February 18-21, 2015; Paris, France; Day #4 Highlights - Draft

Executive Highlights

We have bid au revoir to Paris, and with it, the final learning-packed half-day of ATTD 2015. In just four hours, the conference closed out with T1D Exchange data, loads of learning on automated insulin delivery, commentary on NightScout, new data on Medtronic’s Harmony sensor, and much more. Please see below for our top highlights of day #4 along with some exhibit hall coverage.

1. Dr. Desmond Schatz (University of Florida, Gainesville, FL) provided a comprehensive overview of pump/CGM use in the US pediatric population, courtesy of the T1D Exchange. As of January 2015, fewer than 20% of US children are meeting ADA guidelines for A1c <7.5%. Depressing, indeed.

2. A valuable session on UVA’s DiAs artificial pancreas system provided a whirlwind update on past studies and the exciting 2015 roadmap (including an 11-month trial!). In a notable conversation with us outside the session, UVA’s Dr. Boris Kovatchev shared much more about the team’s approach to regulatory and commercialization.

3. Ms. Stephanie DuBose (Jaeb Center for Health Research, Tampa, FL) presented data from the T1D Exchange suggesting that hypoglycemia unawareness and glycemic variability, not lower A1c, are the main factors underlying the high rates of severe hypoglycemia in older adults with type 1 diabetes. An upcoming study will investigate whether CGM could reduce this risk – we can’t imagine that it wouldn’t, especially with more accurate sensors.

4. Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA) provided an update on the Diabetes Technology Society’s Surveillance Program for Cleared Blood Glucose Monitors and announced a new DTS-sponsored program to develop a cybersecurity standard for connected medical technologies. Of note to #wearenotwaiting watchers, he gave some commentary on Nightscout.

5. New data on Medtronic’s redundant electrochemical Harmony sensor showed a 9% MARD vs. YSI and 10% vs. BGM in a week-long study; the system can weight the signals from the two sensors differently in case one is determined to be off-target.

6. In the afternoon we heard a positive perspective on Afrezza’s clinical profile, including its hypoglycemia benefit and potential as a closed loop adjunct, but more mixed thoughts on the challenging path towards oral insulin.

7. Dr. Roy Harper (South Eastern Health and Social Care Trust, Dundonald, UK) presented positive results from a pilot evaluation (n=126) of Hygieia’s d-Nav device: a 1.7% A1c reduction after six months (baseline: 9.2%). d-Nav is an insulin titration device with built in BGM capability.

8. Dr. Neal Kaufman (DPS Health, Los Angeles, CA) stressed that randomized controlled trials are a particularly poor way to examine digital telehealth interventions in diabetes, as the technology under investigation is often obsolete by the time the study reports results.

9. Dr. Stephanie Amiel (King’s College London, London, UK) delivered a well-received talk about a potential neural link between insulin resistance and altered food-related behavior.

10. We highlight exhibit hall happenings from Abbott, Bayer, Cellnovo, CeQur, Dexcom, Medtronic, and Roche.

Top Ten Highlights

1. Dr. Desmond Schatz (University of Florida, Gainesville, FL) provided a comprehensive overview of “where we are” with pump and CGM use in the US pediatric population. Dr. Schatz prefaced his talk by putting the disappointing status of glycemic control in this population in context by sharing updated numbers from the T1D Exchange. As of January 2015, fewer than one in five US children is meeting ADA guidelines for glycemic control (A1c <7.5%) despite an average annual income of ~$80,000/year in this population. Even in the best group, 26-49 year olds, a mere 31% are at goal – and these are the best centers in America! Given the technology available, Dr. Schatz stressed that the dismal numbers are in part a reflection of the underutilization of pumps and CGM. However, he did note that penetration is on a promising trajectory, based on recent (new to us) data from the T1D Exchange. Particularly in pediatric patients <13 years old, Dr. Schatz showed that pump use has risen substantially over the past five years, increasing from 43% to 62% in those <6 years and 54% to 65% in those 6-12 years. [The Exchange enrolls patients at the best centers, so US-wide use is probably lower – to what extent we’re not sure.] On the CGM front, penetration is still just 5% in T1D Exchange participants <26 years vs. 22% for those >26 years. Dr. Schatz emphasized that the ADA’s statement on CGM use (“a useful tool” for adults > 25 years) is a conservative approach that “needs to be extended” to younger children. This recommendation is likely based on the findings from the landmark (albeit six-year-old) JDRF CGM Trial. In addition, we believe devices need to get much smaller for greater pediatric use. Last, Dexcom only received pediatric approval for the G4 Platinum last year, so the market should grow to a greater extent moving forward, especially as remote monitoring and connectivity improves. Dr. Schatz called out the lack of regulatory scrutiny on insulin infusion sets (“almost none”), expressing “surprise” that such a key feature of the technology’s safety would be overlooked. For more details, please see our report below.

2. A valuable session on the DiAs system provided a whirlwind update on past studies and the exciting future of the system. In the detailed discussion below, we enclose more on our top learnings from some of the smartest minds in automated insulin delivery: (i) UVA’s Dr. Boris Kovatchev shared major studies planned for 2015 and presented a meta-analysis of 62,173 hours of DiAs use from 18 completed and ongoing studies; (ii) JDRF’s Dr. Aaron Kowalski expressed concern in Q&A over the feasibility of an overnight-only system (vs. 24-hour/day); (iii) Dr. Hans De Vries (Academic Medical Center, Amsterdam, Netherlands) presented encouraging new data from a 36-patient, two-month, at-home hybrid closed-loop study in Europe; (iv) UVA’s Dr. Stacey Anderson shared that the ongoing DiAs study testing three-weeks of overnight and three-weeks of 24/7 closed-loop at home has secured a five-month extension (she also shared some encouraging preliminary results); (v) Stanford’s experience closing the loop at camp reminded us that control groups are doing much better than their baseline A1c’s would suggest; and (vi) Dr. Claudio Cobelli suggested that some of the newer, faster insulins could “really mak[e] a difference” in closed-loop control, based on some early modeling work he has done.

