February 14-17, 2018; Vienna, Austria; Days #3-4 Highlights – Draft

Having scattered from Vienna and found our ways back to good old San Francisco, we’re back with our extensive report from Days #3-4 of ATTD. Boy were these two days packed with updates and new data – we hope you have as much fun reading and learning as we did putting it together!

For a refresher before diving into our highlights below, check out our write-ups from day #1 and day #2. A very big thank you to the organizers of ATTD for putting on yet another stellar meeting – we hear that there were over 400 more people at ATTD this year than 2017 – that growth reflects massively increased interest and activity in this field, up well over 10% in one year alone to just under 3,000 people!

Automated Insulin Delivery (AID) Highlights

1. Roche First Company to Join JDRF Initiative to Accelerate “Open Protocol” automated insulin delivery, big interoperability win!; Value-Based, Integrated Diabetes Care Update

While displaying a slide that read, “Shaping the future of diabetes with JDRF,” Roche Global Head of Diabetes Care Mr. Marcel Gmünder announced that his company has “decided to join the JDRF effort for an open DIY artificial pancreas environment.” Roche is the first company to publicly disclose participation in JDRF’s initiative to accelerate “Open Protocol” AID systems, meaning that it will provide seamless, secure, interoperable connectivity with other devices and smartphone apps (e.g., Bluetooth). This is a major interoperability win from one of the largest device players in diabetes, and it represents the first domino to fall in JDRF’s plan to facilitate an open AID ecosystem. Ultimately, it could drive a whole new paradigm of device innovation, where companies supply components/algorithms and patients can mix and match the devices/software they like best. Roche is currently developing a 180-day AID system with Senseonics’ implantable Eversense XL sensor and a TypeZero algorithm, and may also have plans in store for automating delivery from the Solo patch pump (set to launch in the EU in 2018). It’s impossible to know how quickly, but we hope other companies will follow Roche’s plunge – especially Tandem, who seems well suited for this initiative. Patients will win in a plug-and-play ecosystem, meaning industry should too – the ecosystem has potential to add far more value to devices than each company on its own. There are still questions pertaining on regulatory, liability, and postmarket surveillance, but JDRF has said it will coordinate with regulators and legal advisors to develop a predictable FDA pathway and frameworks. FDA’s Drs. Courtney Lias, Stayce Beck, and team have made it clear that AID component interoperability is a big goal – we can’t wait to see how quickly it happens and which companies drive the most innovation in this new paradigm. We firmly believe in this direction, given the pace at which CGM and algorithms are moving, the impressive efforts in the DIY community, and how hard it has been for others besides Medtronic to push commercial AID systems over the finish line. We’re elated to see Roche take the first plunge!

• Roche’s value-based, device-agnostic, integrated diabetes care offering with coaching and provider decision support is starting to resemble some other emerging programs (e.g., Onduo, Virta, Diabeter), which all take on patient care and sell outcomes to payers and employers. The slide below shows how Roche intends to market the mySugr bundle (already reimbursed by major German payers and with US payers to come online in 1H18) as device agnostic – the slide depicts Medtronic, Dexcom, Abbott, and Roche devices – and with insights for population management and EHR integration. The presentation also didn’t mention A1c once, but focused on “True relief” and more time-in-range – nice! On the healthcare provider side, Roche VP Mr. Tim Jürgens introduced the professional platform, a “hub to drive decision support and therapy efficiencies” set to launch later this year. A short walk-through displayed features including: (i) pattern detection; (ii) “meaningful” graphs (data download for different devices; AGP has been licensed); (iii) connectivity (one single hub for all devices); (iv) communication (more intensive monitoring of patients between face-to-face visits); and (v) population management (patients prioritized based on critical situation; real-time tracking of test strip consumption). We remain optimistic about this new-look, innovative Roche, though spinning up the business models is where the real test will be. As different companies and programs develop and iterate new care models, what best practices will surface? How will traditional modes of healthcare delivery adapt?
  
  • “We need to give tools to healthcare providers. We need to move toward outcomes-based models, demonstrate that we are actually improving outcomes. Bringing relief to patients, which is the ultimate outcome, because with relief, you can be sure that the medical outcome is coming as well.”
  • “We are a firm believer in bringing partners together. No company has all answers and tools for a patient. We are able and willing to bring those things together.”
  • “We cannot have a situation where payers are not able to pay for the relief that is needed.”

• Enthusiasm for mySugr, which now has over 1.4 million users (higher than the ~1.2 million users shared by Mr. Anton Kittelberger on day #1) bubbled over in Roche’s presentations. Mr. Gmünder called the app “our main interface and basis for interaction with patients” and “an engaging and intuitive tool,” and Mr. Jürgens added that it is “the leading management platform” and “basically a little clinic in the hands of the patient.” Wow! So great to hear and we salute again the creators …

2. Dr. Kowalski on DIY AID: “Just Do It!...Don’t be Scared. It’s Time to Move to a New World of Diabetes Treatment. It Works.”

A fired-up Dr. Aaron Kowalski made an extremely compelling call for patients and providers to “cross the chasm” and take up hybrid closed loop: “Many are on other side of the chasm. People are there saying, “it’s far, it’s scary.” But I’m over here, and let me tell you, it’s beautiful … To take a line from Nike, just do it! Come on guys, it’s time. Don’t be scared. It’s time to move to a new world of diabetes treatment. It works. You’ve seen the presentations here. We can do this. Your patients need this. It’s safe. It’s going to happen. What are we waiting for?” “People ask, ‘is it dangerous? Unregulated?’ My answer is this train has left the station. If this community [DIY] is driving, innovating, and meeting patient needs, then we as an industry, clinicians, and scientists, need to find a way to make this as safe as possible and how to learn from it.” To ensure that the audience could trust that he was putting his money where his mouth is, Dr. Kowalski proceeded to display his personal Dexcom Clarity AGP featuring his past three months on Loop: The profile was flat, with a mean glucose of 145 mg/dl, time-in-range of ~72%, time <70 mg/dl of 4.4%, and time >180 mg/dl of 24%. He pointed out, “overnight control with a hybrid closed loop is absolutely stunning. It is powerful.” We truly appreciated this assertive, open move, and heard audible gasps from the audience as the JDRF’s Chief Mission Officer so directly and personally endorsed the DIY movement (while backing it up with strong glycemic data) – Adam and Kelly and many others have seen similar outcomes (though we point out this does require using old pumps – Kelly’s is broken, with no way she knows of to fix, while she waits for another closed-loop trial). Importantly, Dr. Kowalski emphasized that JDRF will not waver in its support of “traditional” industry development of closed loop systems – the DIY systems are, no doubt, driving innovation. He also discussed JDRF’s Open Protocol initiative, possible consequences of automated insulin delivery in the clinic, how the field should learn from CGM’s evolution, and payer considerations/remaining hurdles. He concluded the impassioned talk similarly to the way he opened it: “Think about the need for automated insulin delivery. We have to do it.” We heartily agree and are especially impressed with not only the far lower percent of time in hypoglycemia that 670G closed loop and DIY patients are experiencing, but also the “soft landings” that we hear about and beautiful night-time traces in particular.

• On JDRF’s initiative to accelerate “open protocol” automated insulin delivery, Dr. Kowalski said: “JDRF is trying to chart how to make an open platform more accessible. I call it bringing it above the table. Right now, it’s below the table. We have already retained experts to help us with the legal and regulatory changes – let’s be honest, this is totally outside the box. But I think there’s an opportunity here. The DIY community has blazed a trail that we hadn’t anticipated, and it’s great.” We were delighted that in the very next talk, Roche Diabetes management announced that it would be JDRF’s first industry partner to sign on to the open protocol initiative (see above) and we expect many more.
• Dr. Kowalski pointed out that conversations in the clinic change when the patient is on automated insulin delivery: “With my doctor, the discussion is different than it used to be. There’s a lot of conversation around meal-time issues. I personally think you’ll see many more people taking up pump therapy. I’ve personally had many friends move from MDI to pump to go onto 670G or a DIY system. The benefit of pumps now that we’re automating insulin has gone up dramatically. I think you’ll see a large increase in pumpers, which could cause a challenge for some clinics.” We agree with Dr. Kowalski and have been calling AID the killer app for pumps (and CGM) for some time (see 2013 ADA!). It is true that some centers may not specialize in or have the capabilities for high-volume pump training, but we ultimately view increased pump adoption (particularly with automated insulin delivery) as a huge positive. Providers should be able to spend less time discussing abstract A1c numbers, guessing at patient behavior and glycemia, and manually adjusting insulin doses, and be able to spend more time fine-tuning food/exercise responses and discussing issues that really matter to patients (therapeutic, technological, psychosocial and beyond).
• Dr. Kowalski astutely noted that we can learn from CGM’s evolution – while a lot of people gave up on CGM after early issues with accuracy and on-body burden (and general clunkiness, we would add), they have advanced to the point that they speak to phones/watches, are smaller, more accurate, easier to operate and wear, and implantable versions are coming down the pike. “We can knock barriers down.” This lesson – along the lines of the famous sentiment that we overestimate change that will happen in two years and underestimate change that will happen in 10 – illustrates Dr. Kowalski’s optimism that hurdles to optimal automated insulin delivery, such as insulin kinetics, on-body burden, biologic variability, exercise, etc., will be cleared.
• Despite positive early reimbursement of the 670G in the US, Dr. Kowalski urged for additional demonstrations of efficacy (through larger RCTs) and greater appreciation of outcomes beyond A1c on the payer side. He noted that Medtronic is already conducting an n=1,500, one-year outcomes study with 670G, which we hope will be a driver of greater reimbursement for all systems. He also posited, “what is the value to the payer of sleep?” It is crucial to quantify the benefits of outcomes metrics such as time-in-range and other PROs/quality of life of metrics in order to more clearly define the value proposition of hybrid closed loop. 

Q: Any challenges that you experience, things that may be challenges for patients on DIY automated insulin delivery?

Dr. Kowalski: I think building the system is not that difficult. You need to appreciate that you’re still in control of your diabetes. My system modifies basal rates, so even if you forget your phone, the worst you’ll get is basal rate running above or below where it should be for 30 minutes. My rate of hypoglycemia and area under the curve is dramatically reduced – it takes much less treatment, this isn’t anything we haven’t shown in trials, but living it for 15 months is pretty exceptional.

Q: What is the greatest bottleneck to achieve good control?

Dr. Kowalski: I certainly appreciate that insulin action is too slow. Faster-acting insulin is a big JDRF focus and it’s hard. I’ve been experimenting with Afrezza and closed loop, and I’ve found it pretty effective. Sensors are not a problem. The other thing is carbs, if you’re eating you’ll go high with sub-cutaneous insulin. My challenges are around mealtime insulin. I like to eat carbs, what can I say?

3. Omnipod Horizon: 75%-85% time-in-range in 54-hour meal and exercise studies; 74% time-in-range in interim 5-day hotel results (n=11)

Between an appearance at a corporate symposium and an oral presentation, Dr. Bruce Buckingham shared the results of four Omnipod Horizon feasibility studies, including: (i) a 54-hour study with meal bolus challenges; (ii) a 54-hour moderate intensity exercise study with variable setpoints in adults; and (iii) a 5-day/4-night study across age groups including patients on MDI (preliminary data). As the slide shows below, mean glucose and time-in-range data has been strong in the adult cohort across the board, with mean glucose ranging from 136-155 mg/dl, percent time 70-180 mg/dl ranging from 73%-85%, time <70mg/dl ranging from 0.6%-1.8%, and time >180 mg/dl ranging from 14%-26%. The summary slide only includes data from adults, but Dr. Buckingham emphasized multiple times how much he appreciates that Insulet is investigating its system in adults, adolescents, and pediatrics from the very beginning. Indeed, Horizon has just been tested in 4-5-year-olds, and 2-4 year olds will be studied soon – whoa! At JPM 2018, Insulet guided for a “probably 2020 timeframe” for launch of Horizon, on the later side of 3Q17’s “end of 2019”/“early 2020” expectation. Given the impressive study progress to date, this timing feels doable – perhaps even conservative. All in all, well over 100 patients with type 1 diabetes have undergone >7,700 hours of hybrid closed loop therapy with Horizon running on a tablet thus far. Longer terms studies are underway, evaluating the MPC algorithm under free-living conditions with extended use in patients of all ages. We assume once the hotel study wraps up, a pre-pivotal will commence, paving the way for pivotal (perhaps in 2019?) Dr. Buckingham sang the praises of former protégé Dr. Trang Ly and the Insulet clinical team throughout his talks, concluding that “they did really well.” Read on for details from the presented studies, plus intent to bring the Insulet-Glooko partnership to Europe.

• A five-day/four-night hotel study is underway, and topline outcomes were alluded to in a summary slide: mean glucose of 150 mg/dl, ~74% time-in-range, ~2% time <70 mg/dl, ~25% time >180 mg/dl, and ~5% time ≥ 250 mg/dl. These are consistent with prior studies in more supervised settings – great news for Insulet. We’re not sure about the demographics of this tested cohort yet, though a summary slide indicated that the study would be conducted across age groups and include MDI patients. The latter point is big, as it’s Insulet’s target market (~80% of new users) and the most important area to demonstrate benefit for market expansion. 

• The new 54-hour meal study (n=12 T1Ds) tested Omnipod Horizon over three meals (30-90 grams of carbs) with a 100% bolus compared to various real-life scenarios: missed bolus, 130% bolus, and extended bolus (50% upfront and 50% extended over four hours for a high-fat dinner). Overall, the system coped well with uncertainty: mean glucose was 153 mg/dl over the 54-hour study (134 mg/dl at night); time-in-range (70-180 mg/dl) was 76% overall and 93% at night; and time <70 mg/dl was only 0.6%. Since Omnipod Horizon does not issue automatic correction boluses, it did take some time to recover from the missed lunch bolus: 3.8 hours to get back to <180 mg/dl vs. 1.5 hours with a 100% bolus. Average glucose rose to 192 mg/dl with no bolus vs. 141 mg/dl with the 100% bolus (time >250 mg/dl: 0% vs. 10%). That said, this is to be expected of any hybrid closed loop fully reacting to an unannounced meal, and it’s notable that no hypoglycemia was experienced when the bolus was missed. This was also the first we can recall seeing an extended bolus used in a hybrid closed loop study, an important test since this strategy is often used in pediatrics – when kids’ eating habits are unpredictable, it’s safer to give an extended bolus that can be cancelled if all the food is not eaten. Horizon coped well with the extended bolus and the 130% bolus, providing a roughly similar average glucose and time-in-range as the 100% bolus. Overall, this was a solid showing for the algorithm as it moves to being stress tested in real-life meal conditions.

