The First Step Toward An Artificial Pancreas: An Interview With Aaron Kowalski

On January 13th, the Juvenile Diabetes Research Foundation announced a non-exclusive partnership with Animas Corporation to develop an automated, partially closed-loop system to help control blood sugars — the first step toward a fully automated “artificial pancreas.” For anyone with Type 1 diabetes this was, of course, promising news — offering a tantalizing glimpse toward a future where a machine might actually be able to take care of diabetes for you.

As part of the partnership, JDRF will contribute $8 million toward the goal of making a partially closed loop system commercially available within about four years. The system will integrate an insulin pump from Animas, a Johnson & Johnson company, with technology from DexCom, a leading manufacturer of continuous glucose monitors. Aaron Kowalski, Ph.D., Assistant Vice President of Glucose Control at JDRF and Research Director of the JDRF Artificial Pancreas Project took the time to talk about this week’s announcement and explain how this new system might work.

What does the term “artificial pancreas” mean? How does it apply to this new system?

It’s important to set the proper definition here because a lot of people think different things when you say the words “artificial pancreas.” The official goal of what we at JDRF call the “Artificial Pancreas Project” is a closed loop mechanical system that would mimic what a human pancreas does — in other words, it would maintain normal blood sugar levels with no intervention on the part of the user. That’s not what we’re talking about here. We’re talking about a semi-closed loop system that would automatically dose insulin some of the time. It would be the first system ever to do so and would, I’d say, be the first big step toward this ultimate goal of a full artificial pancreas. But it’s not an artificial pancreas yet in the most fully realized sense of the word.

The Artificial Pancreas

So how would it work compared to what we have now — that is, continuous glucose monitors that can tell us our blood sugar levels in near-real time, and insulin pumps, which replace the need for shots (but can’t determine dosing)?

Right now we have an open loop system, which means that we have insulin pumps and continuous glucose monitors but they don’t talk to each other. Instead, we, the users, look at our glucose levels and take into account a lot of things — exercise, food, stress — and try to make a good decision about how much insulin to give ourselves. (Note that fewer than 50% of people with diabetes achieve target glucose levels today and even intensively managed patients, fingersticking seven or more times a day, spend less than 30% of the day in the “normal” blood sugar range.)

With this first generation system, the person with diabetes will still do most of this. But in times of impending low blood sugar or high blood sugar, the system will have the ability to automatically intervene. We call this a “treat to range” approach, meaning that the system is trying to minimize exposure to glucose levels above and below a certain number. The range is probably going to be below 70 mg/dl on the low end, and around  180 mg/dl on the high end.

How would it do this?

If the system predicted you were going to go below 70 mg/dl it would shut off insulin delivery, and if it predicted you were going over 180 mg/dl, it would give additional insulin — the goal would be for the system to automatically bring you back within the target.

The reason we’re taking this approach now is that we want the system to be as safe as possible. It’s only the first step toward a fully automated system — but I think it could be a very clinically important step.

It takes a while for insulin to kick in — and, conversely, once it’s in your body it stays active for a while. How will this system be able to react quickly enough?

It’ll be predictive. The systems use complex algorithms that are much, much more sophisticated than what’s in our continuous glucometers now, and they are able to say with far better accuracy when you’re going to be low or high.

70 mg/dl to 180 mg/dl is a pretty wide range — people without diabetes rarely, if ever, go above 120 mg/dl. Is the proposed target range tight enough? Doesn’t this system risk keeping people’s glucose levels unnaturally high for long periods of time?

A year and a half ago we did a clinical trial with continuous glucometers that found that the average child with diabetes spends 12.5 hours above 180 mg/dl every day and almost one hour below 70 mg/dl. The average adult spends nearly nine hours above 180 mg/dl and over an hour below 70 mg/dl (about 85 minutes on average). Keep in mind, these are patients at the best diabetes centers in the United States — which makes the statistics pretty dramatic, actually.

So my philosophy going in is that okay, here’s what we’re dealing with right now — which is lots of excursions beyond the “normal” boundaries. I don’t think we’ll be able to eliminate 100 percent of lows. But say we can eliminate 50 percent automatically. In my mind, that’d be a pretty good start.

