Associations of maternal insulin sensitivity during pregnancy with childhood central adiposity in the Genetics of Glucose regulation in Gestation and Growth cohort

Abstract

Background:

Childhood obesity has been associated with prenatal exposure to maternal hyperglycemia, but we lack understanding about maternal insulin physiologic components that contribute to this association.

Objectives:

Evaluate the association between maternal insulin sensitivity during pregnancy and adiposity measures in childhood.

Methods:

In 422 mother-child pairs, we tested associations between maternal insulin sensitivity measures at ~26 weeks of pregnancy and child adiposity measures, including dual-energy X-ray absorptiometry body composition, and anthropometry (body mass index, waist circumference) at ~5 years. We used linear regression analyses to adjust for maternal age, ethnicity, gravidity, 1^st trimester body mass index, and child sex and age at mid-childhood.

Results:

In early pregnancy, maternal mean age was 28.6 ±4.3 years and median body mass index was 24.1 kg/m². Lower maternal insulin sensitivity indices were correlated with greater child adiposity based on anthropometry measures and on dual-energy X-ray absorptiometry total and trunk %fat in univariate associations (r= −0.122 to −0.159). Lower maternal insulin sensitivity was specifically associated with higher dual-energy X-ray absorptiometry trunk %fat (n=359 for Matsuda; β= −0.034±0.013; P=0.01) after adjustment for covariates, including maternal body mass index.

Conclusions:

Maternal insulin sensitivity during pregnancy may contribute to increased risk for higher offspring central adiposity in middle childhood.

INTRODUCTION

Childhood obesity is a pressing public health problem. In 2015–2016, 18.5% of children and adolescents in the United States have obesity [1]. Physical and mental health consequences of childhood obesity present during childhood, adolescence, and in adult life [2]. Obesity in children has been associated with cardiovascular diseases and type 2 diabetes later in life, among multiple other adverse outcomes [3]. More specifically, central adiposity in children has been associated with cardiovascular risk later in life [4]. As evidence regarding the adverse consequences of childhood obesity develops [2], there is increasing interest in the detection of early life and prenatal risk factors associated with childhood excess adiposity, globally and/or centrally located.

The landmark Hyperglycemia and Adverse Pregnancy Outcomes (HAPO) and its Follow-up study demonstrated the association of maternal hyperglycemia during pregnancy with greater neonatal adiposity, and with higher offspring whole-body adiposity and waist circumference at 10 to 14 years old [5,6]. Maternal hyperglycemia is caused by an imbalance between excessive lowering of insulin sensitivity and insulin secretion capacity during pregnancy. Low maternal insulin sensitivity is also characterized by an adverse metabolic and lipids profile that may contribute to prenatal programming by altering the placental fuels transfer during in utero development. A few studies have showed that maternal insulin sensitivity is specifically associated with neonatal adiposity [7,8]; however, limited literature has investigated maternal insulin sensitivity and later childhood adiposity measures. Further, if offspring adverse adiposity measures are related specifically to maternal insulin sensitivity in pregnancy, this may support additional targeted interventions to modulate this exposure during pregnancy, which is possible via lifestyle (physical activity) or pharmacologic approaches.

Objectives

To address this gap, we have investigated the association between maternal insulin sensitivity and glycemic measures with offspring adiposity at around five years of age in the Genetics of Glucose regulation in Gestation and Growth (Gen3G) cohort. We were particularly interested in associations between maternal insulin sensitivity and central adiposity in children, given that central adiposity is associated with an increased risk of cardiometabolic adverse outcomes over the life course [9]. We hypothesized that lower maternal insulin sensitivity is associated with offspring higher central adiposity measures assessed by dual-energy X-ray absorptiometry (DXA) in mid-childhood in the Gen3G cohort. We additionally investigated the associations of maternal insulin sensitivity, insulin secretion, and glycemic measures with other offspring adiposity and anthropometric measures at this age.

METHODS

Study design

This observational study of the prospective Gen3G cohort examined associations between maternal metabolic measures during pregnancy and offspring anthropometric outcomes at ~5 years of age.

