February 5-8, 2014; Vienna, Austria; Days #3-4; Highlights – Draft

Executive Highlights

Hello from Vienna, where ATTD 2014 just wrapped up. Our Days #3-4 highlights report contains a lot of nuanced discussion on the artificial pancreas from the field’s major luminaries. As we hang up our conference badges and board our plane back to San Francisco, we feel we have a much clearer sense of where the field stands, more specifics on the challenges, and a clear view on where the field must go in the coming years in order to ensure success for patients and industry. See below for discussion of some of the meeting’s key themes: striking data and outpatient progress on nocturnal closed-loop control; crossing the chasm from academic research to commercial products; and broader discussion on the commercialization/implementation of closed loop, including extensive conversation on patient selection, study design, and related topics.

The meeting’s final two days had no shortage of inspiring outpatient closed-loop data – trials from the University of Virginia, Stanford, the DREAM consortium, and BU/MGH all took center stage. A number of studies on nocturnal closed-loop control had similar messages – it works, it’s safe, and is moving ahead full steam on longer-term home studies. Of course, it will be key to see whether industry sees a major opportunity here, and whether the funding is there to move ahead on commercializing a product. Notably, JDRF asked five diabetes technology companies (Animas, Dexcom, Medtronic, Roche, Tandem) to present their views in a dedicated symposium on the artificial pancreas. Unfortunately, many with whom we spoke were underwhelmed by the ambition and outward commitment from insulin delivery companies; we believe some of this could stem from: how busy the companies are on pushing forward their main “here and now” business, as we know all are under major profitability pressure; not wanting to disclose the strategy they have on the automated insulin delivery/closed loop front; and a desire to focus on on real-time learning at this meeting than on making promises where it is still not necessarily possible to commit. That said, we do understand why some may feel stuck between a rock and a hard place – it would be better if their base businesses were stronger so they had more to invest. We truly believe automated insulin delivery/closed loop will be the “killer app” for both pumps and CGM, and investment without certainty of exactly where and how fast the field is moving will be necessary. It’s absolutely a race to the finish line at this point, and the company that gets a user-friendly and workable closed-loop device to market stands to gain share in a big way. We also acknowledge that companies may view patients as overly demanding in this new arena where much is uncertain; it is one reason why we like the idea of closed loop at night (where there are few disturbances and how well this works should be very clear) and a broad “range” during the day, rather than 24/7 “complete” closed loop to start. On a related note, one of the best sessions we attended was on AP product development; in this session, key opinion leaders weighed in on some of the key debates and their thoughts on what we need right now – see below for thoughts from Drs. Moshe Phillip, Howard Zisser, Roman Hovorka, Boris Kovatchev, and Eric Renard.

Overall, one of the most exciting parts of ATTD was seeing the sheer number of people attending (over 2,100 attendees this year), with every session jam-packed and many full of interesting discourse, new data, and visions of the future (while the “hard core” diabeterati that attend every single conference on this topic felt there was a lot of overlap, we felt it was a pretty good mix since many ideas were new since a year ago, which for many attendees was the last meeting they attended on this subject). The number of groups vying for patient participants, research funding, regulatory expertise, and industry partners has expanded from a year ago; there is great urgency on the part of many surrounding the quest to closed the loop and there’s also more noise in the field now. The field is moving closer and the challenges are also becoming clearer.

We also feel sure leaving Vienna that in addition to questions surrounding what patients want and which patient segments will benefit the most, that it is very clear that the biggest and most important question is how the benefits of automated insulin delivery and the closed loop will be conveyed to the payers around the world (see more in our themes below). As one very respected researcher said to us very succinctly, “Build it and will they come?” is not as big a question as “build it and who will pay?” Gaining access to these advances is an area where we believe JDRF in particular will be a major help; there is no other organization that had such a big impact on CGM reimbursement, though we know of course most of those gains were in the US. From our view, once there is an automated insulin delivery system that works, we would like to see a major long-term outcomes trial begin so that the world’s payers will not be able to deny the impact.

Read below for the final two days’ top 10 highlights, followed by detailed discussion and commentary.

Top 10 Highlights

1. As we leave ATTD 2014, a few key themes come to mind regarding automated insulin delivery. These include striking data and outpatient progress on nocturnal closed-loop control; the translation from academic research to commercial products; and discussion of critical ancillary questions (e.g., patient selection, study design, reimbursement, etc.).

