Professor Roman Hovorka: how a mathematician built the artificial pancreas for type 1 diabetes

Roman Hovorka is Professor of Metabolic Technology Research at the University of Cambridge. He began his career as a mathematician, and is now one of the world’s leading artificial pancreas researchers, developing a device that can deliver insulin automatically according to glucose levels.

As a mathematician, what led you to become interested in type 1 diabetes?

In the late 1990s, I used mathematics to model how the body responds to insulin delivered by injection or pump, and to find an optimal insulin delivery rate.

As a result of this work, I realised the extent of the challenges type 1 diabetes poses in terms of day-to-day variability of insulin needs – and that without closed-loop, achieving optimum glucose control is burdensome for some and impossible for others.

Following that, I developed the first generation artificial pancreas algorithm between 1997 and 2000, and the second generation from 2006 to now.

What benefits have you seen so far in people using the artificial pancreas?

Our clinical study data show a broad range of benefits including reduced HbA1c, reduced mean sensor glucose, reduced hypoglycaemia burden, and increased time in range.

We also observed that users have an improved quality of life demonstrated by improved sleep, less time spent on diabetes management, reduced family stress, greater independence for children and teenagers, and greater meal freedom.

What difficulties have you faced in developing the artificial pancreas algorithm?

The challenges include variations between individuals, and even day-to-day variations in the same person, but also the safety aspects – for example, if sensor glucose levels are temporarily incorrect. The algorithm needs to be able to cope with all these aspects.

Exercise also continues to be challenging, even with closed loop. We offer an ‘Ease off’ mode, which minimises the risk of hypoglycaemia by increasing target glucose and assuming high insulin sensitivity for the period when ‘Ease off’ is activated.

In spite of these challenges, we now have one algorithm which is licensed for use by all people with type 1, from age 1 up, including during pregnancy.

What improvements are you still making?

We are continually improving the app, for example by connecting to Dexcom Clarity/Follower to increase remote monitoring capabilities.

It’s not just about glucose management but also about reducing the burden of management and integrating into people’s lives.

We would like to increase the number of options around managing levels after meals. We’re continuing to monitor the feedback from our users.

What is the future of the artificial pancreas? Do you think a fully automated and interoperable artificial pancreas is possible?

At the moment, the user still needs to give a manual bolus for meals, and the closed loop manages insulin delivery overnight and between meals.

But some people are already using fully automated closed loop – we have teenagers who do not bolus and still get decent glucose control.

Faster insulin analogues, and multihormone closed loop systems including, for example, pramlintide, may help us to move to fully closed loop without the user needing to bolus for meals.

Interoperability will happen gradually, although right now the big players are working on proprietary solutions.

However, our app has been designed to connect to different glucose sensors and insulin pumps, and we continue our discussion with sensor and pump manufacturers to expand connectivity.

I also think that steps to reduce the device burden, such as using patch pumps, or sensors that can be placed on top of the insulin cannula, would increase appeal and adoption.

This interview was originally published in Spanish in Revista Diabetes, the magazine of the Spanish Diabetes Association.