top of page

The Artificial Pancreas: How Closed-Loop Systems Are Automating Diabetes Care


Introduction

In a healthy person, the pancreas performs a quiet, remarkable balancing act every minute of every day. It senses the amount of sugar in the blood and releases just enough insulin to keep those levels in a safe, narrow range — never too high, never too low. For the hundreds of millions of people living with diabetes, that automatic system has broken down. According to the International Diabetes Federation, roughly 589 million adults worldwide now live with diabetes, a number projected to keep climbing for decades.

For people with type 1 diabetes, whose bodies make little or no insulin, survival depends on replacing that hormone by hand. Traditionally, this has meant a relentless routine: checking blood sugar, counting the carbohydrates in every meal, calculating a dose, and injecting or pumping insulin — then doing it all again hours later, often through the night. The central question that has driven diabetes engineering for half a century is deceptively simple: could a machine take over the job the pancreas can no longer do?

That machine now exists. Known formally as an automated insulin delivery (AID) system and informally as an “artificial pancreas,” it links a glucose sensor, an insulin pump, and a control algorithm into a single self-adjusting loop. These systems do not cure diabetes, and they are not fully autonomous yet. But they represent one of the most consequential shifts in diabetes care in a generation — and recent years have seen them expand from a niche tool for type 1 diabetes into a technology approved for a far broader population.

What Is an Artificial Pancreas?

Despite the dramatic name, an artificial pancreas is not an implanted organ. It is a coordinated trio of devices working together in what engineers call a “closed loop.” Understanding the three parts makes the whole system easier to grasp.

The Three Building Blocks

The continuous glucose monitor (CGM). A small sensor sits just under the skin and measures glucose in the fluid between cells, typically every one to five minutes. Instead of a single snapshot from a finger prick, it produces a continuous stream of readings, revealing not just the current level but the direction and speed of change.

The insulin pump. A compact, battery-powered device delivers rapid-acting insulin through a tiny tube or patch worn on the body. Unlike injections given a few times a day, a pump can deliver insulin in continuous micro-doses, adjusting moment to moment.

The control algorithm. This is the brain of the system — software that reads the sensor data, predicts where glucose is heading, and tells the pump how much insulin to deliver. It is the algorithm that turns two separate devices into something that behaves, at least partly, like a working pancreas.

The degree of automation varies across the clinical spectrum of these devices:

  • Open-Loop Systems: The CGM and pump function independently without direct algorithmic communication. The user remains the sole controller, viewing sensor trends and manually adjusting all insulin doses.

  • Hybrid Closed-Loop (HCL) Systems: These systems represent the current commercial standard. The algorithm automatically adjusts background (basal) insulin delivery around the clock and delivers automatic correction doses when hyperglycemia is detected. However, because of the pharmacokinetic lag of subcutaneous insulin absorption, the user must still manually announce meals and deliver a bolus for estimated carbohydrates to prevent postprandial glucose spikes.

  • Fully Closed-Loop (FCL) Systems: These systems operate autonomously, adjusting to meals, physical activity, and stress without requiring manual user input or carbohydrate counting. Research-grade FCL systems perform well in controlled trials, but managing rapid postprandial glucose excursions remains a clinical challenge due to the slow onset of subcutaneously administered insulin compared to physiological insulin release into the portal vein.

Commercially Available Systems

Several hybrid closed-loop systems are now FDA-approved and in everyday use. While they share the same basic architecture, they differ in design, the sensors they pair with, and the populations they are cleared for.

Medtronic MiniMed 780G. Approved by the FDA in 2023, this advanced hybrid closed-loop system automatically adjusts basal insulin and delivers automatic correction doses every five minutes. It uses meal-detection technology to help compensate when a user forgets to dose for food or underestimates a meal.

Tandem Control-IQ. Running on the t:slim X2 pump and built on an MPC algorithm, Control-IQ became, in 2019, the first FDA-approved automated system that did not require fingerstick calibration. It predicts glucose roughly 30 minutes ahead and adjusts delivery accordingly, with an intensified target overnight.

Insulet Omnipod 5. Cleared in 2022, Omnipod 5 was the first automated system delivered through a tubeless, wearable “pod” with no separate handset required and no calibration. Its tubeless design has been especially appealing to children and people who dislike wearing tubing.

CamAPS FX. A smartphone-based app developed in the United Kingdom, CamAPS FX pairs with compatible pumps and sensors and is notable for its use across young children and during pregnancy, populations where tight glucose control is especially important.

Real-world comparisons of these systems generally show all of them improving glucose control over older methods, with differences between them tending to be modest and influenced heavily by how each system is set up and used.

Measuring Success: Glycemic Outcomes and Metrics

For decades, diabetes care was judged mainly by HbA1c, a blood test reflecting average glucose over about three months. It remains important, but an average can hide a turbulent reality — wild swings between highs and lows can produce the same HbA1c as steady, well-controlled glucose.