• Nocturnal closed-loop was the focus of multiple talks at ATTD 2014, with a number of researchers (University of Virginia, DREAM, Cambridge) noting their belief that this is the next step after predictive low glucose suspend. We would note this is in contrast to previous years (and the original 2009 JDRF artificial pancreas roadmap) that put a treat-to-range/hypo-hyperglycemia minimizer as the next move forward in automated insulin delivery rather than a singular nighttime focus. At this meeting, many researchers (e.g., Drs. Boris Kovatchev, Roman Hovorka, Moshe Phillip) have presented impressive, long-term (up to six weeks), outpatient data demonstrating that overnight systems are safe and highly effective in bringing down glucose levels and reducing the risk of hypoglycemia. Notably, studies are also beginning to show that closed-loop overnight has positive ripple effects the following day. The risk-benefit of such systems seems much more favorable, particularly because current insulin therapy overnight is particularly suboptimal; for example, Dr. Hovorka highlighted that nocturnal insulin requirements are highly variable within the same patients from night-to-night (ranging from double the pre-programmed basal rate to half the pre-programmed basal rate within the same week), as well as throughout the night. Researchers also emphasized that overnight closed-loop systems face fewer “disturbances” such as meals and exercise and stress, critical given the current speed of subcutaneous insulin to deal with these challenges. This may also be useful to prove how well automated insulin delivery works “naked,” before adding the external variables that all systems will have to battle during the day. Too, we believe that some patients and even some providers may have unrealistic expectations from initial 24/7 systems; food and exercise and stress will obviously impact control and it may be best to work in increments, improving control at night first while reducing risk of hypoglycemia (but not necessarily hyperglycemia, to the same extent), during the day. We also would be cautious about extremely well managed patients – while their usual care won’t approach closed-loop insulin delivery overnight, we may hear from some patients (and parents) arguments about the early daytime algorithms. That said, even in these patients, closed-loop control stands to reduce all the effort and burden that good management requires.

• Another clear theme from this year’s ATTD could be termed the “academic/commercial chasm” – how will closed-loop systems in the academic research setting translate to commercialized products? Some researchers now have systems that are well tested and characterized in the research setting (e.g., DREAM, BU/MGH’s Bionic Pancreas, UVa’s DiAs) but it’s unclear how these will translate into commercial products. Ultimately, most believe that a control algorithm running on a smartphone is valuable for research purposes, but not feasible for a final commercial product. With that in mind, the next step seems to be integrating algorithms (either from academia or industry) into pump/CGM systems. But this brings up a vast array of questions:

  • Will insulin delivery companies (Medtronic, Animas, Tandem, Roche) in-license academic technology?

  • How hard will it be for multiple companies to work together to get a closed-loop system approved (e.g., Dexcom CGM + another company’s pump)?

  • Will academic groups start their own companies to build commercial products? If so, how will regulators perceive such efforts, especially for manufacturing and validation?

  • What will it take for regulatory bodies to be comfortable with interchangeable closed-loop components (i.e., choose your devices and plug in whatever algorithm you want – Cambridge, DREAM, UVa, etc.)?

  • On the other hand, will approvals hinge on unified systems?

• Historically, the ultimate translation from academic research to commercial products has always been an open question, and certainly one that JDRF and the Helmsley Charitable Trust have worked to address through industry partnerships (e.g., Medtronic, BD, Dexcom). Now, it’s clear that the academic research has advanced far enough; the next question is what are the subsequent steps and how will groups get there. For example, how could academia think about the following?

  • What components make up their first-generation systems?

  • Are the components “exclusive”? With future generations, can more components be used?

  • What will FDA demand in terms of validation for manufacturing?

  • Are handmade systems acceptable for research?

  • What will FDA require in terms of proving manufacturing capabilities?

  • Who will fund these needs? How much funding is necessary?

  • How clunky will the early systems be? What will satisfy patients? How broad is the range of acceptability in form factor for the earliest systems? 

• With research progressing quickly and more robust closed-loop systems moving to the outpatient setting, other ancillary questions are coming up with increasing frequency:

  • What patients are optimal for different closed-loop systems?

    • Should trials enroll more patients with severe hypoglycemia? How would this be defined?

    • How will newly diagnosed patients cope with closed-loop technology?

  • Given the potential for broad use and millions of patients exposed, what’s the best way to evaluate the safety of closed-loop systems?

    • How should regulators balance comprehensive safety evaluations without being overly burdensome?

