DiabetesMine Fall 2020 Innovation Days

November 5-6, 2020; Virtual; Day #2 Highlights – Draft

Executive Highlights

  • DiabetesMine’s Fall 2020 Innovation Days wrapped up Friday afternoon. See our top highlights below and make sure to check out day #1, as well.

  • Dexcom CEO Kevin Sayer gave a keynote on the “future of CGM,” focusing on type 2s and software innovation. While Dexcom’s next-gen CGM, G7 will initially launch with 10-day wear, the G7 will ultimately be 15-day wear, as we heard on the 3Q20 call. Lots of good discussion here!

  • In DIY, we heard from Scott Hanselman and Marquette Trishaun on their LoopLink project, a smaller and longer-range bridge with the same functionality as RileyLink. Finally, the meeting wrapped up with product demos from Alertgy, Diabits, January.ai, Quin, and Ypsomed. We love meeting with smaller companies; some of these companies we had met at ATTD and had been particularly impressed – see inside for more.

  • Many diabetes-related device submissions will be delayed by about 90 days due to COVID-19. A presentation from new FDA Division of Chemistry and Toxicology Devices Director Dr. Kellie Kelm was a highlight of day #2. Dr. Kelm took the new position in July, following the promotion of Dr. Courtney Lias. Notably, Dr. Kelm shared that many of the Division’s resources had been shifted towards the Division of Microbiology and Virology Devices, the division in charge of regulation for COVID-19-related testing. This shift is expected to delay most diabetes-related submissions by “about 90 days.”

  • DiabetesMine Innovation Days Day #1 – Diversity and inclusion took center stage on day #1 and there was lots of learning shared: over half of BIPOC with diabetes don’t feel represented in marketing, 0% Black representation in recent AID pivotals; more on digital health.

Another successful iteration of DiabetesMine’s Fall 2020 D-Data Exchange wrapped up last week – Amy Tenderich and team don’t stop, not even during COVID-19. See our top highlights from day #2 below.

Top Four Highlights

1. Dexcom CEO Kevin Sayer on “The Future Of CGM”: G7, Type 2s, APIs, and Improved Software

In Friday’s keynote, “The Future of CGM,” Kevin Sayer, Dexcom CEO, provided an overview of Dexcom’s CGM achievements and a look ahead. Mr. Sayer emphasized the importance of CGM and remote technologies in the midst of COVID-19 highlighting an increased usage of Dexcom Clarity in clinics throughout the pandemic. Unsurprisingly, Mr. Sayer’s remarks focused on two big frontiers for CGM, type 2s and improved software.

  • Mr. Sayer briefly discussed G7, emphasizing that they company “will get the product out to 15 days [wear].” As we learned during Dexcom’s 3Q20 earnings call, G7 will initially launch as a 10-day product, as survival rates for the sensor out to 15 days are in the ~70% range, a figure Dexcom is not comfortable with for its users. As of 3Q20, that device was undergoing “approval support trials.” Notably, during Q&A, Mr. Sayer confirmed that as of now the device will continue to allow for DIY use in the same way as the G6 sensor.

  • During his presentation, Mr. Sayer spent a significant amount of time arguing in favor of CGM for patients with type 2 diabetes. Specifically, Mr. Sayer referenced a study from Intermountain Healthcare in which 99 patients with non-intensive type 2 diabetes were matched with 49 receiving standard care and 50 using Dexcom’s G6 for 6 months. Preliminary results of the real-world study indicated G6 use was associated with “~$5,000” in annualized savings per patient bolstering the claim that CGM is cost-effective in non-insulin intensive populations. Additionally, Mr. Sayer also cited data from United Health Group’s Level2 program indicating CGM can improve glycemic control in non-insulin intensive type 2 patients. Mr. Sayer also discussed Dexcom’s recently announced partnership with Lilly to co-market G6 and Lilly’s rapid-acting inulin Lyumjev, with a particular focus on primary care.