• In a notable conversation with us outside the session, Dr. Kovatchev shared much more about the UVA team’s approach to regulatory and commercialization. The researchers met with the FDA on February 5 and received notably positive feedback on two fronts: (i) a closed-loop control algorithm does not need to reside on an insulin pump – it can reside on an outboard device, such as a smartphone or even a smartwatch; and (ii) academic trials can be counted for product approval. As we understand it, the UVA team does not intend to commercialize the DiAs system alone; the researchers will invite industry partners (i.e., pump and sensor companies) to bring it to market. The goal is to build a near-commercial system for one more large trial to support regulatory approval, hopefully through the NIH’s $20 million outstanding call for outpatient artificial pancreas studies.

3. Ms. Stephanie DuBose (Jaeb Center for Health Research, Tampa, FL) presented data suggesting that hypoglycemia unawareness and glycemic variability are the main factors underlying the high rates of severe hypoglycemia in older adults with type 1 diabetes – not a lower A1c. The case-control study evaluated 201 patients in the T1D Exchange registry 60 years or older with a duration of type 1 diabetes ≥20 years, none of whom had used CGM. The investigators categorized patients based on whether they had experienced severe hypoglycemia in the past 12 months and used lab tests (e.g., A1c, C-peptide levels), psychosocial assessments, and blinded CGM to identify the key factors that differed between the two groups. The group that had experienced severe hypoglycemia had significantly higher rates of hypoglycemia unawareness (58% vs. 25%; p<0.001) and high glycemic variability (38% vs. 12% in at least the 75^th percentile; p<0.001) compared to the controls. The percentage of patients with both hypoglycemia unawareness and high glucose variability was also significantly higher among cases than controls (24% vs. 5%; p=0.003). Notably, there was no meaningful difference in average A1c (7.8% vs. 7.7%; p=0.06) or the percentage of patients with an A1c <7% (26% vs. 28%; p=0.44) between the two groups, suggesting that at least in this population, tight glycemic control was not associated with a greater risk of severe hypoglycemia. We’re glad to see this since we have not agreed with the whole “oh, just increase their A1c target” as a means to make things safer for older people with diabetes. On a positive note, Ms. DuBose noted that she and her colleagues are planning a study investigating whether CGM could reduce the risk of severe hypoglycemia in these patients, explaining that this is “what the FDA would want to look at to get Medicare to approve these devices in this population.” We assume this means FDA could eventually put pressure on CMS to reimburse personal CGM for those on Medicare, provided the data is positive. We are very glad to see more research being done in this area, as we strongly believe that greater access to CGM would allow many older patients with longstanding diabetes to minimize the risk of severe hypoglycemia without sacrificing glycemic control.

4. Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA) provided an update on the Diabetes Technology Society’s Surveillance Program for Cleared Blood Glucose Monitors and announced a new DTS-sponsored program to develop a cybersecurity standard for connected medical technologies. On the former, Dr. Klonoff noted that the program has not yet received the requisite funding to begin testing meters, which was slated to begin in mid-2015 (per his comment’s at the September update). If funding is not received, the program will pursue a different model in which members of industry would volunteer to have their meters tested. In return, companies would receive DTS’s “seal of approval” that Dr. Klonoff hopes would become a kind of de-facto approval status – our thoughts on this model are shared in the full report below. Dr. Klonoff also highlighted the cybersecurity of connected medical devices (e.g., insulin pumps) as another safety issue and announced the launch of a new DTS initiative to develop standards for such technologies. The program comes out of conversations with the FDA, stemming both from concerns regarding intentional (i.e., grassroots efforts to access data) and malicious efforts to hack devices. We would agree that as wirelessly controlled medical device become the standard in our field, the risks associated with confidentiality and integrity of data have grown exponentially. A timeline was not provided though we assume we will hear more on this front later this year. Dr. Klonoff gave some commentary on Nightscout – see our report below.

5. We got a closer look at Medtronic’s fifth-generation “Harmony” sensor, a redundant approach incorporating signals from two sensors. Harmony involves two parallel planar electrochemical sensors; as we understand it, this is the fifth generation sensor in Medtronic’s pipeline (following Sof-sensor, Enlite 1, 2, and 3). A seven-day study evaluated the sensor in 15 adult type 1s (each wore two identical Harmony sensors) – on two days accuracy was assessed with a standardized meal test and YSI, and throughout the trial comparisons were made vs. a study blood glucose meter. The sensor achieved a MARD of 9% vs. YSI during the meal tests and a MARD of 10% vs. the study meter for the entire duration of the trial (~3,000 evaluation points), one or two percentage points better than the MARDs (vs. BGM) of the individual sensors. Presenter Dr. Amin Sharifi (St. Vincent’s Hospital, Melbourne, Australia) argued that sensor redundancy also extended the lifetime of the sensor, though the benefits (less than half an additional day of use and ~3% longer display time) were pretty marginal. What was more impressive to us was how seemingly intelligent the Harmony system was – in one case in which one sensor veered off course, the system terminated its signal and preserved accurate readings. JDRF and Helmsley were listed as supporters of the study; the two foundations announced a partnership with Medtronic in 2012 focused on an orthogonally redundant sensor that will combine electrochemical and optical sensing mechanisms in one sensor. That has not moved forward quickly at all, and we wonder if Medtronic has now focused effort on solely electrochemical redundancy. A key question with redundancy is manufacturing cost, and for Medtronic, when this sensor would come to market – a 9.0% MARD is what Dexcom’s G4AP algorithm has now, and Gen 6 should be roughly similar with fewer calibrations and a longer wear time.

6. An afternoon tag team from two Profil scientists featured a positive perspective on Afrezza and recognition of the challenges in developing oral insulins. Dr. Eric Zijlstra (Profil, Neuss, Germany) began by comparing Pfizer’s previous inhaled insulin Exubera to a mayfly – an insect that undergoes a long developmental phase but lives for just one day (this got some laughs). Although we heard doubt at the 2014 FDA AdComm whether Afrezza could obtain a hypoglycemia claim, Dr. Zijlstra argued in favor of a hypoglycemia benefit given that the reduction in hypoglycemia was seen in both well-controlled and poorly controlled patients. [As noted in our report on the FDA approval, the label only includes the type 2 diabetes hypoglycemia and weight data, ignoring the benefits in type 1.] Dr. Zijlstra characterized the Sanofi partnership as a win-win, and expressed enthusiasm about the potential to use Afrezza as part of an artificial pancreas system to provide initial postprandial control – something that is badly needed until faster-acting insulins are developed. The UCSB team has presented encouraging data on this front in the past, and we hope to see more discussion of this. Al Mann has commented that the ideal combination for might be a basal insulin patch pump with Afrezza.

