Optimizing Automated Insulin Delivery Systems for Pregnancy

Abstract

Automated insulin delivery (AID) systems have revolutionized modern diabetes care outside of pregnancy, but none of the AID systems currently available in the U.S. are approved for use during pregnancy, none have glucose targets low enough to achieve the stricter fasting glucose targets recommended during pregnancy, and none have algorithms that were designed to respond to the amplified oscillations in glycemia that occur in pregnancy or the progressive changes in insulin resistance observed over the course of gestation. Despite these limitations, many women elect to continue using AID off label during pregnancy based on consideration of individual clinical factors and preferences. This article presents some commonly encountered challenges to off-label AID use and CGM interpretation during pregnancy, along with suggested best-practice workarounds to optimize the care of pregnant individuals with diabetes using AID.

Automated insulin delivery (AID) systems automatically adjust basal insulin delivery based on system-specific control algorithms, in response to preset or customized glucose targets, current glucose values, and trends derived from continuous glucose monitoring (CGM) data (1). Some AID systems additionally deliver automatic correction doses for hyperglycemia to augment the automated basal insulin delivery. In their current form, these devices are known as hybrid closed-loop systems because they still require users to announce meals (1), typically by entering estimated carbohydrate intake into a bolus calculator that suggests a prandial insulin dose based on the programmed insulin-to-carbohydrate ratios (ICRs), with a correction determined by an insulin sensitivity factor (ISF) when hyperglycemia is present. Outside of pregnancy, AID systems have revolutionized modern diabetes management through improvements in time spent in the glucose target range (time in range [TIR]), time spent in the hypoglycemia range (time below range [TBR]), A1C, and psychosocial outcomes (2). Accordingly, AID systems are now recommended as the preferred form of insulin delivery for nonpregnant people with type 1 diabetes and other insulin-deficient forms of diabetes (2).

Although AID systems are widely used outside of pregnancy, none of the AID systems currently available in the U.S. are approved for use during pregnancy, none have glucose targets low enough to achieve the stricter fasting glucose targets recommended during pregnancy (3), and none have algorithms that were designed to respond to the amplified oscillations in glycemia that occur in pregnancy or the progressive changes in insulin resistance observed over the course of gestation (4). Thus, a major limitation to off-label AID use in pregnancy is that the lower limit of programmable glucose targets—100 mg/dL (5.6 mmol/L) for the Medtronic MiniMed 780G system, 112.5–120 mg/dL (6.3–6.7 mmol/L) in Sleep Activity mode for the Tandem t:slim X2 with Control-IQ Technology (Control-IQ) system, and 110 mg/dL (6.1 mmol/L) for the Insulet Omnipod 5 system—is higher than the fasting glucose target range recommended during pregnancy (70–95 mg/dL [3.9–5.3 mmol/L]), as shown in Table 1. For this reason, off-label AID use may lead to fasting hyperglycemia.

Table 1.

ADA-Recommended Blood Glucose Goals in Pregnancies Associated With Type 1 or Type 2 Diabetes (3)

Measurement	Value
Fasting glucose	70–95 mg/dL (3.9–5.3 mmol/L) and either:
1-hour postprandial glucose	110–140 mg/dL (6.1–7.8 mmol/L) or
2-hour postprandial glucose	100–120 mg/dL (5.6–6.7 mmol/L)

Table 2.

CGM Metric Targets for Pregnancy Complicated by Type 1 Diabetes (48)

CGM Metric	Target for Pregnant Individuals With Type 1 Diabetes
TARp: percentage of readings and time >140 mg/dL (>7.8 mmol/L)	<25% (<6 hours)
TIRp: percentage of readings and time 63–140 mg/dL (3.5–7.8 mmol/L)	>70% (>16 hours, 48 minutes)
Level 1 TBRp: percentage of readings and time <63 mg/dL (<3.5 mmol/L)*	<4% (<1 hour)
Level 2 TBRp: percentage of readings and time <54 mg/dL (<3.0 mmol/L)	<1% (<15 minutes)
CV (glycemic variability), %†	≤36%

*Includes percentage of values <54 mg/dL (3.0 mmol/L).