  • Mr. Sayer dedicated the second half of his presentation to answering questions many of which revolved around plans for Dexcom’s API. Currently, Dexcom does support an API, but due to FDA regulations, real-time data is not available. According to comments from Mr. Sayer, Dexcom is “currently in the process of submitting” real-time API to the FDA, but no timeline was given for submission. Pivoting from APIs, Mr. Sayer also used the question and answer portion of his presentation to emphasize Dexcom’s software related goals moving forward. Most notably, Mr. Sayer acknowledged that the majority of Dexcom’s time and investments have gone into developing sensor hardware with a smaller emphasis on the supporting ecosystem of software and patient experiences. However, now that Dexcom has reached a point where its technology provides accurate and robust CGM data, Mr. Sayer indicated the company is working to advance algorithms to improve sensor reliability and analytic capabilities. We heard similar comments on the 3Q20 call, in which Mr. Sayer referenced future  software offerings aimed at personalizing the CGM experience for different people, helping to improve engagement and ultimately, outcomes. Finally, Mr. Sayer also indicated the company hopes to implement increased automation for re-ordering sensors and simple tech support to improve user experiences.

2. Updates from the FDA: Leadership Changes; Expect Delays for the Next 90 Days; the Potential of Over-the-Counter iCGMs; the INSPIRE Questionnaire

The FDA Division of Chemistry and Toxicology Devices Director Dr. Kellie Kelm spoke on Friday, sharing updates on the division’s leadership, its functioning and bandwidth during the pandemic, and what the division views as potential next steps in diabetes technology innovation and approval. The Division has seen several changes in its leadership in the last few months. Dr. Kelm took over the role as Division Director from Dr. Courtney Lias in July, following Dr. Lias’ promotion. Dr. Marianela Perez-Torres has filled in Dr. Kelm’s previous role as Acting Deputy Director. Within the Diabetes Diagnostic Devices Branch, specifically, Dr. Leslie Landree has taken over as Acting Chief, following the departure of Dr. Alain Silk earlier in the year. Dr. Landree’s focused on BGM and A1c-measuring devices, while Acting Diabetes Diagnostic Device Branch Team Lead, Dr. Yiduo Wu leads the team focusing on AID systems, CGM, pumps, controllers, and algorithms, meaning that he’s the FDA regulator focusing on much of the new technology in diabetes. Drs. Landree and Wu have both been at the Agency and in diabetes for a length of time (Dr. Wu since 2012 and Dr. Landree since 2009).

  • Dr. Kelm shared that regulatory delays should be expected for at least the next 90 days, as resources and personnel shift away from the Division of Chemistry and Toxicology to work on COVID tests and Emergency Use Authorizations (EUAs). As a reminder, diabetes regulation falls under the Division of Chemistry and Toxicology Devices (DCTD), which falls under the Office of In Vitro Diagnostics and Radiological Health. The division in charge of reviewing COVID-19 testing, the Division of Microbiology and Virology Devices (DMVD), is also under In Vitro Diagnostics and Radiological Health; understandably, some of Chemistry and Toxicology’s resources have been shifted to DMVD. Already the DMVD has reviewed >2,400 Emergency Use Authorization applications and >770 Pre-EUA for SARS-CoV-2 tests, and “unfortunately, the number of applications continues unabated, and the backlog has grown because unfortunately, we just haven’t put enough staff towards the review of these EUAs. So recently, within the last weeks, the Office has decided to prioritize these EUAs and help support the Division of Microbiology and Virology Devices with more resources from the other divisions. Our division is now left with only about 12 reviewers, managers, and medical officers to support our division’s activities.” Per Dr. Kelm, the DCTC will continue to review clinical trials (Investigational Device Exemptions), Breakthrough Device designations, and study risk determinations. However, the change in resources will delay PMAs, 510(k)s, submissions, and pre-submissions for “about 90 days.”

  • Interoperability was a major focus of Dr. Kelm presentation, including the potential for over-the-counter iCGMs. Dr. Kelm shared that the FDA is open to the potential of over-the-counter iCGMs. Given the strong accuracy and reliability required for iCGM clearance and the obvious benefits for the diabetes community in terms of accessibility and affordability, the OTC path under iCGM regulation is a solid option for companies seeking OTC CGM approval. However, Dr. Kelm noted that companies would need to show that people can use the system successfully only by using the instructions. 

