June 5-9, 2015; Boston, MA; Full Report – Closing the Loop and Insulin Delivery – Draft

Executive Highlights

This document contains our coverage of closing the loop at ADA 2015. Immediately below, we enclose our themes on the category, followed by detailed discussion and commentary.

This year’s meeting made it very clear that closing the loop is inevitable – no longer is this a question of “if,” but a question of “when” and “how.” Academic investigators (Cambridge, UVA, Bionic Pancreas) are pushing quickly to longer, larger, and more real-world studies (some with an eye towards regulatory approval), while Medtronic has set the industry pace by initiating a pivotal study of the 670G. Competition is pushing thinking and dialogue forward aggressively, and we can hardly wait for the host of longer-term, real-world data coming in the next one to two years.

With no major new data, more nuanced questions came to the fore at ADA 2015: (i) What’s the appropriate control group in a pivotal study – sensor-augmented pumpers, MDI+SMBG, or some combination? (ii) Insulin-only vs. multi-hormone?; (iii) What’s the right device form factor – can an artificial pancreas algorithm reside on a phone?; (iv) How does the regulatory path look, particularly regarding an FDA review of manufacturing/quality systems; (v) Is there a business model? (vi) What do payer’s think?

These questions are fleshed out more fully in our themes below, followed by detailed coverage of an excellent psychosocial workshop on the artificial pancreas; the JDRF/NIH Closed-Loop Consortium Meeting; several symposia related to closing the loop; new closed-loop data in oral presentations and posters; and company updates.

Talk titles highlighted in yellow were among our favorites from ADA 2015; those highlighted in blue are new full report additions from our daily coverage.

Themes

• More than ever, ADA made it abundantly clear – closing the loop is now inevitable! And, it’s a question of “WHEN” and “HOW,” not “IF.” With Medtronic beginning its pivotal study of the 670G, other groups moving ahead or exploring their options (e.g., Animas, Bigfoot, Bionic Pancreas, Insulet, Tandem, Type Zero), academic investigators gearing up for very large scale efficacy trials, and a clear regulatory path, there seems to be less doubt that hybrid closed loop is coming near-term (Dr. Aaron Kowalski cited a ~2017-2018 timeframe). Indeed, UVA’s insightful and super-smart Dr. Stacey Anderson summed up the main session of artificial pancreas oral presentations with a succinct characterization of the status of systems: “SAFE during the day. HIGHLY EFFECTIVE overnight.” Dr. Kowalski also emphasized the field’s progress, noting that technical feasibility is now proven in all six stages of the original JDRF roadmap – it’s now time for commercial development and regulatory studies to support approval. Moving forward, Dr. Kowalski’s revised AP roadmap (Diabetes Care 2015) has just three steps: (i) low glucose suspend; (ii) predictive low glucose suspend; and (iiia) insulin-only closed loop or (iiib) multi-hormone closed-loop. The real challenges ahead are adoption and reimbursement – or as Dr. Kowalski suggested, appealing to people with diabetes (& loved ones), clinicians, and payers via three key metrics: glycemic control, “burden”, and value.
  
  • Competition is pushing the thinking and dialogue forward aggressively, and we can hardly wait for the host of longer-term, real-world data coming in the next one to two years. Rapid-fire presentations from Drs. Moshe Phillip (DREAM), Steven Russell (MGH), Roman Hovorka (Cambridge), and Boris Kovatchev (UVA) showed each group’s future planned large-scale efficacy studies, which may heavily depend on the NIH’s $20 million initiative, “Advanced Clinical Trials to Test Artificial Pancreas Device Systems in Type 1 Diabetes” – as we understand it, the one to three award recipients are expected to be notified quite soon. We were impressed to see how actively and ambitiously the academic groups are thinking about these bigger trials – an equally big set of questions surrounds how manufacturing and commercial devices will work (see below).
• The most illuminating artificial pancreas discussion centered on practical and hard-hitting questions: (i) What’s the appropriate control group in a pivotal study? (ii) Insulin-only vs. multi-hormone?; (iii) What’s the right device form factor and burden vs. risk tradeoff? (iv) Will organizations have manufacturing in place to submit a PMA application? (v) What’s the right business model? (vi) What do payers think? There is still a major difference of opinions on the right control group (MDI vs. SAP) and remaining uncertainty regarding device form factor, manufacturing, business models, and payers. The insulin-only vs. bi-hormonal debate is in the stay-tuned phase right now, as more comparative data is coming. That said, we would certainly like patients to have access to both – we don’t think one closed loop product will be right for every patient that could benefit from one.
  
  • A key point of debate emerged during the main closed-loop oral presentations session: “What is the appropriate comparator group in a pivotal closed-loop study?” Cambridge’s Dr. Roman Hovorka asserted that the appropriate comparator is sensor-augmented pump (SAP) therapy – closed loop should have to prove itself against the best therapy we have now. Conversely, MGH’s Dr. Steven Russell felt strongly that closed-loop studies should compare to individuals’ therapy at baseline, whether that’s MDI/SMBG, MDI+CGM, pump only, or pump+CGM. We think they are both right, and the much harder question is choosing which is more important to focus on initially. Close Concerns’ Adam Brown raised this question again at the invite-only JDRF/NIH Closed-Loop Consortium Meeting (contact here to get on next year’s list), and it sparked an even heartier debate and clear camps on both sides of the issue. DREAM’s Dr. Moshe Philip answered from an academic perspective, suggesting that SAP as the control group makes the most sense, since the studies are scientifically testing whether the addition of a closed-loop algorithm to SAP makes a difference. Dr. Russell again heartily disagreed, advocating for studies that are real-world enough to test effectiveness in a broad population of type 1s (especially the majority of type 1s that are on MDI) –  “This is not an academic exercise, he said, “We give people systems to improve their diabetes control. We don’t want to limit ourselves to the people using pumps.” Dr. Aaron Kowalski pointed out that the question has very important academic implications, since journal reviewers will criticize the choice of control group (probably no matter what). Others emphasized that the payer implications could be significant, since reimbursement and coverage will depend on the pivotal study design and included populations. Jaeb’s Dr. Roy Beck eventually stood up and admitted that both sides have a valid case (“I could easily write a point/counterpoint”), but he leaned towards Dr. Russell’s view – the current reality is that only ~15% of people in the T1D Exchange (at most) use sensor-augmented pumps. To get the penetration everyone is hoping for, AP studies need to expand beyond sensor-augmented pump users alone. We agree – pivotal trials should ideally enroll a broad spectrum of type 1s, including those on MDI/SMBG, pumpers, and those on SAP.
    • Medtronic’s MiniMed 670G pivotal study is a single-arm trial, yet another potential answer to this question! Medtronic is clearly trying to get something to market fast, rather than proving how much better it is vs. a comparator in a randomized study. This is smart from a timing perspective (first to market), though presumably means reimbursement at the same rate as the MiniMed 530G. That the FDA is allowing a single-arm study to serve as the pivotal trial also reflects how far the regulatory environment has come.  
  • ADA continued the fascinating insulin-only vs. bi-hormonal debate. We’ve certainly heard a range of opinions in recent months (e.g., Keystone 2014, DTM 2014, ATTD 2015, ENDO 2015), though true head-to-head data is really needed to show what incremental value glucagon adds to the equation. ADA was notable in that regard, providing fresh perspective from the newer-to-the-field McGill University team. In pooled results from two small randomized, crossover, overnight trials comparing insulin+glucagon to insulin-only closed-loop, the addition of glucagon reduced time spent in hypoglycemia by two-thirds (from 3% to 1% - a low base, but many patients would call that meaningful, and real-world rates of hypoglycemia are probably higher). Meanwhile, Drs. Ed Damiano and Steve Russell mentioned their own plans to do an exciting Bionic Pancreas glycemic set point study comparing an insulin-only configuration to an insulin+glucagon configuration head-to-head – we first heard this at GTC Bio in April, though we’ve since learned these will be quite robust studies, testing a variety of glycemic targets (100, 115, 130, 145 mg/dl are being considered) and offering some exciting comparative data. A separate Stanford set-point study will test an insulin-only version of the Bionic Pancreas at several algorithm set points, and will do so over a longer period of time. We think this data is critical to meaningfully advance the conversation on this topic. Perhaps our favorite commentary on this debate came from Yale’s Dr. Stuart Weinzimer in the Closed-Loop in Youth Symposium’s Q&A: “I think it’s a false argument to compare one against the other. What I would like to see are options for patients with diabetes and clinicians. I want to see an insulin-only system and a bihormonal system on the market. Then clinicians can choose and help patients choose what is best for them. So let the market decide.” Hear, hear!
  • What is the ideal form factor? Should a closed-loop algorithm reside on a smartphone? At the Psychosocial AP Workshop, the FDA’s Dr. Stayce Beck said that the Agency is “open to the idea of” an AP algorithm residing on a smartphone, but “there are lots of challenges” to hammer out – the risk mitigation sounds quite challenging, and it might even require working with phone companies to ensure that the algorithm takes precedence over other smartphone functions (that sounds like a nightmare). We’re not sure how open Apple and Samsung and LG and other device makers would be to that. BU’s Dr. Firas El-Khatib further pointed out that phones are very fluid – an algorithm might get approved with mitigated risk, but the phone will change if it is a personal device (e.g., downloading apps and new updates). Alternatively, AP developers could lock the smartphone and strip out all its functionality, but then the integration is arguably not as useful – patients will have to carry another phone. Of course, some may prefer the user interface on a stripped-down phone to a pump, so there are multiple solutions here. We do think the ideal convenience solution – a personal phone running an artificial pancreas algorithm – is possible to create and something that we expect to see happen down the road. Certainly, Dexcom has moved ahead with its Gen 5 CGM system – five years ago, no one would have thought a smartphone could function as the primary display device for CGM, but if all goes as the company expects, that could be on the market by the end of the year. From a product perspective UVA (Type Zero) and Bigfoot have alluded to AP systems potentially driven by smartphone algorithms; we assume Animas, Insulet, and Tandem are thinking more about algorithms embedded into their pumps.
    • The question of form factor could have critical implications for product development and market uptake of AP systems. As Dr. Aaron Kowalski noted in Q&A, “We’re going to butt up [i.e., max out] on the ability to improve glycemic control with artificial pancreas systems. The driver will be form factor and minimizing the burden of wear. The phone is an obvious component ... Where will people with diabetes be willing to take the risk for the benefit of a reduction in burden? ... How much do people with diabetes want it? What are the mitigations you can build in? The person with diabetes should have some voice in there.”
  • The field seems quite comfortable with (and even excited about!) the regulatory path for artificial pancreas systems. Equally important is the accompanying review of manufacturing and quality systems. FDA’s Dr. Stayce Beck pointed out that PMA submissions can’t just bring pivotal trial data – sponsors must have manufacturing capabilities in place. Notably, the FDA is working with JDRF on different manufacturing models (and presumably regulatory review) to enable “component artificial pancreas systems” – Dr. Beck said this is “not a small part” of the process, but it is also “not nearly as exciting” as the science. However, she urged the community to start shifting the conversation, since so many things go into an FDA review besides a positive pivotal study. The JDRF/NIH Closed-Loop Consortium Meeting left it a little unclear how (and if) academic groups will go from large-scale studies to regulatory PMA submissions. This is a big question for many groups – some, like the Bionic Pancreas team at MGH/BU have been less clear about who will manufacture the dual-chamber prototype pump they have been building (Tandem has de-prioritized this project, per its 1Q15 call), while others are more clear (Medtronic clearly has this question answered, along with customer service and infrastructure). It’s also a bit unclear if some of the upcoming large-scale trials will actually serve as true “pivotal” studies, since regulatory submission requires a final commercial-ready device. This is certainly the intent of the Medtronic/DREAM, Bionic Pancreas, and UVA (Type Zero) teams; we’re not sure about Cambridge. We agree with Dr. Beck that manufacturing is not a small issue, nor is the ability to go to market and offer customer support for a class III medical device – previous PMA and commercialization experience obvious helps. UVA has established a company, Type Zero, to do this, but the aforementioned steps are a tall order for any company, let alone a startup.
  • Medtronic has moved ahead of the competition and initiated a pivotal study of the MiniMed 670G, with other companies following along. Still, there are some tough business questions to answer. At the Psychosocial AP Workshop, type 1 advocate and industry consultant Ms. Marie Schiller said that closed-loop systems will “probably not” command a premium – “You either have to be at par or charge less.” This will presumably depend on the pivotal data that is generated, but it’s certainly a tall order for smaller pump companies to live up to: the R&D investment for an artificial pancreas system is significant, and pricing at parity may be tough economically speaking. In one sense, Ms. Schiller’s comment is not a surprising sentiment, considering payer frustration over diabetes (see payer expert Mr. Aaron’s David’s talk below) and increasingly constrained healthcare dollars; still, artificial pancreas systems have great potential to simultaneously improve A1c and significantly drive down or completely avoid severe and costly short-term events. That has never been seen before in type 1, and consequently, we assume some data might eventually support a premium. Would payers pay more for a system that results in a significantly better A1c AND a far lower rate of severe hypoglycemia? Presumably yes, but that will likely need to be shown over time in longer-term studies, since such events are unlikely to arise with enough regularity in a pivotal trial. Ms. Schiller further suggested that building a revenue model is something industry “is struggling” with. For standalone pump companies (i.e., not Medtronic), she called the necessary collaborative effort “one of the biggest hurdles,” since none have an in-house CGM. These are tough questions to address, and have probably contributed to why companies like Animas, Insulet, Roche, and Tandem have not moved faster on the closed-loop front.
  • ADA 2015 brought a mix payer views on the artificial pancreas, ultimately suggesting more education and engagement is needed. Most encouraging was highly regarded endocrinologist Dr. Kenneth Snow of Aetna (formerly of the Joslin), who had crystal clear advice for the entire room at the JDRF/NIH Closed-Loop Consortium Meeting: artificial pancreas systems may show benefits on A1c, hypoglycemia, and burden, and the data must clearly show that. This seemed pretty reasonable to us. Dr. Snow used the helpful analogy of buying a mouse trap – there are $3, $5, and $30 mouse traps, but when you go to buy a mousetrap at the store, there is no data to point out which is better. In that case, the consumer probably just buys the cheapest one. The same is true with the artificial pancreas: Aetna will pay for the “better mouse trap” if there are data to clearly suggest that. On the other hand was payer expert Mr. Aaron Davis, who highlighted near-non-existent payer understanding of severe hypoglycemia (mostly anecdotes, since there is no real data); major lack of consensus on unmet needs in type 1 diabetes; and low payer understanding of the different insulin pumps. Perhaps most surprising were some of the direct payer quotes from primary payer research:
    • “Plans may underestimate hypoglycemia. I led a team looking at the issue a few years ago. We were asking what hypoglycemia events cost, and we really don’t know.”
    • “There is more hypoglycemia risk in type 1, but I don’t have any data for type 1s. I appreciate that it can be disastrous, but I don’t know what we can do about it.”
    • “We recently did a survey of our insulin pump patients. 50% didn’t understand key pump features. We want to know more about what is going on with this technology.”
• Payers are clearly most concerned about cost-saving devices, not simply products that reduce mild/moderate hypoglycemia. The challenge for manufacturers is that proving reductions in severe hypoglycemia causing hospitalizations is a very long trial indeed. We hope these trials are eventually undertaken, though it’s hard to imagine a pivotal study could bring such data, unless trials are enriched with those at high risk of severe hypoglycemia. We also believe measures of hypoglycemia need to be more specific and standardized in artificial pancreas trials (i.e., severe hypoglycemia needing assistance, moderate hypoglycemia, etc.) – this could go a long way towards giving payers much better information.

Detailed Discussion and Commentary

Working Group on the Psychosocial Aspects of the Artificial Pancreas

This valuable workshop on the artificial pancreas was headlined by patient-friendly optimism from FDA’s very impressive Dr. Stayce Beck; caution from the renowned endocrinologist Dr. Anne Peters; a balanced industry perspective from consultant Marie Schiller; and a surprising and distressing lack of payer awareness about type 1 diabetes and what severe hypoglycemia actually costs. Though the workshop was technically focused on the psychosocial aspects of the artificial pancreas, it really hit on a wide diversity of perspectives addressing the real world challenges as products advance to commercial readiness.

Payer Perspective

Aaron Davis (Partner, Jupiter Life Science Consulting, Los Angeles, CA)

Mr. Aaron Davis gave one of the most insightful closed-loop presentations of ADA, summarizing primary payer research (n=10 US payers, medical and pharmacy) from Metabolic Markets (2014). The results were fascinating, indicating virtually non-existent payer understanding of severe hypoglycemia (mostly anecdotes, since there is no real data); major lack of consensus on unmet needs in type 1 diabetes; considerable payer frustration over type 2 diabetes; and little understanding of the different insulin pumps. Mr. Davis said that payers are in a challenging spot right now, as spending has increased but outcomes have only modestly improved. From our recent conversations with payers, we know the challenge has increased from some elements of healthcare that have nothing to do with diabetes, such as the unexpected costs of drugs related to other illnesses like hepatitis C. Notably, Mr. Davis urged the audience to more effectively communicate who is likely to benefit the most from a new technology. A direct payer quote on the type 2 slide was particularly striking: “[The] unmet need is not more therapies. Our problem is getting patients to comply and doctors not letting patients “alibi” out of it.” [While we wouldn’t put it exactly like this, we do believe that lack of attention to behavioral elements of diabetes, along with mental health challenges (not just depression but “distress” as characterized by Dr. Bill Polonsky and friends) make it very challenging for patients to adhere to a complex set of drug, technology, exercise, and food goals. We are not particularly surprised to hear HCPs say they can’t control all of this, though we haven’t heard the phrase “alibi out of it” before – that’ll go down in our books! We do think creating more resources for coordinated and integrated care is critical, and technology-related tools could play an important role in scaling the success of HCPs.]

