Clinical Approaches to Managing Glycemia with the iLet Dosing Decision Software

Abstract

The iLet Bionic Pancreas (iLet) is an FDA-approved automated insulin delivery system designed to manage glycemia through use of the iLet Dosing Decision Software. It uses real-time glucose data from a compatible continuous glucose monitor to automatically adjust and deliver insulin without the need for manually calculating, programming, and adjusting insulin therapy settings. The iLet autonomously determines all insulin dosing. For this context, autonomous insulin delivery is referring to the fact that all system dosing is independently determined by the system without requiring baseline clinician determined settings, but the system does require qualitative meal announcements by the user. It is shifting the paradigm for glycemic management toward simplification and expanding technology access to a broader population. The shift from adjusting settings to behavioral education represents a return to the art of medicine for the clinician. While the clinician no longer needs to adjust settings, they are not powerless when faced with dysglycemia. Clinical interventions are education focused. In this review, we provide clinical guidance for diabetes management with the iLet Dosing Decision Software (Beta Bionics, Irvine, CA, United States). Other systems with similar autonomous functionality and learning capability will likely be available in the coming years.

Keywords

automated insulin delivery continuous glucose monitoring diabetes autonomous insulin delivery pump therapy type 1 diabetes glycemic management

Introduction

The Diabetes Control and Complication Trial (DCCT) and United Kingdom Prospective Diabetes Study clearly demonstrated strong association between glycemic outcomes (A1c) and diabetes complications.^1,2 Unfortunately, the optimal glycemic outcomes in the intensive insulin treatment arms in the DCCT were achieved with increased incidence of severe hypoglycemia. With significant advances in automated insulin delivery systems, achieving glycemic goals without significant hypoglycemia is now feasible and without increasing significant burden on users.³

The progress in insulin delivery started from low-glucose and predicted low-glucose suspension of insulin delivery to algorithm-driven semiautomated insulin delivery based on sensor glucose with manual meal and correction boluses.^4–6 Ongoing advances include autocorrection boluses, scaling correction dosing based on the rate of glucose change, multiple glucose target set points and the latest, a fully adaptive system available commercially in the United States.^7,8 A number of studies have shown improved glycemic outcomes without increasing risk for hypoglycemia with automated insulin delivery (AID) systems. Hence, the American Diabetes Association (ADA) Standards of Care recommend AID for all individuals with insulin-requiring diabetes.³

Despite strong evidence, the use of AID systems in the real-world, while increasing, remains low.⁹ Barriers to using AID include lack of knowledge, provider inertia, cost, and perceived complexity in initiating AID systems. Progressively more autonomous insulin dosing is shifting the paradigm for glycemic management toward simplification and expanding technology access to a broader population. The shift from adjusting settings to behavioral education represents a return to the art of medicine for the clinician. While the clinician no longer needs to adjust settings, they are not powerless when faced with dysglycemia. Clinical interventions are education focused. In this review, we discuss management strategies for the clinician managing individuals using the iLet Dosing Decision Software (Beta Bionics, Irvine, CA, United States).

Background/Evidence

The iLet Bionic Pancreas (iLet) is an FDA-approved AID system designed to manage glycemia through use of the iLet Dosing Decision Software. It uses three adaptive closed-loop algorithms along with a compatible continuous glucose monitor (CGM) to automate insulin delivery, autonomously titrating all therapeutic insulin, including basal, correction, and prandial insulin.¹⁰ Current adaptive closed-loop algorithms, such as the iLet system, describe the algorithms as closed in that the dosing is independently determined by the system versus influenced by adjustable background insulin settings. User interaction is still required to initiate qualitative meal announcements for the iLet system to adapt to the user’s mealtime insulin needs. The next stage in AID development would be optional or have no need for meal announcements.

In the 13-week pivotal randomized controlled clinical trial in 440 subjects with type 1 diabetes, iLet use led to a 0.5% absolute reduction in A1C compared with control treatment (the participant’s usual insulin delivery method), with no increase in hypoglycemia and with improvements in diabetes distress and self-management burden.^10,11 A sub-analysis showed that use of the iLet led to similar improvements in A1C among both white and minority participants.¹² In research conducted on the psychosocial impact of iLet use, pivotal trial subjects and their families reported less worry, less time talking about diabetes, less friction with loved ones, and more freedom to enjoy their lives.¹³ Adult study participants reported significant decreases in fear of hypoglycemia and in diabetes distress.

In ongoing real-world analysis, commercial iLet users have a higher baseline A1C than participants in the pivotal trial and achieve larger decreases from baseline A1C to iLet glycemic management indicator (GMI) values.¹⁴ Derived from CGM data, the glycemic management indicator (GMI) provides an average estimate of a person’s blood glucose levels over a specified time period. GMI is frequently used in real-world analysis of data from users of diabetes technology including CGM and AID, to approximate A1C. Recent research reports that the change in GMI is influenced by the baseline A1C of the participants and it underestimates the true change in A1C; thus it may not fully capture the efficacy of an intervention in real-world studies or other research.¹⁵ The GMI equation has been updated to overcome the stated limitation and, hopefully, the updated GMI may serve as a useful marker for clinical and real-world studies in the future.^16,17

How the iLet Dosing Decision Software Works

The iLet Dosing Decision Software is currently utilized by the iLet Bionic Pancreas(^™) but could be utilized in future alternate controller enabled pumps. Clinicians should be aware that software updates are frequently made in diabetes technology and it is important to stay current with ongoing updates. The system determines insulin doses every 5 min in response to glucose values from a CGM, insulin dosing history, and user input (e.g., meal announcements and, when CGM glucose values are not available, manually entered blood-glucose values). The system does not require any information about the user’s total daily dose of insulin, basal or long-acting insulin requirements, or insulin correction factors. The system has three algorithms running in parallel: an adaptive basal insulin algorithm, which continually adapts to each individual’s basal metabolic need for insulin; an adaptive correction algorithm, which provides doses that are required above and beyond the basal metabolic needs; and an adaptive meal dose algorithm which provides insulin in response to a meal announcement (Fig. 1). The only user input is qualitative meal announcements when eating, and the system will adapt to the user’s identified insulin needs for each meal type. All basal and correction dosing is automated by the system.¹⁸

FIG. 1.

iLet Dosing Decision Software. CGM, continuous glucose monitoring.

