Meet Jennifer Schneider, an orthopedic surgeon and mother to a now-preteen daughter with type 1 diabetes.
Despite her mellow, down-to-earth manner, she's also a high-powered Silicon Valley entrepreneur, one of the pioneers working to get a functional Artificial Pancreas system to market ASAP.
A Guest Post by Jennifer Schneider
It is 2003, I am an orthopedic surgeon with a 2-year-old, and I can barely reach over the table to operate on my patients given that I am 34 weeks pregnant. My husband Dain has planned for every eventuality. Life is going to get even more chaotic. But it’s also going to get even better with the arrival of our son.
Over the weekend, our daughter Taylor is delighting in her new skill of drinking from a cup. As with any new skill, it’s yet to be perfected so I am cleaning up a lot of spilled milk. It’s not clear how much she’s actually drinking, but as soon as the glass is empty she’s asking for another.
By Sunday morning, I realize that I’m also changing wet diapers, a LOT. I’m concerned. I call my roommate from residency who is a pediatrician, and she confirms that Taylor needs to go to the ER immediately. What happened next will be familiar to many of you with type 1: blood tests, hospital stay, then back home completely overwhelmed with unpredictable and wildly fluctuating blood glucose numbers, carb counting, and round-the-clock BG checks. Even with my medical training, this is complicated and unbelievably exhausting.
In our case, Taylor was on such tiny doses of insulin that the only way to measure it in a syringe was to dilute the insulin. And no, you can’t get diluted insulin from the pharmacy so we were instructed to mix our own insulin, combining dilutent and Humalog. Dosing diluted insulin is yet one more step in a complex process: first calculate the dose based on carbs and BG; second, convert from insulin units to unit markings on the syringe; third, mix it with the right dose of NPH. This requires uninterrupted concentration that is difficult to find with a 2-year-old running around and new baby on the way. I couldn’t believe we were expected to manage blood sugars with this imprecise and error prone system for a 2-year-old! We transitioned to an insulin pump as soon as we could.
Through JDRF, Brave Buddies (an online group long before Facebook), and (the newer Palo-Alto based support group) CarbDM, we found an incredible T1D community. We joked that T1D was the best club that we never wanted to join. The years went by and despite all of it, Taylor thrived -- school, sports, friends. The amazing T1D community supported us, and I am grateful every day.
As a D-Mom and a physician, I avidly read the diabetes medical literature. I was particularly intrigued by the promise of a closed loop, also known as the Artificial Pancreas. The data around closed loop was compelling. Even with early, less accurate sensors, controlled clinical studies showed that algorithms were effective, especially at night. However, it was equally clear to me that the early sensors were not ready for prime time. Our personal experience with early sensors was an exercise in futility. The readings were inaccurate, the calibration process was cumbersome, and the huge needle made it miserable to insert. For closed loop to work, the sensors needed to improve.
Fast forward to late 2012. The Dexcom G4 Platinum, an accurate and reliable sensor had just been approved. Unfortunately, we learned about it the hard way. Taylor, now 11 years old, had woken up with a blood glucose in the 60s -- not terrible, but not comforting. She drank some juice, and started getting ready for school, but I had an intuition that something wasn’t right. I kept a close eye on her. While she was brushing her hair, she suddenly collapsed with a seizure. It was frightening to see. After an emergency shot of glucagon, we headed to the hospital. We are incredibly fortunate to live close to Stanford Hospital and to have had a long and close relationship with the amazing Dr. Bruce Buckingham. He was kind enough to meet us in the ER and made sure that Taylor was going to be fine.
That’s when he recommended the Dexcom G4.
From the moment that we started using the G4, our lives were changed. For the first time in a decade, Dain and I had peace of mind. But as great as this was, it raised the question: why was I still getting up in the middle of the night just to enter numbers from the sensor into the pump? Sensor accuracy and reliability had leapt forward. Pump technology was solid. The closed loop algorithms had undoubtedly achieved a proof of principle. Where was closed loop?
I literally asked this question of everyone I met: academics, members of industry, diabetes advocates, and funding groups. I attended conferences. Closed loop research was brimming with activity. It seemed to be on the cusp, but only in the academic world. That’s when I met Tom Peyser.