• “Payers recognize hypoglycemia risk, but don’t know the details very well.” Mr. Davis said that payers understand severe hypoglycemia is dangerous and costly, but there is no data beyond anecdotes to understand the issue. Payers believe the severe events in type 1 are likely small in numbers, but can be a significant cost per event. Direct quotes from the research:
  
  • “Plans may underestimate hypoglycemia. I led a team looking at the issue a few years ago. We were asking what hypoglycemia events cost, and we really don’t know.”
  • “There is more hypoglycemia risk in type 1, but I don’t have any data for type 1s. I appreciate that it can be disastrous, but I don’t know what we can do about it.”
  • “Hypo is a big risk, which is where education comes in. But what is my absolute risk. If I have a million member plan, how many severe hypo events will I see in a year?”
  • Payers sound like they do not speak to patients or HCPs very much, or assess records from emergency rooms. It also sounds like they are not aware of the T1D Exchange research on this issue.
• Mr. Davis believes there is “low payer understanding” of type 1 diabetes. We were glad he was so candid about this; in particular, he cited primary payer research (n=10 US payers, medical and pharmacy) from Metabolic Markets (2014) that asked, “When thinking about type 1 diabetes, what are the unmet needs?” Notably, more payers said “no unmet need” (n=3 mentions) than “no cure” (n=2) – a stark reminder of the lack of consensus and knowledge around type 1. The rest of the responses were equally distributed across the board: closed-loop (n=3); better insulins (n=2); postprandial glucose issues (n=2); need to simplify (n=2); avoid hypoglycemia (n=2); improved infusion sets (n=1). He did not offer many details on the methodology, but presumably the 10 payers could make more than one choice, as there were 17 responses. We were quite surprised to see that 30% of the payers actually said there is “no unmet need” in type 1 diabetes, despite all the spending on total cost of care. It was also notable to see three votes for the closed loop as an unmet need – though we would love to see an artificial pancreas, we’re not sure it fits the category of “unmet need” at this point. To the list we would add clinical decision support (e.g., how much insulin to take), mental health/distress, and overworked/underpaid healthcare providers.
• In type 2 diabetes, “the payer community is more direct, but they are challenged.” Regarding a similar question– “What are the unmet needs in type 2 diabetes?” – responses bucketed into just three areas: compliance (n=6); clinical inertia (n=4); and arresting the disease (n=2). Again, it was distressing to see what was not mentioned: medication side effects, overworked/underpaid HCPs, the behavioral challenges/stigma for type 2 patients, or lack of access to care.  Mr. Davis noted that this payer view of type 2 diabetes has been very consistent over the years. However, payers are now grappling with increased spending and only modestly improved outcomes – presumably they are also not feeling great about the increased exposure (much higher number of patients) and the fact that patients overall are living longer and costing the system a great deal of money. Overall, Mr. Davis said there is “considerable payer frustration in type 2 diabetes.” From this conversation, we understand why payers would be nervous – just the number of patients – but to say there is “considerable frustration” is unfortunate, since there is no shortage of ideas to try.  
• Payers want to know more about advances in insulin delivery beyond features/benefits. However, there is said to be generally a limited (or low!) understanding of the different pumps available today, and payers see little differentiation among pumps from an outcomes perspective. The latter is unsurprising to us, since most of the companies have not done studies to show the different outcomes. Some verbatim quotes from the payer research included one that was quite enthusiastic about the closed loop:
  
  • “I’m very interested in what’s going on with the artificial pancreas. We will be wiling to pay for it with a few caveats: like proper patient selection and knowing who are the right doctors to support it.”
  •   “The usual key questions come to mind: do we get better outcomes than with the existing technology – my hypothesis is yes, but what’s that incremental benefit vs. the increased cost?”
  • “We recently did a survey of our insulin pump patients. 50% didn’t understand key pump features. We want to know more about what is going on with this technology.”
•  “Payers are under an assault of new technologies all the time. If we don’t communicate who are the best patients, they will take the path of least resistance: I get to choose.” Mr. Davis urged the audience to communicate who is best for a new technology. Payers will base coverage decisions on who benefits the most based on the available clinical data.

Artificial Pancreas – The Regulatory Approvals Perspective

Stayce Beck, PhD, MPH (FDA, Silver Spring, MD)

FDA’s Dr. Stayce Beck gave a very positive talk on the potential of incorporating patient perspective into approval of artificial pancreas systems. She highlighted a new Draft Guidance on Patient Preference Information (released in May) – notably, such information could support labeling claims and meaningfully affect the risk-benefit assessment of closed-loop systems. The document explains what manufacturers should consider when choosing to collect patient preference information, as well as how such information can be incorporated into device labeling for patients and HCPs. It sounds incredibly exciting the direction that this could take; Dr. Beck said there are “lots of ways” to incorporate such information into regulatory decisions – for example, scared parents waking up multiple times at night to test blood glucose vs. an automated system that eliminates overnight hypoglycemia but raises A1c by 0.5%. It’s fantastic to hear the FDA speaking this language, which parallels the Agency’s highly encouraging February blog post on the approval of EnteroMedics’ implantable obesity device (“Listening to Patients’ Views on New Treatments for Obesity.”) We had been encouraged by that approach, though do not see that as a product that will be embraced by HCPs or patients – the AP is of course much different. We believe patient-reported outcomes will be key for closed-loop devices, since burden reduction is a major benefit of AP systems that should be incorporated into regulatory reviews – to what extent it will be weighed vs. harder outcomes will be interesting to see. Dr. Beck also said in Q&A that the FDA doesn’t have any preference on whether an artificial pancreas algorithm runs on a phone or pump; the former is more challenging, however, since the software mitigation and testing is much greater – still, this represents another sign of a forward-thinking FDA, for which we are very happy. It’s also another sign to us how the FDA drug side is so much different from the device side.

• “We are interested and open to incorporating patient preference information. But it must be done in a validated, quantifiable way.” There’s often a lot of “anecdotal evidence” on patient preferences, which is challenging for the FDA – “We’re a science-based agency, and we must make decisions that are defendable and based on scientific, quantified, validated data.” Interestingly, she said that companies sometimes don’t meet their endpoints, and at the last minute, appeal to patient preference information. Dr. Beck cautioned against collecting such information after studies end. To aid sponsors in collecting valid patient preference information, the agency has released three guidance documents:
  
  • May 2015: Draft Guidance on Patient Preference Information. 
  • November 2013: Draft Guidance on Medical Device Development Tools (MDDTs). This is a pilot program for FDA qualification of tools like patient-reported outcome assessments. MDDTs can be qualified for a specific context of use, and the Agency is currently accepting proposals.
  • December 2009: Final Guidance on Patient-Reported Outcome Measures.
• Dr. Beck highlighted the recent approval of EnteroMedics’ implantable weight loss device as an example of incorporating patient preference information – see our detailed report on the Blog Post, “Listening to Patients’ Views on New Treatments for Obesity.” We were impressed at the time, as the device lacked efficacy, but patient perspective played an important in getting approval. We don’t know enough about the advocacy effort, however, we don’t expect this product to be particularly commercially successful – we would want to make sure that patients do not receive any blame for the product not doing well.

How Can We Manage Expectations And Support HCPs To Provide Optimal Support For AP Systems In Routine Care?

Anne Peters, MD (USC, Los Angeles, CA)

Dr. Anne Peters provided a cautious HCP perspective on closed-loop systems (not hers), fearing clinicians may not be equipped to handle the technology: “Why adopt a largely unproven, highly complicated, very expensive, potentially dangerous technology” when most patients turn down CGM, when most providers don’t download BGMs, and when many people with type 1 are doing okay on current therapies?” Dr. Peters showed a metaphor to illustrate the point – a humorous video of a self-driving car that goes awry (an artificial pancreas), leaving the car’s passenger (the HCP) completely helpless and frightened. It was a hit with the crowd and accentuated two of her final points: (i) lack of fellowship training in diabetes technology; and (ii) a responsibility question mark when things go wrong with systems (Device manufacturers? AP researchers? Endocrinologists? CDEs?). Dr. Peters’ concluding thought in Q&A summarized her experience fighting for patients, but also caring for the underserved in LA: “I’ll offer the artificial pancreas to everyone once we have it. I’ll fight like crazy for them to get it. The magic is in the people who will need it the most. But they’re not going to be the first to get it.” We do ask what “doing okay on current therapies” means and whether that is using A1c only or other points as well. 

Challenges and Opportunities from an Industry Perspective

Marie Schiller (Schiller Advisors, Boston MA)

Type 1 advocate Marie Schiller shared an industry perspective on closing the loop, discussing key challenges and opportunities. On the bright side, the technical/development/regulatory risk is starting to decline, she said, though companies seem to be struggling with the business/revenue model. Ms. Schiller does not believe closed-loop systems will command a premium – “You either have to be at par or charge less.”

• Key opportunities:
  
  • Technical/development risk is starting to decline: The regulatory path is “more fleshed out,” she said, in large part thanks to efforts from JDRF and HCT.
  • Ability to have a greater impact. “Management teams get it – they see patient views.”
  • Capture MORE patients for longer periods of time: If closed-loop systems are prescribed right upon diagnosis, companies can put patients on systems sooner. Ms. Schiller ballparked US pump penetration at “somewhere in the 35-50% range” in type 1 (we haven’t heard that high end before), and “not at all” in type 2 (at DTM 2014, Dr. Bruce Bode estimated that there are ~100,000-150,000 type 2 pumpers in the US). Ms. Schiller estimated a potential artificial pancreas market of ~ 8 million patients on basal-bolus MDI regimens in the US, Canada, and the EU markets combined (“A big opportunity”).
  • Increased revenue from new components: Ms. Schiller used the term line extensions” to suggest companies adding new closed-loop products to what they currently offer.
  • New business models: As an example, Ms. Schiller highlighted Medtronic’s recent split into three business units: one is service and solutions, while the others are intensive- and non-intensive diabetes management. Closed-loop systems bring an opportunity to create new business opportunities beyond just selling hardware, she said. This is also how the Bigfoot Biomedical team has talked about their automated insulin delivery system.
• Key challenges:
  
  • Technical: CGM accuracy, concentrated insulin, risk mitigation.
  • Cost of goods sold and business models: Ms. Schiller characterized “the ideal product” as a system with the lowest possible cost that can provide the most value. She said closed-loop systems will “probably not” command a premium. Though medtech has made money on premiums in the past, “We cannot do that any more. You either have to be at par or charge less.” Ms. Schiller said that building a revenue model is extremely challenging and “industry is struggling a bit” with it.
  • Regulatory and Integration: Bringing all the components together is challenging, as the FDA considers artificial pancreas products a “system” – pump, CGM, algorithm, meter, human interaction. Aside from Medtronic, this is not something that companies have the infrastructure to internally take on. Later, Ms. Schiller called the necessary collaborative effort “one of the biggest hurdles.”
  • Portfolio considerations: Drug delivery units in companies are often not specific to one therapeutic area. For instance, an infuser could be used in rheumatoid arthritis, cirrhosis, and other disease areas. Something that is best for diabetes might not be best for other areas. This is an interesting perspective we had not heard before, though we’re not sure it applies to most of the pure-play diabetes organizations working on the artificial pancreas – the exception is Medtronic, though they clearly have a very specific roadmap ahead. 
  • Greater competition. More than ever, there are a range of companies looking to bring hybrid closed loop system to market in the 2017-2018 time frame.
  • Features and IP: Closed-loop algorithms are complicated, and companies must decide how they’re going to design from user and IP perspectives. 
  • Outcomes: Payers go by A1c only, unless companies can prove why other outcomes matter. We’d note that this has changed on the regulatory side, as the final artificial pancreas guidance does accept time-in-range as a valid endpoint.
  • System/Physician economics: Going into a physician’s office, there must be a clear value proposition as to how the economics of an office will change once automated insulin delivery is available. “We need to go out and do a full systems review. We need to build those outcomes into our studies. For example, when I go out on Ed’s system, I spend 10 minutes with a patient instead of 20 minutes. The system costs 10% more, but we just saved you 30%. Every time we ask a clinical question, we need to ask an economic question.”
  • Distribution models. Ms. Schiller highlighted that regional differences are especially key.

Panel Discussion

Dr. Henry Anhalt (T1D Exchange, Boston, MA): I really enjoyed your presentation, Stayce. How do you balance scientific evidence vs. patient preference information? Does that just come down to a label discussion? If you talk about change in A1c going up, but a decrease in psychosocial burden, how do you weight one against the other?

Dr. Beck: We haven’t approved an artificial pancreas device, so we haven’t put it into practice. We’re talking a lot internally. It depends on what the study is and what the investigator is saying the device can do. Are you saying that with an artificial pancreas, you don’t ever have to think again about diabetes? Or just that it can minimize hypoglycemia? We’re having conversations and talking about it before hand. I always recommend to sponsors, “Come early, be loud, and stay late. Come talk to us about it before you start the study, so you really are actively looking at it.”

Dr. Katherine Barnard (University of Southampton, UK): data. What data do you need?

Dana Ball: We recognize that we need validated standardized tools to measure what we think we’re going to need to get approval. We need to build a body of evidence to support patient and clinician needs. The Trust has projects to fund the development of quality of life scales across all stages – the pediatric setting, young adults, and older adults. We have to think very carefully and tailor tools – it’s not going to be a one-size fits all approach. It’s a two-prong approach – collect the data in a good enough way, and at the same time, work as a community to create new tools.

Q: There is a balancing act between usability and cost. Would the FDA approve anything less than perfect?

Dr. Beck: Definitely. Let’s not let perfection be the enemy of good. We don’t expect this will be a perfect device. We do want to look at the risk profile. When it isn’t going to work, are there mitigations that can be incorporated into the device? For instance, when the device goes offline or when devices aren’t communicating, it slips to open loop and maintains basal levels. We need to build that into the system to keeps disaster away. But it doesn’t have to be perfect.

Dr. David Kerr (Sansum Diabetes Center, Santa Barbara, CA): I’m nervous that the artificial pancreas is going to be a California hipster technology. And the majority of people who will benefit will struggle to access it.

Dr. Anne Peters: The most needy patients are my underserved patients, and they don’t have access. The mortality rate is much, much higher. People don’t have enough money for food. The ones who are going to use it [artificial pancreas] are the early adopters – they fight me for technology. One of my patients was in the LA Times in an article about Afrezza. She loves it. She’s a type 1 on an insulin pump and has an A1c of 6.8%. There is nothing about her that makes me to put her on another drug. She convinced me that she wanted it. She believes it helps her life. I cannot predict who that’s going to be. People self-select. I’ll offer this [artificial pancreas] to everyone once we have it. I’ll fight like crazy. The magic is in the people who need it most. But they’re not going to get it first.

Dr. Kerr: Who is the ideal population for a pivotal study?

Dr. Beck: Everybody. We don’t necessarily want you to pick people with a 7-9% A1c, no complications, and no cardiovascular problems. In a study, you see who it does and doesn’t work for. Maybe it’s not appropriate for someone with a low A1c. Maybe it is, but that’s part of the purpose of the pivotal study – to determine who should use the device and if there are certain populations that shouldn’t. You would label the device with that information and why it shouldn’t be used in those populations.

Mr. Davis: From a payer perspective, the payer world is looking for guidance about who is the right patient for this type of technology. Absent guidance from industry and thought leaders, they get to define it. The opportunity right now, as data is developing, is identifying who is the best fit. We need to start to decide now who is the most appropriate population. Otherwise, the payer community gets to decide that for you.

Q: What are you current thoughts or concerns on phones controlling artificial pancreas systems?

Dr. Beck: We get asked this a lot. The FDA doesn’t have any preference on whether the algorithm is in a phone or the pump. What we do have preference for is that whatever is controlling it is tested and looked at in different ways. The phone has different risks than the pump. There might be a technological way to deal with the risks, but we do have to think about a lot more things. What about a phone call? What if I’m playing Candy Crush? What if i don’t get text messages – that would be a problem. The artificial pancreas system needs prioritization in the phone. Some of that has to be worked out with phone companies. It’s thinking through those situations and ways to address it.

Dr. Jill Weissberg-Benchell (Children's Hospital of Chicago, IL): I have a patient needs question. If I’m a three-year old, what I might want is fundamentally different from a 13-year-old, from a college student, from a 60 year-old. How do we move forward if everyone needs something different?

Ms. Schiller: We’re not looking at the label as an approved artificial pancreas for kids 6-9 years old, females, with an A1c 8-9%, and both parents at home. [Laughter] But Aetna might. That’s what industry is doing – obviously looking at target product profiles and target patient populations. Sometimes you start seeing it fall to the lowest common denominator.

Mr. Davis: It’s a very good question, and a very important one. One example of payer policies is what I showed for BCBS Alabama and Aetna. For sensor-augmented pumps, it said if you hit our pump and CGM policy and you’re over 16, we’ll approve this. They probably pulled age 16 out of the air. Somewhere, someplace, 16 seemed right. Is that the right person? There might be no guidance and no data to support it. Back to the original question – if it’s approved for people with type 1 diabetes, it’s on the manufacturers and community to say, “This is the right technology for everybody.”

Dr. Firas El-Khatib (Boston University, MA): I’m struggling with the idea that the FDA could potentially approve an artificial pancreas driven by a phone, provided risk is mitigated on the phone. How do you deal with that? We all have phones. But this is a fluid situation. The phone gets approved and risk is mitigated, but then the phone changes if it is personal phone. If it’s not and you lock and freeze the phone, how useful is it to carry another phone? This can turn tragic quickly if another app fringes on one, or there is an update to the phone, or the phone company releases the next phone.

Dr. Beck: I didn’t say it was easy. We’re open to the idea. Personally, I would prefer to carry around one phone. I agree that there are lots of challenges, and how the realities of it would work or not. There are ways that you could work with the phone companies to do some testing. If you’re one of their developers, there is some way to go through a pre-specified test program. We are open to that idea of setting a pre-specified test plan. A lot of it is about system setup. I still haven’t updated from iOS 6 to 8 – I don’t have enough memory on my phone left to do it. Maybe there is a way to make it so it stops working and can only do open loop.

Dr. Aaron Kowalski: This is an important topic. We’re going to butt up [i.e., max out] on the ability to improve glycemic control with artificial pancreas systems. The driver will be form factor and minimizing the burden of wear. The phone is an obvious component. Marie and I often discuss this, “Where will people with diabetes be willing to take the risk for the benefit of a reduction in burden?” I think we coddle people with type 1. Bryan Mazlish’s wife and son have been using an artificial pancreas running on a phone for two years. How much do people with diabetes want it? What are the mitigations you can build in? The person with diabetes should have some voice in there.

