February 18-21, 2015; Paris, France; Day #3 Highlights - Draft

Executive Highlights

Hello from ATTD, where there was no signing of yearbooks to be had today – it’s gone all digital! Today, we bring you our top 10 highlights (along with a few honorable mentions) from a mix of automated insulin delivery, drugs, glucose monitoring, data management, and treatment guidelines. Although in aggregate, this didn’t quite capture the greatness of days #1 and #2, it was plenty great.  

1. As we mentioned to you last night after we got the early scoop, Dr. Moshe Phillip (Schneider Children's Medical Center, Petah Tikvah, Israel) announced that the DREAM consortium has received a CE Mark for its overnight MD-Logic closed-loop algorithm, marking the software as the first closed-loop algorithm to receive regulatory approval. The team has also formed DreaMed Diabetes to commercialize the algorithm. Now, we’ll be waiting for the first system …

2. Dr. Michael Rickels (University of Pennsylvania, Philadelphia, PA) presented a phase 3 study (n=75 adults) comparing Locemia’s exciting intranasal glucagon powder for severe hypoglycemia to standard intramuscular glucagon injection. This was an excellent presentation though we found chairperson questions a bit questionable.

3. In a collective discussion on priorities in technology, Dr. Aaron Kowalski (JDRF, New York, NY) and Mr. David Panzirer (Helmsley Charitable Trust, New York, NY) emphasized the importance of increasing patient access and including the patient’s voice early on. Dr. Kowalski introduced his idea of the diabetes “scorecard” as a method to incorporate patient, provider, and payer voices when assessing different therapies. Patient perspective, as we know, is so integral to adherence and uptake.

4. Dr. Trang Ly (Stanford University, Stanford, CA) presented the first study we’ve seen on the MiniMed 670G hybrid closed loop system using the Enlite 3 sensor. The algorithm was not tuned aggressively enough, though overnight control was excellent and new studies with a modified algorithm are ongoing.

5. Roche’s Dr. Matthias Axel Schweitzer shared in Q&A that the company does not intend (in the short-term) to create open APIs that allow patients to access their own data. He said a regulated market and concerns over patient liability handicap these efforts.

6. A CGM study on Sanofi’s Toujeo (insulin glargine U300) highlighted the formulation’s PK stability vs. Lantus, especially towards the tail end of the 24-hour window. We love seeing CGM studies, which really help us get at the real value.

7. Preliminary results from Dr. Lori Laffel’s (Joslin Diabetes Center, Boston, MA) CGMi trial suggest that the biggest barriers to more CGM use are fundamental aspects of the devices themselves; as Dr. Thomas Danne put it, “the biggest problem with CGM devices is they’re not useful.” This was a bit extreme in our view, but food for thought on future directions given his comments on the success of FreeStyle Libre.

8. Dr. Irl Hirsch (University of Washington, Seattle, WA) advocated for higher A1c targets in patients with longstanding type 1 diabetes and at high risk of hypoglycemia. We think that is so right on though we’d also love to see advocates simultaneously advocate for glycemic dependent therapy where possible.

9. Dr. Guillermo Arreaza-Rubin (NIH, Bethesda, MD) provided a list of NIDDK’s program priorities in diabetes technology. Notably, conversations with the T1D Exchange are ongoing to potentially design reimbursement studies, and his team is working with FDA on guidance.

10. The traditional rapid-fire ATTD Yearbook session provided a comprehensive update on the past year of scientific literature in diabetes drugs and devices; the Yearbook is available free online.

Honorable Mentions: Dr. Roman Hovorka on the Bionic Pancreas and unsupervised home closed-loop; GoCarb, a mobile app for counting carbs with a single photograph; Dr. Jennifer Sherr on using additional hormones in closed loop study.

Top Ten Highlights

1. After our pre-announcement yesterday, Dr. Moshe Phillip (Schneider Children's Medical Center, Petah Tikvah, Israel) announced that the DREAM consortium has received a CE Mark for its overnight MD-Logic closed-loop algorithm, marking the software as the first closed-loop algorithm to receive regulatory approval. The approval is a notable milestone for the closed-loop field as whole, and now the question is when and how patients can get it – and when the first system will receive approval! The “academic” group’s commercialization strategy and timeline remain key unknowns moving forward, though we understand that the consortium recently established a new company, DreaMed Diabetes, for this purpose. The company is presently exploring ways to best leverage the technology, primarily through two primary commercialization strategies: (i) placing the algorithm on a piece of hardware, selling directly to patients, and allowing consumers to connect the software directly to their pump (we’re a bit amazed this would be allowed); or (ii) licensing the algorithm to a pump company, who could integrate it directly (though we wonder whether this would require another CE Mark). We suspect pump companies will much rather have algorithms on pumps rather than Androids … On the US front, the consortium has plans to submit a PMA application to the FDA for the entire system, though timing is unclear, presumably because a system isn’t set. In other positive news from the DREAM consortium, Dr. Revital Nimri presented a host of never-before-seen data, including: (i) the latest interim results from the consortiums three-month overnight study; and (ii) recent findings from a four-day fully closed-loop study. More details below.

