"This is a discussion," says Boris Kovatchev, 48, waving at a whiteboard on the wall of his University of Virginia office. There are no words, only scrawled equations and mathematical doodles. But "discussions" like this one, from the mathematician's brainstorming session with colleagues earlier, could help transform the lives of those with type 1 diabetes. Kovatchev has devoted much of his career to diabetes, which killed his father, and his team is among the leaders racing to develop an artificial pancreas.
The idea is to marry two existing devices. One is a pump that infuses insulin into the bloodstream through a tube. The other is a continuous glucose monitor—generally a sensor below the skin that reads blood-sugar levels and a transmitter that wirelessly sends the data to a receiver.
The human pancreas meets the body's immediate need for insulin and even anticipates it from, for example, the smell, sight, or thought of food. Conventional insulin pumps, on the other hand, are dumb: They dispense the hormone according to a preset program or when triggered manually, not by assessing the body's needs. Kovatchev and others working on faux pancreases want to make the pumps smart, using a sophisticated computer algorithm that interprets the data from the sensor and instructs the pump precisely how much insulin to spit out. Creating such algorithms is Kovatchev's task.
It's a thorny assignment, but Kovatchev believes his team is up to the job of crafting algorithms that control blood glucose under most circumstances. Last year, the Food and Drug Administration accepted computer simulations he and colleagues ran using programmed models of 300 real type 1 diabetics in lieu of traditionally required animal trials. Within a few months, the FDA approved a small inpatient trial on humans.
So far, 20 type 1 diabetics have had their blood glucose levels maintained for up to 16 hours, including overnight, by an artificial pancreas using an algorithm Kovatchev customized for each participant. It wasn't totally automated: Insulin could be dispensed only with a study physician's OK, an FDA safety stipulation. Still, the algorithm's insulin-dosing instructions led to a drop of nearly 80 percent in the rate of overnight blood sugar crashes. Outpatient studies are planned. Kovatchev's eventual goal is a reliable, safe system that operates without human intervention. "That's the main thing," he says. "To [ease] the burden of thinking about control and just automate."