Diabetes Forecast

The Artificial Pancreas: Now and in the Future

For years, an artificial pancreas was a pipe dream. Now, we’re closer than ever to systems that will automatically adjust background insulin. Here’s what’s in the works—and what’s still to come.

Illustration by Kaitlin Graves

This story was updated on April 27, 2017.

Since she was diagnosed with type 1 diabetes almost 40 years ago, Alecia Wesner has spent a lifetime managing her condition: Crunching numbers, counting carbs, and trying on a variety of wearable devices to help keep her blood glucose under control. “My numbers are within a recommended target, but I work like crazy at that,” she says.

So when Wesner, 43, was asked to participate in trials for a hybrid artificial pancreas insulin delivery system, the New York City–based lighting and jewelry designer jumped at the chance. The system, called the Diabetes Assistant (DiAs), is an automated blood glucose management tool, what some researchers are referring to as a closed-loop system or bionic pancreas.

Over the course of a few years, Wesner has spent weeks going to study visits at Mount Sinai Hospital in New York City and sleeping in hotels while hooked up to an artificial pancreas prototype with teams of doctors watching her blood glucose levels rise and fall in real time. She has since used the systems—made up of an insulin pump, a continuous glucose monitor (CGM), and a smartphone application running a program that lets the two communicate and administer insulin—while training for cycling events.

Designed to mimic the function of its namesake organ, the system calculates a person’s insulin needs based on glucose readings, activity, carb intake, sleep, and other factors. Then it automatically adjusts and delivers basal doses of rapid-acting insulin around the clock. That’s an especially big deal at night, when people with diabetes have to wake up in order to treat highs or lows.

Where the artificial pancreas most noticeably differs from the real thing is in mealtime insulin delivery. In the trial Wesner took part in, the experimental system calculated mealtime boluses based in part on her current and trending insulin and glucose levels, but she had to count carbs and deliver the bolus dose by pushing buttons, just like using a regular insulin pump.

Wesner says the systems have gotten smaller, more streamlined, and more reliable since she started participating in the trials. They’re essentially modified existing pumps and CGMs, worn on a belt or in a pocket, that interact with a person’s body via an infusion set and a sensor inserted under the skin. For someone who’s lived with diabetes almost her entire life, the promise of a device that could reduce some of the daily burden of the disease is tremendous. “It’s one thing to hear about technology that could change your life,” she says. “It’s pretty different to actually wear it.”

Many people with type 1 and type 2 diabetes could have access to more than a prototype soon. In September 2016, the Food and Drug Administration (FDA) approved the first partial artificial pancreas system—Medtronic’s MiniMed 670G—for sale in the United States. As of press time, it’s expected to release in late spring. At least five other companies are working on their own systems, hoping to bring them to market by 2020.

The first devices to hit the market will be so-called “partial” artificial pancreases, requiring user input and monitoring for mealtime boluses and automatically adjusting only basal insulin. A few years from now, developers hope to release systems that will manage all insulin delivery automatically: The user can essentially strap it on and then forget about it.

Doing the Math

When it’s working properly, the pancreas is a wonder of balance and sensitivity. Inside the hot dog–sized organ, millions of beta cells monitor the levels of glucose in the bloodstream. When glucose climbs too high, the beta cells release insulin, a hormone that prompts cells elsewhere in the body to absorb and store the glucose and shuts off the release of glucose by the liver. If blood glucose goes too low, a different set of cells in the pancreas—called alpha cells—pump out the hormone glucagon, which tells the liver to release stored glucose.

Diabetes is what happens when the beta cells in the pancreas stop working as they should. Without the beta cells to sense rising (or falling) blood glucose levels and to release insulin accordingly, people with diabetes are forced to rely on substitutes—from finger sticks and insulin pens to CGMs and insulin pumps. These require users—typically people with type 1—to master sometimes-tricky tools and techniques and carry around one or more devices all the time.

And then there’s the math. The latest generation of CGMs may provide accurate, real-time readouts of glucose levels. But it’s still up to users to program basal rates, count carbs, and reactively decide to deal with highs and lows—instead of avoid them in the first place.

“The burden of diabetes self-care with technology has become more challenging,” says Carol Levy, MD, an endocrinologist at Mount Sinai Hospital in New York City who heads the Icahn School of Medicine’s Artificial Pancreas Research Program, which uses the DiAs system—an earlier version of the TypeZero system designed by University of Virginia researchers—in its trials. “You have all this data, and the patient has to figure out what to do with it on a day-to-day basis.”