Participants and setting

The study population included women from Gen3G, a cohort from Quebec, Canada, and their offspring after the women gave birth. The Gen3G cohort was formed to aid in developing a deeper comprehension of the biological, environmental, and genetic factors of glucose regulation in pregnant women, and their effects on fetal and offspring development. We recruited 1024 pregnant women without preexisting diabetes at first trimester screening from January 2010 to June 2013 from health care facilities at or affiliated with Centre Hospitalier Universitaire de Sherbrooke (CHUS). Participants were demographically representative of the greater population of the region; and a more comprehensive description of the cohort has been previously published [10]. Each study participant provided informed written consent prior to joining the cohort, and the study was conducted in concordance with the Declaration of Helsinki.

We collected measurements from mothers at visit one (V1) during 5 to 16 weeks of gestational age, visit two (V2) during 24 to 30 weeks of gestational age, and at birth. At the time of publication, we had collected offspring data up to ~5 years of age.

Variables and measurements

V1 measurements aligned with mothers’ first trimester prenatal clinical blood sampling. V2 aligned with the oral glucose tolerance test (OGTT) in line with second trimester gestational diabetes mellitus (GDM) universal screening recommendations. We collected demographic data, medical history, standardized anthropometric measurements, and lifestyle data via questionnaire during both visits. Trained research staff measured weight with calibrated scale and height with standardized stadiometer, following research protocols. We calculated first trimester body mass index (BMI) using weight divided by squared height (kg/m²). Alongside the clinically indicated blood draw at V1, we collected additional blood samples that were drawn (for the majority of women) during the 50g glucose challenge test (GCT – performed as part of research protocol) to obtain glucose 1h post-GCT. We excluded women from the study at V1 if they had known pre-pregnancy diabetes or screened positive for existing diabetes (HbA1c >6.5% or 48 mmol/mol; or post-GCT 1h-glucose >10.3 mmol/L) as well as those with non-singleton pregnancies. At V2, again, we collected additional blood samples at fasting, 1h, and 2h points of the 75g-OGTT to measure insulin and C-peptide for each time-point, in addition to glucose. We excluded women during pregnancy if they experienced miscarriage, abortions, health issues that did not allow for further participation, or if they moved or no longer wished to be a part of the study.

We collected offspring age and sex at birth using medical records, and at the ~5-year follow-up visit using questionnaires. Additional clinical information at birth was gathered from electronic medical records. Anthropometry measurement procedures at ~5 years are detailed below.

Laboratory measurements

We performed blood draws, then rapidly processed samples and prepared aliquots of plasma. We measured glucose levels via the hexokinase method (Roche Diagnostics; CHUS biochemistry laboratory) as soon as samples were collected. We measured HbA1c, cholesterol, HDL (high-density lipoprotein), and triglyceride at the CHUS biochemistry lab. We stored additional samples at −80°C until we performed biomarker measurements. At that point aliquots were thawed, and then centrifuged at 6000 × g for 10 minutes at 4°C. We measured insulin and C-peptide levels via multiplexed particle-based flow cytometric assays (Human Milliplex MAP kits; EMD Millipore) from the previously frozen plasma samples.

Insulin Sensitivity/Resistance Indices calculations

Using glucose and insulin values during the second trimester OGTT, we determined estimates of insulin sensitivity through the Matsuda Index, as validated against clamps in pregnant women [11]. Based on a prior HAPO publication [12], we calculated C-peptide insulin sensitivity index, an alternate estimate using C-peptide measured during 2^nd trimester OGTT. We also calculated HOMA-IR (reciprocal of insulin sensitivity) using the homeostasis model [13]. Using OGTT glucose and insulin levels, we calculated Stumvoll first-phase estimate as reflection of insulin secretion, using the original validated formula [14] confirmed as valid in pregnancy [15] against intravenous glucose tolerance testing.

Main outcomes: Adiposity measurements at 5-year follow-up

At the 5-year follow-up visit, we measured weight (kg) using an electronic scale (Rice Lake Weighing systems), and measured height (cm) using a calibrated stadiometer (Seca). We calculated weight z-score and BMI z-score adjusted for sex and age using WHO (World Health Organization) AnthroPlus software (WHO Reference 2007 for children 5–19 years old). We measured waist and hip circumferences using a Shorrflip Tape© measuring tape and rounded to the nearest 0.5 cm. Trained research staff measured triceps, biceps, subscapular, and suprailiac skinfold thickness (mm) using a calibrated skinfold caliper (AMG Medical). Each measure was taken twice, and measured a third time if the first two measures differed by more than 10 percent. We calculated total skinfold thickness (mm) by summing triceps, biceps, subscapular, and suprailiac skinfold thickness measurement means. We measured trunk and total fat mass measures using whole-body DXA with a Horizon DXA System (Hologic). Some children were unable to keep a leg, arm, or their head steady for the duration of the DXA scan; if the movement of a limb impacted imaging, measurement from the opposing limb was used (20% of scans had arm and/or leg duplicated). Each DXA image involving a child that moved during measurement was reviewed manually by 2 research staff to determine when duplication of a limb was necessary, and if it was appropriate to include the scan overall (with consultation with lead investigators when there were uncertainties); none were deemed necessary to exclude [16]. We defined body regions using the manufacturer’s model of Hologic software (version 5.5.3.1).

