Epidemiology and Therapeutic Strategies for Women With Preexisting Diabetes in Pregnancy: How Far Have We Come? The 2021 Norbert Freinkel Award Lecture

The field of diabetes in pregnancy has witnessed tremendous changes over the past 30 years both in epidemiology and management. The worldwide explosion of cases of type 2 diabetes has posed great public health challenges, and its changing demographics, with a shift to younger individuals, including women of childbearing age, has created the potential for increased perinatal adverse outcomes for both mother and child. In this review, I will examine how we have tackled these issues through gaining a greater understanding of the epidemiology of diabetes in pregnancy and by testing new therapeutic strategies both for type 1 and type 2 diabetes.

The prevalence of preexisting diabetes (type 1 and type 2 diabetes) in pregnancy has grown enormously in the last two decades (1–6), primarily due to a marked rise in the prevalence of type 2 diabetes in pregnancy. One of the first studies to describe this phenomenon was a study examining the prevalence of diabetes in hospital deliveries in the U.S. between 1994 and 2004, where an increase of 367% was noted in women with type 2 diabetes (3) (Table 1). In a population-based cohort study of 1.1 million women who delivered in Ontario between 1996 and 2010, we found that the rate of both gestational diabetes and preexisting diabetes during pregnancy had more than doubled over this 14-year span, going from 2.7% to 5.6% in women with gestational diabetes and from 0.7% to 1.5% in women with preexisting diabetes (5). In women over 30 years old, 1 in 10 had some form of diabetes in pregnancy. While the data available did not allow us to differentiate between women with type 1 versus type 2 diabetes in our study, we suspected, given the marked rise in prevalence of type 2 diabetes in the general population and in our clinics, that the growth in numbers was mostly due to a rise in type 2 diabetes in pregnancy. This was corroborated by other cohort studies (1,2). In a population-based cohort study in Scotland, while the prevalence of type 1 diabetes in pregnancy did increase by approximately 40%, the rise was 90% in women with type 2 diabetes (1). In a population-based study in Sweden in similar years, the prevalence of type 2 diabetes in pregnancy rose by 111% (2). The reason for such dramatic increases in prevalence coincided with marked increases in type 2 diabetes in general (7), thought to be related to rising rates of obesity, unhealthy diet, and sedentary lifestyle, creating engineered diabesity. Also contributing, however, was a lowering in the age of onset of diabetes (8), leading to more women having diabetes in the child-bearing years.

Indigenous populations appear to be at highest risk of type 2 diabetes in pregnancy. According to the Australian Institute for Health Welfare, Indigenous women of Australia are 10 times more likely to have type 2 diabetes in pregnancy than non-Indigenous women (9). In the Pregnancy and Neonatal Diabetes Outcomes in Remote Australia (PANDORA) cohort study, which looked at a cohort of Indigenous and non-Indigenous women in the Northern Territory of Australia, 28% had type 2 diabetes in pregnancy compared with only 3.8% in non-Indigenous women (10). Indigenous women with type 2 diabetes in pregnancy also have been found to have worse pregnancy outcomes than non-Indigenous women with type 2 diabetes (11). Clearly this is a population that deserves further attention.

Mothers with gestational diabetes are at increased risk of developing type 2 diabetes. While many small postpartum follow-up studies have been reported, a large population-based study had not yet been done. We undertook a large population-based study of all deliveries in Ontario between 1995 and 2002, with a median follow-up of 9 years. We found that, among the 659,000 deliveries, 21,823 had gestational diabetes. The prevalence of diabetes postpartum was 19% after 9 years (12). In a subsequent article, we extended our follow-up to 15 years (1994–2008) and examined over 1 million deliveries in Ontario. We found that the cumulative probability of developing diabetes over this period was over 40% (13). We also found that the diagnosis of preeclampsia or gestational hypertension was an independent risk factor for the development of diabetes postpartum, and in those with both preeclampsia/gestational hypertension and gestational diabetes, the risk was additive.

The disturbing observation of a dramatic rise in the prevalence of diabetes in pregnancy underscored the importance of defining its effect on maternal and fetal health.

