Diabetes in Pregnancy for Mothers and Offspring: Reflection on 30 Years of Clinical and Translational Research: The 2022 Norbert Freinkel Award Lecture

In his 1980 Banting Award lecture, “Of Pregnancy and Progeny” (1), Norbert Freinkel hypothesized that abnormal nutrient mixtures in the mother could disrupt fetal development and lead to long-range changes in offspring, including behavioral, anthropometric, and metabolic teratology—the concept of “fuel-mediated teratogenesis.” Research over the ensuing 40 years has largely supported his original ideas, providing a wealth of specific information on the breadth and depth of long-term health effects related to maternal hyperglycemia and metabolic aberrations in pregnancy. Those effects mimic a double-edged sword, affecting both mothers and their offspring in a vicious cycle that can impact multiple generations. The growing body of knowledge in this field has helped to increase public health awareness of the associated risks and served as a base for development of strategies to improve maternal and child health.

However, gestational and preexisting diabetes continue to impart significant morbidity to mothers and offspring, and rising rates of gestational diabetes mellitus (GDM), type 2 diabetes (T2D), and type 1 diabetes (T1D) in women of reproductive age amplify the impact of these conditions within and across generations. In this article, I provide an overview of work that my colleagues and I have done to better understand the health impacts of diabetes during pregnancy in four areas: 1) future diabetes risk after pregnancies complicated by GDM, including racial and ethnic disparities; 2) pathophysiology of GDM and subsequent diabetes in one racial and ethnic group, namely, Hispanic women; 3) diabetes prevention and β-cell preservation following GDM; and 4) multiple potential developmental effects of maternal diabetes on offspring that vary by the timing of exposure and severity of maternal diabetes during pregnancy.

My work in the field of GDM research started in the 1990s while working closely with two pioneers of GDM research, Dr. Thomas Buchanan and Dr. Siri Kjos, at the University of Southern California (USC). Together we made several novel findings by analyzing data collected by Dr. Kjos from a clinical cohort of high-risk pregnancies at the Los Angeles County/USC Medical Center. A large majority of the patients were Hispanic, and GDM was diagnosed by National Diabetes Data Group criteria (2). We found that these women had almost 50% risk of developing diabetes within 5 years after the index pregnancy. Elevated glucose levels during oral glucose tolerance tests in pregnancy or postpartum, along with a relatively early gestational age at diagnosis of GDM, predicted the highest risk of future diabetes (3,4). In addition, one subsequent pregnancy tripled the long-term diabetes risk and each 10 lb of weight gain during follow-up doubled the diabetes risk. These two effects were independent of each other and all other identified risk factors (5). Furthermore, progestin-only contraception while breastfeeding added additional risk, while combination hormonal contraceptives had no detrimental effects (6–8). These findings provided important information relevant to patient management. They also provided hints into the pathogenesis of diabetes after GDM. Greater hyperglycemia during and after pregnancy identified women who were already close to diabetes; relatively little deterioration was required to develop diabetes during follow-up. However, data for individuals with weight gain, individuals who used progestin-only contraception, and, in particular, individuals who had one additional pregnancy suggested that insulin resistance may be accelerating the decline in β-cell compensation that leads to diabetes. That suggestion was borne out in subsequent studies discussed below.