• The Omnipod Horizon system performed well in response to exercise, with an average time-in-range (70-180 mg/dl) of ~85%. The entire study (n=12) took place over the course of 54 hours in a supervised hotel setting and featured two 45-minute sessions of moderate-intensity exercise with setting adjustments 90 minutes pre-exercise. On day one, the adjustment consisted of a setpoint adjustment (raised from ~110 mg/dl at start to ~150 mg/dl 90 minutes pre-exercise), and on day two, the adjustment consisted of a basal rate reduction (down to 50% 90 minutes pre-exercise). For the duration of the 54-hour study, mean glucose was 136 mg/dl, with ~85% time-in-range, just 1.4% <70 mg/dl (0% overnight; impressive for an exercise study), ~14% >180 mg/dl, and ~2% ≥250 mg/dl. Temporary reduction in basal rate and raised blood glucose setpoint strategies performed well in terms of mean glucose during and after exercise, but there is a visual difference in the glycemic traces (below): With basal rate reduction, there is a slight drop followed by rebound hyperglycemia, while with setpoint adjustment, participants went a little lower initially, but then had less pronounced hyperglycemia later. Seven subjects required carbs during the setpoint adjustment, while just four did with the basal rate reduction – a reminder of how hard exercise is for AID. Of course, the exact adjustments could be further refined, personalized, and tweaked based on individual differences and type/duration of exercise, but this preliminary study shows the algorithm is safe during physical activity. Noted Dr. Buckingham, “This was actually pretty good exercise. One guy in the group, in college, he ran a four-minute mile. So he wanted to run six-to-seven miles in the 45 minutes. The device was on tablet, and I didn’t have a team that could keep up with him carrying tablet, so we put a guy on a bike that followed him around as he ran laps around a lake. There were pools of sweat around these people. We really worked these people.” We initially reported on this topline data at Keystone and ADA.

• Horizon’s 36-hour safety and feasibility study was published in DT&T (Buckingham et al.) on February 12^th. In all age groups, the system performed very admirably with respect to primary endpoints of time <70 mg/dl and time ≥250 mg/dl, ranging from 0.7%-2% and 4%-7%, respectively. Time in range also hovered in the 70%-73% range, on par with performance of the Medtronic MiniMed 670G in its pivotal trial and the real-world setting. Glycemic profiles were generally flatter throughout the day with hybrid closed loop, though high overall mean glucose and postprandial highs (see adult graph below) have presumably been areas of focus in subsequent algorithm refinement. Preliminary results from this paper were presented a year ago at ATTD 2017.

• Following a presentation on real-world Omnipod use trends and data obtained through Glooko, Medical Affairs Director Dr. Jennifer Layne said, “we look forward to having more data to present to you in the near future and hopefully bringing Glooko to the EU.” As of EASD 2017, Insulet-provided Glooko was in 2,800+ US clinics, with over 50,000 users uploading. The partnership will remain an important asset over time, helping Insulet keep up with the likes of Medtronic’s CareLink. An entrance into Europe makes sense as Insulet looks to go direct beginning on July 1, and Glooko and Insulet both stand to benefit – Glooko through sheer penetration, and Insulet through data collection and evidence generation.

4. Bionic Pancreas with Zealand’s dasiglucagon: non-inferior to reconstituted Lilly glucagon; moving to bridging study in iLet device

Dr. Steve Russell (MGH) presented encouraging safety results (n=10) comparing Zealand’s pumpable, stable dasiglucagon vs. reconstituted Lilly glucagon in the iPhone-driven Bionic Pancreas (Tandem pumps). The random order, crossover, in-clinic, eight-hour study intentionally induced hypoglycemia via excess basal insulin and postprandial exercise, stressing the algorithm to test Zealand’s glucagon. Time <60 mg/dl was 13% with dasiglucagon vs. 18% with Lilly glucagon (not significantly different), while time in 70-180 mg/dl was 71% with dasiglucagon vs. 66% with Lilly glucagon (not significantly different). Mean glucose (110 mg/dl vs. 101 mg/dl) and total glucagon dose (0.69 mg vs. 0.86 mg) were also not significantly different. Dr. Russell noted how the glucagons were really tested here – outpatient studies typically use about 0.5 mg per day of glucagon, while this study used ~25% more than that in just eight hours. The most common adverse events were hypoglycemia and nausea, which were comparable between the study arms. Transiently elevated white blood cell count was noted at the end of both visits, but this is reportedly expected with glucagon dosing – and will be key to observe over a longer period in the bihormonal pivotal study. Zealand’s glucagon was dosed without occlusions or infusion set reactions, though obviously the bridging study will provide a better test of that over more days of wear. The team still has many steps left before commercialization – successfully testing dasiglucagon in the commercial iLet device, pivotal studies (next year – April start for insulin-only, June start for bihormonal), manufacturing, human factors, FDA submission, etc. – but this feasibility data certainly green lights dasiglucagon as a viable path forward for the bihormonal configuration of the Bionic Pancreas.

• Dr. Russell showed the bihormonal configuration of the iLet Gen 4 device (first unveiled at Friends for Life 2017), which will be the commercial device used in the pivotal studies (slated to start next year). Dr. Russell confirmed that the 1.6 ml insulin cartridge will have both manual fill and NovoRapid PumpCart (prefilled) options, while the Zealand glucagon cartridge will only come in a 1 ml prefilled format (4 mg of glucagon) – assuming 0.5 mg of glucagon are used per day, this means it would last 5-6 days or possibly more. The infusion set tubing and site will be packaged separately, enabling patients to keep the same cartridge/tubing, but change the site out every 2-3 days.
• The study design to induce hypoglycemia was clever: patients wore the bionic pancreas, in addition to remaining on their own pump. To induce hypoglycemia, the basal rate was doubled on a patient’s own pump, unknown to the bionic pancreas. Participants then ate lunch and received a full bolus through their own pump (also unknown to the bionic pancreas). When glucose started to rise, the Bionic Pancreas would dose additional insulin on top of the hidden bolus. Participants then cycled after lunch, causing hypoglycemia.

5. NIH-Funded AID Study Updates: FLAIR Medtronic Advanced HCL Simulation (TIR 83%), iDCL Protocols, Beta Bionics Study Timing, Dan05 (Hovorka) Recruitment on Hold

NIDDK’s Drs. Guillermo Arreaza-Rubin and Andrew Bremer chaired a crowded session during which PIs of the four major NIH-funded closed loop trials provided status updates. While the FLAIR study (Medtronic Advanced Hybrid Closed Loop vs. 670G) will not commence until later this year, Dr. Rich Bergenstal showed a trace depicting the potential of automatic correction boluses to blunt postprandial hyperglycemia (no surprise there). He also showed data from 40 simulated patients on advanced hybrid closed loop, where time-in-range was a strong 83%. This trial is running well behind the original plan to start in late 2017, perhaps as Medtronic gets the device ready. Dr. Boris Kovatchev gave timing and study design updates on all four iDCL study protocols (including protocol 4, which will investigate Harvard’s enhanced MPC algorithm on a mobile system with Dexcom CGM and either a t:slim or Omnipod pump). Dr. Steve Russell provided a very granular timing update for Beta Bionics’ insulin-only (bridging study beginning in May, pivotal beginning in April 2019) and bihormonal (phase 2b study with dasiglucagon beginning in July, pivotal to begin June 2019) systems – these are also a bit back of the previous plan to start the trials at the “beginning of 2019.” Finally, Dr. Hovorka shared that his Dan05 pediatric closed loop study has enrolled two subjects, but recruitment is currently on hold. It is great to see all of these progressing following their announcement just over a year ago. Dr. Bremer also announced a slew of type 1 diabetes funding opportunities – just released February 15 and enabled by the hard-won Special Diabetes Program – that may be of interest to the ATTD community, including “clinical, behavioral, and psychological research testing current and novel closed loop systems,” “treating diabetes distress to improve glycemic outcomes in type 1 diabetes,” and more. Read details below on each study:

• The Prof. Moshe Phillip-Dr. Rich Bergenstal co-PI’d FLAIR study comparing the 670G to “advanced hybrid closed loop” (670G + Fuzzy Logic including automatic correction boluses) will begin “later this year when all of the systems are built and ready to go” – back from the original plan to start in late 2017  In preliminary simulations of 40 patients, advanced hybrid closed loop has reduced mean glucose by ~10 mg/dl (152 to 142 mg/dl), lifted time between 70-180 mg/dl to ~83% (up from 76%) largely by shaving off the highs (time >180 mg/dl decreased from~24% to ~16%) and with little impact to lows (a similar ~0.6% with 670G vs. ~0.7% with advanced hybrid closed loop). These improvements were conferred by an average of 10+ auto correction boluses per day. Not only does the advanced hybrid closed loop provide earlier and automatic correction boluses, but a slide also laid out that it “liberalizes constraints on algorithm parameters,” and uses sensor glucose values non-adjunctively – we’d assume Medtronic will need to use a next-gen sensor for this system, as Guardian Sensor 3 is only approved adjunctively. And while it is ambitious to take on the 670G as a comparator (unlike the other NIH-funded studies, which don’t compare their systems to existing automated insulin delivery options), there is a fair amount of space for improvement, particularly in curbing postprandial highs. Dr. Bergenstal showed an “n of 1” trace (provided by Dr. Thomas Danne) in which advanced hybrid closed loop cut the duration of a postprandial high after an unannounced meal in half, from six hours to three hours, while also diminishing its amplitude. We’d guess this product is still more than a year from market, and it could be on roughly similar timing to Tandem’s Control-IQ system with TypeZero’s automatic correction bolus feature (launch in 1H19).

• Dr. Boris Kovatchev broke the International Diabetes Closed Loop (iDCL) trial down into its component parts, with the most details we’ve ever heard on the setup for protocol 4 (Harvard’s mobile closed loop system), data from the training protocol, and a slew of timing updates. See this table for a detailed reminder of all of the protocols – it is difficult to keep track of all of the moving pieces.
  
  • Protocol 4 with Harvard’s “Enhanced Control-to-Range (eMPC)” will be PI’d by Sansum’s Dr. Jordan Pinsker and commence at the “end of this year.” The algorithm will be on a mobile device, controlling either Tandem’s t:slim X2 or Insulet’s Omnipod (TBD), with input from Dexcom CGM. The three-month study will enroll 126 subjects, randomized 1:1 to closed loop and SAP, at seven US sites: UVA, Harvard/Joslin, Mount Sinai, Mayo Clinic, Barbara Davis Center, Stanford, and Sansum. Like all of the NIH-funded AID trials, protocol 4 will be coordinated by the Jaeb Center. Primary outcomes are superiority in time <70 mg/dl and non-inferiority in time >180 mg/dl.
  • The results from the n=40 iDCL training protocol were presented in poster form at ATTD, and Dr. Kovatchev highlighted the 73% time-in-range, 76% time-in-range overnight, and 0% median time <50 mg/dl. While there were no differences in system performance with the Roche Spirit Combo or Tandem pumps, there was a notable difference in pump connectivity with peripheral devices, so the iDCL team opted to proceed with the Roche pump in protocol 1 (Roche Spirit Combo pump, Dexcom G5 CGM, TypeZero algorithm), for which the last completer (of n=106) is expected on July 1^st.
  • Dr. Kovatchev shared that protocol 2, the EU pivotal trial of the Roche-Senseonics-TypeZero 180-day closed loop system is expected to begin soon. It was previously slated to begin in 1Q18, wrap up later this year, and support a CE mark. This study’s design was also featured in an ATTD poster.
  • Dr. Kovatchev said the iDCL study group hopes to begin the pivotal study of the Tandem t:slim X2/Dexcom G6/TypeZero Control IQ embedded system (protocol 3) this May, conclude at the end of this year, and continue on with the extension phase. We were not previously aware of an extension phase, but Tandem has most recently projected for a 1H19 launch, suggesting that FDA approval will hinge on the phase that occurs this year. In the protocol 3 pilot (pre-pivotal study of the Tandem/Dexcom embedded system), n=5 participants spent a “very good” ~87% time in 70-180 mg/dl overall, with 94% time-in-range overnight. Time in hypoglycemia was “minimal,” with ~3% overall and just 1% overnight. Notably, there was no time spent >250 mg/dl, and time >180 mg/dl was limited to 5% overall. Wow! Dr. Kovatchev was extremely excited about this system in Dexcom’s symposium on Day #2, a major win for the companies – we have high hopes for this product and pivotal trial.
• MGH’s Dr. Steve Russell gave very specific timing updates for Beta Bionics studies moving forward, headlined by an insulin-only pivotal trial beginning in April 2019 and a bihormonal pivotal study to begin in June 2019 – both are back from the previous plan to start at the “beginning of 2019.” The insulin-only pivotal will be followed by an expected 4Q19 PMA submission, which should allow for what has previously been set at a 1H20 launch. Launch timing for the bihormonal system is more up in the air, as timing of the PMA submission depends on the length of the dasiglucagon exposure study required by FDA. While the effects of chronic exposure are still a question mark, Dr. Russell is excited by the drug’s acute performance and safety: “One of the problems with dual hormone is we don’t have a stable pumpable glucagon, or I should say, we didn’t have a stable pumpable glucagon. Dasiglucagon is stable for more than a month at 40 degrees in a pump. It behaves comparably to reconstituted Lilly glucagon. Now we have a stable pumpable glucagon we can take forward into pivotal trials.” (See the early data above testing dasiglucagon in the iPhone drive system.) Prior to the start of the pivotal studies, an insulin-only bridging study is expected to commence in May (using the iLet integrated device, and the group’s first type 1 diabetes study closed loop study enrolling MDIs as well as pumpers), and the Zealand bihormonal phase 2b study will begin in July (first use of the iLet integrated device in bihormonal mode). The team has pushed back trial and submission timelines several times, so we are surprised to see them providing such granular projections; on the other hand, that could mean the path ahead is more lined up, especially on the hardware development side. We are excited that dasiglucagon is a viable candidate and could make the bihormonal vision a reality. Plus, it will be terrific to see how the single and dual hormone systems compare in the pivotals.
• Dr. Roman Hovorka’s six-month, n=130 Dan05 study of closed loop (modified Medtronic 640G, Enlite 3 CGM, and Cambridge control algorithm on Android phone) in children and adolescents with type 1 diabetes has recruited two subjects, but recruitment is now on hold as investigators await manufacturing of the study device. Dr. Hovorka’s fingers are crossed that the study can start this coming May and wrap up August 2019, but he specified that this is “contingent on a few things coming together.” The comparator arm is pump therapy with or without CGM, though FreeStyle Libre Pro will be used to capture glucose data in the control groups when necessary. Notably, secondary outcomes of psychosocial impact and health economics will be performed at Stanford and UCLA, respectively, with the latter aimed at supporting reimbursement. 

Questions and Answers

Dr. Guillermo Arreaza-Rubin: What are the advantages and disadvantages of embedded (i.e., in the pump) vs mobile (i.e., on a phone) systems?

Dr. Kovatchev: Embedded are simpler, more compact, and faster to commercialization. Mobile provides the option of component interoperability if you run from a phone, devices can be interchangeable…that’s more in the future, but we’re working towards it as well.

Dr. Russell: Of course connectivity can be a problem. You’re always going to have connectivity to the CGM to worry about. One of nice things with fully-integrated systems is you lose the issue of connectivity with a pump. But the disadvantage is you need a fully-integrated system, and in our case, building that has taken time.

Dr. Hovorka: In principle, a mobile system can be more flexible to update if the app is on the App Store. Some of these aspects are possible advantages, but I agree that, especially for children, integrated is beneficial.

Dr. Bergenstal: Not much to add, but I will just say that the phone is not always with you, and if you use pumps to deliver your insulin, your pump is always with you.

Q: In your experience, why do you think people prefer bihormonal to insulin-only?

Dr. Russell: They like that they have even less hypoglycemia than with insulin-only. They also take less carbs; with insulin-only, they occasionally take carbs, and they have to take very few that are medicinal (low correction) with bihormonal. And they just feel a greater sense of confidence that they are not going low, for things like exercise, particularly.

Q: Is there a plan to use new insulins in these systems?

Roman: It’s an interesting question whether these systems will need a new regulatory pathway or new trials if they include new insulins. I think the algorithms are flexible enough to deal with faster insulins and there would be no need to change them – there is so much more variability between people than there is due to a ten-minute shift in insulin speed.

