For people living with type 1 diabetes, it pays to be good at math. Keeping blood glucose in a healthy range requires a constant stream of calculations: counting the carbs in each meal, and then finding the right dose of insulin to balance them out. But new clinical trial results from an automated insulin delivery system from Beta Bionics suggest that computation can be reliably outsourced to a machine — potentially putting better blood glucose control in the hands of more patients.
In the last several years, automated insulin delivery has helped many patients with type 1 diabetes avoid harmful dips and surges in their blood sugar. Sometimes called an ‘artificial pancreas,’ the systems, aided by a glucose sensor, algorithmically determine insulin doses that are delivered continuously through a pump. But the systems aren’t perfectly autonomous: Users still have to manually input the carb counts of their meals before eating, so the system can deliver the right dose of medication.
Beta Bionics’ system, the iLet bionic pancreas, takes one more step toward full automation. Instead of reporting carb counts, users input whether an upcoming meal has more, less, or the same number of carbs than normal. “Then the system learns what they mean by that, how much insulin they need, and it will automatically give a portion of that insulin upfront in response to that announcement,” said Massachusetts General Hospital’s Steven Russell, who co-developed the system before it was licensed to Beta Bionics.
On Wednesday, results from the device’s pivotal trial were published in the New England Journal of Medicine, following the company’s submission to the Food and Drug Administration earlier this year. Participants using the iLet over 13 weeks saw improvements in blood glucose, with hemoglobin A1Cs that dropped from 7.9% to 7.3%, while participants receiving standard care saw their A1Cs remain steady at 7.7%. The time iLet users spent in a target blood glucose range also increased by 11% over the control group.
Those results alone mirror — but don’t exceed — the performance of existing automated insulin delivery systems. But by making the technology even easier to use, the device’s creators hope that the iLet will eventually complement existing systems from Medtronic, Insulet, and Tandem Diabetes, which are currently used by roughly 20% of people with type 1 diabetes.
Jen Sherr, a pediatric endocrinologist at the Yale School of Medicine, said the impact of that simplified reporting could be significant. “As a person living with diabetes and having the privilege to treat individuals with diabetes, we’re constantly thinking about: What exercise did I just do? What stress am I facing? Am I feeling ill? What are the carbs I’m going to eat? How do those carbs impact me? What type of fat was in that food?” said Sherr, who wrote an accompanying editorial about the trial in NEJM.
“I think about my diabetes 24/7, and I think these systems just allow us to offset some of that,” she added.
If authorized, the iLet would also be the only system that determines every dose of insulin algorithmically, without the ability for users to modify their own doses. “That’s very, very different than all those other systems,” said Russell, who directed the iLet pivotal trial. “And the hope is that it will make it possible for it to be used by a broader swath of the population with type 1 diabetes.”
It’s also why the system won’t be a fit for every person with diabetes. Some patients will be uncomfortable with ceding control to a computer. And many participants who came into the trial with relatively good control of their diabetes, with A1Cs below 7%, actually saw their numbers increase slightly when they used the iLet. The question, said Sherr, is how those users feel about the tradeoff: Is it more important to regain some time and quality of life, or to get their A1C as low as possible?
For patients who have so far been taken out of consideration for automated insulin delivery, though, that trade-off may be far simpler.
Existing systems are less likely to be used by patients with especially poor blood glucose control, and those whose care is managed by a primary care physician. That’s because their setup requires the time and expertise to calculate basal insulin rates, insulin-to-carbohydrate ratios, and insulin sensitivity — and providers have to trust that a patient or their family can use the device successfully before they agree to prescribe it. “We fear putting patients on advanced technologies when people can’t even prick their fingers,” said Nelly Mauras, the trial’s investigator at Nemours Children’s Health in Jacksonville, Fla.
But the iLet’s initial insulin settings are calculated using the patient’s weight alone — widening the range of possible prescribers. “You don’t need to be a board-certified endocrinologist or diabetologist to get someone on this,” said Bruce Buckingham, a pediatric endocrinologist at Stanford University School of Medicine who has been an investigator on many clinical trials for automated insulin delivery systems, including this one.
And the system’s success in a diverse group of diabetes patients could make providers more comfortable prescribing to those they previously might have ruled out. The iLet trial enrolled young and old participants regardless of their current form of insulin delivery, without a cap on A1C, and with an emphasis on socioeconomic and educational diversity. Hispanic and non-Hispanic Black participants each made up 10% of the study population, with 74% identifying as non-Hispanic white; 45% were female.
“We deliberately went after a more representative cohort of the real world,” said Ed Damiano, who developed the system with Russell and co-founded Beta Bionics. Historically, that hasn’t been the case — in studies of other automated systems, Buckingham said many participants were “already extremely interested in the technology.”
Keisha Bird, lead study coordinator at Nemours Children’s Health in Florida, saw the benefits across multiple patient groups. Typically, when she starts patients on insulin pumps and injections, she sees a “huge learning curve.” But with the iLet, “that learning curve was decreased by far.”
“It didn’t matter what background, what device they came from. Once we set it up and put in their weight, it was pretty much like they were on autopilot,” she said.
Not full autopilot, though. Patients, providers, and device developers still want to see an insulin delivery system that closes the loop between blood glucose readings and insulin delivery with no input from the user at all. Buckingham said that may come in the next aggressive generation of device development. Already, automatic insulin delivery has improved by leaps and bounds over its first iterations — cutting down on overnight alarms that interrupt sleep, as well as improving automated correction doses of insulin when blood glucose values stray too far out of the target range. “In this line of progression,” said Buckingham, “we’re going to see more and more systems that will ask less and less.”