San Francisco, California
London, United Kingdom
San Fernando Valley Section
Engineering in Medicine and Biology Society Chapter
Fraser Cameron, Ph.D.
Postdoctoral Research Fellow, Rensselaer Polytechnic Institute
Clinical Trials of a Closed-Loop Artificial Pancreas with Large Unannounced Meals
Closed-loop control of blood glucose (BG) levels in people with type 1 diabetes can reduce patient burden and the incidence of complications, particularly if meals do not need to be announced. We tested a multiple model probabilistic predictive controller (MMPPC) on four preliminary patients, revised it and tested six primary patients. Each admission lasted for 32 hours with five unannounced meals containing 1 g/kg of carbohydrate (CHO).
The closed-loop therapy used an Abbott Navigator CGM and Insulet insulin pump with the MMPPC implemented through the UCSB artificial pancreas system. Therapy began at 9 AM with unannounced meals at 9 AM, 1 PM, 5:30 PM, and 9 AM and 1 PM the next day. The patients had a mean (±SD) HbA1C of 7.3±0.6%, age of 28±5 years, total daily dose of 43±13 U, and weight of 74±13 kg. The controller was initialized only with the patient’s total daily dose and daily basal pattern.
The MMPPC algorithm explicitly estimates and predicts BG uncertainty stemming from past and potential future meals, endogenous glucose production, and insulin sensitivity. Insulin boluses are calculated to lower predicted BG levels until there is a roughly 3% risk of BG levels below 80 mg/dl. At night, the MMPPC targets a BG level of 100 mg/dl with attenuated control providing smooth corrections.
On a 24-hour basis, the primary patients had mean reference/CGM values of 161/142 mg/dl, with 63/78% of time spent between 70 and 180 mg/dl. Two preventable system failures led to a manual bolus (per subject settings) and an underestimation of active insulin which subsequently required CHO intervention for hypoglycemia. One other CHO treatment was given for a nocturnal glucose of 66 mg/dl with a rate of change of -0.25 mg/dl per min. For the 30 unannounced meals the mean pre-meal, post-meal maximum, and 3-hour post-meal values were 139/132, 223/208, and 168/156 mg/dl respectively.
The MMPPC was tested in-clinic against repeated, large, unannounced meals and maintained good control overnight and during meals.
Fraser Cameron received a Ph.D. from the Department of Aeronautics and Astronautics at Stanford University in 2010. He received a Master's from the same department and a Bachelor's of Applied Science from the University of Waterloo in 2002. He is currently a Juvenile Diabetes Research Foundation Postdoctoral Research Fellow at Rensselaer Polytechnic Institute in Troy, NY. His research is also supported by the NIH/NIDDK. His research includes fault detection and control as applied to regulating glucose levels in people with T1DM. Clinical aspects of this research are carried out in conjunction with Bruce Buckingham at Stanford University and the Barbara Davis Center for Diabetes in Colorado.
Note: An NDA is required to enter the Medtronic facility. Please contact the organizer to obtain a copy to bring to the event.