We conducted linear regression analyses to adjust for potential confounding variables (see statistical methods section), including maternal BMI, which we considered as the major potential confounder in the current analyses. Maternal BMI is an integral confounder as it is strongly correlated with both maternal insulin sensitivity and offspring BMI. We selected other covariates included in our multivariable models based on our theoretical model containing potential confounders supported by previous published literature [16,17] and known precision variables (child sex and age at adiposity measurements). We compared characteristics at first trimester from participants that were included in this analyses with characteristics of non-included participants (Supplementary Table 1), to assess potential selection bias.

Included participants

Our study population included 422 mother-offspring pairs for which we had at least one child anthropometric measure (measured by our staff using standardized procedures) at the 5-year follow-up visit and at least one glycemic related measure at V2 (glucose levels during OGTT, HbA1c, and/or insulin sensitivity/secretion indices). Everyone in our included sample had measured weight and BMI data. As DXA measurement requires individuals to stay static ≥5 minutes, not all the five-year-old children were able to comply; further, as the DXA procedure emits a minor bit of radiation, some parents decided against their child’s participation. Thus, our maximal sample size for DXA measurements was 377.

Statistical methods and variables

We reported normally distributed continuous variables as mean ± SD (standard deviation), and non-normally distributed continuous variables as median and interquartile range (IQR); and we reported categorical variables as percentages. We log-transformed non-normally distributed variables. We used Pearson pairwise correlations for all available child adiposity variables from the five-year follow-up visit and glycemic related measures at V2 (glucose levels during OGTT, HbA1c, and insulin sensitivity indices) to compare strengths of correlations between maternal glycemic measures with child adiposity measures (across all measures). Our primary adiposity outcome was DXA trunk percent fat; and secondary adiposity outcomes included z-score weight, z-score BMI, waist circumference, hip circumference, sum of mean skinfold thickness and individual skinfold thickness (tricep, bicep, subscapular, suprailiac), DXA total percent fat, DXA total lean mass, DXA total lean mass + BMC, and DXA total mass.

We conducted linear regressions for our main exposure of insulin sensitivity indices (Matsuda, C-peptide) and testing association with offspring adiposity outcomes, presenting first unadjusted coefficient estimates (Model 1). We subsequently performed multiple linear regressions on the exposures and outcomes to adjust for maternal age, ethnicity (white or non-white), and gravidity (first pregnancy or not first pregnancy), child sex, and child age at time of outcome measurements (Model 2). Finally, we further adjusted our regression models for maternal BMI (as a continuous variable) measured in the first trimester of pregnancy (Model 3). As an exploratory analyses, we also tested associations between maternal insulin secretion (Stumvoll first-phase estimate) and child adiposity measures using linear regression models. Given that any insulin secretion estimate must be interpreted while considering the degree of insulin sensitivity that the beta-cell is responding to, all models investigating Stumvoll were adjusted for Matsuda (including Model 1). We considered P-values <0.05 as significant.

We explored potential sex interaction in regression analyses of insulin sensitivity indices with adiposity outcomes. We considered potential interaction if P <0.10 for the interaction term and showed the stratified analyses if potential interaction was detected. We were not able to evaluate potential interaction by race/ethnicity due to our primarily European sample; and we were not able to evaluate potential interaction by age due to the uniformity of child ages at the five-year follow-up visit. All analyses were performed using R (version 4.1.1).

RESULTS

Participants

Our population included 422 mother-offspring pairs from the Gen3G cohort for which we had measurements of offspring weight from the five-year follow-up visit and at least one glycemic related measure at V2. For other childhood adiposity outcomes, we included as many individuals as data available for each outcome of interest, ranging from 359 to 422 mother-offspring pairs across correlations. First trimester characteristics of participants who were included in this analyses were similar to characteristics of participants non-included (Supplementary Table 1).