2. Dr. Michael Rickels (University of Pennsylvania, Philadelphia, PA) presented a phase 3 study (n=75 adults) comparing Locemia’s exciting intranasal needle-free glucagon powder for severe hypoglycemia to standard intramuscular glucagon injection. The randomized, crossover trial at eight T1D Exchange clinics administered IV insulin until blood glucose <60 mg/dl. Five minutes after stopping insulin (mean nadir was ~46 mg/dl), glucagon was given – a 3 mg intranasal dose at one visit and a 1 mg intramuscular dose at the other visit. Notably, Locemia’s intranasal glucagon was non-inferior to intramuscular injection: all subjects responded with an increase in blood glucose, although 1% of the Locemia arm and 0% of the intramuscular arm failed to achieve the primary outcome (an increase in blood glucose to >70 mg/dl OR an increase >20 mg/dl from nadir within 30 minutes post-administration – this seemed like an unconventional endpoint, but it was agreed upon with the FDA). Median time to achieve the primary outcome was 10 minutes for intramuscular injection vs. 15 minutes for intranasal delivery. The five minute delay in pharmacodynamics for intranasal delivery was fairly inconsequential in our view, given the time it takes to prepare a traditional glucagon kit vs. the needle-free, single-button push delivery of the Locemia device (see a picture on Twitter here). Nausea and vomiting occurred at a similar frequency, though transient head/facial discomfort were more common with intranasal glucagon (25% vs. 9%) – not a big issue given the circumstances under which this would be given. The Locemia team is doing other studies, including a usability study (n=200) and pediatric study. We assume a launch could happen as early as late 2016 or early 2017, assuming regulatory goes well. This technology could be a very big deal for patients because it would calm down caregivers – the current glucagon solutions from Novo Nordisk and Lilly are hard to use and basically both represent a major overdose. Nausea after use isn’t uncommon, not are super high blood glucose scores (which often starts the entire cycle all over again of high blood glucose leading to low blood glucose). To say the current systems are under-optimized is fair – it will be interesting to see what happens here as we believe the market could be far bigger than it is today with a better product.

3. In a discussion on priorities in technology, Dr. Aaron Kowalski (JDRF, New York, NY) and Mr. David Panzirer (Helmsley Charitable Trust, New York, NY) emphasized the importance of increasing patient access and including the patient’s voice early on. Dr. Kowalski highlighted that the ultimate success of the artificial pancreas is making sure people with diabetes achieve better outcomes, which is accomplished through expanding patient access. He emphasized that moving forward, JDRF’s funding will focus on “breaking barriers” with regulatory bodies, payers, and clinicians so that the artificial pancreas can become a less costly system available to more people. This is of course also a sign that artificial pancreas technology has adequately progressed to shift major focus to issues of reimbursement and access. Following Dr. Kowalski, Mr. Panzirer noted that the Helmsley Charitable Trust (HCT) shares JDRF’s vision and that “the single most important thing that needs to be worked on is including the voice of the patient from day one,” citing the relatively low penetration of insulin pumps and CGMs. In addition, he emphasized that primary care providers must be better educated to treat type 1 diabetes, as the majority of people with diabetes do not have access to an endocrinologist. It was valuable to hear both Dr. Kowalski and Mr. Panzirer on the same page, as the HCT and JDRF represent two of the most influential funders and voices in type 1 diabetes.

• During Q&A, Dr. Kowalski introduced his idea of the diabetes “scorecard” as a method to incorporate patient, provider, and payer voices when assessing different therapies. He commented that the field focuses too much on A1c, while burden and cost are often disregarded early on in the development of a therapy. Therefore, his scorecard evaluates therapies based on three factors – glycemic control, burden, and cost – which would be assessed by patients, providers, and payers. This scorecard’s potential to help consider different stakeholder’s views, according to Dr. Kowalski, could help provide guidance in driving the field forward. Certainly, it sounds quite aligned with what the Bigfoot Biomedical team is working on, and we look forward to a more formal document summarizing the scorecard.

4. Dr. Trang Ly (Stanford University, Stanford, CA) presented the first study we’ve seen on the MiniMed 670G hybrid closed loop system using the Enlite 3 sensor. The seven-day camp study randomized 20 patients to either hybrid closed loop (MiniMed 670G – patients still bolus for meals) or threshold suspend (MiniMed 530G). Surprisingly, there were no statistically significant differences in any of the glycemic parameters between the groups. This was in part due to remarkably excellent control in the 530G group (73% time in 70-180 mg/dl vs. 70% on the 670G), perhaps reflective of camp care – we were just pointing out last night at a conversation at dinner why we don’t think RCT are necessarily always the most valuable thing to look at – necessary but not sufficient, I guess we’d conclude. Dr. Ly did not specifically show the time in range by time of day, but said the overnight control with the 670G was “excellent.” One issue seems to be the aggressiveness of the algorithm, which we believe may be tightened for future studies – it makes sense due to safety, of course, to have this be conservative upfront. The Enlite 3 sensor came in with a MARD of 12.6% vs. fingersticks and a median ARD of 9.7% (n=742). Overall, the integrated system seems very seamless and low burden (algorithm built into the pump + Enlite 3 CGM; picture here) – participants were in closed-loop for 93% of the time, and sensor values were obtained 99% of the time. Connectivity has been a major Achilles heel in artificial pancreas research, so it’s exciting to see Medtronic making improvements on this front. The algorithm sounds like a work in progress at this point, and slated improvements include additional adaptability, allowing patients to enter their own carb:insulin ratio, and more. In a later presentation, we learned that there is an ongoing 12-day study of the MiniMed 670G. More system details below.

