Skip to Content

An Artificial Pancreas

A device that reads glucose levels and delivers insulin may be close at hand.

Today, people with diabetes have a range of technologies to help keep their blood sugar in check, including continuous monitors that can keep tabs on glucose levels throughout the day and insulin pumps that can deliver the drug. But the diabetic is still responsible for making executive decisions–when to test his blood or give himself a shot–and the system has plenty of room for human error. Now, however, researchers say that the first generations of an artificial pancreas, which would be able to make most dosing decisions without the wearer’s intervention, could be available within the next few years.

Artificial pancreas: Scientists are pairing continuous glucose monitors, such as the device pictured here (white device, top), with insulin pumps, such as the one pictured here (pagerlike device, bottom), to create an artificial pancreas for people with diabetes. In this commercial system by Medtronic, the glucose monitor wirelessly transmits data to the pump via a meter (not pictured). However, the user must still decide how much insulin he needs and dose it out himself. In an artificial pancreas, specially-designed algorithms would calculate how much insulin is required, and how quickly, and then signal the drug’s delivery without human intervention.

Type 1 diabetes develops when the islet cells of the human pancreas stop producing adequate amounts of insulin, leaving the body unable to regulate blood-sugar levels on its own. Left unchecked, glucose fluctuations over the long term can lead to nerve damage, blindness, stroke and heart attacks. Even among the most vigilant diabetics, large dips and surges in glucose levels are still common occurrences. “We have data on hand today that suggests that you could get much better diabetes outcomes with the computer taking the lead instead of the person with diabetes doing it all themselves,” says Aaron Kowalski, research director of the Juvenile Diabetes Research Foundation’s Artificial Pancreas Project.

At its most basic level, an artificial pancreas consists of three components: a continuous sensor to detect glucose levels in real time, a miniature computer that can take those readings and use an algorithm to predict what will happen next and determine how much insulin is necessary to keep the levels steady, and an insulin pump driven by the computer that doses out the appropriate amount of the drug.

Two of the components–insulin pumps and continuous glucose monitors–are already on the commercial market (the latter received marketing approval by the U.S. Food and Drug Administration just a few years ago). “In the near term, you could probably create a pretty robust system with today’s technologies,” says Kowalski, whose group has spearheaded a coalition aimed at bringing an artificial pancreas to market as soon as possible.

Members of the consortium are experimenting with variations of this closed-loop system, so named because the computer algorithm connects the insulin pump and the glucose monitor, closing the loop. Perhaps the person closest to developing a commercial system is Roman Hovorka, a principal research associate at the University of Cambridge, in the U.K., where he leads the Diabetes Modelling Group. His first closed-loop study examined the effectiveness of the system when used overnight, during the hours when blood-sugar levels are likely to drop precipitously and complications can occur. “I want to move to an approach that could be commercialized, and the simplest is just to close the loop overnight, at a time when one cannot do too much about insulin anyway.”

Hovorka used two devices, both commercially available. The first, a continuous glucose monitor, consists of a subcutaneous sensor that measures glucose levels in tissue beneath the skin and a device that communicates wirelessly with the sensor to download its data. The second is the pump itself, a pager-size device with an insulin reservoir that delivers the drug through a thin tube to a subcutaneous needle. Hovorka and his collaborators added an algorithm that not only put the pump and sensor in communication with each other, but also took the (sleeping) user out of the picture by determining precisely how much insulin to mete out every 15 minutes.

When tested in 12 children with type 1 diabetes, the closed-loop system brought the kids’ blood-glucose levels into the target range 61 percent of the time, up from 23 percent for those who followed their normal routine. “With the closed loop, we are able to avoid the extremes–the extreme bad low and the extreme bad high,” Hovorka says. He’s currently working with device makers in the industry to create a marketable commercial product.

Technologically, the remaining obstacles for researchers are those of refinement–for example, constructing algorithms that are exquisitely honed to predict in which direction glucose levels are moving and at what rate. Other researchers are working on sensors that can monitor blood glucose over an extended period of time (currently, sensors must be replaced every three to eight days) and with improved accuracy.

Despite the fact that much of the technology is on the market, researchers must still prove to the FDA that their system is safe when combined with the algorithms, and that if anything goes wrong–if a sensor goes wonky or the insulin pump clogs up–the computer can sense it and either set off an alarm or turn the whole system off.

“You don’t have to get the perfect system to make a tremendous advance and make it considerably easier to live with diabetes,” says William Tamborlane, chief of pediatric endocrinology at Yale School of Medicine, who invented insulin-pump therapy in the late 1970s. As a clinician, he’s more interested in seeing these incremental advances make their way to the patients than in waiting for a perfect system to be created. “We now have sensors that can say what the blood sugar’s doing every minute,” Tamborlane says. “And we have insulin pumps that can change how much insulin it gives on a minute-to-minute basis. We have the technology right now to come pretty close to what might be considered the ultimate solution.”

Keep Reading

Most Popular

This new data poisoning tool lets artists fight back against generative AI

The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models. 

The Biggest Questions: What is death?

New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.

Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist

An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.

How to fix the internet

If we want online discourse to improve, we need to move beyond the big platforms.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.