July 6-7, 2016; Bethesda, MD; Day #1 Highlights – Draft

Executive Highlights

Greetings from Bethesda, Maryland, where Day #1 of the Fourth Artificial Pancreas Workshop at NIH is in the books - this year it’s co-sponsored by the Diabetes Technology Society, FDA, the Helmsley Charitable Trust, JDRF, and NIH. An aura of nostalgia has surrounded the meeting thus far – JDRF’s Dr. Aaron Kowalski recalled the inaugural meeting back in 2005, how the artificial pancreas was mostly just a concept at that point, and how data was incredibly scarce (back then, it was the original Guardian and Dexcom sensors – the CGMS Gold and the STS!). Dr. Bruce Buckingham expressed amazement that in just 11 years, talk has shifted from conceptual feasibility to quality of life. Indeed, this is an exciting time to be in the field as it’s on the cusp of commercialization.

The search for the next frontier in automated insulin delivery (AID) research represented a major theme of the day. Our own Adam Brown urged the entire field to think about this in Q&A: “Where should this field direct its resources over the next five years? If we can really define that specifically, it has huge implications for how funders allocate resources, and how product design changes. Adoption is about broader concerns than taking time-in-range from 80% to 95%. And everyone in the field would be bummed if AID was just for people already on pumps and CGM. What do we need to do to make this technology more widely available?” Some argued in favor of psychosocial research to boost patient acceptance and manage expectations of closed loop technologies, others for provider/patient education, many for broader study inclusion, and yet others for design and algorithm optimization. We would want to see all of this prioritized to at least some degree. See our top highlights below.

1. Testing automated insulin delivery in more diverse populations was the biggest priority that emerged in today’s discussion: those with high A1c’s, MDI users, elderly and very young patients, those with frequent severe hypoglycemia, type 2 patients, pregnant women, underserved populations, etc. This is still a major hole in the field’s data gathered to date, but we expect it to change as comfort grows and funders begin mandating these studies.

2. The MiniMed 670G pivotal study (presented at ADA) was criticized for the lack of a control group, limiting any conclusions about the system’s efficacy. The single-arm study design was defended as a safety-focused trial optimized for getting to market quickly, not proving efficacy.  

3. FDA’s Dr. Alberto Gutierrez highlighted Medtronic’s recently submitted MiniMed 670G/Enlite 3 hybrid closed loop. The Agency is “clearly excited” about it and will bring “all efforts to the table to get it approved as quickly as [it] can.” It’s nearly unprecedented in our view to hear this kind of enthusiasm from the FDA in advance of a product being approved – to hear the Agency not only give its intention but to discuss timing broadly.

4. A strong session on the psychosocial side of AID drove one view home: a system’s quality of life impact is just as important as how well it controls glucose. This is often lost, it was said, in the engineering- and glucose-centric discussion of AID. Of course, the FDA may well differ on this – the DCCT and other studies have shown the impact of lower A1c on long-term control while we haven’t heard as much about the impact of better quality of life on outcomes. We’d love to see more such designed, studied, and funded to address this.

5. “I continue to be concerned about the hype associated with automated insulin delivery. We may raise an expectation that is simply unattainable,” noted Dr. Katherine Barnard. Others were less concerned about managing expectations, arguing that those new to technology may see better efficacy (relative to open loop) and have lower expectations for what devices can actually do.

6. The legendary Dr. Bruce Buckingham (Stanford University, Palo Alto, CA) highlighted the incredible challenges of closing the loop in young people, argued that intensive inpatient studies in the very young should not be used, and expressed optimism that adolescents are “ideal” candidates for AID.

7. Dr. Rich Bergenstal (International Diabetes Center at Park Nicollet, Minneapolis, MN) began his talk on closed loop in older patients in an unorthodox fashion, by stating his six-word conclusion with emphasis: “We need a lot more data!” He estimated that only ~20 patients >65 years have ever been studied on closed loop! That was a shockingly low number to hear.

8. Cambridge’s Dr. Roman Hovorka argued in favor of an insulin-only artificial pancreas, noted widely varying outpatient RCT results (algorithms and study population/setting really matter, he emphasized), and laid out two future directions to expand the field (higher and lower complexity).

9. MGH’s Dr. Steven Russell provided an expansive view of multi-hormone closed-loop control, arguing persuasively for the value of glucagon and highlighting “promise” for use of pramlintide.

10. Multiple speakers expressed concerns about training healthcare providers to handle AID, compensating them for their time, and whether PCPs and NPs can even handle new diabetes technology.

