First, what’s the status of the race towards an artificial pancreas?
I’ve been quietly watching from the sidelines for nearly the past two years as work on artificial pancreas algorithms and hardware has gotten much closer to becoming reality. Excitingly, we are now just under two years from having a wide selection of artificial pancreas options on the market! (pending FDA approval of course)
The soonest looks like it will be Medtronic’s MiniMed 670G hybrid closed-loop system, which is currently in pivotal trials (pivotal or Phase 3 trials are the phase required by the FDA before devices can be marketed and sold to the public), with a potential launch in April 2017. Following closely is TypeZero’s “inControl” system.
Tandem, Insulet (of OmniPod fame and the only tubeless artificial pancreas trial device I’m aware of), Bigfoot Biomedical and Beta Bionics are all aiming for pivotal trials in 2017 and would therefore potentially launch in 2018.
Short of a cure, the release of these products is something I’ve dreamed about for years. No longer having to check my blood sugar 8 to 12 times per day. No longer having to precisely measure the carbohydrate content of my food. No longer having to painstakingly calculate the correct carb:insulin ratio. And no longer having to worry as much about what I eat in general and constantly have to think about diabetes.
I was lucky to hear about a study going on at Stanford for the artificial pancreas about 6 months ago and ended up taking part in a trial for the Beta Bionics version. That’s the same Beta Bionics, led by Ed Damiano, which just secured $5M for the planned 2017 pivotal trial to take their iLet Bionic Pancreas to market. If you don’t know the Beta Bionics story, Ed gave an excellent TED talk that also serves a good intro to the concept of an artificial pancreas:
3 Weeks on a Bionic Pancreas
The Stanford based study that I participated in used the trial hardware from Beta Bionics — an iPhone 4 based system that is the predecessor for the improved iLet that will end up being the final product. The iPhone based system was clunky, oversized and as a result of integrating an old Dexcom G4 receiver and Tandem pump faced a number of connectivity challenges. However, what’s really important is the algorithm underlying the hardware that is the brains of the entire bionic pancreas.
Each of the implementations of the artificial pancreas leverage different algorithms and each has its own set of assumptions and input mechanisms. The Beta Bionics bionic pancreas system is one of the simplest, requiring only your weight and then, on a meal by meal basis, the type of meal (breakfast, lunch, dinner or snack) and size of meal (ranging from small bite through normal to extra large). Over time, the system learns what a “normal” meal is to you for each type and how to bolus insulin correctly to cover your needs. Other algorithms, such as the Medtronic 670G, will still require you to put in a carbohydrate amount to bolus whereas that is not needed with the Beta Bionics system.
The system is constantly learning from what you tell it. You never need to estimate exact carbs — only compare and answer “is this a normal amount of carbs for this meal for me? less than normal? more than normal?” Based on what the system learns about you and what it observes after meals through the CGM data coming in every 5 minutes, it will adjust future meals of that type and size accordingly.
It is worth noting that there is a limitation in that the system only learns when given four uninterrupted hours after a meal. If you eat again or inject again within four hours it won’t be able to learn from that meal. In today’s world of frequent snacking this sometimes meant extended periods where the bionic pancreas never properly learned how to manage certain meals for me.
The bionic pancreas smartly compensates for any trends it observes. If you start to trend high, it will start injecting tiny boluses to bring you back down into range. And conversely, if you start to trend low, it will suspend insulin delivery so that your blood sugar can rise. Note that the iLet is fully capable from a hardware perspective of being a dual-hormonal system and releasing glucagon to raise blood sugar as well. However, glucagon is not yet FDA approved for small, consistent doses and also current glucagon mixes can’t remain stable for more than 24 hours. Until both those problems are solved (lots of work is being done as you read this), all artificial pancreas systems on the market systems will be insulin only.
The upside and the downside to the simplicity of the Beta Bionics system is the lack of any control over more specific settings. This means that when the system gets it right it is magical and incredibly relaxing. It’s not possible to fully put into words the burden the system took away when I no longer had to constantly keep a portion of my brain dedicated to thinking about my blood sugar and tweaks I needed to make. Going into a meeting and trending high? I know the bionic pancreas will handle it, so not a problem! Forgot to bolus for that little snack? My blood sugar will run a bit high a bit longer since I didn’t give the system any heads up, but ultimately it handled that well too!
That said however, especially for someone who, like me, keeps a close eye on blood sugar levels and is constantly tweaking and correcting, it can be frustrating to watch while the system reacts more slowly than I would have to correct a high blood sugar.
The thing you’ve got to understand is that even if the system were perfect, with today’s technology blood sugars wouldn’t be perfect.
The first issue is that CGM sensors are inserted just under the skin, in the interstitial fluid. While you can get accurate blood sugar readings from the interstitial fluid, the readings are about 15 minutes delayed when compared directly to the blood. Accuracy is rapidly improving in CGMs but this lag issue is going to remain and means that any system relying on CGM data will be slightly behind.
