Subtypes of Gestational Diabetes Mellitus Are Differentially Associated With Newborn and Childhood Metabolic Outcomes

Gestational diabetes mellitus (GDM) affects ∼8% of pregnancies in the U.S. and is associated with short- and long-term adverse outcomes in mothers and offspring (1,2). GDM results from an imbalance between pregnancy-related increases in insulin resistance and secretion superimposed on existing predispositions (2,3). Recently, GDM subtypes were described based on a predominant defect in insulin sensitivity or secretion or a mixed subtype of both defects (4,5). Different GDM subtypes appear to be associated with different short-term adverse pregnancy outcomes, although there are inconsistencies across studies in the observed associations (4–10). GDM overall is associated with adverse metabolic and anthropometric outcomes in children, including higher postload glucose levels, lower insulin sensitivity and disposition index, and higher odds for obesity and impaired glucose tolerance (11,12), but whether GDM subtypes are differentially associated with childhood outcomes is not known.

The Hyperglycemia and Adverse Pregnancy Outcome (HAPO) study demonstrated the association of GDM with short-term adverse pregnancy outcomes (13), while the HAPO Follow-Up Study (FUS) demonstrated associations of GDM with child anthropometric and glucose-related traits, including obesity and impaired glucose tolerance (12,14). The goals of the current study were to use HAPO and HAPO FUS data to extend previous studies of the association of GDM subtypes with adverse newborn outcomes and childhood anthropometric and glucose-related outcomes within a large, diverse cohort. We hypothesize that GDM subtypes are differentially associated with newborn and childhood adverse pregnancy outcomes.

The HAPO study was a population-based observational study conducted from 2000 to 2006 that examined associations between maternal hyperglycemia and newborn outcomes in a large multinational cohort (15). All pregnant individuals at a center were eligible to participate unless they met exclusion criteria (16). Participants underwent a standard 75-g oral glucose tolerance test (OGTT) at ∼28 weeks’ gestation; fasting, 1-h, and 2-h plasma glucose samples were collected. Individuals in whom the 2-h plasma glucose reading was diagnostic of diabetes were excluded. OGTT results were blinded to participants and caregivers. Maternal height, weight, and blood pressure measurements were collected at the OGTT. A standardized questionnaire was administered to collect data on smoking, alcohol use, history of diabetes among first-degree relatives, and demographic information. Plasma glucose levels were measured at a central laboratory using enzymatic methods. Maternal insulin sensitivity was calculated per Radaelli et al. (17) and insulin secretion per Stumvoll et al. (18,19), as described previously (7). Cord blood samples were collected at delivery and exported to the central laboratory to measure serum C-peptide and glucose levels. Newborn anthropometric measurements, including birth weight and sum of skinfolds (SSF), were measured within 72 h of birth using calibrated equipment and standardized methods across field centers (15). The protocol was approved by the Institutional Review Board (IRB) at participating sites, and each mother provided written informed consent for herself and her child. Medical records were abstracted to collect details of prenatal, delivery, and neonatal care.

GDM was defined retrospectively using International Association of Diabetes and Pregnancy Study Group (IADPSG)/World Health Organization criteria (20) and glucose levels obtained during the HAPO pregnancy OGTT. GDM subtype classification criteria have been described (21). Briefly, pregnant individuals were categorized as insulin-resistant GDM if their insulin sensitivity was in the lowest 25th percentile of women with normal glucose tolerance (NGT) with preserved insulin secretion (i.e., >25th percentile). Insulin-deficient GDM was defined as insulin secretion in the lowest 25th percentile of women with NGT with preserved insulin sensitivity (i.e., >25th percentile), while mixed-defect GDM was defined as being in the lowest 25th percentile for both insulin sensitivity and secretion. Individuals with neither insulin sensitivity nor insulin secretion in the lowest 25th percentile were considered unclassified GDM.

Secondary analyses categorized GDM subgroups according to timed OGTT measurements: 1) fasting glucose–only GDM (i.e., only fasting glucose level exceeded GDM diagnostic threshold); 2) postload glucose–only GDM (i.e., 1-h and/or 2-h glucose but not fasting glucose exceeded diagnostic threshold); and 3) both fasting glucose and postload glucose GDM.