Descriptive data

Characteristics of included participants can be found in Table 1. At V1, the mothers’ mean age was 28.6 years (SD=4.3), median BMI was 24.1 kg/m² (IQR 21.7; 28.2), and 35.3% were primigravid. During the second trimester of pregnancy, median Matsuda index was 7.64 (IQR 5.36; 10.95). At follow-up, offspring median age was 5.3 years (IQR 5.1; 5.5) and mean BMI z-score was 0.22 (SD=0.94). About half (47.2%) of offspring were female. Offspring median trunk percent fat was 25.2% (IQR 22.8; 28.9), and median total percent fat was 29.9% (IQR 27.3; 33.4), as measured by DXA. Other adiposity measures are included in Table 1.

Table 1.

Characteristics of Gen3G mother-child pairs during pregnancy and child at ~5 years

	N	Mean ± sd or Median [IQR] or N (%)
Maternal measures – 1st trimester
Age, years	422	28.6 ± 4.3
Ethnicity, European descent	422	407 (96.4%)
Gestational weeks at 1st trimester visit	422	9.4 [8.1 ; 11.9]
Gravidity, Primigravid	422	149 (35.3%)
Familial history of diabetes	422	83 (19.7%)
First trimester BMI, kg/m2	422	24.1 [21.7; 28.2]
Glucose 1h post-50g, mmol/L	391	5.4 [4.6 ; 6.5]
Maternal measures – 2nd trimester
Gestational weeks at 2nd trimester visit	422	26.3 [25.9 ; 27.0]
2nd trimester BMI, kg/m2	419	26.8 [24.4 ; 30.3]
Glucose fasting, mmol/L	418	4.2 [4.0 ; 4.4]
Glucose 1h post-OGTT, mmol/L	418	7.2 [6.1 ; 8.2]
Glucose 2h post-OGTT, mmol/L	416	5.7 [4.9 ; 6.6]
HbA1c, %	422	4.99 ± 0.29
HOMA-IR	406	1.29 [0.89 ; 1.95]
Matsuda index	402	7.64 [5.36 ; 10.95]
C-peptide insulin sensitivity index	403	90.1 [68.2 ; 124.7]
Stumvoll (secretion first-phase estimate)	405	1101.2 [914.4 – 1288.6]
Child measures – 5 years
Sex, girls	422	199 (47.2%)
Age, years	422	5.3 [5.1 ; 5.5]
Weight, kg	422	19.1 [17.7 ; 20.7]
Weight for age z-score	422	0.14 ± 0.99
BMI, kg/m2	422	15.6 [14.8 ; 16.4]
BMI for age z-score	422	0.22 ± 0.94
Waist circumference, cm	421	53.5 [51.9 ; 55.7]
Hip circumference, cm	421	57.8 [55.6 ; 60.1]
Tricep skinfold thickness, mm	417	11.0 [9.0 ; 12.8]
Bicep skinfold thickness, mm	419	6.3 [5.0 ; 8.0]
Subscapular skinfold thickness, mm	416	5.6 [4.8 ; 6.9]
Suprailiac skinfold thickness, mm	417	6.1 [4.9 ; 8.3]
Total skinfold thickness, mm	416	29.3 [24.6 ; 34.9]
DXA total fat, %	377	29.9 [27.3 ; 33.4]
DXA trunk fat, %	377	25.2 [22.8 ; 28.9]
DXA total lean mass, g	377	12 477 ± 1 720
DXA total lean mass + BMC, g	377	13 139 ± 1 785
DXA total mass, g	377	18 602 [17 253 ; 20 177]

BMI: Body Mass Index; HOMA-IR: homeostasis model of insulin resistance; DXA: dual-energy X-ray absorptiometry; BMC: Bone Mineral Content

Correlations between maternal glycemic markers and childhood adiposity (Supplementary Table 2)