Dr. Russell: Our system has a built in PK parameter, so it makes assumptions about insulin absorbance and clearance. Because of between-people variability, we need to make a conservative estimate. It would be advantageous if didn’t have to adjust it for people – if it did, and the patient switches to slower-acting insulin, then you have a problem. But if you do have to adjust it, then you have to think about an auto-detection method, and that adds complexity.

Dr. Hovorka: In our system, we also have a system that adapts and learns the PK. For those with long PK, the system knows it and can be more conservative and vice versa.

CGM Highlights

1. HypoDE Study: CGM Reduces Hypoglycemia in MDI-treated Type 1s with Impaired Hypoglycemia Awareness by 72%; Another Win for CGM in MDI!

In the very first randomized controlled trial to examine the efficacy of real-time CGM in MDIs with type 1 diabetes and impaired hypoglycemia awareness, CGM significantly reduced hypoglycemia incidence (≤54 mg/dl for at least 20 minutes) by 72% (p<0.0001). Severe hypoglycemia was nearly halved with CGM (p=0.02), a major win for cost-effectiveness. The Dexcom-funded HypoDE study (n=141 completers), conducted by Dr. Lutz Heinemann et al. and published in The Lancet during ATTD, was a six-month, 12-center trial in which participants were randomized to receive either 22 weeks of real-time CGM (Dexcom G5; n=75) or continue with SMBG (n=66). CGM results were collected over an additional four-week follow-up phase, during which the control participants wore the masked Dexcom G4 system. (All participants wore masked Dexcom G4 over a four-week baseline phase.) Study participants were eligible if they (i) had type 1 diabetes for at least one year and (ii) had experienced at least one severe hypoglycemia event requiring third-party assistance in the previous year or reported a score ≥4 on the Clarke assessment. The graphs below depict the primary outcome, with controls in panel A and the RT-CGM group in panel B: Mean number of hypoglycemia episodes per 28 days fell from ~11 to <4 in the CGM group, while staying largely flat at 14 in the control group. Mean nocturnal hypoglycemic events per 28 days also fell from 2.3 to 1.0 in the CGM group, while climbing from 2.4 to 2.7 in the controls. Regarding overall glycemia, time ≤54 mg/dl significantly decreased in the CGM group, falling from 1.7% to just 0.3% (-20 minutes per day) as compared to the control group, which saw a non-significant decrease from 2.5% to 2.3% (-3 minutes per day). Time spent ≤70 mg/dl in the CGM group also dropped significantly from 4.9% to 1.6% (-47 minutes per day), while the control group saw a non-significant decrease from 6.9% to 6.4% (-7 minutes per day). The incidence of severe hypoglycemia events requiring third-party assistance in the control group (1.18) was roughly twice that seen in the CGM group (0.64; IRR 0.36), with 39 events recorded in the control group and 24 in the CGM group – this reached statistical significance (p=0.02), is a major win for cost effectiveness, and confirms the severe hypoglycemia findings from IN CONTROL (EASD 2016). Importantly, the hypoglycemia improvements were not achieved at the price of hyperglycemia, as no significant changes were observed in A1c for CGM (-0.2%; baseline: 7.6%) or the control group (-0.1%; baseline: 7.4%). Glycemic variability measured by coefficient of variation (CV) improved quite significantly from 39% to 34% in the CGM group, while the control group remained flat at 41%. (Dr. Rich Bergenstal frequently cites CV ≥36% as an unstable CGM profile, consistent with Dr. Louis Monnier’s 2017 publication in Diabetes Care.) Despite this improvement, the CGM group did not see a meaningful increase in time between 71-180 mg/dl (58% to 59%) as time >180 mg/dl increased ~3% (absolute increase) in both groups. Aligning with previous work, there were no significant differences in the self-reported hypoglycemia unawareness score between study groups – oddly, both groups improved by ~40%.

• Dr. Rich Bergenstal wrote a stirring article accompanying the study’s publication in The Lancet, noting that “this is the type of study needed to expand the indications for use of CGM in populations with a strong need to improve HbA1c and reduce glucose variability and hypoglycemia.” In the 2018 ADA Standards of Care, CGM indication for those with hypoglycemia unawareness was graded with C-level evidence – we’re hopeful that the HypoDE results will serve to increase the strength of this important recommendation. In the comment, Dr. Bergenstal also pointed to CGM indications as an adjunct to pump therapy/in closed loop, in type 1 patients on MDI, in pregnancy (CONCEPTT), and more and more in type 2 diabetes (see Onduo and Unitedhealthcare/Dexcom’s n=10,000 type 2 CGM pilot), pointed to CGM as a useful instrument in cost-effective telehealth, and said it “might be the best example of diabetes precision medicine widely available today.” Wow! We have certainly seen evidence for CGM's efficacy in a number of populations come a long way, and hope to see the user base follow.

**Panel A depicts the control group (no RT-CGM) while panel B depicts the experimental group (RT-CGM)

• Patient-reported outcomes indicated a positive effect of CGM on hypoglycemia-related distress and satisfaction with CGM in this population. Fear of hypoglycemia dropped from a mean score of 55 to 42.2 on the HFS for patients in the control arm (~23% change) vs. 53 to 37 for patients in the CGM arm (~30% change). Although this just missed the threshold for statistical significance with p=0.07, Dr. Norbert Hermanns noted that CGM showed an “impressive reduction in the burden caused by hypoglycemia fears.” (And minimizing these fears is not at all trivial, given that 79% of type 1 patients lower their insulin dose after a severe hypo, pushing further away from optimal diabetes management.) Overall diabetes distress dropped from a score of 2.5 to 2.1 among control participants (16% improvement) and from a score of 2.6 to 2.0 among CGM participants (23% improvement) on the DDS scale (non-significant p=0.203). Dr. Hermanns also presented findings on the various DDS sub-scales, however, highlighting a meaningful difference between CGM and SMBG when it came to hypoglycemia-related distress, specifically (score of 3.0 at follow-up with SMBG vs. 2.7 with CGM, p<0.05). The Glucose Monitoring Satisfaction Survey was administered for all HypoDE participants. Mean score improved from 3.7 to 4 among SMBG patients (~5% change) and from 3.7 to 4.3 among CGM patients (~16% change), revealing a statistically significant benefit to CGM (p<0.01).
  
  • The fact that there were eight dropouts in the control arm and none in the experimental arm could also suggest greater satisfaction and quality of life with CGM. This seems beyond a statistical anomaly, and one of the strongest indicators that type 1 MDIs with impaired awareness of hypoglycemia enrolled in this study liked wearing CGM. A crossover design would’ve been icing on the cake in answering this question, but the study is quite convincing as is.
• The positive results of the HypoDE trial add to the mounting evidence demonstrating CGM’s benefits in MDI. Dr. Hans DeVries, who provided independent commentary on the study, noted that HypoDE’s “important results” might influence reimbursement – we certainly hope so. Although the GOLD and DIaMonD studies of CGM in MDI likely excluded patients with hypoglycemia unawareness, use of CGM was linked with significant decreases in hypoglycemia in those treated with MDI. The IN CONTROL study presented at EASD 2016, which specifically investigated CGM in those with problematic hypoglycemia (but did so in a mixed population of MDI and pump users) found CGM significantly reduced severe hypoglycemia (14 events vs. 34 events; p=0.03). As the HypoDE publication mentions, since the majority of those with type 1 diabetes are treated with MDI, the HypoDE findings combined with these prior results have both clinical and health-economic implications. The now-repeated finding that MDI plus CGM yields very strong clinical benefits is exciting, especially given the cost-effectiveness of this approach relative to a pump without CGM. We’d love to see a head-to-head study comparing CGM plus MDI with decision support vs. automated insulin delivery with a pump/CGM – would there be a major difference in outcomes? How would payers assess the cost-benefit? It would also be interesting to investigate the role alarms and smarter retrospective analytics play in the reduction of hypoglycemia – e.g., will insights like “You are likely to be low today” have additional benefit over current threshold alarms?
• Despite the significant difference in severe hypoglycemia “requiring third-party assistance,” a sub-category showed no difference: the incidence of severe hypoglycemia requiring “medical intervention” (3 in control group vs. 5 in CGM group; p=0.59). However, given that these events are quite rare (e.g., paramedics coming), it’s likely the study was not powered to detect such differences. King’s College London’s Dr. Pratik Choudhary raised this question during Q&A: Sensor-detected hypoglycemia that the patient experiences (and overcomes) at home doesn’t overtly cost the payer anything, so did HypoDE “just miss the mark” on relevance to payers by not accumulating enough data on hospitalizations? Dr. DeVries responded with notable optimism, explaining that reimbursement is driven by a whole host of data and other factors (not one study alone), and arguing that “some form of CGM” is eventually going to become standard of care for all patients with type 1 diabetes. “Even if there are people who will go to great lengths to dismiss all available data, at some point, someone is going to make the right decision and say now we have enough evidence.” Indeed, an EASD poster from Belgium estimated a cost reduction of €345,509 (€911 saved per person) with reimbursement for 515 CGM users (type 1 pumpers) – the body of evidence supporting CGM, for patients, providers, and payers, is growing. 

2. Dexcom Clarity: data from 50,000 users; EHR integration live at CHLA; mobile app launching OUS in 2018; insulin, meal, exercise data coming

We got a slew of updates on Dexcom Clarity in its first major oral presentation and dedicated corporate symposium: real-world usage/outcomes from 50,000 users (more monthly logins is associated with more time-in-range, less hyperglycemia, and a lower average glucose); EHR integration screenshots (piloting now at Children’s Hospital LA); early Clarity screenshot concepts that add insulin, meal and exercise data; adoption of the mobile app; and clinic uptake. The data management platform is now available in 13 languages and 25 countries, with the greatest usage in the US. Notably, the new Clarity mobile app has made a big difference for usage: 30% of registered Clarity users logged in to the website quarterly vs 50%+ of registered users who have engaged with the mobile app in the past 30 days. The Clarity mobile app is expected to launch outside the US in 2018. Impressively, almost 19,000 people have signed up for the new Dexcom Clarity mobile app/email notifications (see our coverage) – no surprise there since it really reduces hassle and changes the paradigm of diabetes data (passively collecting and sending trends automatically every week). Dexcom is also building out the app to add all the capabilities available on the website. It was a bit unclear how large the Clarity user base, though it’s certainly over 83,800 accounts in the US, 48% of whom are sharing data with their clinic.

• Dexcom’s Dr. John Welsh presented the first Big Data analysis from Dexcom Clarity – web login trends vs. glycemic data from 50,000 US G5 iPhone users in August 2017. As we’ve come to expect with diabetes data, 59% of these users did NOT log in to Clarity last August, 16% logged in once, 13% logged in 2-3 times, and 13% logged in 4+ times. The most impressive trend was for time-in-range (70-180 mg/dl), which ranged from 55% in those who did not log in to Clarity to 64% in those who logged in 4+ times last August (p<0.0001). (The average time-in-range in this 50,000-patient cohort was 57%.) The linear trend was also true of mean glucose (176 mg/dl with 0 logins vs. 161 mg/dl with 4+ logins; p<0.0001) and time in hyperglycemia (41% with 0 logins vs. 32% with 4+ logins; p<0.0001). Notably, there was not much of a signal between Clarity logins and time <70 mg/dl, which makes sense to us – the real ability to avoid time-in-hypoglycemia comes from the real-time alerts. Dr. Welsh emphasized that these are associations and not an assertion of causality, a point we’d strongly echo – it’s likely Clarity logins are a good marker of diabetes motivation, engagement, familiarity with technology, comfort with data, etc. Still, the results parallel Dexcom’s separate ATTD poster on screen views and glucose outcomes, a positive. While Abbott and Medtronic have been faster to share real-world data (on FreeStyle Libre and 670G, respectively), we’re elated to see Dexcom adding to the discussion. This data will be very important for getting a pulse on how patients are doing in the real world.

• Dexcom’s Dr. Nate Heintzman show a screenshot of the EHR-integrated Clarity experience, now live at Children’s Hospital LA – “clinicians there love it.” A clinician “orders” Dexcom data within the EHR, just like ordering a lab test. There is no extra log in on the website, a huge win for time-stressed clinicians. The data appears directly within a patient’s EHR profile. Dr. Heintzman said he is very focused on this project this year, with hopes to bring it to other clinical sites.

• We saw the first-ever early concept designs of Dexcom Clarity with insulin, meal, and exercise data – these may change during human factors, but it’s great to see these are well along in development. The insulin and meal data are displayed in histograms above the modal day plot and time bucketed across the day. No timing was shared on a launch. Seeing these data streams in Clarity will be fantastic (especially insulin), though managing the complexity will rise with the number of streams coming in – otherwise it can be easy to get lost in reports and difficult to piece together what is driving what. Luckily small toggle switches on the bottom of the graph will allow users to turn insulins, exercise, and meals on/off. Plus, decision support will hopefully be layered on top pull out clear patterns.

• Dexcom’s Gary Cohen also showed Clarity use on the clinic side – 83,800 “newly created patients,” of which 84% were invited to share their data with the clinic (70,099) and 48% are currently sharing their data (40,132 acceptances). This has seen a nice uptick in the past year. The Clarity OUS base seems to be ~10-fold smaller, with 8,668 users created in clinics – part of this relates to the smaller Dexcom OUS user base and the shorter time that Clarity has been available OUS. As we understand it, the number of “newly created patients” understates the total Clarity user base, since an account can obviously be created outside the clinical setting.
• Dr. Heintzman shared an update on Dexcom’s public API, which now has more than 500 registered third party developers that have created over 400 prototype apps. Dexcom’s data partners have launched ten apps that use the API, including One Drop, Glooko, Tidepool, Nutrino, and Rimidi – for a complete current listing of Dexcom Data Partner apps, see the App Gallery here. Dr. Heintzman showed how easy it is to set up for a user – from the Glooko app, for instance, a user goes to settings, taps on the Dexcom account, logs in to Dexcom (via OAuth2), and authorize Dexcom to send data to the partner app. We hope to see this public API generate a thriving CGM data app ecosystem, especially as it hopefully moves to real-time data. For a complete current listing of Dexcom data partner apps, click here.
• “Sometimes you miss some things when you just look at AGP. You need to look at 2-3 reports to find everything.” Dexcom’s Keri Leone made this point with an example: a pattern of hypoglycemia that was not noticeable in AGP, but was very clear in Dexcom’s more brightly color-coded modal day plot. We like that mantra – HCPs and users should look at a few reports (if possible) to triangulate views on what’s going on.

3. Medtronic Last Major CGM Company to License Ambulatory Glucose Profile (AGP) Report, According to IDC’s Dr. Deborah Mullen

Medtronic has reportedly licensed the one-page standard Ambulatory Glucose Profile (AGP) report, according to IDC’s Dr. Deborah Mullen – hooray! With this news, Medtronic joins Abbott, Dexcom, Glooko/Diasend, and Roche as the fifth major player to offer AGP to patients and providers – notably the last of the major CGM companies to do so. As a reminder, Dexcom was the most recent company to license the glucose data visualization report last June for integration into Dexcom Clarity, which is already live on the web and mobile app versions. We’re glad to see Medtronic now on board, a win for the field. We have heard murmurs that, even though companies have licensed the AGP, they still may have different specs that could confuse users. For example, the margins around the median trace may be the intended 25%/75% and 10%/90% for one company, but different for another. As noted above, it’s also going to be important for companies to continue innovating on reports, as AGP shares a lot but the single graph won’t catch everything.   