• Dr. Tim Heise’s (Profil, Neuss, Germany) presentation provided a balanced view of the benefits and challenges associated with oral insulin. Oral delivery provides a more physiologic route that should lower hypoglycemia risk and reduce weight gain. The harsh pH of the stomach and presence of proteases pose formidable challenges; recent phase 2 data on Novo Nordisk’s oral GLP-1 agonist OG217SC showed almost 300x lower bioavailability with oral delivery (though very positive effects on A1c). Lilly’s peglispro (which is liver-selective) was recently delayed due to liver safety questions, and we wonder if oral delivery (which also involves first-pass liver processing) may have the same risks. Dr. Heise characterized clinical data on oral insulin from Emisphere, Biocon/BMS, and Oramed as slightly lackluster, with weaker efficacy and/or inter-patient variability. Dr. Heise continues to believe that there is lots of potential in oral insulin (especially oral basal insulin), but there remains much room for headache and frustration as well.

7. Dr. Roy Harper (South Eastern Health and Social Care Trust, Dundonald, UK) presented results from a pilot evaluation of Hygieia’s d-Nav device, an insulin titration device with built-in blood glucose meter capability. The single-center study in Northern Ireland enrolled 126 patients with diabetes (122 with type 2) treated with insulin for at least a year – 51% on basal/bolus, 42% on premixed insulin, and 7% on basal insulin only. Patients were instructed to use the device to test their glucose before every insulin injection and when they suspected hypoglycemia; the device would then recommend an insulin dose (if needed) as well as providing weekly insulin dosage updates based on blood glucose patterns with the aim of achieving a “balance” between reducing glucose and avoiding hypoglycemia (we hope to hear more specific details about how the algorithm works in the future). Among the 96 active participants, average A1c dropped significantly from 9.2% at baseline to 7.8% at the first follow-up visit (3-5 months) and 7.5% at the second (6-12 months); there was also a significant increase in the percentage of patients with an A1c ≤7.5% (12.6% at baseline; 45.1% at the first follow-up; 60.2% at the second follow-up). In many cases, the glycemic improvements likely resulted from increased insulin dosage; the average total daily insulin dose increased from 84 units at baseline to 153 units after 26 weeks. On the negative side, participants experienced an average 3.5-kg weight gain, and 40% reported an increase in mild hypoglycemia; there was no significant increase in severe hypoglycemia.

• Dr. Harper portrayed d-Nav as a service that can help patients manage the frequent dose adjustments necessary for successful insulin therapy without extensive support from clinicians – something that is sorely needed given the limited time most providers have with patients. The group is currently conducting a larger evaluation (n=260) in an attempt to “regionalize” the product throughout Northern Ireland; Dr. Richard Bergenstal is also initiating an NIH-funded study that could support uptake in the US. Dr. Harper said that the reception among patients has been very positive thus far, suggesting that weight gain and hypoglycemia have not been major barriers in most cases. Given the framing of d-Nav as a pseudo physician replacement, we are curious to see how amenable HCPs are to the device if it becomes more widely used – we imagine that some would embrace it as a useful tool to support greater self-management, while others might perceive it as infringing on their control over patients’ treatment regimens.

8. Dr. Neal Kaufman (DPS Health, Los Angeles, CA) stressed that randomized controlled trials are a poor way to examine digital telehealth interventions in diabetes. He pointed out that such studies are so time-intensive that the technology is often obsolete by the time results are reported. He did acknowledge that RCTs have value – particularly in understanding what drives patients to technology – but emphasized that “technology itself moves too fast.” He suggested that paradigms of investigation need to change, such that the field accepts short-term economic analyses showing cost-savings in self-selected populations enrolled in telehealth programs as opposed to waiting for longer-term A1c and complications outcomes. In fact, Dr. Kaufman noted that studying reduced hospitalization rates over just a 12-month period has shown significant cost benefit (up to $364 per patient) in those involved in such programs. Ultimately, we believe his vision for the future of telehealth provides valuable food for thought, especially because many innovative healthcare startups are not experienced in the vagaries of funding, conducting, and publishing RCTs. We believe that digital interventions have great potential to reduce the financial burden on the healthcare system and improve patient self-management, particularly by facilitating more convenient follow-up, just-in-time education, real-time feedback, and increased attention.

9. Dr. Stephanie Amiel (King’s College London, London, UK) gave a well received talk on a potential neural link between insulin resistance and altered food-related behavior. Dr. Amiel – an expert on hypoglycemia – was intrigued to research this topic after discovering that circulating insulin levels impact a substantial portion (~20%) of glucose metabolism in the brain. While it is increasingly clear that obesity and metabolic dysfunction are associated with changes in neural circuits involved in appetite regulation, she said that the direction of the relationship has not been demonstrated – “we don’t know what drives what.” Dr. Amiel and her colleagues conducted a series of studies examining brain activation in insulin sensitive vs. insulin resistant subjects with comparable BMIs of ~25 kg/m². fMRI and other neural imaging techniques showed that these non-obese, insulin resistant individuals displayed abnormal activation patterns in brain regions associated with reward and with responses to anticipating and eating food. Importantly, when the insulin resistant group underwent an insulin-sensitizing treatment regimen with lifestyle interventions and metformin, many of these patterns became comparable to the control group.

• Dr. Amiel also presented results from a neuroimaging study comparing obese patients who had undergone bariatric surgery with BMI-matched controls who had not. As expected, the post-surgical patients were less insulin resistant and displayed greater activation of brain regions associated with appetite inhibition. Interestingly, they also had abnormally increased activity in brain regions associated with reward salience; applying similar findings from studies of hypoglycemia unawareness, Dr. Amiel hypothesized that a lack of emotional salience around the rewarding aspects of food might be a component of the altered appetite/satiety signaling seen in obesity, suggesting that the sort of behavioral interventions used to increase awareness of hypoglycemia might be helpful in this area as well. While far more research must be done to understand the neural mechanisms underlying metabolic dysfunction, we wonder if these intriguing findings could hold potential for new approaches that address altered neural pathways at much earlier stages.

10. Our team visited the exhibit hall, which featured 25 companies this year including Abbott, Bayer, Dexcom, Medtronic, Roche, and many more. Highlights from select companies are featured in the appendix below – stay tuned for more in our full report.

