As a result, he says, patients may experience many hours of very high or low sugar levels before returning to a normal state.
The UD system more precisely controls blood sugar by constantly predicting the patient's need for insulin. Based on a mathematical model of the human glucose-insulin system, the algorithms analyze data from past events to forecast future insulin requirements.
And, because the algorithms are "linear," or simplified to approximate the function of the gut, the pancreas and other portions of body systems, they could be maintained on a tiny computer chip, Doyle says.
His approach, based on "model predictive control with state estimation" (MPCSE) algorithms, effectively reduced peak glucose levels by 44 percent, in computer simulations, compared to algorithms published in scholarly literature.
The system also reduced by 80 percent the "overshoot," or degree to which blood-sugar levels rose above a targeted range of 81 mg/dl, compared to patients with uncontrolled diabetes, Parker says. Small delays in receiving data from glucose sensors didn't seem to impair the performance of the UD system, which demonstrated a settling time of about 4.5 hours.
Doyle predicts that "smart" implantable insulin pumps won't become available to patients for another three to five years, even with FDA approval. But, he says, research presented at the AAMI meeting confirms the viability of this promising new technology.
Someday, he says: "Automatically controlled, implantable insulin delivery systems will no longer be science fiction."