Relieving the Burden

The flood of numbers may help explain some curious patterns researchers have noticed when it comes to who uses technologies such as CGMs and pumps, the building blocks of artificial pancreas systems. In a study published last year in Diabetes Care, researchers found that nearly half of all people with type 1 diabetes in a group of about 15,000 were reluctant to use devices because they were a “hassle.” Interestingly, the study showed that older users were more comfortable with using technology to manage their diabetes, while younger people—those between 18 and 25—were the least likely to wear CGMs or insulin pumps.

The artificial pancreas products should make things easier. In theory, the idea is simple: Combine a continuous glucose monitor with an insulin pump. When the CGM senses glucose levels rising, it sends a signal to the insulin pump to dose insulin. When it senses dropping glucose levels, it decreases or suspends insulin delivery.

At the system’s heart is a complex formula, called an algorithm, built into the device or stored on a smartphone in the form of an app. “It’s a very sophisticated computer program,” Levy explains. The algorithms take in data—how long insulin lasts, how fast glucose is rising, how a person reacts to insulin—and calculates how much insulin or, in some models, glucagon to deliver.

The first devices to market won’t do this all by themselves. In its press release heralding the Food and Drug Administration approval of the first hybrid closed-loop insulin delivery system, the type 1 research advocacy group JDRF dubbed the system an “artificial pancreas.” The term has been eagerly adopted as shorthand for a variety of systems going through testing for safety and efficacy right now.

But Medtronic’s system, along with Insulet’s Omnipod Horizon, are what Levy calls hybrids: They still ask users for carb gram input before meals to calculate how big of a bolus to deliver, for example. Which means carb counting won’t be a thing of the past for a few more years, at least.

That’s part of the reason not everyone likes the term “artificial pancreas.” Boston University biomedical engineer Ed Damiano, PhD, calls his iLet prototype a “bionic pancreas”—reflecting, he says, its fully-automated nature.

Stanford pediatrician Bruce Buckingham, MD, on the other hand, is sticking with “closed-loop system.” He argues that “artificial pancreas” implies something much closer to a replacement organ than the reality, which is a wearable device and sensor plus an infusion set, all of which have to be changed regularly.

And Bigfoot Biomedical, a California-based company founded by four fathers of kids with type 1 diabetes, prefers the term “automated insulin delivery.” “Our system does require user input,” says Bigfoot spokesperson Melissa Lee.

The systems may represent an additional cost for consumers, although manufacturers aren’t sure how much more. Insurance coverage is also an open question until more devices are on the market. “Are insurance companies going to pay for these? We don’t know the answer yet,” Levy says. She argues that the improved blood glucose control the devices provide saves money in the long term by heading off complications.

Doctors say it’s at night when these early systems really shine. While nighttime insulin requirements tend to be more predictable because people don’t eat—and the body is at rest—during sleep, the threat of low blood glucose at night is something that people with diabetes dread. “I spend a lot of time having things beep at night, either because I’m high or low,” says Wesner of her current continuous glucose monitor. She then has to decide whether it’s a false alarm or necessary to treat with insulin or fast-acting carbohydrate. “The thought of having something that would manage that is really appealing.”

The first generation of artificial pancreas devices—such as Medtronic’s 670G—will go a long way toward helping people with diabetes sleep better. The JDRF, which has been lobbying for fast-tracked approval for the systems for years, sees the arrival of the devices as a major win. “People who have participated in artificial pancreas clinical trials have not only attained better overall glucose control but have experienced the relief of sleeping through the night and waking up in the morning with blood glucose levels within target range,” Aaron Kowalski, PhD, JDRF’s chief mission officer, told Diabetes Forecast. “That’s an improvement in quality of life.”

In tests, people with type 1 diabetes using artificial pancreas devices stayed within their desired blood glucose range much more often overall than people who weren’t using the systems. Pilot studies in people with type 2 who depend on insulin show that artificial pancreas devices kept their blood glucose levels within recommended ranges as well.

“This is a major historical milestone,” says Buckingham, who has worked with children as young as 7 to test the Medtronic device and several others. “It gives people more security and will make them sleep a lot easier. Waking up in the 100 to 140 mg/dl range is a pretty good way to start the day.”

So what’s taken so long?

Smartphones and Sensors

Developing safe, reliable algorithms is a huge challenge. It’s also important to remember that a mistake or glitch in the software can have serious or even deadly consequences: Too much insulin can drive blood glucose levels dangerously low, resulting in hypoglycemia, coma, or death.