What about people who already are under good control — that is, those who don’t spend a lot of time above 180 mg/dl or below 70 mg/dl. How much will the system help them?

I think the system will hold the most power for the people with the worst glycemic control. So if you have a hemoglobin A1c of 5.5 right now, you probably won’t get too much benefit unless you’re spending a ton of time low. But with that said, everyone sleeps part of the day, and that’s a time when people — even those under good control — are very vulnerable to excursions outside of their target range.

Also, I’ve looked at thousands of CGM traces and can say that even well controlled people spend quite a lot of time out of target. Take me, for example — I’m a scientist, I’ve had diabetes for 25 years, I live and breathe this stuff professionally, and I don’t have an A1c under 6. The example I always use is a time about 2 years ago when I went to a meeting in San Francisco and spent a day golfing. I had a talk the next morning so I tweaked my basal down a bit before bed because I was worried about having walked for five hours — and I woke up at 350 mg/dl inexplicably. And I thought, well, if I’d spent 6 hours at 200 mg/dl instead of 350 mg/dl, that would have been a big improvement.

On the flip side, with our current systems your insulin pump still pumps insulin into you even if your sensor is warning you that you’re very, very low. It’s nutty. And we can fix that. We can fix that right now. Let’s not let perfection be the enemy of the good here — we can probably lop off a lot of these low and high blood sugars.

What needs to happen before we can get to a true artificial pancreas?

We need a couple things, including faster-acting insulin, redundant sensors, and more safety features.  But in the meantime, let’s take what we have and make it better.

I understand that users would still have to give themselves mealtime boluses. But how would the system work with basal rates? Would users still program those themselves?

I think we’ll start with basal rates that are user-defined, but the system will then will give quick feedback on how to manually improve some of the things you do with your pump as you go through the day — like basal rates, insulin-to-carb ratios and correction boluses.

I understand that the system would help predict and correct highs and lows. But how would its abilities be different from the predictive alarms on today’s continuous glucometers?

Just think about any computer and how many variables it can look at simultaneously compared to the human brain.  The new systems will incorporate multiple algorithms that will allow the systems to crunch all the numbers, take into account a bunch of different variables, and start to get smart about what’s going on at the individual level. These next generation systems are able to do amazing, amazing stuff that people simply can’t do — even if you’re a very smart person — and give individualized recommendations on how to improve blood glucose control.

Will these algorithms be in place when this new system comes out?

What the exact product looks like remains to be determined — but in our academic consortium, that’s the type of research going on.

What about glucagon? Will the semi-closed system incorporate it alongside insulin as a way to prevent lows?

Not yet. We’re doing that research in our academic consortium. But in order for that to happen two very important things need to exist that don’t right now. One is a pump that would be able to pump two hormones. That doesn’t exist, and in fact, very few people are even working on that because no one sees a market for it right now. And second, we need a pumpable form of glucagon. In our research, we use glucagon that’s been reconstituted from the powder that you have in an emergency glucagon kit. But to do that in a pump you would need it to be FDA labeled, it would need to be more stable than it is right now, and it would need to be pumpable. So while the results of the studies we’ve done with glucagon are very, very compelling — and integrating it is a big future goal of ours — it’s simply not possible tomorrow.

With the partnership we just entered into, on the other hand, the pieces of the puzzle already exist. The Dexcom sensor exists, the Animas pump exists, and the two companies are putting those devices together right now. They’re planning to market an open loop version of their joint system this calendar year. So to answer your question about glucagon — issues still remain, like who’s going to build a dual-chambered pump, and I don’t want to wait when I think a quasi-closed loop system using just insulin can have an impact right now. It’s a less elegant solution, but I think it’s a very clinically significant solution.

What makes this week’s announcement different from what was going on before? I thought JDRF had been working on the artificial pancreas project for several years.

The Artificial Pancreas Project started internally in 2005 and we launched it publicly in 2006. We’ve seen unbelievable study results in the academic research sites we work with, but the big news this week is that we finally have a company that has committed to saying okay, we think this can be done — we think this can be a product and we can deliver it to people with diabetes and they can benefit from it.

