Balancing Act

The morning after his first full night with the artificial pancreas, Ben Yarmis is sitting up in his hospital bed and smiling. He feels pretty good, he says. There was a glitch during the night—one of the glucose sensors moved out of place as he tossed and turned in his sleep—but attending staff were able to resolve the issue quickly. “It’s odd getting accustomed to it,” he reflects. “I don’t have to do something I’ve been doing for the past however-many years. I can’t wait for it to be released.”

In his most optimistic moments, Weinzimer predicts that an artificial pancreas might hit the medical marketplace within five years. But he knows that’s an aggressive estimate. One reason is that the U.S. Food and Drug Administration will be vetting the device with extreme scrutiny. Automated devices carry a high risk of liability; if a patient died while wearing the mechanism, the manufacturer could be in for years of costly lawsuits. Weinzimer thinks the FDA will look more favorably on the artificial pancreas if devices for partial automation of insulin delivery—which are already on the market—gain wide acceptance. “In terms of a product,” he says, “we’re not going to have a home run. It’s going to be a series of singles and doubles.”

The Paradigm Veo, which Medtronic launched last fall in over 50 countries (not including the United States), represents the first step in this progression. The Veo includes both an insulin pump and a continuous glucose monitor, and it will withhold insulin if the patient’s glucose levels drop too far. But patients must program the insulin pump themselves. And they still have to check their own blood sugar regularly.

So far, diabetics have given the Veo good reviews. But the artificial pancreas is not ready to follow it into mass production. One hurdle is the problem Yarmis experienced with a sensor that shifted at night; in most trials, patients’ sensors fell out at least once.

Medtronic’s technicians plan to address the issue in future prototypes. “The best fail-safe is to have some redundancy,” says Cesar C. Palerm, a Medtronic principal scientist who came to Yale–New Haven to observe Yarmis’s trial. That might involve multiple sensors; if one sensor showed glucose levels rising and others did not, the system could be programmed to disregard the first sensor’s input. Alternatively, a differently designed sensor might be less likely to pull free from the patient’s skin.

Weinzimer is still reviewing the exercise trial and hasn’t finalized his conclusions. Still, he’s optimistic: “Preliminary data suggest this device can really help reduce the risk of low blood sugar reactions that frequently accompany exercise,” he says. He’s now planning another critical step toward taking the artificial pancreas live: a trial to establish whether patients can safely use it outside the hospital.

As the trials progress, Weinzimer and Medtronic expect to make an assortment of tweaks to the device. They’re considering adding a function that will let users lower their insulin dosage before they start exercising, as well as warning alarms that will alert users if the glucose sensors stop functioning accurately.

For Yarmis—who is now mulling medical sensor development as a future career option—the day when Weinzimer and Medtronic have the device ready for production can’t come too soon. Wolf feels the same way. “To say the artificial pancreas is a step up is an extreme understatement,” he says. “The idea of never having to worry about monitoring—that’s close to a cure.”