If the apparent need for insulin decreases (e.g., a change in health condition resulting in increased insulin sensitivity, an increase in physical activity, or the effect of non-insulin medications), the insulin dosing aggressiveness will respond with relatively smaller doses. If the apparent need for insulin increases (e.g., a change in health condition resulting in decreased insulin sensitivity, a reduction in physical activity, the effect of non-insulin medications, or intercurrent illness), the insulin dosing aggressiveness will respond with relatively larger doses. The adaptation of meal doses is entirely separate from the adaptation of basal and correction insulin doses and is intended to track and continually update user insulin requirements around the three available meal types (breakfast, lunch, and dinner). Each algorithm references both historical and real-time data to determine dosing for the user. The basal and correction algorithms are most heavily influenced by the previous rolling 24 h. The meal dose algorithm references the insulin needs for meal announcements of each announcement type (breakfast, lunch, and dinner) over the last seven meal announcements of that type.

To fully remove the dosing decision burden from the end user in an insulin only AID system, it is important that the system receives high quality data to determine appropriate therapeutic dosing. For example, should a user experience a hypoglycemic event, it is important that the user recognizes the need to return to euglycemia versus hyperglycemia, as the latter could trigger correction insulin dosing and negatively impact learning in an autonomous system.

A case study (Fig. 2) demonstrates the importance of a fully adaptive system in times of rapidly shifting insulin needs.

FIG. 2.

Case study illustration of total daily dose and time-in-range before, during, and after glucocorticoid use. CGM, continuous glucose monitoring; TDD, total daily dose; TIR, time-in-range.

Other systems differ from the iLet in their approach to glycemic management. Commercially available insulin delivery systems in the United States automate insulin differently and require somewhat different strategies for glycemic management versus a learning autonomous system. Systems such as the Insulet Omnipod 5 (Insulet Corporation, Acton, Massachusetts, United States), Minimed 780 g (Medtronic Diabetes Minimed, Northridge, California, United States), Tandem Control-IQ+ (Tandem Diabetes Care, Inc., San Diego, California, United States), and Sequel Twiist powered by the Tidepool Loop algorithm (Sequel Med Tech, Manchester, New Hampshire, United States) require clinicians to enter settings, either individually or through a setup wizard, such as basal insulin rates, insulin sensitivity factors, carbohydrate-to-insulin ratios, total daily insulin dose, or some combination of these. The systems can automate insulin delivery within set parameters around the static settings. For meals, the user is generally asked to enter an exact carbohydrate amount or a fixed carbohydrate amount so the system can calculate a bolus dose. Achieving glycemic goals may still rely on user-initiated correction boluses when hyperglycemia occurs. The static settings in these systems require intermittent adjustment by a clinician when the user’s insulin needs exceed the maximum ability of the system to adjust insulin delivery based on the system parameters. The degree of required adjustments varies by product. Some systems also have the capability of adjusting basal insulin dosing based on changes in the user’s total daily dose of insulin. Users can achieve glycemic goals by using any automated insulin delivery system, but the user and clinician experiences will differ depending on the amount of system interaction required by both the user and clinician to achieve those outcomes.

Initiating the System

Patient identification

AID systems are recommended by the ADA as the preferred insulin delivery method over injections and other pump-based systems in people with diabetes, with early initiation of AID therapy noted to be beneficial.³ Before initiating therapy, clinicians should confirm that individuals and their families clearly understand their responsibilities for safe device use. Screening questions support assessment of the patient’s and/or caregiver’s comprehension of the iLet system’s function, required maintenance, and safety considerations (Table 1). An essential learning objective is to facilitate understanding of the transition from manual diabetes management, or AID systems requiring more user/clinician interaction, to an adaptive system that learns from daily glucose patterns and responds to insulin needs in real time, relying on minimal user interaction.

Table 1.

Screening Questions/Concepts to Assess Readiness to Use the iLet and Ability to Use It Safely

Topic	Screening questions/Concepts
Expectations	• How do you think the iLet will make it easier to manage your diabetes? • How do you think the iLet will affect your glucose? • Do you realize the pump will be supplying both your long- and fast-acting insulin? • Why is it risky if the system delivers insulin for a high glucose on your behalf and then you also take an external dose of insulin?
Carbohydrate awareness	• Which types of foods contain carbohydrates? • Which types of foods do not contain carbohydrates?
Potential to misuse meal announcements	• What is your comfort level around allowing a device to dose insulin on your behalf? ◦ If comfortable, move forward. ◦ If uncomfortable, explore the topic. To move forward, determine if the person recognizes and agrees with the need to provide the system with accurate data. If this causes such discomfort that the person would be unable to use the device safely, do not proceed.
Ability to maintain device/Understand DKA prevention	• Describe the risk of going without insulin for an extended period of time. • Do you feel willing and able to maintain an insulin delivery device, including changing the infusion site when needed and keeping the device charged and filled with insulin?
Overtreatment of hypoglycemia or unnecessary treatments due to fear	• How often do you worry about hypoglycemia? • At what number do you typically feel low blood glucose symptoms? • Tell me how you typically respond to lower blood glucose?