In a population-based cohort study of 995,990 deliveries in Ontario from 2002 to 2015, we found that, compared with women without diabetes, women with preexisting type 1 or type 2 diabetes (n = 16,283) had an almost 3-fold increased risk of preeclampsia/gestational hypertension and preterm birth, 6-fold increased risk of congenital anomalies, 20-fold increased risk of LGA, and 1.6 times the risk of perinatal mortality (14).

Women with either type 1 or type 2 diabetes have adverse pregnancy outcomes; however, there are some differences in characteristics and outcomes between them (15). Compared with women with type 2 diabetes, women with type 1 diabetes have a longer duration of diabetes, more diabetic retinopathy and nephropathy during pregnancy, higher HbA_1c, and more large-for-gestational-age (LGA) infants and preterm births (15). On the other hand, women with type 2 diabetes are more likely to have prepregnancy obesity, are more likely to be in the lowest deprivation quintile, and have more chronic hypertension (15,16). Women with type 2 diabetes begin pregnancy with increased insulin resistance, and this increases during pregnancy. Hence, they often require very high doses of insulin to reach the pregnancy glucose targets. Their insulin doses tend to be much higher than those in women with type 1 diabetes. This can lead to discomfort, high cost, and noncompliance.

Women with type 2 diabetes come less often for prepregnancy care and, despite lower HbA_1c, have more stillbirths and perinatal mortality than women with type 1 diabetes and a similar rate of congenital anomalies (15). In a large, population-based study in the U.K. of 17,375 women with type 1 and type 2 diabetes in pregnancy seen in 2014–2018, more women with type 2 diabetes were in the most deprived income quintile and displayed fewer markers of pregnancy preparedness, including taking 5 mg folic acid and avoiding potentially teratogenic medications like statins and angiotensin-converting enzyme (ACE) inhibitors or angiotensin receptor blockers (16). Compared with women with type 1 diabetes, women with type 2 diabetes had more infants who were small for gestational age (SGA) and had more perinatal deaths (stillbirths and neonatal deaths). Independent predictors of perinatal mortality in the whole cohort included not achieving HbA_1c <6.5% in the third trimester, being in the lowest deprivation quintile, and having type 2 diabetes. The underlying mechanism for this increased perinatal mortality is not known; however, several factors associated with perinatal mortality are clustered in women with type 2 diabetes, including advanced maternal age, obesity, chronic hypertension, and increased social/economic deprivation. According to the World Health Organization, social determinants of health, which include social and environmental factors, account for 50–60% of health outcomes and health care disparities (17). These factors may very well be playing an important role in the adverse outcomes seen in women with type 2 diabetes during pregnancy.

Increasingly we are seeing women with undiagnosed type 2 diabetes during pregnancy. These women are often diagnosed with gestational diabetes; however, they have more adverse outcomes than women with gestational diabetes. Early studies from South Africa showed that women with unrecognized type 2 diabetes had the highest prevalence of perinatal mortality, higher than those with known type 2 diabetes (18). In a more recent retrospective population-based cohort study, we looked at women with undiagnosed type 2 diabetes, which we defined as those with gestational diabetes but who were found to have diabetes in the first year postpartum and therefore presumably had undiagnosed diabetes in the pregnancy (14). We found that those with undiagnosed type 2 diabetes had rates of preeclampsia/gestational hypertension, congenital anomalies, and perinatal mortality very similar to those of women with preexisting diabetes, and these rates were much higher than those seen in women with true gestational diabetes. The strongest predictors of undiagnosed type 2 diabetes included early gestational diabetes diagnosis (under 20 weeks), a previous diagnosis of gestational diabetes, and chronic hypertension. Others have also found those diagnosed with gestational diabetes <12 weeks had pregnancy outcomes similar to those with preexisting diabetes (19). These women need to be identified early and given the same obstetric monitoring as women with a preexisting diagnosis of type 2 diabetes. Given the importance of identifying this high-risk group first diagnosed in pregnancy, the World Health Organization, American Diabetes Association, and others have made a recommendation to categorize these women who have glucose levels similar to those diagnostic of diabetes outside pregnancy with a distinct category separate from gestational diabetes (20–24). They are diagnosed with “diabetes in pregnancy” or “overt diabetes in pregnancy,” prompting caregivers to assess for chronic diabetes complications and observe closely postpartum for hyperglycemia.