• During her presentation, Dr. Mullen highlighted the value AGP has in the clinic: ~90% of patients and providers say it helps them see new patterns, and almost 98% say it lets them find new opportunities for improving glucose management. She also shared plans to make the AGP visualization “a little bit prettier” over the next few months – we’re excited to see what they come up with, as better and easier-to-use visuals could be a big plus for engagement and usability.

4. Abbott Orals Detail Use of FSL in Belgium (Natl Reimbursed Market) and in At-Risk Individuals Over Six Months

Dual Abbott oral presentations detailed use of FreeStyle Libre in Belgium (where the system has been fully nationally reimbursed since mid-2016) and the longitudinal impact on “high-risk” patients, respectively. Of note, Abbott’s Dr. Tim Dunn foreshadowed both of these analyses at DTM, and also suggested that further analyses could identify strategies of Libre usage (around hypoglycemia, pre/post-meal, insulin dosing, and exercise) that beget glycemic success. We were impressed by these two orals, but are especially looking forward to that around strategies – what are the best practices for FreeStyle Libre usage to derive the most glycemic benefit with the least amount of effort/burden?

• Dr. Tim Dunn compared FreeStyle Libre use behavior and glycemic trends in Belgium (16,000 readers, 43 million monitoring-hours, ~5,000 patient-years of use) with that of all countries (~238,000 readers), finding similar results. Daily scans per reader were roughly on par, with Belgian users averaging 10 scans per day, and all countries averaging 13 scans per day. Belgium’s curves of daily scans vs. eA1c and time >180 mg/dl were nearly identical to the overall user base, but curiously, time-in-range and time ≤54 mg/dl vs. scan frequency curves were somewhat offset (see below). For some reason, the degree of improvement per scan (the slope, per sé) was roughly the same, but the Belgian cohort consistently spent 30-45 minutes per day less in-range at each scanning level, and 10-15 minutes more per day ≤54 mg/dl at each scanning level. Overall the results are not drastically different and support full reimbursement of FreeStyle Libre, but it would be interesting to dig deeper into the time-in-range and hypoglycemia discrepancies. Is it that baseline diabetes care in Belgium is worse than that in other FreeStyle Libre markets? Are there dietary or other cultural factors at play? Are there high levels of stigma sometimes preventing patients from taking action when they see (trending) out-of-range glucose values? Could there be incentives dynamics related to having zero out-of-pocket costs (unlikely, since scanning frequency is similar)?
  
  • This analysis immediately reminded us of a compelling EASD 2017 poster sharing 12-month healthcare outcomes from a national Belgian reimbursement study of CGM in 515 insulin pump users. Most notably, the percentage of patients experiencing a hypoglycemia/DKA hospitalization declined by a striking 75% following CGM initiation, and CGM drove a 64% decline in the percentage of patients experiencing a work absence. This sort of study would be a significantly bigger lift for Abbott, but we’d love to see broader outcomes, including health economics, included in these data sets, which would also make for a much more compelling story to pitch to payers.

• Shortly after Dr. Dunn’s talk, Abbott’s Dr. Sujit Jangam followed with a presentation of FreeStyle Libre data from high-risk individuals using the system over six months. The data set was populated by ~6,800 readers, ~67,000 sensors, and 21 million monitoring hours between September 2014 and May 2016. High-risk individuals were defined as individuals in the highest tertile of time >180 mg/dl (high hyperglycemia risk), and the same criteria was used for time <70 mg/dl and time ≤54 mg/dl to define high hypoglycemia risk. In all groups, scanning frequency started at ~20/day, and by two months, had decreased to ~15/day, which was sustained for the duration of the six months. In individuals with high risk of hyperglycemia, time >180 mg/dl decreased from ~54% of the day to ~50% of the day (-1 hour/day); Most of the changes occurred during the first two weeks of use and were sustained out to six months. The improvement was more drastic in higher- and medium-frequency scanners than in low-frequency scanners: high-frequency scanners saw a ~90-minute/day improvement in hyperglycemia vs.low-frequency scanners who saw a 45-minute/day improvement. Time <70 mg/dl in those at high hypoglycemia risk dropped from 14% of the day to 10% of the day (-45 minutes); Again, most of the improvements occurred in the first two weeks and were sustained out to six months. Interestingly, the improvement was similar between higher- and lower-frequency scanners. Last, time <54 mg/dl in those at high risk for hypoglycemia decreased from 6% to 4%, a 22-minute improvement that largely occurred within the first two weeks of use and was sustained to six months – we’d guess Abbott’s next-gen system with continuous communication (and presumably alarms) will further reduce what is still fairly significant dangerous hypoglycemia. Similar to previously-presented data showing scan frequency to be correlated to time ≤54 md/dl, the improvement was more marked in individuals that scan with higher frequency. The most surprising revelation from the analysis is that time <70 mg/dl is improved similarly in people at high risk of hypoglycemia regardless of scanning frequency.

Questions and Answers

Q: What is the least number to scan to have good results? We can say to the patient at least one per hour?

Dr. Dunn: In that eight to ten range, as with typical plateauing of relationship. But there’s a lot of variability, some people are more effective than others. Yes, one per hour would be pretty great.

Q: Any data regarding exercise?

Dr. Dunn: In theory, we don’t have that data yet. The device does have a logbook for insulin, food, and exercise, so we hope to address that.

Q: Does distribution of scanning throughout the day matter? If someone clustered in morning, compared to someone scanning every three hours, maybe adjust for that?

Dr. Dunn: Yes that falls in the category of more personalized analysis, what is working…we need to do that analysis, but overall we see people scanning mostly during the day, maybe one or two overnight.

Q: What is the cause of reduction in scanning over time?

Dr. Jangam: As with any new system, there’s a lot of excitement as people figure out the right mode of operation for them, but in our data, even though we start with 20 per day, even by six months, they’re still at 14-15 scans per day, which is still a lot of scans.

5. 9.5% One-Cal MARD for Senseonics Eversense in PRECISE II US Pivotal; Eversense Now Remote Monitoring App Cleared for Use in EMEA

During Roche’s symposium, Senseonics CMO Dr. Lynne Kelley shared an unpublished sub-analysis from the US PRECISE II pivotal study of the Eversense implantable CGM: Solid MARD of 9.5% vs. YSI, compared to 8.8% vs. YSI with two calibrations per day. The excellent two-cal accuracy in hypoglycemia (≤80 mg/dl) of 9.6 mg/dl is also largely maintained, with MAD rising just to 10.2 mg/dl with one-cal. Excluding hypoglycemia, MARD is 8.2% with two calibrations per day and 8.9% with one calibration per day. No zero-calibration data was broken out, though we’d be curious to see it, particularly since the sensor has 90 days during which the signal could potentially drift. Importantly, with just one calibration per day (as many users would certainly use it), Eversense still meets the much-publicized ~10% MARD recommended by Kovatchev et al. for non-adjunctive use. Eversense is now in 14 countries (13 in the EU) and still under FDA review, with an Ad Comm yet unannounced but previously expected in 1Q18. Notably, some patients in the real world are now on their seventh 90-day sensor and there have only been “single-digit numbers of complains of skin reactions in the commercial launch.”

• In a study conducted at LMC Healthcare in Toronto (n=36), an impressive 78% [CI: 65%-93%] of the 180-day Eversense XL sensors made it the full six months without having to be extracted. As can be seen in the Kaplan-Meier survival curve below, 95% of sensors were working at 90 days and 94% of sensors were working at 120 days – what presenter Dr. Alex Abitbol called “fairly robust performance.” Of the eight sensors that were removed prior to 180 days: One participant withdrew consent one day after insertion; One participant’s sensor was removed at 93 days due to connectivity issues; and there were a number of early sensor replacement alarms after day 130 (one on day 136, one on day 142, three on day 146, and one on day 165). For a first-to-market long-term implantable sensor in the real-world, this is a very strong performance, and we imagine it will only improve. According to Dr. Kelley, people actually forget they’ll need to change out their sensor, so Senseonics has added a phone notification at 150 days to remind patients it is time to make an appointment to change their CGM – talk about reduced burden!

• Dr. Abitbol wasn’t at liberty to share accuracy data from the LMC study in the form of MARD, but he could share the consensus grid error analysis (below). Compared to YSI, the sensor performed very well, with ~93% of readings falling in Zone A and ~6% in Zone B.

• CE Mark for Eversense Now, an app enabling remote monitoring, was expected shortly as Senseonics is “waiting on a daily basis” – and just came in as we were putting the final touches on this report! Eversense users in EMEA markets now have the option to invite up to five others to remotely view their real-time glucose readings and alerts. This feature will be a huge plus and will help appeal to the pediatric cohort. The updated Eversense app and the Eversense Now app will both be available for free on the App Store later this month (development for Android is in progress)
• Endocrinologists have now been trained on the insertion procedure at over 375 centers. Dr Steve Russell shared his early positive views last December at IDF. Each insertion, done by endocrinologists with little advanced procedural training, reportedly takes just 5-10 minutes and it generally takes one to three tries for a doctor to become comfortable with the procedure. This aligns with what we have heard from KOLs in the field who have complimented the simplicity and accuracy of Eversense. Scaling manufacturing will be key for the company to demonstrate – our sense is the launch is still fairly limited.

6. FDA’s Dr. Stayce Beck on Advances in CGM, device interoperability (“key to patient choice”), label claims, and accuracy

FDA’s Dr. Stayce Beck shared valuable perspective on Advances in CGM technologies, including product claims, accuracy, and interoperability. The latter was a focus of the talk, reiterating FDA’s vision to have interoperable components that would allow plug-and-play AID systems. “The anticipated pace of AID innovation challenges the current regulatory framework for medical devices. Every time one of the components is modified, it has to come to FDA with a new submission. Users can’t mix and match systems that meet their needs…” She highlighted how lack of device communication and partnership contracts (e.g., between pump and CGM companies) have really slowed down the field – something FDA hopes to address with more interoperability. Dr. Beck added that “the key to patient choice is interoperability” and “interchangeability of AID components with an AID system would enable innovation and more rapid AID commercialization.” There’s no question FDA is sold on this concept, but the specific path forward remains a bit fuzzy, now 1.5 years after Dr. Lias first introduced this idea at the 2016 D-Data Exchange. Dr. Beck did allude to some possible progress: “we think we’re at a point where there is a technical solution to help with this” and “FDA is working on facilitating active discussion to openly solve these challenges.” These could imply a public meeting will happen at some point, a set of communication standards that will be recommended, or both. This would be amazing for patient choice, in line with JDRF’s open protocol initiative. As noted above, Roche announced it is IN on this JDRF initiative, a lead we hope other companies follow. Other insights from Dr. Beck:

• “We don’t care how the device works, we just care that it does work and it does what it claims to do. We look at inputs and outputs, meaning it can be a black box.”
• “FDA is agnostic on device claims – we want advancements so that the American public can benefit. We even want to help you get there.”
• “How do I get to say that my device can do ___? Do a risk analysis, design a study to show that your device can do ___. Complete the study in the population who will use your device. Have data to support that your device can do ___. Submit to the FDA.”
• “Risk analysis: what are the potential failures? What could go wrong, in an imperfect world, from both a tech and user perspective? What is the root cause of these failures? What are the potential effects of the failure? For instance, if I want CGM data to go to a phone, what happens if get a call? What if I close out of the app? What if I’m playing a game? What happens if a person thinks it is working? What happens if the phone is on mute, and they thought they would still get alarms from CGM? What can be done to mitigate the risks or prevent the failures?”
• In terms of CGM accuracy, FDA focuses on “agreement.” Compared to the lab based comparator method, what percentage of CGM data points are within 15% or 20% of reference? Dr. Beck showed the Dexcom G5 label as an example, where 93% of points are within 20%. FDA gets concerns when the values are very different and could be dangerous – for instance, when the CGM reads 40-60 mg/dl and the lab comparator reads over 300 mg/dl. “This means 7-8 times in this study, the CGM indicated the person was low, but in fact was very high. In those situations, we’d investigate further. Was this only one person, or did this happen to multiple people? We use all that information to decide how to approve a device.”
  
  • Dr. Beck encouraged developers to look at approved devices’ Summary of Safety and Effectiveness Data” (SSED) for a good sense of what FDA really cares about.

7. Extended iHART CGM: Nearly Two Hours Less Time <70 mg/dl Per Day in FreeStyle Libre Users Who Switched to Dexcom G5 for Eight Weeks

Imperial College of London’s Dr. Monika Reddy presented extended data further supporting the results from the head-to-head iHART CGM study comparing the Dexcom G5 and Abbott FreeStyle Libre in type 1 adults with impaired hypoglycemia awareness. Following the eight-week trial randomizing patients to receive either the G5 or FreeStyle Libre, all participants were given the option to continue with the Dexcom G5 for an additional eight weeks. 16 patients originally on the G5 and 20 patients originally on FreeStyle Libre chose to continue with the G5. Results showed that those who switched from FreeStyle Libre to the G5 significantly decreased time <70 mg/dl from 11% to 4% (-1.7 hours/day), while those initially on the G5 showed no significant difference in time in hypoglycemia, decreasing slightly from 6% to 5% (-11.5 minutes). Those who switched from FreeStyle Libre also reported significantly improved time-in-range, increasing from 60% to 67% (+1.8 hours/day). Again, those who were originally on the G5 did not show significant changes in time-in-range. There were no significant changes in A1c over the extended eight-week period for either group, indicating that glycemic control was not ceded for the hypoglycemia improvements. The data serve to further confirm the results of the original iHART CGM study, suggesting that for those with impaired hypoglycemia awareness, G5’s alarms can be incredibly beneficial in reducing hypoglycemia and improving time-in-range. As we noted in our initial coverage of the eight-week study results, head-to-head device trials like this are very valuable and will help the field determine recommended devices in sub-populations, especially as great CGM options continue to proliferate.

Pump Highlights

1. Immediate Pump Therapy Initiation Results in 1.1% Lower A1c in Four-Year Observational Study

Ms. Deborah Foskett (Insulin Pump Angels, Australia) presented compelling data indicating significant A1c reductions in children that initiate pump therapy within one month of their type 1 diabetes diagnosis (n=38) as compared to those who waited one year (n=37). At the end of the 48-month study, early initiators reported significantly lower A1c (6.8%) as compared to late initiators (7.9%). Specific baseline A1c values were not provided, but from the graph below, it appears the early initiators demonstrated substantially higher baseline A1c values, which very likely impacted the subsequent decrease. Still, given that the results were maintained over four years, Ms. Foskett makes a good point that immediate initiation of pump therapy offers tremendous potential value in children with type 1 diabetes. We wonder how the results would look for CGM or AID initiated within month of diagnosis! There were no differences in the rate of severe hypoglycemia events between groups, nor were there significant differences in quality of life between groups, with both early and late adopters reporting high satisfaction. Of course this study is not free from confounds – i.e., what kind of patient/family waits to initiate pump therapy vs. starting right-off-the-bat – but it is a useful data point.