5. Roche’s Dr. Matthias Axel Schweitzer shared in Q&A that the company does not intend (in the short-term) to create open APIs that allow patients to access their own data. Dr. Schweitzer acknowledged that this needs to happen in the future, but that both a regulated market and concerns over patient liability handicap these efforts. Indeed, he noted that Roche is instead focused on delivering “safe and proven” solutions and, in redirecting Brandon Arbiter’s (Tidepool) question, he painted a grey picture for fans of data liberation, connectivity, and open-source solutions, commenting that open formats are “a long, long ways off.” We will be interested to see where this goes; there has certainly been significant value that has come from putting data in patient hands and allowing patients to solve real-world problems as seen in the success of NightScout – that said, now that NightScout is more “official” we also don’t know where service, liability, etc. will land. Horizontal systems allow for clinically relevant innovation and scaling that is hard to achieve in the current, closed proprietary medical-device-manufacturer-as-software-developer model. While we are optimistic about all the work happening on interoperability and while some would note Roche and Medtronic are a couple of the only remaining companies that have not partnered with Tidepool, they certainly are among the biggest.

6. We saw some encouraging results on Sanofi’s new basal insulin contender Toujeo (insulin glargine U300), including CGM data on the new formulation’s PK stability. The multicenter phase 2 CGM study enrolled 59 type 1 diabetes patients who were randomized to Toujeo or Lantus and crossed over between morning and evening injections, wearing CGM for the entire duration of the study (this designed really enabled others to assess the value in a way that A1c alone just doesn’t allow). During the final two weeks of each crossover period (when the insulin dose was fixed) there was no significant difference between time-in-range (80-140 mg/dl) between the Toujeo and Lantus groups, although there did appear to be a small trend in Toujeo’s favor. However, an analysis of the last four hours pre-injection showed less than half the change in glucose with Toujeo vs. Lantus. CGM traces included in the slides really showed the difference in intra-day PK stability with Toujeo vs. Lantus, especially with morning dosing. Toujeo also showed a 5%-15% improvement in various standard deviation-based measurements of glycemic variability over Lantus. The CGM data definitely showed some benefit with Toujeo over Lantus, though and particularly in the early weeks when there is more stability, this is important since that is a common week for patients to otherwise decide to drop therapy (this was the time there appeared to be less nighttime hypoglycemia). Big picture, we were encouraged to see CGM used in an insulin trial and hope more manufacturers do such studies in the future as it makes it easier to assess results. We look forward to seeing more qualitative data as we believe this also has a major impact on adherence.  

• We also saw full one-year data from the phase 3 EDITION I trial – these results were presented as posters at ADA and EASD last year but this was the first time we were seeing them on the big screen. The six-month primary results found equal A1c reductions of 0.83% from baseline (8.2%) with both formulations and a reduction in hypoglycemia at some time points with Toujeo. We were intrigued to see that in the second six months, Lantus’ efficacy began to wane very slightly while Toujeo’s held firm, leading to a modest but statistically significant 0.17% difference between the arms (p=0.007); mean fasting plasma glucose levels and weight gain at the end of the trial were not significantly different between groups. There was either a significant decrease or a non-significant beneficial trend in hypoglycemia, depending on how hypoglycemia was defined – it would’ve been terrific to see CGM in the full trial so time in hypoglycemia could be assessed.

7. Dr. Lori Laffel (Joslin Diabetes Center, Boston, MA) shared preliminary results from the CGMi trial that is evaluating barriers to greater CGM use in pediatric patients. The 24-month randomized controlled trial is comparing the effectiveness of a “family-focused behavioral teamwork intervention” vs. standard education and support in the initiation of CGM in youth age 8-17 with type 1 diabetes. The primary endpoint is change in A1c at 12 months. The trial is ongoing with an estimated completion date of April 2015, but Dr. Laffel presented preliminary data on CGM use during the first year. Notably, 72% of patients found to be eligible during screening declined to participate in the trial, and half of them cited reluctance to use CGM as their specific reason for declining – quite a sobering indication of the barriers to greater uptake. On a slightly brighter note, out of the 130 patients who did enroll, 70% were still using CGM after 12 months. In an attempt to identify the factors underlying the high rates of CGM discontinuation (or “dis-continuous glucose monitoring,” as Dr. Laffel put it) among youth, the investigators asked the families of all enrolled patients to identify the key challenges they had experienced with the device. The three factors cited much more frequently by the group that discontinued CGM were difficulty managing glucose alarms, wearing the sensor, and carrying the receiver – in other words, fundamental aspects of the device (perhaps THE most fundamental!) rather than technical glitches.