Ms. Schiller: Today, I’m all open loop. If I can have four hours and then the phone shuts down, it is incremental benefit.

JDRF/NIH Closed-Loop Consortium Meeting: “The Last 100 Yards”

This invite-only annual meeting brings together closed loop researchers and industry for learnings and nuanced discussion on the state of the field. The evening began with JDRF’s Chief Mission Officer/VP of Research Dr. Aaron Kowalski, who shared a comprehensive overview of AP highlights from the past year. Subsequently, Drs. Roman Hovorka (Cambridge), Boris Kovatchev (UVA), Steven Russell (MGH/BU), and Moshe Philip (Schneider Children’s Medical Center) presented their plans for large efficacy studies of artificial pancreas systems – these proposals are summarized in a table below. A panel discussion followed, including added representation from FDA, Medtronic, Animas, and Dexcom. A panel of providers, patients, and payers closed out the night, sharing perspectives on expectations and adoption of artificial pancreas systems.

The Last 100 Yards

Aaron Kowalski, MD (JDRF, New York, NY)

Dr. Aaron Kowalski opened the annual JDRF/NIH closed-loop night (“The Last 100 Yards”) with an inspirational, gracious, and passionate lecture. His impressive 20-minute review provided a comprehensive summary of closed-loop progress from ALL the major players in the field – Stanford, UCSB, Barbara Davis Center, Bionic Pancreas, DREAM, UVA, and Cambridge, just to name a few! He turned to running for an analogy, opening his presentation suggesting that a commercialized hybrid closed-loop has the finish line in sight and must make the most of its final adrenaline rush to drive at the nuanced questions that remain. The field, in his view, has demonstrated technical feasibility in all six stages of the JDRF roadmap, but is left with commercialization questions regarding adoption and reimbursement. His words were a reminder that improving patient outcomes – reducing burden and simultaneously improving glycemic control and hypoglycemia – is both the great challenge and the great promise of this technology. Dr. Kowalski declared emphatically that “the standard of care right now is completely unacceptable!” He cited very impactful data on overnight PLGS presented by Dr. Bruce Buckingham earlier at ADA – the 3,000+night study showed that those in the control arm were spending 5% and 8% of nights (depending on the age group with a glucose <60 mg/dl for two hours or more. PLGS reduced that to just 1% and 3%, respectively. Said Aaron, “This data has really motivated me. It is really powerful.” There are major hurdles ahead, but Dr. Kowalski was optimistic in thinking about the challenges that remain ... or in his words, this footrace’s “last 100 yards.”

• Dr. Kowalski provided a comprehensive look of closed-loop progress from ALL the major players in the field. He broke down his analysis both by the academic groups involved in closed-loop research and the major sites playing home to studies. We bring you his summary below:

Major Players in Closed-Loop Research

Group/Site	Recent Work/Progress
UVA (DiAs)	12 research centers around the world 3 control algorithm implemented: USS Virginia, RCM (Pavia); Zone MPC (UCSB) 320 people with type 1 diabetes participating in multi-center clinical trials 118,470 hours (14 years) of DiAs system use Overnight CLC in Outpatient and Home Setting (with NIH) Multicenter Outpatient Control-to-Range (JDRF CTR3)
Stanford	Hotel based studies to evaluate initial settings in the Medtronic MiniMed 670G hybrid closed-loop system Detection of Sensor and Infusion Set Failure Control-to-Range Multicenter study of prolonged outpatient use of a closed-loop system Predictive Low Glucose Suspend Bionic Pancreas DiAs Advisor to restore hypoglycemia unawareness Use of CGM, Closed-loop systems and Behavioral Support in Toddlers Probabilistic Closed-loop Control (MMPPC) to Handle Unannounced Meals Control-to-Range Closed-loop in a Diabetes Camp Setting Development of a drug eluting infusion set 670G in a diabetes summer camp – 2014 (Medtronic) 640G for predictive low glucose suspend (Medtronic) Extended wear infusion set (Dexcom)
UCSB/Sansum	Semi-automated artificial pancreas with MannKind’s technosphere insulin Control to Range @ Home PID vs. MPC Clinical Trial Risk Mitigation and Fault Study Fault Detection with Stanford RO1-DK085628: Exercise detection and classification DP3-DK104057: Pediatric design and evaluation (UCSB & Yale) DP3-DK101068: Implantable artificial pancreas AP Database
Yale	Closed-loop work with: Insupatch; hyaluronidase; liraglutide; SGLT2 inhibitors, dual SGLT1/2 inhibitors; Hotel Study (with Stanford and BDC) Exercise/Snacking Study Young Child closed-loop study (with UCSB) Collaborations in Development: Princess Margaret Hospital; Schneider Children’s Hospital; University of Cambridge; Illinois Institute of Technology
Montréal/McGill	Pump vs. dual hormonal vs. single hormonal Dual hormonal vs. single hormonal Closed-loop and carbohydrate counting
Oregon Health and Science University	Bihormonal artificial pancreas research
Illinois Institute of Technology (Ali Cinar)	CGM fault detection Exercise classification module Module for detecting rapid glucose increases and insulin bolusing
Cambridge	Three free living studies completed Two home studies underway One home study in preparation
DREAM	Closed-Loop Studies At Home Over the weekend 60 hours at home study ongoing Two weeks 24/7 closed-loop at home ongoing 3 Months overnight closed loop study at home ongoing
Australia (Dr. Tim Jones)	Australian at Home Overnight Closed Loop Study Predictive Low Glucose Management RCT Home Trial Hybrid Closed Loop (Medtronic) in free living conditions at home
AP@home	2x2 month dinner-to-breakfast home trial 2x3 month 24/7 home trial
Bionic Pancreas	Insulin + glucagon vs. insulin-only set-point study (MGH) Insulin-only set-point study (Stanford) 12-month Pivotal Study

Large Efficacy Studies for Artificial Pancreas Systems

Moshe Phillip (Schneider Children's Medical Center, Petah Tikvah, Israel), Steven Russell (MGH, Boston, MA), Roman Hovorka (University of Cambridge, UK), and Boris Kovatchev (UVA, Charlottesville, VA)

Rapid-fire presentations from Drs. Moshe Phillip, Steven Russell, Roman Hovorka, and Boris Kovatchev shared updates on each group’s future efficacy studies that are coming soon. We heard never-before-details from each speaker and were impressed to see how actively and ambitiously the academic groups are thinking about large-scale efficacy studies (in some cases with an eye towards FDA approval). Competition is really pushing the thinking and dialogue forward aggressively, and we can hardly wait for the host of real-world data the coming 12-24 months will hold. It is going to be a truly exciting couple of years for the artificial pancreas. Please note that some of these studies have already begun (Project Nightlight), some are planned for the near future (Bionic Pancreas pivotal, IDCL), and some do not have public timelines (DREAM pivotal). We summarize updates from each group below:

Planned Large Closed-Loop Efficacy Studies

	Study Length/Size	Design	Comparator	Endpoint	Commercialization plan
DiAs – “Project Nightlight”	n=84; 11 months	Patients will cycle through SAP therapy, nighttime, 24-hour, and nighttime (a second time) closed-loop control	SAP vs. overnight vs. 24/7	A1c, risk of hypoglycemia, patient preference	Established a company, TypeZero to commercialize algorithm either on a smartphone or built into a pump
DiAs – “International Diabetes Closed-Loop Trial” (IDCL) Pivotal trial of commercial grade artificial pancreas built by Type Zero and in collaboration with pump companies	N=240; 6 months	Randomized patients 2:1 to closed-loop control vs. SAP; stage one: control-to-range vs. SAP; stage two: enhanced control-to-range vs. SAP	SAP therapy	A1c, incidence of hypoglycemia
Bionic Pancreas Pivotal Study To obtain FDA approval for the device and a chronic use indication for Xeris’ glucagon	12 months; n=480 Xeris glucagon Dual-chamber Pump	Randomized 2:1, parallel Bionic Pancreas vs. usual care arms; no remote monitoring or restrictions on foods/exercise	Usual care (CSII or MDI)	A1c, % time spent < 60 mg/dl during the last three months of each arm	???
Cambridge	12 months; n=130	Randomized patients 1:1 on 24/7 closed-loop control vs. SAP therapy; no remote monitoring or restrictions on foods/exercise	SAP therapy	???	Working with pump and CGM companies
DREAM pivotal study	6 months; n=?	“Parallel studies looking at both safety and efficacy”	SAP therapy	% of readings < 70 mg/dl; time within 70-180 mg/dl; A1c at six months	Licensed to Medtronic
Medtronic US pivotal study	3 months; n=150 Now Recruiting	Single arm study; two-week run-in on pump+ CGM followed by 3 months on the 670G	SAP therapy (during run-in period)	Safety; A1c; insulin dose; weight; time spent with closed-loop on vs. off; time spent in different glucose ranges	Commercialize internally

Panel Discussion – Large Efficacy Studies for Artificial Pancreas Systems

Roman Hovorka, PhD (University of Cambridge, UK), Boris Kovatchev, PhD (UVA, Charlottesville, VA), Steven Russell, MD, PhD (MGH, Boston, MA), Moshe Philip, MD (Schneider Children's Medical Center, Petah Tikvah, Israel), Stayce Beck, PhD (FDA, Silver Spring, MD); Chip Zimliki, PhD (Medtronic, Los Angeles, CA), Krishna Venugopalan, PhD (Animas, West Chester, PA), Tomas Walker (Dexcom, San Diego, CA)

Dr. Aaron Kowalski: Steven, you laid out the clinical trial design. When we did the JDRF CGM trial, we thought about this a lot. You’re going for a dual outcome – it’s bold. You’re shooting for the moon, and I commend you. The preliminary data suggests you can do this. The concern is that you have people that are already there, so you’re already putting yourself behind the eight ball. You may be treating people with an A1c of 6.5%. You have an A1c outcome, which is logical. But can you talk about that decision?

Dr. Steven Russell: Quite a few people in our trials have already had good glucose control. In most of the trials, half had an A1c <7%. So it’s primarily those in poor glucose control who see a reduction in mean glucose. For those in good control, we reduce hypoglycemia. Our population of patients so far has been pretty sophisticated folks at the time of enrollment. We think when we can enroll a much more general population and people with higher A1cs – not necessarily those who hear about the trial and fly across the country to participate. That will make it easier to see reductions in mean glucose. In terms of a reduction in hypoglycemia, that’s what takes more power. In the multicenter home use study, we reduce time in hypoglycemia from 1.9% in usual care to 0.6%. It was still a very large reduction and highly statistically significant.

Q: Do you have any reactions to that? Any aspects of this that you don’t think they are capturing with the studies they have described?

Dr. Krishna Venugopalan: I think it’s good to see longer-term studies being driven by academic investigators. It allows us to understand what the true solutions are going to provide. Extended data is essential to access, not just approval. It’s very exciting to see that. We’re very excited to be working with some of these individuals that are moving this forward.

Mr. Tomas Walker: I’m happy to see the focus on reducing hypoglycemia. The impact of hypoglycemia cannot be overstated. Despite the focus on severe hypoglycemia, the impact we’re seeing on non-severe hypoglycemia at night is really validated by trials. I’m happy to see addressed.

Dr. Stayce Beck: Everybody talked about how they are willing to talk to us. Each system might have different efficacies, but each of them will be able to show what their system can do.

Dr. Chip Zimliki: I applaud Stayce and the FDA for being so flexible. Medtronic has taken an aggressive approach to accelerating our artificial pancreas. We are focused on safety endpoints – the critical factors such as DKA and severe hypoglycemia. While I applaud those looking at long-term trials, we want to get a first generation product out as soon as humanly possible.

Ms. Hutton: Can you talk about the user interface and how important that is? In these large efficacy studies, how are you capturing the importance of the patient perspective in designing such systems and capturing that in clinical studies?

Dr. Roman Hovorka: I think that for the success of closed-loop systems, human factors are as important as the algorithms. If people are not willing to use the systems, you are not going to get the benefits. We are working with psychologists and human factors engineers. I think the challenge is that we don’t have the right metrics to capture some of these things. We’re trying to do the best we can, but I don’t think we have the best tools to understand how people perceive these systems.

Dr. Boris Kovatchev: We have had the advantage of using a relatively unchanged interface in ten different centers across the world. The interface was originally designed using focus groups and psychologists and human factors engineers. In these three years of use, we’ve collected a lot of data. We’ve figured out what people would like to have changed and are aiming for a refined version.

Dr. Moshe Phillip: You don’t need many studies to understand that patients want these systems as simple as possible and want systems that minimize the number of devices.

Q: You used the metaphor of the last 100 yards. As we get closer, could we run into the paradox where you keep halving the distance and not quite getting there? What are the barriers and potential pitfalls that we might run into?

Dr. Kowalski: Very, very good point. I think I see that question from both sides. I do commend industry – Medtronic really threw down the gauntlet. Medtronic is really pushing the entire industry towards these first systems. But it’s just one finish line, and there are going to be multiple. The other side of that equation is that I worry about being overly conservative. One of the interesting things, as we approach that goal line, is whether we are going to alarm too much. Are we not going to give people enough control, or be too conservative and risk averse? The expectation setting is critical. How much do you alarm?

Dr. Beck: I think what we talk about is the clinical study. What are the endpoints? And then there is manufacturing. We’re working with JDRF on different manufacturing models for different component systems; it’s not a small part, and it’s not nearly as exciting. But we need to start shifting our conversations.

Mr. Walker: From the standpoint of a CGM in the artificial pancreas, the alarm has to be actionable. We need to change how we think about them: alarms vs. action alerts. Alarms lose value when you cannot do anything about the alarm. It’s well reported in the literature across aviation and healthcare. Alarms have to mean something. It’s an action alert.

Dr. Kowalski: You know this very well at Dexcom. If you give too many action alerts, particularly in teens and young adults, it really becomes an issue. Where is the balance between safety and dis-incentivizing people from using these systems?

Dr. Zimliki: We use a risk-based approach. That’s what Medtronic is doing. You need a prioritization of alerts, so that some will supersede others. That needs to be done automatically. That’s the purpose of what we are trying to do.

Dr. Russell: However, even if these systems are actionable, it would be better if you had a system that would prevent the low from occurring in the first place.

Mr. Brandon Arbiter (Tidepool, Palo Alto, CA): Some of the questions we talk about at Tidepool are: who are the early adopters of this technology? How do we manage the chasm between early adopters and everyone else? I think it’s interesting to consider whether certain people are predisposed to certain systems.

Dr. Russell: I do think that we’ve seen early adopters enrolling in these trials so far. We’re trying to avoid that now by making sure one-third of our pivotal trial patients are MDI users. If you make a system that is simple enough, then you should be able to draw in people that are more than early adopters. You’d have to think that a simple system would be more attractive than a sensor-augmented pump.

Mr. Venugopalan: Let the user determine the extent of control and the extent of sharing. A key part with these solutions is disrupting the pump business model of sticking with something for 4-5 years. It becomes something with upgradeable features and remote upgrades. More diagnostics. You may be able to tailor solutions to individuals and iterate much more rapidly. It’s not just innovation from an algorithm perspective – it’s innovation from a business model perspective.

Dr. Kovatchev: The system will have to be adaptable in real time. One person may want something today and a different configuration tomorrow. You need to provide a different solution at different times. That has to be embedded in system design.

Dr. Hovorka: I think the design of clinical studies is difficult. I think it’s interesting to consider those who are putting pregnant women on closed-loop bot during and following delivery. That’s very real life.

Adam Brown (Close Concerns, San Francisco, CA): What is the appropriate comparator group in a closed-loop pivotal study? Is it MDI users on SMBG – the majority of people with type 1 – or is it the best that we have right now: sensor-augmented pump therapy?

Dr. Phillip: I would say that you have to design the control group in a way that allows you to test the primary outcome of your study. If you have a control group that is two steps removed – such as Dr. Russell’s that adds a sensor and closed-loop – then you don’t know if final effect is because of the sensor or because of the closed-loop system. You’ve got to make sure your outcome reflects what you want it to reflect.

Dr. Russell: This is not an academic exercise. We give people systems to improve their diabetes control. We don’t want to limit ourselves to the people using pumps – that’s only 1/3 of the people with type 1 diabetes. People have chosen not to use a pump or a sensor, or maybe their HCPs believe they are not good candidates. Those are exactly the people who need an automated system to take care of them. Most people with type 1 diabetes are on MDI. I don’t think we should take that person and give them a run in on sensor-augmented pump therapy – they never felt they wanted it in the first place. You try and do that and you have a very long run in period, and you still don’t know how to implement that therapy. To use that as a comparison group doesn’t make sense. What benefit could you provide to patients that are actually candidates to use the devices? It’s not an academic question about which part is due to the control algorithm. What if people don’t want to use sensors and pumps because they don’t get enough benefit?

Dr. Kowalski: Steve, Steven. I know you’re a reviewer. You know that all the major journals will critique that approach. Medtronic faced this in their NEJM paper.

Dr. Russell: This has been a major concern. But if we want to get approval for this device in the general population, we need to include MDIs. We do our studies for patients, not peer reviewers.

Comment: Steve, we’ve done exactly what you’ve said. We took patients on MDI or on pump or on SAP, and we gave them a seven-day run-in. We saw no difference in time-in-range or adaptation whether they used a pump, a sensor, or MDI. If you look at the published data, you cannot tell which is which.

Dr. Hans De Vries (Academic Medical Center, Amsterdam, The Netherlands): I think both are right. If you just look at CGM development, we’ve seen initial trials in CGM looking at pump users in the control group and then we saw comparisons to sensor-augment pumps in the control group. We’ll see the same thing again here. You might need to convince the payers by comparing to the optimal treatment. After all, why would they pay for something new if they don’t have the optimal treatment? But you’ll also have patients who aren’t on pumps and CGMs, but want to get on the whole package.

