Navigating Automated Insulin Delivery for Type 1 Diabetes Management During Pregnancy

Abstract

Achieving pregnancy-specific glucose targets is difficult in pregnant individuals with type 1 diabetes (T1D), and the rates of complications for mothers and their infants remain high. Currently marketed automated insulin delivery (AID) systems are hybrid closed-loop (HCL) systems in which basal insulin delivery (with or without automated correction boluses) is driven by algorithms, and users are required to initiate meal boluses. For non-pregnant people with T1D, HCL therapy has established benefits for glycemic outcomes and quality of life. While none of the currently available HCL systems were designed for pregnancy-specific glucose targets and outcomes, preliminary data suggest that the use of HCL systems may result in improved glycemia during pregnancy. There is an accumulating body of literature examining HCL systems in pregnancy, although there are still limited data regarding the impact of HCL systems on perinatal outcomes. Many individuals conceive while using clinically available HCL systems and may be hesitant to discontinue use during pregnancy, and clinicians may consider HCL therapy for pregnant individuals who are struggling to meet recommended glycemic levels during pregnancy. We therefore offer guidance on how to counsel patients on the risks and benefits of HCL therapy in pregnancy, how to identify appropriate candidates for HCL therapy in pregnancy, and how to manage commercially available HCL systems off-label throughout gestation.

Introduction

The identification of insulin 100 years ago and the ongoing refinement of its delivery in humans ever since have allowed individuals with T1D to carry pregnancies to term. Achieving and maintaining euglycemia before and during pregnancy reduce risks for spontaneous abortion, congenital anomalies, large for gestational age birth weight, cesarean delivery, neonatal hypoglycemia, and Neonatal Intensive Care Unit (NICU) admission. In addition, achieving adequate glucose levels can reduce the risk for hypertensive disorders of pregnancy, cardiac, and renal complications. ¹ As a result, current guidelines set relatively strict glycemic goals for pregnant individuals with a goal of reducing hyperglycemia without significant hypoglycemia.

Recommended glycemic targets in pregnancy include an HbA_1c <6 to 6.5%, fasting glucose value <95 mg/dL (5.3 mmol/L), 1-hour postprandial value <140 mg/dL (7.8 mmol/L), and 2-hour postprandial value <120 mg/dL (6.7 mmol/L).^1,2 With the increasing use of continuous glucose monitoring (CGM), the International Consensus on Time in Range recommended pregnancy-specific goals for time spent in pregnancy-specific target glucose range (psTIR), above range (psTAR), and below range (psTBR) (Table 1). ³ There is an accumulating body of evidence demonstrating that increased psTIR is associated with decreased risk for large for gestational age (LGA) birth weight and neonatal complications.^4-6 Other CGM-derived glycemic variability markers have also been linked to adverse perinatal outcomes.^7-9 The CONCEPTT study demonstrated how difficult it is for pregnant individuals with T1D to achieve recommended targets in pregnancy, with only 61% to 68% of participants achieving the recommended time in range (TIR) at 34 weeks’ gestation. ⁵ The initial studies demonstrating the benefit of CGM use in pregnancy found that mean glucose was higher than the suggested targets (trimester-specific values ranging from 121 to 137 mg/dL [6.7-7.6 mmol/L]^4,6 and 128-140 mg/dL [7.1 to 7.8 mmol/L]),^5,6 with psTIR significantly less than the recommended goal of >70%.^4-6 These data demonstrate that even under optimal circumstances, it is difficult to achieve these glucose levels using conventional therapies such as multiple daily injections or sensor-augmented pump therapy (SAPT). In addition, CGM data suggest that it may be necessary to achieve a mean glucose <126 mg/dL and psTIR >55% to 60% by 10 weeks’ gestation to achieve a normal birth weight. ¹⁰ Real-life data indicate that only 15% to 16% of pregnant individuals with T1D achieve the target HbA_1c level of 6.5% during pregnancy. ¹¹

Table 1.