†Recommendation for general population. TARp, pregnancy-specific time above range.

However, for some pregnant individuals, the benefits AID may offer, including improved overall glycemia, increased pregnancy-specific TIR (TIRp), decreased glucose variability, and/or decreased hypoglycemia, may outweigh the risks of fasting hyperglycemia. Current guidance recommends that select use of AID systems without pregnancy-specific glucose targets or a pregnancy-specific algorithm may be considered in pregnant individuals with type 1 diabetes when used with assistive techniques and under the guidance of an experienced health care team (3). Guidance on appropriate counseling of pregnant individuals about potential risks and benefits of off-label AID use and individualized decision-making about whether to use commercially available AID systems in pregnancy is available elsewhere (2,3,5).

Despite the limitations, many women elect to continue using AID off label during pregnancy based on consideration of individual clinical factors and preferences. Often, this decision stems from a reluctance to relinquish the glycemic and quality-of-life benefits derived from AID use pre-pregnancy or difficulty achieving pregnancy glucose targets via conventional sensor-augmented pump therapy. In some individuals, TIRp is lower when using manual versus automated mode of their AID system. In these people, conventional therapy may lead to greater divergence from glycemic targets than the algorithmic AID system limitations described above. It is worth noting that many pregnant individuals using conventional therapies, including CGM, do not reach the recommended TIRp until the third trimester, if at all (6,7). A major obstacle to the achievement of the lower and narrower glycemic targets recommended during pregnancy is the amplified risk of maternal hypoglycemia in the first and early second trimesters, with increasing insulin resistance occurring after 20 weeks’ gestation (8).

To date, clinical trials evaluating the off-label use of the AID systems commercially available in the U.S. have not shown improved TIRp overall (9,10), although one study showed decreased hypoglycemia (9). We await the results of the CIRCUIT trial using Control-IQ technology in pregnancy (11). Preliminary findings appear positive (12). If an individual woman is able to achieve increased overall TIRp with off-label AID use, even if at the expense of fasting hyperglycemia, it is conceivable that off-label AID use could lead to improved pregnancy outcomes in some individuals who are unable to achieve pregnancy glucose targets using conventional therapies.

Further data are needed to better understand the impact of off-label AID use on pregnancy outcomes. Furthermore, high priority must be placed on obtaining access in the U.S. to the CamAPS FX AID system, which has pregnancy-appropriate glucose targets, efficacy in improving TIRp and other glycemic outcomes during pregnancy (13), and approval for use during pregnancy in other parts of the world. Similarly, future randomized trials of an investigational AID system developed by a consortium from the U.S. that is customized for pregnancy (14) should be prioritized.

We present here some commonly encountered challenges to off-label AID use during pregnancy, along with suggested best-practice workarounds for each. We also highlight the need for further advancement—through increased access to technologies that are tailored to and/or approved for use during pregnancy—to optimize and advance the care of pregnant individuals with diabetes.

Basal Insulin Delivery in AID Mode: Pregnancy-Specific Limitations and Optimization

A major limitation at the crux of concerns regarding off-label AID use is the inability to achieve the fasting glucose target range of 70–95 mg/dL (3.9–5.3 mmol/L) that is recommended during pregnancy (3). Of note, current American Diabetes Association (ADA) guidelines advise that, among women without diabetes who have early abnormal glucose metabolism with dysglycemia (defined as a fasting plasma glucose [FPG] of 110–125 mg/dL [6.1–6.9 mmol/L] or an A1C of 5.9–6.4% [41–47 mmol/mol]), treatment may be intensified if FPG levels are mostly >110 mg/dL (>6.1 mmol/L) before 18 weeks of pregnancy because this degree of dysglycemia is thought to identify individuals who are at higher risk of adverse pregnancy and neonatal outcomes (15).