Dr. Roy Beck (Jaeb Center for Health Research, Tampa, FL): We’ve worked with two groups here. I could write a point-counterpoint on this, because I think you could do both scientifically. What we really want it a three-arm study. But ultimately, I come around to Steve’s point. In the T1D Exchange data, at most 15% of people at the top centers are using sensor-augmented pump therapy. The reason for that is not all about insurance. It’s because of the reasons we’re all alluding to – the hassle factor. If you could pop this on and use it, there’s a really expanded population for it.

Dr. Yogish Kudva (Mayo Clinic, Rochester, MN): What lessons have we learned and not learned from the past year? Let’s think about the use of STAR 3 to approve devices. It has been sub-optimal. I think that there are issues of making sure providers are educated. The other issue is the payer issue. One of the payer issues is that when you have a refusal that has come back from the insurance company, it is an opinion that has come from someone in the middle. That is the real world in the US. There is a middle-man. I think in the US that’s an important issue. The third point is that we know a lot about patient factors with regard to device use. But what about the MDI use? I don’t think we know enough about that yet.

Dr. Kowalski: I think you are right. All of this sums up into patient access and improved outcomes. The components of patient access are very complex but these studies play a very important role, as my team will attest. We sit with these payers and they will criticize the design of these trials and that does have downstream affects on decisions. So those are important clinical questions. I think the glass-half-full view is that the collaboration between this community helps, and I’m convinced that these systems are the tipping point in terms of improved control and reducing burden concomitantly. I have to believe payers will catch on.

Panel Discussion – Adoption of Artificial Pancreas Systems

Jessica Roth (JDRF, Washington, DC); Adam Brown (Close Concerns, San Francisco, CA); Kenneth Snow, MD (Aetna, Boston, MA); Paul Radensky, MD (McDermott Will & Emery, Washington, DC); Jill Weissberg-Benchell, PhD (Feinberg School of Medicine, Chicago, IL); Brian Herrick (JDRF, New York, NY)

JDRF’s Jessica Roth and Close Concerns’ Adam Brown co-moderated this panel of payers, providers, and patients, who discussed expectations for and adoption of various artificial pancreas systems. The payers on the panel – the very articulate endocrinologist-turned-payer Dr. Kenneth Snow of Aetna and Dr. Paul Radensky of McDermott Will & Emery (a law firm specializing in reimbursement) – had crystal clear advice for the entire room: artificial pancreas systems may prompt benefits on A1c, hypoglycemia, and burden, and to get it reimbursed, the data must clearly show that. Dr. Snow used the analogy of buying a mouse trap – there are $3, $5, and $30 mouse traps, but when you go to buy a mousetrap at the store, there is no data to point out which is better. In that case, the consumer probably just buys the cheapest one. The same is true with the artificial pancreas: Aetna will pay for the “better mouse trap” if there are data to clearly suggest that. Psychologist Dr. Jill Weissberg-Benchell and type 1 patient Brian Herrick emphasized the artificial pancreas’ quality of life benefits (e.g., less worry, a parent that doesn’t have to wake up at night), and the payers made it clear that those softer endpoints are indeed seriously considered in the payment decision; however, the data must be rigorously collected with validated instruments to prove it’s true (we are not sure how this will be done, though HCT has funded a group of researchers to develop such approaches). Adam pointed out the challenge in that tradeoff – capturing powerful, transformative stories about the impact quality of life and summarizing them with what can seem like cold, quantitative statistics (e.g., an 8.6 on a 1-10 worry scale). Both payer representatives cautioned against being deemed a “convenience” benefit alone, a weak argument to justify payment.

• This panel emphasized that engaging payers early and often will be critical for properly designing artificial pancreas pivotal studies. And this time is now. Indeed, Dr. Roman Hovorka wondered whether any learnings from the JDRF CGM study could be applied to reimbursement for artificial pancreas technologies – Dr. Kowalski agreed that they could, highlighting how hard JDRF worked early to seek input from payers in planning the study. What was interesting to hear, however, was that different payers weighed different endpoints more heavily (e.g., A1c vs. hypoglycemia). That suggested that any ultimate study design is really going to be a compromise, meaning it may not satisfy all stakeholders (patients, payers, clinicians) or goals equally.
• We left the panel and the entire evening with a strong appreciation of the nuances in pivotal study design; we hope academic investigators and companies are heavily leveraging JDRF’s expertise and network to do this properly.

Quotable Quotes

• Dr. Kenneth Snow (Aetna): “Does technology improve outcomes? If it does, please prove it to me – we’re happy to cover it if you do. Lower A1c, les hypo, less distress – we all know what needs to be shown. If you can’t show me, I won’t cover it.”
• Brian Herrick (JDRF, type 1 patient): “For the first system that works, I’m going to do everything in my power to get it. And I’m not the only one that feels that way.”
• Dr. Paul Radensky: “Through the work with Medicare and Medicaid we’ve seen a willingness to accept quality of life outcomes. However, it’s tough if you do not come first with science outcomes. Time in range is something they are willing to look at – I looked before a home anti-coagulant monitoring, where having numbers in range was important. I don’t think it’s an either/or – it’s both sets of outcomes that you need.”
• Dr. Jill Weissberg-Benchell: “I think we first need to improve the lives of people that live with diabetes, which means a real-time everyday perspective. And that’s not A1c, because that’s not a real-time everyday perspective. Can a child sleep through the night? Can a parent sleep through the night? Is there less conflict between parents and children? Is the child a better friend? That to me is the most powerful thing that a closed-loop system can provide.”
• Dr. Snow: “A disclaimer. I’m an endocrinologist. I’ve had patients die in the middle of the night. So that does hit home and I understand the issue. But I’m not going to claim that all of my colleagues will and that all payers will. There is a significant value that goes beyond A1c and beyond hypoglycemia. All of these things are very important and should be quantified as part of a study. Provide me with the evidence.”
• Dr. Carol Levy: “I’m the person who has to fill out the forms to get reimbursement. But no one ever asks for psychological factors. They always ask for numbers.”
• Dr. Snow: “I think the harder challenge is when the innovation is moving faster than the literature. You might have four studies that are in print and coming out eventually, and yet the data that is currently out there is outdated; that is much harder to do. Some people come to us and say, ‘Hey here is the result of the study before it is in print.’”

Joint ADA/JDRF Symposium: Closed-Loop Technology in Youth – Real Estate, Alarms, and Challenges

The Artificial Pancreas Road Map Update – Too Slow, Too Fast, When Does My Child Get an Artificial Pancreas?

Aaron Kowalski, PhD (Chief Mission Officer/VP, Research, JDRF, New York, NY)

Dr. Aaron Kowalski’s AP roadmap presentation mixed outrage with optimism, highlighting what he characterized as incredible unmet need and the very promising status of artificial pancreas systems (including multiple slides on Bigfoot, Bryan Mazlish). Part one of his presentation used T1D Exchange data to highlight how few people with type 1 are meeting glycemic goals – this data is still quite sobering, and Dr. Kowalski made a major point that it’s not due to “compliance” (i.e., patients even testing 7-8 times per day are still struggling). Part one concluded with a stark slide centered on two big bolded words to summarize how quickly artificial pancreas efforts are moving: “TOO SLOW!!” Dr. Kowalski was optimistic, however, and deliberately did NOT show his six-step roadmap – “We’ve achieved technical feasibility across all steps,” he said, and now it’s time for (i) commercial development; (ii) regulatory approval; (iii) reimbursement; (iv) clinical adoption; (v) people with diabetes access/benefit; and (vi) improved outcomes. JDRF sounds very focused on all these buckets, particularly reimbursement. Dr. Kowalski noted that “multiple companies” are in commercial development and are targeting 2017-2018 launch of hybrid systems (he did not specify, but we know of at least Medtronic, Bigfoot, Bionic Pancreas, Tandem, and potentially Insulet, Animas, TypeZero, Cambridge, and Roche). Dr. Kowalski was equally positive about the FDA pathway to market, which he characterized as “very favorable.” Moving forward, Dr. Kowalski has proposed a “Diabetes Scorecard (Diabetes Care 2015) to judge the value of artificial pancreas systems beyond A1c – scorekeepers are the people with diabetes (& loved ones), clinicians, and payers, and the categories are glycemic control, “burden”, and value.

• Dr. Kowalski’s Diabetes Care paper  now proposes a three-step roadmap to artificial pancreas systems, with a bifurcation occurring at step three: automated insulin delivery (AID) vs. multi-hormone (MH) systems. The first two steps – low glucose suspend and predictive suspend – are already available and commercialized as the MiniMed 530G/Veo (Worldwide) and MiniMed 640G (EU, Australia). This is in contrast to the initial model: (i) low glucose suspend; (ii) predictive low glucose suspend; (iii) hypoglycemia/hyperglycemia minimizer; (iv) automated basal (hybrid closed loop: manual meal bolus); (v) fully automated closed-loop (insulin-only); and (vi) fully automated closed-loop (multi-hormone).
  
  • The “next steps” column (see below) makes it clear that multi-hormone approaches still have much to prove. Overall, the paper takes a somewhat skeptical eye towards bi-hormonal systems, particularly glucagon. This is not that surprising, as Dr. Kowalski has long questioned the incremental benefit that glucagon might add over insulin-only systems. We look forward to head-to-head trials that will bring data to the discussion.

Dr. Kowalski’s Revised AP Roadmap
System	Next Steps
1. Suspend	Clinical adoption
2. Predictive Suspend	US approval, reimbursement, clinical adoption
3a. Automated insulin delivery (AID)	Faster insulin action, miniaturization, integration, individualized control
3b. Multi-hormone (MH)	Glucagon: soluble, pumpable glucagon; chronic glucagon exposure studies; dual-chamber pump development, algorithm finalization, head-to-head vs. AID Amylin: co-formulation vs. dual-chamber pump, ratio determination, algorithm finalization, head-to-head vs. AID

• Dr. Kowalski devoted two slides to a Diabetes Mine guest column by Sarah Mazlish (Bigfoot), which underscored the reduced “burden” artificial pancreas systems can provide (even to those in very good control like Sarah). Though he acknowledged artificial pancreas systems might be more expensive than other therapies, Dr. Kowalski believes their “value” will be worth it, since outcomes will be so much better. [We’d note that it’s still hard to know how the cost of commercialized artificial pancreas systems will stack up to current therapies; what the clinical data will look like vs. current therapies; and what fraction of patients will switch to closed loop systems. We do believe the value will be there for many patients, but it’s hard to know what fraction of type 1s that is, and how it will change over time with subsequent generations.]

Pediatric Patient Safety and Patient Products – The FDA Perspective

Stayce Beck (FDA, Silver Spring, MD)

The FDA’s Dr. Stayce Beck provided strikingly optimistic commentary on the pathway to a pediatric closed-loop system – “[We want] a fully approved system available for anyone that needs it.” We would like her to advocate before CMS when the time comes! Dr. Beck opened her presentation emphatically, noting that closed-loop approval in adults and pediatrics can – and should, in the FDA’s eyes – occur in parallel. She acknowledged the widely held misconception that an adult approval must precede pediatric approval – indeed, she stated that “that is the opposite of what we want.” Dr. Beck stressed a host of differences that introduce nuance into the transition of closed-loop systems into pediatric populations: differences in body fat, activity level, glucose variability, hormones, and capacity for decision-making between adults and children. In light of these differences, she suggested that closed-loop systems must be tested separately in children in order to pave the way for approval – and also that the FDA is very willing to consider this need. Of course, she acknowledged the FDA’s view that appropriate mitigation techniques, such as remote monitoring, are paramount in this more vulnerable population. She also stressed that such safety measures do not devalue the data in the FDA’s eyes, quelling concerns that the use of such techniques could hamper confidence in results or imply a distrust of the system. Dr. Beck advocated for more minimal exclusion criteria for pivotal studies in general, stressing that the FDA would rather identify patient sub-populations that should not use closed-loop systems through clinical trials, as opposed to off-label use leading to disastrous results in the real world. We’re pretty much ready to start a Stayce Beck fan club – she just continues to impress in public presentations with her patient-friendly perspective.

Pathway to the Pediatric Bionic Pancreas

Ed Damiano, MD (Boston University, MA)

Dr. Ed Damiano’s comprehensive overview of the Bionic Pancreas in pediatrics shared an updated timeline on the group’s pre-pivotal studies – notably, two glycemic set-point studies will take place starting in August. An MGH study will test insulin-only and bi-hormonal configurations of the bionic pancreas head-to-head – the same participants will use both configurations for at least one set-point (130 mg/dl), though the team will also expects to test other target levels (insulin-only 145 and possibly 115 mg/dl; bi-hormonal: 100, 115, 130 mg/dl). A separate Stanford set-point study will test an insulin-only version of the Bionic Pancreas at several algorithm set points, and will do so over a longer period of time. [We first reported the insulin-only news in less detail at GTC Bio in April – though we had been fairly surprised at the time, it makes complete sense given the uncertainty around multiple elements of glucagon.] We see this as a necessary move to counter the group’s critics, especially those asserting that the system doses insulin and glucagon too aggressively and would be unable to fail safely. It also speaks to the team’s willingness to let the data speak for itself – they could easily just move ahead to the bi-hormonal study, assuming glucagon availability. Last, we believe it could be a mitigation strategy in case an approved stable glucagon does not come in time for the Bionic Pancreas pivotal study (2016-2017). ADA certainly continued the fascinating insulin-only vs. bi-hormonal debate, and we’ve heard a range of opinions in recent months (e.g., Keystone 2014, DTM 2014, ATTD 2015, ENDO 2015); ultimately, true head-to-head data is needed to show what incremental value glucagon adds to the equation. Our favorite commentary came from Yale’s Dr. Stuart Weinzimer in Q&A: “I think it’s a false argument to compare one against the other. What I would like to see are options for patients with diabetes and clinicians. I want to see an insulin-only system and a bihormonal system on the market. Then clinicians can choose and help patients choose what is best for them. So let the market decide.” Amen to that.

• As a reminder, the Bionic Pancreas team still hopes to conduct its pivotal study in 2016-2017, with FDA premarket approval slated for 2017-2018 and commercial launch in 2018. We saw results from the recently completed 11-day multicenter study for the Bionic Pancreas at AACE 2015; Dr. Damiano showed these again at ADA, but we did not see anything new.

Practical Translation of Closed-Loop Technology to the Pediatric Clinic – The Clinical Diabetes Educator's Perspective

Laurel Messer, RN, CDE (University of Colorado, Aurora, CO)

Ms. Laurel Messer shared updated results from a predictive low-glucose suspend study (PLGS) in adults and children. (This was not Medtronic’s MiniMed 640G but rather a custom algorithm replacing the Veo’s low-glucose suspend feature.) The study enrolled patients 3-45 years old (n=unknown) who were randomized to either CGM alone or PLGS overnight for 42 days. Results indicated that PLGS reduced clinically significant hypoglycemia (for > two hours) by 74% (patients 15-45 years old), 62% (11-14 years old), and 53% (4-10 years old). Ms. Messer attributed the difference in efficacy to disparities in adherence – 4-6 year olds wore the sensor almost one fewer day than other age groups due to skin issues and CGM usability. Within the same group, 64% experienced skin reactions and 36% had moderate/severe reactions to the device. Ms. Messer stressed that such human factors have big implications for the penetration of closed-loop technology in this younger population and called for more dialogue on this front.

• Broadly, Ms. Messer stressed that there is very little information about the usability of closed-loop technologies in children and that expectations are unrealistically high. As an example, Ms. Messer cited a survey in which a majority of patients reported feeling reluctant to start pump therapy (58%), but also reported that they would have “no barriers” to using closed-loop technologies (79%). Ms. Messer suggested that such irony is presently par for the course and that the field needs to start thinking harder about the human factors side of artificial pancreas technologies.
• Ms. Messer stressed that different pediatric subpopulations have unique needs in terms of closed-loop human factors design. In the youngest children (patients 3-10 years old), she noted that the artificial pancreas will not be able to decrease the care burden until trust with parents is established. There are also needed CGM improvements to better deal with skin sensitivity. In older adolescents, Ms. Messer suggested that there are challenges associated with increased autonomy in this age range that tends to lead to more frustration and greater burnout risk, highlighting that education around calibration is key (for all patients, we would add).

Panel Discussion

Q: What is the status of a soluble glucagon? And is there potential to integrate an accelerometer into the pump? Could that be integrated into the pump to avoid hyperglycemia for a 10-second sprint and hypoglycemia with prolonged exercise?

Dr. Ed Damiano (Boston University, MA): I’ve never been a big fan of using an accelerometer for automatic glucose control. In my own personal experience, if my son is playing basketball in practice, his blood glucose will go down. If he’s doing the exact same thing but competing in a game, his blood glucose goes up – he is tense, excited, and there’s hormonal activity. These are the same scenario – the accelerometer wouldn’t know the difference. There is such variability that it’s not a good read. The ultimate signal we want to control is glucose – that should be the only read. And the results reading only glucose seems to bear that out.

The soluble glucagon is a big one – it’s a challenge that there’s no stable, pumpable glucagon that is FDA approved. There are some platforms out there that have shown stability out to more than a year, and some up to two years. Xeris is one; Zealand is working on an analog. There are other groups out there. What we’re doing right now has no commercial path forward – we have no expectation of using reconstituted Lilly glucagon. It’s not chemically stable long enough. The challenge can be solved in more than one way, and there are good candidate drugs out there. We have talked about the roadmap with the FDA and what it would look like for one of the drugs out there. Some are into phase 3, some are just in phase 1. I’m optimistic we’ll see them coming along very quickly.

Q: Is it necessary to have glucagon? Even in young kids? Can the Bionic Pancreas work insulin-only?

Dr. Ed Damiano: We have yet to test insulin-only, so I would say to come back to me in a few months. My intuition is that in kids, more than anyone else, we need glucagon. We’ve evolved to use glucagon. To preserve the spontaneity in life, we have to preserve glucagon. We use it. And it is the first line of defense for people without diabetes. So it is definitely something I want to see in an automated system in the future. We could introduce insulin-only systems to bridge the gap to glucagon systems.

