Advantages and Challenges of Automated Insulin Delivery Use in Pregnancy: Physiology Considerations

Introduction

Pregnancy is a unique stage of life characterized by continuous maternal metabolic adaptations from conception to delivery. The metabolic adaptations result from critical maternal-fetal-placental crosstalk to specifically curate and support the nutrient and energy requirements necessary for fetal development, growth, and maturation. ¹ Glucose metabolism is particularly important for reproduction; uterine glucose dysmetabolism (i.e. hyperglycemia, hyperinsulinemia, and inflammation) has been associated with decreased implantation, spontaneous abortions, congenital anomalies, and placental insufficiency. ² Understanding the dynamic nutritional and energetic requirements of pregnancy can inform the use of current diabetes technology tools effectively, recognition of their limitations, and development of future diabetes technology to support pregnant people with diabetes management.

Current evidence-based dietary reference intakes (DRIs) for pregnancy have focused on adequate micro and macronutrient intake to support metabolic and fetal energy (brain) needs. ³ Understanding the importance of glucose metabolism for embryo, decidualization, and placental implantation, the current recommended dietary allowance of carbohydrates (recommended daily allowance, RDA 175 g/d or 45%-65% of total energy intake) do not account for the placental glucose consumption which would increase the RDA for carbohydrates to 220 g/d. ⁴ Although the nutritional balances and energy demands change across gestation as fetal requirements evolve, dietary intakes of macro and micronutrients for pregnancy are recommended upon recognition of pregnancy, ideally preconception. This can be a challenge for people with diabetes transitioning to pregnancy, many of whom significantly restrict carbohydrate to achieve better glycemic control.

Automated insulin delivery (AID) systems may help support individuals with diabetes as they adjust to more balanced macronutrient consumption, require higher amounts of insulin, and confidently minimize hypoglycemic events with intensive glycemic management. However, limitations of currently available AID systems also present unique challenges for pregnant individuals using them for glycemia management.

Early Pregnancy Considerations

Prior to implantation, the embryo is completely dependent on maternally secreted growth factors and nutrients such as pyruvate and lactate with a gradual rise in glucose consumption until the embryo enters the uterus with a sharp increase in glucose utilization. ⁵ Glucose entry into the uterine cavity is tightly regulated and much lower than plasma glucose levels, with average glucose concentration in uterine fluid measuring approximately 3.15 mM (57 mg/dL). ⁶ In addition, the insulin secretory response increases in early pregnancy, independent of changes in insulin sensitivity. ⁷ There is a significant reduction in fasting plasma glucose as early as 6 weeks gestation that may also be attributed to increased insulin secretion, luteal-placental shift of progesterone production, higher growth differentiation factor 15 (GDF15) levels, or mild energy restriction from nausea. ⁸ This is also seen commonly among pregnant people with type 1 diabetes (T1D) often requiring a reduction in insulin requirements in early pregnancy. ⁹ Hyperglycemia or glucose dysmetabolism, particularly during this period, can lead to impaired embryo development, decidualization, and implantation.

Pregnancy may be complicated by nausea and hyperemesis gravidarum (HG) in severe cases, which lowers food intake and challenges consistent, reliable consumption of meals. Growth differentiation factor 15 is a transforming growth factor-beta superfamily protein that is highly expressed in the liver, kidney, and placental trophoblasts. Severity of nausea and vomiting of pregnancy has been shown to be associated with the interaction of fetally derived GDF15 and the pregnant person’s sensitivity to this peptide. ¹⁰ Levels of GDF15 are higher in pregnant people with nausea/vomiting of pregnancy and HG than in those without these symptoms. Food intake suppression is mediated through the GDF15 receptor GFRAL (glial cell line-derived neurotrophic factor receptor alpha-like) which heterodimerizes with an RET (rearranged during transfection) co-receptor in the hindbrain. Likely through ligand-induced desensitization, the nausea severity improves generally after the first trimester. In addition, chronic conditions (thalassemia), medications (metformin), and exercise have been shown to elevate GDF15 with a reduction in HG, but more studies are needed to understand the dose-response, metabolic effects, and efficacy of these potential interventions in pregnancy.^11,12 The biologic functions of GDF15 are continuing to be discovered, but it plays a role in glucose, lipid, and iron metabolism, all of which are important for pregnancy and fetal development.