HAPO FUS was conducted from 2013 to 2016. Of 15,812 eligible mother-child pairs, 4,834 participated (12). Data were analyzed from 4,160 individuals with a fasting glucose measurement and at least one other OGTT measurement or reported having diabetes on treatment and were not excluded due to type 1 diabetes as determined by antibody testing. Details of HAPO FUS have been reported (12). Briefly, height, weight, skin folds, and percentage body fat (by air displacement plethysmography [BOD POD]) were measured at the HAPO FUS visit. Child’s age, first-degree family history of diabetes, and menstrual history for girls were collected from the parent via questionnaire. A 2-h OGTT was performed after an overnight 8-h fast. Samples were drawn for glucose and C-peptide measurement at fasting, 30-min, 1-h, and 2-h. If the participant self-reported diabetes on drug therapy, only fasting glucose was collected. Glucose was measured by hexokinase in the Clinical Chemistry Laboratory of Northwestern Memorial Hospital on a Beckman-Coulter Synchron LX analyzer. C-peptide was measured in the Comprehensive Metabolic Core at Northwestern University. Offspring insulin sensitivity was estimated using a modified Matsuda index that included OGTT glucose and C-peptide levels (22). The insulinogenic index was calculated using C-peptide levels and defined as the ΔC-peptide (0–30 min, nmol/L)/Δglucose (0–30 min, mmol/L) (23).

The HAPO FUS protocol was approved by each center’s IRB. All mothers gave written informed consent for their child, and children assented where required by the local IRB. There was an external Observational Study Monitoring Board.

Multiple linear and logistic regression analyses were used to assess associations of GDM pathophysiologic subtypes and glucose-based subgroups with outcomes in newborns and children. Continuous outcomes for newborns included birth weight, SSF, and cord C-peptide, and dichotomous outcomes included birth weight, SSF, and cord C-peptide >90th percentile and neonatal hypoglycemia as defined in the original HAPO study. Model 1 for newborns included the following covariate adjustments: field center, maternal age, height, mean arterial pressure, gestational age, parity (0/≥1), smoking (yes/no), and alcohol consumption (yes/no) from the HAPO pregnancy OGTT. Model 2 added maternal BMI at the time of the HAPO pregnancy OGTT to model 1 covariates.

Continuous childhood outcomes included percentage body fat, fasting glucose, glucose levels at 1-h and 2-h post-OGTT, insulin sensitivity (Matsuda index), insulinogenic index, and BMI z-score, and dichotomous outcomes included impaired glucose tolerance, percentage body fat >85th percentile, and obesity defined by age- and sex-specific BMI cutoffs from the International Obesity Task Force using Asian-specific cutoffs for Asian children and international cutoffs for all others (24). Model 1 for childhood outcomes included the following covariate adjustments: field center, child age, sex, family history of diabetes in first-degree relatives, and maternal age, height, mean arterial pressure, gestational age, parity (0/≥1), smoking (yes/no), and alcohol consumption (yes/no) from the HAPO pregnancy OGTT. Model 2 added maternal BMI at HAPO OGTT to model 1 covariates.

All analyses used R 4.4.0 software, with statistical significance set at nominal P < 0.05. Results are reported as adjusted mean differences for continuous outcomes and odds ratios (ORs) for dichotomous outcomes with 95% CIs. All estimates were made relative to the “no GDM” reference group.

The data sets generated during and/or analyzed in the current study are available from the corresponding author upon reasonable request.

The association of pathophysiologic GDM subtypes with newborn outcomes was examined in a cohort of 7,970 mother-offspring dyads from 13 HAPO Study field centers (Table 1). Of the 7,970 mothers, 1,241 (15.6%) had GDM. Among the 1,241 participants with GDM, 289 (23.3%) had insulin-deficient GDM, 742 (59.8%) had insulin-resistant GDM, 131 (10.6%) had mixed-defect GDM, and 79 (6.4%) had unclassified GDM. For association with childhood outcomes, 4,160 mother-offspring dyads from 10 of the 13 centers that participated in the HAPO FUS were examined (Table 1). Among the 4,160 mothers, 589 (14.2%) had GDM during the HAPO pregnancy, with 120 (20.4%) categorized as insulin-deficient GDM, 368 (62.5%) as insulin-resistant GDM, 69 (11.7%) as mixed-defect subtype GDM, and 32 (5.4%) as unclassified GDM. Outcomes for the cohort that included all participants with GDM have been previously reported (12,14). Given the small size of the unclassified GDM subtype, subtype analyses focused on individuals with insulin-deficient, insulin-resistant, and mixed-defect GDM. Compared with individuals with insulin-deficient and unclassified GDM, those with insulin-resistant GDM had higher BMI, mean arterial pressure, and insulin secretion and lower insulin sensitivity, while individuals with mixed-defect GDM had values intermediate between the insulin-resistant and insulin-deficient subtypes. Individuals with mixed-defect GDM had higher fasting and postload glucose levels than the two other GDM subtypes. Compared with individuals with insulin-deficient GDM, individuals with insulin-resistant GDM had higher fasting but lower postload glucose levels.

