This article is part of “Innovations In: Type 1 Diabetes,” an editorially independent special report that was produced with financial support from Vertex.
Edward Damiano carved his life into precise 90-minute intervals. In 2000 his then 11-month-old son, David, developed type 1 diabetes when an autoimmune response in his tiny body attacked the beta cells in his pancreas, which manufacture and secrete the hormone insulin. No beta cells meant no insulin. If Damiano or his wife, Toby Milgrome, a pediatrician, didn’t give David injections of the hormone, the baby’s cells could no longer use glucose, a vital energy source. Within hours David could be in a coma or dead. With their son’s pancreas no longer functioning, Damiano and Milgrome had to take over the organ’s work by measuring every gram of carbohydrate David ate and dosing the right amount of insulin. To ensure his son didn’t receive too much or too little of the lifesaving medicine, Damiano checked David’s blood glucose every hour and a half, rain or shine, day or night.
When David first became ill, Damiano, a bioengineer at Boston University, kicked it old school to monitor his son, accumulating a set of spiral-bound notebooks in which he or Milgrome logged every drop of insulin and morsel of food. Each day the couple flipped over a new page and started again, building a complex calculus of sugar grams per sip of Juicy Juice measured against units of insulin and blood glucose that would allow David to thrive. Even for a physician-scientist team, the work was grueling and relentless. Damiano also knew that he was one of the lucky ones.
On supporting science journalism
If you’re enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
“The hardest part of diabetes management is making all those decisions. And if you give yourself a little too much insulin, you could end up in the ICU,” Damiano says.
While Damiano waited for David’s blood glucose to tick up or edge down, his engineering brain crunched the problem. There had to be a better way, he thought. The more widespread use of commercially available continuous glucose monitors (CGMs) by 2004 meant someone with diabetes could get a minute-by-minute measurement of their blood glucose without turning their fingers into pincushions. Users of insulin pumps, which deliver doses of insulin for meals and steady doses in the background, could tether their CGMs to their pumps and start to automate some parts of their insulin delivery. In September 2016 Medtronic first paired CGMs and insulin pumps in what scientists call a hybrid closed-loop system that could automatically adjust insulin delivery based on a person’s blood glucose levels, except at mealtimes, which still needed to be manually programmed. The devices were a major breakthrough, but many people with diabetes still found managing blood glucose to be a continuous struggle.
“Management of type 1 diabetes is like driving a car 24/7 on a curvy mountain road with no breaks even when you’re asleep. So if you could take some of that burden off, it could make a huge difference,” says biotechnology entrepreneur Bryan Mazlish, co-founder of biopharmaceutical company Surf Bio and of Bigfoot Biomedical, which produces tools to help people manage diabetes.
In 2015 Damiano co-founded Beta Bionics to see whether he could ease the burden for his son and others with type 1 diabetes by creating an artificial pancreas—a completely automated, closed-loop system that functioned like a real, healthy pancreas. In May 2023 the U.S. Food and Drug Administration approved the company’s iLet device, an adaptive closed-loop system that requires only a patient’s body weight to start insulin delivery. Instead of counting carbs, users can input “small,” “medium” or “large” for meal sizes. In January 2025 Beta Bionics went public.
The company is one of several outfits currently inching toward the development of a fully automated insulin-delivery system, considered an artificial pancreas, that can track and alter blood glucose levels through insulin, much like the body does on its own in healthy individuals. Clearing the final hurdles for regulatory approval won’t be easy. Successfully programming a device to make the same minute-by-minute adjustments as the human body will require a close marriage of biology and technology. Until now, most advances in type 1 diabetes tech have allowed the rich to get richer, so to speak, says Steven Russell, an endocrinologist at Massachusetts General Hospital and chief medical officer at Beta Bionics. People who already had healthy control over blood glucose could make improvements, but those who were struggling often continued to struggle.
“We can allow people who hadn’t traditionally had good glycemic control to achieve it,” Russell says. “That means almost everybody can get good glucose control on it regardless of where they’re coming from.” Although the iLet costs less than other closed-loop systems, retailing at $3,500, those without health coverage may still struggle to afford the device and CGM supplies it requires.