2. Pickup Meta-Analysis Shows ~0.4% A1c Benefit and ~24 Units/Day Insulin Dose Benefit with Pump vs. MDI in Type 2s; Greatest Benefits in Type 2s with High Baseline A1c/Insulin Dose

Dr. John Pickup reviewed findings from his now-published meta-analysis on pump therapy in type 2 diabetes (Diabetes Care, May 2017), showing favorable reductions in A1c (~0.4%) and insulin dose (~24 units/day) as compared to MDI. These final results build on the preliminary data shared at last year’s ATTD, which was comprised of five studies randomizing 590 patients in total on MDI to either continue MDI or switch to an insulin pump. Mean A1c treatment difference was ~0.4% in favor of pumps (95% CI: 0.86%-0.05%), though Dr. Pickup underscored that the greatest benefit appears in people with higher baseline A1c. In the OpT2mise trial (n=331), mean A1c at baseline was 9%, and individuals in the pump arm experienced 1.1% A1c decline over six months vs. 0.4% in the MDI arm (estimated treatment difference of 0.7%, p<0.0001). In contrast, baseline A1c in Raskin et al. and Herman et al. was only 8.1% and 8.3%, respectively. In Raskin et al., A1c dropped 0.6% with pump vs. 0.5% with MDI after six months; in Herman et al., A1c dropped 1.7% with pump vs. 1.6% with MDI after one year, for a similar estimated treatment difference of ~0.1% (noticeably smaller than the 0.7% difference in OpT2mise). Dr. Pickup further highlighted the A1c optimization period in OpT2mise, which ensured that only people poorly controlled on MDI (A1c between 8%-12%) were randomized – these individuals thus experienced minimal additional benefit from continuing MDI and likely contributed to the larger estimated treatment difference. In RCTs without optimization, participants randomized to MDI continued to see some A1c-lowering over the course of the study. Mean insulin dose in Dr. Pickup’s meta-analysis was reduced by 0.25 units/kg with pump vs. MDI (95% CI: 0.31-0.19), and by 24 units/day (95% CI: 30.55-17.45). Once again, switching to pump therapy produced greater benefits in people with higher baseline insulin dose. Dr. Pickup described a ~21 units/day treatment difference between pump vs. MDI for a baseline insulin requirement of 90 units/day, which he said grows to a ~36 units/day treatment difference – “a tremendous improvement” – for those with a baseline insulin requirement of 150 units/day. He pointed out that most diabetes care providers probably do see a high proportion of type 2 patients with insulin requirements around 150 units/day, suggesting that there are many good candidates for pump therapy in the real-world type 2 diabetes population. There were no significant differences in body weight or BMI across all five studies in Dr. Pickup’s meta-analysis. The overall estimated treatment difference on body weight was 0.08 kg (0.18 lbs) with pump vs. MDI (95% CI: -0.33 kg-+0.48 kg), and for BMI was 0.00 kg/m² (BMI data was only available from four RCTs, not including Wainstein et al.).

• Dr. Pickup also showed how pump therapy is cost-effective compared to MDI for people with type 2 diabetes and high baseline A1c. He pointed to a study conducted in The Netherlands, which reported an ICER of 18,610 euros for a pump vs. MDI in a patient with A1c of 10%. ICER rises to 35,837 euros for someone with A1c of 9.5%, and to 62,895 euros for someone with A1c of 9% (more expensive than the typical 50,000 euro per QALY threshold endorsed by many payers). We found this to be quite compelling sub-population analysis, and in terms of expanding access to insulin pumps for the type 2 patient population, we have to start somewhere – pumps should be duly considered for individuals with higher starting A1c/insulin dose, who not only stand to benefit the most from switching to pump therapy, but also likely contribute the most to healthcare costs. That said, Dr. Pickup emphasized that various health systems around the world will approach the cost of insulin pumps in type 2 differently. How will the more cost-conscious healthcare systems view pump reimbursement, and will pumps be covered indefinitely or only until patients achieve some target A1c where cost-effectiveness may diminish? For now, this remains an open question. Cheaper disposable (yet fully-featured) pumps developed for the type 2 population like Insulet/Lilly’s U500 and U200 Omnipods as well as BD’s Swatch patch pump may help make a dent in costs while retaining benefits like connectivity. The type 2 pump market is severely underpenetrated, with other type 2 pumps/patch insulin delivery devices like Valeritas’s V-Go, Cequr’s PAQ, and J&J’s bolus-only OneTouch Via slow to roll out (the latter two haven’t launched at all, despite many years of development). Competitive pricing, outcomes, and cost analyses will continue to be key in getting insurance companies on board.
• The meta-analysis covered Raskin et al., Herman et al., Wainstein et al., Berthe et al., and Reznik et al. (OpT2mise).

3. Pump Therapy Has ICER of 47,834€/QALY for Type 2 Patients in Finland (Based on OptT2Mise)

Medtronic’s Dr. Alexis Delbaere presented fascinating results from a simulated cost analysis study determining the cost-efficacy of pump therapy for type 2 diabetes patients in Finland. With a 0.32 improvement in QALY, the incremental cost-effectiveness ratio of pump therapy was shown to be 47,834€/QALY, allowing pump therapy to just squeak in under the bar for cost-effectiveness in Finland, which reports an acceptable willingness to pay of ~50,000€/QALY. The IQVIA Core Diabetes Model was used to calculate diabetes-related complication incidence and associated costs, including treatment and productivity, for pump and MIDI users. Clinical data were derived from the randomized controlled trial OpT2mise (n=331), which showed pump therapy (n=168) to result in a significantly greater A1c reduction of 1.1% (baseline: 9.0%) at six months as compared to MDI (n=163), which resulted in an A1c reduction of 0.4% (baseline: 9.0%). Diabetes-related complication and intervention costs were sourced from Medtronic (the MiniMed 640G pump was used) and from published and official tariffs specific to Finland. Not surprisingly, the cost analysis revealed pump therapy to be roughly twice as expensive as MDI, with higher annual direct costs partially offset by savings due to expected reductions in diabetes-related complications. Direct treatment cost for pump therapy totaled 3,749€ in the first year of use, decreasing slightly to 3,494€ in the following years due to reduced need for patient training. MDI annual costs were estimated to be stable at 1,806€. Compared to MDI, pump users saw estimated cumulative incidence reductions after five years for eye diseases (-24%), renal diseases (-28%), ulcers/amputations (-17%), and cardiovascular diseases (-5%). In fact, average lifetime costs/patient for complications and management totaled 51,000€ for pump users and 60,000€ for MDI patients – a 15% reduction. We find such cost analyses incredibly important, especially in the type 2 population, and we wonder if they will serve to convince payers to expand pump reimbursement in type 2s with high baseline A1c’s – particularly given the monthly cost of expensive type 2 diabetes drugs. This observational type of study raises the question of whether or not the type of patient who would end up on a pump would have similarly low costs had he/she stayed on MDI – ultimately, this is a question for RCTs and crossover studies to answer, though they are not without their pitfalls too.

Drug & Nutrition Highlights

1. Sanofi’s Real-World LIGHTNING Study: Toujeo Significantly Reduces Severe Hypoglycemia by >60% vs. Lantus and Levemir, On Par with Tresiba

Results from Sanofi’s real-world LIGHTNING study were presented on a poster and during a corporate symposium at ATTD. The company’s next-gen basal insulin Toujeo reduced severe hypoglycemia risk by more than 60% vs. Lantus (p=0.009) and vs. Levemir (p=0.002), while severe hypo rates were similar with Toujeo vs. Novo Nordisk’s next-gen Tresiba (p=0.370). LIGHTNING analyzed electronic medical records of 10,458 patients with type 2 diabetes who switched from one basal insulin to another. After propensity score matching, the mean estimated event rates of severe hypoglycemia per 100 patient-years were 3.6 for Toujeo vs. 9.7 for Lantus (p=0.009), 3.6 for Toujeo vs. 15.1 for Levemir (p=0.002), and 3.4 for Toujeo vs. 5.3 for Tresiba (p=0.370). A1c-lowering was similar across all three comparisons, ranging from -0.50% to -0.89% (baseline 9.1%-9.2%), indicating that reductions in hypoglycemia did not occur at the expense of greater hyperglycemia. Our key takeaway from this data is that the hypoglycemia benefit associated with next-generation basal insulins (both Toujeo and Tresiba) translates from RCTs to the real world – this is exciting for patients, and should hopefully be compelling to payers as well. During Sanofi’s accompanying symposium on basal insulin, UCSD’s Dr. Jeremy Pettus commented that real-world studies add confidence to the scientific vigor of RCTs, demonstrating that clinically-meaningful outcomes are maintained outside the structure of a randomized controlled trial. Dr. Pettus expects to be seeing more of these studies, and indeed, Sanofi has launched a robust real-world evidence campaign around Toujeo, which we learned about in-depth at EASD 2017. The company issued a press release at the beginning of ATTD to call attention to this real-world read out.

• LIGHTNING hints that Toujeo has real-world impact comparable to Tresiba, although head-to-head RCTs reporting later this year will shed additional light on the relative advantages/disadvantages of the two advanced basals. From a baseline 9.2%, A1c declined by a mean 0.8% for people switching to Toujeo vs. 0.89% for people switching to Tresiba (p=0.591), and hypoglycemia frequency was similar at 3.4 events/100 patient-years vs. 5.3 events/100 patient-years, respectively (p=0.370). Sanofi’s BRIGHT study compares Toujeo vs. Tresiba head-to-head in a randomized controlled setting, and thus far has shown a similar proportion of patients (~50%) reaching A1c goal <7% after 24 weeks (baseline A1c 8.6%-8.7%). The hypoglycemia findings from BRIGHT aren’t yet available, but we’ll be looking for these at ADA 2018. Novo Nordisk has its own head-to-head RCT ongoing with hypoglycemia as the primary endpoint, and this trial is expected to complete in October 2018. Of note, Tresiba commercially outperformed Toujeo by a fair margin in 2017 (Novo Nordisk’s product sales growing 80% YOY to $1.1 billion, Tresiba’s product sales rising 26% YOY to $926 million). We’ve heard some thought leaders endorse Tresiba as a flatter basal insulin vs. Toujeo overall, with greater flexibility around missed doses, based on their clinical practice. Regardless, we want to emphasize that both these next-gen products offer substantial advantages to patients above and beyond what was available before, as evidenced by Toujeo’s highly-significant hypoglycemia benefit vs. Lantus and Levemir in this real-world trial.
• As with any real-world study, there are some limitations to LIGHTNING, most notably that previous basal dose and reasons for switching weren’t available from the Humedica database. Data was gathered on patients between April 1, 2015 to December 31, 2016 from >50 healthcare systems. LIGHTNING defined severe hypoglycemia as inpatient or emergency room visits due to low blood glucose (≤70 mg/dl or coded ICD-9 or 10). These results corroborate DELIVER-D findings, from a smaller (n=1,620) real-world study that looked at patients switching from Lantus specifically to Toujeo or Tresiba.

2. Dr. Eran Segal’s Personalized Nutrition Project: Microbiome-Based, Predictive, Individualized Diet Advice Based on CGM Diet Study

Weizmann Institute’s Dr. Eran Segal provided a fascinating overview of his work on the Personalized Nutrition Project, an initiative designed to uncover personal nutrition recommendations using insights from the microbiome. From a data science perspective, this research has been a massive undertaking: Across 800 study participants, Dr. Segal’s group used CGM data to measure glucose response to nearly 50,000 meals. From there, the researchers developed a machine-learning algorithm that integrated the CGM data, dietary intake, anthropometrics (BMI, A1c, etc.), and microbiome composition to produce a prediction of each individual’s post-prandial response to a different meal. The researchers then designed an individualized “good” and “bad” diet for 100 of the original participants based on the algorithm’s assessment of their microbiomes. In a blinded CGM study, the algorithm-recommended diet resulted in significantly lower post-prandial glucose responses – see the very cool TED Talk here and the 2015 Cell paper here. Although this research was conducted in a non-diabetes population, we see big potential implications for this predictive microbiome-based diet to minimize post-prandial excursions in people with type 1 and 2 diabetes alike, as well as the prediabetes population. To this end, Dr. Segal’s predictive algorithm was recently licensed to DayTwo, which makes an app that provides personalized decision support and nutrition advice for blood sugar management based on customers’ microbiome samples. Last February, DayTwo and Dr. Segal’s lab teamed up with the Janssen Human Microbiome Institute to conduct further translational research on the gut microbiome, with the ultimate goal of refining DayTwo’s app to make even more actionable microbiome-based nutritional guidelines. While we believe that both the novelty of the field and complexity of the gut microbiome limits how quickly we will see any concrete microbiome-based therapies, we are intrigued by its potential in the near-term to inform this kind of decision support technologies paired with CGM and other biometric data – an area of high unmet need in the realm of nutrition, where we still have so very much to learn.

ATTD Yearbook

The 2018 ATTD yearbook session shared 12 chapters, covering the year of publications in diabetes technology and therapeutics. Dr. Tadej Battelino noted that last year’s version received 12,000 downloads, led by three popular chapters: digital health, CGM, and new insulins. See summaries of each chapter below, or get the full yearbook here.

Continuous Glucose Monitoring in 2017

• Dr. Bruce Bode reviewed the “busy year” in CGM studies, covering DiaMonD and REPLACE-BG. He reviewed the type 1 (JAMA 2017; ADA 2016), type 2 (Ann Int Med 2017; ATTD 2017), and pump extension (Lancet D&E; ATTD 2017) phases of DiaMonD, noting that MDIs can benefit from CGM regardless of diabetes type or level of education. In particular, Dr. Bode emphasized the A1c and time-in-range benefits in both DiaMonD cohorts; the ~3 SMBGs/day in DiaMonD at baseline; and the very strong compliance with CGM in this study. He also covered the REPLACE-BG study, which confirmed the safety of non-adjunctive use of Dexcom CGM. The CGM chapter includes 16 total articles and can be found here.

Closing the Loop

• Harvard’s Dr. Eyal Dassau reviewed the closing the loop chapter, spending the most time on the landmark NEJM 2016 study testing Cambridge’s system in pregnancy for up to 3+ months, including during labor – read our coverage of that here. Dr. Dassau was very impressed with the results (as we were), noting the outstanding reduction in variability (see the compelling modal day profiles here): “This is what technology can do across the board.” He noted the importance of bringing closed loop to broader populations – especially pregnancy – as they are often excluded in current studies (particularly in the US). The chapter also includes several other major papers in the closed loop field, including the 670G pivotal study outcomes in adults/adolescents (JAMA 2016 and DT&T 2017); three Cambridge team studies, including a compelling one in type 2 diabetes that we first saw a couple years back (Diabetes Care 2016, Lancet D&E 2017, Lancet D&E 2017) Bionic Pancreas multicenter study (Lancet 2016), among others. The Closing the Loop chapter includes 15 total articles and can be found here.

Using Digital Health Technology to Prevent and Treat Diabetes

• Dr. Neal Kaufman reported on the digital health chapter – the most downloaded of the previous year – which reviewed 19 articles. Even before delving into discussion of data, Dr. Kaufman expressed concern with the types and quantity of articles that were available for review; There were “only three intervention studies worth sharing…not a good idea,” along with 10 systematic reviews, which detailed outdated studies performed ~five years ago. And echoing a debate the field has been negotiating of late, Dr. Kaufman pointed out that there are over 15,000 diabetes apps out there, with very few researched. “The evidence for effectiveness and generalizability to the larger population is not really there. As you can imagine, those trials are difficult. Yet there were a number of analyses conducted and published in the past year, showing “modest” and “not really prolonged” improvements in glycemic control, physical activity, and weight loss. Interestingly, Dr. Kaufman highlighted TEXT4DSM, which implemented a texting intervention in three diabetes programs that existed in three low-resource countries (the Democratic Republic of Congo, Cambodia, and the Philippines). The study was large (n=1,440 patients, presumably with type 2 diabetes) and the text intervention included messages pertaining to healthy eating, physical activity, medications, foot care, and more, but did not result in a difference in the proportion with well-controlled diabetes after two years (although there was a decrease in foot wounds in the experimental group). He suggested that the lack of positive impact “implies caution about mHealth outcomes,” but acknowledged a number of possible reasons for no impact. On the one hand, we found it curious that the one study Dr. Kaufman chose to feature used a relatively simplistic intervention and had negative results, but on the other hand, this may reflect the dearth of randomized, controlled, peer-reviewed outcomes in this area. We expect the slew of newer developments, such as personalized coaching and insulin titration, will slowly make its way into the literature and change this summary meaningfully in the years to come.  Download the full chapter here.
  