In considering potential risks and benefits of off-label AID use among pregnant individuals with type 1 diabetes, one must consider whether the fasting hyperglycemia that may result from use of commercially available AID systems could increase the risk of adverse pregnancy outcomes. Conversely, off-label AID use could lead to improved pregnancy outcomes if improved TIRp and/or decreased pregnancy-specific TBR (TBRp) outweighed any adverse impact of fasting hyperglycemia. Thus, it is incumbent on health care providers (HCPs) to thoughtfully discuss with each pregnant individual with diabetes the controversies, potential risks and benefits, and uncertainties surrounding off-label AID use in pregnancy.

It is crucial for HCPs to be aware of which AID systems allow adjustment of basal rates in automated mode. Basal rate adjustment is allowed with the Control-IQ system, but adjustment of basal rates is not possible in automated mode with the Omnipod 5 or MiniMed 780G AID systems. Although users are able to adjust basal rates in the Control-IQ system in automated mode, the algorithm decreases basal insulin delivery if glucose is predicted to be <112.5 mg/dL (6.3 mmol/L) in 30 minutes and stops basal insulin delivery if glucose is predicted to be <70 mg/dL (3.9 mmol/L) in 30 minutes, both in the standard automated mode and when using the system’s Sleep Activity feature. It is suggested that pregnant individuals using the Control-IQ system use Sleep Activity throughout the entire day during pregnancy (i.e., set a sleep schedule from 12:00 a.m. to 11:59 p.m. every day) to enable use of the lowest glucose target range possible in this system throughout the entire day and night (16). It is important to inform users of this system that automatic correction boluses will not be delivered while Sleep Activity is activated, so corrections for hyperglycemia must be manually administered via the bolus calculator according to the programmed ISF. Note that automatic correction boluses are delivered up to once per hour at 60% of the dose calculated per the ISF if the future predicted glucose is expected to increase to >180 mg/dL (10 mmol/L) (17), whereas conventional correction doses administered via the bolus calculator enable delivery of the full correction dose based on the current glucose level. Review of pump data on the pump’s software system (18) enables one to compare actual versus programmed basal insulin delivery. A suggested best practice is to program basal rates higher than delivered basal insulin (e.g., increase basal rates by ∼20–25%) after ∼20 weeks’ gestation (16).

For systems that do not use programmed basal rates in automated mode (MiniMed 780G and Omnipod 5), users may adjust the glucose target, but basal insulin modulation is determined solely by the system-specific control algorithm. During pregnancy, it is recommended to use the lowest glucose target available (100 mg/dL [5.6 mmol/L] for the MiniMed 780G and 110 mg/dL [6.1 mmol/L] for the Omnipod 5). Although programmed basal rates are not used in automated mode for these systems, it is important at each visit to evaluate basal rate settings, which are used in manual mode, and compare the programmed manual basal rates to the current average basal insulin being delivered in automated mode. This information is readily obtainable from the respective software systems (19,20). If a person chooses to change to manual mode or if automated mode is unexpectedly stopped (e.g., because of sensor malfunction), a large discrepancy between manual basal rates and current basal insulin delivery could lead to unexpected hypoglycemia or hyperglycemia.

It has been suggested to administer small amounts of “phantom carbohydrates” (i.e., entering carbohydrates into the bolus calculator when none are eaten to enable a supplementary insulin bolus) when using some systems (10,21) as a possible workaround to lower glucose in the setting of the immutable (and higher-than-ideal) glucose targets mandated by commercially available AID systems. However, this practice should be limited to small carbohydrate entries and implemented with caution because of the risks of both hypoglycemia and rebound hyperglycemia, which can follow automated insulin reduction or suspension resulting from a rapid decrease in glucose.