    • What’s next? Dr. Kelm offered a slew of possibilities moving forward, including more standalone algorithms (iControllers), remote phone-based bolusing for insulin pumps, app interfaces for AID systems, additional CGM options, and more device options for type 2s. Overall, she noted, “the FDA would love to see more options in the interoperable device space to provide more choices to people with diabetes.”

  • The Insulin Dosing Systems: Perceptions, Ideas, Reflections, and Expectations (INSPIRE) questionnaires were also highlighted. As a reminder, the INSPIRE questionnaires were developed by Dr. Katharine Barnard-Kelly (Barnard Health) and other leading psycho-social experts and were qualified as a Medical Device Development Tool (MDDT) in June, the first diabetes-related MDDT and the sixth ever MDDT. The INSPIRE questionnaires can be used to qualitatively assess the impact of AID systems on the psychological functioning and quality of life in youth with type 1 diabetes (ages 8 to 17), as well as their parents and caregivers, and in adults with type 1 and their partners. Dr. Kelm urged viewers to speak to the FDA about using the questionnaires in studies and during Q&A, suggested that it could be used as a study endpoint and in technical assessments and testing.

Friday’s afternoon session with the DIY community’s Scott Hanselman (the sixth ever Looper) and former NASA engineer Marquette Trishaun (Hip Science) was very interesting. The two helped create a new bridge option for people on Loop, LoopLink. LoopLink, like RileyLink, is a Bluetooth-to-radio frequency bridge that allows DIYers’ smartphones and insulin pumps communicate. In fact, LoopLink is a fork of RileyLink (RileyLink’s hardware and firmware are open source on Github) and preserves the same firmware and functionality as RileyLink. LoopLink is smaller and thinner than RileyLink (see picture below) and has approximately double the range. LoopLink is collecting emails for beta testers right now and hoping to ship to test users sometime this month – sign up for the beta list here. Mr. Hanselman is working with Nightscout to build and distribute LoopLinks (at cost) with the hopes for broad availability in December 2020. For now, the LoopLink will only be compatible with Omnipod pumps (Eros, 433 MHz) and be $125 shipped. This is about even with the cost of RileyLink. Back in February 2020, Jaeb’s John Lum estimated a global Loop user base of ~9,000, with 2,585 Medtronic RileyLink shipments and 6,449 Omnipod RileyLink shipments.

  • Throughout his presentation, Mr. Trishaun expressed his appreciation for the work of the DIY community. With RileyLink being open source, he estimated that “90%” of the work was already done in creating LoopLink, allowing him and Mr. Hanselman to create a product in just a year. Mr. Trishaun also discussed the potential for using microcontrollers, rather than smartphones or Raspberry Pis, to host the AID algorithm. With this set-up, the system would be more power-efficient and allow users to be outside the range of their smartphone and still be in closed loop. Mr. Trishaun outlined an obvious use case from his experience coaching youth basketball, as such a system would allow people with diabetes to run up and down the court without worrying about carrying their smartphones around.

4. Product Demos: Alertgy, Diabits, January.ai, Quin, and Ypsomed

  • Alertgy founder and CEO Marc Rippen presented his company’s work on developing a non-invasive CGM, DeepGluco. DeepGluco takes the form of a wristband and uses a dielectric sensor with radio frequency (16-800 MHz) to measure glucose levels. As a composite signal, the device is measuring a mix of blood and interstitial glucose levels, though weighted more heavily towards to blood glucose, due to the large number of veins and arteries in the wrist. Changes in dielectric properties of blood with glucose concentration is documented in the literature, though the changes are quite small (Yilmaz et al., 2019). Still, Mr. Rippen shared that Alertgy has already run a small study and is currently starting a “secondary trial” with the Cleveland Clinic. The company has already raised “about $3 million” in seed and bridge funding and is currently working on a raise, presumably, Series A. On cost, Mr. Rippen shared that the prototypes cost “under $100” for the build and additionally, the device does not need to be replaced every few weeks. Finally, Alertgy aims to have its device on the market by 2023, though Mr. Rippen noted that his goal is to work with a bigger company in the space to help get his product to the market earlier.



