Artificial pancreas: an (almost) automated diabetes treatment?

Interview with Prof. Christoph Stettler, Medical Director and Chief of Staff, Bern University Hospital for Diabetology, Endocrinology, Nutritional Medicine & Metabolism (UDEM), (Inselspital)


The treatment for diabetes is very time-consuming for patients: they need to regularly monitor blood sugar levels, take medication and inject insulin. Poor self-management may result in a dangerous lapse in blood glucose levels. Yet external factors can also contribute to diabetes being out of control. An artificial pancreas system could offer relief.

Photo: Smiling man with suit and tie - Prof. Christoph Stettler; Copyright: Tanja Laeser/Inselspital

Prof. Christoph Stettler; ©Tanja Laeser/Inselspital

In this interview with, Prof. Christoph Stettler explains how this type of system works and how a clever algorithm regulates blood glucose levels - and is even meant to predict the future.

Prof. Stettler, what is an artificial pancreas?

Prof. Christoph Stettler: In the body, the pancreas regulates -among other things- blood sugar by producing the two hormones, insulin and glucagon. In the case of type 1 diabetes, the pancreas no longer produces insulin because the beta cells have been destroyed. Subsequently, blood sugar levels rise. If we want to replace the pancreas with a device, we need to simulate an organ that is extremely well "built" because not only does it release hormones but it also continuously "measures" blood glucose levels to respond to changes - regardless of whether we eat, exercise or sleep. Simulating this function is not a trivial task.

The artificial pancreas is a device system that actually consists of three components: first, a pump that releases insulin via a subcutaneous catheter; secondly, a sensor that continuously measures blood glucose in the subcutaneous fatty tissue; thirdly, a control module with an algorithm that interprets the sensor data and controls the pump to where it always releases the right dose of insulin.

We want to use a novel patch pump for this that is being adhered to the skin. From here, a short catheter is directly placed into the subcutaneous fatty tissue. The pump is technically very sophisticated and uses a brand-new type of motor. We have already tested it in a clinical trial. Today, the control module could be a cell phone for example. The entire system is called a "closed-loop system".

Photo: Products and devices that patients need to treat their diabetes; Copyright:

Measuring, injecting insuline, noting values, estimating carbohydrates and carrying an emergy set with glucagon at all times: This is how complicated the self-treatment of diabetes is. An artificial pancreas could help here; ©

What type of diabetes patient would actually benefit from this type of device? After all, there are several methods to self-manage your diabetes.

Stettler: Technically, all diabetics can benefit from this system. Not just type 1 but also type 2 diabetics, where there is initially no lack of insulin, can all benefit from glucose sensors and insulin pumps. Sooner or later, these patients also need to regularly measure their blood sugar levels and adhere to a more or less complex treatment with pills and insulin injections. At least, this time-consuming self-management is partially automated with a closed-loop system.

What are the limits of previous pumps?

Stettler: Today’s pumps are purely dispensing devices. From the start, they are programmed to dispense a certain amount of insulin as a regular dosage throughout the day. The wearer can subsequently manually request additional insulin to compensate for a meal. However, in doing so, many patients regulate their blood glucose levels at a suboptimal rate. A clever algorithm could do a better job.

At night, a pump that is linked to a sensor via an algorithm is already able to better control blood glucose levels than we could do manually. Studies have shown this. The next step is to also implement this type of control during the day when patients move, eat or experience stress. All of this impacts blood sugar levels and limits the function of the pumps.

Having said that, closed-loop systems also have their limitations: although today's sensors are very accurate, they actually always measure at the wrong spot. Changes in blood sugar levels always arrive with a delay between five and 15 minutes at the subcutaneous fatty tissue. Patients would actually already need to inject insulin before a meal to avoid an increase in blood sugar. Even with today’s rapid-acting insulin types, they are technically too late if they inject after a meal.

Photo: Woman takes a picture of a meal on a plate; Copyright: Vvoennyy

Many people are already taking pictures of their meals anyway. Using the app from the "GoCARB" project, diabetics can jump on this bandwagon to estimate carbohydrates contained in a meal; © Valery Vvoennyy

How can this problem be solved?

Stettler: I think we should work on hybrid systems. This is also the approach we are pursuing. The system per se requires a basic setting. When something is added to this, like a meal, for instance, the patient needs to administer a little more to have enough insulin in the body. The system should also take the blood sugar progression into account and how much insulin is still active in the body to receive and implement this intake in a hybrid mode.

The algorithms that connect the components of the artificial pancreas are very complex. They need to include the delay in the subcutaneous fatty tissue and respond as if they were able to predict the future based on the values from the past.

So this is not a fully automated system and I still need to supply the device with data?

Stettler: That is correct. It is very important to estimate the amount of carbohydrates in your meal. Patients frequently underestimate these values considerably - by 30 to 50 percent. We developed an app for this during the “GoCARB” project that was funded by the EU. Patients can use the app to take a picture of their meal. The app then creates a three-dimensional model and calculates the amount of carbohydrates on your plate. Patients can use this as input to operate their pumps.

What are the next stages of development for the artificial pancreas?

Stettler: We currently continue to work on the algorithm to where it will be able to project blood sugar levels as far in advance as possible, perhaps up to 30 minutes, to detect a trend and actuate the pump at the right time.

The first experiments with these types of devices always take place "in silico", meaning via simulation. Here we use artificial patients, that being large databases with data from real patients, to test the algorithm. The next step is to then test the system with the patch pump and algorithm in standardized situations in human beings and see how well the system works in everyday situations. We are now able to start with this process.

Photo: Timo Roth; Copyright: B. Frommann

© B. Frommann

The interview was conducted by Timo Roth and translated from German by Elena O'Meara.