  • Our team’s Clinical Diabetes article on challenges in digital health was the first article reviewed in the chapter. The paper was written by Varun Iyengar (a Close Concerns Associate alum), Alex Wolf (a diaTribe alum), Adam Brown, and Kelly Close

Insulin Pumps

• Reviewing pump therapy, Dr. John Pickup identified six themes from the past year: (i) Comparative effectiveness of MDI vs. pump therapy (are pumps as beneficial as some had traditionally thought?); (ii) Factors that determine effectiveness of pump therapy (further evidence that poor control on pump therapy is not associated with fear of hypoglycemia, but psychological issues); (iii) The role of pumps in type 2 diabetes (a meta-analysis from Dr. Pickup’s group shows pumps are most effective in those with highest baseline A1c and insulin dose; see above); (iv) safety and reliability of pump therapy (surveys show reliability of modern pumps continues to be a concern, and SAP may be the most unreliable); (v) ultra-fast-acting insulin in pumps (potential for improved control); and (vi) PLGS pumps (evidence for reduction in hypoglycemia, including after exercise, though longer-term studies in adults with severe hypoglycemia are still needed). He dug in deeper on the first theme, pointing to the somewhat controversial REPOSE results suggesting that glycemic outcomes in type 1 patients were no different on pumps and injections (though satisfaction was higher in the pump group) – see our coverage from IDF 2015 here, including an insightful interview with Dr. Irl Hirsch. There are a host of positive and negative factors to consider, said Dr. Pickup: On the positive side, the REPOSE study was robust (large, and long) and subjects were not highly-selected, but on the other hand, the study had a small number of patients with severe hypoglycemia, A1c difference was significantly different for those with baseline A1c ≥8.5%, there was a large center difference in pump efficacy, etc. He also pointed to the DIaMonD type 1 pump extension phase (Lancet D&E; ATTD 2017), which found that glycemia is further improved by switching to pumps in type 1s already on CGM, though biochemical hypoglycemia was increased. The whole pump chapter can be downloaded here.

Self-Monitoring of Blood Glucose

• Kicking off the packed ATTD Yearbook session, Dr. Irl Hirsch reviewed the “important, but older technology” that is SMBG. A study published in JAMA Internal Medicine looked at the policy impact in Ontario, Canada of limiting test strips in 2013: 400 strips/year for those receiving drugs known to cause hypoglycemia, 200 strips for all other patients, and 3,000 strips/year for those using insulin. The time series analysis (2008-2015) showed the policy reduced costs by 20% (24 million Canadian dollars), without a negative impact on emergency department visits or A1c (JAMA Int Med, 2017). Dr. Hirsch questioned whether the costs of SMBG will continue to decrease in both high- and low-income countries, especially given these Canadian results. More importantly, and Dr. Hirsch’s “big concern,” he wondered how strip quality might be affected if SMBG use declines alongside increasing CGM penetration. Dr. Hirsch detailed a Swedish study (BMJ Open Diabetes Res Care 2017) surveying 314 type 1 patients not using CGM. Only 44% of respondents indicated that they take a fingerstick four or more times/day, leaving more than half of the participants with an SMBG frequency below the recommended rate – no surprise there, given T1D Exchange data and dQ&A data (see Adam’s presentation from day #1). Top reasons for not testing included forgetting, a lack of time, and self-consciousness. Unfortunately, Dr. Hirsch believes SMBG frequency is unlikely to improve. The SMBG chapter reviews 14 key articles and can be found here.

Diabetes Technology and the Human Factor

• Dr. Alon Liberman summarized 2018’s “Diabetes Technology and the Human Factor” chapter, touching on two themes: locus of control and barriers to device use. “Locus of control” refers to whether individuals believe that they have control over the outcomes/events in their lives. When applied to diabetes, people with an internal control orientation believe that success or failure are attributable to their own efforts. A 2017 study in Diabetic Medicine found that internal locus of control “plays a significant role in achieving tight glucose control” in type 1 adults on pumps. The correlations in the actual paper are not that strong (r=0.18 or less), though they are statistically significant and do link locus of control to A1c and severe hypoglycemia. Dr. Liberman also reviewed two psych studies on devices, one on barriers to uptake (Tanenbaum et al., Diabetes Care 2016) and another on clinicians’ views on barriers to adherence (Tanenbaum et al., JDST 2017). We liked some of the questions included in these studies, which serve as a bar for devices to meet: “Diabetes technology has made my life easier”, “Diabetes technology has made managing my health easier”, and “I am lucky to live in a time with so much diabetes technology.” This chapter includes 11 total articles and can be found here.

Technology and Pregnancy

• Dr. Jennifer Yamamoto (filling in for Dr. Helen Murphy) emphasized that there is still substantial room to improve glycemic control in pregnant women with preexisting diabetes. She reviewed a study (Diabetologia 2017) analyzing 2015 national audit data (n=3036 women) showing only 14% of those with type 1 diabetes and 37% of those with type 2 diabetes meet the A1c target of <6.5%. Pregnancy preparation remains less than ideal, with high numbers of complications: Prevalence rates for congenital anomalies (46.2/1000 births for type 1, 34.6/1000 births for type 2) and neonatal death (8.1/1000 births for type 1, 11.4/1000 births for type 2) were unchanged as compared to 2002/2003. However, stillbirth rates were nearly 2.5-times lower, dropping from 25.8/1000 births to 10.7 and 29.2/1000 births to 10.5 for type 1 and type 2 mothers, respectively. On a more positive note, Dr. Yamamoto detailed Dr. Murphy’s study investigating closed loop insulin delivery during pregnancy in women with type 1 diabetes (NEJM 2016; read our coverage). She found it particularly telling that 14 out of the 16 participants chose to continue on closed loop therapy throughout their pregnancy and noted the system performed “remarkably well.” There were no hypoglycemia events lasting over 20 minutes either 24 hours prior to delivery or 48 hours post-delivery. Women reported 87% time-in-range 24 hours prior to delivery and 74% time-in-range 48 hours post-delivery. Dr. Yamamoto was particularly excited by the ability for closed loop to adapt to the changing demands of pregnancy, including dynamic insulin resistance, glucose utilization, steroid use, and labor and delivery. The technology and pregnancy chapter reviewed 10 articles and can be found here.

Diabetes Technology and Therapy in the Pediatric Age Group

• Dr. David Maahs gave a crisp review of five pediatric diabetes trials published in the past year, demonstrating steady progress in diabetes tech (from PLGS to day-night closed loop in poorly controlled adolescents and young children), but emphasizing that there is still more work needed to bring these benefits to more children worldwide. He began with Battelino et al.’s 14-day PLGS study in children with type 1 diabetes, which showed that predictive insulin suspension resulted in fewer hypoglycemia events, but at the expense of more time spent >140 mg/dl. Spaic et al.’s follow-on paper showed that addition of predictive hypoglycemia-hyperglycemia minimizer to PLGS is safe, feasible, and effective overnight (increasing time-in-range and lowering mean glucose). Next, Tauschmann et al. demonstrated that day-and-night hybrid closed loop therapy in adolescents resulted in higher time in range, lower mean glucose, and less hypoglycemia in adolescents with type 1. Dr. Maahs again cited Dr. Tauschmann et al.’s paper describing free-living home use of 24-hour hybrid closed loop in sub-optimally controlled adolescents with type 1 to be safe, feasible, and glycemia-improving (without a concomitant increase in hypoglycemia). And lastly, Del Favero et al.’s camp study investigating closed loop in 5-9-year-olds showed that closed loop is feasible and safe, though a 3x decrease in time-in-hypoglycemia was counterbalanced by a decreased time-in-range and increased mean glucose. Download the full chapter here.

Advances in Exercise, Physical Activity, and Diabetes Mellitus

• As Dr. Michael Riddell astutely noted, the fitness and diabetes fields are beginning to merge. Dr. Riddell found a whopping 1,334 papers on exercise and diabetes in the yearbook timeframe, but limited himself to ten for the ATTD Yearbook chapter on advances in exercise, physical activity, and diabetes mellitus. In one study (DTT 2017), circuit-based exercise was found to be less risky than continuous exercise, resulting in a less-dramatic drop in blood glucose. However, Dr. Riddell cautioned that there is still substantial individual variability, with certain participants showing a rise in blood glucose following higher-intensity circuit-based exercise. He reviewed a second study (Diabetes 2017) suggesting that high-intensity interval training (HIIT) reduces hypoglycemia awareness in type 1 diabetes patients following exercise. He found these results “a little bit alarming,” creating yet another perfect storm for people with type 1 diabetes to deal with. Interestingly, participants scored higher on a cognitive performance test following HIIT – even more intriguing, the difference in cognitive score was found to be significant only in the sub-group with type 1 diabetes and no history of impaired hypoglycemia awareness. Exercise will continue to be a hurdle for closed loop systems, given the extreme inter- and intra-patient variability, though we’re optimistic about decision support and smarter algorithms here – since most people to the same types of exercise, pattern recognition should be possible.

New Insulins, Biosimilars, and Insulin Therapy

• Dr. Thomas Danne presented on faster, longer, safer, and biosimilar insulins. First, faster. He highlighted Novo Nordisk’s Fiasp (faster-acting insulin aspart) as the first-in-class ultra-rapid mealtime insulin. Although pre-meal injection leads to the best outcomes, Dr. Danne alluded to the fact that Fiasp can be injected up to 20 minutes after you start eating (it’s the first injectable bolus approved without a pre-meal dosing recommendation, which affords extra flexibility around meals). Next, longer. Dr. Danne mentioned Novo Nordisk’s Tresiba (insulin degludec) and Sanofi’s Toujeo (insulin glargine U300), pointing to a CGM study (from Dr. Rich Bergenstal) that confirmed Toujeo’s flatter profile vs. Lantus. He said a direct head-to-head comparison of the two advanced basals is “highly-desirable,” so that HCPs can better understand the differential advantages/disadvantages to decide which insulin is right for which patient. These head-to-head RCTs are coming – Sanofi’s BRIGHT and a Novo Nordisk-sponsored comparison trial – and could be part of next year’s ATTD Yearbook. Third, safer. DEVOTE provided reassurance around Tresiba’s CV safety, in addition to showing hypoglycemia benefit vs. Lantus. Lastly, biosimilars. Compared to other disease areas, diabetes is behind in bringing biosimilars to market, according to Dr. Danne. That said, 2017 was an important year for biosimilar insulins, and the Yearbook features studies of Lilly/BI’s Basaglar vs. Lantus as well as the SORELLA studies of Sanofi’s Admelog vs. Humalog. The “new insulins” chapter includes 17 total articles and can be found here.

New Medications for the Treatment of Diabetes

• Dr. Satish Garg discussed new medications for the treatment of diabetes that emerged this past year, including basal insulin/GLP-1 combos and SGLTs for type 1. The new fixed-ratio combination class, comprised of Sanofi’s Soliqua and Novo Nordisk’s Xultophy, saw “modest uptake” in its first full year on the market. In the words of Dr. Garg, these products are “making their way,” though volume/sales trended below expectations in 2017. He forecast that SGLT inhibitors may become an option for people with type 1 diabetes by next year, pointing to inTandem3 (Sanofi/Lexicon’s sotagliflozin will be filed with FDA in the “very near future”) and DEPICT 1 (AZ’s phase 3 study of dapagliflozin in type 1), which both read out at EASD 2017. Dr. Garg focused much of his short presentation on MannKind’s Afrezza, and this particular Yearbook chapter includes numerous papers on the inhaled mealtime insulin (Postgraduate Medicine 2016, Cardiology in Review 2017, World Journal of Diabetes 2016, etc.). Dr. Garg reviewed the FDA’s label update to Afrezza to reflect an ultra-rapid-acting profile, which came through in October 2017, and he shared an optimistic outlook on how this product could improve postprandial glucose control in people with diabetes. The full chapter on new medications, highlighting 23 studies, is available here.

Immune Intervention for Type 1 Diabetes

• Dr. Desmond Schatz highlighted a 2017 study on type 1 diabetes immune intervention, emphasizing efforts to identify responders to immunotherapy and define immune markers of type 1. Malmegrim et al. found that immunological balance of autoreactive and regulatory T cells was associated with C-peptide levels following autologous hematopoietic stem cell transplantation (AHSCT). In this study (n=21), longer duration of insulin independence after AHSCT was associated with higher regulatory T cell counts, and the subgroup of patients with lower frequencies of autoreactive islet-specific T cells had superior responses to stem cell transplant (remained insulin-free for longer, presented with higher C-peptide). Dr. Schatz discussed the concept of endotypes, the different potential disease mechanisms that can be associated with the phenotype of type 1 diabetes, and he explained how research is now moving to identifying these. Better understanding these etiologies should enable personalized immune therapy, as researchers continue piecing apart the immunological mechanisms leading to type 1 diabetes. The Immune Intervention for Type 1 Diabetes chapter includes six articles and can be found here.

Exhibit Hall

Abbott

Not surprisingly, Abbott’s large, welcoming booth advertised FreeStyle LibreLink, LibreLinkUp, and LibreView, all now available in 12 EU countries for Android and iOS (launched just before ATTD). It was great to see the app focus in an Abbott booth, which has historically shown only readers. Representatives were unable to provide updates on when we might expect to see these features in the US, nor could they comment on a potential pediatric indication for the FreeStyle Libre in the US (it is already available for pediatrics in the EU). A non-randomized, single-arm study evaluating the safety and effectiveness of FreeStyle Libre in pediatric populations (4-17 years-old) is listed as currently recruiting on Clinicaltrials.gov. It’s likely this study will serve as the basis for a pediatric FDA submission, in which case Abbott likely won’t launch in pediatrics until late 2018.

Ascensia

The Ascensia booth highlighted the Contour Next One BGM and new features of the paired Contour Diabetes app, including “my patterns” pattern detection, test reminders, and Contour Cloud compatibility (patients can now sync their glucose results with Glooko so that providers can view them). All of these features are already live in Germany, and to arrive in other markets by the end of April (no comment on US timing). The reps were also not able to comment on the Voluntis partnership, which was previously expected to launch in 4Q17, though the companies are still working on it.

BD

At BD’s Type 2.0 Lab, attendees voted on which areas of diabetes technology warranted prioritization. A screen in the booth displayed real time results. We stopped by fairly early in the meeting, but at the time a tube-free insulin delivery device was in first place, holding 21% of votes – great news for BD! Dose capture was in second place (17% of votes) and flexible dosing was in third (15% of votes). As expected, booth representatives were unable to comment on the current status of BD’s infusion set partnership with Medtronic, international plans for BD’s type 2 smart patch pump Swatch, or news on the decision to halt the smart pen needle program (per the company’s 4Q17 call).

