Fully Closed Loop: Automated Insulin Delivery Takes the Next Step

Current makers of automated insulin delivery (AID) systems, university researchers, and some members of the diabetes community are testing and refining next-generation systems that allow you to skip meal announcements and other user input.

The approach reduces diabetes management stress, said experts who research the systems and users who have tried them. 

"The freedom of not announcing meals has a profound effect," said Marc Breton, associate director for research at the University of Virginia’s (UVA) Center for Diabetes Technology. 

Multiple diabetes technology makers have fully closed loop systems in development, while some open source diabetes community members are using existing autonomous systems modified to eliminate the need for meal announcements. Both users and researchers have years of experience using these systems full-time with apparently good results. 

Closing the loop

In most hybrid closed loop systems, the user is expected to count the number of carbohydrates in a meal. That number is then entered into a bolus calculator, which determines the mealtime dose; this is based on how sensitive the person is to insulin using a ratio of how many carbs one unit of insulin typically covers.  

There is currently one FDA-approved closed loop system (Beta Bionics’ iLet) that uses an adaptive closed loop algorithm that only requires a user’s body weight but no additional user input. However, Beta Bionics does recommend meal announcements to help control blood sugar spikes after meals.  

In a fully closed loop system, no carbs are counted, no bolus is entered, and no meal announcements are needed. Instead, as blood sugar rises after a meal, relatively large doses of insulin are delivered.

The algorithms used in fully closed loop systems vary, using a range of different methods to identify and counter rises in blood sugar after meals. Some fully closed loop systems identify the rate of acceleration of the post-meal rise, and the ratio determining the bolus can be adjusted depending on how the user typically responds.

Some systems use an initial, small dose of insulin, followed by larger doses to respond to the rise in post-meal blood sugar. Breton called this "bolus priming."

"A small amount of insulin will open your capillaries and make the next dose of insulin much more efficient, which lets the algorithm get ahead of the blood sugar rise from a meal," he said.

Delivering larger insulin doses can introduce safety concerns, which are addressed, in part, by dialing back the aggressiveness of the system when food isn't typically being consumed, for example, overnight. 

Ideally, the system will deliver insulin quickly enough to bring blood sugar back into range after the meal without causing hypoglycemia (low blood sugar). Introducing insulin doses earlier is meant to avoid "stacking" insulin, which is when too many doses are given too close together, causing blood sugar to crash.

A break from carb counting

The UVA Center for Diabetes Technology recently announced a research partnership with Tandem to develop fully closed loop AID systems, which have been tested at the university since 2017 (Tandem's current Control-IQ AID system was developed based on research done at UVA).

Breton said he considers the need to announce meals a hindrance for people with diabetes who could benefit from automated insulin delivery.

"That hurdle is something that shouldn't be in an automated insulin delivery system," Breton said. "And we have the tools to make it go away. If we can get rid of that constant focus on, 'What am I eating, how am I doing, what is it going to do to me?' then we can offer something really important to people who live with the disease."

As algorithms become more powerful and increasingly capable of reducing blood sugar spikes in a fully closed loop setting, it's unclear if most people will allow the systems to be autonomous or whether some will still want manual control. Breton said more research is needed to see how people respond to eliminating meal announcements.

"We always assess outcomes," Breton said. "So far it has been overwhelmingly positive. It is definitely something people care a lot about."

Fully closed loop in the real world

David Burren, who has lived with type 1 diabetes since 1982, uses a fully autonomous AID system that routinely keeps his blood sugar well in range. 

Burren is co-chief of Nasence Biomed, a technology company that is working on an autonomous closed loop AID system developed in part based on his experience.

About seven years ago, Burren, who lives in Melbourne, Australia, started using OpenAPS, an open source, hybrid closed loop AID system. Over the years he refined his settings, including basal rate and insulin sensitivity factor, and got good results. He saw that part of the success was due to the software’s ability to detect unannounced meals and make corrections.

"I realized a lot of what the system was doing was cleaning up when I forgot to do things, such as declare snacks," he said. "So I figured I’d see what happened when I didn’t declare a meal at all."

He continued to tweak his settings, and in early December of 2020, he gave his last bolus. In late January of the following year, he stopped announcing meals or manually bolusing. "I haven’t looked back," he said. "The fourth anniversary of that last meal announcement is fast approaching."

Initially, his time in range dropped, as did his time below range (less than 70 mg/dL). Still, his endocrinologist commented that his results were better than most of her other patients. He continued to tweak his system settings until he was satisfied.

Recent results show his time in range at about 97% and time in tight range (70–140 mg/dL) at about 90%. He doesn't eat a low-carb diet, which many current AID systems would likely handle well without meal announcing. He said he typically eats about 200-250 grams of carbs a day.  

A choice vs. a necessity

In general, the more information an AID system is given, the better the results will be. Those interviewed for this story said that even as the systems move toward fully closed loop, some people will always prefer to announce meals, count carbs, and manually bolus – and they will likely see better results. 

"I would get less blood sugar variation if I put the work in, counted carbs, and announced all meals," Burren said. He added that manually bolusing would tighten things up even more. “But the point is, I get results that are good enough without having to bother."

