In this second installment of a two-part interview with Aaron Kowalski, PhD, assistant vice president of treatment therapies for JDRF, he discusses continuous glucose monitoring, the possibility of a fully automated artificial pancreas system, and when Americans might see the next phase of the project come to market.
By: John Parkinson, Clinical Content Coordinator, 
One of the fascinating components of the ongoing artificial pancreas project has been its democratic, collaborative nature. Behind-the-scenes there have been various moving parts to the project including the ebbs and flows of ideas coming from different industries, researchers addressing patient and clinician concerns, and other design and enduser considerations that must be taken into effect when working on such a significant project.
For example, with numerous players from various industries providing their expertise and concerns, many things have to be weighted and prioritized. Yet, even with so many moving parts, and what could be cause for inertia has actually generated simultaneous studies or what Dr. Kowalski (pictured, left) calls “parallel research”. This means while one set of researchers are working on a phase one project, simultaneously, another group of researchers are working on a phase two portion of the project.
In this interview with Dr. Kowalski, he discusses continuous glucose monitoring (CGM), patient criteria for who might be an ideal candidate for an artificial pancreas, and news about when Americans with diabetes can expect to see the technology on the market. With delays from the time of when CGM does the reading to when it reports it, can we ever expect to take the human element out of the equation fully with the artificial pancreas or as you have said, should it be more characterized as full automation of the insulin pump as “some” of the time with people with diabetes still doing measurements and monitoring their insulin and glucose part of the time?
Dr. Kowalski: With today’s CGM technologies no one thinks it’s possible to fully take the person out of the loop.
JDRF has been involved in CGM studies, and in our one study we showed that if you have an A1c of 7, you spend about nine hours a day above 180 mg/dL. The kids who came into our study who were between eight and 14 years old, they had an average A1c of about 7.8. Now these kids come from many of the best clinics in the United States. For people with this reading they spend about 12.5 hours a day with high blood sugar.
One of things we are focused on right now is artificial pancreas systems where the extreme highs and lows would be eliminated. We think with today’s sensors, you could knock that out. You would still be monitoring your diabetes, but if a child misses a bolus at school, or a person with type one is asleep, or someone is just not focused on his diabetes at the moment, this system will help reign you back in.
We are doing studies on this right now, and they are going incredibly well. It works and it is safe with today’s technology. The question is what do we need to do to make it more automated to restore normal blood sugars? In partnership with the Helmsley Charitable Trust, we just launched a big initiative which is focused on next generation CGMs.
One of the challenges in full automation is we need to make sure there is no erroneous data that could potentially lead the technology to overdose someone with insulin. With today’s sensors, sometimes you do see data that is incorrect; it is rare, but we cannot have a system do that in a fully automated closed loop technology.
Today’s CGMs are fantastic. We know they perform very well from all the studies. I have used a CGM for over five years now, and I’m a huge proponent. The accuracy is not as big an issue as reliability and error detection. We have been talking to nuclear and aeronautical engineers as well as chemical engineers in how to make sure the data is high enough quality. 

