Artificial Pancreas System Shows Promise in Humans

In the clinical trial at MGH, 11 adults with type 1 diabetes were hooked to an experimental version of the software-controlled, closed-loop system for 27 hours and supplied with three carbohydrate-rich meals.
In the clinical trial at MGH, 11 adults with type 1 diabetes were hooked to an experimental version of the software-controlled, closed-loop system for 27 hours and supplied with three carbohydrate-rich meals.

To maintain safe blood sugar levels, people with type 1 diabetes must submit to a demanding daily regimen that requires frequent monitoring of blood glucose (BG) and dosing of insulin via injection or infusion pump. Even with recently developed continuous glucose monitoring (CGM) technology, the diabetic or caregiver must still double-check BG levels and administer insulin several times a day. Because the process is so labor-intensive, the type 1 diabetic community — nearly two million strong in the U.S. — has long sought an automated, “closed-loop” system that regulates BG with minimal human intervention.

Now investigators at Boston University and Massachusetts General Hospital have conducted the first clinical trial of a closed-loop system prototype developed by Associate Professor Edward R. Damiano (BME) and Senior Research Associate Firas H. El-Khatib (BME). The system uses decision-making software to pump insulin and glucagon (a blood sugar-raising hormone) beneath the skin based on BG readings every five minutes. The team — the first to complete a human trial of a closed-loop system using both insulin and glucagon — described the system’s performance in the April 14 edition of Science Translational Medicine
 
“Over time, our automated system would not only reduce the decision-making load on type 1 diabetes patients and their caregivers, but would also help keep patients’ blood glucose levels within a much healthier range than open-loop systems,” said Damiano. “As a result, they would experience far fewer low blood sugar episodes and avoid chronic high blood sugar levels associated with cardiovascular disease, kidney failure, retinal damage and other long-term complications.”

Based on a series of experiments they conducted on diabetic pigs since 2005, Damiano and El-Khatib’s closed-loop system uses a control algorithm to receive BG data from a standard, FDA-approved CGM sensor every five minutes and dose either lispro, a fast-acting insulin analog (for elevated BG), or glucagon (for low readings) with a standard, FDA-approved insulin infusion pump. In effect, the system works to emulate the endocrine pancreas, continually producing these two hormones to regulate blood sugar.

Emulating the Pancreas
In the clinical trial, the research team studied 11 adults with type 1 diabetes, each hooked to an experimental version of the closed-loop system for 27 hours at MGH and supplied with three carbohydrate-rich meals. Consisting of an intravenous BG monitor, infusion pumps to deliver insulin and glucagon through abdominal tissue just beneath the skin and a laptop containing the control software, the system compared each BG reading with a 100 milligrams/deciliter (mg/dL) “ideal” target, and computed insulin and glucagon doses aimed at minimizing the difference between the reading and the target.

Six of the subjects exhibited a near-normal aggregate mean BG of 140 mg/dL, which is significantly better than the 150-160 range recommended by the American Diabetes Association for people with type 1 diabetes. But the five others experienced episodes of hypoglycemia—low blood sugar caused by too much insulin—that required the subjects to drink orange juice to correct.

The researchers determined that because these five subjects absorbed insulin at a much slower rate than modeled by the software, the system had administered excessive doses of insulin, thereby inducing hypoglycemia. By nearly doubling the expected absorption time in the software, they prevented hypoglycemia in repeat 27-hour experiments in the same participants, and achieved an aggregate mean BG of 166 mg/dL — still close to the ADA’s recommendation.

The Seven-Year Goal
Damiano’s interest in developing an automated, closed-loop control system emerged in 2000 when his then-11-month-old son, David, developed type 1 diabetes.

“When he was a baby, I checked his blood sugar with finger stick measurements 15 times a day,” he recalls. “Two years ago I got him on to a CGM that allows me to put a receiver in my bedroom with alarms to wake me up to give him juice when his BG hits 75, or insulin when it reaches 140. In seven years David goes to college, and he won’t get the kind of control I can provide for him.”

Damiano’s goal is to produce a practical artificial endocrine pancreas for outpatient use by that time. In a follow-up study this spring, he and his collaborators will test a standard portable infusion pump and standard CGM on 36 adult and pediatric subjects for about two days. The study will add exercise sessions and a small pre-meal “priming bolus” of insulin aimed at reducing overall mean BG. Eventually the software, currently stored on a laptop, will be condensed onto an integrated circuit on the pump.

The study was supported by grants from the Juvenile Diabetes Research Foundation, the Wallace Coulter Foundation, the Charlton Fund for Innovative Research in Diabetes and the National Center for Research Resources.