Continuous Glucose Monitoring Metrics for Pregnant Individuals With Type 1 Diabetes.

CGM metric	Pregnancy target
psTIR: % of reading at time 63-140 mg/dL (3.5-7.8 mmol/L)	>70% (>16 h, 48 min)
psTAR: % of readings and time >140 mg/dL (>7.8 mmol/L)	<25 % (<6 h)
ps Level 1 TBR: % of readings and time <63 mg/dL (<3.5 mmol/L)	<4% (<1 h)
Level 2 TBR: % of readings and time <54 mg/dL (<3.0 mmol/L) a	<1% (<15 min)
Glycemic variability (%CV) target a	≤36%

PS, pregnancy specific; CGM, continuous glucose monitoring; TIF, time in range; TAR, time above range; TBR, time below range; CV, coefficient of variation.

^aMetrics are not pregnancy specific.

Automated insulin delivery systems

There are four hybrid closed-loop (HCL) systems currently available in the United States including the Medtronic MiniMed 780G, Tandem Control-IQ, Omnipod 5, and the iLet Bionic Pancreas that are the focus of this review.^12-15 Table 2 provides an overview of each of these systems including an overview of the glucose algorithms and factors associated with each system that can be modified by the user.^16-18

Table 2.

Commercially Available Hybrid Closed-Loop Systems in the United States.

Automation mode	iLet Bionic Pancreas	Medtronic 780G a	t:slim X2 & Mobi	Omnipod 5
		SmartGuard	Control-IQ	Automated mode
Parameters required to start the system	Body weight	Body weight and total daily insulin	Body weight and total daily insulin	Body weight and total daily insulin
Glucose prediction	Basal insulin is initialized using patient’s weight. Basal insulin adjusts every 5 min based on CGM glucose trends and adapts over time based on analysis of daily glucose patterns	“Auto Basal” is calculated from total daily insulin and adjusted every 5 min based on CGM glucose trends aiming for the target glucose value	Increases or decreases the programmed basal rates every 5 min based on a 30 min prediction of CGM glucose, aiming for the target glucose range	“Adaptive Basal” calculated from total daily insulin, which is adjusted every 5 min based on a 60 min prediction of CGM glucose, aiming for the target glucose value
Auto correction	Yes	Auto correction boluses every 5 min	Auto correction boluses every 1 h	No
Target glucose options	“Usual,” “Lower,” “Higher”	100, 110, 120 mg/dL	Target range: 112.5-160 mg/dL 112.5-120 mg/dL (sleep mode)	110, 120, 130, 140, 150 mg/dL
Basal rate adjustment?	N/A a	No	Yes	No
Carb ratio adjustment?	N/A a	Yes	Yes	Yes
Correction factor adjustment?	N/A a	N/A	Yes	Yes
Active insulin time adjustment?	N/A b	Yes (lower AID is preferred)	No, fixed at 5 h	Yes
Correction bolus target adjustment?	Yes, same as algorithm target	No, fixed at 120 mg/dL	No, fixed at 110 mg/dL	Yes, same as algorithm target
Are extended bolus available?	N/A a	No	Yes (extend up to 2 h)	No
Special features	None	Temp target c	Exercise activity d Sleep activity e	Activity features f
Current compatible CGM devices	Dexcom G6 & G7, Freestyle Libre 3 Plus	Guardian 4	Dexcom G6 & G7, Freestyle Libre 2 Plus	Dexcom G6 & G7, Freestyle Libre 2 Plus

^aThere are still some users of the Medtronic 670G/770G, but we have not included these devices in our review since they are being phased out of the market.

^bThere are no pump settings programmed into the iLet, as all insulin delivery is automated by the algorithm without the use of any programmed pump settings.

^cChanges target glucose to 150 mg/dL to reduce auto-basal delivery for the chosen duration and disables autocorrection.

^dChanges target range to 140-150 mg/dL to reduce basal delivery (setting is manually started by user).