Meta-analysis conducted by others showed that women with GDM had an approximately sevenfold higher long-term diabetes risk than women without GDM (9). Moreover, GDM was known to be rising in frequency across all racial and ethnic groups (10), in parallel with the epidemic of obesity and increasing trend in T1D and T2D in children and adolescents (11). However, less was known about whether the long-term diabetes risks after GDM are the same across racial and ethnic groups. Using electronic medical records from Kaiser Permanente Southern California (KPSC), we found that 10.3% of the 139,334 singleton pregnancies delivered at KPSC between 1995 and 2008 had GDM by the Carpenter and Coustan criteria (12,13). The rate was highest for women who self-identified as Asian and Pacific Islander (17.1%), followed by self-identified Hispanic (11.4%), non-Hispanic White (NHW) (7.4%), and Black (6.9%) women (13) (Fig. 1A). We assessed incidence rates of diabetes by race and ethnicity for up to a decade after delivery of GDM pregnancies and age-matched and race- and ethnicity-matched control participants who delivered in the same year. As expected, all the GDM groups had a much higher rate of diabetes during follow-up than their counterparts who did not have GDM (Fig. 1B). However, the pattern of variation in long-term diabetes rates among racial and ethnic groups with GDM did not parallel the variation in their rates of GDM during pregnancy. Specifically, Black women had the lowest rate of GDM during pregnancy but the highest rate of diabetes over the next decade. Asian and Pacific Islander women had the opposite pattern, with the highest rate of GDM but the second lowest rate of subsequent diabetes. Hispanic women demonstrated the best concordance, with the second highest rate of GDM and the second highest rate of subsequent diabetes. NHW women had a GDM rate that was nearly as low as that of Black women combined with the lowest rate of diabetes during follow-up (Fig. 1C). Compared with counterparts who did not have GDM, the relative risks (95% CI) of developing diabetes associated with GDM were 9.9 (7.5–13.1) for Black, 7.7 (6.8–8.7) for Hispanic, 6.5 (5.2–8.0) for NHW, and 6.3 (5.0–7.9) for Asian and Pacific Islander women after adjusting for potential confounders. Adjusting for maternal obesity measured by prepregnancy BMI reduced the risk estimates associated with GDM for all racial and ethnic groups but did not explain the differences of future diabetes risk across racial and ethnic groups (13). Thus, this large observational study revealed important racial and ethnic differences between GDM risk and future diabetes risk. The discrepancies suggest different pathophysiologies among groups. Mechanistic studies are needed to understand the differences in more detail, along with influences of genetics, environment, lifestyle, and other factors that may contribute to these differences.

Pregnancy is normally attended by progressive insulin resistance that begins near midpregnancy and progresses through the third trimester due to increased maternal adiposity and the insulin-desensitizing effects of hormonal products of the placenta (14). This is a normal pathophysiologic response of pregnancy that is thought to ensure adequate nutrient supply from the mother to meet the needs of a growing fetus (15).

To understand the pathophysiology of GDM and future diabetes risk, our group designed a detailed study of glucose regulation in Hispanic women with GDM diagnosed by using the criteria of the Third International Workshop Conference on GDM (16). We recruited 175 Hispanic women during the third trimester of pregnancy, 150 with GDM and 25 without GDM, and they were frequency matched on age, prepregnancy BMI, and gestational age to the GDM group. All women had a 100-g oral glucose tolerance test, euglycemic clamp at physiologic hyperinsulinemia, and frequently sampled intravenous glucose tolerance tests (IVGTT) between 28 and 34 weeks gestation. Compared with their non-GDM counterparts, women with GDM had significantly lower glucose clearance, higher glucose production, higher levels of free fatty acids at baseline of the clamp, and less stimulation of glucose clearance and suppression of glucose production and fatty acid levels compared with controls (17). The GDM group also had a marked reduction in the disposition index from the IVGTTs, a measure of acute β-cell compensation for the degree of insulin resistance (17). These results demonstrate that GDM in Hispanic women is characterized by all the major defects in glucose regulation, i.e., insulin resistance in muscles, liver, and fat and reduced β-cell compensation, observed in patients with T2D, albeit to a lesser degree. Testing of a large subset of these women ∼6 months after pregnancy revealed that their defect in β-cell compensation was similar to that observed during the third trimester (−69% vs. −62%, respectively) (14). Similar patterns were observed for Black women (18) and White women (19). Thus, GDM in large part represents the detection of pre-existing defects in β-cell compensation for chronic insulin resistance in relatively young women.