Matsuda index showed generally stronger correlations with truncal and total percent fat estimated by DXA (r= −0.159 and −0.146) compared to maternal glucose levels during OGTT or HbA1c. Lower insulin sensitivity indices were also associated with higher BMI z-score (r=−0.122 for Matsuda; r=−0.186 for C-peptide index;), waist circumference (r=−0.099 for Matsuda; r=−0.131 for C-peptide index) and skinfolds centrally located (subscapular, suprailiac). We observed that HOMA-IR showed a similar pattern of association with childhood adiposity markers (but in the opposite direction, given it reflects maternal insulin resistance – the reciprocal of insulin sensitivity), providing further consistency of our findings with coefficient of correlations generally similar compared to the ones observed for Matsuda and C-peptide indices. Higher post-OGTT glucose values (1h or 2h) tended to be associated with higher percent fat (total and truncal), and lower lean mass (all estimated by DXA). Additional correlations between maternal first and second-trimester insulin and glycemic measures with child adiposity measures at the five-year follow-up visit can be found in Supplementary Table 2. In addition, we presented the cross-sectional correlations across child adiposity and body composition measures at 5 years of age to show the inter-relationships between the various measures (Supplementary Table 3).

Linear regression analyses of maternal insulin sensitivity indices in the second trimester of pregnancy with child adiposity at ~5 year of age

We presented the adjusted models for associations of child adiposity measures at the 5-year follow-up visit with maternal insulin sensitivity estimated by Matsuda index in Table 2, and by C-peptide index in Supplementary Table 4.

Table 2.

Associations of maternal insulin sensitivity (Matsuda index) in the second trimester of pregnancy with child adiposity measures at the ~5 year follow-up visit

Outcomes	Matsuda index (log)
	N	Model 1				Model 2				Model 3
		Beta	SE	P	Adj R2	Beta	SE	P	Adj R2	Beta	SE	P	Adj R2
Weight	402	−0.016	0.012	0.21	0	−0.018	0.012	0.13	0.06	0.003	0.013	0.79	0.10
z-score weight	402	−0.122	0.086	0.16	0	−0.119	0.086	0.17	0	0.035	0.092	0.71	0.04
BMI	402	−0.020	0.008	0.01	0.01	−0.019	0.008	0.01	0.01	−0.002	0.008	0.77	0.07
z-score BMI	402	−0.197	0.08	0.01	0.01	−0.194	0.08	0.02	0.01	−0.018	0.085	0.83	0.07
Waist circumference	401	−0.012	0.006	0.048	0.01	−0.012	0.006	0.04	0.04	−0.002	0.006	0.79	0.07
Hip circumference	401	−0.011	0.006	0.08	0.01	−0.011	0.006	0.05	0.05	−0.001	0.006	0.93	0.10
Mean tricep skinfold thickness	397	−0.021	0.023	0.36	0	−0.019	0.022	0.38	0.08	0.013	0.023	0.58	0.10
Mean bicep skinfold thickness	399	−0.030	0.029	0.31	0	−0.024	0.027	0.38	0.12	0.025	0.029	0.39	0.15
Mean subscapular skinfold thickness	396	−0.069	0.028	0.01	0.01	−0.066	0.026	0.01	0.12	−0.026	0.028	0.35	0.14
Mean suprailiac skinfold thickness	397	−0.067	0.035	0.06	0.01	−0.064	0.033	0.05	0.12	−0.024	0.036	0.51	0.13
Sum of mean skinfold thickness	396	−0.043	0.025	0.08	0.01	−0.04	0.023	0.08	0.13	0	0.025	0.99	0.17
DXA total fat, %	359	−0.035	0.012	0.006	0.02	−0.034	0.011	0.002	0.26	−0.021	0.012	0.07	0.28
DXA trunk fat, %	359	−0.044	0.015	0.003	0.02	−0.043	0.012	0.0007	0.30	−0.034	0.013	0.01	0.30
DXA total lean mass, g	359	6.34	156.75	0.97	0	−38.62	143.35	0.79	0.17	157.77	154.44	0.31	0.19
DXA total lean mass + BMC, g	359	18.25	162.65	0.91	0	−29.22	148.81	0.84	0.17	173.72	160.35	0.28	0.19
DXA total mass, g	359	−0.015	0.013	0.23	0	−0.019	0.012	0.14	0.06	0.002	0.013	0.86	0.10

Model 1: Unadjusted

Model 2: Adjusted for maternal age, maternal ethnicity (white y/n), gravidity (first pregnancy y/n), offspring sex, offspring age (at the time of outcome measurements)

Model 3: Adjusted for maternal age, maternal ethnicity (white y/n), gravidity (first pregnancy y/n), offspring sex, offspring age (at the time of outcome measurements), and maternal BMI