11. Now that the MiniMed 670G is under FDA review, Medtronic and DreaMed are already moving to incorporate automatic correction boluses into the follow-up “Advanced Closed-Loop System.” A feasibility study was completed in June, testing the DreaMed MD Logic algorithm with the MiniMed 670G’s hybrid closed loop algorithm. A bridging study is on the docket too.

12. Following the initial announcement at ADA, Diabetes Care has published multiple critical papers on AID including “Outcomes Measures for Artificial Pancreas Trials: A Consensus Report” by Dr. David Maahs et. al. (this author list is a “who’s who” of the AID world and we’re extremely impressed by the breadth of the group) and “Design Considerations for Artificial Pancreas Trials” by Dr. Steve Russell and Dr. Roy Beck. These appear in a larger section in this month’s Diabetes Care, “The Artificial Pancreas in 2016: A Digital Treatment Ecosystem for Diabetes.” We salute the ADA into putting so much work into this section.

13. Moving forward, control algorithm research is very focused on personalization and adaptation, including variable targets and algorithm aggressiveness that moves flexibly with patients over time.

14. Notably, we did not hear a single person today lament CGM’s accuracy or reliability – a testament to how far sensors have come, and how much we can do with what is available now.

Top Highlights

1. Testing automated insulin delivery in more diverse populations was the biggest priority that emerged in today’s discussion: those with high A1c’s, MDI users, elderly and very young patients, those with frequent severe hypoglycemia, type 2s, pregnant women, etc. Based on data from the T1D Exchange (Miller at al., Diabetes Care 2015), many argued that patients with high A1c’s (adolescents) and highly hypoglycemia-prone patients (those with 40+ years diabetes duration) are probably the most likely to benefit from automated insulin delivery; unfortunately, they are generally not included in studies (Dr. Rich Bergenstal estimated only ~20 patients over 65 years have EVER been studied with closed-loop!). This is obviously a solvable problem, and partially reflects the field’s caution to test systems in well-controlled patients before moving on to more challenging populations. It also brings us optimism: will current studies undersell the benefits of this technology, as patients are often doing “incredibly well” at baseline (A1c’s ~7%)? (Editor’s note - this raises yet again how the field characterizes patients’ success – an A1c below 7% is certainly positive if it is accompanied by low hypoglycemia; often it isn’t.) We include some of our favorite quotes on this topic below.

• David Panzirer: “The folks in the room represent a small faction of type 1 diabetes. The voice of most people out there has not been heard, specifically adults with T1D who are not seen by and endocrinologist. How do we build these automated insulin delivery (AID) systems for them? We are guilty of drinking our own Kool-Aid. We must incorporate the adult populations, and perhaps as funders we should force studies to recruit adults who are technology naïve and outside the diabetes clinics. Even if we have to pay extra money, we need to get the voice of those who are clueless about this technology – we don’t know what these people really want. To me, if this was a business, it would be an absolute failure. We have to do the market research across all segments if this is going to succeed. Look at Afrezza: in my opinion one of the reasons for their struggles is they didn’t do the market research, they didn’t educate clinicians, and they didn’t ask patients about tradeoffs. We’re going down the same path with automated insulin delivery: if we don’t do our homework, this will only work for a small proportion of the population.” We certainly agree that expanding the group of patients for trials would be positive – today, that may be challenging since there is so much competition within centers to get into these trials. Broader populations could of course be mandated. There were also additional challenges related to Afrezza, particularly related to reimbursement.
• Dr. David Maahs: “We need to be studying kids with A1c’s of 9%, 10%, 11%. We don’t know what we don’t know, and they have potential for the greatest benefit. Thinking back on how we rolled out pumps has changed in last 10-15 years. Previously, we were more restrictive as to whom we put on a pump; now we are broader. It would sure be nice to have data on these people who are struggling with diabetes and have 9% A1c’s, especially before this goes out to bigger populations, where we could have some surprises.” This was terrific to hear and we do point out that multiple patients with lower A1cs are also often “struggling” – no doubt, however, the field would benefit from broader populations tested, particularly teens. Market research on this group could be useful to see if patients in this group would like to be part of trials and then to see what motivators could be found.
• Dr. Roman Hovorka on type 2 diabetes: “There is huge potential. Commercially it might even be more of a way to go forward, and we can earn some money to use in type 1 diabetes. The challenge is providing cheaper, simple pumps and reducing the cost of CGMs. But I think that is happening. The burden of hypoglycemia in type 2s is big, and if we make it simple for an older population to use, it is there. I know this is a type 1 diabetes meeting, but it is interesting to emulate this in type 2.” It was interesting to hear the meeting characterized as a “type 1” meeting; though virtually all of the conversation surrounded type 1, the name of the meeting “Fourth NIH AP Workshop: Testing and Adoption of Current and Emerging Technologies” doesn’t actually specify a type of diabetes.
• Dr. John Pickup: “I would stress the need for testing in appropriate groups. Patient groups that are going to use and benefit from these systems – for instance, those with severe hypoglycemia – have hardly been evaluated at all. There is a danger if a commercialized product is unleashed and has not been properly tested in that group. Then there might be big trouble.”
• Dr. Roman Hovorka on older patients: “Our first elderly subject used closed loop at home for three weeks and was a 70 year-old. It was uncomfortable for her to be at 6 mmol/l (~110 mg/dl), as she is used to running higher. She didn’t like the system and went off it. We might need to think differently about this population. Their psychosocial issues and targets might be different. This is anecdotal, as we don’t have systematic data.” We were glad to hear this characterized as “anecdotal” since we would not want this view broadly generalized to anyone who is 70 years old! We also wonder whether the patient’s experience may have been different depending on how it was originally presented to her.