    • How will companies perceive the liability of commercializing closed-loop technology?

    • What length of pivotal study is necessary to understand a system’s risk, given the potential for millions of patient-years of exposure once a device is commercialized?

  • How should patient expectations be set? What is optimal?

    • What term(s) should be used – “automated insulin delivery” for the early systems, “closed loop” for the later systems? Where do the terms “artificial” and “bionic pancreas” fit in?

  • Are healthcare providers ready for closed-loop insulin delivery?

    • What role should remote monitoring play?

    • Should commercialized closed-loop technologies include remote monitoring by default and let patients/providers choose whether to use it?

    • Will clinical trials undersell the benefits of closed-loop systems, as patients in these trials are typically motivated and well managed at baseline?

    • How will closed-loop insulin delivery systems be priced?

      • What will payers demand for higher reimbursement? How will A1c improvement be perceived? How will less mild/moderate/severe hypoglycemia be perceived?

      • How will payers evaluate anecdotal patient experience and quality of life?

2. Dr. Boris Kovatchev (University of Virginia, Charlottesville, VA) presented a round up of four outpatient closed loop studies conducted in the last twelve months, using the DiAs artificial pancreas system. In the first outpatient study, hypoglycemia was reduced by 50%, in the second, in a summer camp setting, overnight hypoglycemia was eliminated; in the third, post-prandial excursion due to missed meal boluses was greatly diminished. These three studies set the scene well for a more detailed description of the fourth – an outpatient study using an overnight closed loop system designed to ‘reset’ the patient to a ‘flat’ 120 mg/dl glucose by 7am each morning. Personal testimony from Kelly Close, who took part in the study and wrote an article about it in diaTribe, was that the system dramatically improved her diabetes management during the five days in which she wore the system at night (“being able to fix hypoglycemia at night through CGM is a godsend; being 120 mg/dl through the night for multiple hours straight is far better due to the impact during the day”). While she acknowledged that she couldn’t prove the absence of a “trial effect,” she called the technology transformative, forecasting once approved, pending reimbursement, it would be disruptive for anyone at high risk of severe hypoglycemia. In the last four hours of the night, closed loop control delivered 82% time in target zone (80-150 mg/dl) compared with 39% for open loop control. As might be expected, control the next day was correlated with overnight control. The technology used in this study was a modified Android smartphone controller, Accu-Check Combo pump (Roche), and Dexcom G4 Platinum CGM. Over the last year, 86 patients experienced 60 days and 208 nights in outpatient closed-loop control with no serious adverse events and no system failures. Dr. Kovatchev commented, “Now it’s time to move on.” More details below.

3. Dr. Trang Ly (Stanford University, Stanford, CA) presented results from a pediatric inpatient (n=16) and a diabetes camp (n=21; n=50 nights) study testing an overnight closed loop system (OCL). The OCL consisted of a Medtronic Enlite CGM, MiniMed Revel 2.0, and a proportional-integral-derivative algorithm with insulin feedback (PID-IFB) operating on an Android smartphone. In the camp study, participants were randomized to OCL or SAP on the first night and then alternated therapy each night. Participants spent significantly more time in range (70-150 mg/dl) when on OCL (66%) than when on SAP (46%; p<0.005). When OCL lasted at least six hours, the results for hypoglycemia reduction were some of the best we can ever recall seeing – significantly less time spent <70 mg/dl (~1% vs. ~20%), <60 mg/dl (<1% vs. ~10%), and <50 mg/dl (~0% vs. ~5%). Participants’ mean glucose was higher on OCL than on SAP (128 mg/dl vs. 110 mg/dl), however this was statistically insignificant (p=0.129). Turning to the inpatient study, participants were admitted to the hospital for a 22-hour overnight period, and performed two periods of exercise to simulate camp conditions. The OCL was started in 13 of 16 nights (81%), and the percent time in range was 63%. Dr. Ly characterized the Enlite sensor as performing “reasonably well” in the inpatient study with a MARD of 14.1% (n=247). However, the Enlite did “poorer” in the camp study with a disappointing MARD of 19.2% (n=798).

