Last year the FDA approved the first artificial pancreas for general use. The technology is designed to automatically monitor and inject insulin as needed, in patients with type 1 diabetes. Now a team at Harvard has developed an even more sophisticated system that connects to a patient's smartphone and learns their habits, enabling glucose levels to be kept within a certain healthy range.

Previous treatments for type 1 diabetes have generally involved a patient monitoring their blood glucose levels and then manually triggering an insulin pump at certain times when readings are too high. These insulin pumps can also be triggered to inject insulin when a patient eats a meal.

Artificial pancreas systems combine these two processes into what is called a "closed-loop" system. A glucose monitor is placed under the patient's skin, alongside the insulin pump, allowing the device to independently control glucose levels.

Several clinical trials have shown these artificial pancreas systems to be very successful in maintaining a patient's glucose levels. Removing the need for a patient to be constantly monitoring their own blood sugar is also a major plus.

One challenge these systems have faced is that there is a small delay of up to two hours between the injection of insulin and it reaching a peak level in the bloodstream. Some researchers are exploring faster acting types of insulin but the team at Harvard has taken a more algorithmic approach to the problem.

Schematic showing components of the closed-loop system. The insulin pump and Dexcom G4 Share AP CGM receiver with 505 algorithm were connected wirelessly via bluetooth to the Diabetes Assistant (DiAs) smartphone device.(Credit: Doyle Group, Harvard SEAS)

Using a strategy based on a model-predictive control algorithm, the artificial pancreas is directed to keep a patient's glucose levels to within a broad range instead of trying to maintain a specific fixed point. As well as being able to respond to live glucose readings, the algorithm can "learn" from a patient's daily cycles, delivering insulin based on a variety of variables such as sleep, meals, physical activity and metabolism. This allows the system to ostensibly get ahead of any major blood glucose changes making the device more proactive and less reactive than previous artificial pancreas models.

"This is by far the longest duration trial we have conducted, and it is a testament to the robustness of the algorithm that our key performance indices were maintained from our earlier, shorter trials," says Frank Doyle, lead on the recent clinical trial, which displayed successful results in controlling the glucose levels of 30 patients with type 1 diabetes.

The system still needs to go through more trials before it reaches the general public, but this kind of adaptive artificial pancreas looks set to make the lives of those suffering from type 1 diabetes significantly easier.

The research was published in the journal Diabetes Care.

Source: Harvard

View gallery - 2 images