  • Amir Hayeri (Bio Conscious) presented the Diabits app which features predictive glucose capabilities out to 60 minutes and predictive nocturnal hypolycemia nine hours in advance. Diabits integrates CGM data with three machine learning algorithms that compete to best predict glycemic outcomes. Diabits’ three models are: (i) a base model trained only on CGM data; (ii) an ICE model that integrates glycemic responses to insulin, carbs, and exercise; and (iii) a neural net model based on conglomerate data from all Diabits users. These three models compete to most accurately predict glucose with whichever of the three models that has been the most accurate in the prior 30 minutes predicting the next 30 minutes of glycemic data. Between the three models, Mr. Hayeri explained the personalized Basal and ICE models always outcompete the neural net model indicating that personalization improves predictive capabilities. Currently Diabits needs 5-7 days of baseline CGM data to make accurate glucose predictions 60 minutes in advance.

January AI

  • Noosheen Hashemi, Founder and CEO of January AI, presented her app which provides patients with glycemic predictions based on their food and exercise behaviors. January AI is targeted at the non-insulin intensive type 2 population and works by tracking patient CGM data for two weeks to train an AI algorithm that is then able to predict the user’s glycemic response (without the need for real-time CGM data) to any of the 16 million foods in the January AI food database. In the January AI app, users are able to identify foods with a simple photo or a more advanced manual input of foods and ingredients. The app then uses an AI algorithm to predict the glycemic index of foods and provide users with a prediction for their own glucose levels that includes tools for staying in range such as specific length of time to walk following a meal. January AI also functions as a telemedicine provider connecting users with doctors who are able to prescribe CGMs and work to help improve glycemic control.

  • Founder and CEO Cyndi Williams presented Quin, a free UK-based mobile app co-created with people with diabetes that helps people on MDI to assist in insulin titration and dose timing, based on what’s worked for them in the past. Using the app, 75% of users have fewer hypoglycemic and hyperglycemic events, and 67% of users spend more Time in Range. Quin asks users to manually enter their meals (small, usual, or large, optional carb count, and time eaten), basal and bolus insulin doses, and hypoglycemia treatments in the app. When users enter their meal, they can look at their past experiences eating that meal and find the experience that was most similar, using that experience as a model for their current insulin dosing decision. Because it relies on past data, Quin generally suggests that users to log for three minutes/day for three weeks to start gaining insights from their past experiences. Quin helps also users predict their glucose levels in the next five hours and sends users reminders when their recorded insulin changes phase, when they planned to eat, and when they scheduled their basal and bolus insulin doses. The app also includes trend arrows on the “What’s Ahead” screen, showing the impact that insulin and meals are currently having on their glucose levels, as well as insulin on board. The app is available for iOS devices (App Store; 4.7/5 stars) in the UK and Ireland (the Android app is currently in the works), launches new features monthly, and has 2000+ downloads. During Q&A, Ms. Williams reiterated that the app is not for those on insulin pumps and noted that while Quin is not currently integrated with smart pens, they currently take the data from Apple Health and are hoping to eventually integrate directly with smart pen data. Quin was CE-Marked in 2017, and the FDA pre-submission and NHS pivotal study are slated for 2021. We met with Ms. Williams at ATTD and found ourselves immediately immersed – she is yet another entrepreneur who is watching very closely for what can be enabled in our arena and we look very forward to hearing more from her and the team.


  • Rounding out the five demos, Todd Durniak (Ypsomed US/Canada General Manager) presented updates on Ypsomed’s Ypsopump and Ypsomed Loop Program – it’s great to see this longtime diabetes leader from BD in the field and we’re excited to see what Ypsomed can make possible. Much of what we heard was discussed in Ypsomed’s F1H21 update last Tuesday. As a reminder, YpsoPump is small (“the size and profile of your standard business card”; 1.6 cm thick), the battery lasts 30 days, has a glass 160 U insulin reservoir that can be exchanged separate from the infusion set, and comes with an app that offers bolus calculation functions. We got a look at the pump via Ypsomed’s new Explorer App, which allows potential users to explore the pump’s functions.  Mr. Durniak also offered an update on the timeline for Ypsomed’s AID system, Ypsomed Loop Program. Ypsomed expects approval for Ypsomed Assist in 1H2021, Ypsomed Dose in 2H2021, and Ypsomed Loop in “mid-2023.”


--by Hanna Gutow, Katie Mahoney, Albert Cai, and Kelly Close