For the glucose-based GDM subgroups, individuals with fasting glucose–only and both fasting and postload glucose GDM had higher BMI and lower insulin sensitivity compared with women with postload glucose–only GDM (Supplementary Table 1). For analyses of newborn outcomes, 81.0% of mothers with insulin-deficient GDM were in the postload glucose–only subgroup, and 14.2% were in the fasting glucose–only subgroup. In contrast, among mothers with insulin-resistant GDM, 45.4% were in the fasting glucose–only subgroup, 34.5% were in the postload glucose–only subgroup, and 20.1% were in the both fasting and postload glucose subgroup (Supplementary Table 2). The distributions among glucose-based GDM subgroups were similar for mothers with child outcomes (Supplementary Table 2).

Newborns of mothers with insulin-resistant and mixed-defect GDM had higher cord C-peptide levels compared with newborns of HAPO participants with normal glucose tolerance (NGT) during pregnancy (Fig. 1 and Table 2). In contrast, cord C-peptide levels in newborns of individuals with insulin-deficient GDM were similar to levels in newborns of mothers with NGT. Consistent with these results, insulin-resistant and mixed-defect GDM were associated with an increased OR for cord C-peptide levels >90th percentile (OR 2.31, 95% CI 1.83–2.93, P < 0.001 and OR 1.87, 95% CI 1.10–3.18, P = 0.02, respectively) in the fully adjusted model. In contrast, risk was not higher in newborns of mothers with insulin-deficient GDM (OR 0.97, 95% CI 0.59–1.60, P = 0.92) (Fig. 1). For neonatal hypoglycemia, the OR in newborns of mothers with insulin-resistant GDM was 2.03 (95% CI 1.16–3.55, P = 0.01) and 0.54 (95% CI 0.13–2.25, P = 0.40) in mothers with insulin-deficient GDM. The OR for neonatal hypoglycemia in mixed-defect GDM was similar to that in insulin-resistant GDM, but the 95% CI crossed the null (OR 2.08, 95% CI 0.63–6.89, P = 0.23). In analyses using glucose-based GDM subgroups, all three subgroups were significantly associated with cord C-peptide levels and cord C-peptide >90th percentile (Supplementary Table 3). The fasting glucose–only subgroup (OR 2.61, 95% CI 1.42–4.82, P = 0.002) but not the other two subgroups was associated with neonatal hypoglycemia.

In the fully adjusted model, insulin-deficient, insulin-resistant, and mixed-defect GDM were all significantly associated with higher birth weight and SSFs compared with offspring of mothers with NGT during pregnancy (Table 2). Similarly, all three GDM subtypes were associated with higher odds for birth weight and SSFs >90th percentile (Fig. 1). Odds for birth weight >90th percentile were similar in offspring from the insulin-resistant (OR 1.64, 95% CI 1.30–2.06, P < 0.001) and insulin-deficient (OR 1.55, 95% CI 1.06–2.25, P = 0.02) GDM subtypes, while odds for birth weight >90th percentile were higher in the mixed-defect GDM subtype (OR 2.69, 95% CI 1.71–4.24, P < 0.001). For SSFs >90th percentile, the ORs were similar for the insulin-resistant (OR 1.85, 95% CI 1.46–2.35, P < 0.001), insulin-deficient (OR 1.77, 95% CI 1.23–2.56, P = 0.002), and mixed-defect (OR 1.81, 95% CI 1.08–3.03, P = 0.03) GDM subtypes. Similar results were obtained for the glucose-based GDM subgroups; all three were associated with higher birth weight and SSFs as well as each of these measures >90th percentile (Supplementary Table 3 and Supplementary Fig. 1).