It is increasingly recognized that in utero exposure to diabetes increases the risk of obesity and type 2 diabetes in offspring. However, children of women with type 2 diabetes, especially but not exclusively those from Indigenous populations, appear to be at the highest risk. In a large cohort study of 467,950 offspring in Manitoba, Canada, followed for up to 30 years, they found that diabetes exposure was independently associated with incident type 2 diabetes in the offspring after controlling for sex, maternal age, socioeconomic status, birth weight, and gestational age (25). There was a gradient of risk in both First Nations and non-First Nations individuals, with the highest risk in those exposed to mothers with type 2 diabetes, followed by those exposed to gestational diabetes, and lowest in those whose mothers did not have diabetes in pregnancy. First Nations offspring had the highest risk of type 2 diabetes. The authors suggest that although diabetes in utero clearly plays a role in the development of diabetes in the children, inequities in income, education, food security, and prenatal care continue to promote these adverse outcomes through the generations.

These epidemiological studies helped define the extent and the urgency of this global problem. As clinicians, we were faced with the challenge of how to positively affect these outcomes.

One avenue was the use of technology to improve maternal and fetal outcomes, including better blood glucose monitoring, faster-acting insulins, and improved obstetric and neonatal care.

The use of continuous glucose monitoring (CGM) outside pregnancy has shown improvements in glucose time in range (TIR), reductions in hypoglycemia, and improved quality of life. Patients with type 1 diabetes in pregnancy stood out as a population that may particularly benefit from this new technology.

Prior to the Continuous Glucose Monitoring in Women With Type 1 Diabetes in Pregnancy Trial (CONCEPTT), only one large study had been done looking at the use of CGM in pregnancy, and they did not show improvements in HbA_1c or outcomes, likely due to only intermittent use of CGM (26). In CONCEPTT, along with my co-principal investigator, Helen Murphy, we sought to determine whether the use of continuous real-time (RT)-CGM would improve glycemic levels and, secondarily, if it would improve pregnancy outcomes (27). In this study, 215 women who were pregnant, and 110 women planning pregnancy, were randomized in 2 parallel trials to receive RT-CGM to be worn continuously, along with self-blood glucose monitoring, or standard care (self-blood glucose monitoring alone) (27). We found a significant difference in HbA_1c at 34 weeks of gestation between the groups (difference of 0.19%) and a significant difference in time spent in the pregnancy target range (63–140 mg/dL or 3.5–7.8 mmol/L). CGM users spent 7% more TIR (68% vs. 61%) and less time above range (TAR), while time spent being hypoglycemic was low and comparable (3% vs. 4%). In addition to improved glycemic levels, we found improved neonatal outcomes. Women using CGM had significantly fewer LGA infants, fewer neonatal intensive care unit (NICU) admissions over 24 h, and fewer cases of neonatal hypoglycemia. Infants in the intervention group also spent 1 day less time in hospital.

An observational cohort study in Sweden helped to corroborate these findings. In a real-world observational cohort study of 186 pregnant women with type 1 diabetes, investigators were able to obtain CGM readings from 92 women using RT-CGM and 94 women using intermittently scanned CGM during pregnancy (28). They found that, similar to CONCEPTT, a 5–7% TIR was associated with a reduction in LGA and a composite of adverse outcomes.

Several secondary analyses have been done using the rich CONCEPTT data, and these have contributed to our understanding of type 1 diabetes in pregnancy.

In a recent study by Scott et al. (29), CGM data from the CONCEPTT and Swedish studies were combined in an effort to determine glucose metrics across pregnancy and their relationship to LGA infants. They found that glucose levels diverged early, by 10 weeks of gestation, between those mothers with and those without an LGA infant. Mean CGM glucose concentration was lower and time in the pregnancy target range was higher from 10 weeks of gestation and onwards in women with normal versus LGA infants, mostly due to mealtime highs. This study suggests that an emphasis on achieving the pregnancy glucose targets from very early in pregnancy, starting at 10–12 weeks of gestation, and an emphasis on correcting postprandial hyperglycemia is needed to achieve desired neonatal outcomes.