2. The MiniMed 670G pivotal study was criticized for the lack of a control group, limiting any conclusions about the system’s efficacy (see our ADA 2016 coverage). Everyone in the room was excited that the first commercial hybrid closed loop is now under FDA review (submitted just after ADA), though Medtronic’s Dr. Benyamin Grossman and investigator Dr. Rich Bergenstal had to defend the single-arm study design in Q&A: a safety-focused trial optimized for getting to market quickly, not proving efficacy. Dr. Grossman said quickly, on the spot – “The FDA told us, ‘We don’t care about efficacy’” – eliciting disapproving headshakes from the Agency’s Drs. Courtney Lias, Alberto Gutierrez, and Stayce Beck in the back of the room. Then, Dr. Roy Beck jumped in: “Just to clarify: I’m certainly positive the FDA cares about efficacy. What I think happened is that there was a path to get labeling that didn’t require an RCT to show efficacy, and that’s the path that was taken [FDA nods in back of room.]. My personal view is that it was an unfortunate path. However, it does get the 670G through the process faster.” This is a challenging call for companies, and while it’s easy to criticize Medtronic for doing an uncontrolled study, we appreciate the goal of getting this system to market faster – it is up to FDA to assess if that is enough information though we very much appreciate “safety first” before approval and efficacy after. Notably, Medtronic is planning a rigorous outcomes study (n=1,000) that will enroll a spectrum of type 1s (wide A1cs and ages) and randomize patients to three groups for six months: pump alone, sensor-augmented pump (no automation), or the MiniMed 670G. While as noted, we like the move to put the focus on safety pre-market and efficacy post-market, it will be interesting to see how other companies pursue this – not all will have the resources for a 1,000-person outcomes study.

• Dr. Roy Beck highlighted the importance of a control group, citing Dexcom’s DIaMonD study presented at ADA: even in the control group of MDIs on usual care, A1c improved by 0.4% with very few visits and no devices added (“If that was an intervention, we’d say it was pretty successful”). Dr. Beck clarified that this isn’t to say the MiniMed 670G isn’t efficacious, but based on the pivotal study, no firm conclusions can be drawn about how it performs.
• Dr. John Pickup also weighed in with his concerns on proper AID study design: “The technology is fantastic and constantly changing. But we have to be careful not to let technology run away from the need for careful clinical investigation. I would caution the need for proper RCTs, rather than observational studies, particularly in testing of commercialized products. I would also caution on the difference between short-term evaluations and long-term evaluations. In what we’ve seen with CSII, there is a big difference between what happens in six months and a year and what happens over five years – those who respond and the groups who don’t respond at all. Who are the non-responders, and what’s the psychological profile of those who do better or worse with the artificial pancreas?” This was interesting commentary and we are interested to know what the differences in cost and time would be if RCTs were required. We caution that RCTs often overstate the impact of “regular” care and overstating the value of RCTs could also be a danger since the “control group” often has elements of life “controlled” that won’t be controlled in “real life.”