Neither insulin-deficient nor insulin-resistant GDM was associated with a difference in childhood fasting glucose levels compared with offspring of women with NGT, but the mixed-defect GDM subtype was associated with significantly higher fasting glucose levels (Fig. 2 and Table 2). In contrast, 1-h glucose in offspring of mothers with both insulin-resistant and insulin-deficient GDM was higher compared with offspring of mothers with NGT, while the adjusted mean difference of 1-h glucose levels in offspring of mothers with mixed-defect GDM (11.82, 95% CI 4.70–18.94) was greater than that in insulin-resistant (7.40, 95% CI 4.05–10.75) and insulin-deficient (6.34, 95% CI 0.89–11.78) GDM. For 2-h glucose levels, the adjusted mean differences for the mixed-defect (6.62, 95% CI 1.76–11.48, P = 0.008) and insulin-resistant (3.86, 95% CI 1.61–6.11, P < 0.001) GDM subtypes were significant, but the adjusted mean difference for offspring of mothers with the insulin-deficient GDM subtype compared with mothers with NGT was not. The adjusted mean difference in childhood insulin sensitivity between offspring of individuals with insulin-resistant GDM and mothers with NGT during pregnancy was significant. The other two GDM subtypes were not associated with insulin sensitivity, and none were associated with a difference in the insulinogenic index (Table 2). Among the glucose-based GDM subgroups, the adjusted mean differences for 1-h and 2-h glucose, insulin sensitivity, and insulinogenic index were significant for the postload glucose GDM subgroup, while the adjusted mean differences for fasting and 2-h glucose were significant for the both the fasting and postload glucose GDM subgroup (Supplementary Table 3). The fasting glucose–only GDM subgroup was not associated with any of the continuous glycemic measures.

The final glycemic outcome examined was impaired glucose tolerance (Fig. 2). Both insulin-resistant (OR 2.21, 95% CI 1.50–3.25, P < 0.001) and mixed-defect GDM (OR 3.01, 95% CI 1.47–6.19, P = 0.003) were associated with a higher risk of childhood impaired glucose tolerance compared with women with NGT during pregnancy, while insulin-deficient GDM was not associated with a higher risk (OR 1.28, 95% CI 0.61–2.72, P = 0.52). For the glucose-based GDM subgroups, postload glucose–only GDM (OR 2.04, 95% CI 1.34–3.11, P = 0.001) and both fasting and postload glucose (OR 3.44, 95% CI 1.95–6.07, P < 0.001) GDM subgroups, but not the fasting glucose–only GDM subgroup, were associated with higher risk of impaired glucose tolerance (Supplementary Fig. 2).

Significant differences in childhood percentage body fat and BMI z-score between offspring of mothers with different GDM subtypes and offspring of mothers with NGT during pregnancy were not seen (Table 2). Compared with offspring of mothers with NGT, offspring of mothers with insulin-resistant GDM had a significantly higher risk of percentage body fat >85th percentile (OR 1.35, 95% CI 1.01–1.80, P = 0.04) (Fig. 2). Significantly higher risk of percentage body fat >85th percentile was not observed for offspring of mothers with insulin-deficient (OR 0.93, 95% CI 0.52–1.67, P = 0.81) or mixed-defect (OR 1.26, 95% CI 0.67–2.38, P = 0.47) GDM. Offspring of mothers with insulin-resistant GDM also had a higher risk of obesity (OR 1.53, 95% CI 1.13–2.08, P = 0.007) compared with offspring of women with NGT during pregnancy, while a higher risk was not observed for offspring of women with insulin-deficient (OR 1.08, 95% CI 0.58–1.99, P = 0.81) or mixed-defect (OR 1.33, 95% CI 0.67–2.68, P = 0.42) GDM. For glucose-based GDM subgroups, offspring of mothers in the fasting glucose–only GDM subgroup were at higher risk for obesity compared with offspring of women with NGT (OR 1.67, 95% CI 1.11–2.49, P = 0.01), but none of the glucose-based GDM subgroups were associated with higher risk of percentage body fat >85th percentile.

HAPO showed that GDM defined using IADPSG/World Health Organization criteria was associated with newborn birth weight, percentage body fat, and cord C-peptide >90th percentile, while the HAPO FUS demonstrated association of GDM with higher 30-min, 1-h, and 2-h glucose together with lower insulin sensitivity, insulin secretion, and disposition index in offspring aged 11–14 years (12–14). GDM was also associated with a higher risk for childhood obesity and impaired glucose tolerance (12,14). GDM subtypes have been described and confirmed in multiple cohorts (4,5), and we recently reported that the metabolic and genetic architecture of the GDM subtypes differ (21). The current study expanded on prior reports by demonstrating that GDM pathophysiologic subtypes are specifically associated with cord C-peptide levels and neonatal hypoglycemia in newborns and with childhood glycemic and anthropometric outcomes, including impaired glucose tolerance and obesity.

