Skip to Main Content

The artificial pancreas

Yale Medicine Magazine, 2016 - Winter


The artificial pancreas is the marriage of a sensor that monitors glucose levels and a pump that sends insulin into the bloodstream as needed.

Last May, 15-year-old Jacob Conte was scaling a rock wall at City Climb Gym in New Haven. Hand over hand, grabbing onto holds protruding from the wall, he made his way up the artificial cliff. Underneath his clothing—indeed, under the skin on his abdomen—he wore a small sensor that transmitted his blood glucose level readings every five minutes to a smartphone device in his backpack. Conte has type 1 diabetes and requires regular infusions of insulin. The smartphone told an insulin pump affixed to his belt when and how much insulin to deliver. The pump—smaller than a smartphone—injects insulin under the skin of the abdomen through a short thin tube.

Maintaining his blood sugar control requires constant vigilance and dozens of decisions each day. When Conte plays football or snowboards, he has to avoid a drop in blood sugar that could make him dizzy or pass out. Before eating, he tests his blood sugar and takes insulin, but if food is delayed, the insulin might kick in too soon, causing his blood sugar to drop. Nighttime—when more than half of hypoglycemic (low blood sugar) emergencies occur—is especially perilous. If he wakes up with symptoms of hypoglycemia, he tests his blood sugar and eats a snack. “It’s kind of a burden,” he said. “Everything in my life revolves around it.”

Handling his diabetes is about to change. The climbing session was part of a four-day clinical trial in which Conte and four other teenagers with type 1 diabetes took those sensors and pumps—which when combined with a dosing algorithm are known as an “artificial pancreas”—for a test run. Rather than a biomedically engineered organ made of tissue, the artificial pancreas is a system of devices and software that transmits his blood glucose level readings to the device that controls insulin delivery. Wearing the sensor and pump doesn’t interfere with Conte’s normal activities—including climbing the rock wall, playing football, or sleeping—and the devices are worn under his clothing, so nobody knows they’re there. Yale researchers have been studying the artificial pancreas for the last decade, but this spring was the first time it was tested in pediatric patients outside the hospital—here and at Stanford University and the Barbara Davis Center for Childhood Diabetes in Denver.

Closing the loop

Most of the 35 million people around the world who have type 1 diabetes inject themselves with insulin several times a day or use an insulin pump. The pump, which is programmed to deliver tiny doses of insulin, has advanced diabetes treatment. It sometimes works with a continuous glucose monitor (CGM) that has a sensor like the one Conte wore. Both devices are approved by the Food and Drug Administration (FDA) but work independently of each other. The artificial pancreas closes the loop between the two (it’s sometimes referred to as a closed-loop system) by allowing the pump to adjust insulin delivery every five minutes in response to the body’s glucose levels. “The ability of the system to self-adjust automatically while patients go about their daily lives would really be transformative in the lives of people with diabetes,” said Stuart Weinzimer, M.D., professor of pediatrics at the School of Medicine. Weinzimer and his colleagues have been working on an artificial pancreas for the last 10 years, with support from the Yale Center for Clinical Investigation. The device tested last spring, developed by Medtronic, is being tested at 10 centers (of which Yale is one) in a phase III clinical trial that is expected to lead to FDA approval in 2017.

The path to the artificial pancreas began in the late 1970s, when researchers found better ways to monitor and control blood glucose levels. With the advent in 1977 of the hemoglobin A1C test, known simply as A1C, doctors could analyze blood sugar control over a period of two to three months, providing a longer view than isolated blood glucose readings taken during clinic visits. The insulin pump—first tested in a clinical trial at Yale in 1979—more closely resembled the way the pancreas produces insulin by delivering small doses throughout the day with larger doses at meals. By the early 1980s, home blood glucose monitoring had replaced urine testing, and synthetic human insulin began to replace animal insulin. But diabetes remained difficult to manage. “In those days, when we started to be able to measure more accurately how well controlled our patients were, most of the numbers would have been viewed today as totally unacceptable,” said Robert Sherwin, M.D., the C.N.H. Long Professor of Medicine, who was on the team that first tested the insulin pump. The pump, which gave patients better glycemic control, gained traction in the early 1990s when the landmark Diabetes Control and Complications Trial showed that controlling blood sugar levels reduced such diabetes-related complications as blindness, kidney failure, and neuropathy. The next step was to replace a finger stick every few hours with a sensor that would measure glucose continuously. “As long as insulin delivery was not linked to changes in blood sugar, no regimen would be perfect,” said William Tamborlane, M.D., professor of pediatrics (endocrinology), another member of the Yale insulin pump team.

