Maternal overweight is not an independent risk factor for increased birth weight, leptin and insulin in newborns of gestational diabetic women: observations from the prospective ‘EaCH’ cohort study

The ‘EaCH’ approach is a prospective observational cohort study in mother-child dyads with maternal GDM and/or overweight/obesity, with the primary focus on early developmental origins of unfavorable health outcomes through pre- and/or early postnatal exposure to a ‘diabetogenic/adipogenic’ environment. The primary aim of ‘EaCH’ is to identify metabolic, hormonal, nutritional, epigenetic and other causal factors/determinants that are critically associated with pre- and neonatal acquired disease susceptibility. The overall goal is to deliver new mechanistic and/or preventive approaches and strategies to reduce disease risk in affected offspring.

The study includes a final GDM cohort of 205 mother-child dyads (75% of eligible cases, Fig. 1), from initially 562 women with GDM and delivery between June 2007 and December 2010 at the Clinic of Obstetrics, Charité – Universitätsmedizin Berlin, Campus Virchow-Klinikum, Germany (Flow chart, including exclusion criteria, shown in Fig. 1). A major exclusion criterion was non-sufficient German/English speaking, due to a relatively high proportion of immigrants attending our clinic. Research design and methods were conducted in accordance with the Declaration of Helsinki, revised in 2004 [22], and approved by the local Ethics Committee (EA2/026/04). All participants provided written informed consent before inclusion into the study.

Respective data were collected at the woman’s first visit at our clinic by personal interview via standardized questionnaire. Parental educational background, current employment status (before maternity leave) and job description were recorded and used for socio-economic status (SES) categorization, as applied previously [23, 24]. The resulting SES categories were further assigned to lower (categories: unemployed, manual worker, non-manual worker without university degree) or higher SES (category: non-manual worker with university degree).

Paternal height and weight was recorded via standardized questionnaire at enrollment and the BMI (kg/m²) was calculated. Maternal height and weight before conception, weight change in the 1st, 2nd and 3rd trimester as well as the last weight measured within one week prior to delivery were abstracted/calculated from the ‘Mutterpass’. This ‘Mutterpass’ is a standardized maternity record/documentation in Germany, containing essential information about pregravid health/conditions, regular screening/medical examinations throughout pregnancy as well as clinically important perinatal data [25]. Prepregnancy BMI was calculated using pregravid weight and height and then categorized according to WHO criteria (normal weight: 18.5–24.9 kg/m², overweight: 25.0–29.9 kg/m², obese: ≥30.0 kg/m²). Total gestational weight gain (GWG) was calculated as the difference between prepregnancy weight and nearest weight to delivery and categorized in line with the Institute of Medicine (IOM) guidelines from 2009 using the labels inadequate, adequate and excessive weight gain, respectively [26]. As an indicator of the genuine maternal weight gain during pregnancy, net GWG was calculated by subtracting birth weight and placental weight from total GWG.

Gestational diabetes screening was conducted generally between the 24th–28th week of pregnancy (mean ± SD: 25.6 ± 3.8) using 75-g oral glucose tolerance test (oGTT) according to the guidelines of the German Society of Gynecology and Obstetrics at the time of recruitment [27]. Respective guidelines were based on the Carpenter and Coustan criteria [28]. GDM was diagnosed if two values were equal to or above either capillary fasting glucose (FG) ≥90 mg/dL (5.0 mmol/L) and/or 1-h glucose ≥180 mg/dL (10.0 mmol/L) and/or 2-h glucose ≥155 mg/dL (8.6 mmol/L). Subsequent diabetic care was provided according to above mentioned guidelines, aiming especially for the following glucose targets: FG 60–90 mg/dL (3.3–5.0 mmol/L) and 1-h postprandial glucose (PPG) ≤140 mg/dL (7.8 mmol/L). Among all GDM cases, 63% (n = 129) were treated by diet and moderate physical activity only, while 37% (n = 76) received additional insulin therapy to achieve glycemic control. Insulin-sensitizing drugs (e.g. metformin) were not applied here. Overall, women were instructed to monitor blood glucose at least 4-times daily, i.e., morning FG and three 1-h PPG values, using the Accu-Chek glucose meter (Roche Diagnostics, Mannheim, Germany). From these individual profiles, mean FG (based on three values) and mean PPG levels (based on nine values) were calculated from three days within the 32nd and 36th week of gestation, respectively. As marker for long-term glycemic status, glycated hemoglobin A1c (HbA1c) was assessed in the third trimester (mean ± SD: 33.2 ± 2.5 weeks’ gestation), and, additionally, at delivery by standardized high-performance liquid chromatography (Variant II System, Bio-Rad, Hercules, CA, USA) at the clinic’s central laboratory. Based on recent guidelines from the American Diabetes Association [29], an HbA1c value above 6% was used as indicator of poorer metabolic control for evaluations.