3. Despite concerns voiced about the single-arm trial, FDA’s Dr. Alberto Gutierrez enthusiastically highlighted Medtronic’s recently submitted MiniMed 670G/Enlite 3 hybrid closed loop system. The Agency is “clearly excited” about it and will bring “all efforts to the table to get it approved as quickly as [they] can.” This echoes Dr. Courtney Lias’ enthusiasm for the MiniMed 670G during the Live Artificial Pancreas Webinar last month, and subsequent positive comments expressed at the D-Data Exchange at ADA. Dr. Gutierrez hoped that by the next AP Workshop in two years, there will be many approved devices available so that “focus can shift toward prevention and cures instead of management.” This statement was strikingly optimistic – the devices and algorithms will still require ample design optimization and access work in the coming years – but we liked how lofty this vision is. Of course, first-gen AID is not going to solve all diabetes management problems, but we’re glad to see FDA is motivated to get system to market quickly and learn from early systems that are approved.

4. A strong session on the psychosocial side of AID drove one widely-voiced view home: a system’s quality of life impact is just as important as how well it controls glucose. This is often lost in the engineering- and glucose-centric discussion of AID. Said psychologist Dr. Katherine Barnard (BHR Limited, UK), “This is not about glycemic control. It’s about, ‘How can I use this to improve my life or to improve the burden of diabetes?’” We very much appreciated this view although from a reimbursement perspective or even regulatory perspective, quality of life has not yet been elevated to this view (though it is acknowledged more frequently on the device side vs. drug side of regulatory agencies). Drs. Barnard and Lori Laffel highlighted the ongoing INSPIRE project to develop standardized measures for assessing the psychosocial impact of automated insulin delivery; the goal is still to have these ready by Fall 2016, consistent with the timing Dr. Korey Hood shared at ADA. The project has included an impressive 60+ focus groups, 89 individual interviews, and over 400 participants to date (200+ hours of transcription with adults, children, and adolescents with diabetes; parents; and partners). The standardized measures for assessing the psychosocial impact of automated insulin delivery will be piloted in clinical trials this summer according to the researchers. Notably, the powerhouse four-person team (Drs. Barnard, Hood, Laffel, and Jill Weissberg-Benchell) has been working with the FDA to ensure these measures will be useful in regulatory reviews – this was terrific to hear from a patient perspective and we will be following this closely to better understand what FDA will be requesting and how the data requests will be standardized. There was much praise for the vision behind this Helmsley-sponsored project: enable consistent and critical appraisal of patient-reported outcomes across clinical trials, inform patients and prescribers as systems come on the market, and assist payers and regulatory bodies in assessing AID systems. It will be phenomenal to combine these measures with standard glycemic efficacy measures, offering a far more holistic view of what AID offers – even if a system shows a small incremental advantage in a pivotal trial, it might still revolutionize quality of life. We would like to see this captured in guidance.

• Dr. Barnard: “Devices must be discreet, small, accurate, effective, reliable, flexible, and easy to use. ‘Allows me to have a normal life. Gives me freedom in my life. Let’s me not have to think about diabetes all the time. Gives me a better quality of life. Makes me happier and less exhausted with diabetes.’” Although it is challenging to capture in trials the impact of being “happier and less exhausted” we would love to better understand how higher patient engagement impacts short-term and long-term outcomes.
• The psychosocial session also emphasized why post-market learning is critical for understanding the psychosocial impact of AID: (i) we will not be able to foresee all the psychological issues associated with these systems until they are used by more patients for a longer period of time; (ii) people will inevitably test systems’ limits by doing “crazy things” (e.g., workarounds to make them more aggressive by arbitrarily inflating the open-loop settings, taking manual boluses and not informing the system); (iii) many people will have to “unlearn” over-learned behaviors, particularly those already using sensor-augmented pump therapy (e.g., treating hypoglycemia with food when an AID system has already taken action). We thought these were very valuable points and a reminder of Dr. Astro Teller’s outstanding and highly relevant talk on Google’s self-driving cars from SXSW 2015.
• These psychosocial measures will cover a lot of fascinating ground: perceived benefits and downsides associated with AID; acceptance; trust; usability; cost concerns; impact on social relationships; impact on family members; impact on perceived burden of disease; and more. Drs. Barnard and Laffel gave more examples of questions than we heard at ADA: “An AID system will help me worry less about diabetes”; “If the AID system improves my A1c, than I will put up with...meal announcements, exercise announcements, having to count carbs, having to check my blood glucose levels”; “I worry that the AID system will fail”; “I will panic if the AID system fails”; “I worry I will check the AID system is working many times per day.”