The best diet for diabetes in pregnancy remains controversial. High levels of carbohydrates are thought to contribute to higher glucose levels, while high-fat diets may contribute to fetal adiposity. In a substudy of CONCEPTT by Neoh et al. (30), the dietary intakes of women with type 1 diabetes were described. They found that while total carbohydrate intake was consistent with guidelines in pregnancy, 50% of the carbohydrates were from nonrecommended sources such as sugar, confectionery and preserves, biscuits, and cake. Diets were also found to be higher in fat than recommended, low in fiber, and with inadequate consumption of fruits and vegetables. Efforts are needed to improve the diet content in women with type 1 diabetes. Further research is needed to help determine the ideal maternal diet to optimize glucose levels and pregnancy outcomes in women with diabetes.

Given the development of several new glycemic markers, Meek et al. (31), in a secondary analysis of CONCEPTT, examined whether CGM measures could be used to predict pregnancy outcomes and how they compared with predictions using some of the newer glycemic markers along with HbA_1c. Other glycemic markers examined included glycated CD59, 1,5-anhydroglucitol, fructosamine, and glycated albumin. Using blood samples from 157 CONCEPTT participants, we found that the CGM measures TIR/TAR and HbA_1c did as well as or better than the other markers in predicting pregnancy outcomes such as preterm birth and LGA. CD59 did well in predicting NICU admission and neonatal hypoglycemia in the second trimester, while HbA_1c predicted these outcomes best in the third trimester.

We know that glucose levels are an important determinant of infant weight but not likely the only determinant. Placental growth factor (PlGF) and soluble fms-like tyrosine kinase (sFlt-1) are angiogenic factors that are currently being used as biomarkers to assess the risk of preeclampsia (32). PlGF is a proangiogenic factor secreted by the placenta that enhances the effects of vascular growth factors. Soluble fms-tyrosine kinase reduces levels of PlGF and vascular endothelial growth factor, thus acting as an antiangiogenic factor (32). PlGF and sFlt-1 have also been used as surrogate markers to represent placental health. We wanted to determine if there was an association between placental function, as measured by angiogenic factors, and offspring birth weight z scores in women with type 1 diabetes. In the secondary analysis of CONCEPTT by Bacon et al. (33), we found that the relationship between PlGF levels and birth weight z scores differed according to maternal glucose levels. In women with a lower HbA_1c or higher time in the target range, birth weight z scores were reduced to normal with increasing PlGF levels (i.e., a healthier placenta). However, in those with higher HbA_1c or shorter time in the target range, infants were heavier with increasing PlGF (healthier placentas) but lighter with reduced PlGF (unhealthy placentas). Therefore, in the milieu of higher glycemic levels, an infant weight of “normal” may actually reflect inappropriate growth where suboptimal placental function should be considered and looked for.

An economic analysis using CONCEPTT data has shown that use of CGM during pregnancy is cost-effective (34,35). In their U.K. study, Murphy et al. (34) developed a model to estimate National Health Service (NHS) England costs associated with use of RT-CGM during pregnancy using CONCEPTT data. Using government costs for the devices, it was found that if the NHS paid for CGM for all women with type 1 diabetes during pregnancy, the NHS would save 9.6 million pounds per year, representing a 40% reduction in cost compared with the cost of self-blood glucose monitoring alone. This was primarily driven by a shorter NICU stay for infants of mothers using CGM. An economic analysis by Amed et al. (35), applied to CONCEPTT data using health care costs from 3 Canadian provinces, showed similar trends. Given these data from CONCEPTT, several countries are now paying for CGM use during pregnancy in women with type 1, including the U.K., Denmark, Sweden, and Australia.