4. In a session packed with new data and exciting results on the benefits of the artificial pancreas, Dr. Tadej Battelino (University Children’s Hospital, Ljubljana, Slovenia) outlined the impressive results from the 24-patient, six-week, at-home, crossover study of the “GlucoSitter” (DREAM 4). The 24 participants (12 adolescents) had an average baseline A1c of 7.9% and a mean age of 21 years. Compared to use of sensor-augmented pump (SAP) therapy, use of the MD-Logic AP improved time in range between 70-180 mg/dl (78% vs. 60%; p <0.05). Additionally, use of MD-Logic AP led to a significantly lower mean glucose level (146 mg/dl vs. 160 mg/dl; p <0.05), as well as a quarter as many total hypoglycemic events <60 mg/dl (24 vs. 100; p <0.05). Dr. Battelino also broke out the impressive results of patients prone to hypoglycemia (A1c <7.5%; n=8); within this subgroup, patients on closed loop experienced significantly fewer readings >240 mg/dl than patients in the SAP arm, fewer hypoglycemic events (<60 mg/dl), less AUC for readings <60 mg/dl, and significantly fewer readings <60 mg/dl (0.7% vs. 4%) (p <0.05 for all).

• With the exception of the treatment period, the study design was the same as the four-week study: a crossover, double-blind study with a two-week washout period. Participants in the MD-Logic AP arm used SAP therapy during the day and MD-Logic AP for only the evening meal and overnight period, with no additional living restrictions (including late-night snacking! Indeed, Dr. Phillip sometimes comments that “nighttime is not a sleepytime” for many patients). To make these results even more impressive, none of the patients in the MD-Logic AP arm experienced severe hypo adverse events, including severe hypoglycemia or DKA. Dr. Battelino expressed his confidence in the MD-Logic AP, highlighting that in the 1,200 (!) nights of the four-night and six-week trials there were no serious adverse events and the system reduced the duration, magnitude, and rate of nocturnal hypoglycemia.

5. Dr. Ed Damiano had three presentations on the Bionic Pancreas on the third day of ATTD (a morning plenary on the Summer Camp study and two talks in a Tandem-sponsored session dedicated to the Bionic Pancreas). Dr. Damiano expanded on the outpatient study results shared on Day #2 of ATTD – the bionic pancreas improved mean glucose vs. usual care (133 vs. 159 mg/dl [Beacon Hill] and 142 vs. 158-189 mg/dl [Summer Camp]) with a simultaneous reduction in mean time spent <60 mg/dl (1.5% vs. 3.7% and 1.3% vs. 2.2%). The results project to an A1c reduction of 0.9% in Beacon Hill (~6.2% A1c vs. ~7.1% with usual care), rising to an impressive 1.6% A1c reduction in the summer camp study (~6.6% A1c vs. an 8.2% at baseline). In his day three presentations, Dr. Damiano provided more detailed and nuanced discussion of the trials (including individual patient data); addressed concerns over glucagon during a fascinating Q&A in the Tandem session; and shared his vision for the next year. A four-center, 12-day study will start this spring and run until early 2015 – notably, participants (students/employees at MGH, UMass, UNC, Stanford) will have lots of latitude, including sleeping at home and going about their normal work weeks. In addition, Dr. Damiano and his team will conduct another five-day camp study this summer in 24 pre-adolescents (6-11 years old). Overall, some at ATTD seemed skeptical of his dual-hormone approach (e.g., Drs. Howard Zisser and Aaron Kowalski), mainly due to control algorithm concerns (e.g., glucagon enables much more aggressive insulin dosing, but what if that system fails?) and the absence of a stable glucagon. Regarding the latter, we’ll be interested to see how things progress with Xeris over the coming year, as they are farthest along to our knowledge.

6. Dr. Sybil McAuley (St. Vincent’s Hospital, University of Melbourne, Australia) presented the first human study (n=18 total) of Medtronic’s orthogonally redundant sensor (electrochemical + optical), a project supported by JDRF and the Helmsley Charitable Trust. The seven-day study compared the redundant sensor to a single electrochemical sensor (we assume Enlite); involved in-clinic (vs. YSI) and home portions (vs. SMBG); calibrated sensors four times per day; analyzed data after the study, but using a prospective algorithm; and evaluated several iterations of the redundant sensor over the course of six months (four patients wore the latest configuration). The system is clearly still a work in progress and the results were solid but not impressive. The biggest advantage of redundancy came in greater reliability – display time rose from 94.6% with the single electrochemical sensor (n=18) to 98.7% with the latest redundant sensor configuration (n=4). Dr. McAuley emphasized that this translates to one hour per day of additional display time. Overall accuracy improved marginally with the redundant sensor (from 13.9% to 12.3%), though day six accuracy improved markedly (16.2% to 10.6%). In-clinic MARD was 9.6%, an incremental improvement over 11.0% with the single electrochemical sensor. Based on these data, overall accuracy/reliability seems roughly on par with Dexcom’s G4 Platinum (MARD: 13.2%; 97% data capture). At this point, the major downsides of the redundant system are the need for four calibrations per day (this will go down over time, we assume) and the size of the on-body device (it appeared to be the size of a second-gen Insulet OmniPod). Future work will focus on continued development of the optical sensor, as well as use of multiple electrochemical sensors combined with the optical sensor. Study design details are in the appendix.