5. “I continue to be concerned about the hype associated with automated insulin delivery. We may raise an expectation that is simply unattainable,” noted Dr. Katherine Barnard. Others were less concerned about managing expectations, arguing that those new to technology may have higher ability to improve (over open loop) and lower expectations for what devices can actually do. Said Dr. Rich Bergenstal, “I’m a little less pessimistic and worried about expectations. In the experience I have had, we start off with amazing patients who are tech savvy. They say, “See if you can beat this, I’m taking 22 shots a day. And in my experience, we do. But they have to be taught to let go, take a deep breath, and let the system work. But several have been very happy. We also had a few patients in the [670G] pivotal trial that were not so intensively managed, and it was much easier to train them, because they weren’t untraining and unlearning. Their expectations were, ‘I was afraid of technology, but I’ll give it a try. Wow, this was better than I ever imagined.’” Still, many pointed out that “unrealistic expectations” about what AID will do will actually increase stress, burnout, and lower quality of life – far more negative than just disappointment and discontinuation. From our view, it’s a lot about presentation and we believe if the communication with patients is there about stress, burnout, user interfaces that may not live up to expectations, that will be very positive – it will be important to inform early users what the systems do not do and what future hopes and plans are. Dr. Lori Laffel noted that patients still have a long list of continuing concerns about accuracy, false alerts, worries about failures, multiple devices and insertions, frequent site changes, tubing (“an extraordinary challenge”), carrying backup supplies, non-waterproof devices (“an extraordinary burden”), discretion, compactness, and unobtrusiveness. Managing expectations was also a big concern raised at ADA, and we’ll be interested to see how Medtronic does it with the MiniMed 670G as it comes to market – that product may set the tone for years to come. We strongly believe this is addressable and are very grateful that so many researchers and patients and patient families have raised the importance of this.

• Dr. Barnard: “People expect it to do what is says on the tin. We have all this great engineering and biomedical excellence, but people don’t care much about what goes into the box; they just care that the box works.”
• Dr. John Pickup: “We need to bear in mind the need for absolute reliability. We don’t want a repeat of insulin pump therapy where there are still a lot of malfunctions occurring. As CSII gets more sophisticated, it has not been matched by reliability of the technology. I want to see that built into evaluations.” We’d love to better understand how this would be built into evaluations – the evaluations would need to be especially large to capture of all this.

6. The legendary Dr. Bruce Buckingham (Stanford University, Palo Alto, CA) highlighted the incredible challenges of closing the loop in young people, argued that intensive inpatient studies in the very young should not be used, and expressed optimism that adolescents are “ideal” candidates for AID. To start, children are generally diagnosed with diabetes at a very young age (<15 years), so the artificial pancreas, if implemented at diagnosis, might be given to those as little as 0-2 years old. This population is at high risk for ketosis and hypoglycemia, has limited skin “real estate”, irregularities in food consumption patterns, frequent infections, and, most notably, miniscule insulin requirements. To illustrate that last point, a one-year-old infant weighing 9 kg (~20 pounds) might require ~0.08 units per hour of basal insulin … if that rate is slightly raised or lowered by 0.025 units per hour, that represents a physiologically significant ~33% change! This problem might be addressable with diluted insulin and/or a pump that delivers insulin in smaller doses, but that is certainly a challenge for closing the loop. Adolescents pose a whole different set of constraints: hormones, underdeveloped frontal lobes, frequently missed meal boluses, activity, self-image, and huge appetites. He shared one flabbergasting anecdote of one teen in an AID study, whose carb consumption varied from 160 to 971 grams per day – seriously, almost 1000 g of carb in a day! In both groups, Dr. Buckingham said, closed loop can be very beneficial – in fact, single and dual hormone systems have been shown to reduce adolescents’ estimated A1cs to 7% and 6.5%, respectively, far better than they are doing in the T1D Exchange (~9%). Dr. Buckingham concluded with a persuasive argument against inpatient trials in toddlers and young children: It is hard for them to go through intensive medical procedures (e.g., YSI monitoring) and be away from home, yet when they are in outpatient studies, they are virtually always supervised and monitored by their parents (“very controlled settings”). Instead, new systems should be tested for reliability in adolescents and adults in an inpatient/hotel setting, and trials in young children should be done in an outpatient setting at home. It’s a great point for companies and the FDA to keep in mind.