Tom is the former VP of Science for Dexcom. In 2014 we started meeting regularly to discuss how to expedite commercialization of a closed loop system. Tom had reviewed the entire body of closed loop literature for a paper published in The Annals of the New York Academy of Sciences. We concluded that work was needed to translate the academic algorithms to a commercial product. Many open questions remained. Just a few examples include: how to mitigate the small but potentially dangerous risk of sensor error, how to start on a closed loop system, and how to transition between open and closed loop.
Together we started a company, Mode AGC (Automated Glucose Control) in Palo Alto, focused on addressing these questions and with the intent of working with pump companies to integrate the algorithm into their products. Tom had led Dexcom’s involvement with closed loop studies around the world observing numerous studies first-hand. While there are many excellent academic groups, Tom’s experience, combined with his comprehensive literature review, led him to recommend that we reach out to the Doyle Lab at University of California, Santa Barbara. We partnered with Frank Doyle, PhD and Eyal Dassau, PhD (pictured) and licensed their most recent algorithm.
Our team now had sensor expertise, user perspective, and algorithm expertise. Tom has an exceptional understanding of the sensor and played a major role in helping Dexcom improve the accuracy of their CGM with G4 Platinum and G4AP. I understand the professional pressures on healthcare providers: EMRs (electronic medical records); constraints around documentation, coding and reimbursement; large patient panels with diverse goals, expectations and challenges, and very limited time per patient. I also understand the many facets of type 1 diabetes management along the continuum from toddler through teen years. Dr. Doyle and Dr. Dassau are world-renowned control theory and closed loop algorithm experts. Just last month, Dr. Doyle was appointed Dean of the School of Engineering and Applied Sciences at Harvard where he and his team will continue to work with us on commercializing closed loop.
The algorithm that we have licensed uses model predictive control (MPC) to automate glucose control. MPC is a control theory that was originally used in chemical plants and oil refineries and, since the 1980s, has been used for complex engineering across a wide range of industries. It is safe. It is proven. And it fits the biggest problem with dosing insulin, namely, that insulin delivered now takes effect in the future. The algorithm determines, in real time, the optimal micro-bolus of insulin so that the predicted glucose an hour in the future is on target. This process is repeated every five minutes, 288 times a day. It reminds me of the Wayne Gretzky quote about skating to where the puck is going, not where it has been.
Dr. Doyle developed the world’s first MPC algorithm for insulin dosing over 20 years ago and has been improving it ever since. The current version of the algorithm is in a clinical study with 30 volunteers. I was fortunate to observe the study recently and was impressed to see the algorithm automatically dose insulin to cover a 65-gram pasta meal. I watched the blood glucose tracing go from around 100 mg/dL pre-meal back to around 100 mg/dL post-meal without the user doing anything. It took a few hours, and the peak was about 270 mg/dL so it wasn’t perfect, but it was pretty amazing to watch the BG tracing come right back down to the target without a manual bolus. I marveled at the contrast to what happens now with a missed meal bolus: 65-grams would result in a BG of 400 mg/dL.
In many ways, this first-generation closed loop product will work a lot like the current pump and sensor system: boluses, infusion set changes, and troubleshooting. But there will be one big difference: blood glucose control. This system will automate nighttime glucose and assist the user during the day. We don’t see the system as a self-driving car – it’s not just set and forget – but for the vast majority of users this will be a game-changer, offering dramatically tighter glucose control without lows.
Decades of research have gone into the development of the components of a closed loop system: algorithms, sensors, and pumps. The final product will need to carefully integrate all three components. At Mode AGC, we’re focused on this integration, and translating basic science to a product. The next step is for a pump company to carry this across the finish line, and when this happens, it will be a win for the T1D community.
Wow Jennifer, thank you for sharing your vision and plans! This -- alongside the many other efforts from industry players Medtronic, Animas, OmniPod and Tandem, to the startups like Bigfoot Biomedical and Type Zero Technologies -- gives us great hope for a closed loop system sooner rather than later.
*** UPDATE Feb. 25, 2016: *** Insulet has announced that Jennifer's early-stage startup Mode AGC is working to develop an algorithm for the OmniPod Artificial Pancreas, and work on that system is currently in clinical trials during 2016. Official news release is here.