That’s why developing a system that reliably and safely mimics the body’s own insulin delivery system is fiendishly complicated. Ed Damiano, PhD, should know: He’s been trying to develop one for 17 years, ever since his then11-month-old son, David, was diagnosed with type 1 diabetes.

Almost immediately after his son’s diagnosis, Damiano—now a professor of biomedical engineering at Boston University—dedicated himself to solving the problem. The goal was obvious: an insulin delivery system that could monitor and adjust his son’s glucose levels automatically as they both slept.

How to get there wasn’t as clear. Damiano began by developing algorithms to manage the balance between insulin, glucagon, and blood glucose. It was a tremendously optimistic project to undertake. At the time, there wasn’t an accurate or reliable CGM on the market. Until recently, the glucose readings from most commercially available CGMs could be thrown off by something as simple as Tylenol. That’s not a huge problem when backed up by regular finger sticks, but it’s a major barrier to a true artificial pancreas system that would run without user input.

When Damiano started, a desktop computer was needed to run the programs, which he tested on pigs in a lab. The first tests in people were conducted in hospitals because participants had to be connected to bulky computers.

Fifteen years later, that’s all changed. “The artificial pancreas was really waiting on the mobile phone industry, as well as the sensor,” Damiano says. “Now we have smartphone technology with an app ecosystem that made it possible to just drop these programs on your phone.”

Damiano says he’s on schedule to get FDA approval for his iLet device in early 2019, in time for his son’s sophomore year of college. The system would manage basal insulin like other artificial pancreas devices, but with an added bonus: It won’t require manual bolus doses, instead sensing the post-meal rise in blood glucose and adjusting insulin accordingly.

There’s much work to be done, of course. “It’s still an inelegant system,” Damiano says. “It’s a device you carry with you 24-7. It’s plagued by needing battery power, needing to change infusion sets, cartridges, and sensors. But it’s the best solution we have today. Ultimately, there will be a biological cure for diabetes, and this device will be the bridge to that cure.”

Nuts and Bolts

The artificial pancreas is groundbreaking, but it’s still at the gawky adolescent stage: It depends on current insulin formulations and pump delivery methods. With that in mind, smart people and companies around the globe are focusing on these two areas ripe for advancement.

Better insulin infusion. Artificial pancreas systems rely on infusion sets to ferry insulin into the body, but those come with their own challenges: variations in absorption at different skin sites, tissue damage from poor site rotation, kinks in cannulas, and adhesive failure. Proper insertion technique and site management are crucial—but are subject to human error. And until longer-wear products are a reality, users still need to change their set every few days.

Faster insulin. The rapid-acting liquid insulin analogs currently on the market are speedy. But especially with the absorption issues mentioned above, they can’t quite compete with first- and second-phase insulin release by a healthy pancreas in response to eating.

Different Technologies

Here are some of the different systems researchers hope to roll out in the next few years:

Hybrid Closed-Loop: This setup fully automates basal insulin doses but still requires carb counting  and input from users to confirm correction insulin doses and mealtime boluses. Because such systems are the least ambitious, they’re the first to make it through the Food and Drug Administration’s clearance process. The first device in this category to hit the market is Medtronic’s MiniMed 670G. Also in development: Insulet’s tubeless Omnipod Horizon Automated Glucose Control System, Bigfoot Biomedical’s Bigfoot Smartloop, and Tandem Diabetes Care’s inControl.

Closed-Loop: When they’re commercially available, possibly in early 2019, the algorithms in these devices will be able to sense and bolus for mealtime blood glucose surges and other irregularities by themselves. The Beta Bionics iLet is an example of a closed-loop system.

Dual-Hormone Systems: Damiano and others are working on artificial pancreas systems that can administer both insulin and glucagon, the hormone that tells the liver to release stored glucose into the bloodstream to replenish low blood glucose levels. Ideally, such devices will be able to prevent both blood glucose highs and lows, further reducing the burden of care for people with diabetes. But these are still several years from commercial approval. “A bi-hormonal system is more complicated to develop,” Damiano says, “but it provides a simpler end technology to the user.” One hurdle: There’s no commercially available or FDA-approved stable liquid glucagon—yet.

In the May/June 2017 issue’s “Building a Better Pancreas” article, we incorrectly reported that Alicia Wesner took part in trials for Medtronic’s hybrid artificial pancreas system. She was involved in trials using the Diabetes Assistant artificial pancreas system.



Take the Type 2
Diabetes Risk Test