Artificial pancreas passes first test in patients with type 1 diabetes

The first evaluation of a fully automated artificial pancreas in 10 patients shows that the system can safely regulate glycemia in patients with type 1 diabetes even following a meal challenge, without prior meal information, said Howard Zisser, MD.
     Short of a physiologic cure, the best hope for the management of type 1 diabetes may lie in continuous glucose sensing and automated insulin delivery to simulate the glucose control loop that is automatically regulated in the nondiabetic person. The fully automated closed-loop system combines 2 subcutaneous continuous glucose monitors and 3 continuous subcutaneous insulin infusion pumps with a sophisticated control algorithm. All of the insulin calculations and delivery are done automatically.
     First, Zisser and colleagues at Sansum Diabetes Research Institute in Santa Barbara, California, generated a personalized model and insulin-on-board control algorithm from 3 days of data, including meal information, from subjects with outpatient insulin pumps and sensors. During the actual evaluation of the system, 10 patients who had type 1 diabetes for more than 20 years were studied with the artificial pancreas for 8 to 10 hours each, during which time 1 or 2 meals consisting of 35 g of carbohydrate were consumed. The closed loop was initiated at the glucose concentration at the time the patients came in for the study. The aim was to determine if the system could maintain euglycemia, defined as a target glucose level of 110 mg/dL (±30 mg/dL), and then to overcome the meal.
     “What we wanted to do with this project was to run a fully automated pancreas where we brought the patient in, flipped the switch, and stood back and didn’t do anything else but observe from a safety standpoint,” said Zisser, Director of Clinical Research and Technology at Sansum, and Adjunct Professor, Department of Chemical Engineering at the University of California, Santa Barbara. “We used small meals…just to prove the concept.”
     The controller successfully brought subjects back to the euglycemic range, he said. The system recognized all of the unannounced meals and gave appropriate meal boluses of insulin. The average percent time in the target glucose range (80 to 180 mg/dL) was 77%, with one episode of mild hypoglycemia.
     “We met our goals and did not have to interrupt delivery or intervene in any of these patients,” said Zisser.