From JDRF’s perspective, we now have a chance to take what we learned in academia and actually capture it in a product that will help people with diabetes. I think that’s what’s really important — we have a timeframe for driving this through the FDA and into the hands of patients.

Are other companies working on closed loop systems as well?

In our academic consortium, we’ve used devices from a variety of different companies. We hope this will be the first of multiple partnerships in the world of artificial pancreases. Our ultimate goal is to see multiple companies succeed here so that people with diabetes have as many options as possible.

What’s the soonest you expect this first generation system to be commercially available? Is four years realistic?

Having lived with diabetes for so long, I hesitate to put time frames on things. But the agreement between Johnson & Johnson and JDRF is a four-year plan. And as I said, the components needed to make this successful are on the table right now — we’ve got the pump, the sensor and the algorithms. What’s more, we’re already doing trials on the academic side of the artificial pancreas project that show that there’s good proof of concept, and that this can not just work, but work very, very well. So the question becomes what it’s going to take to get a real product — and I think the biggest challenge for us will be getting approval from the FDA.

The good news on that front is that we’ve been working with the FDA for the past four years and the FDA has named the artificial pancreas project a critical path initiative, meaning it’s a high-priority focus area for the agency. I think JDRF and the FDA obviously want the same thing: a device that has the highest degree of safety and efficacy.

What about design — both of this system and of diabetes systems in general. In other words, when are we going to have the diabetes equivalent of the iPhone?

Design is another area we’re going to focus on in the very near future. I think everyone appreciates that, at the moment, the design of diabetes devices is lacking. Hopefully we can start an arms race to see smaller, easier to use and more comfortable devices. Because you know, for me, I’m willing to have all this junk in my pocket. But for a mom who has a 3 year old with diabetes, it’s almost impossible. We have to make it easier — and it’s coming. We’re hopeful we can get really innovative companies to collaborate with us and do it better.

What’s the best way people can support this project and make it happen quickly?

The best way to get involved is to go to the JDRF website and sign up for information on the Artificial Pancreas Project — http://www.artificialpancreasproject.com. There’s tons of information on there, and you can sign up so that you get notified of clinical trial opportunities. Fundraising is also obviously critical to help us move research forward. Advocating is another key thing we need folks to do so that when there are challenges at the FDA or with Congress we can have a community of people pushing on the project’s behalf.

Can you allocate money specifically for that project?

Yes. [click here to donate

What hasn’t been reported on this week — what are people missing or getting wrong that you want to point out or clarify?

I think where you led off is a main source of confusion — which is, what is an artificial pancreas? We can’t fully automate a system right now, but I think this stage-wise approach is a key component of getting to a real artificial pancreas.

I also really hope this device will help with the daily grind of living with diabetes. Anyone with diabetes can appreciate that the sheer task of managing the disease. Reducing part of the burden would be a big step forward — you’ve got priorities in life and diabetes often takes a lot of time that you wish you could spend on other things.

What makes you personally excited about this project? What message do you have for people with diabetes?

I have Type 1 and my brother does as well and I think what makes me the most excited is the idea of reducing lows and reducing — or maybe even eliminating — the risk of diabetic complications in the future.  Also, easing the burden for teenagers and kids and parents of kids with diabetes — I mean, to me, that’s just so huge. I don’t know that we’ll remove the burden completely right out of the gate, but I think we can have a big impact. Having seen a lot of this data, I personally just can’t wait. I think it’s possible, and I think we’ll get it done.

Catherine Price is a regular contributor to ASweetLife, she writes the blog The Reluctant Diabetic

Comments (7)

  1. Robert Scheinman
    Robert at

    Exciting stuff. I wonder if one algorithm will work for everyone or if certain lifestyles will require different algorithms.

  2. Mandy at

    I agree with Robert, this is exciting stuff. What this truly comes down to is the interface and the algorithm of the device. Something extremely intuitive and programmable by the user is what’s needed, in fact, you mentioned the iphone. An application that connects through wifi or bluetooth to the device would be phenomenal.

Add a comment

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>