Patient onboarding for success

When starting the iLet, the user should have a current body weight available. All device-related safety topics are addressed during initial training. In addition to device maintenance-specific topics, the user receives education on helping the system to efficiently adapt to their specific insulin needs.

For the first few days, users are instructed to eat meals that have their “Usual” amount of carbohydrate for each meal type and to announce them as “Usual for me.” The system will perform best if allowed the opportunity to fully adapt to the “Usual for me” amount of carbohydrate before the user begins to choose “More” or “Less” options (Fig. 3). During the initialization period, users should be advised to avoid impacting glycemia with snacks/meal announcements in the 4-h postprandial window. The meal algorithm has adapted to the appropriate dose once postprandial glucose values are at goal.

FIG. 3.

Meal announcement guidance.

Keys to Achieving Optimal Outcomes

Addressing dysglycemia through coaching and education

The continuous learning of an autonomous insulin dosing system adds importance to these key factors to managing glycemia: device/site maintenance, responding to alerts, and appropriate use of meal announcements. Addressing dysglycemia requires patient education rather than setting adjustments. The clinician must assess and address any gaps in these key factors. The clinician should use the Bionic Report to evaluate these three factors. The report provides an overview of glycemic outcomes, meal announcement us, and recent insights that can assist the clinician with identifying areas of success as well as behavioral education opportunities. During this time the clinician should also assess any diabetes distress signals that could impact outcomes and quality of life.

Device and site maintenance

All AID users must keep the device charged/powered, filled with insulin, and connected to CGM. It is also important to be sure the infusion site is viable. If the iLet loses connection with the CGM, the system will continue insulin delivery and only adjust basal or correction dosing if a blood glucose value is entered. Clinicians should instruct patients to request replacement sensors to retain for backup in the event a sensor fails or does not last its expected time frame. All AID users should be encouraged to proactively order supplies and carry back-up supplies. Counsel on tips for addressing skin problems or difficulty with keeping the sensor on, potentially using barrier preps, tackifiers, over-tapes, or adhesive remover wipes as necessary. Patients should be advised to change infusion sets every 2 to 3 days, or if suspected of being compromised, and to rotate sensor insertion sites (arms, hips, buttocks, abdomen). A compromised infusion site that is unaddressed could negatively impact learning in a system such as the iLet that uses historic data to adapt insulin dosing.

Responding to alerts

All iLet alerts and alarms are designed to be actionable to minimize alarm fatigue (Table 2). Accordingly, users/caregivers should be instructed not to turn off important alarms and to always respond to alerts/alarms indicated on the device. Caregivers can receive glucose alerts on accompanying applications. Advise avoidance of preventative hypoglycemia treatment, unless receiving a predicted low glucose alert. Since the iLet will suspend insulin delivery prior to/during hypoglycemia, smaller treatments are needed and the user should not use the pause feature as a way to treat hypoglycemia. As with all AID systems, the user should aim to return the glucose to euglycemia to avoid triggering any correction insulin related to overtreatment of hypoglycemia. If the iLet system is regularly influenced by unnecessary unannounced rapid acting carbohydrate consumption it could cause the system to deliver more insulin than the user truly needs.

Table 2.

Appropriate Responses to Glucose Alerts

Alert and definition	Appropriate response	Setting options
Glucose falling quickly CGM glucose is <100 mg/dL and is falling at 2 mg/dL per minute or more.	Treat with 5–10 g of rapid acting carbohydrates.	On/off
Low glucose CGM Glucose is <75 mg/dL	Treat with 5–10 g of rapid acting carbohydrates.	On/off
Urgent low soon CGM Glucose is predicted to be <54 mg/dL within 20 min.	Treat with 10–15 g of rapid acting carbohydrates.	On/off
Urgent low glucose CGM Glucose is <54 mg/dL.	Treat with 10–15 g of rapid acting carbohydrates.	Fixed to on
High glucose Indicates CGM glucose has been >300 mg/dL for more than 90 min.	Clear/resolve any alerts/alarms that could stop insulin delivery (e.g., occlusion alarm). Change the infusion site. Users should follow their Ketone Action Plan. If an insulin injection is prescribed for elevated ketones, the iLet should be disconnected for 90 min to avoid stacking of insulin.	On/off

CGM, continuous glucose monitor.

Appropriate use of meal announcements

Patients should be instructed to announce meals solely based on the carbohydrate amount of their meal/snack, not on protein, fat, or calorie content or the quantity of food. The carbohydrate amount is relative to each user and their eating habits. Clinicians should review the user’s eating history to identify the meals they eat most often for each meal type (“Usual for me” meals) as well as their level of carbohydrate awareness. “Usual” meals are foundational to the system’s meal adaptation (Fig. 3). The clinician can use the Bionic Report to assess the appropriate use of meal announcements by focusing on timing, size, frequency, and postprandial glycemia (Table 3).

Table 3.