As a young endocrinologist in the 1980s, David Klonoff was typically greeted with a full waiting room at the diabetes clinic at the University of California, San Francisco. Some had seeing-eye dogs because years of too-high blood glucose had caused retinopathy, which, in some extreme cases, caused blindness. Others carried scars and marks on their forearms from dialysis treatments for kidney disease, which can result from cumulative damage to fragile blood vessels in the kidney from living with elevated blood glucose levels. Some had a fungal infection called mucormycosis, also known as black fungus, which can infect the nasal passages, sinuses, lungs and skin of people who have chronic high blood glucose or are immunocompromised. Severe cases can cause disfigurement and death. Still others bore amputations after diabetes complications led to peripheral neuropathy, a condition that blunts sensation in the extremities. Without nerve sensation, the patients weren’t aware of festering ulcers and infections, which therefore went untreated until amputation was the only option. Given that Klonoff’s patients had to guess at the appropriate insulin doses, perhaps it wasn’t so surprising. Klonoff likens the technology of the time to the Wright brothers’ first aircraft prototypes as compared with today’s jet engines.
Most of these patients devoted hours every day to managing their condition. The problem wasn’t a lack of effort but rather a lack of appropriate tools. With no way to test their blood glucose levels, people with diabetes had to make an educated guess about how much biosynthetic insulin to inject. And because too much insulin could be fatal, they had to err on the side of too little insulin and adjust on the fly. Not surprisingly, blood glucose control was often suboptimal. The advent of home-based finger-prick testing in the 1980s and faster-acting biosynthetic insulins made a tremendous difference in reducing complications, but Russell and other doctors still saw great disparities among their patients. Those lacking good medical literacy and adequate time and resources to devote to managing their disease continued to struggle.
Although the development of automated insulin pumps helped to remove some of the burden, many patients found simply staying alive to be a permanent, unpaid, full-time job.
“It’s impossible to be awake and on top of this optimally—understanding the physiology of insulin action and insulin duration and the impact of specific meals and foods. It’s not feasible,” says Carol Levy, director of the Mount Sinai Diabetes Center in New York City. And she would know—Levy has lived with type 1 diabetes for more than 50 years.
Mazlish knew what to expect when his son turned out to have type 1 diabetes. After all, his wife also had the disease, and she began to educate her husband as soon as their son was diagnosed. Mazlish watched as his wife programmed her insulin pump to deliver both dribbles of basal insulin throughout the day and a large bolus with meals. If her blood glucose dropped too low, she had to eat something carbohydrate-laden to raise it. To Mazlish, a finance quant turned life sciences entrepreneur, the work seemed amenable to automation with a computer program.
To help his son, Mazlish wrote a bespoke algorithm that would automatically adjust the insulin delivered by the pump based on the child’s blood glucose levels. “It was really life-changing. We could live much more freely, and it gave peace of mind to all of us,” Mazlish says.
Other tech-savvy patients were innovating in similar ways, creating computer codes and programs to automate insulin delivery and ease their own burden. These efforts demonstrated that such an approach could work, but not everyone in the field was sure that an algorithm could accurately adjust insulin delivery based on glucose levels, says Boris Kovatchev, an engineer and director of the University of Virginia Center for Diabetes Technology.
That is until widespread use of CGMs by people with type 1 diabetes in the mid-2000s and the subsequent integration with insulin pumps created the first realistic hope that endeavors by Kovatchev, Mazlish, and other biomedical technorati could result in a usable device.
“Management of type 1 diabetes is like driving a car 24/7 on a curvy mountain road with no breaks even when you’re asleep.” —Bryan Mazlish Surf Bio and Bigfoot Biomedical
In December 2005, scientists, engineers and physicians gathered on the campus of the National Institutes of Health in Bethesda, Md., for the first-ever workshop about the prospects of building an artificial pancreas. Some of the attendees weren’t convinced the effort was feasible, Kovatchev says. Physicians and engineers alike fretted about the high stakes if a system malfunctioned. Others preferred to focus their attention on creating a cure for type 1 diabetes, not just building more bells and whistles for an existing treatment.
Kovatchev, however, thought that people with diabetes could benefit from an artificial pancreas and that the potential for error would be significantly less than what resulted from patient guesswork.