Full automation is coming, but it is going to take a little bit of evolution of the sensors, and that is something we are funding. We just announced a very large $20 million initiative with the Helmsley Charitable Trust. You can take a child who has a 7.8 A1c for 12.5 hours above 180 mg/dl, and take him down to 4 hours above 180 mg/dl. That would be very meaningful.
Conversely, if you take those 80 to 90 minutes when the child is below 70mg/dl and knock that down to 15 minutes below 70 mg/dl a day that can also be very meaningful.
Interestingly enough, when we have done diabetes focus groups, most people don’t want to give up complete control that would be associated with a fully automated system.
A good analogy is airplanes and pilot control of the plane. Today, the pilots primarily rely on the autopilot. The pilot is the secondary person who controls the airplane as you land, and we are talking about the big commercial airbus planes which carry 400 people on them. However, pilots can still take off the autopilot and fly the plane manually.
Right now, the auto pilot is becoming very good in the artificial pancreas; it’s not ready for full automation yet, but we are getting there. A person with diabetes will still be able to take control of their diabetes even with artificial pancreas. This way, if a person with diabetes can see something doesn’t seem right, they can make an executive decision. This is how we are thinking about this. What are some of the variabilities (e.g. age, physical activity, food intake) in the algorithms you must take into consideration when developing the artificial pancreas?
Dr. Kowalski: The biggest challenge right now is the insulin. Insulin delivery for people is so incredibly different than a pancreas making insulin. It works so slowly compared to the pancreas.
For starters, for people without diabetes, when they see food, their brains tell their bodies to start making insulin. Secondly, the brain tells the alpha cells to stop making glucagon and it tells the liver to stop making glucose as a person gets their system primed up.
As the food comes in, your body secretes insulin and works almost immediately because it is secreted into the blood and you don’t get high blood sugar. People with diabetes don’t have a first phase of insulin secretion. Some people with diabetes will dose insulin before they eat, but it really is not the same as it is being dosed into the skin. Then we are trying to count carbs, and the insulin doesn’t peak into action until 90 minutes later. As such, we have an initiative around directing insulin, and that is going to be a big component.
In talking about people on a personal level, the composition of meals is a big issue. Probably the biggest issue that people worry about the most is exercise, so how do we make sure the artificial pancreas system senses exercise because people can go low during physical activity. Interestingly, this is an area where we can make good headway on because you can monitor peoples’ heart rates.
Certainly for people with diabetes, right now food is a huge issue. However, meal composition for the closed loop--because the system is watching on a minute to minute basis--will be lower in terms of concerns, but there are so many variables we know there are pretty big challenges.
To counter, the proper comparison is can we do better than what people with diabetes are doing now with their control and management? We see in our human studies over and over again that these artificial systems can perform like a pancreas yet outperform people every single time. It is unbelievable and it is significant. I think that is the proper way to frame this. We would love to have our pancreas working like we used to, but we can make a machine that makes it better than what people with diabetes are doing right now. What do you see as the single biggest challenge to this technology becoming a reality?
Dr. Kowalski: The thing we are looking at most is failure detection. It is a big issue. For example, critical systems engineers look at determining how to detect if something is going wrong. One of the key ways to do that is to have redundancies built into the system.
Right now, our CGM devices do one reading, so how do we know that this is really good? We look at blood sugar and we calibrate. Redundancy is the key element to detect if things are working properly and to make sure that number you think you have with the reading is the actual number. 
This is going to be a core element of another initiative, and the question is can we accelerate the delivery of sensors that have very strong failure detection? Interestingly failure detection may be less important than sensors being used in the present open loop as opposed to being used in the closed loop. In the present open loop use, a person can do a check with their meters. With the closed loop, you are going to have a product that is going to be dosing automatically without a person being involved, so we want to be very confident that those numbers are real. I think it is coming and we are trying to stimulate this by funding research. What do you think will be the ideal criteria for a good candidate for the artificial pancreas?
Dr. Kowalski: What we have seen with CGMs is you need to wear your device and be engaged with your diabetes to do well with CGM. Teenagers are often not paying attention and we have seen this in trials. With the closed loop, the idea is the person is going to need to wear the device and they are going to need to be responsible for keeping an eye on things.
One of the things I’m cautiously optimistic about is that some of the people, who struggle with diabetes, may benefit the most here. For example, if we can get kids with high A1cs who are not paying attention to wear these systems and alleviate that burden, it could be tremendously impactful. Are there any medical conditions or contraindications that would eliminate a potential candidate?
Dr. Kowalski: Given that I’m a scientist and not a physician, I think I’m not the best person to answer this, but from our studies, people with mental illness who are not able to troubleshoot the systems and recognize issues are not going to be suitable. Is there any exciting news to report in terms of clinical studies or progress with the artificial pancreas?
Dr. Kowalski: The one I’m excited about is a new trial we just launched called Control to Range, which will try to eliminate the highs and lows. It is being run out of the University of Virginia, but we also have research sites in Santa Barbara, Israel, France, and Italy.
Here in the U.S., the studies are being done by Boris Kovatchev, PhD, from the University of Virginia, and Howard Zisser, MD, from the Sansum Diabetes Research Institute in Santa Barbara, Calif.
This trial is game-changing, because ultimately, this is going to be the first product that is going to come to market. JDRF has a relationship with Animas on this concept. The data is going to be very exciting and it will be used in the first outpatient study. Can you provide a prediction as to the timeframe for when the artificial pancreas might be ready for use?
Dr. Kowalski: The Medtronic MiniMed Paradigm Veo is out on the market internationally. [This product was mentioned in the last interview with Dr. Kowalski.]  Predictive hypoglycemia can be done tomorrow, but  we have regulatory issues with the FDA to resolve first. They want outpatient studies, and the studies with Drs. Kovatchev and Zisser will happen next year. This is real world research with prototype systems.

And with this guidance on December 1, if we can get a clear understanding and a reasonable pathway, then we have the end in sight. The devices exist, the algorithms exist, and it is a matter of testing them. The preliminary tests have all been great. It is not ten years from now. If you are a company, this is commercial development time.
I would be very disappointed if people aren’t using these systems in 2015, and I hope it is sooner.


Part one of this interview with Dr. Kowalski can be found here.