A variety of maternal parameters, including parity, history of GDM, comorbidities (e.g. gestational hypertension), mode of delivery, gestational age at birth, were abstracted from the standardized ‘Mutterpass’ and medical records, respectively. Gestational hypertension and preeclampsia were diagnosed according to the guidelines of the German Society of Gynecology and Obstetrics [30]. Mode of delivery was defined according to HAPO study criteria [31], i.e., spontaneous, vaginal-operative, primary or repeat Cesarean section (CS). Gestational age was estimated based on woman’s last menstrual period in combination with subsequent ultrasound examinations at first trimester.

Information about infant’s sex, preterm birth (< 37 weeks’ gestation), placental weight, birth weight, length and head circumference was abstracted from medical records. Placental weight was measured immediately after delivery with umbilical cord attached. Additional anthropometric outcomes included relative birth weight (g/cm), ponderal index (g/cm³ × 100), and macrosomia (defined as birth weight ≥ 4000 g). Furthermore, based on sex- and ethnic-specific birth weight for gestational age percentiles [32‐34], infants were categorized into small-for-gestational age (SGA, <10th percentile), appropriate-for-gestational age (AGA, 10th–90th percentile) or large-for-gestational age (LGA, > 90th percentile). Occurrence of further adverse neonatal outcomes, e.g. hypoglycemia and/or shoulder dystocia, was retrieved from medical records.

Venous umbilical cord-blood was collected immediately after delivery and plasma samples stored at − 80 °C. Commercially available radioimmunoassays were used for analyses of insulin (Cat# 10624, Radim Diagnostics, Pomezia RM, Italy), C-peptide and leptin (C-peptide: Cat# RIA-1252, leptin: Cat# RIA-1624, DRG Instruments, Marburg, Germany) according to manufacturers’ instructions. The 90th percentile cut-offs in this study were the following: insulin 34.2 μU/mL, C-peptide 2.2 ng/mL, and leptin 29.6 ng/mL.

Descriptive statistics are presented as means ± standard deviations (SD) or numbers and percentages, respectively. All continuous variables were tested for normal distribution using Kolmogorov-Smirnov test and evaluation of distribution plots. Group comparisons were analyzed by one-way ANOVA followed by Tukey’s post hoc test or Kruskal-Wallis test followed by Dunn’s post hoc test and Fisher’s exact or Chi-square tests, respectively. Spearman’s correlation coefficients (r) and/or linear regression models (unstandardized B coefficients with 95% confidence intervals [95% CI]) were calculated to assess relationships between paternal and/or maternal anthropometric/metabolic parameters and neonatal anthropometric and cord-blood hormone outcomes. For regression analyses, normal distribution was achieved for respective variables by transformation using the method described by Templeton [35]. Furthermore, categorical variables were dichotomized, e.g. mode of delivery was grouped into deliveries with vs. without CS etc., if required. Overall, collinearity was controlled by bivariate correlation (r < 0.7) and variance inflation factor (< 2.5). The following parameters were tested as independent variables for their association with birth outcomes. Anthropometry: maternal prepregnancy weight and BMI, GWG in 1st, 2nd and 3rd trimester, total and net GWG; Metabolism: FG, 1-h and 2-h glucose, and area under the curve (AUC) of glucose at oGTT, mean FG and PPG at 32nd and 36th weeks’ gestation, HbA1c in 3rd trimester and at delivery. Birth outcomes were: placental weight, birth weight, relative birth weight, cord-blood insulin and leptin. All regressions are shown as crude and adjusted models. Model I included general parameters: maternal age, SES, ethnic origin, smoking in pregnancy, parity, height (except for pre-pregnancy BMI), type of therapy (diet vs. insulin), gestational age, mode of delivery, and infant’s sex. To evaluate independent associations of maternal anthropometry and metabolism with birth outcomes, further adjustment models were applied. For maternal anthropometry, model II contained all parameters of model I plus mean FG and PPG at 32nd and 36th weeks’ gestation and HbA1c partum. Accordingly, the applied model II for maternal metabolism included all parameters of model I plus prepregnancy weight and net GWG. In addition, forward step-wise regression analyses were carried out to identify categorical predictors of anthropometric/hormonal birth outcomes, considering paternal and maternal parameters (max. n = 118–153). In line with previous studies in this research area [5, 36], adjusted semipartial correlation coefficients (Sr²) were calculated for each individual (potential) predictor entered into the final models. These coefficients estimate the contribution of each factor to the total variability of the outcome variable. Finally, owing to small numbers of unfavorable categorical anthropometric and cord-blood hormone outcomes logistic regression analyses (odds ratio, OR) were just performed for LGA (n = 41 in the total cohort), using the same adjustment models as described above. Additionally, percentages of relevant maternal anthropometric and metabolic variables/exposures in newborns with unfavorable anthropometric and cord-blood hormone outcomes were compared with those in reference groups that had no respective outcomes. In general, missing data analysis according to Little’s MCAR test was performed and indicated random distribution. A P-value < 0.05 was considered significant (two-tailed). All calculations were performed using SPSS v. 24.0 (IBM Co., Armonk, NY, USA).