Managing Meal Announcements

Usual for me	Most meals should fall into the “usual for me” carbohydrate amount category.
Less	If seeing >40% “Less” meal announcements in any meal type: Confirm the user is not adjusting the meal size to “Less” for a “Usual” amount of carbohydrates related to a fear of hypoglycemia. This can cause the “Usual” dose to adapt upward to a level that is too high and therefore could be unsafe for the user’s “Usual” amount of carbohydrate when used again. Stop the behavior and allow the meal to adapt downward at the next announcement, being mindful of hypoglycemia; consider performing a factory reset if the meal has adapted upward to a dose that is no longer safe for use. If using the “Less” option frequently for a more varied carbohydrate intake, typically or for snacks, no intervention is needed.
More	If seeing >40% “More” meal announcements in any meal type: Confirm the user is not adjusting the meal size to “More” for a “Usual” amount of carbohydrates in an effort to manipulate insulin dosing. This will keep the iLet from being able to adapt the “Usual” meal dose to the user’s actual needs. Stop the behavior and return to the recommended use of the meal announcement feature. If using the “More” option frequently for a more varied carbohydrate intake typically or for larger snacks, no intervention is needed. This might occur more often during holidays/vacations.
Using meal announcements to correct hyperglycemia	Using a meal announcement to correct hyperglycemia could put the person at risk for severe hypoglycemia now as well as in the future by causing the meal dose to adapt incorrectly. Stop the behavior and return to the recommended use of the meal announcement feature.
Postprandial hyperglycemia or hypoglycemia	Confirm that the user is carb aware, is appropriately categorizing meals, and is not announcing meals late. If yes, for a few days, the user should allow a 4 h window after meal announcements (avoiding additional meal announcements or carbohydrate-containing snacks) to allow the system an opportunity to learn the user’s current needs. Ultra-rapid-acting insulins may improve postprandial glycemic outcomes for some users¹²
Announcement timing	Meal announcements should be made no more than 15 min prior to the meal, and no more than 30 min after eating begins. If it has been more than 30 min, the system’s correction dosing will already be active.
Not using/under using meal announcements (missed boluses)	Meal announcements allow for optimally timed insulin dosing. Not using this feature puts the person at a higher risk for hypoglycemia as the system tries to correct for the uncovered carbohydrates. Encourage meal announcement, even if only willing for meals with the highest carb amount. If the user is not using/under using the meal announcement feature despite eating carb amounts that would warrant announcements, consider reducing the CGM target to “Lower” to allow earlier correction dosing during a hyperglycemic excursion.

Meal adaptation applies across all dose sizes for a given meal type. When the system learns that a higher (or lower) dose is needed for that meal type, it automatically scales dosing and the “More” and “Less” doses will remain 50% more and 50% less than the “Usual” dose as the doses scale. If user behavior leads to inappropriate adaptation, a factory reset may be needed to restart learning (see Table 4).

Table 4.

Special Considerations for Managing Glycemia

Exercise	Users should always have hypoglycemia treatment available during exercise. Exercise recommendations should be individualized. Insulin on board can be referenced in the device history under Algorithm Steps. Preloading with unannounced carbohydrates prior to exercise while connected to the system could trigger correction insulin delivery and is not recommended. If preloading with carbohydrates, the user should be disconnected from the iLet. The user can disconnect and pause insulin delivery up to 30 min prior to activity but should be mindful of the glucose value and consider reconnecting and resuming before the system would command a large correction dose of insulin (e.g., resuming at a glucose level of 150 mg/dL vs a glucose level of 250 mg/dL). Increasing the CGM target by one step for the time of day when regular physical activity occurs can be considered if that activity typically induces hypoglycemia. For example, raise the CGM target from lower to usual for the morning hours when the user takes a daily long walk. This guidance is appropriate when first starting the iLet and for ongoing use.
Sick days	The iLet adapts autonomously and continuously to the user’s insulin needs. Therefore, in most cases, there is no need to make any adjustments to the iLet during illness. If the user experiences dysglycemia during an illness, an adjustment to the CGM target can be considered. The target should be returned to the initial setting once the user has returned to their pre-illness baseline. The clinician should always provide an alternative/backup insulin delivery plan as well as a plan for ketone management to the user.
Glucocorticoids	Typically there is no need to make any adjustments to the iLet based on its continuous autonomous adaptation to the user’s new insulin needs, but lowering the CGM target can be considered during glucocorticoid use. As insulin needs return to baseline, the iLet will adapt dosing downward as well.
Hospitalization	The iLet is not indicated for use in this setting. If used, care should be taken to be sure that meals are announced considering what the “usual” carbohydrate amount for meals was prior to admission. If CGM accuracy is compromised due to health status or medications, it is not safe to use the iLet. Consider any tests and appropriateness of wearing both iLet and CGM devices. If the iLet is shut down for the duration of the stay, it will require the charger to restart upon hospital discharge. The system will maintain learned data for approximately 1 month after the battery is depleted.
Clear liquids/Procedure prep/NPO (nothing by mouth)	The iLet adapts autonomously and continuously to the user’s insulin needs. Therefore, in most cases of clear liquids/procedure prep/NPO, there is no need to make any adjustments to the iLet. If carbohydrate-containing liquids are being consumed but slowly sipped (vs. consuming all at once), meal announcements would not likely be necessary. If prolonged fasting is recommended, raising the CGM target could be considered.
Gastroparesis	Users with gastroparesis should initially announce meals with the typical guidance of up to 15 min prior to the meal, but not more than 30 min after eating has started. If the user is experiencing hypoglycemia related to delayed carbohydrate absorption, the clinician can consider guiding the user through announcing a few minutes into the meal or just after, but before the system would deliver correction insulin.
Shift work/Unpredictable daily routine	Those with unpredictable routines should be encouraged to plan for device maintenance. These users should also be instructed to be consistent in the logic they use in announcing meals. For example, the first meal of the day is consistently designated as breakfast regardless of the time of day or the meal content. Alternatively, meals can also be announced based on the meal content, for example, pancakes are consistently labeled breakfast regardless of the time they are eaten.
Manual override of the system	Currently there is no safe way to override the system’s dosing without negatively influencing the system’s learning. The adaptation capability of the system should provide coverage for most metabolic scenarios that result in shifting insulin needs. Should concerns arise outside of this, the system can be reset (see Factory Reset below) or suspended, and the user instructed to follow a backup plan.
Factory reset	If a Factory Reset is performed, all the adaptation of the iLet algorithms will be lost. The user’s weight will have to be entered to restart the iLet, and adaptation will restart without any memory of previous use. If this function is being performed due to unmet glycemic goals or concern that the system has adapted using inaccurate data (such as misuse of the meal announcement feature), the clinician should be sure to cover all appropriate education topics prior to the reset to ensure that the system is able to adapt appropriately to the user’s needs after resetting. If meals have adapted up to an unsafe dose and the meal doses are no longer safe for use, factory resetting the system would be a reasonable intervention. If the basal or correction algorithms are not appropriately adapted due to frequently overtreating hypoglycemia or other activity, the system can typically adapt to the user’s true needs after a day or two of appropriate system use without the need for a factory reset.
Concentrated and regular insulin	Humalog® (insulin lispro) U200: The iLet has not been tested with or indicated for use with U200 insulin. The system cannot detect insulin concentration and was designed for use with U100 insulin.¹⁸ Should the clinician choose to use U200 as an off-label alternative for users with very high insulin needs, it is important that the system is either started new with U200 or a factory reset is completed prior to use with U200. A user should not switch back and forth between U100 and U200 concentrations of insulin. If U200 is used after adaptation has already occurred with U100 insulin, the risk of severe hypoglycemia is very high. Regular human insulin (RHI) U100 or U500: The iLet has not been tested with or indicated for use with RHI U100 or U500. The system assumes U100 rapid-acting insulin pharmacokinetics. Use of slower-acting RHI will cause inaccurate insulin-on-board calculations, increased insulin stacking, and a very high risk of severe hypoglycemia.¹⁸ The extended action time of RHI U500 amplifies these risks even further. These insulins should not be considered as off-label alternatives.