^eNarrows target range to 112.5-120 mg/dL and prevents autocorrection boluses (this can be programmed to a sleep schedule or manual start/stop).

^fChanges target glucose to 150 mg/dL and decreases dose by ~50% to reduce adaptive basal delivery for the chosen duration.

Outside of pregnancy, HCL is recommended for individuals with T1D due to established glycemic benefits.^{13,14,16,18,19} A recent review of clinical trials and real-world studies in non-pregnant populations using HCL therapy found that the TIR improved by 5% to 10% with HCL therapy across studies, and the TIR was 70% to 75% for most systems. ²⁰ In addition, HCL systems have potential psychosocial benefits.^21,22 While it is conceivable that select off-label use of clinically available HCL systems could lead to overall improved glycemia in pregnancy, the trade-off is that it is incompletely understood whether the achievement of overall psTIR and psTAR targets is associated with improved outcomes compared to achieving specific fasting and post-prandial targets. Reduction of hypoglycemia is particularly relevant during early pregnancy, a time in which the risk for hypoglycemia is increased.^2,23

Rationale for off-label use of hybrid closed-loop during pregnancy

There are several points that provide a rationale for off-label use of HCL therapy during pregnancy. First, achieving optimal glucose levels during the preconception period and pregnancy is important but difficult with multiple daily insulin injections. Second, HCL therapy has improved glycemic levels and quality of life in non-pregnant individuals, and more pregnant women are conceiving while using these systems and wish to continue. Finally, expert guidance is needed to help providers and pregnant individuals manage off-label HCL use during pregnancy since data are limited.

Clinical trials evaluating use of hybrid closed-loop not commercially available during pregnancy

The Closed Loop in Pregnancy (CLIP) Studies 1 to 4 were the pioneering studies evaluating HCL therapy in pregnancy using glucose targets adapted to pregnancy-specific glucose targets.^24-27 The CLIP 1-4 studies used an investigational system that is not available in the United States. This system is now modified and available commercially in Europe as the Cambridge Artificial Pancreas system (CamAPS FX) with a CE mark for use in pregnancy. The initial proof-of-concept studies demonstrated the safety of this HCL system in pregnancy.^24,25 The CLIP-3 study was a randomized crossover trial of 16 pregnant individuals with T1D designed to compare overnight closed-loop therapy to SAPT. ²⁶ In this trial, the target glucose was set to 108 mg/dL (6.0 mmol/L). The authors demonstrated a significant improvement in the primary outcome for this study, which was the percent time that overnight glucose levels were within the psTIR (74.7% vs 59.5%, P = .002). In addition, mean overnight glucose levels were lower (119 mg/dL vs. 133 mg/dL [6.6 mmol/L vs. 7.4 mmol/L], P = .009) with HCL therapy compared to SAPT without increased hypoglycemia. The authors then conducted the CLIP-4 Study, which was an open-label, randomized crossover trial of 16 pregnant women with T1D that compared glycemic outcomes for home day and night use of HCL therapy versus SAPT for 28 days separated by a washout period. ²⁷ Target glucose levels were 104.4 to 131.4 mg/dL (5.8-7.3 mmol/L) depending on fasting versus postprandial status. The use of HCL resulted in significant reductions in hypoglycemia, with reduced time spent <63 mg/dL (3.5 mmol/L) and time spent <50 mg/dL (2.8 mmol/L) as well as fewer hypoglycemic events and less overnight hypoglycemia.