• Dr. Thomas Danne (Diabetes Center for Children and Adolescents, Hannover, Germany) echoed similar themes during a subsequent panel discussion, saying “the biggest problem with CGM devices is they’re not useful.” He noted that the enormous early success of Abbott’s FreeStyle Libre supports the theory that hassles like false alarms and calibration are likely the main barrier to greater use of CGM. He positioned future closed loop systems as the best solution to these problems, urging the audience to “work on FDA to get big studies in the American market so we finally bring less burden to patients.” We agree that Libre’s incredibly positive reception (read more about it here) has been an excellent illustration of the value of reduced hassle factor. The big question is if closed-loop devices will be the killer app for CGM and pumps; some key barriers (per our day #2 report) include price and a perceived loss of control over diabetes.

8. Dr. Irl Hirsch (University of Washington, Seattle, WA) advocated for higher A1c targets in patients with longstanding type 1 diabetes and a high risk of hypoglycemia. He explained that while there is certainly data suggesting benefits from intensive glycemic control, the evidence is somewhat mixed – the relationship between tight control and reduced mortality is not as clear as many had predicted, and the falling rates of microvascular complications in recent decades may be a result of more effective insulin regimens (i.e., routine use of mealtime insulin and CGM) as well as lower A1c targets. He also discussed the limitations of A1c as the sole indication of glycemic health, noting that ~14% of patients in his practice have misleading A1c values due to various confounding conditions.

• Given these limitations, Dr. Hirsch argued that there is scant evidence to support extremely strict control for patients with a duration of diabetes >20 years and that the risk of hypoglycemia should not be ignored; for example, he cited data showing a ~20% risk of severe hypoglycemia (defined as seizure or coma) per year for patients with a duration of diabetes >20 years. Dr. Hirsch noted that he spends more time persuading older patients with hypoglycemia unawareness and low A1cs to set higher glycemic targets than he does trying to implement more intensive control. We certainly agree that the risks of hypoglycemia should not be underestimated and that less stringent targets may be appropriate for some patients. However, we do wonder whether greater access to and use of CGM in this population could allow more patients to achieve tighter targets without undue risk of hypoglycemia – come on CMS! We also think patient choice should be part of this – while DCCT didn’t enroll many people over 65, we also haven’t seen data showing no correlation between tight A1cs and complications. Due to the young age at which many get type 1, there are heaps of patients with duration of diabetes over 20 years and we suspect many would not necessarily want to just take a higher target for the same reason – there isn’t data showing that a higher target will serve them well (and as noted, there are other ways to avoid severe hypoglycemia if a person has access). We do acknowledge if cost is a problem, and if using more low-cost glucose strips and testing more often isn’t possible, a higher target sounds reasonable (even though some patients may not want them). Dr. Robert Gabbay noted in the same session that risk of hypoglycemia (rather than history of hypoglycemia) should influence A1c targets, and use of CGM (which can significantly reduce that risk) is not incorporated into most guidelines on setting individualized goals, which is a shame from our perspective.

9. Program Director Dr. Guillermo Arreaza-Rubin (NIH/NIDDK, Bethesda, MD) provided a list of the NIDDK’s program priorities in diabetes technology, which included artificial pancreas, human islets, bio-artificial/encapsulation strategies, and the imaging of beta cells. Dr. Arreaza-Rubin highlighted that the NIDDK is focusing on technologies to support the newly established Human Islets Research Network (HIRN) in order to develop strategies to protect and replace functional beta cell mass. In addition, the program is researching new technologies for the advancement of cell replacement research such as bio-artificial and encapsulation, of which Dr. Arreaza-Rubin noted NIDDK is working to complement HCT and JDRF’s work in this area. He also highlighted that significant progress has been made in the imaging work of beta-cells in better understanding the cells’ natural history, inflammation, and engraftment. Other priorities included developing new technologies for complications and assessment of risk (for earlier diagnosis and prognosis of type 1 diabetes) as well as research into automated-mechanical hormone replacement therapies such as artificial pancreas systems – we’d note that automated insulin delivery was not a major focus of this talk, though it’s clearly on the NIH’s map in a big way. Notably, NIDDK has been collaborating with the FDA, specifically in discussing recently published guidance (for low glucose suspend and artificial pancreas systems) as well as in participating in collective workshops.

• Dr. Arreaza-Rubin also highlighted during Q&A that he has been in conversations with Mr. Dana Ball (Execute Director, T1D Exchange & CEO, Unitio) to design studies that can better convince payers for reimbursement. In discussing the challenges with the CMS on various devices, Dr. Arreaza-Rubin pointed out the importance of garnering solid evidence appropriate for the Agency. He emphasized that this is essential in geriatric populations, as CGMs lose coverage in Medicare. For more on reimbursement and regulatory struggles in diabetes technology, see our ATTD Day #1 highlights.