CGM Targets and Special Considerations

After prioritizing patient education and proper system use, adjusting the CGM target can be considered to address dysglycemia. The Usual target (120 mg/dL) is recommended for most patients when starting. The Lower target (110 mg/dL) may be helpful during times of the day that meals are eaten. This can help reduce postprandial spikes by responding faster to a rise in glucose. Increasing the target may be helpful if there is a pattern of drifting overnight hypoglycemia. The higher target (130 mg/dL) is typically not necessary except in cases of extreme insulin sensitivity. Follow up is essential to assessing the appropriateness of the CGM target. Users consuming fewer carbohydrates, especially simple carbohydrates, typically achieve lower mean glucose levels than those with higher-carbohydrate diets. When users make major dietary changes, the iLet requires time to adapt to new meal patterns.¹⁰

Table 4 outlines scenarios that require special consideration when managing glycemia for iLet users. The system’s ability to deliver all required correction insulin doses and to autonomously adapt changes the way clinicians should manage these scenarios for those on the iLet versus other AID systems. The guidance below is based on the iLet pivotal trial (exercise) and an understanding of how the system responds to changing glycemia and insulin needs. In the 2024 Position Statement, The Use of Automated Insulin Delivery Around Physical Activity and Exercise in Type 1 Diabetes, EASD and ISPAD recommend that iLet users reduce the meal announcement carbohydrate amount for meals consumed prior to physical activity.¹⁹ This approach has not been formally studied, and the impact on the system’s meal adaptation should be considered as a glycemic excursion postprandially would prompt both correction insulin dosing and meal adaptation upward if using a “Less” announcement for what would normally be announced as a “Usual” carb amount. Additional research on the impact of autonomous insulin delivery in these scenarios is needed to develop more specific guidance.

Potential Barriers

Potential barriers to achieving optimal outcomes include the system’s dependence on continuous and accurate CGM data and the user’s ability to recognize when their symptoms do not match the CGM glucose. It is also critical for the user to maintain blood glucose monitoring supplies in the case of CGM unavailability. The system’s ability to learn from historical data allows for the optimization of insulin dosing. However, if the system is frequently provided with low-quality data related to the user’s attempts to override the autonomous dosing, the system’s learning and subsequently insulin dosing are compromised.

Other potential barriers include candidate selection. In addition to using the screening questions in Table 1, clinicians should consider the user’s ability to safely use the product. Anyone who cannot independently maintain and operate the device and has insufficient caregiver support should not be considered as a safe candidate for this technology. The initial learning phase can be a challenging experience particularly for those individuals who experience some level of anxiety as they relinquish management tasks to the system. If the individual will not be able to allow the system to correct glucose excursions at its measured rate but is compelled to intervene/override the system to return to euglycemia faster, they are likely not good candidates for the technology. The iLet is not indicated for use and has not been tested for use in certain populations such as pregnancy and those on hemodialysis. Other AID systems may be more appropriate for these populations. Successful use of any AID system requires assessing and addressing the social drivers of health.

Ongoing Implications for Clinical Practice

Expanding technology access to nontraditional candidates and prescribers

As noted, CGM systems and insulin pumps, in particular AID, are now recommended as the standard of care for all persons with T1D.³ A serial, cross-sectional analysis of data from the Optum Labs Data Warehouse analyzed data from 186,590 youth and adults with T1D in the United States from January 1, 2009, to December 31, 2023. A rapid increase in the use of diabetes technology and notable improvements in glycemia among youth and adults were reported during the 15 years studied, however, the prevalence of achieving glycemic targets remained low, and racial, ethnic, and socioeconomic differences grew over time. The prevalence of achieving glycemic targets and the use of diabetes technology were lowest in Hispanic, non-Hispanic Black, and Medicaid-insured youths and adults with the differences persisting or even increasing over time.²⁰ A systematic review of observational studies published between 2017 and 2024 reported significant differences in diabetes technology prescription and use by race/ethnicity, with White non-Hispanic patients having the highest prescription rates (average 56.3% [range 12%–79%]), followed by patients that identify as Hispanic (average 28.8% [range 4%–76%]), and then Black (average 21.3% [range 3%–52%]).²¹

With the iLet’s capability to provide substantial, rapid improvements in glycemia without increasing the risk of hypoglycemia, including in those with high baseline A1C levels and across racial and socioeconomic backgrounds,¹² it has the potential to serve as a powerful public health intervention, including among high-risk, underserved populations.^10,22 Significant barriers remain to widespread implementation, including cost, insurance coverage, digital access, supply maintenance, and the availability of longitudinal support for underserved populations.