Honorable Mention

Diabetes educator and therapist Mr. Joe Solowiejczyk (President and Founder, A Mile in My Shoes) provided an inside look into his interactive e-book, “A Type 1 Diabetes Guide to the Universe.” The e-book, available in the Apple App Store, integrates the psychological and family aspects of diabetes to help youth and their families learn to manage the disease as independently as possible. Mr. Solowiejczyk noted that the purpose is to “make it feel very similar to going to the doctor’s office,” which is accomplished through various embedded videos (which can be viewed offline) of both Mr. Solowiejczyk as well as of families and children from the US, Canada, and Australia. The e-book also includes a comprehensive range of chapters on diabetes management (e.g., insulin doses, meal planning, and exercise). Mr. Solowiejczyk highlighted that the response to this educational tool has been “tremendous” with its blend of high-touch and high-tech components. In addition, he called for any investors or collaborators who are interested in helping expand “A Type 1 Diabetes Guide to the Universe,” as the e-book is currently only available in English. Looking forward, Mr. Solowiejczyk plans to write similar books on the relationship of diabetes to celiac disease, sports and extreme sports, as well as ADHD/Asperger’s syndrome, with his ultimate long-term goal being an artificial intelligence version of himself that can have virtual conversations with patients. Recognizing the need for better education tools for both patients and their caregivers especially, we applaud Mr. Solowiejczyk for authoring this e-book and taking a unique and more engaging approach to education. As the field continues to suffer from a shortage of providers, we believe that tools like these are critical to help patients manage their own care more and more independently.

Detailed Discussion and Commentary

Parallel Sessions: Safety of Medical Devices for Diabetes

A Pediatrician’s View from Across the Pond – Pumps and Sensors

Desmond Schatz, MD (University of Florida, Gainesville, FL)

Dr. Desmond Schatz provided a comprehensive overview of “where we are” with pump and CGM use in the US pediatric population. Dr. Schatz prefaced his talk by putting the disappointing status of glycemic control in this population in context by sharing updated numbers from the T1D Exchange. As of January 2015, fewer than one in five US children is meeting ADA guidelines for glycemic control (A1c < 7.5%) despite an average annual income of ~$80,000/year in this population. Given the technology available, Dr. Schatz stressed that the numbers are in part a reflection of the underutilization of pumps and CGM in this population. Indeed, he associated pump use directly to better glycemic control by sharing recent data from the T1D Exchange. As we saw at ATTD 2011, the average A1c of pump users in the database is ~0.6% lower than that of injection users. On the CGM front, penetration is still just 5% in T1D Exchange participants <26 years vs. 22% for those >26 years. Indeed, he emphasized that the ADA’s statement on CGM use (“a useful tool” for adults > 25 years) is a conservative approach that “needs to be extended” to younger children.

• Dr. Schatz shared updated data from a subset of the T1D Exchange (n=16,791) on the proportion of US type 1 patients not at goal. The trend was incredibly depressing given how much better the tools have gotten. As a reminder, the target A1c for patients <18 years is 7.5%; the target A1c for those >18 years is 7.0%. Young adults continue to represent that most challenging population to manage with fewer than one in five patients ages 13-25 in goal.

Table 1: Proportion of US Type 1s at A1c Goal

• On a more positive note, Dr. Schatz shared T1D Exchange data showing that US pump penetration has been on an encouraging trajectory since 2010. Though certainly not representative of the whole US type 1 population, the data reveal the gradually increasing penetration of insulin pumps in the type 1 population over the last few years. Use of pumps in pediatric patients <13 years old has increased quite substantially over time, rising from 43% to 62% in those <6 years and 54% to 65% in those 6-12 years. This was pretty marked movement in just four years and we wonder how this breaks down by company – Insulet has historically been very strong in pediatrics, and we wonder how the company has fared vs. Medtronic, Animas, Roche, and Tandem. 

Table 2: Insulin Pump Use by Age in the T1D Exchange

• The presentation also featured updated numbers on the penetration of CGM in the US type 1 population as of the past three months. It is disappointing to see the limited uptake of CGM (5%) in the young adult population (ages <26), a number that has not increased since the T1D Exchange first enrolled patients (per numbers provided by Dr. Irl Hirsch at AADE 2014). We believe this is due to several factors: younger patients have less on-body real-estate and awareness to manage their own diabetes; devices have not gotten accurate enough until recently; remote monitoring and connectivity has only recently gotten convenient, with efforts like Nightscout and the soon-to-be-launched (March) Dexcom Share receiver.

Table 3: CGM Use by Age in the T1D Exchange

• Though the utility of CGM is already well established, Dr. Schatz reaffirmed this conclusion using DKA data in CGM users vs. non-users. Data came from the past three months in the T1D Exchange. In <13 and >26 year olds, CGM users have approximately half the rate of DKA vs. non-users.

Table 4: Prevalence of DKA – CGM vs. Non-CGM Users in the T1D Exchange

• Dr. Schatz also shared T1D Exchange data on the relationship between mean A1c and insulin pump use. This data has remained remarkably consistent since Dr. Irl Hirsch’s presentation at ATTD 2011. The real-world data seems to roughly corroborate what Medtronic and others have found in RCTs of pumps.

Table 5: Mean A1c – Insulin Pump Therapy vs. Injection in the T1D Exchange

• Dr. Schatz acknowledged that pre- and post-market surveillance methods for pumps in the US are woefully inadequate. As we have heard in the past, he called out the lack of regulatory scrutiny on insulin infusion sets (“almost none”), expressing “surprise” that such a key feature of the technology’s safety would be overlooked. Similarly, he called for a more systematic and transparent approach to collecting adverse event data, acknowledging that the current MAUDE database is inadequate. That said, Dr. Schatz highlighted the recent online publication of the ADA/EASD Position Statement on Insulin Pumps as a step in the right direction.
• Dr. Schatz stressed that there are a number of factors that need to improve with CGM as well: (i) improved accuracy and reliability; (ii) redundancy beyond glucose oxidase; and (iii) calibration-free devices. We’d add much smaller and better connected devices to that list. We’re not sure if redundancy beyond glucose oxidase is really a key needed improvement – many artificial pancreas experts have said that a MARD <10% is accurate enough to run closed loop, and Dexcom’s G4AP algorithm is already there.

Questions and Answers

Dr. Tadej Battelino: Although it is strict, the FDA seems to understand that providing good technology is part of the deal. Is that fair to say?