With the increased nutritional demands, physiologic challenges such as nausea and metabolic changes (i.e. increased insulin secretion, increased insulin sensitivity, and lower fasting glucose) in early pregnancy, pregnant patients with diabetes are vulnerable to hypoglycemia in the first trimester. The majority of pregnant people with T1D experience hypoglycemia in early pregnancy with a peak incidence between 10 and 15 weeks of gestation. ¹³ Individuals who have a history of severe hypoglycemia before conception, ≥10 years since diabetes diagnosis, and higher insulin requirements are risk factors for severe hypoglycemia in early pregnancy. ¹⁴ Automated insulin delivery systems currently available in the United States have not been shown to improve pregnancy-specific 24-hr time in range (TIR; 63-140 mg/dL range), mean glucose and time spent in hyperglycemic ranges compared to other insulin management strategies; however, they have been demonstrated to improve nocturnal TIR and hypoglycemia, making them particularly useful for avoidance of hypoglycemia in this stage of pregnancy.^15-17

Although pregnancy-specific continuous glucose monitoring (CGM) and blood glucose targets are lower than nonpregnant populations, the rates of adverse perinatal outcomes are significantly higher than pregnant people without diabetes.^18-20 Even a 5% improvement in percent TIR is associated with a clinically meaningful improvement in neonatal outcomes among pregnant people with T1D.^21-23 Moreover, normative glucose data among pregnant people without diabetes suggest our current glucose targets and CGM metrics are not aggressive enough to achieve a normoglycemic environment. ²⁴ Although AID systems have been very successful in the nonpregnant population to improve TIR (70-180 mg/dl), often to >70%,^25,26 current AID systems have not been designed to address the lower targets or metabolic complexities of pregnancy; see Table 1. One hybrid closed-loop (HCL) therapy system, CamAPS FX, has been extensively studied in pregnancy and has received U.S. Food and Drug Administration (FDA) approval but is not currently available in the United States. ²⁷ Therefore, pregnant individuals who may benefit from continuing AID systems available in the United States in pregnancy will need to be “assisted” with off-label techniques to optimize glycemic control in pregnancy (Table 2). ²⁸

OPEN IN VIEWER

Table 1.

Potential advantages and limitations of automated insulin delivery systems in pregnancy and postpartum.

Advantages	Limitations
Automatic insulin suspension to reduce hypoglycemia during high-risk periods in the first trimester and postpartum	Target glucose thresholds are above pregnancy-specific goals and may suspend insulin delivery in ideal pregnancy glycemic ranges
Automatic basal delivery adapts to changing insulin needs over time in the second and third trimesters	Most systems require frequent manual adjustments to augment prandial insulin dosing, which may increase up to four-fold during gestation
Potential to achieve greater or equal pregnancy specific time in range with lower patient burden	Current rapid-acting insulins used in AID systems are slower than is ideal to meet intensive postprandial glycemic targets
Potential to achieve intensive intrapartum glycemic targets with low patient and clinician burden	Increased risk of diabetic ketoacidosis with pump therapy compared to injections
Potential to reduce postpartum hypoglycemia	To accommodate the rising insulin resistance, insulin cartridges, and pump sites require more frequent changes
Potential to blunt glucose fluctuations with breastfeeding	AID algorithms may not adjust to the rapid changes in insulin sensitivity, rapid reduction in insulin requirements, and vulnerability to hypoglycemia with breastfeeding in the postpartum period

OPEN IN VIEWER

Table 2.