All outcomes are log-transformed except z score weight, z-score BMI, DXA total lean mass, and DXA total lean mass + BMC

Matsuda index was log-transformed

Adjusted R²: Goodness-of-fit measure adjusted for the number of predictors

Lower maternal insulin sensitivity (Matsuda index) in the second trimester was significantly associated with higher offspring DXA trunk percent fat (n=359; β= −0.044 ± 0.015; P=0.003) and total percent fat (n=359; β= −0.035 ± 0.012; P=0.006) at ~5 years in the unadjusted model (Model 1), in line with our hypothesis. Maternal insulin sensitivity (Matsuda index) was also significantly associated with many other child anthropometric measures at 5 years, including BMI z-score, waist circumference, and subscapular skinfold thickness (Table 2). After adjustment for maternal age, ethnicity, and gravidity, these associations remained fairly similar (see Table 2, Model 2). After further adjusting for maternal BMI in the first trimester (see Table 2, Model 3), most associations were attenuated, but association remained significant with DXA trunk percent fat (n= 359; β= −0.034 ± 0.013; P=0.01), though the strength of the association was attenuated. We did not find significant associations between maternal insulin sensitivity and DXA total lean mass, DXA total lean mass + BMC, or DXA total mass using either the crude or adjusted models.

When using the C-peptide index as maternal insulin sensitivity estimate, we found similar associations (Supplementary Table 4) to what we observed with Matsuda. For example, in Model 2, we observed lower maternal insulin sensitivity (C-peptide index) in the second trimester was significantly associated with higher offspring DXA trunk percent fat (n= 360; β= −0.046 ± 0.016; P=0.003) and DXA total percent fat (n= 360; β= −0.035 ± 0.014; P=0.01) at ~5 years. Further adjustment for maternal BMI substantially attenuated the coefficient estimates, however DXA trunk percent fat remained significantly associated with maternal insulin sensitivity assessed by the C-peptide index (Supplementary Table 4, Model 3).

Additional analyses

Linear regression analyses of maternal insulin secretion (Stumvoll) in the second trimester of pregnancy with child adiposity at ~5 year of age.

To further explore other refined glycemic physiologic measures as potential drivers of the association between prenatal dysglycemic exposures and child adiposity, we also conducted association analyses between maternal insulin secretion (Stumvoll) and child adiposity measures. In our linear regression analyses, we found that higher maternal insulin secretion was associated with greater lean mass estimated by DXA in all models (Supplementary Table 5). We did not find statistically significant associations with any of the other child adiposity or anthropometric measures.

Investigating potential sex interactions

We detected potential interaction by sex in our analyses of maternal insulin sensitivity indices and for some of the childhood DXA adiposity outcomes (see Supplementary Table 6). When stratified by sex, we found that the associations between Matsuda index and DXA trunk percent fat were present in girls (n=172; β= −0.058 ± 0.023; P=0.01), but not significant in boys (n=187; β= −0.017 ± 0.016; P=0.29). We observed the same trends in sex-stratified results for Matsuda associations with DXA total percent fat (n=172 females; β= −0.049 ± 0.019; P=0.01), (n=187 males; β= −0.001 ± 0.014; P=0.92); and when using maternal insulin sensitivity estimated by C-peptide index (see Table 3). We did not detect significant interactions between sex and maternal insulin sensitivity indices for any of the other anthropometric measures in the overall sample.

Table 3.

Associations of maternal insulin sensitivity (Matsuda and C-peptide indices) in the second trimester of pregnancy with child trunk % fat and total % fat (estimated by DXA) at the ~5 year follow-up visit, stratified by offspring sex

	Boys				Girls				Boys				Girls
Outcome	Matsuda index								C-peptide index
	Model 3				Model 3				Model 3				Model 3
	N	Beta	SE	P	N	Beta	SE	P	N	Beta	SE	P	N	Beta	SE	P
DXA trunk fat, %	187	−0.017	0.016	0.29	172	−0.058	0.023	0.01	188	−0.013	0.021	0.54	172	−0.065	0.027	0.02
DXA total fat, %	187	−0.001	0.014	0.92	172	−0.049	0.019	0.01	188	−5.41E-05	0.019	1.00	172	−0.046	0.023	0.04

Model 3: Adjusted for maternal age, maternal ethnicity (white y/n), gravidity (first pregnancy y/n), offspring sex, offspring age (when outcome measure was taken) and maternal BMI 1^st trimester (log)