A: You’re right. We have seen a change in the FDA. They really want to work with us. They recently down-classified secondary display of CGM data. That was a surprise to me. I think with all of this that there’s a change in FDA attitudes and approaches.

Q: You showed that people have better becomes with CGMs and pumps when used. How do we encourage more use?

A: I think it’s education. I think that very often we talk about technology and its wonder and don’t spend enough time explaining all the features of the pump and benefits and what could go wrong. I think that patients are often surprised when things go wrong. We don’t usually explain that well enough, and I think that if we were to spend more time going through these things, more people would be using these devices and less quick to abandon them.

The Need for Postmarket Surveillance of Clerned Blood Glucose Monitors

David Klonoff, MD (Mills-Peninsula Health Services, San Mateo, CA)

Dr. David Klonoff pointed out the need for programs to improve the safety of inaccurate blood glucose monitors and of wirelessly connected diabetes devices. He opened with a summary of the current state of postmarket surveillance of blood glucose monitors, highlighting that cleared devices do not always function as well as they did to become cleared. He stressed that inaccurate meters are risky for making treatment decisions and particularly for calibrating CGMs. On this front, Dr. Klonoff provided an update on the Diabetes Technology Society’s (DTS) Surveillance Program for Cleared Blood Glucose Meters. As we understand it, the program has not yet received the requisite funding to begin universal testing of all meters on the US market, which was slated for mid-2015 (per his comment’s at the September update). Dr. Klonoff also highlighted the cybersecurity of connected medical devices (e.g., insulin pumps) as another safety issue and announced the launch of a new DTS program to develop a cybersecurity standard for such technologies. The program comes out of conversations with the FDA, stemming both from concerns regarding intentional (i.e., grassroots efforts to access data) and malicious efforts to hack devices. On the latter front, we would agree that as wirelessly controlled medical device become the standard in our field, the risks associated with confidentiality and integrity of data have grown exponentially. DTS will begin by forming a steering committee to discuss what such a cybersecurity standard would look like. A timeline was not provided though we assume we will hear more on this front later this year.

• Dr. Klonoff laid out an ambitious slew of milestones for DTS’s Surveillance Program for Cleared Blood Glucose Meters, including: (i) the completion of the BGMs testing protocol this winter; (ii) the certification the document; and (iii) raising money for testing.
• Dr. Klonoff also described a new model for the program in which members of industry would volunteer to have their meters tested. For example, Company A would provide funds to have Company A's meters tested, and Company B would provide funds for Company B's meters to be tested. In return, companies would receive the DTS’s “seal of approval” that Dr. Klonoff hopes would become a kind of de-facto approval status. We do wonder how feasible this model is considering that: (i) there are no clear consequences for company’s that do not pass DTS’ program; (ii) there is no indication that payer’s would remove cheaper devices from formularies given that those devices have been FDA approved; and (iii) the BGM environment is already strapped for cash. The need for improved post-market testing is undeniable, and we applaud Dr. Klonoff for persevering on this front despite the fact that industry has not fully bought into the concept.
• Dr. Klonoff raised some concerns over recent grassroots efforts to hack connected medical devices (e.g., NightScout, #DIYPS) – the commentary is in line with other remarks we have heard on his front as some doctors have expressed genuine concern about the reliability of improvised technologies that have not gone through standard regulatory review.  Some – not necessarily Dr. Klonoff – may be upset that laymen can devise these high-tech gadgets on their own, usurping the role of the experts to develop, promote, and explain the latest breakthroughs. From our view, parents will always do what they think is best for their children, particularly given the danger of insulin.  That said, Dr. Klonoff was speaking from the perspective of safety and we are not surprised he did not get into the politics.

Twelve Years of Collaborative Transatlantic Research on Artificial Pancreas

This impressive session provided a whirlwind update on all things happening with the DiAs system around the world. Below, we enclose some of our top learnings from this session: (i) Dr. Boris Kovatchev (UVA) shared major studies planned for 2015 and presented a meta-analysis of 62,173 hours of DiAs use from 18 completed and ongoing studies; (ii) Dr. Aaron Kowalski (JDRF) expressed concern in Q&A over the feasibility of an overnight-only system (vs. 24-hour/day); (iii) Dr. Hans De Vries (Academic Medical Center) presented positive new data from a 36-patient, two-month, at-home hybrid closed-loop study in Europe; (iv) Dr. Stacey Anderson (UVA) shared that the ongoing DiAs study testing three-weeks of overnight and three-weeks of 24/7 closed-loop at home has secured a five-month extension; (v) Stanford’s experience closing the loop at camp reminded us that control groups are doing much better than their baseline A1c’s would suggest; and (vi) Dr. Claudio Cobelli suggested that some of the newer, faster insulins could “really mak[e] a difference” in closed-loop control, based on some early modeling work he has done.

• Dr. Boris Kovatchev highlighted DiAs’ history (introduced in 2011) and future, including major studies planned for 2015:
  
  • UVA’s “Project Nightlight,” an 11-month study testing alternating treatment modalities, will start sometime after March 1 (NIH-4). The approach is discussed in more detail below, but patients will cycle through sensor-augmented pump therapy, nighttime, 24-hour, and nighttime (a second time) closed-loop control. In addition to what approach works best, the idea is to see what patients prefer. 
  • Reducing hypoglycemia and reversing hypoglycemia unawareness with one month around the clock closed-loop control at home (NIH 3: UVA, Stanford). It’s tremendous to see a study in hypoglycemia unawareness, a population that stands to benefit significantly from automated insulin delivery. We wonder how large the study will be and if it will be possible to show a benefit on severe hypoglycemia.
  • A four-center trial will test UCSB’s Adaptive Zone Model Predictive Control (NIH 5: UCSB, UVA, Mayo, Padova). Dr. Kovatchev did not share more details on the length, design, or start date of this study.
• Dr. Kovatchev also presented a meta-analysis of 62,173 hours of DiAs use from 18 completed and ongoing studies (n=302). Results shown below are weighted by study, which have occurred in the US, France, Italy, Holland, and Israel and tested three different control algorithms (USS Virginia, Pavia’s RCM, and UCSB’s Zone MPC) – as a reminder, the DiAs framework allows investigators to swap algorithms in and out.
  