Guideline recommendations regarding the use of CGM in pregnancy have also been changed in response to CONCEPTT findings. As of December 2020, the National Institute of Health and Care Excellence updated their guidelines to offer CGM to “all pregnant women with type 1 diabetes to help them meet their pregnancy blood glucose targets and improve neonatal outcomes” (36). In Canada, recommendations were updated to state that during pregnancy all women with type 1 diabetes “should” use RT-CGM to improve glycemic control and outcomes (37). The International Consensus on Time in Range (38) used the CONCEPTT results, and those of other analyses, to make recommendations regarding the time that should be spent in the pregnancy glucose target range of 3.5–7.8 mmol/L. Given that CONCEPTT women using CGM achieved an average TIR of 68% with improved outcomes, the consensus committee suggested that women should aim for >70% of their time in the pregnancy target range throughout pregnancy. Do women achieve this goal? A study by Tundidor et al. (39) examined how many women achieved >70% TIR in CONCEPTT and found that overall, only 7.7% achieved this in the first trimester, 10.2% in the second trimester, and 35.5% in the third trimester. In CONCEPTT, more CGM users achieved the TIR goal than those using only capillary glucose monitoring, but still only 44% achieved this goal by the third trimester. It is important to note, however, that during CONCEPTT, CGM TIR targets were not routinely used by patients and clinicians. In a recent study looking at CGM measures during pregnancy, only 28% of participants achieved a cumulative >70% TIR throughout pregnancy (40). We need to find ways to support patients in achieving their target goals, and we must do so sooner in pregnancy. If CGM cannot get women to their glycemic target goals, can pumps improve the time spent in range?

Pumps have been associated with improved HbA_1c, reduced hypoglycemia, and improved quality of life. However, older randomized trials failed to show a benefit when pumps were used during pregnancy, and more recently, meta-analyses of pump use versus multiple daily injections (MDI) showed that pump use was associated with increased weight gain and larger infants (41). We examined differences in glycemic levels and outcomes in women using pumps versus MDI in the CONCEPTT participants in Feig et al. (42) and found that pump users spent 5% less TIR in the second trimester and 6% more TAR. In a study by Scott et al. (43), functional data analysis was done to look at these differences further. It was found that in the second trimester pump users were spending 12 h per day with higher glucose levels. This difference in glycemia in the second trimester may explain the finding of more gestational hypertension in the pump users and more NICU admissions and neonatal hypoglycemia in their infants.

Why would there be such a difference in glycemic levels between pump and MDI users, particularly in the second trimester? We wondered if pump users were more flexible with their diet, leading to worse glycemia. In a study by Neoh et al., the diets in pump versus MDI users in CONCEPTT were examined (44). In this study, there was no difference in total energy intake or in carbohydrate, protein, fat, or fiber intake. The average daily carbohydrate intake was 190–200 g per day, and there was no difference in the quality of carbohydrates or in their distribution throughout the day between pump and MDI users. We hypothesize that women and their caregivers find it harder to keep up with increases to the bolus doses during the second trimester, when one sees marked increases in insulin resistance. Mathiesen et al. showed that the bolus doses increase by 400% during pregnancy (45). The carbohydrate-to-insulin ratios may be more confusing for pump users to change compared with increasing premeal doses in those using MDI, and there may be a fear of hypoglycemia with significant changes to the carbohydrate-to-insulin ratios. We have now focused on the education of patients and caregivers to help them understand the need for changing the bolus doses aggressively in the second trimester and show them how to confidently and efficiently make those changes.

If CGM and pump therapy will not get us to our glycemic goals, what about “artificial pancreas systems” or “hybrid closed-loop systems”? The hybrid closed-loop artificial pancreas system (APS) measures glucose by CGM, which is then plugged into an algorithm. Based on predicted glucose levels in the near future, the algorithm changes insulin levels given by insulin pumps. The system is hybrid because it still depends on the operator to assess and advise the pump of the carbohydrates that will be ingested and the bolus for those carbohydrates. A meta-analysis of 25 randomized trials of APS in the outpatient setting in the nonpregnant population showed that APS resulted in a significant improvement in the time spent in the target range of 3.9–10.0 mmol/L and led to a significant reduction in time spent being hypoglycemic (<3.9 mmol/L) (46). These findings give hope to the pregnancy population; however, there are several challenges that need to be addressed in pregnancy. These include the increase in hypoglycemia in early pregnancy, the marked rise in insulin resistance in mid-pregnancy, the delayed time to action of insulin, which worsens as pregnancy progresses, and the very tight range of glycemic levels needed to prevent adverse pregnancy outcomes.