The reduction in burden for the user potentially makes the system more accessible to a broader population, including those who struggle with the complexities of insulin management, including carbohydrate counting. Those capable of insulin management but experiencing burnout are potential candidates if they are willing and able to allow the system to assume their glucose management, which may involve more deliberate relinquishing of closely held personal control for some people more than for others. The reduced user interaction compared with other systems may lead to AID adoption by a wider range of persons with type 1 diabetes^10,13,22 and be able to be deployed by a broader spectrum of clinicians including primary care,²³ thus helping to address disparities in technology adoption and outcomes.^24–28

AID systems are typically provided in the endocrinology clinic due to the intensive setup and programming required and the need for trained staff and adequate time and resources for ongoing support of most systems. However, three out of four counties in the United States do not have endocrinologists practicing within their borders. Around half of adults with type 1 diabetes rely on primary care for diabetes management.²⁹ It is vital to be able to offer technology including AID in primary care to meaningfully expand technology access for people with type 1 diabetes regardless of where they live. A randomized clinical trial including 40 subjects with type 1 diabetes compared deployment of the fully adaptive iLet in primary care versus endocrinology and in-person versus virtually. Of the participants, 97% achieved an average glucose < 183 mg/dL, representing an A1C level < 8%, while 64% achieved an average glucose < 154 mg/dL, representing an A1C level < 7%. An overall average glucose reduction of 17 mg/dL and a time-in-range increase of 11% with no change in sensor-measured hypoglycemia using iLet compared with usual care was reported. Similar outcomes were seen between initiation in endocrinology versus primary care or in-person versus virtual care.²³ A larger trial is currently enrolling. Trial designs such as this should be considered for other automated insulin delivery systems.

The evolving role of the clinician

As clinicians transition from manual settings adjustments as a primary focus of care to guiding patients in effective use of autonomous insulin delivery systems, they must embrace the reduced need for clinician intervention that these systems both allow and require for optimal performance. Clinical focus should shift toward supporting patient–system interactions and reinforcing appropriate behaviors related to when and how to engage with the system. This approach can improve time-in-range from baseline without increasing time-below-range. It may involve redefining success for each individual, achieving an acceptable level of glycemic management for the individual with less effort while improving the quality of life.³⁰ It is essential to set realistic expectations and to emphasize both the glycemic benefits and reduced treatment burden, while acknowledging that relinquishing direct control over insulin dosing may initially feel uncomfortable for both clinicians and patients who are historically accustomed to a high level of control over insulin dosing.³¹ The clinician’s role shifts to explaining how user–system interactions affect glycemia and safety, including downstream effects on algorithm adaptation beyond the user’s immediate intent, and to guiding patients toward safer strategies to achieve desired outcomes.

This is a partnership between the patient and the device, not a handoff. It requires learning how to share control safely. With less time required during clinical encounters to adjust insulin dosing, more time can be devoted to assessing and addressing psychosocial issues or other aspects of care. The clinical encounter is focused on the individual and their experience rather than settings and numbers.

Since ongoing adjustment of insulin therapy settings is not required with the iLet system, the diabetes care and education specialist (DCES) can play an expanded role in the ongoing management of the system in partnership with the user. Prescriptive authority is not required for providing the education regarding dosing behaviors that influence the functioning of the adaptive closed-loop algorithms. The DCES can shift from teaching advanced carbohydrate counting and math skills to person-centered conversations focusing more on healthy food choices and activity as part of an overall healthy lifestyle. This potentially opens mental space to problem-solve other aspects of diabetes management and potentially lessens the risk of maladaptive eating behaviors or eating disorders.³² Precise carbohydrate counting is helpful with some AID systems to optimize glycemic outcomes. There are additional strategies, such as fixed carbohydrate entry, that could simplify the process of meal dosing, but this still requires numeracy skills. It is recognized that various methods for teaching simplified carbohydrate counting have been used over the years.³³ The iLet allows for a simplified approach, requiring only carbohydrate awareness for its qualitative meal announcements.

Clinicians must adopt a broader, person-centered definition of success. While A1C remains a valuable metric, it should be complemented by time-in-range, glycemic variability, and person-reported outcomes such as treatment satisfaction, perceived burden, and quality of life. For many patients, success may mean fewer interruptions in daily life, less cognitive load, and reduced diabetes distress, even if glycemia is stable rather than improved. This approach allows clinicians to align care goals with what matters most to patients, promoting sustained use and trust in the technology as well as engagement with their self-management.

Conclusion

As automation advances, there is a need for targeted professional development to equip clinicians with the knowledge and confidence to support users effectively. As the clinical role shifts from manual insulin titration to interpreting device outputs and supporting behavioral adaptation, clinicians will need training in understanding how user input, daily routines, and life contexts influence algorithm performance and glycemia. Training must extend beyond technical operation and include strategies for expanding technology access to a broader population, expectation-setting, and addressing psychosocial dimensions such as trust in automation and shifting roles in self-management.