Dr. Aaron Kowalski (JDRF, New York, NY): I think this is an important area that we need to focus on: the benefit of insulin alone vs. bihormonal systems. What we are seeing is that insulin-only approaches will work. And they will work very well. One of the things I just wrote abut is that we need to be careful about the comparator when we think about diabetes. Some have been critical about artificial pancreas systems because they claim that these systems won’t be like islet cells. If you’re comparing against islets, you are doomed to fail. But if you are comparing against the standard of care today, you can do better. Insulin-only systems can do better and the data bears that out. I do think that glucagon can add a level of control. I just think that there is a level of dysregulation that we need to understand.

Dr. Damiano: It’s not like it’s one or the other.

Dr. Kowalski: No I agree. It’s not one or the other.

Dr. Stuart Weinzimer (Yale University, New Haven, CT): I think it’s a false argument to compare one against the other. What I would like to see are options for patients with diabetes and clinicians. I want to see an insulin-only system and a bihormonal system on the market. Then clinicians can choose and help patients choose what is best for them. So let the market decide.

Q: Do you think faster insulin would help your systems? And second, if I just ate at a restaurant and don’t want to wait for a cab and decide to walk to a hotel, I would often get hypoglycemic. That’s one of the variables that really frustrates people. In your population, were you seeing that glucagon was enough to rescue patients in that situation? I know people were trying all crazy kinds of things. Do we just need faster insulins?

Dr. Damiano: I really want faster insulin. It is going to make our system better. However, I would agree that what we have now is good enough. It’s good enough not just for hybrid systems but even for a fully automated system. What we have is good enough for really good glucose control. We published a paper to show exactly that. We took an adolescent population and looked at half without meal announcements and half with meal announcements. The difference between the groups was only ~13 mg/dl. So even if you forget meal announcements, we have a system that works to bring adults to ~7.0% and teens to ~7.5% without hypoglycemia. Despite that, I do want to see insulin get faster. It makes bihormonal systems better. It makes insulin-only systems faster. It makes these systems more realistically approximate physiological conditions.

Q: Can you talk more about whether an accelerometer would help?

Dr. Damiano: I don’t think an accelerometer is predictive of glycemia. There is nothing more challenging than our summer camp setting. In Beacon Hill, we had one woman walk 35 miles in five days. In the multicenter study, people could use gyms as much as they wanted. Our system dealt with active people just fine.

Q: Have you ever followed up with patients who were very active the following week and seen how the system adjusts to less activity? Can it adapt again?

Dr. Damiano: The timescale of adaptation is about 18 hours. However, this varies depending on the insulin usage. For example, we had one patient who was receiving 0.5 units/kg/day and adaptation was almost instantaneous. However, in patients that received more insulin, adaptation took a bit longer. It adapts pretty darn quickly. I think it handles dramatic changes quite well.

Q: As a camp doctor, I’m interested in how much down time there was during your studies when a site was out?

Dr. Damiano: Typically, if we lost a sensor – which happened a lot – you would have a two-hour period, maybe up to three hours, with no CGM data. The system learns the normal basal rates and invokes those even when the CGM sensor is offline.

Q: Can you predict any differences in parental buy-in between insulin-only and bi-hormonal systems?

Dr. Messer: We’re excited to work in the future with insulin and glucagon in the Bionic Pancreas. My initial concern was the burden of two pumps. The hassle was very concerning. More studies need to be done in pediatrics, with insulin-only systems, and with bihormonal systems.

Dr. Damiano: I don’t know of any studies that have compared the buy-in of insulin-only vs. bihormonal systems. I just want to clarify that the system we use now uses two infusion sets. However, we’re building a dual-infusion set that you would have to change every 72 hours. It will be stainless set; it will have separate catheters that will use separate paths. It will require no more skin real estate than a current infusion set.

Dr. Messer: Still, it’s actually the CGM that causes the most problems with body real estate. The CGM is still really a major player here in terms of burden.

Dr. Weinzimer: Looking at user factors and quality of life, there’s some great data that Kat Barnard collected, but we have to be very cautious. Most studies are 5-10 days per patient at most. No matter how burdensome it is the device is, 5-10 days is very different from 3-6 months. We’re going to get a much better idea of the burden when we start doing pivotal studies.

Dr. Kowalski: it’s a very, very important point. You’ll see the insulin-alone systems come to the market first. I’m almost certain. I believe we’ll have to do the studies on incremental benefit of bihormonal control to see whether the glycemic benefit is distinguishable from an insulin-only system or liraglutide or amylin. We’ll have to look at the glycemic benefit, the severe hypoglycemia, the mean blood glucose, the time-in-range. Are those endpoints distinguishable enough from insulin-alone, given the greater cost and size? Where do we put effort in the future to drive systems that are more adoptable? Part of that is glycemic control, but part of that is burden. Glucagon systems may mean blood glucose is the mean, but patients have a much lower risk of severe hypoglycemia. It could be tremendously beneficial and very much worth the cost. These first systems are going to set the benchmark. We’ll have to drive from there.

Q: Such a gap exists between the patients that do and do not have access to technology. Dr. Kowalski, I wonder where you see the JDRF’s role in advocating for payers to come to the table?

Dr. Kowalski: The ultimate metric for the success of closed-loop technologies is the number of patients that achieve better outcome. This is not just a tool for wealthy people. It’s for everyone. The JDRF has evolved quite a bit since I first joined. We are just about to kick off a major initiative with the Helmsley Charitable Trust that focuses on downstream issues for the artificial pancreas: clinical adoption and access. We’ve already had a number of discussions with payers on closed-loop technologies. I’m cautiously optimistic. I do think that we’re going to pivot from therapies that are driving incremental improvements to a big step forward in glycemic control. If you look at the changes in these studies, they are quite dramatic. We’re going to see both the burden and glucose control improve concomitantly. Just remember, CGM has only 10% penetration because of the many barriers. Patients do not think the incremental benefit is worth it. But with closed-loop systems, the incremental benefit is worth it. This is going to be a multifaceted campaign, because the ultimate metric is people with diabetes getting better.

Q: In pediatrics, there is very little real estate. Were any bionic pumps inserted in lipohypertrophic areas?

Dr. Damiano: No, not intentionally.  We inserted in the abdomen. Dr. [Bruce] Buckingham [Stanford University, Stanford, CA) suggested that sensors work well even in lipohypertrophic areas.

Dr. Laurel Messer (Barbara Davis Center, Aurora, CO): We actually found that sensors worked better in those hypertrophic areas.

Dr. Damiano: However, skin real estate is definitely a concern for pump therapy, whether automated or not. Closed-loop systems are only a bridge to a cure; they can’t solve all the problems of pump therapy. You have to be connected to a device. It’s a contraption. I’d love to see Doug Melton step up and show us a cure tomorrow. But this is a much better solution than what we have today. And this handles everyone. It’s an umbrella for everyone with type 1.

Q: In adult studies, did you look at food intake? If didn’t have hypoglycemia and hyperglycemia, were they just eating less?

Dr. Damiano: The 9,000 calorie-anecdote was on the bionic pancreas, which wasn’t usual care. We did not capture data on calorie intake. In the real-life setting, you don’t want to capture all the data or it’s not real-life. We did in Beacon Hill, and there wasn’t any difference compared to usual care. We have no reason to believe that subjects ate less.

Q: I work on closed loop tech for pregnant women in the UK. I completely agree with Dr. Beck that children and other marginalized groups need to be part of the studies. This includes children, pregnant women, and people with multiple co-morbidities. I understand that the FDA’s view is to focus on low-risk populations first in safety studies, but I worry this delays technology. How can we fast track this technology and make sure it gets from research to the clinic to the people who need it most?

Dr. Stayce Beck (FDA, Silver Spring, MD): We’ve been working a lot with different investigators to design studies that best protect patients but are also as real-world as possible. There will always be a need for some in-clinic studies first to make sure people are protected. This includes testing with more vulnerable populations before going to market. We do want to ensure that once the device is on the market, it can be safely prescribed off-label for anyone an HCP chooses. We want to make sure the device is studied to some extent in those populations.

Symposium: Diabetes Devices and Technologies – The Patient Experience

How a Device Becomes a Device

Adam Brown (Close Concerns/The diaTribe Foundation, San Francisco, CA)

Please see Adam’s slides here, which pinpoint five principles that helped the iPhone succeed, apply them to diabetes devices, and discuss key barriers – the visuals and stories really make the outline below come to life.

• Adam highlighted five key principles that helped the iPhone succeed: (i) Makes my life easier, solves a need, subtracts hassle; (ii) User-experience-obsessed, iterative, feedback-driven design; (iii) No instructions needed; “it just works”; (iv) Significant improvement over existing options; and (v) Fun, addictive, “awesome!”
• Adam applied this five-part framework to two diabetes device case studies – Insulet’s OmniPod and Cygnus’ GlucoWatch, discussing why one has largely succeeded commercially (despite recent business challenges for the US OmniPod business) and why the other failed quickly following launch.
• In the second half of his presentation, Adam shared seven principles why devices often fail to meet all five criteria: (i) Device development often focuses only on safety/efficacy to pass the FDA (Necessary, but not sufficient!); (ii) The regulatory bar is much higher for solving really important problems (e.g., like how much insulin to take); (iii) Not enough attention is paid to making it fun, addictive, and awesome; (iv) Lots of focus on “high tech” – sometimes the simplest solution is profound (e.g., BD’s AutoShield Duo Pen needle); (v) an over-reliance on outdated business models. Companies must disrupt themselves. (e.g., example of Kodak inventing the digital camera in 1976 and burying the technology); (vi) First gen devices aren’t great in any domain. We may be in the early days of where devices could go...(i.e., we must have patience); (vii) Too much focus on incremental instead of game changing improvement. (How about 10x better?).
• To build a great device, Adam told attendees to remember the acronym PLANE: Patient and Provider Feedback (early, often, and relentlessly); Life Easier, Less Hassle; Addictive, Awesome; No instructions needed; Exceed Existing options (10x!).

The Human Side of the Artificial Pancreas

Jill Weissberg-Benchell, PhD, CDE (Northwestern University, Chicago, IL)

Dr. Weissberg-Benchell discussed very interesting psychosocial research on the 40 participants in the multicenter trial of the bionic pancreas. Most notable were the slides qualitatively documenting patient perceptions – the take was overwhelmingly positive on less worrying while sleeping, greater peace of mind, relaxing, and trust in the device – see the data below. We’re very excited to see so much more work going into the real-world, human factors, psychosocial side of the artificial pancreas, and as Dr. Weissberg-Benchell mentioned, it’s fortunately happening before these products are on the market. Along with Drs. Katherine Barnard and Korey Hood, Dr. Weissberg-Benchell has a grant to examine closed-loop human factors in greater detail. We hope the learnings are leveraged as pivotal studies take place to make products as great as they can be, to manage sky-high provider expectations, and to make clincians’ lives easier.

Table 1: Experience of Bionic Pancreas – Positive Aspects

	% Agree or Strongly Agree
It helped me worry less about having a low while sleeping	78%
I had greater peace of mind while wearing the device	73%
It helped me to worry less about high blood sugars	73%
IT helped me to relax, knowing that unwanted change in blood sugar would be addressed automatically	73%
It helped me worry less about low blood sugars	73%
By the end of the study, I trusted the device to manage my blood glucose	68%
I found it hard to trust that the bionic pancreas could control my blood sugars	17%
Using the device was more trouble than it was worth	33%

Table 2: Experience of Bionic Pancreas – Negative/Bothersome Aspects

	% Agree or Strongly Agree
It was a big bother having to change the glucagon every day	78%
Carrying around all of the equipment was a burden	75%
It was uncomfortable to wear all of the necessary equipment	60%

Questions and Answers

Q: Will we always need glucagon?

A: Only 2 of the artificial pancreas studies are bi-hormonal. As a psychologist, I don’t have an opinion of which is better; a single system won’t meet the needs of everyone; so let’s get them all to market and give people a choice.

Q: What did the caregivers say?

A: We didn’t interview them yet, but have a Helmsley grant to do further studies.

Q: What did you think about the results of the WHO-5 (that quality of life did not improve much)?

A: It’s not really a sensitive enough measure (only 5 questions).  The more specific questions about impact of the APs are probably more relevant.  We now have an opportunity to define what will be used to define success for APS systems: it shouldn’t be just A1c.

Q: Interesting that people express so much frustration with alarms, even though they are what keep people safe.

A: We haven’t done enough to figure out how to give patients information and control they need to make the alarms work well for them.

Q: Can you give some more background about people who don’t continue use of pumps and CGM?
A: I can show you what data I’m citing – there’s a lot published about discontinuation rates of pumps, especially related to body image.

Q: In the 40 people in the study, was there a difference in people who have had diabetes for different lengths of time?

A: Outcome differences were not changed by duration of diabetes.

Symposium: Novel Clinical Interventions in Therapy That Impact the Management of Diabetes

Predictive Low Glucose Insulin Suspension Reduces Duration of Nocturnal Hypoglycemia in Children without Increasing Ketosis

Bruce Buckingham, MD (Stanford University, Stanford, CA)

Dr. Bruce Buckingham shared encouraging results from an in-home predictive low glucose suspend study (not Medtronic’s MiniMed 640G), which spoke to the promise of this technology in an impressively large and well-conducted RCT in pediatrics – the 81 participants (4-14 year olds) were blinded to whether PLGS was turned on (n=1,724 nights) or off (n=1,696 nights). The percentage of nights with a glucose <60 mg/dl for 120 minutes declined from 8% to 3% (p<0.001) during PLGS in 11-14 year-olds, and from 5% to 1% (p<0.001) in 4-10 year-olds. The magnitude of benefit rose considerably as the duration of nocturnal hypoglycemia increased (e.g., in 11-14 year olds, nights with hypoglycemia episodes >30 minutes declined 36%, while nights with an episode >180 minutes declined 68%). Mean overnight glucose was slightly higher (+7-8 mg/dl) during PLGS in both groups (152 vs. 144 mg/dl and 160 vs. 153 mg/dl), but the tradeoff for hypoglycemia was unquestionably worth it. This is now published in Diabetes Care. It was truly striking to see that patients spent more than 1 out of 20 nights hypoglycemic for more than two hours – Dr. Aaron Kowalski repeatedly highlighted this data through ADA.

Day and Night Closed-Loop Control Using the Integrated Medtronic Hybrid Closed-Loop System in Type 1 Diabetes at Diabetes Camp

Trang Ly (Stanford University, Stanford, CA)

Stanford’s Dr. Trang Ly shared a bit more about the Medtronic MiniMed 670G’s adaptability, courtesy of a deeper dive on the camp study first presented at ATTD. Dr. Ly presented the efficacy of the system by day of wear, which actually showed solid adaptability – overall time-in-range (70-180 mg/dl) rose from ~62% on day one to ~83% on day six, including an impressive jump from ~68% to 96% time in range overnight. Regarding the algorithm, Dr. Ly shared that the system is tailored to adapt “slowly, gradually” to the median over the previous six days, and we look forward to longer studies that will, perhaps, show even greater improvement relative to standard care (As a reminder, the ATTD data showed no statistically significant differences in any of the glycemic parameters between the 670G and 530G groups). As we noted at ATTD, the algorithm is still a work in progress, with future slated improvements to include flexible insulin:carb ratios, changing the correction threshold and target, and more aggressive tuning overall. This is now published in Diabetes Care.

Questions and Answers

Q: In the camp setting, how do you account for physical activity? Isn’t that a huge variable?

A: It is. Physical activity in camp is very different from activity at home. That’s why you have a control group. All the groups participate in all the activities. The biggest challenge is that the insulin delivery is supervised in camp. That really impacts the control and generally helps patients avoid hypoglycemia.

Q: How often are patients disconnected from their pumps? Overall, how long does it take for adaptation to occur? Could longer studies show a greater effect?

A: These are great results for six-day study but longer studies will show where adaptation works over time. It takes the median of previous six days, weighted toward more recent days. It’s tailored to be a slow, gradual adaptation over time.

Meet the Expert Sessions

Patient Perspectives on Closed-Loop Systems and Studies – Panel Discussion

Chris Aldred (The Grumpy Pumper, Southampton, UK) and Kelly Close (The diaTribe Foundation, San Francisco, CA)

A patient-centered panel discussion – led by Stanford’s great behavioral psychologist Dr. Korey Hood and featuring Mr. Chris Aldred (Southampton, UK) and our very own Ms. Kelly Close (The diaTribe Foundation, San Francisco, CA) – brought a healthy dose of both realism and optimism to the closed-loop conversation. Mr. Aldred, a very respected blogger in the UK (“The Grumpy Pumper”), lived up to his name by pointing out some of his concerns that the artificial pancreas may not live up to its billing right out of the gate. Indeed, he pointed out the risk that early on, there may well be unrealistically high expectations and advocated for more conservative optimism in light of the number of questions remaining to be answered. The words served as a strong reminder that not all patients will be automatically excited to use artificial pancreas technology when it arrives, particularly those happy with their current therapies or unwilling to wear something on the body. Kelly shared a few of her thoughts on the challenges of human factors design, acknowledging that connectivity issues have been meaningful in certain studies (this is a solvable problem, she stresses, although not one that was often discussed before last year) and that closed-loop systems in the research setting are of course in “debugging” mode – this is what research is for! With all that in mind, she spoke in glowing terms of the benefits already apparent in these first-gen systems – being brought to a soft landing from hyperglycemia (“virtually no hypoglycemia”) and the huge benefits of overnight control (“I wake up and I have a fair shot at my day … my husband says he always loves me, but he likes me more on the closed loop because of how I can start all these days …”). She posed the question at the end, “is it worth it?” and answered that it was worth it for the right patients – not the ones with particularly high expectations, for example. Ultimately, Mr. Aldred thought that the closed loop would be worth it too. The nuanced discussion underscored the importance of setting patient expectations moving forward: closed-loop technology will be a welcome prospect for many, but presentation in the early days will be important, as will patient selection.

Mr. Chris Aldred: I have had type 1 since I was 25. That was 21 years ago now when I was diagnosed. I’m really keen on technology of many sorts, things that’ll help me make better judgments and make life easier. I haven’t been a part of any trials yet. I have read about this stuff and I’m into thinking about what the right thinking is. To be honest, I’m a bit skeptical. I just won’t use any old thing that’s out there. Technology has got to help my diabetes live with me.