Features of automated insulin delivery systems commercially available in the United States as of June 2025.

Automated mode settings	Medtronic 780G	Tandem control IQ	Insulet omnipod 5	Betα bionics iLet	Twist with tidepool loop
Basal rates a	No	Yes	No	No	Yes
Carbohydrate-to-insulin ratios (ICR) b	Yes	Yes	Yes	No	Yes
Correction factors (CF) c	Yes, but no manual correction boluses in automated mode	Yes	Yes	No	Yes
Active insulin time d	Yes, and affects automated insulin delivery	Fixed at 5 hours in automated mode	Yes, but does not affect automated insulin delivery	No	Yes
Target glucoses or ranges for automation (lowest target in bold)	100 mg/dL 110 mg/dL 120 mg/dL	112.5-160 mg/dL Sleep mode: 112.5-120 mg/dL	110-150 mg/dL (increments of 10 mg/dL)	Lower (110 mg/dl) Usual Higher	87-180 mg/dL
Reservoir capacity (Units)	300	200-300	200	180	300
Desirable features	Randomized control trial demonstrating equivalent pregnancy-specific TIR to standard insulin therapy Adapts insulin delivery based on total daily dose	Ability to program basal rates which drive automated insulin delivery	Tubeless Pharmacy prescription Adapts insulin delivery based on total daily dose	No carbohydrate counting Continuous adaptation	Ability to program basal rates which drive automated insulin delivery Pharmacy prescription
Pregnancy off-label assistive techniques e	Reduce insulin active time to 2 hours Use fake carbohydrates to add manual correction boluses	Escalate programmed basal rates beyond expected insulin delivery	Drive insulin total daily dose up with aggressive bolus settings (ICR, CF, insulin active time)	None yet described, techniques used in other systems may be counterproductive	None yet described
Postpartum recommendations e	Manual mode for 2-6 days prior to going back to automation	Reduce programmed basal rates	Factory reset or manual mode for ~2 pods prior to going back to automation	Use higher target	Reduce programmed basal rates

Yellow—system features that are unable to be able to keep up with the dynamic metabolic adaptations in pregnancy.

a

Even for systems that cannot adjust basal rates, basal rates should be adjusted throughout pregnancy in the event the patient is kicked out of automation.

b

ICR need to be adjusted throughout gestation to keep up with the rising insulin resistance. Patients with AID systems that can deliver ICR can deliver “fake carbohydrates” to deliver more insulin to assist in achieving pregnancy targets. Prebolus timing before meals should increase across gestation.

c

CF should be adjusted throughout pregnancy among AID systems that can adjust the CF to reflect the increasing total daily dose of insulin across gestation.

d

If AID systems can adjust the active insulin time, it should be adjusted to the shortest time period available.

e

Pregnancy assistive-techniques and postpartum recommendations listed are off-label and based on the clinical experience of the authors and other experts in the field. These techniques carry risks, including hypoglycemia, and should be used with caution and only by clinicians and patients experienced in the use of these technologies after weighing risks and benefits.

Late Pregnancy Considerations

Late pregnancy is characterized by a dramatic increase in insulin resistance which requires escalation of insulin dosing beyond what is typical for AID system users outside of pregnancy.^29,30 A longitudinal study in T1D with optimally managed glycemia demonstrates that insulin requirements double between 20 weeks gestation and delivery. ³¹ Most insulin pump system reservoirs hold 300 or fewer units of insulin. While published studies suggest the average insulin requirement in the third trimester among people with T1D is between 65 and 95 units per day,^31-33 it is not rare to require more than 150 units per day, resulting in the need for infusion site changes daily or even more frequently. In patients with high insulin requirements, we sometimes prescribe a basal insulin injection to be used along with the AID system. When insulin requirements exceed 200 units per day, we consider off-label use of concentrated insulins via insulin pump. While U200 insulin has similar kinetics to U100 insulin that allows it to be used off-label for AID, ³⁴ U500 insulin has substantially delayed onset compared to rapid-acting insulins, requiring alternative strategies. ³⁵