There are research gaps to be addressed, including the long-term glycemic outcomes and use in physical activity. Additional topics could include the sustainability of behavioral engagement, the impact on diabetes distress, quality of life over time, disordered eating, and caregiver burden across different life stages. Data on real-world access, use, and outcomes in diverse populations, including those with limited health literacy or numeracy, mental health conditions, or historically marginalized backgrounds, are needed. Integration of AID into broader models of diabetes care is essential for scalability and sustainability. This includes aligning with primary care workflows and leveraging digital health platforms and virtual coaching to provide longitudinal support.

Research is currently underway to explore the use of autonomous insulin dosing in those with cystic fibrosis, pregnancy, and for a longer duration than previously studied in primary care. Device manufacturers and academia continue efforts to optimize algorithm functionality. The future of AID will be impacted by artificial intelligence and improvements in the field of sensing, including but not limited to the dual sensing of glucose and ketones. Future systems may also incorporate additional drug delivery options, such as glucagon.⁷ Wearability and form factor improvements continue to evolve as well. All of these advancements should lead to a better future for people living with diabetes and those that care for them.

Authors’ Contributions

All authors contributed substantially to the design of the work and critical review and revision of the work to a finalized version. All authors approved the final version of the article.

Footnotes

Acknowledgments

Medical writing support was provided by Janice MacLeod, Janice MacLeod Consulting.

Author Disclosure Statement

I.B.H. has a research grant from Sequel Medical Technology and receives consulting funds from Abbott, Roche, and Hagar. D.I. is a speaker/consultant for Dexcom, Abbott, Medtronic, Insulet Corporation, Tandem, Beta Bionics, Novo Nordisk, Lilly, Sanofi, Sequel Medical Technology. P.M. has received honoraria for serving on advisory committees for Novo Nordisk, Biomea Fusion, Dexcom, Insulet, Medtronic, Abbot, and speaking fees for Novo Nordisk, Eli Lilly, Dexcom, Insulet, Medtronic, Abbot, and Beta Bionics. S.M.O.’s institution has received research support from the Helmsley Charitable Trust, Dexcom, Abbott Diabetes, Insulet, and Sequel Medical Technology, as well as honoraria for consulting from Dexcom and Insulet. V.N.S.’ institution has received research support from Dexcom, Eli Lilly, Enable Bioscience, Zucara Therapeutics, DEKA research, Cystic Fibrosis Foundation, Breakthrough T1D and NIH. V.N.S. has received honoraria for speaking, consulting or advising from Sanofi, NovoNordisk, Eli Lilly, Dexcom, Insulet, Tandem Diabetes Care, Sequel Medical Technology, Biomea Fusion, Roche, Abbott Diabetes, and T1D Scout; outside of this submitted work. B.S. has received speaking fees from Insulet Corporation, Beta Bionics, and Medtronic. B.S. has received fees for patient training from Beta Bionics, Insulet Corporation, Medtronic, and Tandem. K.W. has received speaking honoraria from Beta Bionics. K.P.C. is an employee and stockholder of Beta Bionics. J.M. is a consultant with Beta Bionics. S.R. is an employee and stockholder of Beta Bionics.

Funding Information

There was no funding for this work. Article processing charges for open access were paid by Beta Bionics (Irvine, CA).

References

1. Nathan

, Genuth

, Lachin

, Diabetes Control and Complications Trial Research Group. et al. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329(14):977–986; doi: 10.1056/NEJM199309303291401

2.UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes. Lancet 1998;352(9131):837–853; doi: 10.1016/S0140-6736(98)07019-6

3.American Diabetes Association Professional Practice Committee; 7. Diabetes Technology: Standards of Care in Diabetes—2026. Diabetes Care 2026;49(Supplement 1):S150–S165; doi: 10.2337/dc25-S007

4. Boughton

, Hovorka

. New closed-loop insulin systems. Diabetologia 2021;64(5):1007–1015; doi: 10.1007/s00125-021-05391-w

5. Leelarathna

, Choudhary

, Wilmot

, et al. Hybrid closed-loop therapy: Where are we in 2021? Diabetes Obes Metab 2021;23(3):655–660; doi: 10.1111/dom.14273

6. Pauley

, Berget

, Messer

, et al. Barriers to uptake of insulin technologies and novel solutions. Med Devices (Auckl) 2021;14:339–354; doi: 10.2147/MDER.S312858

7. Bergenstal

, Heller

, Breton

, et al. Evolution of the artificial pancreas: Components and integration—CGMs, insulin, and AP systems. Diabetes Technol Thera 2025;27(3_suppl):S9–S20.

8. Sherr

, Heinemann

, Fleming

, et al. Automated insulin delivery: Benefits, challenges, and recommendations. A consensus report of the joint diabetes technology working group of the European Association of the Study for the Study of Diabetes and the American Diabetes Association. Diabetes Care 2022;45(12):3058–3074.

9. Lingen

, Maahs

, Bellini

, et al. Removing barriers, bridging the gap, and the changing role of the health care professional with automated insulin delivery systems. Diabetes Technol Ther 2024;26(S3):45–52; doi: 10.1089/dia.2023.0440

10.

10. Russell

, Beck

, Damiano

, Bionic Pancreas Research Group. et al. Multicenter, randomized trial of a bionic pancreas in type 1 diabetes. N Engl J Med 2022;387(13):1161–1172; doi: 10.1056/NEJMoa2205225

11.

11. Beck

, Russell

, Damiano

, for the Bionic Pancreas Research Group. et al. A multicenter randomized trial evaluating fast-acting insulin aspart in the Bionic Pancreas in adults with type 1 diabetes. Diabetes Technol Ther 2022;24(10):681–696; doi: 10.1089/dia.2022.0167

12.

12. Castellanos

, Russell

, Damiano

, Bionic Pancreas Research Group. et al. The insulin-only bionic pancreas improves glycemic control in non-hispanic white and minority adults and children with type 1 diabetes. Diabetes Care 2023;46(6):1185–1190; doi: 10.2337/dc22-1478

13.