That is why continuous glucose monitoring introduced a more revealing metric: time in range (TIR), the percentage of the day a person spends within a target glucose band, usually 70 to 180 mg/dL. Alongside it, clinicians track time above range and time below range, the latter being especially important because hypoglycemia (low blood sugar) is one of the most dangerous and feared complications of insulin therapy.

By these measures, artificial pancreas systems consistently deliver. In a real-world analysis comparing the three major systems in adults with type 1 diabetes, users who switched to automated delivery saw substantial gains in time in range — in one group, an increase of roughly 21 percentage points after adjustment — along with corresponding drops in time spent too high. Just as importantly, these improvements typically come without raising the risk of hypoglycemia, which is the holy grail of insulin management: better control and greater safety at the same time.

Major Expansion: From Type 1 to Type 2 Diabetes

Until recently, artificial pancreas systems were almost exclusively type 1 diabetes technology. That is changing in a significant way. Because type 2 diabetes is roughly ten to twenty times more common than type 1, and because many people with advanced type 2 diabetes eventually require insulin, extending automation to this population could affect an enormous number of people.

The evidence has been building rapidly. In a randomized controlled trial published in 2025 in the New England Journal of Medicine, adults with insulin-treated type 2 diabetes who used an automated system saw their HbA1c fall by about 0.9 percentage points over 13 weeks, compared with 0.3 points in a control group using continuous glucose monitoring alone. Their time in range rose markedly, and hypoglycemia remained rare. On the strength of evidence like this, regulators and insurers have begun extending coverage of hybrid closed-loop systems to adults with type 2 diabetes, which is a notable widening of who these devices are meant to serve.

This expansion does not mean everyone with type 2 diabetes needs an artificial pancreas. Many people achieve good control with oral medications, GLP-1 receptor agonists, or other therapies. But for those who depend on insulin and struggle to keep glucose stable, automation is increasingly on the table.

The Next Frontier: Dual-Hormone Systems

Every system described so far delivers only one hormone: insulin, which lowers blood sugar. But the healthy pancreas uses two opposing hormones. When glucose drops too low, it releases glucagon, which signals the liver to release stored sugar and raise blood glucose back to safety. An insulin-only system can only ease off the gas; it cannot hit the brakes in the other direction.

This has driven the development of bihormonal, or dual-hormone, systems that deliver both insulin and glucagon. The most prominent example is the iLet Bionic Pancreas from Beta Bionics. The insulin-only version of the iLet earned FDA approval in 2023 and is notable for its simplicity: it requires only the user’s body weight to initialize and does not ask for carbohydrate counts or preset insulin ratios. The company is also developing a bihormonal configuration that pairs insulin with dasiglucagon, a stable, ready-to-use glucagon analog, with the aim of defending more aggressively against lows.

Dual-hormone systems are more complex, requiring two drug reservoirs, more intricate hardware, and a stable pumpable form of glucagon, which has historically been difficult to formulate. But the promise — a system that can both lower and raise blood sugar automatically — brings the artificial pancreas one step closer to mimicking the real thing.

Limitations and Challenges

For all their progress, artificial pancreas systems are not a finished solution, and an honest account must acknowledge where they fall short.

The meal problem. Because injected insulin works more slowly than insulin released naturally, even the best hybrid systems still benefit from users announcing meals. Fully hands-off mealtime control remains a work in progress.

Burden and trust. Wearing two devices, managing alarms, changing infusion sites, and troubleshooting connectivity all take effort. In some real-world studies, a meaningful fraction of users stopped using automated mode over time, often because of alarm fatigue or frustration. Adopting technology successfully also requires education and a willingness to let an algorithm take control.

Cost and access. These systems and their ongoing supplies are expensive, and coverage varies widely. This raises real concerns about equity, since the people who might benefit most are not always the ones who can afford or access technology.

Hardware on the body. Some users dislike the visibility of devices or experience skin irritation at sensor and infusion sites, factors that affect long-term adherence.

Conclusion

The artificial pancreas is one of the clearest examples of how engineering and medicine, working together, can change the daily experience of a chronic disease. In a single generation, diabetes care has moved from manual finger pricks and mental arithmetic to algorithms that quietly adjust insulin every few minutes, often while the user sleeps. The technology has expanded from type 1 diabetes into type 2, from insulin-only to emerging dual-hormone designs, and from cumbersome research prototypes to pocket-sized commercial devices.

References


Assessed and Endorsed by the MedReport Medical Review Board

 
 

©2025 by The MedReport Foundation, a Washington state non-profit organization operating under the UBI 605-019-306

 

​​The information provided by the MedReport Foundation is not intended or implied to be a substitute for professional medical advice, diagnosis, or treatment. The MedReport Foundation's resources are solely for informational, educational, and entertainment purposes. Always seek professional care from a licensed provider for any emergency or medical condition. 
 

Privacy Policy

Editorial Policy

Terms & Conditions

bottom of page