To date, the only randomized trial published on pregnant women using APS day and night was that from Stewart et al. (47), who used the Cambridge APS system. In this small randomized cross-over study of closed-loop versus sensor-augmented pump use in 16 patients, they reported no difference in time spent in the pregnancy target range of 3.5–7.8 mmol/L overall; however, there was a significant reduction in the time spent hypoglycemic (<3.5 mmol/L) and a trend toward more time spent in the target range overnight (P = 0.06).

Currently there are several ongoing randomized trials of APS technology during pregnancy, including the Automated Insulin Delivery Amongst Pregnant Women With Type 1 Diabetes (AiDAPT) trial, using the Cambridge APS FX system (NCT04938557), the PICLS pilot trial, using the Medtronic 670/770G (NCT03774186), the Closed-loop Insulin Delivery In Type 1 Diabetes Pregnancies (CIRCUIT) trial, using the Tandem Control-IQ (NCT04902378), the Closed-Loop Insulin Delivery in Pregnant Women With Type 1 Diabetes (CRISTAL) trial, using the Medtronic 780G (NCT04520971), and other systems currently being developed, such as the Longitudinal Observation of Insulin Requirements and Sensor Use in Pregnancy (LOIS-P) trial (NCT03761615). We look forward to the results of these trials to determine whether these APS can improve achievement of target glucose measures and thus improve pregnancy outcomes.

The management of type 2 diabetes during pregnancy would ideally include prepregnancy counseling, early diagnosis, prevention of metabolic syndrome, and optimization of glycemic levels. While there are many oral and injectable antihyperglycemic agents available for the care of people with type 2 diabetes outside pregnancy, most are not approved for use during pregnancy. Metformin, through its antihyperglycemic properties as well as other metabolic effects, offered a particularly intriguing opportunity in this high-risk group.

Metformin is a biguanide that acts through multiple mechanisms. It increases the AMP-kinase pathway, leading to increased glucose uptake in muscles and decreased gluconeogenesis in the liver. It also acts by decreasing production of cyclic AMP and suppressing mitochondrial complex 1 and the mTORC1 signaling pathway (48). In addition, metformin lowers nutrient intake and body weight by increasing levels of growth/differentiation factor 15 (GDF-15) (49). Metformin is currently recommended as first-line therapy for type 2 diabetes outside pregnancy and has been used in the treatment of women with gestational diabetes during pregnancy (50). Meta-analyses of randomized trials comparing metformin to insulin in women with gestational diabetes have shown numerous benefits, including maternal benefits such as reduced maternal weight gain and decreased hypertensive disorders as well as neonatal benefits such as fewer LGA infants and less neonatal hypoglycemia (50). Alternatively, metformin has been associated with a lower gestational age (by approximately 1 day) and, in some meta-analyses, increased preterm birth, although the etiology is not known. In addition, as described later in this review, there remains concern over the uncertainty over possible (currently unknown) long-term effects on the offspring.

Prior to the Metformin in Women With Type 2 Diabetes in Pregnancy Trial (MiTy), there was a paucity of evidence for the use of metformin in women with type 2 diabetes during pregnancy. There were two randomized trials of metformin use during pregnancy that looked at neonatal outcomes in this population, but these were limited by small sample size and methodological issues.

