships between maternal metabolic parameters with neonatal adiposity in women with obesity differ by neonatal sex: Secondary analysis of the DALI study

Summary Objectives To investigate the importance of time in pregnancy and neonatal sex on the association between maternal metabolic parameters and neonatal sum of skinfolds. Methods This was a longitudinal, secondary analysis of the vitamin D and lifestyle intervention for gestational diabetes mellitus study, conducted in nine European countries during 2012 to 2015. Pregnant women with a pre‐pregnancy body mass index (BMI) of ≥29 kg/m2 were invited to participate. We measured 14 maternal metabolic parameters at three times during pregnancy: <20 weeks, 24 to 28 weeks, and 35 to 37 weeks of gestation. The sum of four skinfolds assessed within 2 days after birth was the measure of neonatal adiposity. Results In total, 458 mother‐infant pairs (50.2% female infants) were included. Insulin resistance (fasting insulin and HOMA‐index of insulin resistance) in early pregnancy was an important predictor for boys' sum of skinfolds, in addition to fasting glucose and maternal adiposity (leptin, BMI and neck circumference) throughout pregnancy. In girls, maternal lipids (triglycerides and fatty acids) in the first half of pregnancy were important predictors of sum of skinfolds, as well as fasting glucose in the second half of pregnancy. Conclusions Associations between maternal metabolic parameters and neonatal adiposity vary between different periods during pregnancy. This time‐dependency is different between sexes, suggesting different growth strategies.


| INTRODUCTION
Globally, in 2017 an estimated 38 million children under 5 years of age presented either overweight or obesity. This is a major public health concern. 1 Hence, strategies have to be developed to reduce this burden and prevent childhood obesity. This requires a thorough identification and understanding of the underlying determinants.
Maternal obesity in pregnancy is linked with the development of neonatal and childhood adiposity. [2][3][4][5] Although there is much interest in the relationship between maternal obesity and childhood adiposity, maternal metabolic factors driving the increase in offspring adiposity have not been fully understood.
The Hyperglycemia and Adverse Pregnancy Outcome study showed a continuous positive relationship of maternal glucose in the second trimester with measures of neonatal body fat. 6 Findings from the Healthy Start Study support this and, in addition, showed temporal changes in the association of glucose and neonatal fat, with stronger associations in the second half of pregnancy compared to early pregnancy. 7,8 However, other studies clearly indicate that the influence of maternal phenotype on foetal growth can already be demonstrated in the early pregnancy period: foetal abdominal circumference, as proxy for foetal overgrowth, is increased in women with obesity and/or gestational diabetes (GDM) already at 20 weeks of gestation or even before 9,10 ; maternal insulin resistance in the first half of pregnancy is related to neonatal fat percentage, 7 and fasting glucose in the first trimester with the risk of a large-for-gestational age baby. 11 Although these studies suggest an influence of the early pregnancy period, surprisingly little is known about temporal relations between maternal metabolic measures and neonatal fat accrual. 12 Female neonates are known to be more insulin resistant, 13,14 which might explain why males are more affected by in utero exposure to gestational diabetes. 9 Therefore, it is mandatory to assess temporal relations of maternal metabolism with neonatal adiposity in a sex-dependent manner.
In this study, we investigated the association between maternal health measures at different times during pregnancy and neonatal sum of skinfolds in a sex-specific manner. We hypothesized that: (a) associations between maternal metabolic parameters and neonatal adiposity differ at different time points, and (b) these temporal differences are sex-specific. We tested these hypotheses in 458 motherinfant pairs in the vitamin D and lifestyle intervention for gestational diabetes mellitus (DALI) study, a pan-European study originally designed for the prevention of gestational diabetes. 15 2 | METHODS

| Design and participants
This is a longitudinal, secondary analysis of the DALI study, which was a multicentre parallel randomized trial conducted in nine European countries (Austria, Belgium, Denmark, Ireland, Italy, Netherlands, Poland, Spain and United Kingdom) during 2012 to 2015. The study was prospectively registered as a randomized clinical trial (RCT) with the primary aim to prevent gestational diabetes mellitus on November 21, 2011 (ISRCTN70595832).
Local ethics committee approval and written informed consent of all women was obtained. Pregnant women with a pre-pregnancy body mass index (BMI) of ≥29 kg/m 2 , <20 weeks of gestation, a singleton pregnancy and aged ≥18 years were invited to participate. 16 Exclusion criteria included diagnosis with early gestational diabetes mellitus, 17,18 preexisting diabetes, and chronic medical conditions.