The increase in insulin requirement in the latter half of pregnancy is disproportionately due to an increase in prandial requirements. Two longitudinal studies found that average increases in basal insulin requirements in insulin pump users between the first and third trimester were 50% while average increases in prandial insulin requirements were four-fold.^33,36 Accordingly, an average carbohydrate ratio would suggest a 5 unit bolus for a breakfast containing 60 g of carbohydrates in the first trimester and a 20 unit bolus for breakfast containing 60 g of carbohydrates in the third trimester. With the exception of the iLet Bionic Pancreas, most AID systems do not automatically provide and adjust prandial coverage, thus these frequent changes in meal requirements must usually be programmed manually to maintain the intensive postprandial glycemic targets demanded by pregnancy (1-hr postprandial blood glucose <140 mg/dL; Figure 1).

OPEN IN VIEWER

Figure 1.

AID benefits and challenges in pregnancy. The dynamic physiologic and metabolic adaptations across gestation require frequent insulin adjustments that may increase up to four-fold by the third trimester and increase the risk for DKA. Current available AID systems in the United States do not have algorithms with glucose targets low enough to achieve pregnancy-specific goals. Although AID systems can increase self-efficacy and may reduce some of the psychological stressors of diabetes, they cannot replace the importance of nutritional health and positive effects of lifestyle behaviors for both glucose metabolism and pregnancy outcomes. Created in BioRender. Valent, A. (2025) https://BioRender.com/2qcczmf

When striving to achieve intensive postprandial glycemic targets in pregnancy, insulin kinetics play a major role. A study in nonpregnant AID and sensor-augmented pump users found that objectively ascertained delayed prandial insulin boluses are associated with both hyperglycemia and hypoglycemia: reduced TIR (70-180 mg/dl), greater glycemic variability (glucose coefficient of variation), and greater time below 54 mg/dl. ³⁷ Highlighting the importance of this issue in pregnancy, a well-designed pregnancy physiologic study demonstrated that the time from insulin injection to peak insulin action substantially lengthens between the first and third trimester. ³⁸ The reasons for this are unknown but could be related to the higher volume of insulin injected or changes in subcutaneous absorption. As a result, individuals requiring meal-time insulin often need to dose rapid-acting insulin 30 to 45 minutes prior to eating in late pregnancy to achieve postprandial glycemic targets. Administering insulin at the time the meal begins often results in early postprandial hyperglycemia due to delayed insulin onset and late postprandial hypoglycemia once the rapid-acting insulin reaches its peak effect. In nonpregnant individuals, extra-rapid insulins such as faster aspart and lispro-aabc have an earlier onset (by 5-13 min) compared to rapid-acting insulins (aspart and lispro).^39,40 A 2023 trial, in which ~20% of participants used non-AID insulin pumps, demonstrated that faster aspart was associated with similar achievement of glycemic targets as conventional aspart, with a lower risk of postprandial hypoglycemia in the second trimester. ⁴¹ Faster aspart has been used in AID systems; at least one trial in nonpregnant individuals suggests that it provides a small (2%) increased TIR (70-180 mg/dL) than conventional aspart insulin. ⁴² We are unaware of studies reporting on the use of faster aspart or lispro-aabc in AID systems in pregnancy, though given the physiologic challenges of diabetes in late pregnancy, innovations in this area would be welcome.

Automated insulin delivery systems account for the current glucose, predicted glucose, and recent insulin boluses (“insulin on board”) in determining the basal insulin delivery. A large preprandial bolus in the setting of euglycemia, as recommended in late pregnancy, will often result in suspension of insulin delivery. While the predelivered bolus may maintain euglycemia in the early postprandial period, when the effects of this bolus begin to wane users may experience late postprandial hyperglycemia 3 to 4 hr after eating related to prolonged insulin suspension. When this occurs frequently, we typically recommend switching the system into manual mode with an appropriate programmed basal rate after the premeal bolus to mitigate these effects.