We continued following these women with oral glucose tolerance tests and IVGTT at 12−15-month intervals for up to 12 years after the index pregnancy unless their fasting glucose reached 140 mg/dL or they were lost to follow-up. Forty-three percent developed diabetes, with an average annual incidence rate of 7.2% (20). Body weight and fasting and postchallenge glucose increased significantly; β-cell compensation measured by disposition index declined significantly (20,21). The strongest factor associated with falling β-cell compensation was weight gain, an association explained statistically by a combination of increasing insulin resistance (40%), falling adiponectin, and rising C-reactive protein (31% combined effect) (21). This study also revealed that low insulin sensitivity and β-cell function at baseline and the rate of their deterioration during follow-up contributed to the risk of T2D development (20). In addition, high habitual daily caloric intake was associated with worsening insulin resistance, falling β-cell compensation, and rising glucose levels independent of weight gain (22). This study provides the longest follow-up of which we are aware, and it uses detailed physiological measurements to characterize the natural history of glucose regulation following pregnancies complicated by GDM. It provides evidence that obesity and related metabolic changes may be important drivers of the β-cell decline that leads to diabetes. Interventions to increase exercise and reduce body weight should reduce diabetes risk, as was demonstrated in the Diabetes Prevention Program (23).

An important observation from this longitudinal study is that the relationship between changes in circulating glucose levels and changes in β-cell function is nonlinear (24). Both fasting and 2h glucose rise slowly with the decline in β-cell compensation over many years (Fig. 2 in Xiang et al. [24]), followed by more rapid increase in glucose per change in β-cell compensation when compensation gets very low (∼10% of normal). These findings define a pathogenesis for T2D in Hispanic women with prior GDM that is characterized by a relatively long-term decline in acute β-cell compensation for chronic insulin resistance, attended initially by slowly rising glucose levels. Only relatively late in this process do glucose levels rise rapidly and into the diabetic range. The implication of these findings is that simply measuring glucose itself may not reveal the underlying stage of diabetes risk.

GDM presents an opportunity for diabetes prevention. Based on the results of our longitudinal studies of glucose regulation described above, we hypothesized that amelioration of chronic insulin resistance would preserve β-cell function and delay or prevent the onset of T2D in high-risk Hispanic women with prior GDM. We randomized 266 Hispanic women with history of GDM to placebo or the insulin-sensitizing thiazolidinedione medication troglitazone in a 1:1 ratio (TRIPOD [Troglitazone in the Prevention of Diabetes] study) (25). During a median follow-up of 30 months on blinded medication, treatment reduced the diabetes incidence by 55% compared with placebo and prevented the worsening of β-cell compensation over time (25,26). Open-label continuation of treatment with another thiazolidinedione drug, pioglitazone (the PIPOD [Pioglitazone in Prevention of Diabetes] study), revealed that pioglitazone stopped the decline in β-cell compensation that occurred during placebo treatment in the TRIPOD study and maintained the stability of β-cell compensation that had occurred during troglitazone treatment in the TRIPOD study (27). The risk of diabetes was lowest in women with the largest reduction in total IVGTT insulin area after 3 months of treatment in the TRIPOD study and after 1 year of treatment in the PIPOD study. These findings suggest β-cell unloading is important in diabetes prevention, a concept that has been used in β-cell preservation outside of pregnancy by treating obesity (28).

Thiazolidinedione treatment in the TRIPOD study was also associated with a 31% reduction in the progression of subclinical atherosclerosis, measured by ultrasound as carotid intima-media thickness (CIMT), in these premenopausal women (29,30). Importantly, reduction in CIMT progression occurred only in treated women who had an increase in insulin sensitivity (29). Continued treatment with open-label pioglitazone slowed CIMT progression in women who had been on placebo in the TRIPOD study and maintained a relatively low rate of progression in women who had been on troglitazone (30). The similarity of findings between the PIPOD and TRIPOD studies supports a class effect of thiazolidinedione drugs to enhance insulin sensitivity, reduce insulin secretory demands, preserve β-cell function, and slow the progression of subclinical atherosclerosis.