Biocorp

Biocorp, whom we first wrote about upon their receiving a CE mark for their Easylog smart pen attachment in December, divulged expectations to launch in the US this summer and in the EU by the end of the year, both with partners. The rep demoed the device, which consists of an oblong piece that clips on to the top of the pen, and a circular piece that fits over the control dial/button (see pictures below). Dose capture is accomplished simply by tracking the turn of the pen’s dial. The rep further indicated that it will cost <$50, be compatible with all major pens (including Lilly, Sanofi, and Novo Nordisk’s), and last for two years. Further, because it will launch with partners, the paired app that we wrote about in December is more of a placeholder that can be replaced or augmented by proprietary software from the partner company. If the technology is deemed accurate and durable (Biocorp is still rather unknown to us), then it might be in a device or pharmaceutical company’s best interest to acquire Easylog.

Biomicro

Singapore-based Biomicro is developing a 12-mm (the size of a grain of rice) implantable sensor/stimulation platform that can be used for a number of different applications, including neurostimulation (for pain) and glucose sensing. A poster boldly touted “ultra-low powered implantable CGM” with factory calibration and a 180-day lifetime. The device would be implanted under the skin, transmit to a reader (large – about the size of a deck of cards at this point) that sits on top of the skin and powers the sensor, which then sends glucose readings to a smart device via Bluetooth. Unlike Senseonics’ reversible glucose-binding fluorescence-based reaction, Biomicro’s sensing mechanism is irreversible, though the rep told us that byproduct buildup over 180 days wouldn’t be an issue. The company is extremely early-stage at this point, with animal trials expected to commence next month.

Cnoga

The Cnoga booth displayed the company’s BGM, consisting of two parts: (i) a traditional fingerstick used for calibration; and (ii) a non-invasive BGM channel personalized to the individual (see picture below). Users calibrate with eight fingersticks/day for the first week and can then test non-invasively. The device has been CE marked for two years with “thousands” of “very happy” users in 10 countries – we haven’t previously heard of it, so presumably uptake has been quite limited. Booth representatives boldly expect 1.5-2 years until a US launch and have just submitted IRBs for two clinical trials in the US. The non-invasive monitor reportedly leverages the same technology as face-recognition software, using optical sensors to generate signals from four LED sources. Representatives emphasized that it is the most accurate non-invasive device on the market: A trial conducted in Germany found that the non-invasive monitor “came close” to meeting the 2013 ISO requirements, with calibration issues at fault for the readings in Zone B. Thorough initial calibration is key for the success of the BGM, as it is driven by self-learning software (i.e. the more readings the device has over a broader blood glucose range, the more accurately it will be able to read non-invasively). It’s a tough time to break into the BGM market, though a non-invasive device – if it actually works – would certainly help Cnoga persevere.

Defymed

In a mid-size booth at the side of the exhibit hall, Defymed displayed its two preclinical intraperitoneal (IP) insulin infusion devices: the ExOlin insulin delivery device and the MailPan Bioartificial Pancreas. ExOlin involves a biocompatible membrane that is implanted in the patient’s abdomen and a subdermal membrane that fields injections. The non-biodegradable device is permeable to insulin, and allows insulin to be delivered to the sub-dermal membrane via injection – this allows patients to continue their normal insulin injection methods (i.e. syringes, pens, and pumps), while still providing the advantages of greater portal vein insulin delivery (similar philosophy to Roche’s DiaPort). According to the Defymed representative, the device is on track to enter its first in-human trial in Europe by the end of the year, and the company hopes to achieve CE marking by the end of 2020. The even-earlier-stage MailPan Bioartificial Pancreas involves a semipermeable, immune protective membrane pouch that can be filled with insulin-secreting cells and inserted into the intraperitoneal space; a subdermal membrane fields injections to provide nutrients to the encapsulated cells, which would otherwise lose vitality because of lack of vascularization with the surrounding tissue. The product is expected to enter phase 1 in 2020. Notably, Defymed is in partnership with Semma Therapeutics for preclinical validation of a version of the MailPan device containing Semma’s insulin-producing stem-cell derived beta cells. Though these early-stage projects are certainly exciting, IP infusion technology is not without steep scientific challenges. As discussed at length at the JDRF-Helmsley Charitable Trust Closed Loop Intra-Peritoneal Infusion Workshop, “biosymbiosis” to avoid a foreign body response clogging the catheters will be a major hurdle for Defymed to overcome.

Dexcom

Dexcom showcased G5 mobile on iOS and Android phones, alongside table-top screens highlighting the web-based Clarity data management platform. Clarity was a bigger focus than ever for Dexcom at ATTD – see above for coverage of the dedicated oral and symposium, which shared glycemic outcomes from 50,000 US users, screenshots of EHR integration, and uptake statistics. Handouts focused on the benefits of CGM in MDI, including ample reprints of the DiaMonD study.

Diabnext

At the Diabnext booth, we learned that the company’s multi-device AI management platform (BGM connector, pen dose capture, connected pill bottle top, activity sensor, weight scale, carb counting app) will first launch in China in March followed by a US launch expected in June – this has been pushed back again, most recently from the expected 1Q18 timeframe shared at CES. The app is being piloted in multiple countries to collect patient feedback and to support reimbursement in Singapore, Australia, and France. In the US, the app will be free with a premium version priced at less than $50/year including special features like graphs and additional insights. Each device will cost $49 with special coupons provided in exchange for patient recommendations. Diabnext hopes to partner with insulin, pump, and BGM companies to pursue US reimbursement from self-insurers and expects to receive France reimbursement “pretty fast.” The company is proposing to the Australian and Japanese governments for reimbursement and eventually hopes to pursue CMS, though we imagine CMS’s gatekeepers will need to see some compelling outcomes data, of which Diabnext has none that we know of at this stage. As a reminder, the “all-in-one” digital diabetes management platform includes: a cable to Bluetooth-enable non-connected meters (Gluconext; think Glooko’s MeterSync Cable); an insulin pen clip attachment to capture doses (Clipsulin; slightly dubious in design); an app that uses AI and pictures to calculate carbs on the plate (Snapcarbs); a connected pill bottle top (Vigicap); a coin-sized activity sensor that can be carried in a pocket (Vigifit); and a wireless body composition scale (Vigiscale).

Dianax

In a small booth at the edge of the exhibit hall, Dianax highlighted its “lab on a chip” technology – a disposable test cartridge that inserts into a connector and can be plugged into a smartphone to reportedly provide fast and easy lab results. The company is currently developing this technology to measure diabetes-relevant parameters such as A1c, C-reactive protein, and hemoglobin/ferritin (important for the monitoring of CKD) at the point of care. With its promise of fast results and anywhere use, we think the Dianax platform could be great for use at-home or in under-resourced areas where it is difficult for patients to reach a clinic for screening/checkups. The project is currently funded by a grant from the EU’s Horizon 2020 initiative, and according to Dianax’s website, a clinical trial of the technology is “ready to start” at the San Raffaele Hospital in Milan.

Glooko

Glooko’s welcoming booth demoed the DreaMed Advisor Pro software running on the web-based Population Tracker – as covered on day #1, the CE-Marked clinical decision support software optimizes pump settings, giving HCPs specific pump basal, insulin:carb, and correction factor adjustment recommendations based on CGM data. The software is also under FDA review. The Advisor Pro user experience takes place within Glooko’s well-designed web-based Population Tracker (see picture below), and DreaMed had a separate booth looking for signups from early adopter clinics. We learned from Glooko that pump and CGM data are actually sent out to DreaMed, analytics are run by DreaMed, and then the recommendation are sent back to Glooko and displayed on the Population Tracker. Once a clinician approves the recommendations, he/she can click “Approve and Share,” which relays the pump settings changes to a patient’s Glooko mobile app. The Glooko team told us over 20 patients are now enrolled in the Helmsley Charitable Trust-funded multicenter study – per the post on ClinicalTrials.gov, the n=112 study is expected to wrap up by this December.

GlucoMe

The GlucoMe booth presented its digital diabetes platform, consisting of a connected BGM, patient-facing app, and provider-facing Digital Diabetes Clinic (DDC). GlucoMe’s BGM has already received a CE mark and the company is working to commercialize in Israel, India, Europe, and Latin America. GlucoMe is finalizing its clinical trial for FDA clearance of the BGM and expects to launch the app and DDC in the US ahead of the device in 3Q18. In a call with EVP John Erickson, we learned that the company expects a 90-120-day FDA clearance, though obviously the bar for BGM accuracy has risen at FDA. The BGM is unique in that it uses patented technology to transmit readings to the app via audio waves, independent of Bluetooth or WiFi (is this more or less susceptible to interference?). According to the booth representative, GlucoMe’s BGM has met the latest ISO accuracy requirements. While the app can only stream GlucoMe’s BGM data, there are plans to eventually integrate with Apple Health Kit and other wearables. The DDC reportedly integrates with EMRs, thereby streaming data from multiple sources, and can also serve as a telemedicine platform. GlucoMe is also boldly developing an insulin pen monitor that will be compatible with any insulin pen, as per the company website. It is still under development and is expected to launch in 2019.

Insulclock

Insulclock has a CE-marked insulin pen cap that tracks insulin dose, timing of dose, what type of insulin has been administered, and temperature. A rep tried to demo the device but unfortunately couldn’t pair it to the app (Google Play; App Store). He also couldn’t divulge the mechanism of dose capture, but it seemed to us to be accelerometer (motion)-based. The pen costs €249 and reportedly lasts an impressive five years. A single charge through the micro-USB port lasts three days, and it takes ~1.5 hours to fill the battery. The device will be available for Lilly’s Kwikpen on March 26^th, Novo Nordisk’s Flextouch on April 9, Sanofi’s Solostar on April 23, and Novo Nordisk’s Flexpen on May 7 – the rep indicated that launches will begin in Spain, followed by other EU countries. The app can pull in glucose data from Apple Health and Google Fit, but conversations with BGM companies for direct integration are ongoing. The rep was excited about an Emory study, PI’d by Dr. Guillermo Umpierrez, of Insulclock in type 2 diabetes: An n=80, 24-week crossover trial with a primary outcome of insulin injection irregularities. Secondary outcomes are related to satisfaction, quality of life, change in A1c, and hypoglycemia. In the intervention arm, participants will use the Insulclock and receive daily information on their smartphone on insulin administration (time and dosing) as well as reminders in the event of missing doses, while participants in the control arm will use Insulclock without feedback; they will then switch over at week 12. We are excited to see results from this trial either in December 2018 or December 2019 (unclear based on projected primary or overall completion dates on CT.gov) which could make a compelling case for reimbursement.

Insulet

In its first European exhibit hall booth, Insulet reps shared tremendous excitement to take over European Omnipod distribution from Ypsomed on July 1. (Ypsomed did advertise the Omnipod on its own booth wall, though its in-house YpsoPump was definitely the focus in booth demos.) Insulet reps emphasized the desire to provide uninterrupted care for Omnipod’s 50,000+ OUS customers. The biggest question from attendees was OUS timing on the new Dash PDM, but there is none to report at this time – the much-improved touchscreen, Bluetooth-enabled controller was submitted to FDA earlier this year, with clearance expected in 2H18 followed by a “limited market release.” Presumably an OUS launch could follow in 2019 (our speculation)…

Insulin Algorithms

Mellitus Health’s Insulin Algorithms exhibited its insulin dose clinical decision support software, which has already received CE marking and FDA clearance. Through the system, BGM readings are sent to a provider-facing dashboard, which generates insulin regimen recommendations “in 15 seconds.” The healthcare provider must then relay any changes to patients. Insulin Algorithms works with “nearly all glucometers” as per the company website and lists several partners, including Abbott, Roche, Ascensia, and Agamatrix. The booth displayed a poster demonstrating that the software resulted in an impressive 1.9% A1c reduction (baseline: 10.0%) at 90-days (n=11). For those who remained in the study for six months (n=17), an A1c reduction of 2.4% points (baseline: 10.0%) was reported by the study’s end. We’ll have to see if Insulin Algorithms can scale effectively and integrate with provider workflow – since providers have to review every recommendation and personally inform the patients of alterations, efficiency will be key to follow (similar to DreaMed’s Advisor Pro). Still, the study results are quite impressive, assuming the algorithm avoids hypoglycemia with this profound decrease in A1c. The booth representative shared that the software is available in 30 “countries and clinics” in Europe and the US. Depending on the clinic and number of patients, the software is either priced at a flat annual rate or by the number of users.

J&J – OneTouch

Apart from the booth reps’ name tags, the OneTouch booth did not have any J&J marketing – at least from what we could tell). The reps themselves weren’t OneTouch employees, and so couldn’t comment on a possible sale of the business despite reporting that many attendees had inquired about it. The booth did highlight the blue (low), green (in-range), and red (high) color indicators of blood glucose values on the OneTouch Verio Flex Bluetooth-enabled BGM. Conference-goers’ sorting ability and reaction timing were assessed in a game where a glucose value would fall from the top of the screen and they’d have to sort it into high, low, or in-range depending on the magnitude. In the second phase of the game, the numbers fell faster, but they were haloed by the color that corresponded to their range. To OneTouch’s point, this made it much easier to mentally categorize the value as in-, above-, or below-range. Animas closed in October, and J&J CFO Dominic Caruso said on the 4Q17 call that his company continues to assess whether “the business is better in someone else’s hands.”

Lilly

The wallpaper was the highlight of Lilly’s exhibit hall booth, showing a timeline of the company’s diabetes efforts, from the 1876 founding to the 2016 CV indication for SGLT-2 inhibitor Jardiance (empagliflozin). Reps also pointed to Basaglar’s US launch in December 2016 as a major recent milestone for Lilly. Indeed, Basaglar (biosimilar insulin glargine) is the first-to-market biosimilar insulin, and has posted remarkably strong sales so far. There was no mention throughout the booth of Lilly’s “Connected Diabetes Ecosystem,” which includes a connected pen and AID system (Dexcom CGM + in-house pump + hybrid closed loop algorithm acquired from Class AP in Montreal).

Medtronic

Medtronic’s large booth was the first visible in the convention center, highlighting the Unomedical’s new Mio Advance infusion set (see below and our day #1 coverage), the Guardian Connect standalone mobile CGM, and the MiniMed 640G. No MiniMed 670G was on display, despite 3Q17 plans to launch outside the US before this April. We enjoyed taking a time-in-range quiz using a Nintendo Wii balance board – great to see Medtronic educating attendees on this topic.

Metronom

Stealth CGM startup Metronom finally made its public debut in the exhibit hall, sharing plans to submit a CE Mark in 2Q19 for its 14-day wear, optical-based, factory calibrated CGM. The sensor technology includes direct oxygen measurement (which can monitor sensor reliability and prompt for a “smart” calibration when needed); dedicated sensors for hypoglycemia, euglycemia, and hyperglycemia; an insertion that is more like a lancet (applicator device like FreeStyle Libre); and a durable transmitter that sends data via Bluetooth to phone and watch apps (it can already go direct-to-Apple Watch, according to the company). Notably, the plan is to manufacture the sensor like a test strip (“reel-to-reel”), enabling lower expected cost of goods and expected EU pricing on par with FreeStyle Libre (second picture below). The first-in-human feasibility data (hand assembled sensors) demonstrated a 9.0% MARD and 94% of points within 20/20 (n=1,641 paired points, range of 50-300 mg/dl) in 20 people with diabetes (n=10 T1D, n=10 T2D). The study used retrospective calibration and obviously was not making sensors at scale, so upcoming studies will need to prove accuracy and reliability on a bigger scale. The sensor also has a unique adhesive that moves with the body. More in-human studies are expected this year in advance of a 1Q19 EU pivotal study. The company definitely plans to secure a commercial partner, and assuming these results and product features hold, we’d have to imagine many are interested. A second-gen concept (prototype) version that is fully disposable and much slimmer on the body was shown on a plasma screen– see the third picture below, which is not final and may change. We first covered Metronom Health in July 2014 – when it raised $4.8 million – though the company has been very quiet up until now. The company was incorporated in 2009, meaning nearly ten-years of work have gone into the product. We were impressed with the team at ATTD, though the path to scaling any CGM is always challenging and winding – can Metronom get everything in place over the next year? Who are the likeliest commercial partners?