• As he often does, Dr. Buckingham emphasized the importance of handwashing and taking the second drop of blood, which greatly increase fingerstick accuracy (and thus, CGM calibration). He showed data from a fascinating 2011 Norwegian paper (Hortensius et al., Diabetes Care) on peeling fruit, highlighting the impact of dirty hands and the value of simply wiping the first drop off when handwashing is unavailable – a good one for patients to keep in mind. Dr. Buckingham also reiterated his data from camp studies of the Dexcom G4 and Enlite sensors, showing significantly elevated MARDs with the same sensors when used outpatient (vs. BGM) vs. inpatient (vs. YSI) – a trend that was remedied with Enlite 3 when patients were instructed to only use the second drop of blood for calibration. We wonder if this could be incorporated in CGM labeling.

Effect of Handling Fruit on Meter BG 

7. Dr. Rich Bergenstal (International Diabetes Center at Park Nicollet, Minneapolis, MN) began his talk on closed loop in older patients in an unorthodox fashion, by stating his conclusion with emphasis: “We need a lot more data!” After a thorough literature review, he and his team estimated that only ~20 patients >65 years have ever been studied on closed loop! As our population ages, he said, it will be more and more important to understand the interactions, both physical and psychological, between artificial pancreas technologies and older adults. These individuals with type 1 diabetes are excellent candidates for closed loop, of course – only ~30% meet the target A1c <7%, and more importantly, ~20% of those with a diabetes duration >40 years had a coma or seizure in the last 12 months. (Dr. Bergenstal noted, however, that diabetes duration, not age, was the main driver of this effect, even more concerning since younger patients can also have a long diabetes duration). On top of that, individuals who had recently experienced severe hypoglycemia were prone to recurrent hypoglycemia and generally more hypo unaware. Dr. Bergenstal urged the field to start testing closed loop in these patients, who stand to benefit tremendously, but concluded with provocative questions for clinicians and researchers to consider: (i) How will older adults handle this technology? How can we evaluate the degree to which they are tech savvy? (ii) How will artificial pancreas work in those with cognitive decline or dementia? (iii) What is the best way to train the elderly and those supporting them?

8. Dr. Roman Hovorka (University of Cambridge, UK) argued in favor of an insulin-only artificial pancreas, noted widely varying outpatient RCT results (algorithms, study population, and setting MATTER), and laid out two future directions to grow the field. He began by answering the question “Why insulin-only artificial pancreas?” First, he pointed out, the insulins we have at our disposal are not perfect, but they are good enough to improve control. Next, insulin has low biological risk, “no unknown unknowns” as he put it, alluding to the potential risks associated with chronic use of glucagon and other therapies. The single hormone design also allows for lower system complexity, which Dr. Hovorka emphasized is crucial for device acceptance. He still believes there is sufficient innovation potential with insulin-only systems, such that control can improve without a concurrent increase in complexity. Dr. Hovorka reviewed a slew of insulin-only outpatient RCTs from DREAM, Medtronic, DiAs, and Cambridge, noting inconsistent results: reduced mean glucose and reduced hypoglycemia, reduced mean glucose and no change in hypoglycemia, reduced hypoglycemia and no change in mean glucose, no change whatsoever, and increased mean glucose and reduced hypoglycemia! His point was that the algorithm and the baseline characteristics, study population, and settings really matter, and we may need to think differently about the strengths and requirements of different populations and systems. Dr. Hovorka proposed two distinct paths for the future of closed loop control: similar/reduced complexity or increased complexity. The path of same/reduced complexity involves simplifying meal dosing, developing faster insulins, completely closing the loop, combining glucose sensing and insulin delivery in a single cannula (something that an EU project and Pacific Diabetes Technologies in Oregon are currently pursuing), pushing for factory calibration, prolonging insulin cannula use, and/or advancing algorithms to make them more adaptive. Alternatively, he argued, the path of increased complexity would include other hormones or additional physiological sensors such as heart rate monitors and accelerometers to facilitate glucose management during exercise.

• Dr. Hovorka did mention that although insulin itself has no “unknown unknowns”, uptake of closed loop systems is an “unknown unknown.” A great point! If a product becomes more complex (say, by the addition of a heart rate monitor), the increased complexity must obviously be balanced by increased efficacy. In that vein, Dr. Hovorka said that, based on data from home studies, people do not want to carry extra devices. More market research of this kind – what people want and are willing to tolerate –  should ultimately inform the trajectory of the field as it moves into gen one and beyond.