  • DiAs has performed tremendously during overnight studies: time spent <70 mg/dl has been reduced by two-thirds (1% vs. 3%), average glucose has improved to 139 mg/dl (vs. 153 during open-loop), and overall time-in-range (70-180 mg/dl) has risen to 78% from 64% during open loop.  
  • During 24-hour closed-loop control, time spent <70 mg/dl has been more than halved on closed loop (1.8% vs. 3.8% on open loop), while time in range (70-180 mg/dl) has improved to 73% on closed loop vs. 66% on open loop. Average blood glucose has dropped slightly during closed loop: 148 vs. 154 mg/dl on open loop.
• Dr. Aaron Kowalski expressed concern in Q&A over the feasibility of an overnight-only system: “A theme here is overnight vs. 24 hour. I look down at Dr. Hovorka, as we have debated this. I worry a little bit that as we display data from these pilot studies, folks not as familiar with the field will not see a big difference between daytime closed-loop vs. open loop. But as Dr. Buckingham showed, these pilot studies are so controlled. The control group is testing frequently and there is often staff nearby. So you see a compression of glucose control that is not representative of the general population. [see below] A key issue is, ‘What is overnight?’ I don’t think there is ever going to be an overnight-only system. Are there scenarios when a closed loop would underperform in a broader sense or lead to bad scenarios during the day?” [Note: This remark reiterated Dr. Kowalski’s back-and-forth with Dr. Roman Hovorka at last year’s EASD Diabetes Technology Meeting.]
• Stanford’s experience closing the loop at camp reminded us that control groups are doing much better than their baseline A1c’s would suggest. Note in the table below – control groups are coming in with a baseline A1c of 8.2-8.4%, though while at camp, averaging glucose values equivalent to A1c’s of 7.1-7.3%! This was an encouraging reminder that automated insulin delivery should ideally do much better in a real-world setting, as the delta between open- and closed-loop should be wider. Of course, this is to say nothing of the types of patients that go to camp or enroll in closed-loop studies – those who are diligent, motivated, and in-touch with their clinical care teams. In addition, many of the patients that would most benefit from closed-loop systems are often not include in these trials (e.g., those with A1c’s of >10%, those with hypoglycemia unawareness, very young patients, etc.). 

• Dr. Hans De Vries (Academic Medical Center, Amsterdam, Netherlands) presented the first data from a 36-patient, two-month, at-home hybrid closed-loop study (one of two final AP@home studies). The crossover trial has randomized patients to begin with either two months of sensor-augmented pump therapy or two months of closed-loop at home (from 8 pm – 8am, dinner included). Dr. DeVries presented preliminary graphs from the first half of the crossover trial. The system uses the DiAs, a Roche pump, the Dexcom G4 (+ the Mazlish box for Bluetooth relay).
  
  • Though he did not share specific numbers, time spent in hypoglycemia appeared to decline by ~50% (p<0.05), and time spent above range declined from ~36% to ~30% (p<0.05). Mean blood glucose was not significantly different between open and closed-loop. This was an intent-to-treat analysis from all patients who completed more than six weeks of the study.
  • Dr. DeVries commented that such a dinner-to-morning closed-loop system might be possible to design into a commercial product– it  could simply be made available on the device (via GPS or Wi-Fi) when patients get home, and disappear when they leave home. We thought this was a novel approach to the question – “What is overnight?” – as it flips the challenge of defining a specific time period to defining a set location (home, where patients spend most nights and risks are lower than during daytime meals, exercise, etc.).
  • This AP@Home study has two one-month extensions planned: four weeks of 24/7 closed-loop (uncontrolled) and four weeks of 24/7 closed loop with an individually adaptive algorithm (uncontrolled).
• Dr. Stacey Anderson shared that the ongoing DiAs study testing three-weeks overnight and three-weeks 24/7 closed-loop at home has secured a five-month extension! The trial is taking place at UVA, UCSB, Stanford, Padova, Montpellier, Israel. Changes requested prior to prolonged are use in progress. The team hopes to add a temporary basal feature and has already switched to the recently approved Dexcom Share receiver (previously, they were using the Mazlish box to send the Dexcom G4 data to the DiAs algorithm running on an Android phone).
  
  • US patients in the study have spent half the time in hypoglycemia during combined closed-loop day + nighttime use (2% vs. 4%), and significantly their improved time-in-target (74-76% vs. 66%). Results are even more impressive overnight, with 0% of the time spent in hypoglycemia (vs. 2% on open loop) and 78-81% time in 70-180 mg/dl (vs. 67% on open loop).
• Based on modeling studies, Dr. Claudio Cobelli suggested that some of the newer, faster insulins are “promising” and “really making a difference” in closed-loop control. He did not specify which, though we wonder if he was referring to data on Novo Nordisk’s FiAsp (ultra-fast aspart), which suggest it’s particularly faster than Novolog in the first hour. In our view, the speed of insulin – which directly affects a system’s ability to deal with meals and exercise – is really the biggest unsolved problem in closing loop.

AP Data Club

Closed Loop Control Modalities in Type 1 Diabetes: Efficacy and System Acceptance

Boris Kovatchev, PhD (UVA, Charlottesville, VA)

Dr. Boris Kovatchev shared his view on why the artificial pancreas is not a “single-function device”, but a platform for technology deployment. He emphasized that closed loop control is one of many possible apps running on a scalable mobile medical network (e.g., sensor-only mode, pump-only mode, safety mode, closed-loop) – patient preferences and signal availability determine the active treatment modality. Dr. Kovatchev shared the design of an 11-month study (starting sometime after March 1) to compare sensor-augmented pump, overnight-only, and 24-hour closed-loop control. This sounds fascinating and we look forward to seeing how patients do, as well as what they prefer. In a notable conversation with us outside the session, Dr. Kovatchev shared much more about the UVA team’s approach to regulatory and commercialization. The researchers met with the FDA on February 5 and received notably positive feedback on two fronts: (i) a closed-loop control algorithm does not need to reside on an insulin pump – it can reside on an outboard device, such as a smartphone or even a smartwatch; and (ii) academic trials can be counted for product approval. As we understand it, the UVA team does not intend to commercialize the DiAs system alone; the researchers will invite industry partners (i.e., pump and sensor companies) to bring it to market. The goal is to build a near-commercial system for one more large trial to support regulatory approval, hopefully through the NIH’s $20 million outstanding call for outpatient artificial pancreas studies.