Levy et al ²⁸ recently evaluated the feasibility and performance of at-home use of a zone model predictive controlled-based closed-loop insulin delivery system customized for pregnancies complicated by T1D (CLC-P). Ten participants were enrolled, and after study sensor wear collecting run-in data on personal pump therapy and two days of supervised training, participants used CLC-P targeting 80 to 110 mg/dL during the day and 80 to 100 mg/dL overnight. Mean percentage psTIR increased 14.1 percentage points (3.4 h/d) compared with the run-in period (CLC-P 78.6 ±9.2% vs run-in 64.5 ±16.3%, P = .002). During CLC-P use, there was a significant decrease in both time >140 mg/dL (P = .033) and the hypoglycemic ranges of <63 mg/dL and 54 mg/dL (P = .037). Nine participants exceeded consensus goals of above 70% psTIR during CLC-P use. ²⁸ This study suggests that a pregnancy-customized closed-loop system has the potential to improve pregnancy outcomes and safety despite tighter target glucose ranges.

Data involving currently available hybrid closed-loop systems in pregnancy

Although none of the currently available HCL systems are approved for use in pregnancy, there are several published case reports and a case series describing successful use of automated mode with the 670G HCL systems during pregnancy,^29-31 and a more recent report using the 780G system. ³² Five small single-center retrospective observational studies have also examined the impact of an HCL system on glucose levels in pregnant individuals with T1D, with four describing use of the 780G system^33-36 and another describing use of the Control-IQ system. ³⁷ All reports suggest that initiation of automated mode is associated with improved glycemia, and available data indicate that increased time in automated mode is associated with improved glycemic indices.

One observational multicenter retrospective analysis of 13 pregnant women using the 780G system (including a twin gestation) demonstrated that the TIR was 54%, 64%, and 66% across the first, second, and third trimesters, respectively, with an overnight TIR >70% throughout pregnancy. ³⁵ Similar results were also described by Albert et al ³³ who evaluated six pregnant women using the 780G system and reported mean TIR was 67%, 65.5%, and 65.5% in the first, second, and third trimesters of gestation, respectively. Dodesini et al ³⁴ examined glucose levels in eight pregnant women who used the 780G during gestation and identified increased TIR in the second and third trimesters. Munda et al ³⁶ found that six women treated with the same HCL system had higher mean TIR values during the second and third trimester (78.6% and 83.6%, respectively) with mean TBR >4% during the first and second trimesters. Finally, Wang et al reported four cases using the Control-IQ system during pregnancy. All participants achieved a mean TIR value >70%, with three out of the four reaching an HbA_1c ≤ 6.1% (43 mmol/mol) by the third trimester. ³⁷ For all of these studies, it is important to note that rates of large for gestational age, neonatal morbidity, and cesarean delivery were high when reported.

The CRISTAL Study was a double-arm, parallel-group, open-label, randomized controlled trial conducted at hospitals in Belgium and the Netherlands. ³⁸ A total of 95 pregnant individuals with T1D were randomly assigned to HCL therapy (n = 46) via the 780G or standard insulin therapy (n = 49). The mean proportion of time spent in the target range was 66.5% in the HCL therapy group compared with 63.2% in the standard insulin therapy group (adjusted mean difference 1·88 percentage points [95% CI = −0.82 to 4.58], P = .17). While HCL therapy did not improve overall time in target range, it did improve overnight time in the target range (adjusted mean difference 6.58 percentage points [95% CI = 2.31 to 10.85], P = .0026), and time below range both overall (adjusted mean difference −1.34 percentage points [95% CI = −2.19 to −0.49], P = .020) and overnight (adjusted mean difference −1.86 percentage points [95% CI = −2.90 to −0.81], P = .0005). Participants assigned to HCL therapy reported higher treatment satisfaction, and there were no unanticipated safety events with HCL therapy. ³⁹

Considerations for hybrid closed-loop therapy in pregnancy

None of the HCL systems currently available in the United States are customized to achieve psTIR of greater than 70%, a recommended target, which creates challenges in achieving overnight, fasting, and premeal glucose values within recommended pregnancy ranges. In addition, insulin resistance increases progressively across gestation ⁴⁰ but this may not be adequately addressed by the currently available HCL algorithms. ⁴¹ Insulin absorption is also delayed as pregnancy progresses, ⁴² but available algorithms do not take this into account. There is a paucity of literature regarding the use of faster-acting insulin analogs such as Lyumjev and Fiasp during pregnancy.