In MiTy we aimed to determine whether the addition of metformin to insulin, in women with type 2 diabetes in pregnancy, would decrease or increase the rate of a composite of serious neonatal outcomes, including pregnancy loss, preterm birth, birth injury, respiratory distress syndrome, neonatal hypoglycemia, and NICU admission >24 h (51). In this international, randomized, parallel, double-masked, placebo-controlled trial, 502 women with type 2 diabetes were randomized between 6 weeks and 22 weeks, 6 days of gestation to receive either 1 g metformin twice daily or placebo, added to their usual insulin regimen. Recruitment took place in 25 centers across Canada and 4 centers in Australia. We found no difference in the composite neonatal outcome between the metformin and placebo groups; however, several other significant differences emerged, suggesting that metformin was having a profound biological effect on maternal/fetal outcomes. The mothers in the metformin group gained significantly less weight, had lower HbA_1c at 34 weeks of gestation, had fewer cesarean sections, and required significantly less insulin during pregnancy (on average 45 units less insulin per day). Infants in the metformin group weighed significantly less (approximately 200 g) and had significantly lower birth weight z scores. Significantly fewer were extremely LGA and fewer were macrosomic (>4 kg), and they showed reduced adiposity measures. The birth weight z score distribution conformed to a normal distribution, with metformin-exposed infants’ normal curve shifted to the left (i.e., they weighed less). Accordingly, we found that more infants in the metformin group were SGA.

How should we interpret these findings? Caregivers and patients should be aware of the many maternal and neonatal benefits but also consider the increased risk of SGA and weigh the risks and benefits. For instance, one may consider not giving metformin to women with a history of small babies or to twins, for whom small babies are expected, or stopping metformin treatment in those developing SGA.

The reason for the increased SGA in the metformin group is likely multifactorial. It may be a direct effect of metformin. Metformin affects the mechanistic target of rapamycin (mTOR) pathway, which could influence nutrient flow across the placenta (48). mTOR is activated by growth receptors such as insulin growth factor, and its downstream targets are involved in protein synthesis via activation of amino acid transporters (52,53). In humans, mTOR activity and its downstream targets are decreased in fetal growth restriction and increased in fetal overgrowth (54). It is hypothesized that disruption of mTOR signaling disrupts placental growth pathways (55). It also may be an indirect effect via the reduced maternal weight gain, improved glycemic levels, and reduced insulin doses. Given this increase in SGA with metformin use, we looked to describe the characteristics of the metformin versus placebo SGA infants and determine the predictors of SGA in women with type 2 diabetes, which would allow us to individualize their care. In a recent secondary analysis of MiTy, we found that the presence of chronic hypertension and/or nephropathy and the use of metformin were independent predictors of SGA (56). The absolute risk of SGA in women with chronic hypertension and/or nephropathy who used metformin was quite high (25%). We suggested, therefore, that until further data are available, and with the intent of limiting SGA, it is reasonable to be cautious in the use of metformin in those specific populations.

More data are on the way. The Medical Optimization of Management of Type 2 Diabetes Complicating Pregnancy (MOMPOD) trial (NCT02932475), a trial of women with type 2 diabetes in the U.S., where metformin or placebo is added to insulin, has finished recruitment, and results should be available by Spring 2023. An individual patient meta-analysis is planned, combining the MiTy and MOMPOD trial data, to give more power to answer these important questions regarding the use of metformin in women with type 2 diabetes.

One of the main concerns with the use of metformin has been the suggestion in some studies that offspring of mothers with polycystic ovary disease (Metformin Treatment of Pregnant Women With Polycystic Ovary Syndrome [Pregmet] trial) (57) or gestational diabetes (Metformin in Gestational Diabetes Trial [MiG] trial) (58) exposed to metformin in utero are at increased risk of overweight/obesity at 4–9 years of age and possibly dysglycemia (59). Researchers hypothesize that metformin is acting to reduce placental nutrient flux by inhibiting the mTOR pathway, leading to fetal programming of obesity and metabolic syndrome later in life, similar to what is hypothesized to occur in infants exposed to undernutrition during pregnancy (60). This effect on offspring exposed to metformin has not been shown universally, however, and outcomes may be different depending on the metabolic environment.

Given the findings described above, the long-term outcomes of children of mothers with type 2 diabetes exposed to metformin is of great interest. The Metformin in Women With Type 2 Diabetes in Pregnancy Kids Trial (MiTy Kids) study has followed the offspring of mothers in MiTy up to 2 years of age, and the findings should be available soon. The Effect of In Utero Exposure to Metformin in 5–11-Year-Old Offspring of Mothers in the MiTy Trial (MiTy Tykes), recently funded by the Canadian Institute of Health Research, aims to follow these children up to 5–11 years of age and compare measures of adiposity, metabolic syndrome, and neurobehavior between the groups.