However, in the near future, experts said, announcing meals will be an option in AID systems – at least for some – instead of a necessity as it is now. 

"It literally changed my relationship with my diabetes when I no longer had to count carbs," said Burren. "I glance at my watch occasionally and get reassurance that the system is working properly, but that’s happening less and less. I’m not watching it and planning what corrections and adjustments I need to do."

What does the research say?

Dr. Charlotte Boughton is the lead author of a study called CLEAR (The Closing the Loop in Adults With Type 1 Diabetes), which tested a fully closed loop system in participants whose A1C was above 8%. The American Diabetes Association recommends most people aim for an A1C of 7% or less.

Boughton is a clinical lecturer in diabetes and endocrinology at the University of Cambridge in the U.K. and works with other researchers to develop algorithms that can adapt to different individuals and reduce the need for carb counting and bolusing.

"I don't think we're that far off from the majority of people with type 1 being able to get to 70%  time in range with a fully closed loop system,"  Boughton said, if used alongside adjunct therapies, which could include GLP-1 receptor agonists and SGLT-2 inhibitors. While the optimal time in range for diabetes isn’t really known, experts generally recommend people with type 1 and type 2 diabetes aim for at least 70% of the day between 70-180 mg/dL. 

The CLEAR study used a fully closed loop AID system developed at Cambridge, called CamAPS HX, which uses an aggressive algorithm initially designed to treat people in hospitals with type 2 diabetes. Cambridge researchers also created a widely used hybrid closed loop system called CamAPS FX, which has been cleared for use by the FDA but is not yet available in the U.S.

 "We were just really curious how this was going to work in people with type 1 because [fully closed loop] is obviously the next step," Boughton said. "It is definitely the way we're going."

After eight weeks of using the AID system, time in range for the CLEAR study participants went from 36% to 50%, without announcing meals. 

Current hybrid closed loop systems have been shown to keep most participants in studies in range about 70% of the time.

Challenges to closing the loop

Boughton said part of the difficulty of creating a fully closed loop system is that individuals react very differently to meals – and to the effects of insulin. 

Some people will do well in a fully closed loop system with an aggressive algorithm, she said, because their insulin metabolism happens quickly and won't lead to hypoglycemia. Others may not do as well with such algorithms. She sees these issues as both a challenge for creating next-generation AID systems and an opportunity for personalization. 

"There's more to it than just the safety constraints and aggressiveness of the algorithm," Boughton said. "You can't predict it when you see someone at the outset – I couldn't tell you who's going to do well on this and who isn't."

For example, some of the CLEAR study participants reached 70% time in range, while others only saw 20%, she said.  

"It comes down to personalization," she said. “You tweak [the settings for the algorithm] when you see what sort of characteristics a person has." 

For those who struggle with multiple daily injections or insulin pumps without an AID component, a fully autonomous system could lead to major improvements in time in range, as seen in several studies. 

"What I suspect will happen with this new generation system is that there are meals that people will just ignore, and the system will handle it perfectly well," said UVA's Breton. 

"And then there are meals where you're eating out with friends, and you're actually going to give a bolus. The new system will be able to take that into account and achieve really excellent glycemic control. It's a matter of choice. Ultimately, as an engineer, that's all I should be in the business of doing. I shouldn't assume the right solution," he said.

The bottom line

Fully closed loop AID systems aim to simplify diabetes management by eliminating the need to announce meals, carb count, and manually bolus insulin. They do so by using more aggressive algorithms that deliver insulin soon after a post-meal blood sugar rise is identified.

Those who have used fully closed loop systems report a sense of freedom, such as a CLEAR study participant who described a major improvement in mood and quality of life: "Diabetes worked around my life rather than my life working around the diabetes."

Studies have shown that users with high A1C levels saw improvements in time in range when moving to a fully closed loop system. However, the time in range in those studies was around 50%, whereas hybrid closed loop systems usually deliver time in range of 70%.

More work needs to be done before fully closed loop systems allow people with diabetes to see similar results to those using current AID systems. Some people with diabetes may respond too slowly to fully closed loop systems; for those individuals, an autonomous mode might serve as an occasional break from meal announcements. Adjunct therapies, like the use of GLP-1 receptor agonists and SGLT-2 inhibitors, as well as faster insulins, could help some who might otherwise struggle to reach time in range goals with next-generation AID systems.

Experts predict fully closed loop systems will be the standard in the future, offering complete freedom from meal announcements for some, while others may choose to announce meals for better results. In the end, these next-generation systems may offer meal announcements, carb counting, and manual bolusing as a choice instead of a necessity.

Learn more about AID systems and related diabetes technology here:

• Tech Watch: The Latest in Diabetes Tech News
• AID Technology Improves Diabetes Care No Matter How You Get It
• AID Systems Hit Impressive Time in Range Targets, Large Studies Show
• Could AID Transform Type 2 Diabetes Care?
• Benefits of Automated Insulin Delivery for Children
• How To Start Using a CGM
• Nightscout: How To Get Started With CGM in the Cloud