Ms. Kelly Close: I write for a newsletter called diaTribe, which is funded by the renowned Helmsley Charitable Trust – I’ve written many pieces in diaTribe about my experiences on the artificial pancreas. It’s so cool to have patients up here talking about our experience and we thank ADA and Korey for asking for our patient perspectives. I’ve been in the Bionic Pancreas and DiAs trials, and they have both been instrumental – but in the Bionic Pancreas trial, I had a nurse helping me 24/7, so that perceived “hassle factor” was lower than it would be in real life. The DiAs study was a little closer to what reality would look like. Of course I would want it for many patients who could benefit, especially those for whom acute short- or long-term complications could be avoided. I think getting patient access to this technology is a pretty big question and not one that is easily answered. If I can quickly ask, what does the audience think is the biggest benefit of the artificial pancreas?

A (CDE): Makes lives easier.

A (CDE): I think the answer is twofold. One of them is from a patient perspective: the relief of the burden of day-to-day diabetes management. The second is the glycemic control portion of it. Patients who have recurrent hypoglycemia or who have A1cs that are 9.0%, 10.0%, or 11.0% might be able to tighten down. There are about 20 other reasons I can think of, but those are the top two that come to mind.

Kelly: Thank you. So from my view, having been lucky enough to try several systems, I can say, the biggest benefit of automated insulin delivery is that it gives me a fair shot at my day. I don’t know how many of you have read Gretchen Rubin’s The Happiness Project, but she speaks about the importance of making your bed every day. Something so simple makes you feel better and have a better day. For me, the artificial pancreas gets me back to zero. That’s where everyone else starts their day. You wake up and you have a fair shot at your day. It makes you feel more normal. We are often – in spite of ourselves – somewhat grumpy. I called my family when I was on the Bionic Pancreas, and I spoke to my husband who said, “you sound more like my Kelly.” We have no idea how much we don’t get to be us because diabetes creates a lot of noise – we’re so busy just thinking about how to deal with our diabetes all the time. We joke about this all the time - my husband says that he always loves me and that he likes me much more on automated insulin delivery.

Q: If you were on a long-acting insulin overnight, how often do you find that you have glucose excursions?

Kelly: Personally, all the time. When you’re on CGM, you ultimately get to fix it and you’re much safer. Diabetes for sure is less costly and less dangerous with CGM, but patients don’t always hear all the alarms, and many still find it challenging to use, and HCPs can find it hard to use. But personally, all the time. CGM lets you fix it much of the time, but you don’t avoid it.

Ms. Aldred: I like my sleep way too much to wake up at night. Because I didn’t grow up with diabetes, I didn’t have a lot of those worries as a kid. Now I know that if I have had certain foods, I will go high. However, I don’t test a lot at night.

Kelly: One of the hardest things about diabetes is that you can be really high and not even realize it. Then you take a rage bolus (I learned that term from the great Kerri Sparling Morrone of Six Until Me, who is sitting right here in the front row) and you are at 40 mg/dl and it’s a huge spiral. Automated insulin delivery has eliminated that. Closing the loop is not going to change whether or not you go high if you eat too much pasta or pizza – of course you will go high! It’s not going to keep your glucose in range 24/7. However, what it does is it creates an amazing soft landing from hyperglycemia, where you get to avoid hypoglycemia. Period. It’s interesting because in my first trial I had a nurse with me 24/7 and I raved about it. In my second trial on the DiAs, it was “in the wild” (a research house). And the one I’m on right now is even tougher – it’s all me taking care of it (and John!). It’s tougher without a nurse right there for sure. So there are some bumps. However, if you remember the days of [Dexcom’s] STS, it was really challenging and few of us felt it was an advanced product. But look how far we’ve gotten. The STS enabled Dexcom’s G4! As research subjects, we have to do trials like the STS. Frankly, current closed-loop systems are pretty hard to use right at this moment; there are connectivity problems and all sorts of other things that nurses took care of in the first bionic pancreas trial I was part of. It’s hard, because researchers are trying to get these systems to work and ensure that they are safe enough and as patients we need to focus on helping de-bug them.

To the question, are they worth it? OF COURSE! Of course they are worth it. For anyone who’s been in a trial, there is no question.

That said, as a community, our expectations are much higher than they used to be. We should be in debugging mode right now. We want patients to be in trials who want to work on making the systems better. I think that’s the stage that we’re in.

Dr. Korey Hood (Stanford University, Palo Alto, CA): What are people skeptical about? There are people who are skeptical and are a little bit worried.

Mr. Aldred: Just to touch on a previous question, I want to address the notion of the burden of diabetes. I think the burden is quite personal. Here’s a hypothetical: If you had a fully working artificial pancreas and users do not have to do anything and its free, but it is the size of a backpack, how many people would actually wear it? Not many, because the burden of the backpack outweighs the burden of diabetes. We can’t treat everyone the same way. Is it worth it? I think so. But my skepticism sits in knowing all the variables that are needed to control my diabetes. How can a system understand all those difference preemptively? How is it going to distinguish an adrenaline high from a carbohydrate high? The former doesn’t require as much insulin. I’m not saying it won’t get to the point where it can do this successfully, but how do we get to that actual end game?

One of the other concerns I have is that we’re going to have first, second, and third gen systems. In the UK, we’re lucky enough to get our pumps funded. However, we don’t get to change our pumps for the next four years. So, if I have one artificial pancreas before the next gen comes out, then I’m stuck with that version for the next four years. This needs to be iterative like our phones, where we can upgrade them. There are lots of challenges here.

Q: Do you wear a CGM? Is it paid for?

Mr. Aldred: Yes. I can’t wear it for more than two days without getting a rash, but I wear it. And if you’re self-funded, you’re going to wear it until it dies and catches fire.

Q: How does a sensor affect your life day-to-day?

Mr. Aldred: I really like it. I wouldn’t wear it if it didn’t outweigh the hassle. I get more benefit than I get from not wearing it and I wear it as long as possible. I tend to know when it’s coming to the end of its useful life. Until it is unreliable, I will leave it on. Whereas with a cannula, I’ll rip it out and change it immediately even if it’s a bit uncomfortable.

Q: Kelly, what systems have you used in these trials and what day-to-day issues have you faced?

Kelly: I think automated insulin delivery is going to be the killer app for CGM. I think the data that can come from CGM and how we can be in zone a lot more of the time can be really powerful. I’ve been in the DiAs study and the Bionic Pancreas. On the latter, I know there has been some controversy about glucagon. However, that was my first exposure to glucagon and it was incredible. It seems to enable a faster recovery from hyperglycemia as it might be a more aggressive algorithm.

People want to know whether I would take that risk: I will take that on. I think making sure your family feels good about that is really important. DiAs does not have glucagon and it had a slower reaction. At night, DiAs is an absolute dream. And during the day, issues like connectivity do come up. However, when I’m out of range, it does bring me down softly. You’re not worried about hypoglycemia and that’s hugely powerful for someone who has experienced a lot of severe hypoglycemia.

Q: Can you talk about the decision to be on MDI vs. a pump?

Mr. Aldred: That’s a very difficult question. I got offered a pump back in the day and I said I don’t want one. But after 15 years of starting days out of range, I gave in.

Kelly: I agree that the bar for using this equipment is high. I only wore a pump when my boss – back in finance – made me do it! We needed to know if MiniMed was a good investment, so she made me wear it for a week. I’ve been on a pump ever since. Would I have gone on that on my own? No way! The perceived hassle in my head was completely different from knowing what it was like to wear it! Also, I’ve been on a CGM and I know the value in it, but again, it is a big hurdle for many, many patients to get over, to put this thing on. We know diabetes is associated with depression, but it’s good to remember that in addition to this, there is also a lot of distress. Patients need to talk to each other about this technology and talk about changes CGM is making in patient’s lives. It’s important to recognize that pumps aren’t perfect. Maybe it will take until the third generation to convince a lot of the hardcore MDI patients. However, in the long run, these devices are going to make life less expensive for payers and less dangerous for patients, when the studies are done that gives data to show that.

Dr. Hood: Some people have asked me whether there is a glycemic benefit to pumps. For some MDI users, they don’t want to move to pumps it’s because they are achieving the kind of control they want without one. They don’t see the value added of going to some other system. When there is a closed-loop system, I think people will see more value added. Hopefully, we can leverage that enthusiasm around it.

Mr. Aldred: I think one of the other issues for me is that the decision was talking about as a “move” from MDI to pump. It was talking about like there was no going back. There will be some people out there, who do not get to “try” a pump and that dissuades them. We don’t give people the change to test drive pumps. I wouldn’t buy a car without test-driving it.

Kelly: I was made to go on a pump, pretty much. I think as patients, we are not really encouraged to, as, say, a civic duty, do the right thing and to keep our glucoses in zone. We all have a responsibility to society, to ourselves, and to our families. I am definitely not blaming patients – not ever! – but I think a little bit more of that responsibility will help.

Q: Kelly, how much SMBG are you doing on closed loop?

Kelly: You do have to do SMBG to calibrate the system, but it is pretty accurate. When I was on the Bionic Pancreas, I think it was 12 times a day. It is a lot less on DiAs, maybe once or twice a day.

Q: Are you not more worried because a computer is doing what it wants to do?

Ms. Close: Oh no! No! I have this pump that has a basal rate and an insulin sensitivity factor and an insulin to carb ratio and all this math. For example, how many of us in the room with diabetes believe our basal rates are perfect? [No one in the room raises their hand.] Right! This is math. Complicated math. This should be done by a computer. This shouldn’t be done by us. We have enough trouble counting carbs. I’m not more worried on these systems. I feel safer. I’m much more worried about access to having the system, and we’ll all have to work hard to address that.

Dr. Hood: At Stanford, we are running an artificial pancreas trial and we ask that question to our patients. By the end, most say that feel more confident in the system across the board. We’re still under 100 patients but that’s still a sizeable population.

Audience comment: I recently just completed the Bionic Pancreas trial. What I found was that does cause some negatives. I was rocking a fanny-pack with a lot of gear, but when it gets commercialized, all that will be gone. You can also physically hear the insulin and glucagon being injected which is nice. It’s really nice to see your glucose come up without you doing anything. I did have some nausea on day one, but that was minimal. It’s still early, but it was a pretty cool concept that is only going to get better and better.

Q: I’ve been a type 1 since 1989. I’m also a tech geek. However, I find that when I’m a switching patient from MDI to pump, diabetes management goes from something that’s been on a shelf and to something that is a part of their lives. It connects them more with their diabetes. What is your experience with that?

Kelly: What was life changing on closed-loop was not having to worry about hypoglycemia. I didn’t think I normally did that. But it is totally different, being on closed loop. I did have to worry about connectively, and we do need to reframe closed-loop conversations to understand that this is going to be a lot of work. I do think we need more people funding this so research can continue to move quickly.

Dr. Hood: The data on CGM in general suggests that there are mixed results. Some people find that it raises their worries and burden, but many think that there’s a great benefit to it.

Q: Have you found that patients gain weight on MDI?

Mr. Aldred: Yes. I’ve put on weight. I probably eat more now that I’m on a pump because I eat when I want to because I feel like I can. MDI was restrictive. In that sense, maybe the technology has helped me put more weight on. But if I did whatever I wanted to and didn’t have diabetes in the first place, then I probably would have put on a lot more weight already.

Q: Can you speculate on certain personality traits and whether they lend themselves to closed-loop?

Kelly: The more worried a patient feels about severe hypoglycemia, that would be a driver to the closed-loop. I think some people feel psyched to be a pioneer and to be a part of change. What I don’t think is a good profile is people who think closed-loop development should move as fast as consumer electronics. That’s just not going to happen. People who should go on the closed loop are people who will be given and who will have reasonable expectations early on and people who want to help move technology forward – also, people with significant needs (e.g., lots of severe hypoglycemia) that the closed loop can address.

Pumps and Continuous Glucose Monitoring in Youth – From Research to Clinical Use

Bruce Buckingham, MD (Stanford University, Stanford, CA)

After highlighting some of the findings from the T1D Exchange dataset, Dr. Bruce Buckingham provided a whirlwind tour of diabetes technology use in children and adolescents. As we have heard more broadly, Dr. Buckingham emphasized that infusion sets are currently the weak point in insulin delivery. He also discussed data from several studies investigating whether insulin brand or needle type affects infusion set health, concluding that there was no difference. On the closed-loop front, Dr. Buckingham explained that an artificial pancreas system will need to be able to detect infusion set failures, and that ideally, infusion sets would need to last for (at least) a week for use in such systems. 

• Dr. Buckingham found no meaningful difference between Novolog and Humalog, or between steel and Teflon needles in a handful of exploratory studies that explore the duration of use of infusion sets. Anecdotally, he commented that there is a strong sentiment among patients that Novolog makes infusion sets last longer and causes fewer occlusions than Humalog. However, the two rapid-acting insulins fared equally well in infusion sets in a small study of 20 subjects; there was no meaningful difference in infusion set survival at one week when either insulin was used and insulin precipitation occurred in every catheter to some degree. Similarly, Dr. Buckingham and his colleagues have also found no difference in infusion set survival at one week in their own studies. Notably, a group of their patients were able to use each infusion set successfully for 6-7 days, with no deterioration in glucose values.

Questions and Answers

Q: Could you tell us about the Bionic Pancreas? How long will it be until it is available?

A: The way we were using it was you had this brick, which had an iPhone and Dexcom and two Tandem pumps. It was a little clunky. Most of the homegrown systems are fairly clunky. People are moving toward making one integrated system. It will receive a signal from the Dexcom. There will be separate cartridges for insulin and glucagon, and they’ll be delivered at one infusion site. It will be much more integrated. The key is to get glucagon approved for pump use. No one has delivered glucagon long-term continuously. I think it’d be great if we could be fairly aggressive with insulin and have glucagon on the back end … the whole key in closed-loop control is adaptability. Everyone is different, so having a system that adapts is going to be key.

Q: We have had poor luck with CGM and getting patients to continue using it. Do you have any thoughts on how we can improve those numbers?

A: I think it’s always a matter of training and expectations and what they can expect out of this system. I think if you have a lot of false alarms and it becomes annoying, it tends to detract from use. Start out where you don’t get a lot of unnecessary alarms, and tell them how to use it so it’s functional for them. The Dexcom does very well in the hypoglycemia range, which has been a problem in the past. I think it’s a real improvement. One case I was going to show was an adolescent using a Dexcom, who was doing two blood tests a day – just enough to calibrate it. The patient based all boluses on the Dexcom. It’s not indicated, but that tends to make people use it more. The Abbott Libre is factory calibrated. In a few years, I think you’re going to see Dexcom and Medtronic have factory-calibrated products. If you have an accurate sensor, you can really use it to a large extent.

Oral Presentations: Moving Towards a Closed-Loop System

First New Year’s Night on Closed-Loop Control (CLC) at Home: Case Reports from a Multicenter International Trial of Long-Term 24/7 CLC (223-OR)

Stacey Anderson, MD (University of Virginia Health System, Charlottesville, Virginia)

Dr. Stacey Anderson presented data from an impressive six-site, multinational, at-home trial testing two weeks of overnight closed-loop and two weeks of 24/7 closed-loop using UVA’s DiAs (Dexcom G4, Roche pump, Android phone). Sensor-augmented pump (SAP) was the comparator, and she showed data from 24/30 participants. Her conclusion summed up the data in seven words – “SAFE during the day. HIGHLY EFFECTIVE overnight.” The primary endpoint of hypoglycemia (time spent <70 mg/dl) was halved overall (4% during SAP vs. ~2% during closed loop; p<0.05); the improvement was very marked at night, where time spent in hypoglycemia improved by more than two-thirds (3% to ~0-1%; p<0.05). Overall 24-hour time-in-range (70-180 mg/dl) improved a bit (66% in open-loop vs. ~73% in the two closed-loop phases of the study), as did overnight time-in-range (62% to ~71-74%). Overall, we would caution that this study was not randomized (the intervention was applied sequentially, pump only, SAP, overnight-only, 24/7 closed-loop), so it’s tough to draw conclusions about how well it works vs. a parallel control group; UVA’s ongoing Project Nightlight study will allow for that comparison. We’d also note that patients were doing fairly well at baseline (mean A1c 7.3%, range: 5.9-8.6%), meaning the ability to improve on open-loop control in terms of time-in-range was probably limited. The hypoglycemia data is certainly impressive.

• You can read Adam and Kelly’s personal experience in this study here – they found that the overnight algorithm was excellent (it treated to a target of 120 mg/dl by 7 AM), and the daytime algorithm erred on the side of conservative in this feasibility study. They both loved the overnight system and were glad to see the daytime version being tested to identify and work out bugs.
• Notably, the study has obtained a five-month extension, which will give longer-term, real-world data on the 10 subjects currently participating. The trial is now using the Share Receiver.

Outpatient Glycemic Control with a Bionic Pancreas in Preadolescents with Type 1 Diabetes (222-OR)

Steven Russell, MD, PhD (MGH, Boston, MA)

Dr. Steven Russell provided a deeper dive on data from the Summer Camp 2014 in pre-adolescents (first shown in October at CMHC 2014) – the findings were very consistent with other camp and adult studies. Full 24-hour data suggested an improvement in mean glucose (168 to 137 mg/dl; p<0.001), a significant reduction in hypoglycemia (time <60 mg/dl: 2.8% to 1.2%), and a strong improvement in time-in-range (70-180: 58% to 81%). Dr. Russell shared – for the first time we can recall – stratified overnight results (11 PM – 7 AM) from the study that showed very impressive nocturnal efficacy: A mean glucose of 122 vs. 168 mg/dl, a dramatic reduction in hypoglycemia (0.6% vs. 2.8%), and a striking 92% time-in-range (vs. 59% in open-loop). Mean insulin usage was identical in the two arms (0.68 u/kg/day) and mean glucagon usage was 0.29 mg/day. On the pivotal study front, Dr. Russell shared a new plan to pursue an indication for the Bionic Pancreas in patients eight years and older.