The use of HCL systems in pregnancy may also create financial challenges due to lack of health insurance coverages, high deductibles, and/or high co-pays. Most private insurance companies cover CGM and insulin pump therapy for individuals with T1D, but Medicaid coverage of CGM varies by state.

The use of HCL may also create quality of life challenges. In CLIP-3 participants reported more sleep disruptions, thinking about diabetes, concern over preventing hyperglycemia, and burdens such as alarm concerns, trust issues, and obsessiveness in checking data. ⁴³ In CLIP-4, pregnant individuals reported worry over technical glitches, CGM inaccuracy, and the burdens of maintaining sensor functioning. ²² Current generations of CGM devices are more advanced than earlier studies, with less potential for technical glitches and improved accuracy. In addition, the current HCL systems may be better able to mitigate hyperglycemia, and for all of these reasons, some of the limitations reported in earlier studies may be less of a concern today. Nevertheless, this highlights the importance of ongoing patient questioning and counseling when utilizing an HCL system in pregnancy.

The decision regarding whether to use HCL during pregnancy must be individualized. Some women may benefit from HCL therapy during pregnancy, but this is not universal. In order to successfully utilize HCL during pregnancy, an individual must be able to engage with diabetes technology reliably and comfortably and maintain a stable and healthy psychocsocial state while using HCL therapy. The HCL therapy may be particularly helpful during pregnancy when an individual is unable to meet glucose targets with alternate methods, those who have high glucose variability, and/or frequent episodes of hypoglycemia, and high stress levels related to diabetes distress or self-care.

Considerations for use of current hybrid closed-loop systems in pregnancy

Considerations across pregnancy and postpartum

Due to the progressive insulin resistance of pregnancy, adjustments to insulin dosing may need to occur weekly and increases up to 20% of current settings may be necessary. Higher weight gain is common in individuals with T1D during pregnancy, ⁴⁴ and therefore, it would be best practice to update weight if weight gain is >5% of baseline with automated insulin delivery (AID) systems such as Tandem Control-IQ and iLet in order to optimize the insulin-dosing algorithm. As a logistical consideration with changing maternal habitus across gestation, placement of the CGM monitor on the back of the arm is preferred. A secondary analysis of the CRISTAL found that intrapartum use of HCL therapy was associated with more psTIR and lower psTAR without increases in psTBR when compared to standard insulin therapy. ⁴⁵ The insulin resistance of pregnancy resolves quickly after delivery, with insulin requirements falling by approximately 50% necessitating rapid adjustment in the postpartum period. Breastfeeding may also be associated with further decreases in insulin requirements. ⁴⁶

Use of the lowest target glucose range available for the system

As shown in Table 2, each system has one or more target glucose or glucose ranges available. For the currently available systems, this would be 100 mg/dL (5.6 mmol/L) for 780G, Control-IQ sleep mode (112.5-120 mg/dL [6.3-6.7 mmol/L]), and 110 mg/dL (6.1 mmol/L) for the Omnipod 5 system. Note that for Control-IQ sleep mode, the glucose range is tighter (112.5-120 mg/dL as opposed to 112.5-160 mg/dL). However, the use of this mode makes automated correction boluses disabled. The use of lower target glucose is associated with better TIR in non-pregnant adults with T1D, and hence, lower target glucose is recommended during pregnancy.

Correct for hyperglycemia and administer extra insulin carefully

The dynamic basal adjustments in HCL help alleviate some hyperglycemia, but this may act slowly for pregnancy. Pregnant individuals using these systems may initiate boluses using their current sensor glucose reading, but this should be approached with caution. Given that HCL systems dynamically adjust basal rates based on predicted changes in glucose, users are advised to avoid entering multiple manual boluses to minimize hypoglycemia. One way that individuals could receive a manual bolus is via input of “phantom carbs,” which is accomplished by entering carbohydrates that will not be consumed. While this strategy has been used, it is unclear whether it improves outcomes. The HCL systems track “insulin on board,” which should reduce the likelihood of insulin stacking, but this could become an issue when individuals with diabetes are bypassing the pump’s recommendations.