13. Weissberg-Benchell

, Vesco

, Shapiro

, et al. Psychosocial impact of the insulin-only iLet bionic pancreas for adults, youth, and caregivers of youth with type 1 diabetes. Diabetes Technol Ther 2023;25(10):705–717; doi: 10.1089/dia.2023.0238

14.

14. Russell

, Selagamsetty

, Damiano

. 947-P: Real-World efficacy of the iLet bionic pancreas in users with type 1 and type 2 diabetes during the first eighteen months of commercial availability. Diabetes 2025;74(Supplement_1):947–P; doi: 10.2337/db25-947-P

15.

15. Montaser

, Abad

, Shah

. Changes in A1C versus GMI across glycemic categories in clinical trials of type 1 diabetes. J Clin Endocrinol Metab 2025;110(12):e4182–e4187; doi: 10.1210/dgaf211

16.

16. Shah

, Xu

, Dabiri

, et al. Association of HbA1c and an updated glucose management indicator (uGMI) with incident diabetic retinopathy in adults with type 1 diabetes: A longitudinal study. Diabetologia 2026;69(3):610–617; doi: 10.1007/s00125-025-06599-w

17.

17. Bergenstal

, Beck

, Close

, et al. Glucose Management Indicator (GMI): A new term for estimating A1C from continuous glucose monitoring. Diabetes Care 2018;41(11):2275–2280; doi: 10.2337/dc18-1581

18.

18.iLet Bionic Pancreas System: Guide for Health Care Providers, Accessed January 7, 2025.

19.

19. Moser

, Zaharieva

, Adolfsson

, et al. The use of automated insulin delivery around physical activity and exercise in type 1 diabetes: A position statement of the European Association for the Study of Diabetes (EASD) and the International Society for Pediatric and Adolescent Diabetes (ISPAD). Diabetologia 2025;68(2):255–280; doi: 10.1007/s00125-024-06308-z

20.

20. Fang

, Xu

, Ballew

, et al. Trends and disparities in technology use and glycemic control in type 1 diabetes. JAMA Netw Open 2025;8(8):e2526353; doi: 10.1001/jamanetworkopen.2025.26353

21.

21. Underwood

, Chirokas

, Keels

, et al. Race and ethnic disparities in insulin pump and continuous glucose monitor use between 2017 and 2024: A systematic review with a focus on health equity. J Health Equity 2025;2:244002.

22.

22. Messer

, Buckingham

, Cogen

, et al. Positive impact of the bionic pancreas on diabetes control in youth 6-17 years old with type 1 diabetes: A multicenter randomized trial. Diabetes Technol & Thera 2022;24:712–725; doi: 10.1089/dia.2022.0200

23.

23. Oser

, Putman

, Russel

, et al. Assessing the iLet Bionic Pancreas deployed in primary care and via telehealth: A randomized clinical trial. Clin Diabetes 2025;43(3):388–398; doi: 10.2337/cd24-0104

24.

24. Ebekozien

, Mungmode

, Sanchez

, et al. on behalf of the T1D-QI Collaborative. Longitudinal trends in glycemic outcomes and technology use for over 48,000 people with type 1 diabetes (2016–2022) from the type 1 diabetes exchange quality improvement collaborative. Diabetes Technol Ther 2023;25(11):765–773; doi: 10.1089/dia.2023.0320

25.

25. Laffel

, Kanapka

, Beck

, CDE10. et al. Effect of continuous glucose monitoring on glycemic control in adolescents and young adults with type 1 diabetes: A randomized clinical trial. JAMA 2020;323(23):2388–2396; doi: 10.1001/jama.2020.6940

26.

26. Ebekozien

, Fantasia

, Farrokhi

, et al. Technology and health inequities in diabetes care: How do we widen access to underserved populations and utilize technology to improve outcomes for all? Diabetes Obes Metab 2024;26(Suppl.1):3–13.

27.

27. Santova

, de Bock

, Lanzinger

, SWEET Study Group. et al. Global inequities in diabetes technology and insulin access and glycemic outcomes. JAMA Netw Open 2025;8(8):e2528933; doi: 10.1001/jamanetworkopen.2025.28933

28.

28. Kanbour

, Everett

. Addressing disparities in technology use among patients with type 1 diabetes: A review. Curr Opin Endocrinol Diabetes Obes 2024;31(1):14–21; doi: 10.1097/MED.0000000000000840

29.

29. Oser

, Oser

. Diabetes technologies: We are all in this together. Clin Diabetes 2020;38(2):188–189; doi: 10.2337/cd19-0046

30.

30. Mora

, Postiglione Cook

, MacLeod

. Diabetes without math as automation advances: A shifting clinical paradigm. Diabetes Technology and Obesity Medicine 2025;1(1):329–339; doi: 10.1089/dtom.2025.0033

31.

31. Nefs

. The psychological implications of automated insulin delivery systems in type 1 diabetes care. Front Clin Diabetes Healthc 2022;3:846162; doi: 10.3889/cdhc.2022.346162

32.

32. Sulik

, Postiglione Cook

, MacLeod

. Meals no longer need to be math problems: Shifting from precise carbohydrate counting to a continuum of carbohydrate awareness as automated insulin delivery advances. Diabetes Technology and Obesity Medicine 2025;1(1):79–83; doi: 10.1089/dtom.2025.0010

33.

33. Pernick

, Rodbard

. Personal computer programs to assist with self-monitoring of blood glucose and self-adjustment of insulin dosage. Diabetes Care 1986;9(1):61–69; doi: 10.2337/diacare.9.1.61