Metformin has improved maternal and fetal outcomes in women with type 2 diabetes, but more needs to be done. The number of women with type 2 diabetes is increasing, and the prevalence of adverse pregnancy outcomes remain high. It is time to study technological advances in women with type 2 diabetes in pregnancy. The International Consensus on Time in Range committee recommends that patients with type 1 diabetes aim for >70% time spent in the pregnancy target range (38), but the appropriate time spent in the target range in women with type 2 diabetes has not yet been determined. Our prospective cohort study, TIMELY (Time Spent in the Target Glucose Range and Maternal and Neonatal Effects in Women with Type 2 Diabetes in Pregnancy), will help determine the TIR and TAR that are needed to achieve the best outcomes in women with type 2 diabetes. Furthermore, randomized trials are needed to determine the efficacy of CGM or intermittently scanned CGM in women with type 2 diabetes.

Women with type 2 diabetes require very high insulin doses during pregnancy, often contributing to marked maternal weight gain. Better guidelines for medical nutrition therapy may help in achieving excellent glycemic levels without excess weight gain and insulin doses. Women with type 2 diabetes are less likely to receive prepregnancy care. More efforts are needed to address the socioeconomic challenges faced by women with type 2 diabetes and the barriers that prevent women from coming for treatment prepregnancy. For those women not planning pregnancy, more effective use of contraception is needed. We know that women with gestational diabetes are at increased risk of progression to type 2 diabetes. Further preventive strategies that address the growing rates of obesity and unhealthy lifestyle are needed to stop this progression. Finally, diabetes begets diabetes, especially in women with type 2 diabetes. Studies have shown that children of women with type 2 diabetes have the highest rates of diabetes themselves and that the in utero environment is independently associated with this development of diabetes (61). Only long-term studies can determine if we can stop this vicious cycle of diabetes begetting diabetes by achieving excellent glycemic levels during pregnancy and/or the reduction of the inflammatory environment in pregnancy by other means (i.e., metformin).

We have learned much from large-scale epidemiological studies that have helped us to define the nature and the extent of the growing clinical challenges faced in the management of preexisting diabetes in pregnancy, whether type 1 or type 2. Our large multicenter randomized trials have further defined the various promising avenues for treatment and helped direct investigators toward future research as we attempt to refine these therapies. We have come a long way in helping women with preexisting diabetes during pregnancy achieve better pregnancy outcomes. Addressing health care inequalities in all women with diabetes in pregnancy will be crucial to further improve health outcomes.

Acknowledgments. I thank my mentors, David Naylor (University of Toronto), Bernie Zinman (Department of Medicine, University of Toronto), and Gillian Booth (Department of Medicine, University of Toronto), for their great guidance. I thank my co-principal investigator on CONCEPTT, Helen Murphy (Division of Maternal Health, St. Thomas’ Hospital, King’s College London), my coinvestigators Kellie Murphy (Department of Obstetrics and Gynecology, University of Toronto) and George Tomlinson (Department of Medicine, University of Toronto), and all the collaborators on CONCEPTT and MiTy for their dedication and inspiration. I thank all those at the coordinating center Clinical Trial Services/Centre for Mother, Infant, and Child Research and ICES at Sunnybrook Health Sciences Centre, without whom this work would not be possible. Lastly, I thank my wonderful family for their unwavering support.

Funding. CONCEPTT was funded by JDRF grant 17-2011-533 and grants under the JDRF Canadian Clinical Trial Network, a public–private partnership including JDRF and the Federal Economic Development Agency for Southern Ontario and supported by JDRF no. 80-2010-585. MiTy was funded by the Canadian Institutes of Health Research, the Lunenfeld-Tanenbaum Research Institute, and the University of Toronto Department of Medicine.

Duality of Interest. D.S.F. reports nonfinancial support from Apotex for drug supplies for the MiTy trial and personal fees from Novo Nordisk for advisory board activities. Medtronic supplied the CGM sensors and CGM systems at a reduced cost.