| Design and procedures
The study was originally designed as an RCT with the following groups, pre-stratified for site: (a) healthy eating; (b) physical activity; (c) healthy eating + physical activity; (d) healthy eating + physical activity + vitamin D; (e) healthy eating + physical activity + placebo; (f) vitamin D; (g) placebo, (h) control. Staff involved with measurements, but not participants, were blinded to the lifestyle intervention.
Both staff and participants were blinded to vitamin D intervention.
For the purpose of this analysis, the data were analyzed as a longitudinal cohort. Maternal measurements took place at baseline (<20 weeks), at 24 to 28 and at 35 to 37 weeks of gestation. Since methodology has been extensively described elsewhere, 16 only variables of interest will be detailed in this manuscript.

| Neonatal outcome
Triceps, subscapular, supra-iliac and quadriceps skinfolds were measured within 48 hours of birth with a Harpenden skinfold calliper (Baty, UK), and the values summed to obtain the primary neonatal outcome measure, the sum of skinfolds. Each skinfold measurement was measured twice and if a difference of more than 0.2 mm was registered, a third measurement was performed and the average of the three was taken. The neonatal age at the measurement was the time between birth and measurements, which was registered in hours.

| Maternal metabolic and adiposity parameters
Maternal height was determined at baseline with a stadiometer (SECA 206; SECA, UK). Women were weighed on calibrated electronic scales (SECA 888 and 877, SECA, UK) at baseline (<20 weeks), 24 to 28 weeks, and at 35 to 37 weeks of gestation. BMI was calculated as weight in kilogram divided by the square of height in metres. Gestational weight gain was defined as the change in objectively measured weight from pre-pregnancy to <20, <20 to 24-28 weeks and from 24-28 to 35-37 weeks. Neck circumference was obtained in a standing relaxed upright position between mid-cervical spine and midanterior neck, to within 1 mm. 19 After fasting for 10 hours, blood was collected and mothers consumed a 250 mL 75 g glucose drink (within a period of 5 minutes).

| Covariates
Information on possible covariates was collected in the baseline questionnaire or from medical files: national site(s) of recruitment, maternal age, gestational age during pregnancy (<20, 24 to 28 and 35 to 37 weeks), maternal ethnicity (European or non-European descent), maternal education (low, medium and high), smoking status at 35 to 37 weeks of gestation (yes/no), pre-pregnancy BMI and gestational age at birth.

| Statistical analyses
All analyses were performed in STATA version 13 for windows (StataCorp LP, College Station, Texas) and a 5% type I error rate was used for the analyses. We have 80% statistical power to detect associations with small effect sizes (f 2 = 0.04) considering the number of participants evaluated (n = 458) and the number of exposures and confounding variables in each of our models.
Skewed variables were log-transformed before analyses (fasting insulin, 1-hour insulin, 2-hour insulin, HOMA-IR, Stumvoll first phase, Stumvoll second phase, triglycerides, leptin and BMI). Independent t tests were performed to evaluate descriptive differences in the mother's metabolic and adiposity parameters for each pregnancy period in relation to their child's sex (results shown in Table 1). Independent t tests also evaluated sex-differences in neonatal skinfolds.    Table S1 presents the cumulative variance of each factor in the three periods of gestation. Fasting glucose, triglycerides, free fatty acids and gestational weight gain did not have a high loading factor for any of the factors, nor did they built a separate factor, and were not included in any of the factors. Thus, they were analyzed as single exposures in relation to neonatal sum of skinfolds.
The association of each resulting factor from the mother's health profile with neonatal sum of skinfolds was assessed using linear multilevel analyses, adjusted for intervention group, site of recruitment, gestational age during pregnancy, gestational age at birth, neonatal age at measurement, maternal ethnicity, maternal education, BMI (except in factor 4 since BMI was part of this factor), maternal age, smoking status and the cluster structure. Sex-effects were estimated by inclusion of an interaction term between the factor and sex. We used robust correction for estimating the standard errors and estimated P-values using the Bonferroni adjustment for multiple testing.
The association between the mother's health profile not included in any of the factors (fasting glucose, triglycerides, free fatty acids and gestational weight gain) and neonatal sum of skinfolds was evaluated following the same procedure abovementioned. Note that we tested the same analysis replacing BMI for gestational weight gain as a confounder. The conclusions were similar, so we only present the results with BMI adjustment.