The association of GDM with adverse newborn and childhood metabolic outcomes has been established (11,12,25), but the association of GDM subtypes with childhood metabolic outcomes has not been reported and provides additional insight into these outcomes. For example, the association of GDM with offspring impaired glucose tolerance in the HAPO FUS (14) appears to be driven primarily by offspring of mothers with insulin-resistant and mixed-defect GDM as opposed to insulin-deficient GDM. GDM overall was not associated with higher child fasting glucose levels in the HAPO FUS (14), but offspring of mothers with mixed-defect GDM had significantly higher fasting glucose levels compared with offspring of mothers with NGT, while offspring of mothers with insulin-resistant and mixed-defect GDM had higher 2-h glucose levels. Consistent with the recent report that maternal insulin sensitivity but not insulin secretion at ∼26 weeks’ gestation is associated with multiple measures of child adiposity at age ∼5 years (26), insulin-resistant but not insulin-deficient or mixed-defect GDM was associated with child percentage body fat >85th percentile and obesity. Mothers with insulin-resistant GDM had higher BMI than the other GDM subtypes, but the associations of insulin-resistant GDM with child percentage body fat >85th percentile and obesity remained significant after adjustment for maternal BMI. There is overlap in genetic variants associated with adult and childhood BMI and obesity (27,28), but the persistence of the significant associations after adjustment for maternal BMI suggest that factors beyond shared genetics and the intrauterine environment characteristic of high maternal BMI likely contributed to the differential association of insulin-resistant GDM with these child anthropometric outcomes.

One potential mechanism underlying differences in the association of GDM with child outcomes is differences in the intrauterine metabolic environment of the GDM subtypes as reflected by differences in the maternal metabolome. As noted, we recently reported differences in the metabolome of insulin-resistant and insulin-deficient GDM (21). Specifically, in the fasting state, only 2-hydroxybutyrate and glycerol were associated with both GDM subtypes, while insulin-resistant GDM was associated with a number of metabolites, including triglycerides, the branched-chain and aromatic amino acids and their metabolic byproducts as well as other metabolites (21). In contrast, insulin-deficient GDM was associated with 3-hydroxybutyrate, fatty acids, and several medium- and long-chain acylcarnitines. Consistent with the maternal metabolome influencing future offspring metabolism, we previously reported that maternal fasting triglycerides and branched-chain amino acids and their metabolic byproducts are associated with cord C-peptide levels independent of maternal BMI and glycemia (29), while others reported that maternal triglycerides are associated with newborn adiposity and birth weight as well as early childhood BMI and adiposity (30–32). Maternal branched-chain amino acid levels have been associated with high and persistently increased offspring BMI between age 1 and 8 years (33). Because higher lipids and specific amino acids can be actively transferred across the placenta, thereby contributing additional fuels in utero and influencing fetal growth and organ development (34,35), differences in the metabolic environment of the GDM subtypes could contribute to differential associations of GDM subtypes with offspring outcomes.

The differential association of GDM subtypes based on insulin resistance and secretion with adverse childhood metabolic outcomes allows for the identification of at-risk offspring. To determine whether this approach provides value over using glucose levels alone to identify at-risk offspring, mothers with GDM were classified based on which time point glucose levels exceeded diagnostic thresholds for GDM during the HAPO pregnancy OGTT. Overall, insulin-resistant and mixed-defect GDM could not be easily defined by one specific glucose phenotype. Consistent with this, there were differences in the ability of the two classification approaches to identify offspring at risk for adverse metabolic outcomes. Offspring of mothers with insulin-resistant GDM were at higher risk of impaired glucose tolerance, percentage body fat >85th percentile, and obesity at age 11–14 years, while insulin-deficient GDM was not associated with higher risk for any of these outcomes. In contrast, fasting glucose–only GDM was associated with higher risk for childhood obesity, postload glucose–only GDM was associated with higher risk for impaired glucose tolerance, and neither subgroup was associated with higher risk for percentage body fat >85th percentile. These findings suggest that subtyping GDM based on insulin resistance and secretion is more effective in identifying a group of offspring at risk for adverse metabolic outcomes during childhood.