9. MGH’s Dr. Steven Russell provided an expansive view of multi-hormone closed-loop control, arguing persuasively for the value of glucagon and highlighting “promise” for use of pramlintide. He focused only a bit on the Bionic Pancreas team’s work, mostly reiterating commentary from ATTD and ADA – relative to insulin alone, glucagon can reduce mean glucose, get more patients to an estimated A1c <7%, reduce/not increase hypoglycemia, and enable a better user experience with more spontaneity. Dr. Russell covered Yale’s work on additional therapies in closed-loop control, highlighting: (i) pramlintide’s “promise” for reducing post-meal excursions; (ii) limited benefit of GLP-1 (liraglutide) for glycemia in the closed-loop setting (though it did spur weight loss and insulin reductions); and hope that SGLT-2s will show a benefit in closed loop (soon to be tested at Yale. The best method of pramlintide delivery – bolus vs. infusion – is unclear, though Dr. Russell wondered if an optimal ratio or co-formulation could be developed. Dr. Aaron Kowalski has long been a proponent of this approach, and we hope as pramlintide goes off patent this might move forward. Looking ahead, Dr. Russell highlighted that Zealand, Xeris, and Adocia are all working on stable glucagons, and the big question is long-term safety of chronic microdosing – this will be evaluated over 12 months (n=100) in the Bionic Pancreas bihormonal pivotal trial, slated to start after the insulin-only trial in 2017 (similar timing as we heard at ADA). Dr. Russell did not mention the plans as of ADA to use Zealand’s glucagon in studies starting in 2H16. The team also has plans to test glucagon in the presence of alcohol and liver disease, key areas to establish glucagon’s safety.

• To quell concerns about over-relying on glucagon infusion, Dr. Russell pointed to several design mitigations in the Bionic Pancreas: (i) defaulting to insulin-only mode when glucagon is not available (e.g., empty chamber, catheter occlusion); (ii) use of steel cannulas and a snake bite infusion set (less susceptible to occlusion or yanking only one cannula out); and (iii) the ability to raise the glycemic target to dose insulin more conservatively.
• Dr. Russell noted the two other insulin+glucagon groups with outpatient studies, Montreal under Dr. Ahmad Haidar and INREDA from the Netherlands, reiterating Dr. Trang Ly’s timing update at ADA: INREDA’s bihormonal system will run a pivotal study in 2016 and apply for a CE Mark in 2017. It is nice to see a purpose built system moving to commercialization, though the earliest version looks clunky, with a pretty large dual-chamber pump and requiring two CGM sensors to be worn. We don’t know much about INREDA; they have published one short study showing increased time-in-range and reduced hypoglycemia.

10. Multiple speakers expressed concerns about training healthcare providers to handle AID, compensating them for their time, and whether PCPs and NPs can even handle new diabetes technology. We were so glad to hear this since we see the exhaustion constantly and believe this topic is very central to the eventual success of the systems. Dr. Rich Bergenstal called it a “huge training issue,” Dr. Henry Anhalt highlighted big issues in PCP education (“Who will support this 24/7? What happens when a patient calls a PCP in the middle of the night?” – we might add, or an endo, or diabetologist, or nurse), and Dr. Robert Vigersky lamented the failure to adequately educate medical students: “Those who go into primary care are totally unprepared for 21^st century management of diabetes. A wonderful result of this conference would be to develop some strategies for improving education in medical schools.” (We agree and also believe that many of those focused on endocrinology itself are also “unprepared for 21^st century management of diabetes”.)  Bigfoot Biomedical’s Jen Block also highlighted her own experience teaching diabetes technology in an NP program, “I can tell you that putting technology into the hands of providers with no experience with it is mind boggling. I welcome all of you to give it a try ... We have missed the boat on designing systems that have potential to be implemented successfully in primary care. We have a lot of work to do ... To hand an insulin pump today to an NP, it takes a really long time to train them. I think we could do things in a way that makes this so much easier.” Dr. Helen Murphy was perhaps the least pessimistic, noting that in one of her closed-loop studies in pregnancy, 50% of patients were MDI users and 90%+ were not on CGM. After training them on pump and sensors, closed-loop training only took an additional 30-60 minutes, she said, the same for experienced SAP users. She encouraged the group not to get too bogged down on training, emphasizing, “If we train our participants well, they will train healthcare providers.” Still, this concern over education and training was a big theme of the day, and it’s clear that there is a great deal of work to do on this front, particularly in clinics where there is not a lot of funding.

• David Panzirer shared another valuable view: “The biggest challenge we face in diabetes technology is the lack of end users. We need an awareness and education campaign, particularly out to primary care physicians. If we build systems for the professional patient, we will have failed. A goal that systems could easily achieve is to get all those double-digit A1c’s down to 8%. It’s getting from 8% to 7% or 6% that will be harder.” From a systems perspective, getting patients to 8% from double-digit A1cs has very high value, of course.