The second issue is that today’s most advanced synthetic insulins peak in 30 to 90 minutes. This means that after the system sees the CGM reading and injects insulin it will take at least 15 minutes for the insulin to start impacting high blood sugar and bringing it down and even longer for it to take full effect.
Even with those considerations, magic is still the only way to describe the fact that I could go to sleep at 180 mg/dL and reliably know that the bionic pancreas would take care of me and I’d wake up at around 120 mg/dL. Across the entire study this was the most universal benefit that resounded with everyone in my study group. Too often sleep is interrupted by hypoglycemia or hyperglycemia and knowing that you had a friend who was watching every 5 minutes and ensuring your blood sugar was in range was incredible!
So would I use it long term?
If the Beta Bionics insulin-only system were on the market today in its final iLet form would I use it? The short answer is in some cases but not always.
I would use the system in a heartbeat for nighttime. Knowing that I can go to sleep and have a friend maintaining my blood sugar perfectly over the course of the night is well worth it for that alone. Waking up within range gives you a perfect start to the day — not only do you feel better but you don’t have to be frustrated dealing with any rollercoaster of blood sugars.
That said, I’m currently using Insulet’s Omnipod, which is tubeless. Aside from the trial I’ve never been on an insulin pump that uses tubing to connect the pump to your body. Between the worry about the tubing catching on things, the hassle of routing the tubing and the fact that the system is not waterproof and has to be removed to swim, it is a no-go for me personally during the daytime.
Additionally, I already aggressively manage my blood sugars and as a result have an A1C that hovers around 6 percent. For me I would love a system that maintained my basal and had the benefits of keeping my blood sugar steady during the day and at night but would still give me the added control of handling meals. This would provide the benefit of the friend who’s always monitoring and helping, but with you still at the reigns.
Due to the insulin-only approach with current systems, the target blood sugars on artificial pancreas systems is going to be between ~120–150, which translates into a slightly higher average A1C given the slower reaction times. For someone who wants a more hands off approach, doesn’t mind tubes, and whose A1C is anywhere higher, however, the system is a complete no brainer. The risk of diabetic complications increases significantly as A1C increases.
Looking to the future
Overall, using the bionic pancreas was a glimpse into the future of diabetes technology. It is incredible to have an independent system that, every 5 minutes, makes decisions about how to better manage my blood sugar. Even today, using the Dexcom G5 I look at my blood sugar maybe every 30 minutes, and often much less frequently. That’s a sample rate increase of at least 6X! That alone provides an incredible amount of room for improvement in control over what I do today.
These next two years are going to be exciting as we watch these systems come to market and immediately start impacting the lives of diabetics everywhere. The impact of bringing someone down from an A1C of 9% or even in some cases 10%+ to the 6–7.5% range these look like they will achieve is the difference between a healthy and long life and one filled with diabetic complications.
Throughout the course of the study there were three key takeaways for me that have led me to better control in the short term:
First, was the fact that CGM data, and blood sugar in general is and always will be a lagging indicator. The only way to keep blood sugars under ideal control is to know in advance what you’re going to eat and what exercise you are going to do and adjust accordingly. I’ve become more mindful of planning my meals and injecting insulin in advance to flatten out spikes in blood sugar.
Second, is the incredible value in the slope of one’s blood sugar. The iPhone trial device displayed the current slope of blood sugar as an actual number e.g. +2mg/dL or -3mg/dL. While the Dexcom G5 shows you arrows, having the actual number is far more actionable. If I’m at 90mg/dL and dropping at -5mg/dL (measured every 5 minutes) and I know I still have insulin onboard then I know I’ll be low (less than 60mg/dL) in 30 minutes and can therefore eat something to avoid that. It’s simple math but it was something I’d never thought of before — having that number gives you predictive knowledge as a leading indicator. For now there is no on-the-market system that displays actual slope values and therefore time before you go low and or high that I’m aware of unfortunately.
Third, with my pump I’d never suspended basal rates entirely. However, watching the system do it regularly to help me avoid lows and maintain my blood sugar made me realize how great a tool that is. Now, when I’m trending low I’ll simply suspend insulin delivery from my pump and avoid or at least lessen the impact of many lows.
Lastly, I was reminded during the study just how incredible the community around diabetes truly is. Ed Damiano flew out to meet our study group and eat dinner with us in person and it was inspiring to hear first-hand why he’s working on this system and his vision for the future. All of the other study participants as well as the team which was running the Stanford study were amazing as well. Dr. Bruce Buckingham, who was running the study, is one of the leading researchers on artificial pancreas systems and his knowledge was enlightening. Everyone helped everyone else and lasting friendships formed that have already far outlasted the conclusion of the study. I’ve written about this before, but it’s always good to get a reminder of the power of the community and the inspiring people within it who are working so hard to bring about change.
This post was originally published on Tim Harsch's personal blog: diabeticallyzen.com.