In addition to the algorithm-imposed limitations on increases to basal insulin while in automated mode, an additional limitation of basal insulin modulation for systems that do not enable adjustment of basal insulin in automated mode is the inability to decrease basal insulin in the late postprandial period. Outside of pregnancy, the concept of a “super bolus” has been advocated if large prandial boluses result in late postprandial hypoglycemia several hours after eating. This phenomenon can occur in the setting of the large prandial boluses needed for high-carbohydrate meals (22) and can be exacerbated during pregnancy, when lower postprandial glucose targets necessitate even larger prandial insulin boluses and prandial insulin requirements increase disproportionately compared with basal insulin requirements (23). Using the “super bolus” strategy (24), also known as “basal-to-bolus switch” (25), the basal rate delivered for 2–3 hours after the meal is decreased by an amount that is added to the premeal bolus, allowing administration upfront of a larger prandial bolus while decreasing the risk of hypoglycemia in the later postprandial period. However, for AID systems in which basal rates cannot be adjusted in automated mode, this strategy unfortunately cannot be used. Furthermore, when automated increases in basal insulin delivery are triggered by immediate postprandial hyperglycemia, subsequent late postprandial hypoglycemia can be further exacerbated. Then, premeal hypoglycemia before the subsequent meal can, in turn, impede the pre-bolusing needed to avert postprandial hyperglycemia after the next meal, and the cycle of glycemic fluctuation continues. When using an AID system that does not permit basal rate adjustments, suggested workarounds are to optimize strategies that improve postprandial glycemia, thus buffering increases in the prandial insulin dose and mitigating intermeal automated basal insulin increases (see the strategies discussed for prandial insulin delivery below).

It is important to note that the CamAPS FX system has a customizable glucose target that can be set within the recommended pregnancy fasting glucose target range and is the only system currently approved for use in pregnancy affected by type 1 diabetes. The CamAPS FX app has a CE (Conformité Européenne) mark for use during pregnancy in the European Union (EU) and United Kingdom, indicating that it has met EU health, safety, and environmental requirements. Although the CamAPS FX app has been approved for use in the U.S. (26), the system itself is not yet available. In light of the significant glycemic benefits demonstrated for this AID system during pregnancy (13), a high priority should be placed on ensuring access to this novel technology, which averts the fasting hyperglycemia that may occur with use of the currently available AID systems in North America by having customizable targets down to 79 mg/dL (4.4 mmol/L) and importantly does so without increasing hypoglycemia. Likewise, a pilot study of an investigational AID algorithm tailored for use during pregnancy has demonstrated marked improvements in glycemia (14). This system, under development by a group from the U.S., is not yet commercially available, and future larger and randomized trials of this system are eagerly awaited. The advancement and increased availability of these novel technologies with targets and algorithms customized for pregnancy will be essential to ensure that individuals with type 1 diabetes may likewise benefit during pregnancy from the technologies that have revolutionized diabetes care outside of pregnancy and during a time of life when achieving optimal glycemia within a tighter glucose range is of the utmost importance for both pregnant individuals and their offspring.

Prandial Insulin Delivery: Pregnancy-Specific Limitations and Optimization

Achievement of postprandial glucose targets is a cornerstone of diabetes management during pregnancy (3,27–29). Compounding the challenges inherent in achieving the lower postprandial glucose targets advised during pregnancy (Table 1) is the progressive and disproportionate increase in prandial compared with basal insulin requirements observed across gestation (23,30) and the delayed prandial insulin effect in late gestation (31).

Although both basal and prandial insulin requirements increase over the course of gestation, prandial insulin requirements increase disproportionately. Basal insulin requirements have been shown to increase by ∼50% over the course of gestation, whereas the ICR needed for adequate prandial insulin becomes three- to fourfold stronger (23,30). Some hypothesize that one reason why conventional insulin pump therapy has not been shown to improve glycemia compared with a multiple daily injection (MDI) insulin regimen across large populations (32) is insufficiently aggressive prandial insulin adjustment across gestation, engendered by the availability of fractional incremental ICR changes in standard pump bolus calculators. Just as one needs to be mindful of the need to strengthen the ICR aggressively with conventional insulin pump therapy during pregnancy, one must also do the same with current AID systems.