Outcomes, Matsuda index and C-peptide index are log transformed

DISCUSSION

Key results

Through our analysis of the Gen3G cohort, we found that lower maternal insulin sensitivity measures and higher glycemic measures during pregnancy were associated with greater offspring adiposity in mid-childhood. We were particularly interested to find that the associations with centrally located fat percentage as estimated by DXA was apparent through the use of two different validated insulin sensitivity estimates (Matsuda and C-peptide indices) and after adjustment for multiple potential confounders, including maternal BMI. We found that lower insulin sensitivity during pregnancy tended to be associated with other measures of central adiposity in mid-childhood, such as waist circumference and subscapular skinfold thickness; as well as global markers of adiposity such as BMI z-score. However, these later associations were substantially attenuated and non-significant when taking into account maternal BMI. The attenuation with maternal BMI is not surprising, given that BMI is strongly correlated with insulin sensitivity inside and outside of pregnancy, as well as with offspring BMI via shared genetics and familial environment. It is also well established that reduction in BMI leads to improvement in insulin sensitivity outside of pregnancy [18] which is partly why some investigators are suggesting pre-pregnancy weight loss as a key intervention to improve maternal and offspring long-term health [19]. Thus, it is notable that the observed associations between maternal insulin sensitivity indices and child truncal fat percentage still persisted after accounting for maternal BMI, which suggests a role of insulin sensitivity in prenatal programming above the maternal weight status.

We focused on insulin sensitivity indices as our exposures as a gap exists in the examination of insulin sensitivity and its associations with mid-childhood adiposity measures. Prior studies of maternal insulin sensitivity during pregnancy and childhood outcomes have investigated cardiometabolic markers [20,21], rather than adiposity outcomes, outside of child BMI and weight. Wahab et al. [20], for example, found elevated maternal early-pregnancy glucose and insulin concentrations were associated with higher glucose and insulin concentrations in offspring at 10 years, though these associations attenuated after adjusting for maternal pre-pregnancy BMI. Similarly, Coles et al. [21], found that insulin sensitivity of mothers with GDM, measured by the Matsuda index, was negatively associated with HOMA-IR in offspring of 1 to 3 years of age. Our study extends on previous literature by examining multiple anthropometric measures and adiposity markers, including overall and central body composition and fat percent by DXA, which is considered the gold standard for body composition and is a highly reliable measure of body fat in children [22].

DXA has been increasingly used in the measure of whole-body and centrally-located adiposity in children [23]. In our analyses, the association between lower maternal insulin sensitivity and greater central adiposity was specifically captured by the truncal fat percentage estimated by DXA. The ability to measure body composition via DXA also allowed us to observe that higher maternal insulin secretion was specifically associated with greater lean mass, but no other anthropometric measures or DXA estimates, supporting the value of detailed body composition evaluation to untangle early determinants of growth and adiposity. Further, Steinberger et al. [24] found that although BMI measurements are highly correlated to DXA measurements, DXA is superior in measuring fatness due to its greater stability and accuracy. Centrally-located adiposity measures, estimated by DXA or using waist circumference, have been found to be stronger predictors of cardiovascular risk factors compared to overall adiposity measurements (e.g. BMI z-score) in mid-childhood, although study conclusions have varied [9,23,25] and prospective studies are particularly limited.

Strengths and Limitations

Our study had numerous strengths. Our prospective follow-up study design allowed for longitudinal data collection over time, including detailed phenotypic measures of insulin sensitivity during pregnancy, as well as DXA measurements in childhood, which are generally considered the gold standard measure of body composition and adiposity distribution in children [27]. We were able to compare different measures of maternal glycemic regulation and multiple anthropometric outcomes in children in the same well-characterized cohort. We adjusted our analyses for multiple potential confounders, including measured maternal BMI in early pregnancy. Our study also had limitations. As our study was observational, we cannot conclude the relationship between maternal insulin sensitivity and offspring central adiposity in mid-childhood is causal. It is possible that other metabolic factors – for example, lipids – associated with maternal insulin sensitivity are responsible for the observed associations. Further, we did not conduct OGTT in the third trimester; the third trimester is a key period of fat deposition and rapid growth of the fetus, and it would have been beneficial to investigate the association between third trimester maternal insulin sensitivity and offspring adiposity. Our cohort was largely made up of women from European ancestry, so our results may not be generalizable to other populations. Given that in general, many contemporary populations of women of reproductive age have higher BMI [26], they are likely entering pregnancy with lower insulin sensitivity, thus similar studies should be conducted in populations with a higher BMI distribution for potential replication of our findings.