• His slide deck also featured a prominent picture of a six-year-old (!) on the Bionic Pancreas, reminding us of how ambitious the camp study really was – the bionic pancreas brick was the size of her chest!
• On the pivotal study front, Dr. Russell shared that the group’s new plan is to pursue an indication for the Bionic Pancreas in patients eight years and older (Dr. Damiano had previously guided for 10 and older). This is great to hear considering that the technology will likely be used off-label in populations where it is not indicated. It’s impressive to see the team going this young – certainly, it’s a population whose families will benefit tremendously from this technology.

Artificial Pancreas Improves Glycemic Control in a Multinight Outpatient/Home Study of Patients with T1D (224-OR)

Sue Brown, MD (UVA, Charlottesville, VA)

Dr. Sue Brown summarized new data from a multicenter overnight study of UVA’s DiAs closed-loop system. The randomized, crossover trial included 44 patients (baseline A1c 7.4%) and compared five nights of overnight (11 PM – 7 AM) closed-loop control in a hotel supplemented by daytime sensor-augmented pump (SAP) therapy to five days of SAP therapy. As hypothesized, analysis of the study’s primary arm demonstrated that overnight closed-loop control did manifest in improved 24-hour glycemic benefits - time-in-target increased from 71.5% on SAP to 78.3% on closed-loop therapy (p = 0.003). This finding was accompanied by a reduction in hypoglycemia as time <70 mg/dl dropped from 4.3% on SAP to 2.5% on closed-loop (p=0.004). Dr. Brown also summarized the results of an extension study in ten patients, who experienced five subsequent nights of overnight closed-loop control at home – Dr. Brown stressed that this was the first time DiAs was taken home in the US! In the home setting, overnight closed-loop marginally increased time-in-target from 71% to 75% and more than halved hypoglycemia from 4.9% to 2.2% (p < 0.05).

• As expected, both primary (n=44) and secondary (n=10) study sub-analysis showed that the 24-hour benefits were driven by the more impressive performance of the closed-loop system overnight.
  
  • Primary: In the multicenter study, mean blood glucose dropped from 155 mg/dl to 137 mg/dl (p<0.001), time-in-target increased from 68% on SAP to 86% (p<0.001), and time < 70 mg/dl dropped from 3.2% on SAP to 0.9% on closed-loop (p<0.001). See Table 1 below.

Table 1: SAP vs. Closed-loop Therapy under supervision

	Overnight			24 Hours
	SAP Therapy	Closed-loop therapy	P-value	SAP Therapy	Closed-loop therapy	P-value
Mean blood glucose	155 mg/dl	137 mg/dl	p<0.001	--	--
Time-in-range (70-180 mg/dl)	68%	86%	p<0.001	72%	78%	p=0.003
Time < 70 mg/dl	3.2%	0.9%	p<0.001	4.3%	2.5%	p=0.003

• Secondary: In the home study, mean blood glucose dropped from 161 mg/dl to 151 mg/dl (ns), time-in-target increased from 63% on SAP to 85% (p=0.07), and time < 70 mg/dl dropped from 3.7% on SAP to 0.6% on closed-loop (p=0.03). See Table 2 below.

Table 2: Overnight SAP vs. Closed-loop Therapy at home

	Overnight			24 Hours
	SAP Therapy	Closed-loop therapy	P-value	SAP Therapy	Closed-loop therapy	P-value
Mean blood glucose	161 mg/dl	151 mg/dl	ns	--	--
Time-in-range (70-180 mg/dl)	63%	85%	P=0.07	71%	75%	ns
Time < 70 mg/dl	3.7%	0.6%	P=0.03	4.9%	2.2%	p<0.05

• Ultimately, the findings are impressive, though not altogether surprising given where the field is now. Overnight closed-loop has been outstanding for year – the big question is whether any first-gen hybrid closed loop will actually be night-only – from what we saw at the JDRF night, it appears unlikely.

Day and Night Closed-Loop Insulin Delivery in Young People with Type 1 Diabetes: A Free-Living, Randomized Clinical Trial (221-OR)

Martin Tauschmann (University of Cambridge, UK)

Dr. Martin Tauschmann presented new data from the Cambridge group’s day and night study of closed-loop insulin delivery in adolescents. The randomized, crossover trial included 12 adolescents (10-18 years of age) and compared seven-days of closed-loop control to seven-days on CGM alone with a goal of assessing safety and efficacy. The findings were very comparable to prior results seen in Cambridge’s adults studies – the intention-to-treat analysis showed that mean time in target (70-144 mg/dl) increased from 53% to 71% and time <70 mg/dl decreased from 2.9% to 1.7%. Much of this improvement was attributed to the nighttime period – said Dr. Tauschmann, “The algorithm was very impressive overnight.” Indeed, mean blood glucose overnight on closed-loop control was 140 mg/dl vs. 175 mg/dl on CGM alone. Differences between the groups during the day were non-significant (169 mg/dl [closed-loop] vs. 185 mg/dl [SAP]) and Dr. Tauschmann was quick to acknowledge, “there is room for improvement” – we point out that much of the reason why the day doesn’t show significant changes is likely due to food, exercise, stress, etc – things the AP can’t meaningfully change due to the slow speed of insulin.  

• Dr. Tauschmann stressed that there were no differences between the groups in terms of total daily dose of insulin - 57.3 units/day (closed-loo) vs. 56.6 units per day (SAP). He noted, however, that there was a shift in distribution with the closed-loop system delivering more insulin as basal relative to SAP therapy.
• Dr. Tauschmann noted that the study was not powered to study hypoglycemia – that said, neither group experienced significant time < 70 mg/dl: 3.3% (SAP) vs. 3.7% (closed-loop). It was a big surprising to see that hypoglycemia was that high in the closed-loop arm, though we imagine most occurred during the day.
• We were impressed with the durability of the group’s closed-loop system, which was active for 91% for the study period. Dr. Tauschmann reported one closed-loop interruption per patient per day that was typically due to connectivity issues. He did acknowledge, however, that patients were asked not to use closed-loop control during strenuous exercise – the group’s goal long-term is to address this though, “[they] are not there yet.” His commentary certainly lends perspective to the impressive results the Bionic Pancreas has seen in the unpredictable camp setting.

Glycemic Control in a Large Cohort of Patients with Type 1 Diabetes (T1DM) Treated with Continuous Subcutaneous Insulin Infusion (CSII) (226-OR)

Lalantha Leelarathna, MD (Central Manchester University Hospital, UK)

Dr. Lalantha Leelarathna presented the results of a recent retrospective study that sought to: (i) evaluate the effectiveness of glycemic control in patients with type 1 diabetes in pumps; (ii) investigate the demographic factors associated with positive A1c changes; and (iii) compare the penetration of different pump models. The study used data from 422 patients at a single center in the UK, who primarily used pumps from Medtronic, Insulet, Animas, and Roche. Results indicated, unsurprisingly, that pump use was correlated with an improvement in glycemic control – the proportion of patients achieving an A1c < 7.5% increased from 19% to 32% after initiating pump therapy. There was also a significant correlation between pre-pump A1c and glycemic improvements. What was unexpected, in Dr. Leelarathna’s view, was the number of patients in poor control – notably, more than a third of patients had an A1c >8.5% and 11% had an A1c >10%. In our eyes, the findings highlight the challenges of managing type 1 diabetes in the real world and provide valuable perspective on the non-representative patients we see in clinical trials.

• Dr. Leelarathna highlighted a handful of additional findings that were thought were notable. We summarize those below:
  
  • Results indicated that there were no differences in A1c change between different pump users (p=0.22) or between those using durable pumps vs. the OmniPod (p=0.54). However, a higher percentage of patients using traditional pumps saw A1c improvements > 0.5% relative to those on OmniPod.
  • The breakdown of pump use by brand was: Medtronic: 42%; Insulet: 30%; Animas: 14%; Roche: 13%. These numbers are skewed significantly away from Medtronic relative to our worldwide full-year 2014 estimates – Medtronic (68%), Roche (10%), Animas (10%), Insulet (10%), and Tandem (2%). We assume whether this is a regional or center-specific preference – Dr. Leelarathna did not offer any commentary on this point.
  • 83 patients changed pumps during the study period with the majority switching from Medtronic to OmniPod. Dr. Leelarathna attributed this trend to the launch of second-gen OmniPod in late 2011 that, in his view, brought a big convenience factor to a lot of patients.

Questions and Answers

Q: When patients were switching from Medtronic to other models, they must have been going from an integrated CGM to another device. Did you account for this at all?

A: Unfortunately, the NHS does not routinely fund CGM. So there was extremely limited CGM use in our population. I don’t think that explains much difference in these patients.

Q: Do you have demographic data on what kinds of patients change to a patch pump from a traditional pump?

A: I can’t answer that at the moment. We are collecting more data though. We’re not sure whether there was a fundamental difference in these groups, and either way, it’s difficult to be sure in a retrospective study.

Q: Did you collect any data on hypoglycemia?

A: Because this was a retrospective study, we were not able to collect hypoglycemia data.

MD-Logic Closed Loop Control (ClC) For Automatic Type 1 Diabetes Meal Management (219-OR)

Revital Nimri (Schneider Children's Medical Center, Petah Tikvah, Israel)

Dr. Revital Nimri summarized new data from two DREAM consortium studies that aimed to optimize meal announcements in the group’s MD Logic closed-loop system (licensed to Medtronic). The first study (n=9) evaluated the safety and efficacy of four different insulin doses (0%, 70%, 100% and 120% of the meal bolus calculated according to the patient’s carbohydrate ratio) on glycemic control, while the second (n=9) determined if increasing the meal bolus by adding a dose equivalent to two hours of basal insulin can improve glycemic control further (known to patients as a “super bolus”). The unsurprising findings of the first study revealed that the optimal dose for an announced meal bolus is between 70% and 100% of the calculated pre-meal bolus – these brought patients’ mean blood glucose to 138 mg/dl and 131 mg/dl, respectively, while minimizing hypoglycemia. Time-in-range was maximized for the normal bolus (> 80% time-in-range) followed closely by the 120% bolus (though this was accompanied by hypoglycemia) and the 70% bolus. Moving to the second study, Dr. Nimri noted that the combined two-hour basal plus bolus dose significantly improved the postprandial glucose profile at both meals tested: breakfast (251 mg/dl vs. 199 mg/dl, p=0.03) and lunch (155 vs. 140, p=0.05). In conclusion, she suggested that the MD-logic bolus calculator should incorporate the two-hour basal profile into its algorithm, though acknowledged that the ideal bolus dose will be more variable depending on multiple day-to-day variables: meal-time, exercise, health, etc.

• Study Design: The two studies included 13 patients total (baseline A1c: 8.0%) with some overlap (nine in each trial). The first study included four sessions of 12 hours for each patient each with two meals, while the second study included two similarly organized sessions. The glycemic index of the breakfast meal was high (40) and that of lunch was low (14).

Questions and Answers

Dr. Roman Hovorka: How much insulin is 80% plus the two-hour basal dose? Does it come to 90%?

A: I don’t have a definite answer.

Dr. Bruce Buckingham: The one thing I’d point out is that at the lunch test in your studies you still have some insulin left over from breakfast bolus. That’s a confounding factor to take into account.

Q: Was carbohydrate done in a supervised manner? I just want to make clear that carbohydrate counting in real life is often less accurate than in studies.

A: We found that the controller can overcome 20% of errors for under- and over-bolusing, so I don’t think the precise accuracy is as important.

Oral Presentations: ADA Presidents Oral Session

Outpatient Overnight Glucose Control with Dual- and Single-Hormone Artificial Pancreas Systems in Type 1 Diabetes: Randomized Controlled Trials (383-OR)

Ahmad Haidar, MSc (Ecole Polytechnique de Montreal, Quebec, Canada)

Mr. Ahmad Haidar presented never-before-seen pooled results of two randomized crossover trials comparing overnight dual-hormone closed-loop therapy to single-hormone closed-loop therapy to conventional pump therapy, showing superior avoidance of hypoglycemia with the dual-hormone approach. The first study – published in The Lancet – compared three nights (11 PM – 7 AM) of each intervention in 33 pediatric patients (ages 9-17) in the camp setting. The second crossover study compared two nights of each intervention in 21 adults and seven pediatric patients at home. Dual-hormonal overnight therapy resulted in significantly less time < 72 mg/dl compared to both alternative treatment paradigms (1.0% vs. 3.1% [single-hormonal] vs. 5.1% [pump]). Perhaps more notably, the dual-hormonal approach resulted in only one hypoglycemic episode requiring intervention in 150 treatment nights (!) relative to 11 events and 29 events in the insulin-only closed-loop and pump therapy arms, respectively. Mr. Haidar showed traces demonstrating that the improvement in dual-hormonal glycemic control came largely during the first-half of the night when glucagon was heavily utilized to counterbalance the insulin stacking of the daytime – indeed, though the single-hormonal systems essentially suspended insulin delivery early on, it was often not sufficient to prevent hypoglycemia. In terms of glycemic control, Mr. Haidar noted that both closed-loop systems achieved mean blood glucoses of 122 mg/dl vs. 140 mg/dl for the conventional pump therapy arm. The findings are notable as the first “pseudo” head-to-head trials comparing insulin-only to insulin+glucagon closed-loop. Certainly, the overnight period is a testing ground, and the really interesting work will occur during the day. We look forward to seeing even more comparative data once the Bionic Pancreas team conducts its two glycemic set point studies (insulin-only and bi-hormonal) starting this August.

• Mr. Haidar stressed that there was not excessive glucagon usage (0.03 mg per night) or adverse symptoms reported in the dual-hormonal arm.
• Notably, there was not perfect parity in the trial designs or in the equipment used – Dexcom sensor + Roche pumps in the pediatric study vs. Medtronic sensor + pumps in the second study – though we still find the results very directionally interesting. Note that these are solely academic systems as opposed to technologies in commercial development.

Questions and Answers

Q: You noted that 35% of nights in the dual hormonal arm did not require glucagon. Do you have predictors for those nights?

A: No. But we wonder if the duration of diabetes was associated with that.

Q: What do we know about the nighttime snack schedule?

A: During the camp, the campers were snacking about one hour before sleeping. Most of the hypoglycemic we saw during the early night hours was related to the snacking bolus. In the home study, we provided a 120 g meal slightly before bed with a bolus on one night. But we did not have the daily snack in home study – that was only in the camp study.

Q: Do you think current insulin is too slow? It seems like patients have to have glucagon in order to compensate for the insulin.

A: When we looked at individual hypoglycemic events, we found that we get decreased insulin delivery before hypoglycemic episodes. We weren’t able to prevent hypoglycemia from previous boluses. However, when you look at the second half of the night, the stacking effect of insulin was not there anymore.

Q: The highest risk for hypoglycemic events was during the first half of the night. Did you try to help patients optimize their insulin delivery during the daytime? 

A: For the camp study, we tried to optimize the basal overnight rates for patients. One limitation is that both studies were using different settings. There was not parity between the camp study and home study.

Oral Presentations: Cardiovascular and Other Clinical Effects of Hypoglycemia

Restoration of Hypoglycemia Awareness and Prevention of Recurrent Severe Hypoglycemia in Long-Standing Type 1 Diabetes: Sustained Benefit and Improved Overall Glycemic Control over 2 Years following Recruitment into the HypoCOMPaSS Multicenter RCT (352-OR)

Stuart Little, MBBS (Institute of Cellular Medicine, Newcastle University, Newcastle, UK)

Dr. Stuart Little presented impressive two-year follow-up results from the HypoCOMPaSS trial, a 24-week 2x2 factorial randomized control trial (RCT) conducted in 96 adults with type 1 diabetes and impaired awareness of hypoglycemia (IAH). The objective of the study was to compare analog MDI vs. insulin pump therapy, as well as to compare SMBG vs. real-time continuous glucose monitoring in patients with impaired hypoglycemia awareness. At the end of the 24-week period, patients returned to routine care – potentially switching back from MDI to CSII or vice versa – though access to CGM was provided to all initially randomized to this monitoring regimen. Two-year follow-up results demonstrated that – provided equal education and attention – hypoglycemia awareness in all patients has continued to improve following RCT completion. Impaired awareness has fallen from 5.1 at baseline to 4.1 at six months and, further, to 3.7 at 24 months (p<0.01 vs. baseline). Significant improvements have also been seen in the severe hypoglycemia rate – 8.9 episodes per patient-year during the 12 months preceding the study to 0.4 over the 24-month study period (n=69; p<0.01). Even more impressively, only 36% of patients have suffered a severe hypoglycemic event over the study period relative to 92% in the preceding 12 months. Despite the reductions in hypoglycemia, A1c still improved at 24 months (7.7% vs. 8.2% baseline; p<0.01). Overall, the findings speak to the value of education and better therapies – what is unclear is how the CGM compared to pump therapy in terms of driving the improvement. Dr. Little concluded that IAH can indeed be improved and that recurrent severe hypoglycemia can be prevented through strategies targeted at avoiding hypoglycemia while improving overall glycemic control. As a reminder, we saw one-year results from the HypoCOMPaSS trial at ADA 2013.

Questions and Answers

Q: Your randomized control trial was run for roughly half a year. However, the results at the end of the half-year persisted for up to two years. What do you think was the mechanism that led to this? Were the pumps the big difference-maker?

A: Yes. Patients avoided recurrent severe hypoglycemia because of the pumps. I wouldn’t say severe hypoglycemia was eliminated, but I do think that the education on hypoglycemia avoidance helped. It’s honestly impossible to tell the impact of pump therapy.

Q: Do you think the results were a consequence of the technology or a behavioral intervention?

A: A mixture of both. I think it was the holistic care.

Q: Some people have the technology and fail nonetheless. We tend to think the technologies are important in treating hypoglycemia. We still have to figure out which combinations of technology + education works best.

A: I agree. In our study, everyone had standardized education at beginning that looked at key aspects of hypoglycemia avoidance. We looked at individuals, looked at times of individual risk,  helped identify times of hypoglycemia risk. We helped patients become more aware of hypoglycemia, especially during the nighttime. All of a sudden, patients had insight into what was happening overnight.

Q: Did you provide an education on how to use bolus calculators? That could have also made a difference.

A: Yes. Good point. I think that all the tools that were used had an impact. 