Sherwin tried to interest medical device companies in developing a sensor, but they weren’t ready to make a financial commitment. Eventually Tamborlane, who was also keen to develop a sensor, began working with Medtronic to test a CGM device that received FDA approval in 1999. In 2002, he recruited Weinzimer to Yale to look at the possibility of combining the sensor with a pump to develop an artificial pancreas.

A study launches a project

While their experience with the pump gave Yale researchers a head start, other centers were also interested in developing an artificial pancreas. In 2006, researchers at the University of California, Los Angeles, published the first study that combined sensor data with an algorithm to adjust insulin infusion automatically. However, there was an inherent delay in insulin delivery: while insulin-producing cells in the pancreas read glucose levels every few minutes and secrete insulin directly into the bloodstream, the probe and the pump work via the fluid surrounding tissue, so there is a delay in both reading blood sugar levels and delivering insulin into the bloodstream. Still, the study, which showed that the system could work in adults, piqued the interest of the Juvenile Diabetes Research Foundation (JDRF), which then launched the Artificial Pancreas Project. The project had two arms: a large clinical trial of CGMs co-chaired by Tamborlane, which showed that the devices improved blood glucose control; and another, involving Yale and four other academic sites, to develop an artificial pancreas system.

Type 1 diabetes is typically diagnosed in children, so the next step was to study a closed-loop system in pediatric patients. In 2008, Weinzimer published the results of a study involving 17 children who used an artificial pancreas in the hospital. A closed-loop system, the study showed, keeps blood sugar stable overnight but can’t deliver insulin fast enough at mealtime. The solution was a manually administered extra dose at meals, a concept that is incorporated in the Medtronic device currently being tested.

Weinzimer presented the findings of the pediatric study at the national JDRF conference. “People were crying,” he said. The artificial pancreas offered hope to parents who stay up at night worrying about hypoglycemia, which can lead to unconsciousness and even death.

Compared to today’s artificial pancreas systems—several are in development, but Medtronic’s is furthest along—the early systems were cumbersome. The sensors were much larger, and a radio transmitter had to be taped to the body and connected to a receiver plugged into a laptop. Today, the components are compact and the systems use wireless technology to transmit sensor readings.

A need to work together

Research is often a collaborative effort, and nowhere is this teamwork more evident than in the development of the artificial pancreas. “At first, we all did our own thing,” said Weinzimer. “Now we’re realizing that in order to demonstrate the safety and efficacy of these devices on a large scale, we need to work together.” Researchers at the University of Virginia; Boston University; the University of California, Santa Barbara; the University of Cambridge; Schneider Children’s Medical Center of Israel, and elsewhere are testing devices and designing sophisticated algorithms to fine-tune different systems. Yale researchers have often been at the forefront of these efforts. Weinzimer showed that the artificial pancreas effectively controls nighttime hypoglycemia, while Jennifer Sherr, M.D., assistant professor of pediatrics (endocrinology), is looking for ways to increase the system’s effectiveness at mealtimes. Says Francine Kaufman, M.D., Medtronic’s chief medical officer and vice president for global medical, clinical, and health affairs, “We rely on an institution like Yale with the capability of their investigators, which is immense, to be a push-and-pull with us.”

The artificial pancreas is based on two devices that already existed, yet at times, progress has been frustratingly slow. Resolving safety issues—such as delivering too much or too little insulin—was a major hurdle. Now increasingly tech-savvy patients are increasing the pressure to bring a system to the market. In fact, the ability to tap into sensor data remotely began with a group of diabetes patients and family members who developed a software program to hack into sensor data via a smartphone or computer.

There is also a pressing need to improve diabetes treatment despite the advances over the last 25 years. The recommended target hemoglobin A1C level is less than 7 percent for adults, and less than 7.5 percent for those under 19, according to the American Diabetes Association. Yet a recent study by the T1D Exchange, a network of more than 70 clinics dedicated to type 1 diabetes treatment and research, showed that the average adult A1C is 8.4 percent, with adolescents averaging 9 percent.

In clinical trials, the artificial pancreas helps patients manage their blood sugar better with less effort. “He had the best control while he was on this closed-loop system than he’s had in the past nine months,” said Nicole Liedke, whose 15-year-old son was part of last spring’s clinical trial.

When the first iteration of the artificial pancreas hits the market, the device won’t be totally automatic. Patients will still have to instruct the pump to provide a dose of insulin before meals to maintain optimal blood sugar control, and the sensor has to be recalibrated twice a day. But for patients like Jacob Conte, letting go of the reins a bit, especially at night, will be life-changing. His mother, Joanne, who also has diabetes, was with him during the clinical trial. “He’s making history,” she said, as tears welled up in her eyes. “For him and for our family, this was an unexpected opportunity of a lifetime.”

Previous Article
From raw idea to finished product
Next Article
If you have a lemon, make lemonade