• “Memo: The artificial pancreas is not a single function device – it is a platform for technology deployment.” Dr. Kovatchev highlighted that various treatment modalities will (and should) be possible with these systems, and closed loop control is just one of many treatment options that can be run on a scalable mobile medical network. Patient preference and glycemic targets at a particular time determine the particular treatment modality, as does signal availability (e.g., if the CGM is not operating). In his view, long-term artificial pancreas use will inevitably require activation of different treatment modalities at different times. What’s more, systems are bound to fail, and as such, they must default to a known state of operation.

• An 11-month, NIH-supported, at-home closed-loop study will test different treatment modalities using the DiAs system (starting sometime after March 1). The study will use the DiAs platform (Android phone running control algorithm/user interface and connecting to a Dexcom CGM and Roche pump). Group A will proceed through the following treatment modalities (10 weeks each) using the DiAs system: sensor-augmented pump, overnight closed loop, day+night closed loop, and overnight closed loop. Group B will do the same four segments, but in a different order: overnight closed loop, day+night closed loop, overnight closed loop, and sensor-augmented pump. In addition to what approach works best, the idea is to see what patients prefer. This should offer interesting perspective on the question of overnight-only vs. 24-hour closed-loop. (We would note that during the overnight-only phase, the system only allows closed-loop control between preset times (e.g., 11pm - 7am) – patients cannot simply run it during the day).
• Dr. Kovatchev contrasted the configuration of the DiAs system to the “very limited” approach taken in the traditional artificial pancreas system configuration. In the latter, the closed-loop user interface, algorithm, and communication functions are built into the pump. Dr. Kovatchev called this approach “very limited” – it’s difficult to get a cloud connection (e.g., remote monitoring, trend recognition, analytics, remote diagnostics), the user interface is limited by the pump’s real estate, and upgrading the system is a challenge. The UVA team prefers an “alternative artificial pancreas system configuration” – a CGM, a pump, an artificial pancreas hub (smartphone), and cloud services. The phone takes the function of closing the loop. This helps overcome the key aforementioned down sides of building the algorithm into the pump. Of course, the phone brings new challenges in (e.g., more devices to carry, more vulnerable to glitches, battery life, durability), so time will tell which approach emerges as quicker and better for patients. 
  
  • As we understand it, the UVA team met with the FDA on February 5 to answer questions about the DiAs system configuration and the possibility of academic trials counted for product approval. FDA gave the UVA team positive answers on both counts! Most notably, the Agency told UVA that a closed-loop control algorithm does not need to reside on an insulin pump; it can reside on an outboard device, such as a smartphone or even a smartwatch (with the appropriate software solutions for safety and security). The major positive is that closed-loop approaches will not need to build new hardware with the control algorithm/interface integrated into a pump. This means any sensor-augmented pump could close the loop with an external device (provided there is rigorous system and software validation). Dr. Kovatchev cautioned that patients will not be able to swap components in and out – the entire system (phone running control algorithm, pump, CGM) will still be reviewed as a Class III medical device. We assume such an approved system would have a stripped down smartphone with very limited functionality.
  • As we understand it, the UVA team does not intend to commercialize the DiAs system alone; the researchers will invite industry partners (i.e., pump and sensor companies) to bring it to market. The goal is to build a near-commercial system for one more large trial. This would take advantage of the outstanding $20 million NIH call to fund large outpatient artificial pancreas projects. UVA intends to apply with a consortium of nine centers (three Europe, six in the US). The application date is April 15.

Questions and Answers

Dr. Bruce Buckingham (Stanford University, Stanford, CA): How about using DiAs in MDIs – those not using pump? Could they use it for safety supervision, without wearing equipment?

A: We have not done the trials, but have plans to do it with Dr. Marc Breton. It is entirely possible to feed MDI information into the system, to consider the profile of long-acting insulin, and still make some recommendations – short of closing the loop.

Oral Presentations

Factors Associated with Severe Hypoglycemia in Older Adults with Type 1 Diabetes

Stephanie DuBose, MPH (Jaeb Center for Health Research, Tampa, FL)

Ms. Stephanie DuBose presented data suggesting that hypoglycemia unawareness and glycemic variability are the main factors underlying the high rates of severe hypoglycemia in older adults with type 1 diabetes. The case-control study evaluated 201 patients in the T1D Exchange Registry aged 60 years or older with a duration of type 1 diabetes ≥20 years, none of whom used CGM. The investigators categorized patients based on whether they had experienced severe hypoglycemia in the past 12 months and used lab tests (e.g., A1c, C-peptide levels), psychosocial assessments, and blinded CGM to identify the key factors that differed between the two groups. Results showed that the cases (the group that had experienced severe hypoglycemia) had significantly higher rates of hypoglycemia unawareness (58% vs. 25%; p<0.001) and high glycemic variability (38% vs. 12% in at least the 75^th percentile; p<0.001) compared to the controls. The percentage of patients with both hypoglycemia unawareness and high glucose variability was also significantly higher among cases than controls (24% vs. 5%; p=0.003). Notably, there was no significant difference in average A1c (7.8% vs. 7.7%; p=0.06) or the percentage of patients with an A1c <7% (26% vs. 28%; p=0.44) between the two groups, suggesting that at least in this population, tight glycemic control was not associated with a greater risk of severe hypoglycemia. Excitingly, Ms. DuBose noted that she and her colleagues are planning a study investigating whether CGM could reduce the risk of severe hypoglycemia in these patients, explaining that this is “what the FDA would want to look at to get Medicare to approve these devices in this population.” We are thrilled to see more research being done in this area, as we strongly believe that greater access to CGM would allow many older patients with longstanding diabetes to minimize the risk of severe hypoglycemia without sacrificing glycemic control.   