10. At the most well attended session of the day, a superstar set of speakers gave rapid-fire presentations to unveil the latest ATTD Yearbook. Covering updates from 2014 for a wide variety of fields across technology, drugs, and more, the Yearbook is now also available for free online on the Diabetes Technology & Therapeutics website. The Yearbook session is always very interesting to hear, although we cannot imagine how difficult it is to design a presentation on a given topic area to fit into a six minute span. We hand it to ATTD’s organizers for putting together a session that treats attendees to quick and efficient updates from a wide variety of fields within diabetes. By the way, for our take on everything that happened in 2014 and what we expect to see this year, see our 2014+2015 Reflections piece!

Honorable Mention

1. Dr. Roman Hovorka (University of Cambridge, Cambridge, UK) summarized 860 nights of unsupervised closed-loop at home from three published studies, reprised his comparison to the Bionic Pancreas’ Beacon Hill study from DTM 2014, and mentioned Cambridge’s four ongoing unsupervised home studies. The latter include two three-month trials and two 1-4 week trials – by March, the team will have an impressive 12 years of data on unsupervised home closed loop! Speaking more generally, Dr. Hovorka stressed the need for adaptive systems that can cope with variable insulin needs (his graph showing the variability in nightly insulin delivery – from 50% to 200% of the pre-programmed basal – “shows why we need closed-loop”); highlighted the need for intention to treat analyses in closed loop studies; and asked for more properly designed studies to compare the benefits of various approaches. We couldn’t agree more. In Q&A, Dr. Peter Chase wondered whether the first commercial systems should only target the night, since the outcomes are so much better. Dr. Hovorka agreed, though said the regulatory bodies are struggling with how to define “nights”; as a result, commercial products will likely need to show they work during the day too. This seems to be the approach Medtronic is taking with the MiniMed 670G: a hybrid closed loop system that will work 24/7, but has the biggest bang for the buck at night. Indeed, Medtronic’s slide yesterday called it “overnight and hybrid closed-loop.” We think this is all about giving patients choice – those who don’t need/want the daytime closed-loop can turn it off; we imagine a significant percentage of type 1s could benefit from the automated nighttime control.

2. Dr. Peter Diem (Bern University Hospital, Switzerland) presented a valuable update on GoCarb, a mobile platform that supports individuals with type 1 diabetes by automatically estimating the grams of carbohydrate in a meal (in near real-time) based on a single photograph. Such a carbohydrate counting system is certainly a patient’s dream, and while the work is still early stage, this system could be the real deal. Indeed, the app moved into the outpatient setting last year and has seen promising results. Dr. Diem shared those findings, which were collected over a nine-day period in patients with type 1 diabetes (n=19) at Bern University Hospital. Patients were given six meals per day and asked to estimate the carbs in each meal on their own before using GoCarb. Results indicated that the mobile platform was significant better than individuals at predicting carbohydrate content (absolute error: 13 grams vs. 28 grams), and 80% of the time, GoCarb error came in at under 20 grams (clinically accurate, according to Dr. Diem). A key question is whether the system will be able to scale to a variety of foods and mixed meals, as we cannot imagine that the selection at the hospital restaurant was particularly diverse. The team’s second clinical trial is scheduled for August 2015. It is great to see movement on this front, since carb counting is incredibly challenging – even for experts – and an effective mobile app would be a huge win.

3. Dr. Jennifer Sherr (Yale, New Haven, CT) gave an interesting talk on using additional hormones in closed loop study, emphasizing that even using the closed loop, exaggerated postprandial excursions remain a problem. Of interest, she discussed a small (n=8) closed loop study in which Novo Nordisk’s liraglutide (GLP-1) was used. Liraglutide was given as a once daily injection in addition to closed loop insulin delivery. In this study, glycemic excursions declined at each meal, insulin dose declined, as did weight. Most study participants had nausea and about half had headaches, though no one quit the study. Overall, the study supported the hypothesis that liraglutide will be an effective adjunct to both open and closed loop therapy in type 1 diabetes. Dr. Sherr also discussed a previous study using pramlintide (Symlin) and closed loop. Patients given pramlintide with closed-loop insulin delivery prompted reductions in glycemic excursions and reduced delays in gastric emptying. Less insulin was also required. We are excited to see more studies using additional hormones in addition to insulin – this is an exciting study to see presented!

Detailed Discussion and Commentary

Oral Presentations

Intranasal Glucagon For Treatment Of Insulin-Induced Hypoglycemia In Adults With Type 1 Diabetes: A Randomized, Cross-Over Non-Inferiority Study

Michael Rickels, MD (University of Pennsylvania, Philadelphia, PA)