Freinkel proposed that pregnancy is a form of tissue culture (1) in which the mix of nutrients and other substances provided from mother to fetus affects fetal development, with both short- and long-term implications for health of the offspring. Potential mechanisms include fetal overnutrition, inflammation and altered cytokines, oxidative stress, immune activation and hyperinsulinemia, and possibly epigenetic changes (31–36).

Newborns being large for gestational age (LGA) is the most common short-term impact associated with maternal hyperglycemia during pregnancy (37,38). To examine this impact across different racial and ethnic groups, we conducted a large birth cohort study including ∼350,000 women with singleton births at KPSC from 1995–2010 and assessed rates of LGA for pregnancies complicated by GDM from ten racial and ethnic groups (Hispanic, NHW, Black, Filipino, Chinese, Asian Indian, Vietnamese, Korean, Japanese, and Pacific Islander) (39). Quantile regression was applied to birthweights with adjustment for gestational age at delivery and parity in each racial and ethnic and sex subgroup, and stratified by birth year to define birth weight percentiles for that subgroup. LGA was defined as birth weight >90th percentile specific to each racial and ethnic group. We found that infants of Black women with GDM had the highest LGA rate (17.2%), followed by infants of Pacific Islander (16.2%), Hispanic (14.5%), NHW (13.1%), Asian Indian (12.8%), and Filipino (11.6%) women with GDM. All other Asian women with GDM had LGA rates comparable to the expected 10% (9.6–11.1%). Compared with NHW women with GDM, the LGA rate was significantly greater for Black women with GDM after adjustment for maternal characteristics except for obesity measures. Further adjustment for maternal prepregnancy BMI and gestational weight gain in the subcohort with available data greatly attenuated the elevated LGA risk for Black women. Our results suggest that variation across racial and ethnic groups in excessive fetal growth associated with GDM can be explained to a significant degree by maternal obesity and gestational weight gain. Studies are needed to test whether reducing obesity and appropriate weight gain will eliminate the disparities in LGA risk associated with GDM.

The investigation of the effect of GDM on obesity in offspring is confounded by maternal obesity, which is itself an important risk factor for both conditions. Excessive gestational weight gain (EGWG) above the Institution of Medicine recommendation and breastfeeding are also important factors affecting obesity risk, especially for infancy and early childhood. We assessed the interplay among four risk factors (GDM, prepregnancy obesity, gestational weight gain, and breastfeeding) and their independent contributions to childhood overweight at 2 years of age. Using electronic medical record data from 15,710 mother–offspring pairs with delivery at KPSC in 2011, we found that women with GDM were less likely to have EGWG than women without GDM (40). Maternal GDM was not associated with childhood overweight at 2 years of age. Maternal preexisting obesity and overweight, EGWG, and breastfeeding ≤6 months were all significantly, independently, and positively associated with overweight in children (40).

We extended this study to assess growth trajectory from ages 2 to 6 years, added maternal preexisting T1D and T2D as potential risk factors, and separated GDM to those who did or did not receive antidiabetes medications during pregnancy (41). Data were from 71,892 children born at KPSC from 2007–2011. Three distinct BMI trajectory groups were identified in the children: group 1 (59% of the cohort) had stable relatively lower BMI; group 2 (35% of the cohort) had stable intermediate BMI; and group 3 (6% of the cohort) had high and increasing BMI over time (41). Being in the high and increasing BMI group (i.e., group 3) was strongly associated with maternal prepregnancy obesity and overweight, modestly associated with maternal T1D, T2D, GDM with medication treatment and EGWG, and slightly associated with relatively short duration of breastfeeding (≤6 months). Individuals with GDM not receiving medication treatment during pregnancy had little association with high and increasing BMI over ages 2–6 years (41).