| Supplementary analysis
Multiple linear multilevel analyses tested the association between each of the maternal measurements during pregnancy and neonatal sum of skinfolds (see Table S2). Sex-effects were estimated by inclu-

| Associations of maternal metabolic parameters with neonatal adiposity
Associations of the four factors, and fasting glucose, triglycerides, fatty acids and gestational weight gain at the three time points with neonatal adiposity are described in Table 2. were associated with sum of skinfolds in girls at <20 and 24 to 28 weeks. Gestational weight gain at <20 weeks was positively associated with boys' sum of skinfolds (Table 2). Table S2 describes the associations of maternal parameters that were included in the factors with neonatal adiposity.

| DISCUSSION
The main aim of the present study was to identify the effect of some maternal metabolism on neonatal skinfolds at different times in pregnancy in a sex-dependent manner. Results confirmed our hypotheses that associations of metabolic parameters with neonatal adiposity change between different time-periods and differ between sexes. We demonstrate a complex pattern of metabolic metabolism and its association with neonatal fat measured as sum of skinfolds, as illustrated in Figure 2. In summary, insulin resistance in early pregnancy is an important predictor for boys' sum of skinfolds, in addition to fasting glucose throughout pregnancy. In girls, maternal lipids in the first half of pregnancy period play a role for sum of skinfolds, as well as fasting glucose in the second half of pregnancy.
In both sexes, the early pregnancy period is relevant for neonatal adiposity, although for different metabolic parameters (ie, insulin resistance in boys and lipid levels in girls). Although a later intervention was effective in reducing neonatal adiposity, 22 our data suggest that ideally maternal metabolism shall be normalized already early in pregnancy to reduce the risk of neonatal adiposity. Hence, interventions beginning prior to or in very early pregnancy might be even more effective than those initiated later, and achieved a 9% reduction in neonatal fat. 22 Our results also indicate that both types of nutrients, glucose and lipids, are related to neonatal adiposity, although associations are time-and sex-dependent. This argues for future time-and sex-specific studies to delineate metabolic pathways between maternal adiposity, insulin resistance, the different nutrients and neonatal adiposity.
The physiological pathways influencing neonatal adiposity originate from maternal insulin resistance, which is higher with increasing maternal BMI. Indeed, in the Healthy Start Study 7 maternal insulin resistance in the first half of pregnancy was an independent predictor T A B L E 2 Multilevel regression coefficients of the association between maternal health profile components and child's sum of skinfolds (mm) by sex in three periods of gestation Note: Adjusted for intervention group, site of recruitment, maternal ethnicity, education, BMI (except when BMI was the exposure), age and smoking status, gestational age during pregnancy (weeks), gestational age at birth (weeks) and neonatal age at measurement (hours), and cluster structure (individuals nested in site of measurement). Besides the aforementioned adjustments, HOMA-index was also an adjustment when triglycerides and fatty acids were the exposures. P values are Bonferroni adjusted. Abbreviation: BMI, body mass index.
of neonatal adiposity. This is in principle in line with our findings of insulin resistance (represented by factor 1) in early pregnancy playing a role, but in our study, this was limited to boys only. Later in pregnancy, the positive association of insulin resistance with neonatal adiposity was found in both boys and girls. The strong determining role of early insulin resistance for neonatal adiposity is not without precedent and extends to the pre-pregnancy period, in which insulin resistance had the strongest association with neonatal fat mass, more so than insulin resistance in late pregnancy. 23 However, neither of those previous studies tested for sex differences in this association or, specifically, in pregnant women with obesity. Two studies that analyzed boys and girls separately 24,25 found associations of insulin resistance late in pregnancy with measures of adiposity in girls, but not in boys, which is not fully in line with our findings. The different finding might be explained by differences in maternal BMI or the participants' level of glucose tolerance, since those other studies were not limited to women with obesity without GDM at baseline, as in our study sample. 24,25 The pathway from insulin resistance to neonatal adiposity might be through one of the nutrients glucose or lipids. 8   F I G U R E 2 Summarized visualization of the degree of importance of some maternal parameters in relation to neonatal sum of skinfolds by sex only one study evaluated exclusively pregnant women with obesity.
Thus, more information is needed to draw any conclusions on the role of insulin secretory response or glucose tolerance for neonatal adiposity.
Maternal adiposity, mostly assessed by the pre-pregnancy BMI, has been related with neonatal adiposity in numerous studies, 4,5,25,31-33 although one reported no association. 34 We found maternal adiposity to be associated with neonatal adiposity in boys throughout pregnancy, but no association was found in girls.
Although, when analyzed individually, maternal BMI in the second half of pregnancy period was associated with the girls' adiposity. Different from many of the previously mentioned studies, we only had women with obesity in our sample, which might have precluded finding strong associations with maternal adiposity. Reports on sex differences in the association of maternal adiposity and neonatal adiposity are conflicting: Some find associations only in boys, 35 but other studies only in girls. 36,37 Prior studies in pregnant women with diabetes have reported an interaction of foetal sex with a number of maternal variables in the prediction of large or small-for-gestational age new-borns but results are not comparable since the outcome variable was not new-born adiposity and maternal variables did not include lipids or insulin sensitivity. 38