The biggest challenge the researchers realized was building a set of algorithms that were sensitive enough to allow minute adjustments to insulin doses and flexible enough to work for millions of patients. It was a challenge that could be overcome only with gobs of data, something neither Kovatchev nor anyone else had.
Like Damiano, Kovatchev approached the diabetes problem from a perspective other than a physician’s—in his case, through mathematical modeling. He needed to figure out how to replicate biology’s intricacies in an automated device that could be used by millions. That was no easy task. For one, everyone’s body responds to insulin slightly differently. “The variation is huge. Everything has to be individualized, and that was a major problem throughout the years,” Kovatchev says.
What’s more, commercially available insulins don’t work as quickly as the hormone naturally produced by the pancreas. Kovatchev and other engineers would need their algorithms to account for insulin already at work in the bloodstream as well as administered hormone that had not yet started to lower blood glucose.
After the first pancreas workshop in December 2005, Kovatchev helped to build the informational foundation for such an undertaking. First he got detailed data on glucose metabolism in healthy people to understand how the body processes glucose from food, how recently consumed sugars interact with the sugars already in the body, and how glucose levels change after meals.
Scientists continued the effort, eventually accumulating data from upward of 7,000 healthy individuals. It was enough to garner FDA approval of the UVA/PADOVA Type 1 Diabetes Simulator in 2008 as an alternative to animal testing in certain preclinical trials and artificial pancreases. An updated version was approved in 2013 with even more parameters to capture the complexity of glucose biology.
Other advances were happening at the same time. In 2008 a team of researchers at the University of California, Santa Barbara, led by chemical engineer Francis Doyle, reported that it had built the first prototype device that would allow CGMs to communicate with insulin pumps, opening the door for systems that could automatically adjust insulin dosing outside of meals. These insulin pumps and CGMs have had a huge positive impact on quality of life, says Jonathan Rosen, director of research at the nonprofit Breakthrough T1D. “Rates of long-term complications have gone down over the years thanks to improved blood sugar control,” he says.
But these achievements were still not considered artificial pancreases: even the most sophisticated devices still required users to manually administer extra insulin to account for meals. And it was these interactions that continued to trip up many users, Russell says. Humans aren’t very good at estimating meal components down to the gram, and measuring every mouthful of food requires intense devotion. In addition, the pumps didn’t function correctly right out of the box; specialized endocrinologists had to help program the devices and make adjustments every few months.
At Boston University, Damiano recognized similar issues as his son grew up and began taking more responsibility for managing his disease. Eliminating the need for manual adjustments—whether for meals or for any other changes during the day—would give people with diabetes a tremendous sense of freedom. Even better would be to forgo the initial programming and adjustments by a physician. By the mid-2010s artificial intelligence and machine-learning algorithms began providing the solutions that Kovatchev and Damiano were looking for.
“I want a system that is democratizing,” Damiano says. Everybody, he emphasizes, should have the ability to access this technology.
For iLet wearers, these solutions mean they can begin using their device immediately. The only information they need to input is body weight. From there, data transfer from previous pumps and the iLet’s own software feed into the AI-driven software to accurately control blood glucose in patients, with only brief meal inputs required by users. The company has reported that within their first year of using iLet, patients have been able to better manage their blood glucose.
Adding neural networks and other AI technology such as digital twins (digital models of real objects, places or people that can be used to simulate responses to a variety of conditions) allowed Kovatchev to re-create CGM readouts from the Diabetes Control and Complications Trial, launched in 1993. These simulations, published in March 2025, were able to accurately predict different diabetes complications and the amount of time someone’s glucose was in a healthy range. When Kovatchev allowed patients to experiment with their digital twin, they could see the likely outcome before it happened. “That was very educational. People loved it,” he says. But Kovatchev also cautions that the safety and security of an artificial pancreas continue to be issues. “These algorithms are black boxes. Nobody really knows what’s going on inside and how they react to different situations. It’s critical to have constraints so it doesn’t do something stupid or dangerous,” he asserts.
To Damiano, the iLet is a game changer, but it’s not the end of the road. He envisions a fully closed-loop system that doesn’t require any user input at all, as well as a dual-hormone device that can administer the blood-glucose-raising hormone glucagon alongside insulin. But the field now is worlds away from those midnight pencil scribbles in a spiral notebook that kept his son alive until better technology came along.