7. A JDRF workshop provided insulin pump and CGM companies with an opportunity to discuss their pipelines and visions for automating insulin delivery. Broadly speaking, presentations from Animas, Medtronic, and Tandem were on the less ambitious side, with no major surprises (and in the case of Animas and Medtronic, very general discussion and no specifics on products or timelines). Of the four presenting pump companies, Roche provided the most detailed and practical outline of plans to create an artificial pancreas. Notably, Roche is working through the steps in the JDRF model, implying that they intend to develop a predictive low glucose suspend product, an overnight control product, and eventually a fully closed loop product. In CGM, Dexcom and Medtronic gave very strong presentations, highlighting efforts to make their existing sensors better, especially through enhanced algorithms. Roche also discussed its in-development CGM, signaling what seems like a clear commitment to this project. Competition certainly fosters innovation, and we hope these five companies will jump in a big way and drive forward to close the loop. As we noted above, the gap between academic research and commercial products is significant; experienced companies will undoubtedly play a big role in some. Detailed summaries of each company’s presentation are discussed below.

8. A session entitled “Towards Product Development of the Artificial Pancreas” offered outstanding perspective from several closed-loop key opinion leaders: Drs. Moshe Phillip on nocturnal vs. 24-hour control (he believes nocturnal is the way to go short-term); Dr. Howard Zisser on product development (“cutting edge”) vs. basic research (“bleeding edge”) – the field can get to overnight closed-loop “fairly quickly,” while multihormonal control is a much bigger challenge; Drs. Roman Hovorka and Boris Kovatchev’s recommendations on priorities for AP development; and Dr. Eric Renard on the potential of closed-loop with intraperitoneal insulin. This was a very strong session on what the path forward looks like – all presentations are summarized below. 

9. In response to the question “can type 1 diabetes be cured?” Dr. Jay Skyler (University of Miami, FL) sighed, “I do not know.”  Dr. Skyler continued to express optimism for a number of approaches progressing through development but cautioned attendees that “hype” and “hope” only differ by one letter. Dr. Skyler expressed frustration with the media for its use of unwarranted headlines like “Breakthrough for Diabetes” that do not respect the emotional investment patients have in cure research. Examples of the transplantation research Dr. Skyler highlighted include xenotransplantation of pig islets, which is to move into clinical trials “soon.” Dr. Skyler noted that phase 1 trials of ViaCyte’s beta cell progenitor encapsulation device, VC-01, could enter clinical trials as early as this spring, which is in line with previous timelines from ViaCyte of 2014 (see our coverage of ViaCyte’s presentation at the 2013 Rachmiel Levine Diabetes and Obesity Symposium here). Additionally, phase 1 trials that are in planning will investigate transplanting a patient’s liver cells that have been transdifferentiated into beta cell (i.e., that have been induced to secrete insulin in a glucose-sensitive manner). Dr. Skyler also noted that research is being conducted on how to best house transplanted cells, and the possibility of enclosing the cells with agents that provide them oxygen or protect them against immune attack. Looking to possible pharmacotherapies, Dr. Skyler remarked that Dr. Doug Melton’s (Harvard University, Cambridge, MA) betatrophin (a peptide that induces beta cell proliferation in response to insulin resistance) could enter human trials “fairly soon.” Indeed, at the Joslin Diabetes Symposium, Dr. Skyler stated that he wanted to test betatrophin in TrialNet (see our coverage of Dr. Skyler’s comment and Dr. Melton’s presentation at the Symposium here).