11. Now that the MiniMed 670G is under FDA review, Medtronic and DreaMed are already moving to incorporate automatic correction boluses into the follow-up “Advanced Closed-Loop System.” A feasibility study was completed in June, testing the DreaMed MD Logic algorithm with the MiniMed 670G’s hybrid closed loop algorithm. Based on a DreaMed poster at this conference (see picture below), a follow-up bridging study already has NIH funding and was granted to IDC (Dr. Rich Bergenstal) and Schneider Hospital (Dr. Moshe Phillip) to perform two projects. The study with the new updatd algorithm will be conducted in Israel at Professor Phillip's institute, and Rich will be doing an update to his AGP to incorporate data from the closed loop system. It is unclear when these studies will begin, and we’re still not sure of the FDA submission and approval expectations for this important algorithm update (adding automated correction boluses on top of the MiniMed 670G basal-only modulation). Medtronic confirmed with us at ADA that this product will be called the MiniMed 690G.

• Notably, a Medtronic slide added University of Cambridge to its list of partners, consistent with Dr. Roman Hovorka’s move to use the MiniMed 640G and an Android phone (running the Cambridge algorithm) in upcoming studies. We wonder if this indicates potential licensing down the road – similar to DreaMed – as the Cambridge team still has not disclosed its commercial path to market.

12. Following the initial announcement at ADA, Diabetes Care has published multiple critical papers on AID including “Outcomes Measures for Artificial Pancreas Trials: A Consensus Report” by Dr. David Maahs et. al. (this author list is a “who’s who” of the AID world and we’re extremely impressed by the breadth of the group) and “Design Considerations for Artificial Pancreas Trials” by Dr. Steve Russell and Dr. Roy Beck. These appear in a larger section in this month’s Diabetes Care, “The Artificial Pancreas in 2016: A Digital Treatment Ecosystem for Diabetes.” We salute the ADA into putting so much work into this section. It’s particularly impressive to see the slew of authors on the standardizing outcomes paper – impressive that they could all agree on this! It is so important that metrics are consistent for research, for payers, for patients, for clinicians, and beyond.

13. Moving forward, control algorithm research is very focused on personalization and adaptation, including variable targets and algorithm aggressiveness that moves flexibly with patients over time (e.g., adapting to transient changes in insulin sensitivity). The Bionic Pancreas algorithm was highlighted a few times for only requiring weight to initialize, using qualitative meal boluses, and adapting over time based on its performance. We expect more algorithms to move in this direction if the goal is to make systems easy enough for non-tech savvy providers and patients to use. There was some discussion about using additional inputs to inform closed-loop control (e.g., heart rate for exercise), but the work here is still very early and not particularly efficacious at this stage. 

• Dr. Bill Tamborlane: “In terms of personalizing, some of our best patients in the Medtronic 670G pivotal were extremely frustrated: the system used a 120 mg/dl set point, and they personally wanted to lower the set point and weren’t allowed to. This will be a step for the future – individualize the program to the patient and not have all these rigid rules. It goes to what Aaron Kowalski was saying: the patient often knows more than we do.” From our view, it’s all about presentation: many would argue that a 120 mg/dl set point is a fine set point to use in trials and that it is important to use one – we can very much see the value of standardization and we believe this just reinforces the importance of presentation. If patients are told going into trials that there is a set point that they might not agree with but that is important for the trial standardization, and that they shouldn’t do the trial unless they can adhere to that, there will probably be fewer unhappy patients. We would suggest envisioning before trials what might frustrate patients and putting that into words so that patients just get this information early and so it can inform their decision whether to participate.
• Dr. Claudio Cobelli: “Now that we have long trials, we have a unique opportunity for algorithm design. Personalization is impossible with short trials. Adaptation over 2-4 day trials is superfluous. With a month, you can play with adaptation ... control should be fed by personalization.”

14. Notably, we did not hear a single person today lament CGM’s accuracy or reliability, a testament to how far the technology has come since even the last artificial pancreas workshop in 2013. While sensors can always improve in accuracy and reliability, it is clear that what we have is “good enough” to drive much better outcomes, and they will only get better as calibrations, cost, and size on the body go down, while wear time and accuracy/reliability continue to improve. Dr. Claudio Cobelli also noted that his team is working with Dexcom and is “on its way to reduce the number of calibrations.” This will be an important component in terms of closed loop usability and safety, and we wonder what the plans are beyond Dexcom’s G6 (1 calibration per day, 10-day wear).

-- by Adam Brown, Brian Levine, and Kelly Close