Dr. Michael Rickels presented a phase 3 study (n=75 adults) comparing Locemia’s exciting intranasal needle-free glucagon powder for severe hypoglycemia to standard intramuscular glucagon injection. The randomized, crossover trial at eight T1D Exchange clinics administered IV insulin until blood glucose <60 mg/dl. Five minutes after stopping insulin (mean nadir was ~46 mg/dl), glucagon was given – a 3 mg intranasal dose at one visit and a 1 mg intramuscular dose at the other visit. Notably, Locemia’s intranasal glucagon was non-inferior to intramuscular injection: all subjects responded with an increase in blood glucose, although 1% of the Locemia arm and 0% of the intramuscular arm failed to achieve the primary outcome (an increase in blood glucose to >70 mg/dl OR an increase >20 mg/dl from nadir within 30 minutes post-administration – this seemed like an unconventional endpoint, but it was agreed upon with the FDA). Median time to achieve the primary outcome was 10 minutes for intramuscular injection vs. 15 minutes for intranasal delivery. The five minute delay in pharmacodynamics for intranasal delivery was fairly inconsequential in our view, given the time it takes to prepare a traditional glucagon kit vs. the needle-free, single-button push delivery of the Locemia device (see a picture on Twitter here). Nausea and vomiting occurred at a similar frequency, though transient head/facial discomfort were more common with intranasal glucagon (25% vs. 9%) – not a major issue given the circumstances under which this would be given. The Locemia team is doing other studies, including a usability study (n=200) and pediatric study.

• This presentation represented the first public display we’ve seen of Locemia’s needle-free, intranasal delivery device for severe hypoglycemia – it looks like a dramatic improvement over existing glucagon kits. The plastic device fits in the palm of the hand and contains a dry powder formulation. To administer the glucagon intranasally requires one step (to be performed by a caregiver in treating an episode of severe hypoglycemia): depress the plunger into the device chamber. That single step compresses the air in the device and dispenses the powder into the nose (similar to nasal spray for a cold). The device contains a 3 mg dose of glucagon, higher than the 1 mg dose in a rescue kit. 
• Locemia is furthest along on the improved glucagon rescue delivery front – both Xeris (auto-injector) and Biodel (auto-reconstitution pen) have yet to begin their phase 3 studies. We assume Locemia could use the 505(b)(2) regulatory pathway, which means the company could potentially come to market as early as late 2016 or early 2017. We do wonder about partnerships and what commercialization will look like.

Hybrid Closed-Loop Control Using The Medtronic 670G And Enlite 3 System In Type 1 Diabetes At Diabetes Camp

Trang Ly, MD (Stanford University, Stanford, CA)

Dr. Trang Ly presented the first study we’ve seen on the MiniMed 670G hybrid closed loop system using the Enlite 3 sensor. The seven-day camp study randomized 20 patients to either hybrid closed loop (MiniMed 670G – patient still bolus for meals) or threshold suspend (MiniMed 530G). Disappointingly, there were no statistically significant differences in any of the glycemic parameters between the groups. This was in part due to remarkably excellent control in the 530G group (73% time in 70-180 mg/dl vs. 70% on the 670G), perhaps reflective of camp care. Dr. Ly did not specifically show the time in range by time of day, but said the overnight control with the 670G was “excellent.” The issue seems to be the aggressiveness of the algorithm, which is being improved for future studies. The Enlite 3 sensor came in with a MARD of 12.6% vs. fingersticks and a median ARD of 9.7% (n=742). The integrated system seems very seamless and low burden (algorithm built into the pump + Enlite 3 CGM; picture here) – participants were in closed-loop for 93% of the time, and sensor values were obtained 99% of the time. Connectivity has been a major Achilles heel in artificial pancreas research, so it’s great to see Medtronic making improvements on this front. The algorithm sounds like a work in progress at this point, and slated improvements include additional adaptability, allowing patients to enter their own carb:insulin ratio, and more. In a later presentation, we learned that there is an ongoing 12-day study of the MiniMed 670G.

• The MiniMed 670G includes a hybrid closed loop algorithm fully integrated into the pump. It uses Medtronic’s fourth-generation sensor (which we have been calling the Enlite 3 though we note the nomenclature seems to vary globally). The pump can operate as a standalone pump with pre-programmed parameters. If CGM information is available, it can also run predictive low glucose suspend (a la the MiniMed 640G). Last, it can run hybrid closed loop, where patients bolus for meals and the pump takes care of things in the background.
• Medtronic is using a PID-based control algorithm with insulin feedback. It targets a set point of 120 mg/dl. The 670G algorithm doses every five minutes and cannot deliver more than a patient-specific maximum insulin limit (e.g., 2.3 units per hour). The algorithm has some adaptability –the major component is in the max insulin limit. There is some adaptability with carb:insulin ratio and ISF as the median total daily dose changes from the preceding six days; however, the effect of that overall on the algorithm is minimal. The carb:insulin ratio is algorithm derived. The system performs automatic corrections when a meter glucose value is above 200 mg/dl.

Plenary: Closing the Loop – Home Studies

Integration of the MD-Logic system into daily management of diabetes

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

Dr. Nimri presented a host of never-before-seen data, including the latest interim results from the consortium’s three-month overnight study and recent findings from a four-day fully closed-loop study.