| Sex and time differences
Conceptually, male and female foetuses follow different growth strategies in utero. 39,40 Mechanisms are not well understood, but male embryos have more rapid cell divisions compared to females, 41 which could lead to a more rapid growth. 39,40 Their higher growth rate might explain why male foetuses are also more responsive to changes in nutrient supply. 39,40 In addition, the peak growth velocity of males seems to be later in pregnancy compared to female foetuses. 39 These different growth strategies might result in differences in nutritional needs, at different times in pregnancy. The results of our study provide strong evidence to support this concept. However, more systematic assessments of time-and sex-dependent associations of maternal metabolism with neonatal adiposity are needed in future studies.
Although foetal fat is mostly accumulated late in pregnancy, 42,43 our data highlight that also the early pregnancy period is important for neonatal adiposity. This finding is in line with other studies, showing that maternal metabolism that led to gestational diabetes diagnosis later in pregnancy, had an influence on foetal abdominal circumference already at 17 weeks of pregnancy, with a more pronounced effect in males. 9 Also maternal obesity was related to accelerated growth at 20 weeks of gestation, 10 without sex differences studied.

| Strengths and limitations
The European representativeness of our study sample with trials sites spread over Europe is a strength, but the restriction to women with obesity can be seen as a limitation of our study. However, since the prevalence of pregnant women with obesity is alarming, 44 studying the consequences of maternal obesity is of public health relevance. A strength of our study is the measurements of maternal metabolic parameters at three time points in pregnancy. Without this elaborate approach, identification of time-dependent effects had not been possible.
DALI study was designed to evaluate the effects of different interventions on preventing gestational diabetes in pregnant women with obesity. Although we adjusted the analyses for the various interventions, we cannot fully exclude a minor influence on the results.
While more direct measures of neonatal adiposity by for example, air displacement plethysmography, would have been preferred, we recently reported an association of sum of skinfolds with cord blood leptin in this cohort, 22 supporting the validity of the skinfold measures as proxy for neonatal adiposity.

| CONCLUSION
Our results show time-dependent associations of maternal metabolic parameters with neonatal adiposity. Importantly, and a novel key result, we demonstrate that this time-dependence varies between sexes. Thus, the study highlights the urgent need for inclusion of neonatal sex in all analyses, not as confounder, but as major determinant and modifying factor. This has been recommended previously. 45,46 The present study can inform future studies to include larger sample sizes, because of the sex differences, but also measurements of maternal metabolism at more than one time point during pregnancy. made substantial contribution to the conception of the DALI study and acquisition of data. All authors revised the manuscript and contributed to the content. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.