Interpretation

We believe that the associations between maternal insulin sensitivity, and specifically percent fat in the truncal region reflecting central adiposity in children, may inform us on how in utero exposures might program the offspring for metabolic health risk across the life course. Maternal hyperglycemia during pregnancy has been associated with greater abdominal circumference as early as during in utero development [27]. Outside of pregnancy, lower insulin sensitivity is associated with type 2 diabetes and adverse lipids profile [28]. During pregnancy, lowering insulin sensitivity is part of physiologic adaptation to feed the fetus; but in some women, excessive low insulin sensitivity coupled with unmatched insulin secretion results in GDM [29,30]. In addition to hyperglycemia, low insulin sensitivity during pregnancy has been associated with adverse lipid profile [31], some of which are actively transferred by the placenta towards the fetus, potentially adding to the ‘fuel teratogenicity’ [32]. It is possible that lower maternal insulin sensitivity leads to greater offspring central adiposity beginning in utero, or programs the peripheral adipose tissue to be ‘biologically less flexible’, leading to central deposition, which then leads to lower insulin sensitivity later on in the offspring’s life [33,34]. This could potentially be contributing to an inter-generational vicious cycle of adverse adiposity and diabetes [35]. It is particularly worrisome that we observed the associations specifically in young girls in our analyses, since they are likely to continue to carry these metabolic risks into their reproductive years and further feed into the cycle in future generations. Our findings from the sex-stratified analyses are in line with prior published literature regarding differential prenatal metabolic programming effects in male and female offspring [36,37], but we remain cautious in our interpretation given the exploratory nature of these analyses.

Though literature on the childhood population is less conclusive, central adiposity has long been linked to adverse cardiometabolic health outcomes in adults [9], such as lower insulin sensitivity and developing type 2 diabetes; central fat distribution has been found to increase risk for these outcomes independently of BMI [38]. It has been hypothesized that visceral fat may be linked to an overactive stress response, which may result in elevated blood pressure and cardiac risk [39]. It is also thought that excess visceral fat releases free fatty acids through the blood stream to the liver, pancreas, heart, and other organs [40]. As these cells are not meant to store fat, the resulting organ dysfunction may impact insulin regulation, heart function, and cholesterol levels, leading to adverse cardiometabolic outcomes [41]. We believe the results of our study will be further bolstered by research examining if mid-childhood central adiposity is associated with these adverse cardiometabolic markers later in childhood or adolescence [42] when we continue to follow-up on Gen3G participants, as well as from reports in other longitudinal cohorts.

Conclusion

In the prospective pre-birth Gen3G cohort, we found that lower maternal insulin sensitivity in the second trimester was associated with higher offspring central adiposity measures, objectivized by DXA trunk percent fat, independently of multiple potential confounders including maternal BMI. These results suggest that maternal insulin sensitivity during the second trimester of pregnancy may play a role in the programming of greater central adiposity in the offspring. Although we are not able to determine a causal impact, our findings support the potential benefits of enhancing insulin sensitivity during pregnancy. Improvement in insulin sensitivity can be achieved via lifestyle approaches such as regular physical activity before and during pregnancy. Pharmacologic agents such as Metformin are also accepted in pregnancy and have shown short-term potential benefits for limiting risk of fetal overgrowth, however, long-term offspring metabolic outcomes may not be as favorable [43]. Further research is also required to understand the mechanisms of how maternal insulin sensitivity may impact offspring adiposity programming — particularly centrally located — as well as the related cardiometabolic risk through childhood and beyond.

ABBREVIATIONS

HAPO

Hyperglycemia and Adverse Pregnancy Outcomes

Gen3G

Genetics of Glucose regulation in Gestation and Growth

DXA

dual-energy X-ray absorptiometry

CHUS

Centre Hospitalier Universitaire de Sherbrooke

V1

visit one

V2

visit two

OGTT

oral glucose tolerance test

GDM

gestational diabetes mellitus

BMI

body mass index

GCT

glucose challenge test

HDL

high-density lipoprotein

WHO

World Health Organization

SD

standard deviation

IQR

interquartile range

BMC

bone mineral content

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.