Equally important as adjusting prandial insulin doses is attention to proper bolus timing. When the peak bolus effect is not optimally aligned with the peak of carbohydrate absorption, early hyperglycemia, followed by late hypoglycemia, can occur postprandially. As pregnancy progresses, there is commonly a delay in the prandial insulin peak effect and resulting glucose disposal (31). As a result, prandial boluses must be administered further in advance of meals with increasing gestation to avoid early postprandial hyperglycemia (from food absorption peaking before the optimal prandial insulin effect) and subsequent late postprandial hypoglycemia (when the peak prandial insulin effect occurs after peak carbohydrate absorption). This phenomenon can be amplified with AID use because the automated increase in basal insulin delivery that is triggered by immediate postprandial hyperglycemia can coincide with the late prandial bolus peak. As a result, adequate pre-bolusing is equally important in AID users to avoid this cycle of glycemic fluctuation. It is suggested that pregnant individuals bolus 15 minutes before eating in early gestation, increasing over the course of gestation to 30–40 minutes before eating later in gestation (31).

It has been observed that users of the CamAPS FX system (designed for and indicated for use during pregnancy) do not need to pre-bolus as much as with conventional insulin pumps or an MDI regimen, meaning that bolusing 10–15 minutes before eating is generally sufficient for CamAPS FX users to minimize postmeal glucose excursions and avert postprandial hypoglycemia (H. Murphy, personal communication). This is likely because of the responsivity of the algorithm, which enables basal insulin adjustments to optimize premeal glucose levels and enhanced mealtime adaptation of the algorithm (33). Adjusting meal composition and order (e.g., pairing lean protein sources with carbohydrates and eating protein before carbohydrates) can also help to mitigate early postprandial glucose spikes (34).

Delayed postprandial glucose excursions can also occur, classically after meals that are rich in protein and/or fat. In this case, a timing mismatch between the prandial insulin peak effect and peak glucose absorption can lead to the opposite postprandial glycemic pattern: early postprandial hypoglycemia (if prandial insulin takes effect before peak carbohydrate absorption), followed by late postprandial hyperglycemia several hours after the meal (when delayed carbohydrate absorption occurs after the prandial insulin dose has already waned). Use of extended boluses, in which part of the bolus is delivered before the meal and part is delivered more slowly over a specified period thereafter, can be helpful in this scenario. However, AID use presents unique challenges because the extended bolus feature is unavailable in some AID systems while in automated mode (MiniMed 780G and Omnipod 5). The Control-IQ system does allow administration of extended boluses. The maximum duration of an extended bolus was conventionally 2 hours in automated mode, but this maximum was increased to 8 hours with the most recent Control-IQ software update (35).

For systems that do not allow extended bolus administration in automated mode, alternate strategies can be used. For the Omnipod 5 system, users can divide the mealtime bolus into two parts, with one part (often half) administered before the meal and the remainder administered at completion of the meal (typically 1 hour later). When the second portion of the bolus is entered into the bolus calculator ∼1 hour later, users can enter solely a carbohydrate entry into the bolus calculator without entering another glucose value, so that insulin on board is not subtracted from the suggested bolus. Splitting the mealtime bolus in this manner can mitigate the hypoglycemia that can occur when the full bolus peaks before absorption of all the carbohydrate, and the delayed peak of the second bolus can lessen the late postprandial hyperglycemia resulting from delayed carbohydrate absorption. Individuals using the MiniMed 780G system are encouraged to deliver 100% of the bolus before the meal and to allow automated basal insulin adjustments with or without automated correction doses to be delivered per the algorithm for any delayed postprandial hyperglycemia. If late postprandial hyperglycemia persists despite these algorithm-driven automated adjustments, a user-initiated correction bolus can be delivered by entering a blood glucose measurement, and in the event that the system does not allow further correction, one could consider adding a phantom carbohydrate entry (limited to a small amount to avoid hypoglycemia). With the CamAPS FX system, a unique feature under the “Add Meal/Slowly Absorbed Meal” function can be used. When this feature is selected for a high-fat or high-protein meal, the user first enters a proportion of the carbohydrate in the bolus calculator to be delivered as a normal bolus and then enters the remainder as a slowly absorbed meal. Insulin for the carbohydrate entered as a slowly absorbed meal is delivered gradually over the ensuing 3–4 hours as needed in response to increasing glucose levels. As a starting point, it is suggested to give 60% of the meal bolus as a normal bolus before the meal and 40% as a slowly absorbed meal bolus, with future adjustments guided by observed glucose trends for that type of meal.