Major parameters across maternal prepregnancy BMI categories are presented in Table 1. Overall, the mean paternal BMI as well as paternal SES were in alignment with those of the mothers, with highest BMI and lowest SES in partners of obese mothers (Table 1). Moreover, there was a positive correlation between paternal and maternal BMIs (r = 0.23, P < 0.01).

Maternal data showed no difference in age across prepregnancy BMI groups (Table 1). Overall, the majority of women of the total cohort was of European origin and belonged to the lower SES category, while significantly more individuals in the overweight/obese groups were of Non-European origin and had higher frequencies of lower SES. There was no difference in smoking during pregnancy between normal weight and overweight (but not obese) women; however, smoking was much more abundant in the obese group. Similarly, parity was higher in obese individuals. Amongst multiparous women, more than one-third had a history of GDM, while a non-significant trend towards higher frequencies of prior GDM was observed in overweight/obese groups compared to normal weight mothers.

The GDM cohort as a whole was overweight, in terms of mean BMI, before pregnancy. Around 40% were classified as normal weight and around 30% as obese. More than half of the women in the overweight/obese groups showed excessive GWG, according to IOM criteria, as compared to only around one-third in normal weight women. Furthermore, there was a continuous decrease in inadequate GWG across prepregnancy BMI groups. Total GWG was lowest in obese women, while normal weight and overweight mothers gained similar amounts of weight during gestation. However, the calculated net GWG, i.e., excluding placental and newborn weight, showed a decrease across all BMI categories (Table 1). Both total and net GWG correlated inversely with prepregnancy BMI (total GWG: r = − 0.30, P < 0.001; net GWG: r = − 0.32, P < 0.001).

In the total cohort, GDM screening revealed that the majority of women exceeded FG and 1-h glucose cut-offs while only one-third surpassed the 2-h glucose threshold at oGTT (Table 1). Among oGTT measurements, only FG and frequencies above its cut-off showed significant increase across BMI categories. In 3rd trimester, an overall decrease of mean FG and PPG values, and, thus, a reduction of frequencies above their respective guideline cut-offs was observed. Of note, application of guideline cut-offs (FG > 90 mg/dL; PPG > 140 mg/dL) regarding 3rd trimester values revealed that to a greater extent FG rather than PPG was less controlled. In line with this, among all glucose values at oGTT, only FG correlated significantly with the respective mean glucose levels later in pregnancy (FG at 32nd weeks: r = 0.40, P < 0.001; FG at 36th weeks: r = 0.27, P < 0.001). Compared to normal weight women, insulin therapy was more abundant in the overweight and obese groups, respectively. Each glucose variable in 3rd trimester, i.e., mean glucose and HbA1c levels, increased significantly with maternal prepregnancy BMI, in contrast to oGTT variables at 24–28 weeks’ gestation (Table 1).

