Automated Insulin Delivery (AID) Guide for Practitioners

The holy grail of insulin delivery for people with diabetes who use insulin is automated insulin delivery (AID) (current term for “artificial pancreas”) through a closed loop system that maintains glucose levels within specified target range. That is, regardless of food consumed, physical activity, sleep, and myriad other factors that impact glucose levels.

While the currently available AID devices, and those at the FDA for review or in development, inch their way towards this holy grail goal, they will not, in the near-future, truly and fully close the loop. Here we define AID systems and their components, draw attention to US Food and Drug Administration (FDA) review processes for AID systems, and enumerate the benefits of these systems to users and caregivers. You’ll also find details on currently available AID systems and get a glimpse into the future. (Note: This article does not cover the Do-It-Your (DIY) systems (OpenAPS and Loop).)

Current and Near-Future AID Systems Defined

Over the last decade and with the brilliant expertise of and collaboration between dedicated experts in bioengineering, mathematics and the medical management of diabetes; steady progress has been made in the evolution towards the currently available AID systems. The availability of and advancement in continuous glucose monitoring (CGM) has made a critical impact on advancing this technology.

The evolution has followed a pathway from the Sensor-Augmented Pump (SAP) to Low Glucose Suspend (LGS) to Predictive Low Glucose Suspend (PLGS) to Hybrid Closed Loop (HCL) to Advanced Hybrid Closed Loop (AHCL) systems, which some refer to as AID systems.1 This evolution tracks the pathway mapped out by Aaron Kowalski, PhD, now JDRF’s Chief Executive Officer, and updated in 2009.^2,3

Today’s AID systems can increase basal insulin delivery to reduce hyperglycemia and suspend insulin delivery to reduce the risk and incidences of hypoglycemia.¹ Most AID systems integrate an algorithm (defined below) that continuously utilizes CGM data to automatically correct glucose levels that are either above or below a specified target glucose range.

The reason that the current AID systems are considered “hybrid,” is because food intake, either by estimating carbohydrate consumption, inputting a meal announcement with relative meal size, or other method; must still be done and a bolus insulin dose given through the pump device by the user.

For example, with the Omnipod5, Insulet’s AID system, users will be advised to enter the amount of carbohydrate they consume in the device’s SmartBolus calculator, where the user will preset specified parameters for bolus dosing. A few studies on various AID systems have shown that when the user is in auto-mode/auto-correct some amount of elevated glucose that occurs due to under bolusing or missed food bolus insulin for a minimal amount of carbohydrate, can be covered.4,5,6

Three Main Automated Insulin Delivery Components

CGM: A CGM device and its resulting data is an essential component. The CGM glucose results (data) are communicated to the device which are then used by the algorithm (defined below).

Insulin Delivery Device: Today there are two types of delivery devices, one that requires the user to attach an infusion set (tubing) from the device to the skin. The other is a pod that attaches to the skin and delivers insulin thru a short, thin cannula. All continue to deliver insulin subcutaneously, as they always have.

Algorithm: The algorithm can be considered the brains of an AID system that is built into the delivery device. There’s variation between the algorithms in AID systems. The various entities tend to be proprietary about these or use brand specific terminology to describe them.7 “An algorithm is a predetermined set of instructions used to tell the pump how to respond to the data (CGM) by providing or suspending insulin delivery,” says Davida Kruger, MSN, APN-BC, BC-ADM, certified adult nurse practitioner and diabetes clinician expert at Henry Ford Health System in Detroit. Kruger offers two examples:

1) the CGM reports to the pump the glucose level is 180 mg/dl. The algorithm assesses, based on the instructions, how much extra insulin a user needs and delivers it without the user taking any action.

2) The CGM provides data showing glucose is trending down in the next 30 minutes and if no change is made in the amount of insulin being delivered, the user will experience hypoglycemia. With this information, the pump automatically suspends insulin delivery and prevents hypoglycemia.

Bevy of Benefits

Anytime that people get relief from making one or many diabetes management decisions or performing daily tasks, it decreases disease burden. AID systems achieve both clinical and quality of life benefits.¹

Clinically, research shows that users of all ages can achieve a higher percent of glucose values within TiR (70 – 180 mg/dL) with less work. In hitting a higher TiR people minimize their Time Below Range (TBR), frequency of hypoglycemia and the need to treat hypoglycemia. Users also experience a lower percent of glucose results above TiR or Time Above Range (TAR). In addition, studies show improvements in their Glucose Management Indicator (GMI)9 result and A1C.

AID researcher and person with T1D for 34 years Jennifer Sherr, MD, PhD, Associate Professor in Pediatrics (Endocrinology), Yale University School of Medicine, shares her observations. “With AID, users are finally able to hand over glycemic variability to algorithmically delivered insulin. This allows users to achieve more targeted A1C levels (goal of <7.0%). And of importance, AID technology accomplishes this while also offsetting disease burden. This has been especially notable in overnight hours and as studies have shown users experience improved sleep quality.”

Other quality of life benefits are less diabetes distress and relief of some daily management burdens, including all the mental math. In sum, people on AID systems find they have more time to focus on other aspects of their lives.

Review Processes of AID Systems

In the US, FDA’s Center for Diagnostics and Radiological Health (CDRH) reviews medical devices through one of several pathways. AID systems have generally been submitted as Class III medical devices which requires the entity seeking the review to submit a Premarket Approval Application (PMA). These applications include results of so-called pivotal studies required by FDA. Read A Look Inside the FDA’s Review Processes for Medical Devices to read how CDRH has revised review processes to allow for diabetes device innovation. Given FDA’s focus on COVID-19 diagnostics and treatments for the last 18 months, several entities noted in their communication that this reality has slowed the review and approval of AID system submissions.

Around the globe entities work with other regulatory authorities. Entities may seek FDA review before doing so with other regulatory authorities or vise versa.

The AID Evolution Continues

AID systems will continue to evolve. The Kowalski pathway in Figure 1 offers insights as to how.^3,4 Each advancement, such as eliminating the need announce food or carbohydrate intake, should make managing insulin-requiring diabetes a bit easier and close the loop further.

In the future you may see the availability of faster-acting insulins that require essentially no lead time before they quickly lower glucose as desired. You’ll likely see multihormone AID systems, however, none are approved to date. Positive studies have been conducted with a system that uses rapid-acting insulin and glucagon¹⁰ and another with the faster-acting insulin FiASP and the hormone pramlintide.¹¹ In addition, there will likely be integration of decision support technologies that use artificial intelligence and machine learning, as noted in the table under the description of Medtronic’s Personalized Closed Loop. Yes, step-by-step AID is advancing!

For references and a glossary of terms, please visit the original post on T1D Exchange.