New Diabetes Management Tech Finally Moves Beyond the Bells and Whistles
Users may welcome the update to personalize automated insulin dosing, but they still await a cure.
The Minnesota state fair almost killed Jamie Marshall at age 14. After sampling countless fried foods on a stick, he couldn’t walk. “My friends gave me a hard time,” he says, “but I was probably close to dying.” Within days, he was diagnosed with type 1 diabetes.
Now 28, Marshall lives in Minneapolis, Minnesota. As he recalls that day at the fair, fear and trauma envelope his face. Unfortunately for Marshall and the nearly 1.6 million Americans with type 1 diabetes, a cure remains elusive. But hybrid closed-loop automated insulin delivery (HCL-AID) systems first hit the market in 2017, using algorithms to put the user’s continuous glucose monitor (CGM) in conversation with their insulin pump. They’re imperfect replicas of a healthy pancreas, but marketing teams still frequently call them “artificial pancreas” systems, which leaves a bitter taste in the mouths of many folks with type 1 diabetes, including Marshall. Plus, new technology often issues a learning curve.
For Audrey Leydon, 64, of Highlands Ranch, Colorado, keeping up with the latest tech since her type 1 diagnosis over 40 years ago has been no small task. When she started an HCL-AID system, she stayed on manual mode. “Then, I grew to appreciate what it does for me. It does have bells and whistles that go off all the time, but to me, it’s pretty remarkable.” Still, Leydon says she will always want some element of control over the device, as she trusts herself more. “After all, we humans are pretty spectacular pieces of work.”
“After all, we humans are pretty spectacular pieces of work.”
Researcher Taisa Kushner agrees. For her Ph.D. at the University of Colorado-Boulder, which she defended in July of this year, Kushner helped personalize HCL-AID systems. The algorithm behind currently-marketed systems relies on a model of glucose-insulin dynamics in an “average” person with diabetes, except that average is often made of data from a homogenous group of white men. And Kushner’s irked. “Everyone’s physiology is so different,” she says. Her team’s improved models enable HCL-AID systems to better plan for and respond to a user’s individual insulin needs.
Kushner’s lightning bolt earrings frame her face, which is already electric with excitement, as she likens algorithms to recipes. “They’re instructions for doing a task.” And when layered on top of one another, algorithms can team up to solve simultaneous problems — just as a cookbook can lead to a four-course meal. So, Kushner has used that space between those high-voltage ornaments of hers to create an algorithm that employs the user’s specific glucose-insulin dynamics — captured by numbers that a physician could enter into the device upon setting it up. The effect? Customized dosage.
Kushner uses computer-based simulations to test the algorithm’s performance on data collected from people with type 1 diabetes, rather than testing it on live people directly. Computer-based simulations were developed to replace animal testing for diabetes research in 2009, but the process hasn’t been truly personalized — until now. Testing with models built from individuals’ data means Kushner can verify how well the algorithm performs across infinite scenarios of meals, insulin doses, and glucose levels. “Human clinical trials can’t anticipate as many scenarios,” she says. “Rigorous computer-based trials help identify ‘bugs,’ and avoid recalls down the road.”
“Human clinical trials can’t anticipate as many scenarios. Rigorous computer-based trials help identify ‘bugs’, and avoid recalls down the road.”
So far, Kushner has tested her algorithm, which improves how the current system is personalized for users, on data collected from 49 people. Her models suggest the new algorithm keeps users in the healthy glucose range 83 percent more often than the first system to hit the market in 2017, the Medtronic MiniMed.
She rips through her PowerPoint, stopping abruptly on the slide with her data. With the current method for “personalizing” the system, blood sugar levels of those 49 simulated people with type 1 diabetes on the Medtronic MiniMed are likely to end up an undesirable, turbulent mess. Meanwhile, simulations using the new algorithm for adjusting the system reveal blood sugar levels locked into a healthy zone. It’s like looking at before and after photos of a poodle that went to the groomer — Kushner has reigned the erraticism of the Medtronic MiniMed way in.
Back in 2017, Marshall tried the Medtronic MiniMed insulin pump, paired with the Enlite CGM, after years of finger pricking and using an insulin pen. His experience confirmed Kushner’s data — he found the system unreliable and uncomfortable, so he ditched it and remained skeptical of these systems. But he was pleasantly surprised by the Dexcom CGM and Tandem insulin pump, which the FDA approved in January 2020.
Now, Marshall admits he has noticed “dramatically better” blood sugar levels and feels some liberation from the “decision-making fatigue” that comes with managing diabetes. But after years of having diabetes tech oversold to him and other users, Marshall hesitates to celebrate. “For decades, we have been told that we are ten years from a cure for diabetes,” he says, “and it feels like we are pretty damn close to reaching the limit of the technology on this.”
“For decades, we have been told that we are ten years from a cure for diabetes, and it feels like we are pretty damn close to reaching the limit of the technology on this.”
For now, the system he uses offers a new blood sugar number on his pump, phone, and smartwatch every five minutes. It also buzzes if he is out of a healthy range. Michael Mayberry, 42, of Chicago, Illinois, uses it, too: “I am probably the healthiest I have been in 10 years, and part of that is because of this technology.” He was diagnosed at 23, but because the systems are designed to alert users, Mayberry says sleeping is the only break he gets from thinking about his diabetes. Clearly, these advancements come at a cost. But at the same time, both users are grateful to sleep through the night without having to check their levels. And Marshall says he wakes up in the healthy zone much more frequently than he ever has in his life.
Kushner’s research would not eliminate the system’s consistent alerts, but rather make the process more personalized and accurate. New models she’s developed could also help current and future systems anticipate a user’s blood sugar levels for up to four hours, which would make it nearly 6–8 times more perceptive than the systems Marshall, Mayberry, and Leydon currently use. But Kushner knows she must re-build trust broken by the first few iterations of HCL-AID systems. So, she’s covering her bases, running more tests with data from different sets of people, and working with physicians and users to better explain how these algorithms work.
Keeping people with diabetes in the healthy zone 83 percent more often is not a cure, but it might be the closest thing available right now. “I want to meet people where they are. To me, computer science can bridge an important gap in bringing people the care they need,” says Kushner. “Care that works.” But the way manufacturers market these products leaves an impact, and companies should be accountable for that — diabetes tech users are not so easily bamboozled. After all, they’ve spent their lives adapting to new technology and management systems.
“I want to meet people where they are. To me, computer science can bridge an important gap in bringing people the care they need. Care that works.”
Over the years, Marshall has become an expert at managing his diabetes. He even returns to the Minnesota state fair each summer. But despite the technological advancements — and personal maturity — that helped him get here, he says that when it comes to Kushner’s forthcoming model, he’ll believe it when he sees it get through human trials.
Still, Marshall marvels at how “not that long ago, if you had diabetes, you just died.” And he does see a world where tech and a cure intersect — an implanted organ, fitted with some type of CGM, that could adjust and react instantly, without user management. Something like a pacemaker for diabetes. “I think you could call that an artificial pancreas. That would be the dream. That would basically end diabetes.”