We further extended our studies to BMI trajectory in offspring from birth to age 10 years (42). Data were from 218,227 singleton children born in 2008–2015 at KPSC. We saw a catchup growth pattern associated with diabetes during pregnancy (Fig. 2A). On average, BMI was 0.1–0.2 kg/m² (∼1% relative to the BMI in the group not exposed to diabetes) lower at 6 months for all diabetes-exposed groups (42). At age 3, the growth patterns started separating with the highest BMI in maternal T1D and T2D groups, followed by the GDM group receiving medication, then the GDM group not receiving medication, and finally the groups with no maternal diabetes. The absolute differences in BMI across these exposure groups increased with age, being ∼0.5 kg/m² (∼3%) higher at age 5 years and ∼1.5–2 kg/m² (10%) higher at age 9–10 years (Fig. 2A and Table 2 in Sidell et al. [42]). By age 7 years, BMI was significantly higher for the GDM group receiving medication compared with the group with no diabetes. Our data showed a hierarchical BMI growth pattern in offspring exposed to different types of diabetes during pregnancy after adjusting for important covariates, including maternal obesity, starting as early as 3 years of age.

Our group is one of the first to carefully delineate potential links between diabetes in pregnancy and neurobehavior developmental disorders in offspring. Using a birth cohort of 322,323 children born at KPSC from 1995 to 2009 with follow-up until 2012, we found that GDM, as a whole, was not associated with a diagnosis of autism spectrum disorders (ASD) in children. However, gestational age at GDM diagnosis was negatively associated with risk of ASD, suggesting that early exposure to elevated glucose during pregnancy was associated with ASD (43). We divided the women with GDM into three tertiles based on gestational age at diagnosis of GDM, rounded to the nearest week for clinical relevance: GDM diagnosed at 26 weeks or earlier (mean of 16 weeks; 30% of women with GDM), after 26 weeks but prior to 30 weeks (mean of 28 weeks; 30% of women with GDM), and 30 weeks or later (mean of 32 weeks; 40% of women with GDM). We found that exposure to GDM diagnosed by 26 weeks gestation was associated with risk of ASD in offspring, but GDM diagnosed after 26 weeks gestation was not (43). Of note, individuals with GDM diagnosed early had higher rates of being treated with antidiabetes medications during pregnancy, but that treatment was not associated with risk of ASD after adjustment for gestational age at GDM diagnosis (43). Preexisting T1D and T2D were both associated with increased risk of ASD in children. The hierarchy of risk was highest for T1D, followed by T2D and then GDM diagnosed at ≤26 weeks gestation (Fig. 2B) (44).

To test whether the association between maternal diabetes and ASD might be related to severity of hyperglycemia in early pregnancy, we analyzed first-trimester HbA_1c levels for an association with ASD risk in children. We found that HbA_1c >6.5% in the first trimester (99% had a diagnosis of T1D, T2D, or GDM) was associated with a 1.8-fold increased risk of ASD in offspring (45). More recent work from my group found that maternal diabetes during pregnancy was associated with increased risk of ASD and associated gastrointestinal disturbances (46), and maternal preexisting diabetes and GDM diagnosed at ≤26 weeks, as well as maternal obesity. were all associated with increased scores on autism screening questionnaires among typically developing children (47).

Taken together, our studies suggest that timing of exposure is important, and abnormal glycemia in early pregnancy plays an important role in a child’s ASD risk. Our observations are consistent with studies showing that the early trimester intrauterine environment (e.g., folic acid supplementation and environmental exposure) is important for the risk of ASD in children (48,49).

Using the KPSC birth cohort, which includes 333,182 children born in 1995–2012, we comprehensively examined risks of another neurodevelopmental condition, attention deficit hyperactivity disorder (ADHD), in association with different types, severity, and timing of exposure to maternal diabetes during pregnancy. We found significant associations between maternal T1D and T2D in pregnancy and ADHD risk in children. For GDM, we found no association overall with ADHD and, unlike ASD, no association between gestational age at diagnosis of GDM and ADHD risk. Instead, we found a significant association between individuals with GDM receiving antidiabetes medication treatment and ADHD in children (50). The ADHD risk hierarchy was individuals with T1D followed by those with T2D and then those with GDM receiving antidiabetes medication. Having GDM but not receiving antidiabetes medication is not associated with ADHD risk (Fig. 3A). Our results suggest that severity of maternal diabetes (T1D vs. T2D vs. GDM requiring antidiabetes medications) influences the risk of ADHD in offspring. The similarity and difference between timing and severity of diabetes on ASD and ADHD risk could represent both overlaps and differences in the etiology of the two conditions.