• On Friday, Dr. Skyler also provided an update on the status of immunotherapy research for type 1 diabetes. Overall, the field appears to be pivoting from testing single agents to investigating combination approaches. Previous studies of the efficacy for single candidates to prevent (e.g., parenteral insulin, oral insulin, nicotinamide, and nasal insulin) type 1 diabetes or intervene in new onset patients (e.g., oral insulin, GAD, neurocrine altered peptide, cyclosporine, anti-CD3 monoclonal antibody, anti-CD20 monoclonal antibody, abatacept, canakinumab, anakinra, and thymoglobulin) have not demonstrated sustained efficacy. Thus, the growing consensus among KOLs over the past several years has been the need to shift research focus (and dollars) to combination approaches. Indeed, two of the planned or ongoing studies Dr. Skyler highlighted involved multiple agents: 1) a trial of sitagliptin and lansoprazole (results will hopefully be presented at ADA), and 2) low dose thymoglobulin and GCSF. For background, at ADA, Dr. Michael Haller discussed the methodology of the ATG-GSCK study and our coverage can be accessed here.

10. The rapid-fire ATTD Yearbook session provided a highly efficient review of the most notable papers published from July 2012 to June 2013 on topics including SMBG, CGM, pumps, closing the loop, insulin therapy, novel type 2 diabetes drugs, and even health IT. The yearbook, which is free to everyone (thanks to the meeting organizers) can be accessed here, and to give you a preview of what it contains we have listed each of the session’s 12 speakers and his/her presentation topic: Dr. Irl Hirsch (University of Washington, Seattle, WA) on SMBG; Dr. Satish Garg  (University of Colorado, Aurora, CO) on new therapies for diabetes management; Dr. Tadej Battelino (University Children’s Hospital, Ljubljana, Slovenia) on CGM; Dr. John Pickup (King’s College London School of Medicine, UK) on insulin pumps; Dr. Eyal Dassau (UCSB, Santa Barbara, CA) on closing the loop; Dr. Thomas Danne (Kinderkrankenhaus, Hannover, Germany) on new insulins and insulin therapy; Dr. Neal Kaufman (UCLA Schools of Medicine and Public Health, Los Angeles, CA) on using health information technology to prevent and treat diabetes; Dr. Walter Pories (East Carolina University, Greenville, NC) on metabolic surgery outside of bariatric surgery; Dr. Jay Skyler (University of Miami, Miami, FL) on immune intervention in type 1 diabetes; Dr. Michael Riddell (York University, Ontario, Canada) on physical activity and exercise; Dr. Shlomit Shalitin (Tel Aviv University, Tel Aviv, Israel) on diabetes technology and treatment in pediatrics; and Dr. Bruce Buckingham (Stanford University, Stanford, CA) on diabetes technology and the human factor.  

Honorable Mentions

1. Dexcom’s Dr. Claudia Graham presented a very interesting model designed as a “starting point” to support the cost-effectiveness benefits of CGM for patients with hypoglycemia unawareness. Results showed that for a US third party payer with 10 million members, giving CGM to all hypoglycemia unaware patients would save between $59 million and $321 million per year (factoring in the cost of CGM). The simple model had a lot of assumptions baked in (see below – all were cited from published literature), but we found the results fairly compelling and highly instructive. Dr. Graham concluded that we need a lot more data in the future: studies of CGM to examine reductions in severe hypoglycemia, more RCTs in the MDI population, more studies in type 2 diabetes, and regional costs of hypoglycemia and hospitalizations. We hope things are moving head on many of these fronts, especially because much of the reimbursement evidence for CGM dates back to early generation systems that were less accurate and reliable (e.g., the JDRF CGM trial). Full model assumptions are listed in the appendix.

2. As lunch was being served, Xeris Pharma’s Mr. Brett Newswanger (Director, Glucagon Products) provided an update on the company’s stabilized glucagon pipeline. Xeris’ leading diabetes-related product, the G-Pen (a stabilized glucagon rescue pen) is to be evaluated in a small pivotal trial (n=30) in 2H14, submitted to the FDA mid-2015, and if approved, launched in the US in early 2016. According to Xeris’ market research, 93% of patients will likely switch to the G-Pen from the current standard of care, reconstituted glucagon rescue pens. Eighty percent of patients reported being more likely to self-inject emergency glucagon using the G-Pen than currently available options. Turning to the rest of the pipeline, Xeris has had three INDs cleared by the FDA in less than six months (Mr. Newswanger pointed out that this is almost one IND per Xeris employee!). The first was for the G-Pen clinical study, which completed in 2013 and has top-line efficacy data available, demonstrating bioequivalence to the comparator Lilly glucagon. The second was for a phase 2a trial of the G-Pen Mini, which is to begin this month under the direction of PI Dr. Morey Haymond at Baylor College of Medicine (the G-Pen Mini can administer micro-doses of glucagon for the treatment of mild to moderate hypoglycemia). The third IND is for a phase 2a trial of G-Pump glucagon administered from an Insulet OmniPod (as a precursor for use in the bi-hormonal bionic pancreas) to begin next month under the supervision of PI Dr. Jessica Castle (OHSU, Portland, OR). Xeris has also submitted an orphan designation request to FDA, for treating congenital hyperinsulinism with its stabilized liquid glucagon in an Insulet OmniPod.