• Other factors that differed significantly between cases and controls included: fear of hypoglycemia (scores of 38.5 and 31.6, respectively, on the Hypoglycemia Fear Survey; p<0.001), frequency of blood glucose monitoring (six times/day vs. five times/day; p=0.02 – not clinically meaningful), and use of beta blockers (40% vs. 21%; p=0.006), though there was no correlation between beta blocker use and hypoglycemia unawareness. A significantly higher percentage of cases also had reduced cognitive capacity as measured by the Montreal Cognitive Assessment (MoCA), but Ms. DuBose noted that this may have been a result of severe hypoglycemia rather than a contributing factor.
• There was no significant difference between the groups for factors including: mean glucose (175 mg/dl for both; p=0.58), time with blood glucose between 70 and 180 mg/dl (51% vs. 52%; p=0.28), daily function, rates of depression, diabetes numeracy, and C-peptide levels. There was a trend toward more overall hypoglycemia in cases vs. controls (65 vs. 43 minutes/day with blood glucose <60 mg/dl), but it was not statistically significant (p=0.12).
• These results open an interesting discussion relative to Dr. Irl Hirsch’s (University of Washington, Seattle, WA) ATTD presentation advocating for higher A1c targets in patients with longstanding type 1 diabetes. Dr. Hirsch argued that given the unacceptably high rates of severe hypoglycemia in this population (~20% risk per year with a duration of diabetes >20 years) and the mixed evidence regarding the benefits of strict control, less stringent targets are likely appropriate for many patients. While we certainly agree that the dangers of hypoglycemia should not be overlooked, these data suggest that tighter control (as measured by A1c) may not be driving severe hypoglycemia. CGM may help, but the lack of evidence from clinical trials confirming that hypothesis has been one of several barriers to Medicare coverage of CGM; our fingers are crossed that the upcoming study Ms. DuBose alluded to can help remedy that.

Questions and Answers

Dr. Robert Vigersky (Walter Reed National Military Medical Center, Bethesda, MD): Did you analyze the results with respect to A1c? In type 2 diabetes, there’s a U-shaped curve that describes the frequency of hypoglycemia related to A1c. Did you look at that?

A: Not yet, but we have a manuscript planned that will include that.

Q: With the variability data, in terms of generalizability, if you define high variability as higher than 75% of the population, that’s less generalizable. Many of these subjects would fit the criteria for starting the 530G. There is a clinical remedy for this. What further studies would we need?

A: We’ve formed a working group for the next study. We want to look at the CGM study and go from there and look at what the FDA would want to get Medicare to approve these devices in this population. It is less generalizable; we defined high variability as a measure to look at the combination of those two factors [glycemic variability and hypoglycemic unawareness], and it was specific to this group.

Dr. Irl Hirsch (University of Washington, Seattle, WA): We’ve published in pediatric and adult patients the relationship between A1c and severe hypoglycemia. In both groups – it’s more striking in pediatrics but not specific to the older population – there’s a U-shaped curve between severe hypoglycemia and A1c. I don’t know this for a fact, but I think the highest A1cs have more hypoglycemia because of decreased frequency of testing. That would be different from type 2 and from the conclusions of the DCCT.

Exhibit Hall

Abbott

Abbott’s distinctive yellow booth greeted attendees at the front right of the exhibit hall. FreeStyle Libre was the focus of the booth, and a slew of representatives offered demos of the new technology. The device launched ~five months ago, and is still clearly generating a lot of excitement judging from the “oohs and aahs” we overheard from attendees seeing the technology for the first time. Representatives noted that reimbursement remains a primary goal moving forward, though emphasized that the fast uptake – which has exceeded manufacturing capacity – is a terrific sign that patients are already appreciating the value proposition.

Bayer

Bayer representatives hinted at a next-generation Bluetooth-connected BGM in the company’s pipeline – this is the first we have heard on this front. Representatives provided a rough timeline, suggesting a launch is “one to two years away.” We have heard very little from Bayer on the pipeline front over the past 12 months, so it is positive to see that the company is still actively investing in R&D despite the fact that the challenging market conditions do not lend themselves to innovation. Reps also highlighted the company’s Next Link 2.4 meter that recently launched with Medtronic’s MiniMed 640G system in Australia. The meter is now available for purchase throughout the EU. As a reminder, the device’s key innovation relative to the first-gen Next Link is the remote bolus feature, a huge convenience win for patients that we think will be well received.

Cellnovo

Cellnovo had a one-manned booth at the edge of the exhibit hall, where we learned that its pump system has launched in both the UK and France. The sales rep noted that the uptake in Europe has been solid with the France launch two months in and the UK launch having begun its phase launch at the end of 2013. The rep did not disclose any updates on progress in the US, although we last heard of a potential FDA filing in 2014. As a reminder, this system consists of three components – a pump/rechargeable battery unit and disposable insulin cartridge, a color touchscreen handset with an integrated meter (LifeScan Vita strips), and a web-based data management system. For more background on Cellnovo, please see our last coverage of the company.

CeQur

Enthusiastic representatives provided us a demo of CeQur’s PaQ insulin delivery device and an updated timeline on both EU and US commercialization efforts. As noted in our Day #2 Highlights, the company is now hoping for a EU launch in 2016 (CE Mark approval came in November 2012), a slight delay relative previous guidance that put a launch in 2015. In the US, CeQur hopes to submit an FDA filing this year and is optimistic for a launch in 2016. As a reminder, the company appointed Mr. Douglas Lawrence as Chief Executive Officer in April 2014, and we are hopeful that his background at BD will help CeQur navigate the complex US regulatory and commercial landscape.

J&J

J&J’s modest booth was devoted exclusively to the Animas Vibe. Reps spoke broadly about the Vibe’s success, touching briefly on its recent FDA approval. When asked to comment on the future of Animas’ pipeline, reps did not divulge any details though noted, “As it is for everyone, the artificial pancreas is the future.” We hope this is true, as J&J has not indicated significant progress on this front besides a couple small inpatient feasibility studies presented at ADA 2013 and ADA 2014.

Medtronic

Medtronic’s large booth showcased the brand new MiniMed 640G, allowing attendees to walk up and play with the pumps. An Australian rep was highly enthusiastic, noting that patients are “delighted” to get the pump, which has a cool factor that sets it apart from “just another medical device. Our ATTD Day #1 report provides extensive coverage of the MiniMed 640G’s layout, features, and menus. See our Day #2 report for the latest clinical data. 

Roche

Roche’s booth gave attendees the opportunity to play around with the Accu-Chek Connect (standalone meter + smartphone app + web portal) and Accu-Chek Insight (next gen insulin pump with BGM remote) systems. The demo of the former featured a look at the sleek new Connect app for iOS that is set to launch in April 2015. (As a reminder, the Connect app is currently only compatible with Android OS). Representatives were silent when asked about progress on the CGM front, though as we understand it, the company’s novel CGM remains at least 12 months away from commercialization.

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