We also examined risks of asthma, depression, and anxiety in children following exposure to different types of diabetes and timing of maternal diabetes during pregnancy. For asthma, we found that the risk of asthma was predominately observed following exposure to maternal preexisting T2D, the risk was small for individuals with GDM receiving medication, and the risk was not increased for individuals with GDM not receiving medication among 97,554 singletons born at KPSC between 2007 and 2011 (Fig. 3B) (51). We did not assess the presence of preexisting T1D due to the small sample size.

For the association of depression and anxiety with maternal diabetes, we found that individuals with maternal preexisting T1D, T2D, and GDM receiving medications were all significantly associated with depression and anxiety from childhood to young adulthood among 439,590 singletons born at KPSC between 1995 and 2015. There was no association with depression and anxiety for individuals with GDM who were not treated with antidiabetes medication during pregnancy. The significant associations were predominantly observed for offspring ages 5–12 and 13–18 years (52).

In summary, our large and multiethnic birth cohort studies show that diabetes during pregnancy can affect a broad spectrum of short- and long-term health outcomes in offspring. The hierarchy of risk, from high to low, is preexisting T1D and T2D, followed by GDM diagnosed early or GDM treated with antidiabetes medications during pregnancy. Individuals with GDM that were diagnosed late during gestation, or who did not receive antidiabetes medications, did not increase these long-term risks in their offspring. Our results suggest that both the degree of glycemic level and timing of exposure matter for long-term offspring health (Fig. 4). Taken together, our results support Freinkel’s prescient proposal more than 40 years ago that fuel-mediated teratogenesis may manifest in many ways, including behavioral, anthropometric, and metabolic teratology (1).

Despite much progress over the past 40 years, many unknowns remain at the intersection of diabetes and pregnancy. Three areas where gains in knowledge could have important implications for the personal care of diabetes during pregnancy should be highlighted. First, we need to understand racial and ethnic differences in the pathophysiology of GDM and subsequent diabetes development, such that targeted approaches specific to each racial and ethnic group can be developed to mitigate the risk. Second, more studies are needed to understand the relative contributions of maternal factors such as obesity, weight gain, hyperglycemia, and hyperlipidemia (15) to perinatal and long-term risks in offspring of mothers with diabetes. Such understanding could have important implications for antenatal management to mitigate risk. Third, much work is needed to understand mechanisms, timing of exposure, and severity of maternal diabetes that predisposes to neurocognitive disorders in children, so that clinical approaches to early detection and treatment (53) can be developed and tested to help women at high risk manage their glucose levels (54) to minimize both short- and long-term impacts to the offspring.

Acknowledgments. I thank my colleague and mentor Dr. Thomas Buchanan for his inspirational education and guidance in diabetes pathophysiology and gestational diabetes as well as for his friendship and support; Dr. Richard Bergman for providing the scholarship to support my initial education and training in diabetes pathophysiology; Drs. Siri Kjos, Richard Watanabe, and Katie Page for their friendship and collaborations; my extraordinary team members at KPSC and the USC GDM Study Group who have made all of this research possible; and my parents, my husband (Jeff), and my children (Marshall and Brandon) for their love and support. Special thanks go to patients who participated in our GDM studies and to patients of Kaiser Permanente for helping us improve care by using the information collected by KPSC electronic health record systems.

Funding. The work reported here reflected work with funding from multiple funders, especially the National Institutes of Health (R01DK46374, R21DK066243, R01DK61628, R01DK100302, R56ES028121, R01DK116858, and R01ES029963) and KPSC Direct Community Benefit funds.

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