• Interim results from DREAM’s randomized three-month overnight study look promising. The trial is evaluating the safety and efficacy of long-term overnight home use of MD-Logic vs. sensor-augment pump (SAP) therapy. The study has reported data from 12/36 patients (six on closed-loop; six on SAP). Patients on MD-Logic have spent 63% time-in-range, with only 1.3% of the time in hypoglycemia (< 60 mg/dl); patients in the control arm have spent only 43% time-in-range despite using more insulin (10.3 units vs. 7.9 units), with 3.4% of the time in hypoglycemia. So far, the data looks largely consistent with previous overnight findings, something the full results from other centers will confirm.
• It is also outstanding to see that the consortium has completed its first day-night closed-loop study. This “over-the-weekend,” four-day trial was new to us and investigated fully closed-loop control for 60 hours compared to SAP at three centers across Europe. The crossover design consisted of a three-week run-in period on SAP before randomizing patients to 60 hours of closed-loop or SAP therapy. Average glucose in the closed-loop arm was 144 mg/dl, with 68% of the time spent in range (70-180 mg/dl). In the SAP arm, average glucose was a bit higher at 159 mg/dl, with 58% of time spent in range. Patients in neither arm experienced much hypoglycemia (0.9% CL vs. 0.8% SAP). We would highlight that these are motivated patients in excellent control to begin with, and MD-Logic was able to bring them safely into even tighter control. On the insulin front, patients on closed-loop used significantly less total basal insulin (54 units CL vs. 62 units SAP) and significantly more bolus insulin (64 units CL vs. 47 units SAP) than those on SAP, though it was notable to see that total daily dose was not significantly different between the two arms.

Update on home closed loop studies and factors affecting outcomes

Roman Hovorka, PhD (University of Cambridge, Cambridge, UK)

Dr. Roman Hovorka summarized 860 nights of unsupervised closed-loop at home from three published studies, reprised his comparison to the Bionic Pancreas’ Beacon Hill study from DTM 2014, and mentioned Cambridge’s four ongoing unsupervised home studies. The latter include two three-month trials and two 1-4 week trials – by March, the team will have an impressive 12 years of data on unsupervised home closed loop! Speaking more generally, Dr. Hovorka stressed the need for adaptive systems that can cope with variable insulin needs (his graph showing the variability in nightly insulin delivery – from 50% to 200% of the pre-programmed basal – “shows why we need closed-loop”); highlighted the need for intention to treat analyses in closed loop studies; and asked for more properly designed studies to compare the benefits of various approaches. We couldn’t agree more. In Q&A, Dr. Peter Chase wondered whether the first commercial systems should only target the night, since the outcomes are so much better. Dr. Hovorka agreed, though said the regulatory bodies are struggling with how to define “nights”; as a result, many believe commercial products will likely need to show they work during the day too. We think the expectations would be easier to meet with a night-only system first though we understand the FDA has asked organizations to go with 24/7. As such, this seems to be the approach Medtronic is taking with the MiniMed 670G: a hybrid closed loop system that will work 24/7, but has the biggest bang for the buck at night. Indeed, Medtronic’s slide earlier in the meeting called it “overnight and hybrid closed-loop.” We think this is all about giving patients choice – those who don’t need/want the daytime closed-loop can turn it off; we imagine most everyone would like and benefit from the nighttime control. That said – we doubt many will turn off the daytime though they will be more critical of it (stemming from “disturbances” like food and exercise and stress!) – that’s why we would rather the night alone be first, so that patients can see the power of it without the “interruptions” and variability of real life. We also think patients may actually eat less responsibly, etc., if they have a closed loop, which will not be helpful in generating good early data.

• Dr. Hovorka pooled data from Cambridge’s two recently published, four-week, overnight home studies of unsupervised closed-loop (Hovorka et al., Diabetes Care 2014; Thabit et al., Lancet Diabetes Endocrinol 2014). The randomized, crossover trials included 24 adults and 16 adolescents and compared CGM alone for 28 days to closed-loop for 28 days – a total of 860 nights of closed-loop were obtain. In the intention to treat analysis (“I really stress that”) mean time in target (70-144 mg/dl) increased from 41% to 59% (p<0.001), and time <70 mg/dl decreased from 2.9% to 1.9% (p=0.014). He noted that 2.9% on control nights translates to just 10 minutes of hypoglycemia, a “very low” number compared to other studies. Mean glucose improved from 157 mg/dl to 142 mg/dl.  
• In summarizing the team’s seven-day day+night home studies (Leelaranthna et al., Diabetes Care 2014), Dr. Hovorka emphasized that most of the improvement came during the night. “The night is the place where closed loop can do best.” There was some improvement during the day, but it was not as pronounced. ON the Cambridge system, patients bolus for meals during the day. As is now the norm with home studies, mean glucose improved (158 to 146 mg/dl), time in 70-180 mg/dl improved (64% to 73%), and patients spent less time below 70 mg/dl (4.3% to 3.1%). This came with 14% less insulin.
• As he did at DTM 2014, Dr. Hovorka directly compared the Beacon Hill Bionic Pancreas trial (Russell et al., NEJM 2014) and his group’s recently published day-night study (Leelarathna et al., Diabetes Care 2014). He expressed concern that the Boston researchers are not only overly aggressive in their use of glucagon, but that insulin-only approaches can provide just as good control as using both hormones. We thought his analysis was thought-provoking, though altogether difficult to unpack, given that the trials were different.
  