The association of GDM subtypes with newborn outcomes has been reported previously. The initial report of GDM subtypes by Powe et al. (4) demonstrated that insulin-resistant but not insulin-deficient GDM was associated with a higher birth weight z-score, large for gestational age (LGA) birth, and neonatal hypoglycemia. A subsequent study by the same group reported that insulin-resistant and insulin-deficient but not mixed-defect gestational glucose intolerance (i.e., an abnormal glucose challenge test result) were associated with higher odds for LGA birth (36). Benhalima et al. (6) reported that insulin-resistant GDM was associated with neonatal hypoglycemia. Results in studies examining association of the subtypes with LGA birth or macrosomia have been conflicting but were generally small in the number of GDM subtype cases and included participants with GDM whose hyperglycemia in pregnancy was treated (6,8–10). Madsen et al. (7) reported in a study that included 6,337 HAPO participants, 2,636 of whom were included in the current study, that the three GDM subtypes were associated with higher birth weight, SSF, and LGA birth. Thus, despite some conflicting results, together, the present and earlier studies suggest that compared with individuals with NGT during pregnancy, GDM, regardless of subtype, is associated with higher birth weight, LGA birth, and greater adiposity at birth as reflected by SSFs.

Cord C-peptide, which reflects fetal insulin secretion, was also examined. A distinct difference between subtypes was the association of insulin-resistant and mixed-defect GDM but not insulin-deficient GDM with cord C-peptide levels and odds for cord C-peptide levels >90th percentile. Madsen et al. (7) also demonstrated association of insulin-resistant and mixed-defect GDM but not insulin-deficient GDM with cord C-peptide levels >90th percentile. One potential explanation is that insulin-resistant and mixed-defect subtypes are characterized by higher fasting glucose levels compared with the insulin-deficient subtype. The HAPO protocol was designed to ensure that participants were in the fasting state during labor and delivery (16); thus, offspring of participants with the insulin-resistant and mixed-defect GDM subtype may have been exposed to higher glucose levels during labor and delivery, resulting in higher fetal insulin secretion.

Another potential mechanism would be shared genetics. We recently reported that variants in BACE2 were associated with maternal insulin sensitivity and C-peptide levels fasting and 1 h after a glucose load at 28 weeks’ gestation (37). In preliminary analyses using offspring genotypes, the lead single nucleotide polymorphism in BACE2 (rs28360503) was associated with cord C-peptide levels (P = 4.89 × 10⁻⁶, unpublished observations). BACE2, which is expressed in pancreatic islet β-cells and pancreas, encodes β-site amyloid precursor protein cleaving enzyme 2 (38,39). Finally, as described above, differences in the metabolome of mothers with different GDM subtypes may also contribute to their differential association with cord C-peptide levels.

Our study had several strengths. First, the cohort was multiancestry and from multiple international sites, enhancing the generalizability of the results. Second, by examining both newborn and child outcomes within the same cohort, the study provides a comprehensive picture of longitudinal associations of GDM subtypes with newborn and subsequent child anthropometric and glycemic outcomes. Finally, participants and caregivers in the HAPO study were blinded to the OGTT results; thus, newborn and child outcomes were not confounded by treatment.

The study also had limitations. While the cohort was large, power was limited in some subtypes, especially for childhood outcomes. Some GDM cases did not fit into a specific GDM pathophysiologic subgroup, but this unclassified subgroup was small, preventing further study. Finally, participants with significant hyperglycemia (1.8% of HAPO study participants overall) were unblinded, resulting in those with severe GDM being excluded.

In conclusion, differential associations of GDM subtypes with newborn and childhood outcomes were observed. In general, offspring of mothers with insulin-resistant and mixed-defect GDM had worse outcomes compared with offspring of individuals with insulin-deficient GDM. This study highlights the potential importance of more careful characterization of women with GDM to better identify offspring at higher risk for adverse short- and long-term metabolic outcomes based on the GDM subtype. This may allow targeted preventative interventions early in life to higher-risk offspring, thereby improving their metabolic health outcomes.

Funding. This study was funded by National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases grants 1U01DK094830 and DK117491, and by Eunice Kennedy Shriver National Institute of Child Health and Human Development grants HD34242 and HD34243.

The study funder was not involved in the design of the study, the collection, analysis and interpretation of data or writing the report, and did not impose any restrictions regarding publication of the report.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.E.O. wrote the first draft of the manuscript. M.E.O., M.-F.H., D.M.S., and W.L.L. were involved in the conception, design, and conduct of the study and the analysis and interpretation of the results. Y.Y. and A.K. and were involved in the analysis and interpretation of the results. J.L.J. was involved in interpretation of the results. All authors edited, reviewed, and approved the final version of the manuscript. W.L.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Cuilin Zhang.