Placental weight was significantly higher in the overweight/obese groups as compared to normal weight women (Table 1). Likewise, birth weight, relative birth weight and ponderal index were higher in overweight/obese individuals. However, instead of a further increase in the obese as compared to overweight women these anthropometric measures plateaued. Moreover, birth weight and relative birth weight were highest in the overweight group. Frequencies of adverse anthropometric outcomes were more abundant in the overweight/obese groups, with the exception of SGA as there was no significant difference across BMI categories. Overweight women had the highest rate of LGA and macrosomia. In obese, the percentage of LGA was similar to the overweight group, however, the rate of macrosomia was between the normal weight and overweight category. In contrast to these observations regarding anthropometric birth outcomes, cord-blood levels of C-peptide, insulin, and leptin continuously increased across BMI categories. Accordingly, the same pattern was observed for frequencies above the 90th percentile cut-offs of these hormones at birth.

Paternal BMI showed no relation to any of the investigated birth outcomes, neither in step-wise regression (see below) nor correlation analyses (Spearman’s r of paternal BMI: vs. birth weight, r = − 0.01; vs. relative birth weight, r = 0.02; vs. cord-blood insulin, r = 0.13; vs. cord-blood leptin, r = − 0.02; all not significant).

Bivariate associations between maternal anthropometry and glycemia revealed that prepregnancy BMI positively correlated only with FG among oGTT measurements (r = 0.36, P < 0.001) but clearly with 3rd trimester glucose variables (mean FG at 32nd weeks’ gestation: r = 0.40, mean FG at 36th weeks’ gestation: r = 0.41, HbA1c in 3rd trimester: r = 0.50, HbA1c at delivery: r = 0.46; all P < 0.001). In contrast, the only relationship of net GWG was an inverse correlation with AUC glucose at oGTT (r = − 0.15, P = 0.037).

Furthermore, prepregnancy BMI, but not net GWG, was significantly positively associated with birth outcomes, i.e., placental weight, relative birth weight, cord-blood insulin and leptin (Table 2, Fig. 2). The relationship between prepregnancy BMI and birth weight was close to significance (r = 0.14, P = 0.050). Glucose variables at oGTT were not related to birth outcomes, with the exception of a weak positive correlation between FG and placental weight. However, mean FG at 32nd weeks’ gestation and HbA1c levels in 3rd trimester and at delivery were significantly positively associated with all birth outcomes, including birth weight (Table 2). Among all anthropometric and metabolic variables, HbA1c at delivery had the strongest associations with birth outcomes, in particular with relative birth weight and cord-blood insulin (Table 2, Fig. 2). Of note, placental weight was an even stronger correlate for birth weight (r = 0.57, P < 0.001) and relative birth weight (r = 0.58, P < 0.001).

The influence of maternal anthropometry on birth outcomes was further investigated in detail by comprehensive regression analyses (Table 3). There were no relationships observed between GWG in 1st and 2nd trimester and net GWG, respectively, and anthropometric birth outcomes. In the crude models, prepregnancy weight status, i.e., body weight and/or BMI, and 3rd trimester and/or total GWG were significantly positively associated with placental weight, birth weight and relative birth weight. However, their initial relations were already diminished after adjustment to general parameters (Model I, Table 3). Further addition of maternal 3rd trimester glucose variables (Model II, Table 3) completely abolished all of these associations. With regard to cord-blood hormones, there were hardly any associations observed with maternal anthropometry. In unadjusted models, only prepregnancy weight and BMI status showed significant positive associations with insulin and leptin, however, these relations no longer persisted even after adjustment only to general variables (Model I, Table 3). Taken together, neither maternal nor paternal anthropometry showed a significant relation to investigated birth outcomes, after consideration of maternal metabolic parameters.