  • In directly comparing Beacon Hill with his team’s work, Dr. Hovorka suggested that insulin-only systems may provide similar control to insulin+glucagon systems. He did acknowledge that the Bionic Pancreas achieved greater reductions in mean blood glucose from the same baseline of 159 mg/dl (133 vs. 146 mg/dl) and greater time in range of 70-180 mg/dl (80 vs. 73%) than those seen in the Cambridge study; however, he stressed that patients on the Bionic Pancreas also spent slightly more time in hypoglycemia than those on insulin-only closed-loop (4.1% vs. 3.1%).
  • Dr. Hovorka also asserted that the comparison to the control group in the Cambridge study was more challenging than in Beacon Hill. First, the Cambridge team used an intention-to-treat analysis (whether closed-loop was functioning or not), while Beacon Hill only considered time when closed loop was operational. Second, the control group in the Cambridge study had real-time CGM always available, while only ~50% of patients in Beacon Hill were on real-time CGM (i.e., patients who did not normally wear a real-time CGM to manage their diabetes were not given one during the usual care arm).
  • Dr. Hovorka noted that in Beacon Hill, participants had a 32% increase in their daily insulin dose during closed-loop. He implied that the increase results from a “stop-and-go” approach to hormone administration that, in his eyes, aggressively doses insulin and depends on glucagon to counteract that “excessive” administration. Dr. Hovorka also pointed to glucagon administration in these patients that exceeded physiological levels (0.8 mg per day, almost a rescue dose).
  • The Bionic Pancreas (BP) did not deliver more insulin than usual care (UC) in three of the team’s four completed outpatient studies (when there was perfect parity between the control arm and experimental group): 2013 summer camp study [0.79 units/kg/day UC vs. 0.82 units/kg/day BP]; 2014 summer camp study [0.68 units/kg/day UC vs. 0.66 units/kg/day BP]; and multicenter study preliminary data in 20 patients [0.69 units/kg/day UC vs. 0.70 units/kg/day BP]. The one exception to this trend was the Beacon Hill study, where the Bionic Pancreas delivered more insulin than in the Usual Care arm. The greater insulin use came in the 11 participants whose mean glucose was >154 mg/dl during usual care, not for the nine participants whose mean glucose was < 154 mg/dl during usual care).
  • In a follow-up conversation at DTM 2014, Dr. Damiano shared two reasons why the bionic pancreas may have delivered more insulin in Beacon Hill: (i) the 11 participants that used more insulin on the bionic pancreas than in usual care may have been underinsulinized under usual care, since all 11 of those subjects had glucose levels >154 mg/dl under their own care; and (ii) different participant behavior on the bionic pancreas (e.g., eating out at restaurants every day vs. usual care eating at home). Indeed, the team has anecdotal evidence that Beacon Hill participants took significant liberties with regard to carbohydrate consumption on the Bionic Pancreas relative to what they did on their own care at home – the participants had been urged to “stress the system,” which is different from everyday care. Even still, insulin utilization in the comparator arm for those subjects who achieved mean glucose levels below 154 mg/dl on their own care were not statistically significantly different from the amount of insulin they received on the bionic pancreas.
• While Dr. Hovorka’s perspectives were thought-provoking, the studies are difficult to compare. Specifics can vary drastically from study-to-study, which makes it challenging to put the stats up side by side and draw conclusions. However, it was another reminder that the value of glucagon continues to be a controversy in artificial pancreas development.

Questions and Answers

Dr. Yogish Kudva (Mayo Clinic, Rochester, MN): In your four studies, are you using adaptive closed-loop, and is that only during basal? How are you handling meals?

A: For meals, we use the standard bolus wizard for meal delivery. We haven’t changed that. We do use more adaptive behavior, as people’s insulin needs change week to week. People who go on skiing holiday, insulin delivery reduced, especially in Austria. One needs to be really responsive to these changes.

Dr. Eda Cengiz (Yale University, New Haven, CT): You showed slide with the percentage of insulin delivery compared to the pre-programmed basal. Did you see associations with glycemic variability, age, C-peptide, or gender?

A: We didn’t look at C-peptide. We tried to look at age and anything else to explain the variability between people. It’s something that we would like to predict – higher variability in insulin needs.

Dr. Peter Chase (Barbara Davis Center, Denver, CO): This may be heresy, but in the old days two years ago, we talked about closed loop in stages, with the night first. With night working so much better, as we go commercial, would we be better to start with nights first?

A: I think we do better on the nights. I think the issue with the regulatory bodies is how to define nights. For commercial products that are night only, it sounds like the FDA still wants you to show how it works in the day. My feeling is that people like it during the day, but it just doesn’t provide as much benefit.

Q: The system purely controls basal? It doesn’t take un-announcement of meals or physical activity into account?

A: The system has the ability to indicate exercise, and the system will take into account. There is a feature in the algorithm that reduces insulin delivery. With the bolus calculator, the system also learns and gets information from the pump to adjust insulin delivery high or low. We are currently doing sub-optimally controlled subjects, and not everyone boluses for meals. We have standard behavior. The system needs to cope with this. People have a life to live.

-- by Melissa An, Adam Brown, Varun Iyengar, Emily Regier, Manu Venkat, and Kelly Close