Considerations When Rapid Changes in Insulin Requirements Occur

Although a major benefit of AID systems is that algorithm-driven automated insulin adjustments can buffer glycemic variability throughout the day (36), AID use may not be advisable in some situations characterized by rapid changes in insulin requirements, such as during illness or supraphysiologic glucocorticoid use (1). During pregnancy, data are lacking regarding the safety and efficacy of AID use after administration of antenatal glucocorticoid treatment for fetal lung maturation or other indications. It has been recommended to proactively increase insulin doses by up to 40% in this scenario (37); thus, use of the pump in manual mode with sizable manual increases in insulin doses may be needed.

Likewise, the rapid decrease in insulin resistance that occurs immediately after placental delivery could complicate uninterrupted AID use postpartum among women using AID off label at the end of pregnancy. Typically, insulin doses are immediately decreased postpartum to levels even lower than pre-pregnancy insulin doses (38). Suggested therapeutic strategies have been described to reduce hypoglycemia risk among AID users in the immediate postpartum period, before the aggressivity of the AID algorithm, which is significantly affected by average total daily insulin dose in the preceding days, has adapted to the new lower postpartum insulin requirements (16,39). These strategies include reducing the basal rates to two-thirds of the pre-pregnancy basal rates (when using Control-IQ, which allows for basal adjustment) and making ICRs 20% weaker than pre-pregnancy and ISFs 20% weaker than pre-pregnancy immediately before cesarean delivery or at the time of pushing during vaginal deliveries. One study using a system that did not allow for basal adjustment demonstrated the safety and efficacy of uninterrupted AID use (MiniMed 780G) intrapartum and immediately postpartum with no severe hypoglycemia. In this study, the pregnant individuals were advised to weaken their ICRs and were given the option to increase their glucose target (40). Because studies evaluating the safety of uninterrupted early postpartum AID use are limited and not available for all commonly used systems, it is essential that future studies assess best practices in the postpartum period, when individuals face an elevated hypoglycemia risk.

Pregnancy-Specific Considerations Regarding Optimal CGM Interpretation in Pregnancy

At the crux of optimal AID use during pregnancy is accurate and efficient CGM interpretation because CGM-derived glycemic patterns are the key data upon which rational adjustments to pump settings are based. Systematic approaches to CGM interpretation outside of pregnancy rely on key CGM metrics and visual displays, namely the ambulatory glucose profile (AGP) report, to efficiently glean essential glycemic patterns (41). Yet, to effectively adapt these approaches to pregnancy, one must adjust the glucose target range to pregnancy-specific values on the CGM software output displays, which is possible on some (18–20,42,43) but not all (44) commonly used CGM software programs (Table 3). Given the established benefits of CGM use during pregnancy (3,45) and current guidance to recommend CGM to all pregnant individuals with type 1 diabetes (3) as well as many pregnant individuals with type 2 diabetes or gestational diabetes mellitus (GDM) (3,46), it is imperative that priority be placed on ensuring universal and simplified access to pregnancy-appropriate standardized CGM output displays on standard CGM software programs. Select practical workarounds for current CGM-related limitations are suggested in Table 3.