In order to identify potential metabolic determinants responsible for these observations, regression analyses with a variety of glucose variables were carried out (Table 4). Glucose values at oGTT showed almost no relationship with any birth outcome. There was just a positive relation between FG at oGTT and placental weight, as already observed in correlation analysis (Table 2). In contrast, even an inverse relation between 1-h glucose and birth weight was observed (Table 4). Both associations, however, were no longer present after adjustment to general variables (Model I, Table 4). On the contrary, 3rd trimester glucose variables were significantly positively associated with anthropometric outcomes at birth. Particularly consistent, mean FG and PPG at 32nd weeks’ gestation as well as HbA1c levels in 3rd trimester and at delivery were significantly positively associated with all anthropometric outcomes in the crude models. Importantly, most of these relations remained constant also after adjustment to general parameters (Model I, Table 4). Further inclusion of maternal anthropometric variables into the adjustment model (Model II, Table 4) abolished some initial associations. However, after adjustment to critical general as well as anthropometric variables (Model II) mean glucose values at 32nd and 36th weeks’ gestation showed clear significant positive relationships with placental weight, while mean PPG at 32nd weeks’ gestation and HbA1c in 3rd trimester and at delivery were significantly positively associated with birth weight and relative birth weight, respectively (Table 4). A similar pattern was observed for critical cord-blood hormones. Again, glucose variables at oGTT were hardly related to cord-blood insulin and leptin. The only association was observed between FG at oGTT and cord-blood leptin, however, this weak positive relationship was only present in the unadjusted model. In contrast, considering 3rd trimester glucose variables, in particular mean FG at 32nd and 36th weeks’ gestation and HbA1c at delivery remained significantly positively associated with cord-blood insulin, even after full adjustment (Model II, Table 4). Likewise, maternal 3rd trimester glycemia, indicated by HbA1c, was significantly positively related to cord-blood leptin, and HbA1c at delivery showed the strongest effects here (Table 4). Adjustment to general and, additionally, maternal anthropometric variables only marginally altered the effect size of these associations (Model II, Table 4).

Forward step-wise regression analyses with key paternal and maternal general, anthropometric, and metabolic parameters in (clinically) relevant categories were then performed to identify major predictors for critical birth outcomes (Table 5). Aside from classical birth weight determining factors, e.g. term delivery, smoking during pregnancy etc., maternal, but not paternal, anthropometric and metabolic variables, respectively, appeared as predictors. However, maternal anthropometry, i.e., prepregnant overweight/obesity and excessive GWG (according to IOM criteria), was positively related only to relative birth weight and explained around 2–3% of the variance (Table 5). On the contrary, maternal glucose during 3rd trimester showed a variety of associations. For example, mean FG at 32nd weeks’ gestation > 90 mg/dL was the strongest factor for placental weight (6% of variance) and a predictor for birth weight and cord-blood insulin (Table 5). A FG value > 90 mg/dL at oGTT was a significant positive factor for cord-blood leptin at birth (3% of variance). Unexpectedly, 1-h glucose value at oGTT above the guideline cut-off (> 180 mg/dL) was associated with lower instead of higher birth weight and relative birth weight, explaining 2–4% of variances. As shown in Table 5, HbA1c at delivery was a particular strong predictor for higher cord-blood insulin and leptin levels (7% of variance).

Additionally, to further illustrate observations for clinically unfavorable neonatal anthropometric (LGA and macrosomia) and cord-blood outcomes (insulin and leptin > 90th percentiles), percentages of relevant anthropometric and metabolic variables/exposures, grouped according to clinical/guideline cut-offs, were compared to respective internal reference groups which did not show these adverse outcomes (Fig. 3). Following this approach, maternal prepregnancy overweight/obesity and excessive GWG only showed significantly higher relative frequencies regarding LGA (Fig. 3). In contrast, parameters of maternal hyperglycemia during 3rd trimester were associated with all unfavorable endpoints. Percentages of 3rd trimester glucose variables above cut-offs, in particular mean FG at 32nd weeks’ gestation > 90 mg/dL and HbA1c > 6% in 3rd trimester as well as at delivery, were significantly higher in offspring with LGA and macrosomia, respectively. Notably, women with macrosomic offspring exceeded 2.5-times more often the cut-off for HbA1c in 3rd trimester (Fig. 3). Furthermore, both cord-blood insulin as well as leptin > 90th percentile showed strong positive associations with maternal 3rd trimester hyperglycemia, but neither with pregravid BMI nor excessive GWG. Percentages above the cut-offs of FG at oGTT and mean FG at 32nd weeks’ gestation (> 90 mg/dL for both) were significantly higher in offspring with cord-blood leptin > 90th percentile. Finally, relative frequencies above the cut-off for HbA1c at both investigated time points were even increased up to 3- to 6-fold in offspring with cord-blood insulin and leptin > 90th percentile, respectively (Fig. 3).