Avoid overcorrection for hypoglycemia

The HCL systems will suspend basal insulin before the onset of hypoglcyemia. In addition, the lower limits of the pregnancy-specific range are 63 mg/dL (3.5 mmol/L). As such, pregnant individuals using an HCL system can often correct hypoglycemia using fewer carbohydrates (ie, 5-10 g as opposed to 15 g). Capillary glucose measurement is recommended before correcting hypoglycemia and again before re-treating for persistent hypoglycemia. Sensor glucose demonstrates a lag compared to blood glucose measurements, ⁴⁷ and falsely low glucose readings can occur due to pressure/compression on the sensor, such as during sleep.^48,49

Avoid prolonged basal insulin suspensions

Prolonged basal suspensions can lead to rebound hyperglycemia, so it is important to view pump reports to assess the frequency and duration of basal suspensions. In the setting where prolonged basal rate suspensions are occurring, pump settings can be adjusted to reduce basal rates for systems that allow basal rate adjustment in automated mode, decrease carbohydrate-to-insulin ratios, lower correction factors, or increase active insulin duration if possible.

Considerations regarding mealtime

The HCL therapy reduces postprandial glucose excursions, but in its current state, it will not completely alleviate postprandial hyperglycemia relating to incorrect or inadequate carbohydrate counting. Pregnancy is associated with a delay between maximum postpranidal blood glucose concentration and glucose disposal ranging from 45 minutes in early to 75 minutes in late gestation. ⁴² This results in prolonged postprandial hyperglycemia in late pregnancy. Optimal administration of mealtime insulin bolus is time dependent, with recommended dosing intervals before meals of 15 minutes in early pregnancy and 30 to 40 minutes in late pregnancy. If a carbohydrate bolus is missed or underestimated, caution should be advised when entering late carbohydrates with HCL therapy, as the system is likely already increasing basal insulin delivery based on the CGM-measured glucose. High-fat meals can lead to delayed carbohydrate absorption, ⁵⁰ which may be addressed with the use of an extended bolus (for systems that offer this feature in automated mode) or by administering part of the bolus before the meal and the remainder 1 to 2 hours later. However, there is a paucity of data indicating whether extended bolus features improve glycemia.

Carbohydrate-to-insulin ratios should be adjusted throughout pregnancy. They may need to be decreased earlier in gestation (late first and early second trimesters) when insulin sensitivity increases and then increased as insulin resistance increases in later pregnancy. ⁴⁰

Managing overnight hyperglycemia

For patients with difficult to manage hyperglycemia with AID, the use of manual mode is one approach that could be considered. In this instance, pregnant individuals could either use SAPT or partial closed-loop therapy overnight if available and reinitiate automated mode each morning. However, the algorithms used by systems such as the 780G and Omnipod are based on the total daily insulin dose, which could be affected by this approach. In addition, this is not an option for systems such as the iLet.

Conclusions

The AID is the standard of care for T1D management in non-pregnant adults. The available data also suggests that HCL therapy could be a powerful tool to improve perinatal outcomes in pregnant individuals with T1D. Given high rates of maternal and neonatal morbidity in pregnancies complicated by T1D, even modest benefits may have a significant long-term impact on maternal and child health.

Future studies will help refine the optimal use of HCL therapy in pregnancies, including much-needed randomized clinical trials comparing HCL therapy to the current standard of care and assessing whether pregnancy-specific algorithms improve perinatal outcomes. It is unlikely that the currently available products will be customized for pregnancy in the near future. We therefore offer an overview of the available devices and literature regarding their use in pregnancy as well as expert guidance on how to optimize their use in pregnancy.