Table 3.

Practical Strategies to Adapt Major CGM Software Systems to Pregnancy-Specific Glucose Targets

CGM Software System	Is Glucose Target Range Displayed on the AGP Adaptable for Pregnancy?	Do Adjustments to Glucose Targets Apply to Individuals or to the Entire Practice?	Can Mean or Median Fasting Glucose at a Specified Time Period Be Assessed?
Dexcom Clarity (42)	Yes*	Individual	Yes (in the “Statistics” tab, click “HOURLY,” which provides mean glucose for 1-hour intervals)
LibreView (44)	No (use of the “Snapshot” tab allows for assessment of TIRp, TBRp, and TARp, but pregnancy-specific glucose targets are not reflected visually on the AGP display)	Individual	No
Glooko (19)	Yes	Individual	Yes (median glucose is visible in hourly increments in the “Summary” view, which is visible via mouse hover over different time segments of the AGP)
Tidepool (43)	Yes*	Individual	No
Medtronic CareLink (20)	Yes	Practice	No

*The pregnancy-specific glucose target range is set to 65–140 mg/dL (the lower limit of the target range is set to 65 mg/dL instead of 63 mg/dL when 5 mg/dL is the minimum increment available on the software program). TARp, pregnancy-specific time above range.

Controversies regarding optimal CGM targets during pregnancy remain, and future research elucidating CGM targets during pregnancy is needed (47). Although TIRp >70% is well established as a primary glycemic target in pregnancy associated with type 1 diabetes (7,48), additional research is needed to determine whether lower glycemic targets and/or a higher percentage TIRp should be recommended in pregnancies associated with type 2 diabetes or GDM. It is similarly unknown whether the goal TIRp should vary across gestation, given that glycemic targets remain difficult to attain in pregnancy and often are not achieved until late pregnancy (6). Emphasis must be placed on sustained achievement of glycemic targets throughout pregnancy to achieve optimal pregnancy outcomes (7).

In addition, future studies should assess the role of additional CGM metrics, including coefficient of variation (CV), a measure of glycemic variability, to be used in conjunction with TIRp as primary glycemic metrics. Reliance on TIRp alone is insensitive to fluctuations in fasting glycemia. For example, if fasting glucose is significantly above the fasting glucose target of 70–95 mg/dL (3.9–5.3 mmol/L) but remains within the overall pregnancy target range of 63–140 mg/dL (3.5–7.8 mmol/L), assessment of TIRp will not signal the need for therapeutic adjustment. Thus, alternative methods must be undertaken to assess fasting hyperglycemia, including assessment of daily glucose profiles and/or measures of mean or median glucose in the overnight or pre-breakfast period.

Defining fasting glucose can be challenging when measured by CGM, which provides multiple glucose values every 1–5 minutes throughout the overnight and pre-breakfast period. Standardization of fasting glucose assessment in pregnancy (defined, perhaps, as the mean of glucose values collected over a standardized time period) would assist both pregnant individuals and HCPs. Incorporation of a mean glucose target, establishment of optimal overnight mean glucose, or use of a separate nocturnal glucose target range could all serve as options to use in concert with TIRp to refine our definitions of optimal glycemia during pregnancy. Optimal CGM metrics and potential targets for these metrics have not been determined to date but merit further study.

Conclusion

When AID systems are used off label in select pregnant individuals based on consideration of individualized clinical factors, it is essential that HCPs be familiar with pregnancy-specific limitations, as well as best practice workarounds to optimize use. We suggest best practice workarounds to facilitate optimization of glycemia with AID use during pregnancy, as well as strategies to enhance CGM interpretation during pregnancy. In the future, development of more pregnancy-specific AID systems and refinement of CGM software programs to consistently reflect pregnancy glucose targets would greatly enhance care for pregnant individuals with diabetes during a time when achievement of glycemic targets takes on the utmost importance for both pregnant individuals and their offspring.