During pregnancy, several physiologic factors contribute to an increased risk of diabetic ketoacidosis (DKA). ⁴³ Diabetic ketoacidosis results from absolute or relative insulin deficiency and resulting disruption of metabolic homeostasis. Insulin resistance, which represents a reduction in the ability of tissues to respond to circulating insulin and results in decreased cellular glucose uptake, peaks in the third trimester and contributes to DKA risk. In DKA, lack of glucose availability at the cellular level induces a counterregulatory response which increases lipolysis, production of acidic ketone bodies, and gluconeogenesis, further elevating glucose levels and leading to an osmotic diuresis which causes hypovolemia. The acidosis, osmotic diuresis, and accompanying renal insufficiency often lead to electrolyte disarray which may be life-threatening. Pregnancy is a ketosis-prone state both because of insulin resistance and because of glucose loss through the urine and uptake by the fetal-placental unit, which, together, have the end effect of decreasing availability of glucose at the tissue level. Moreover, DKA can occur at lower glucose levels in pregnancy because of the urinary and placental/fetal glucose “sinks.” Individuals using insulin pumps typically are only using rapid-acting insulin. Thus, if insulin delivery is interrupted, DKA can quickly ensue in the presence of absolute insulin deficiency, as is present in T1D. Insulin pumps, including those with AID features, are associated with an increased risk of DKA as compared to those using subcutaneous insulin regimens. In some individuals at high risk for DKA, we use long-acting subcutaneous insulin such as glargine or degludec in addition to insulin pump therapy to mitigate this risk.

Intrapartum and Postpartum Considerations

The metabolic demands of labor can be compared to those seen in exercise, with increased glucose uptake by the contracting uterus. ⁴⁴ This may result in decreased insulin requirements in laboring individuals on insulin and may result in hypoglycemia if food is restricted and an additional source of carbohydrate is not provided. In addition, glycemic targets during labor are intensive (<110 mg/dl or <126 mg/dl)^20,45 due to the theory that acute hyperglycemia during labor contributes to both fetal hypoxia and neonatal hypoglycemia.^46,47 Traditional use of a continuous intravenous insulin infusion on a hospital labor and delivery unit requires intensive efforts on the part of hospital staff and the potential for medical error is high. Most laboring individuals have limited oral intake, which removes many of the challenges associated with management of prandial insulin described above. Thus, AID systems seem particularly well suited to glycemic management during labor, but there are limited data on the performance of systems intrapartum. An earlier generation closed-loop system (precursor of CamAPS FX) was tested during labor and delivery in an observational follow-up of a randomized trial. ⁴⁸ While there was no formal comparison group, those using closed-loop (N = 27) achieved 82% TIR (63-140 mg/dL) and had a mean glucose of 124 mg/dL. ⁴⁸ The observational (nonrandomized) continuation of the CRISTAL study found higher TIR (63-140 mg/dL, 71.5% vs 63.1%) in those who continued the AID system in the intrapartum setting (N = 27) compared to those who continued standard therapy (N = 45), with similar rates of maternal hypoglycemia, neonatal hypoglycemia, and neonatal intensive care unit (ICU) admission. ⁴⁹

