OHSU recently announced that WaveForm (AgaMatrix) has licensed, in a non-exclusive agreement, two of its hybrid closed loop control algorithms: its fading memory proportional-derivative algorithm and an MPC algorithm. Both controllers can be run in single- or dual-hormone (insulin + glucagon) mode, but WaveForm has only licensed the insulin-only versions. The algorithms require meal announcements, but automatically deliver correction boluses.
WaveForm also licensed OHSU’s smartphone and cloud-based drug delivery platform – iPancreas – which could enable rapid development of smartphone-based systems. OHSU’s Dr. Peter Jacobs told us that initial WaveForm closed loop studies will be run on iPancreas, though we’re not sure if it will be the basis for the commercial system.
- WaveForm reportedly began testing a hybrid closed loop system, consisting of its own proprietary “Cascade” CGM, the OHSU algorithms, and a pump last month. The plan is to test a “vertically-integrated” system in 2019, consisting of Cascade, the OHSU algorithms, and a “specifically-selected” pump, which WaveForm previously said it intends to ultimately own or have built.
OHSU’s automated insulin delivery (AID) group, led by Drs. Peter Jacobs, Jessica Castle, and Joseph El Youssef, recently announced that Waveform Technologies (the AgaMatrix subsidiary) has licensed two of its control algorithms for an in-development closed loop system for type 1s.
The non-exclusive agreement grants WaveForm access to OHSU’s fading memory proportional-derivative algorithm and an MPC algorithm. Both algorithms require meal announcements, but are capable of automatic correction boluses – competitive with TypeZero. The licensed algorithms are designed to run in either single- or dual-hormone mode, though WaveForm has only licensed the single-hormone versions. Theoretically, they could access the dual-hormone versions at a later date if desired.
In January, WaveForm told us they were in the process of finalizing a license for a basic algorithm and that they intended to modify it in advance of initial studies. OHSU’s engineering group will help WaveForm migrate the algorithms to their platform, and initial trials will leverage OHSU’s iPancreas smart-phone based platform, which was also licensed as part of the agreement.
WaveForm reportedly began testing a hybrid closed loop system, consisting of its own proprietary “Cascade” CGM, the OHSU algorithms, and a pump last month. The plan is to test a “vertically-integrated” system in 2019, consisting of Cascade, the OHSU algorithms, and a “specifically-selected” pump, which WaveForm previously said it intends to ultimately own or have built. The CGM has an ambitious feature set: 14-day-plus wear-time, average one calibration per day (it will launch with one per day in Europe), no receiver (Bluetooth smartphone display), limited interferences (e.g., acetaminophen), one-hour warmup time or less, and “painless” insertion – in four initial studies, it performed with an MARD of 11.1%-13.9%. The company anticipates a CGM CE Mark and EU launch this year (the clock is ticking), and clinical trials to support a “PMA submission” to FDA could begin as soon as 2019. That strategy came before FDA’s creation of a lower-risk 510(k) iCGM pathway, so it’s possible that WaveForm will opt for that route instead – this will still require a clinical accuracy study.
OHSU’s is the latest academic algorithm to be commercially licensed. Most recently, Lilly licensed Class AP’s algorithms (studied in Dr. Ahmad Haidar’s McGill lab); Tandem, Cellnovo, and EOFlow have licensed UVA’s TypeZero algorithms; and Insulet licensed UCSB/ModeAGC’s MPC algorithm. We’re surprised that Dr. Roman Hovorka and co.’s Cambridge MPC algorithm hasn’t yet been snatched up for commercial development – it’s one of the most well-studied systems out there!
Algorithm Details (Q&A with Dr. Jacobs)
Q: What startup info do the algorithms need?
Dr. Jacobs: HbA1c, total daily dose, weight
Q: What are the glucose targets (or ranges) of the algorithms? How do they adapt to the user over time?
Dr. Jacobs: The targets are adjustable in the algorithm. In dual-hormone in our recent studies, the target was set to 115 mg/dl for insulin, and 95 mg/dl for glucagon. In single hormone, the recommended insulin target is set to 130 mg/dl. The algorithms use several adaptation approaches. We have an adaptive learning postprandial hypoglycemia-prevention algorithm (ALPHA) that adaptively updates post-meal basal for postprandial hypoglycemia should it occur. And we have an exercise detection algorithm that works with wearable fitness devices. This feature shuts off/turns down insulin once exercise is detected or announced. For the dual-hormone algorithm, it also adjusts the glucagon target, enabling earlier glucagon delivery and also potentially more glucagon to be delivered during exercise.
Q: When does the algorithm exit closed loop?
Dr. Jacobs: The algorithm will only exit closed-loop mode if there is a problem detected with either the sensor or the pump. The patient can also exit to open-loop mode at any time.
New OHSU Diabetes Care Paper: Single- and Dual-Hormone AID with Automated Exercise Detection
In an outpatient, crossover RCT published in Diabetes Care last month, the team showed that dual-hormone control supplemented by automated exercise detection via wearable sensors (heart rate + accelerometer) reduced hypoglycemia in physically active type 1 adults (we saw initial data from this trial at ADA 2017). The algorithm studied was a modified version of the fading memory proportional-derivative algorithm run on a Nexus phone, the one licensed by WaveForm. With the dual-hormone system during exercise (three moderate-intensity aerobic exercise sessions), individuals spent 3.4% of time <70 mg/dl, compared to 8.3% with single hormone, 7.6% with PLGS, and 4.3% with current care (pre-exercise insulin adjustments allowed. Time in hypoglycemia was the lowest with the dual-hormone system across the entirety of the four-day study as well, and time-in-range (70-180 mg/dl) was not significantly different between the single- and dual-hormone systems (74.3% and 72%, respectively). The question remains whether dual-hormone delivery will be a feasible commercial product; however, we’re very glad to see the comparative research continue!
OHSU Dual-Hormone Research Update
Unrelated to WaveForm, OHSU continues to use Xeris native glucagon (Xerisol) in a three-day outpatient study. The glucagon is stable for the duration of the study and does not need to be changed.
Beta Bionics uses Zealand’s liquid-soluble dasiglucagon for its bihormonal iLet system. We saw positive Zealand/Bionic Pancreas safety results in February, and Dr. Steven Russell’s ATTD talk expected the bihormonal pivotal trial to start in June 2019.
OHSU’s fading memory proportional-derivative algorithm currently uses glucagon as a control equal-and-opposite to insulin – i.e., gas vs. brake. However, in the new dual-hormone MPC algorithm, the algorithm runs an insulin-only MPC during normal operation and a dual-hormone MPC only if glucose drops below a threshold.
At Zealand’s Capital Market Day, Dr. Castle touted new data from her lab (n=20) showing that a dual hormone system reduces the amount of time spent in hypoglycemia 200 minutes post-exercise: 3% (six minutes) with a dual hormone system vs. 8% (16 minutes) with a single hormone system (p=0.009). Again, this kind of analysis is important but does leave us wondering if the incremental advantage is big enough to warrant the higher cost and complexity.
-- by Brian Levine, Adam Brown, and Kelly Close