Oral Presentations: Pediatric Diabetes – Diagnostic Challenges and Effects on the Brain

Effects of Sensor Augmented Pump use on HbA1c and C-Peptide in the First Year of Type 1 Diabetes (383-OR)

Taylor Triolo, MD (Barbara Davis Center for Childhood Diabetes, Aurora, CO)

Dr. Taylor Triolo presented a post-hoc analysis of one-year data from a randomized, controlled study of C-peptide preservation in young people recently diagnosed with type 1 diabetes through the use of hybrid-closed-loop therapy, first presented at ATTD 2013. The trial included 48 patients who received three days of in-hospital hybrid closed-loop therapy within one week of diagnosis, followed by sensor- augmented pump therapy thereafter; the control group (n=20) received standard care with a different set of physicians. As a reminder, no statistical significant differences were seen in A1c or C-peptide after one year, which Dr. Triolo attributed partially to the fact that only 33% of patients used SAP > 5 days/week. Thus, in this updated analysis, the group sought to stratify the data by median CGM use to assess whether more frequent use was associated with lower HbA1c and preservation of C-peptide production. This analysis did reveal statistically significant differences in A1c between the low and high CGM use groups (6.8% vs. 7.9%, p=0.003) along with a marginal improvement in C-peptide (0.50 nmol/L vs. 0.35 nmol/L, p=0.25). However, no statistically significant differences were seen in any of the other metrics studied: age, gender, ethnicity, or A1c at the time of onset. We do wonder whether two-year results might show differences in C-peptide (as occurred in the European ONSET trial of sensor- augmented pump use from diagnosis onward), though the Stanford group has refuted this suggestion in the past.

• We would caution that the correlation between A1c and CGM may be spurious. For example, high CGM users could simply represent a more motivated population that would attain lower A1cs regardless of therapy. Indeed, we think further investigation is needed to fully flesh out such lurking variables.  

Posters

“It Is Definitely a Game Changer”: Closed Loop Technology in the Home Experienced by Adults with Type 1 Diabetes (66-LB)

C Hendrieckx, J Speight, L Poole, A Sharifi, M Loh, J Horsburgh, S Trawley, A Jenkins, K Kumareswaran, R Macisaac, G Ward, P Colman, L Bach, A Kyoong, N Kurtz, B Grosman, A Roy, D O’Neal

Patients (n=10) in a small, randomized crossover study consisting of four nights of home closed-loop therapy vs. normal care (sensor-augmented pump therapy) reported positive qualitative reviews of their time on closed-loop. [The poster did not specify, but we assume this was not the MiniMed 670G integrated system, given the study timing from April 2014-March 2015. Presumably it was an Android phone or laptop-based system.] Specific benefits reported in this poster included: reduced decision-making and human errors, fewer alarms, less glucose variability, and less severe consequences of hypo- and hyperglycemia compared to SAP. When asked about the closed-loop, the most common patient observation was the stable overnight glucose levels that patients described as “unbelievable.” While most reported technical glitches during one or more nights, few safety concerns were raised. Encouragingly, confidence in closed-loop accuracy increased when subjects observed how the system automatically stopped insulin delivery when glucose trended down – an adjustment which patients described as “brilliant” and “impressive” – though patients did note that this might be different for people who are less “tech savvy.” Despite the small sample size and short duration, these positive reviews of closed-loop have us looking forward to hearing future qualitative assessments of patients’ psychosocial experience of the 670G from larger groups of patients. As a reminder, management expects a launch of the MiniMed 670G by April 2017 in the US and April 2018 in the EU.

• This qualitative assessment of patient perspectives was conducted following four consecutive nights of closed-loop control at home. Structured interviews focused on: patients overall experience with closed-loop therapy; the effect of closed-loop control on glucose levels, hypoglycemic, and hyperglycemia; patients understanding of the algorithm; and potential concerns about using closed-loop. Two 10-point scales were used to assess comfort with closed-loop control and confidence in the system’s accuracy.

DiabetesMine D-Data Exchange

Diabetes & The Internet of Things...Sensors, Wearables, Shareable Data: Where Is It All Going?

Sonny Vu (Founder, Misfit, San Francisco, CA)

Mr. Sonny Vu gave the opening keynote, sharing fantastic insights from AgaMatrix and Misfit on product design and digital health. We learned loads from this talk, especially the “Turnaround test” (is something so useful and compelling that you would run home to get it if you forgot it?”); wearables 1.0 vs. 2.0 (we are moving into the latter); what the ideal product design looks like (“combines Harvard Medical School and Playskool” – clinically useful and an amazing, delightful user interface); and where wearables/digital health need to go (passive data acquisition; insights and timely feedback; clinical effectiveness; reimbursable). On the latter, Mr. Vu concluded that feedback is the hardest, but also the area with the most potential to change behavior. [We especially hope to see much better feedback in diabetes, since current devices don’t share encouragement or positivity – e.g., 258 mg/dl vs. “Great job! You’re on a 10-day streak of testing before lunch!”]

• Mr. Vu highlighted the key decades that have revolutionized computing, noting that we’re now in the “Internet of Things” era: 1980s (PC), 1990s (Internet), 2000s (mobile/social), and 2010s (internet of things). The latter encompasses wearable devices, smart home devices, etc.
• Back when AgaMatrix was developing the iBGStar, the company applied the “Turnaround Test” to the design process: “If you’re halfway to work and realize you forgot ____ at home, would you turn around to go get it?” The answer is “yes” for indispensable things like keys, wallet, and cell phone. Passing this test is much harder, however, for things like a glucose meter. AgaMatrix came up with the idea to physically attach the meter directly to the phone, thereby passing the turnaround test. Brilliant! It was fascinating to learn that the company thought about a Bluetooth version of the meter, but deliberately chose not to. “We had to attach the meter to the thing people actually value.” [Of course, once Apple changed the iPhone plug-in to the lightning adaptor, the iBGStar needed a clunky adaptor that really killed the form factor.]
  
  • “Most wearables right now don’t pass the turnaround test.” For instance, if you forget your activity tracker at home, it’s unlikely you’re going to go back and get it. Mr. Vu Proposed two hacks:
    • Hack #1: Make the device so you don’t have to turn around. For instance, an activity tracker that doesn’t require charging, is easy to use, and you wear all the time (e.g., Misfit’s Shine) – that way, you don’t forget it in the first place.
    • Hack #2: Be everywhere in life. For instance, build the sensors into the phone or home, have glucose meters spread throughout your life.
  • We love the idea of the turnaround test as a metric for diabetes products – “Is something so useful and so compelling that it becomes indispensable to be without it?” We believe CGM may be at this point for some current 24/7 users, though it is clearly not there for those not on CGM.
• Wearables are currently in transition from Gen 1.0 to Gen 2.0. Gen 1.0 wearables have been plastic, bulky, not that attractive, and focused on measuring things (activity, sleep heart rate). Gen 2 wearables have better materials (metal, leather, ceramic), require less/no charging, and are designed to be worn for months and life-focused. These next-gen products will bring broader use cases like safety, identification, and controls (emergency monitoring, lock, login, payment, gestures, controlling house functions like lights and air conditioning).
• Why is the internet of things space interesting? Sensors can provide unprecedented data and insights. Great wearable products will give better data continuity, volume, and diversity. That translates into better algorithms and valuable insights.
• What moves us to change behavior? Rewards, connection (with other people), sex, thrill, and play. This framework reminded us of why it’s so incredibly hard to change behavior in diabetes/obesity – often, approaches hit none of the above.
• Why do people wear things? – aesthetics, fashion “(it makes me look good); comfort/protection (“it makes me feel safe”); brand/tribe association (“it makes me feel cool, like I belong”). Wearing something to simply gather more data is often a secondary purpose.
• Where do digital health and wearables need to go?
  
  • Passive data acquisition: “Nothing you have to remember to carry around or do.” Wearables and other ambient sensing solutions will need to provide insights, not j
  • Insights and timely feedback. Instead of just charts and graphs, tell people stuff they need to know. Mr. Vu characterized feedback as the most difficult and the most important.
  • Clinical effectiveness demonstrated in trials and some acceptance by HCPs.
  • Reimbursable: a reasonable pathway.
  • Finite: When can I delete your app?
• “It’s not the customer’s job to know what they want. That’s the difficult work we have to do...It’s about anthropology, it’s not about focus groups.” Mr. Vu shared learnings from AgaMatrix, where the team interviewed hundreds of people with diabetes. “We hardly learned anything. People just told us what we wanted to hear. We got sick of asking people what they wanted. ‘How would you design your glucose meter?’ People would answer, ‘Less blood, make it faster. Make it a better experience.’”
  
  • What proved most useful was sending people home with disposable cameras to take pictures of their real experience with diabetes. This resulted in tons of ideas on improving BGM case design (simply making a larger zipper), better lancing designs, better strip vial designs, etc. Mr. Vu called these areas “unbelievably untapped,” in part because they are “unbelievably unprofitable to tap.” He longed for some company to simply bite the bullet and just innovate in these important areas.

Questions and Answers

Dr. Aaron Kowalski (JDRF, New York, NY): Today, what are you doing to tease out what customers want?

A: For activity trackers, we read Amazon.com reviews. There were 5,000 reviews for Fitbit and Jawbone. We came up with the top five gripes and just said, “We’re going to solve them.” Waterproof, super durable, doesn’t make you look like a dork, and is comfortable. It worked.

Movers & Shakers: Introducing Bigfoot

Jeffrey Brewer (CEO, Bigfoot Biomedical, New York, NY)

Bigfoot CEO Jeffrey Brewer gave a moving closing keynote at Diabetes Mine’s D-Data Exchange, showing the team’s in-development automated insulin delivery system running on a smartphone (the slide showed a bolus calculator screen). The goal appears to be a smartphone interface communicating with the Asante pump and Dexcom’s CGM – ambitious indeed! The product will clearly have a major focus on the entire system, including extensive use of the cloud and remote monitoring (“just as important as on-body automation”). Notably, Bigfoot is even relocating to Silicon Valley to take over Asante’s 45,000 square-foot facility – they are wasting no time indeed (Jeffrey’s home in NYC is already on the market). In addition to the sliver of product details, Jeffrey’s presentation told the story of the company’s founding through his, Lane Desborough, and Bryan Mazlish’s personal experience with type 1 diabetes. The frontier/AP analogy worked very well, as he related the key constituents in settling the Wild West to closing the loop: “trailblazers” (Nightscout, #DIYPS, Lane Desborough, Bryan Mazlish); law and order (the FDA – “necessary and valuable” and open to innovation); doctors (especially PCPs, where most people with diabetes are seen); and the bank (payers – “we must meet their terms, like meeting the terms of a bank loan”). The entire room was noticeably captivated. In a nod to naysayers that believe the 800 lb gorilla is unbeatable (i.e., Medtronic, though he didn’t say it by name), a slide highlighted how former giants like Yahoo, Kodak, and Blockbuster were decimated when newcomers Google, Flickr, and Netflix came into the picture – “I think competition is good; it makes us all better and will deliver better solutions to my son and to all people with type 1 diabetes.” Summed up Jeffrey, “Closing the loop is no longer a mystery; it’s not a puzzle. We just have to do it.”

• “We are developing a system where the phone is the interface. A system that automatically adjusts insulin delivery with everything you need to know, when you need to know it, all on your smartphone. It’s streaming data up to the cloud, and the remote monitoring aspects are just as important as the on-body automation.” The focus on remote monitoring makes a lot of sense from a team-experience perspective (Bigfoot includes Nightscout developers Lane Desborough and Ross Naylor), as well as a safety and healthcare cost perspective – we look forward to seeing what this entails.
• “I’m really excited – we have fundamentally phenomenal components.” Mr. Brewer reminded attendees that Bigfoot recently acquired the assets of Asante (“a company focused on doing the right thing: making it simpler and easier for people with type 1 diabetes”), and just prior to ADA, established a CGM partnership with Dexcom. As far as we know, the Bigfoot product will consist of a Gen 5 Dexcom transmitter sending CGM data to a custom built Bigfoot controller married to the Asante disposable insulin pump body. One thing we’re not sure about is if Bigfoot will pursue a Novolog version of the Asante pump; right now, the Snap is only cleared and designed for Humalog pre-filled cartridges.
• “We’re moving to Silicon Valley, as of today. I just put my house on the market. #ItIsTime.” Mr. Brewer showed a cartoon cityscape containing billboards of major tech companies (Google, Facebook, Twitter), and with a bit of humor, boldly added in the Bigfoot logo. Continuing the Wild West analogy, he said that Bigfoot “embraces the Western ethos” and wants to pioneer the new frontier of automated insulin delivery.
• “In 2002 when my son was diagnosed, I was handed some needles, a couple bottles of insulin, and a handwritten page with a sliding scale for insulin. We were landing planes by autopilot in 2002!” “At the time,” he said, “that’s the best the top endocrinologists could do.”
• “Bryan Mazlish, Bigfoot, has a son and wife that have been benefitting from this technology for the last two years; 30,000 hours of usage. It’s a very simple thing. It gives you less insulin when you need less, and more insulin when you need more. The technology is there. We just need to make it happen for everyone.”

Questions and Answers

Q: Where does Bigfoot stand in terms of openness and data access?

A: We are still working through what that means. But look at the people that created this company. We are the people who have been working in that space. It’s a moral issue and that’s the smart thing to do. Companies that embrace that are more successful. Details are forthcoming. We were building own insulin delivery device until eight days ago. Now we have one, and a wonderful partnership with a leading sensor company.

Mr. Howard Look (Tidepool, Palo Alto, CA): You are the new kid on the block, though you and Lane bring tons of experience from industry. A lot of other device companies are trying to work on closed loop, but it’s taking a long time. What things will you do differently to get to market quickly?

A: The algorithm is solved – we’re not doing anything differently. Our approach might be more commercially friendly, but there are a bunch of different ways to do algorithms. It’s really about the design of the whole system – that’s where the value gets created. How the sensor plays with the pump, the phone, the cloud, the tracking, and monitoring – it’s the whole solution and ecosystem. We’re putting the pieces together in a smart way, like what Apple does – they are not a hardware company. They are an ecosystem company. They created iTunes, which was a tipping point. It was the pricing model to buy songs – the software made the hardware really sing. We have a similar approach to leveraging the hardware.

Q: You are starting from scratch with clinical trials...

A: I don’t know anybody who’s been in a pivotal trial of a closed-loop system.

Q: From a technical point of view, is the sensor there in terms of knowledge and accuracy?

A: It depends on what you want to do with it. It was asked earlier: what’s the perfect solution? Take what the technology can do today, push it as far as you can today, and no more. Deliver something safely. The answer is you can do a whole lot better than people can do themselves. How does type 1 diabetes work? You may have sloppy estimates of carbs, sloppy boluses, and no information between meals when you’re not testing – or you don’t understand CGM data. People are going throughout their lives and stumbling into traumatic health episodes. My own son almost died – he was in the ER and then the ICU for 48 hours. He couldn’t speak his name for 24 hours. After he gave too much insulin, the pump increased the rate before dawn – it was preprogrammed to change. That’s really, really dumb. We can do better than that. Will there be adverse outcomes? Absolutely. But there will be far fewer.

Company Updates

BD Partners with Medtronic to Commercialize FlowSmart Infusion Set

BD and Medtronic announced an exclusive, worldwide collaboration to commercialize BD’s new FlowSmart infusion set. The set will launch in 2016, roughly on par with BD’s timeline to launch in FY16 (October 2015-September 2016). It will be made available for both Medtronic and non-Medtronic pumpers, though Medtronic will control distribution in both cases. As far as we understand it, patients at other pump companies who want the set will either (i) need to become Medtronic customers to access it; or (ii) go through a distributor, who will in turn buy it from Medtronic. While it’s good to see that all Animas, Roche, and Tandem patients will be able to get the set, the exclusive deal is a tough dagger for those pump companies. Sets are a high margin business, and the exclusive deal means those pump customers may now buy the new BD set from Medtronic. See our detailed report for a deeper dive on the deal and the potential impact on patients, Medtronic, BD, other pump companies, Convatec (Unomedical), and HCT/JDRF. We also share our key questions and provide first thoughts after an up close look at the new set.

Medtronic Begins MiniMed 670G Pivotal Study

Medtronic announced a 150-patient, three-month US pivotal study of the MiniMed 670G/Enlite 3 CGM hybrid closed loop (HCL) is now recruiting – see the very informative ClinicalTrials.gov posting here. The primary objective of the pivotal study is to evaluate the safety of the 670G system; secondary endpoints include mean change in A1c, change in insulin dose and weight, time spent with closed-loop on vs. off, and time spent in different sensor glucose ranges (we obtained these from the company directly, as the posting incorrectly implies the primary endpoint is A1c at 12 months). Notably, it is a single arm study – during a two week run-in period, participants will be using the study pump (MiniMed 670G) with only the CGM function activated (i.e., SmartGuard OFF and HCL OFF). Subsequently, participants will wear the 670G at home for three months. The study also includes a short six-day hotel stay phase. We assume the baseline run-in period will serve as the “Control” group, which is far less onerous than the parallel group RCT that could have been requested – kudos to FDA for being so reasonable here! It’s also quite notable to see that the at-home portion of the study is just three months, and expected completion is slated fairly conservatively for May 2016 (11 months from now – we assume it could finish much earlier). That indeed puts Medtronic loosely on track to hit the April 2017 launch timeline (as of JPM 2015), assuming the PMA submission and approval take less than a year – we expect that it could if the MiniMed 640G is approved in the meantime (i.e., largely the same pump user interface, with the exception of the algorithm changes).

• The study inclusion criteria are quite broad, as those with an A1c <10% and ages 14-75 years will be enrolled. It seems like the 670G will be worn day+night in the study, and based on our read of the exclusion criteria, patients won’t have an option of wearing it at night only – we do think this would be a nice option, but understand the rationale for 24-hour wear in the pivotal study. (Whereas in the real-world, patients can choose when and how they want to wear it; for more on this balance, see diaTribe’s test drive on DiAs.)
• This news vaults the company into first place in terms of doing a pivotal study for a hybrid closed loop device. It’s a critical time too, as the competition from other organizations continues to heat up.

 

-- by Adam Brown, Varun Iyengar, Dana Lewis, Jenny Tan, Vincent Wu, and Kelly Close