Postpartum, there is a dramatic increase in insulin sensitivity (i.e. decrease in insulin resistance) that occurs upon delivery of the placenta, predisposing postpartum individuals with diabetes to hypoglycemia.^50,51 In T1D, insulin requirements nadir at 24 to 72 hr after delivery and then increase to a more stable level over the first 2 to 4 weeks postpartum. In contrast to the gradual onset of increased insulin requirements across the latter half of pregnancy, the change in insulin requirement in the postpartum setting is abrupt. Thus, AID systems may differ in their ability to adapt to such a sharp decline in insulin requirements (Table 2). Systems that primarily base insulin dosing on basal settings that can be programmed by the user can be adjusted to immediately provide less insulin. Other systems which take into account recent total daily dose of insulin (averaged over 24 hr to 1 week) may result in hypoglycemia if used in AID mode in the immediate postpartum setting. For this reason, both the CLIMB trial, which randomized postpartum individuals to AID therapy (Medtronic 670 g/770 g) or standard care, and the PICLS trial, which randomized pregnant individuals to AID (Medtronic 670 g in automated mode) or a pump without AID features (Medtronic 670 g plus CGM without automation), waited for 3 to 7 days after delivery to initiate or re-initiate AID therapy.^15,52 However, the observational continuation of the CRISTAL trial which used a later generation device from the same manufacturer (Medtronic 780 g), did not require a period of manual mode use immediately postpartum and found similar glycemic outcomes in those who did and did not use AID postpartum. ⁴⁹ The CLIMB randomized trial (N = 18), but not the PICLS or CRISTAL trials, demonstrated a reduction in hypoglycemia in postpartum women assigned to AID.^15,49,52 A prospective extension study (N = 57) of the AiDAPT trial, which randomized pregnant individuals in the first trimester to the CamAPS FX AID system versus standard care, uniquely found much greater TIR (70-180 mg/dL) from 0 to 6 months postpartum in individuals randomized to AID therapy versus those randomized to usual care, without a difference in hypoglycemia. ⁵³ These data are overall reassuring with regard to postpartum AID use, but more data are needed on the performance of AID systems and the required programming changes in the immediate postpartum transition when insulin requirements drop precipitously.

Notably, glucose is a key substrate for milk production; lactose, the sugar in breastmilk is a disaccharide comprised of both glucose and galactose. This results in glucose uptake by the lactating breast and a 15% to 20% lower insulin requirement in breastfeeding women with T1D compared to prepregnancy. ⁵⁴ There are limited data on the performance of AID systems during breastfeeding, but a secondary analysis of the CLIMB trial (N = 17) demonstrated a small decline in glucose levels in the 3 hr after breastfeeding overnight which was blunted with the use of AID (Medtronic 670 g/770 g). ⁵⁵

Conclusions

Physiologic challenges across early pregnancy, late pregnancy, intrapartum, and postpartum merit special considerations for AID therapy. In early pregnancy, AID systems have the potential to mitigate an increased risk of hypoglycemia in part related to GDF 15-mediated nausea/vomiting and appetite suppression. However, most available AID systems do not achieve recommended intensive targets for early pregnancy that optimize metabolic conditions for embryogenesis, decidualization, implantation, and placental development and function. Thus, innovations in AID systems for pregnancy will need to account for the narrow therapeutic window for insulin dosing to enhance early pregnancy metabolic health. In late pregnancy, escalating insulin doses may exceed maximum daily capacity of insulin pump reservoirs and capacity to deliver prandial insulin. While most systems adapt basal rates quickly enough to keep up with the steady augmentation of insulin requirement in late pregnancy, automated features for prandial dosing are lacking in most, requiring users and their clinicians to make frequent adjustments. Slower insulin action onset as well as AID-related insulin suspensions after prandial boluses may limit the ability to achieve intensive pregnancy postprandial glycemic targets. New AID systems tested with faster or more concentrated insulins may overcome some of these late pregnancy challenges. Automated insulin delivery therapy has been successfully employed in diabetes management during intrapartum and postpartum, despite challenges related to fluctuating insulin needs during this period, though data are limited to a few small studies. Given the burden associated with meeting the physiologic challenges associated with diabetes in pregnancy and postpartum with manual insulin delivery, more innovation and investigation on the use of AID to meet the unique physiologic challenges of pregnancy and lactation are greatly needed.