MyMedBot

MyMedBot is a new tool for remote monitoring of people with diabetes that is currently in a closed, 20-30-patient prototype beta testing stage. Founder and CEO Jacob Arnould, who was inspired to start MyMedBot after two close calls with severe hyper- and hypoglycemia, noted that current remote monitoring solutions are limited in scope and not very easy to set up. With MyMedBot, he and his team are creating an app that integrates with CGM to send automated and manual alerts so that anyone can easily follow or receive notifications from a person with diabetes. Perhaps most importantly, MyMedBot shares information about a person’s location and provides step-by-step instructions on how to help. We imagine this system would be easiest to use if “followers” don’t have to download another app themselves to receive alerts, and we hope this product might eventually offer more peace of mind to people with insulin- or SFU-dependent diabetes (though CGM will obviously be necessary for automated alerts).

Novo Nordisk

One corner of Novo Nordisk’s booth was dedicated to the company’s digital health department, now two-years-old. An “idea box” was set up so that conference attendees could share suggestions for what Novo Nordisk might do in digital health to better support providers/patients – we note that there’s a big need for this sort of input, given that only 43% of HCPs find pharma’s digital support tools useful. ATTD was the first-time debut of the “idea box,” and we hope it sparks even more momentum in Novo Nordisk’s digital health department. Reps discussed the Glooko-partnered Cornerstones4Care (C4C) app, which offers patients personalized diabetes support, automated tracking of blood sugar and exercise, trend recognition, reminders, food/medication databases, and educational information on diet/exercise. C4C is available for free download on the Apple Store and Google Play, offering a fairly similar experience to Glooko. Reps also discussed the connected pen pilot in Sweden: The NovoPen 5 Plus was launched at limited volume to 10 Swedish clinics participating in the trial, and it stores three months of data at a time that patients can discuss with their HCP for individualized advice on diabetes management. Great to hear Novo Nordisk being more public on this front!

Fiasp dominated the majority of the booth. Bright messaging (“from the first bite”) and graphs highlighted faster onset/offset with the next-generation, faster-acting mealtime insulin. Fiasp made waves at this year’s ATTD as the subject of several oral presentations (one-year Onset 1 data, Onset 5 data in pumps, etc.), and the product was launched earlier this month in US pharmacies at parity pricing to NovoLog.

Roche

At the Roche booth, we learned of the company’s plans to eventually merge all components of its digital ecosystem with a smartphone app, currently available for limited download through Roche’s BETAtogether website. The app is still in its earliest development stage, connecting only to Roche’s pumps and displaying basic metrics like battery status and basal rate. The next step will be to add in remote control functions on the app, including bolus advice and basal settings. Once these functions have been added, the app will launch to a greater market. Eventually, Senseonics CGM data will also stream to the app. Booth representatives also noted that Roche’s Solo Patch Pump is undergoing its last clinical trial and will launch in certain EU pilot countries by the end of 2018 – one representative expected an initial launch as early as this Summer or Fall. The timing feels quite ambitious, given how long this has been sitting and that (news to us) it’s still in a clinical trial.The patch pump will initially be controlled with a separate PDM, which the representatives emphasized as being a significant improvement over the previous (clunky) version in both response time and display. Eventually, the patch pump will also be controlled via the app directly on users’ smartphones. Smartphone control would be a huge win for Roche – the Dana pump is the only option currently on the market with this feature, although Ypsomed reportedly is close behind. Booth representatives anticipated a US Solo launch at least two years following the complete EU launch.

Senseonics

At the Senseonics booth, reps ball-parked that there are now >2,000 total Senseonics users, with ~1,500 in Germany alone. While the 180-day Eversense XL has now rolled out in the UK, it won’t be until April that it enters the German marketplace (presumably waiting for the current 90-day lot to run out). We’re not sure about timing for other markets. Reps were unable to comment on US FDA Ad Comm timing, nor an n=300 French reimbursement study, but did indicate that enthusiasm and curiosity about the sensor remained high from booth visitors.

Sooil

The Sooil booth displayed the company’s new Dana U pump, currently under development. The Dana U boasts a full color touch screen controller (it can also be controlled via smartphone app) and uses a normal AAA battery – an improvement from the current pump, which operates via a special AA battery. A booth representative expected a clinical trial for the Dana U to possibly begin at the end of 2018. The company is also investigating integration with a cloud-based system for Dana U. The current Dana pump is available in a remarkable 66 countries across Europe and Asia. The pump can be controlled via Android phones, with a CE mark for iOS devices likely to be received in one to two months, according to the representative. Sooil hopes to get FDA clearance in the future, but the representative believes submission is still one to two years out. Sooil’s largest market is China, in which the Dana is the second most utilized pump (presumably Medtronic is the leader). The company is currently seeking out distribution partners in Canada and Japan to determine whether to pursue regulatory approval.

Theranova

At the sparse Theranova booth, we learned that the company seeks to raise $1.2 million to complete a next-gen closed loop prototype (with sub-cutaneous sensing and use of Roche’s Diaport for proof-of-concept), manufacturing, and acute preclinical studies. A rep told us they have already approached JDRF, NIH, and Helmsley Charitable Trust for funding, and Montpellier’s Dr. Eric Renard concerning the closed loop study program. Eventually, the system will consist of IP glucose sensing (see data from April 2017) and IP insulin delivery. A handout suggests that product development is expected to take 12-24 months, and a CE mark and FDA 510(k) clearance for IP insulin delivery will be pursued in parallel. The company is not only developing an IP-IP closed loop system, but also an IP islet encapsulation system (see picture below).

Unomedical (Convatec)

Unomedical proudly showed commercial versions of the two products we first learned about at ATTD 2017 – the Mio Advance all-in-one, hidden-needle, fully disposable inserter and the Lantern catheter (package as part of “Inset II”). Mio Advance launched here with distribution partner Medtronic (see our Day #1 coverage and the YouTube intro here), and Unomedical shared that it’s available in Europe, the Middle East, and Africa. The rep told us it’s been the “best feedback ever” about an infusion set, a major win for Unomedical and certainly strong competition to BD’s FlowSmart set. The initial Mio Advance launch with Medtronic does not appear to be an exclusive partnership (like BD), though Unomedical has not developed a version for Tandem’s t:lock at this point. However, the set’s website does indicate a luer lock version is available, meaning Roche pumpers could theoretically use it. Regarding Lantern, the catheter with side-cut slits for additional insulin flow, the team remains very excited – the pictures and branding looked great, and a microscope with a screen allowed attendees to view it up close (see below). As we covered last year, Lantern has been designed to allow insulin to flow out of many additional side-cut slits, even when occlusion/kinking  occurs. The positioning of the slits is based on extensive studies and analyses to provide an optimal solution for the functionality of the concept. Lantern’s launch timing depends on what product claims Unomedical goes for – the team said it could possibly launch this year without a product claim, or it might wait to run studies and build more clinical evidence and claims for the set. The ATTD 2017 update called for a Lantern launch in 3Q17, so it’s running at least a couple quarters behind the initial expectation.

Ypsomed

An Ypsomed representative shared that the bolus calculator for the YpsoPump will be available on the MyLife smartphone app in May or June. Unfortunately, representatives were unable to comment on the timing for smartphone control of the YpsoPump, last slated for a 2019 launch, nor did they have any updates regarding Ypsomed’s plans for its patch pump YpsoPod or its participation in a closed loop system. During Ypsomed’s F1H18 call, management noted the company’s plans to launch the YpsoPod in late 2020 to early 2021 and conduct a closed loop pivotal trial with the YpsoPump in 2019. As per a press release in January, the MyLife app is expected to be available in 16 European countries by the end of April.

Startup Showcase

Between sessions on Day #2 and Day #3, we heard 15 10-minute pitch presentations from several start-up companies in the diabetes arena. Session chair Prof. Moshe Phillip characterized this as “the most important part of the ATTD meeting – the future.” Indeed, these sessions were filled with exciting early-stage projects; below, we recap some of the most promising ventures.

• Arecor CEO Dr. Sarah Howell presented new preclinical data on the company’s ultra-concentrated rapid-acting insulin, which completed preclinical studies early in 2018. Arecor’s U1000 candidate showed an onset of action and PK/PD profile comparable to U100 insulin aspart (Novo Nordisk’s NovoLog). Merely increasing the concentration of NovoLog to U1000 confers markedly slower onset of action, which is an unacceptable tradeoff for prandial insulin (when fast onset/offset is critical for precision around meals). In contrast, Arecor combines insulin aspart with proprietary formulation technology Arestat to maintain onset time. Dr. Howell did show that the tail of Arecor’s candidate is longer vs. U100 aspart, and we wonder how this will play out in the first-in-human studies, planned for 2018. Notably, Arecor’s U1000 insulin could have applications in miniaturized insulin pumps/closed loop systems, and it stands to improve quality of life for those who require higher doses of insulin (the company has positioned its candidate for patients with total daily insulin dose >200 units). Dr. Howell also presented data on Arecor’s ultra-rapid insulin, another reformulation of insulin aspart, which has shown faster onset (but not offset) vs. NovoLog in a pig model of diabetes. Beyond this, Dr. Howell pointed out that the ultra-rapid candidate seems at least comparable to Novo Nordisk’s Fiasp (faster-acting insulin aspart) in both onset and offset time (possibly even faster on the front end) – could Arecor’s ultra-rapid insulin offer a meaningful improvement over even Fiasp? We’re intrigued by the possibility, though we understand that this is still very early-stage. Finally, we learned that Drs. Bruce Buckingham, Eda Cengiz, and Thomas Pieber have joined Arecor’s Scientific Advisory Board; having such expertise on board is certainly confidence-inspiring.
• Dr. Muhammad Mujeeb-U-Rahman, co-founder of Integrated Medical Sensors (IMS), highlighted the development progress for his company’s preclinical implantable CGM – which he deems the “world’s smallest CGM.” Born out of Dr. Mujeeb-U-Rahman’s electrical engineering thesis work at Caltech, the sesame seed-sized sensor is the first fully-integrated implanted CGM that uses standard electrochemical monitoring. Despite its small size (3 mm x 0.6 mm), it contains every component of a computer – a tiny processor, an antenna, and a sensor. The three- to six-month-wear sensor can be injected under the skin using a simple needle and interfaces with a wearable wireless transmitter (such as a smart watch) in order to relay the data to a smartphone app. The sensor has demonstrated a MARD of 12.5% in animal studies, and according to the company, human feasibility studies could begin as early as this year after final “chemistry optimization.” Due to its inexpensive underlying semiconductor technology, the sensor is projected to cost just $1/day – a key differentiator from other products on the market. That said, IMS’ product is very early stage in an already robust competitive landscape: Senseonics’s 90-day Eversense implantable sensor has already launched in 14 countries (a variety of EU countries plus South Africa) and remains under FDA review (decision pushed back from 4Q17 to “early 2018”); the 180-day version is slated for a 2Q launch in Europe. Furthermore, Glysens’ one-year implant is awaiting a pivotal trial, and there are additional implantable CGM projects in development by Profusa. And perhaps most importantly, non-implantable sensors are growing more accurate, user-friendly, and discrete – how will the market segment between implanted vs. subcutaneous sensors?

• Mr. Amin Zayani, CEO of Med Angel and winner of the 2017 Novo Nordisk-Lyfebulb Innovation Summit, discussed his company’s smart, waterproof sensor that continually measures the temperature of insulin (or other injectables, including Epipens, biologics, and hormones) in order to ensure that it is safe to use. The Med Angel app sends notifications and alerts to the user, displaying a heart to indicate temperature in one of three colors: (i) green for in range (~36­°-46° Fahrenheit for most insulins), (ii) blue for too cold, and (iii) red for too hot. Users select their insulin from a list within the free Med Angel app (iTunes, Google Play), which supports all available insulin products in the US and Europe. Google Play users have rated the app 4.8 stars (19 reviews) with 500-1,000 reported installs. The product is currently available on Amazon, priced at $45 and earning 3.7/5 stars in 15 customer reviews. As a person with type 1 diabetes himself, Mr. Zayani described his relationship with insulin as “the most important relationship in my life” in that it requires constant, everyday attention. He learned firsthand the importance of keeping insulin in a narrow therapeutic range after he ended up in the emergency room from dosing himself with insulin that had been accidentally frozen and therefore rendered ineffective.
  • Mr. Zayani compared data from 432 MedAngel users to published data from 255 patients on biologic disease-modifying anti-rheumatic drugs, finding that fewer MedAngel users exposed their drugs to temperatures below zero for two hours or longer (13% vs. 25%). As an added benefit, Mr. Zayani mentioned, “even in the event that such accidents happen, users are informed of them and can act accordingly.”

• Dr. Thiago Artioli unveiled the Baci Insulin Infusion System, a 3D-printed insulin pen well-suited for type 1 diabetes patients in low-resource areas due to its low-cost (~$1/pen) and extreme user-friendliness. Hailing from Brazil, Dr. Artioli explained the severe limitations that poverty places on insulin therapy – both in terms of cost (an insulin pump costs 1400% of the monthly income of a low-income person in Brazil) and in terms of usability, given low medical literacy and the complicated dosing regimen for insulin. The 3D-printed Baci pen represents a cheap, simple insulin delivery tool – vastly more patient friendly than a traditional insulin syringe. The pen additionally comes with 3D-printed insulin cartridges which can be made in different colors or in different shapes to help make dosing less complicated for patients with low medical literacy (i.e. “Take two pink squares every morning and take one blue circle with meals.”) According to the company website, the pen accepts any type of insulin refill. We’ll be eager to see outcomes data for the device, as it has the potential to improve access to diabetes care in low-resource settings.

• Dr. Sahan Ranamukha, co-founder and VP of R&D at Microdermics, presented compelling proof-of-concept data on the use of Novo Nordisk’s Fiasp via intradermal injection. Compared to subcutaneous administration, intradermal administration gave significantly faster appearance of Fiasp (faster-acting insulin aspart) in the bloodstream (Tmax of ~5 minutes vs. ~15 minutes subcutaneously), as well as faster offset of action. The insulin reached a similar maximum concentration with both injection routes. While intradermal administration of Fiasp has only been evaluated in a rat model, we’re quite intrigued by these results – this is evidence that an alternative administration method could bring Fiasp even closer to mimicking endogenous insulin secretion, and we hope Microdermics pursues this further. (That said, BD tried to make this technology work and did not push it forward, so the early data will have to be confirmed and scaled commercially.) The company’s intradermal platform uses microneedles to take advantage of the rich vascular (capillary and lymphatic) and immune system presence in the intradermal space. Benefits include improved patient experience (comfort, ease of use), dose reductions, improved drug performance, and the potential for integration with existing drugs with respect to volume, viscosity, and speed. We think this concept could appeal to a wide range of patients, especially children, the elderly, those with serious phobia of needles, and those with motor challenges. Moreover, we would be particularly interested to see whether this platform could be used with GLP-1 agonists: Improved absorption could allow for greater efficacy at lower doses, potentially leading to an improved side-effect profile (i.e. more GI tolerability) and even lower cost.

-- by Adam Brown, Ann Carracher, Abigail Dove, Brian Levine, Payal Marathe, Maeve Serino, and Kelly Close