• Xeris has a multi-stage pump strategy, beginning with a relatively simple glucagon filled pump for the orphan indication of congenital hyperinsulinism. The G-Pump for people at high risk for a severe hypoglycemia event (a closed-loop glucagon-only pump and a CGM) would come next, followed by the G-Pump AP (a dual-chamber bihormonal pump containing glucagon and a currently available fast acting insulin analog) and the G-Pump Fast AP (a dual-chamber bihormonal pump containing glucagon with Xeris’ monomeric insulin (in Xeris’ pipeline) instead of a currently available insulin). The last step would combine Xeris’ glucagon with an insulin-symlin coformulation (also in Xeris’ pipeline). We imagine the latter two are still very early stage R&D projects at this point.

3. Dr. Eda Cengiz presented data from the first patient in Yale’s highly anticipated closed-loop trial testing Halozyme’s PH20 preadministration vs. a PH-20/insulin co-formulation vs. insulin only. The randomized crossover design includes three inpatient study days, one testing each approach. For the first patient, both PH20 approaches showed a solid improvement in postprandial glucose excursions after breakfast and dinner (see data table below). Lunch control was worse with PH20, a finding attributed to algorithm parameters not optimized for ultra-fast insulin. For upcoming patients, the plan is to fine-tune the algorithm to better reflect the faster PK/PD of insulin when PH20 is used. Importantly, the first patient tolerated both PH20 approaches without any site reactions. Dr. Cengiz called the early data “promising.” The team is targeting enrollment of 20 patients – on ClinicalTrials.gov, this study has an expected completion date of December 2015. We look forward to the full results, along with data from the 400-patient phase 4 CONSISTENT 1 trial testing Hylenex preadministration (expected at a medical meeting this year, which we assume will be ADA in June).

4. Dr. J.A. Gomez (University Javeriana, Bogota, Colombia) presented a retrospective, non-randomized, uncontrolled study suggesting that adult type 2 diabetes patients who have failed MDI benefit from using a sensor augmented pump (SAP). The study enrolled 28 adults who had previously been on MDI and had switched to a SAP for at least three months (17% male, mean age 60 years, and diabetes duration 18.1 years). Participants also had to have a history of self-reported moderate or severe hypoglycemia. On average, participants’ A1c dropped a statistically significant 0.9% when they switched from SAP to MDI (baseline of 8.6%; p=0.035). Dr. Gomez did not specify what the mean duration on SAP was at the time of this primary outcome measurement. Additionally, the frequency of severe hypoglycemia was reduced from 1.21 events/patient on MDI to only 0.11 events/patient on SAP. Concomitant to the declines in A1c and hypoglycemia was an insignificant mean weight gain of 1.56 kg (baseline of 72.2 kg; p=0.68), and a significant drop in the mean total daily insulin by 0.33 U/kg (baseline of 1.02 U/kg on MDI; p=0.0024). Looking more closely at patients’ use of the SAP, all participants wore the sensor and used the bolus wizard 80-100% of the time. The A plurality of patients (46.4%) used over five basals and a similar percentage (42.9%) used 3-5 basals. The mean number of capillary glucose measurements was 5.9. Notably, during Q&A, Dr. Gomez explained that in Colombia, people with type 2 diabetes are covered for CGM and pumps if they have failed MDI therapy with insulin analogs. Thus, she indicated that the study participants have been able to continue using the SAP after the study’s conclusion and that most have elected to do so.

Detailed Discussion and Commentary

Closing the Loop

Clinical Results From Transitional And Home Trials Of Outpatient Closed-Loop Control

Boris Kovatchev, PhD (University of Virginia, Charlottesville, VA, USA)

Dr. Kovatchev presented a round-up of four outpatient closed loop studies conducted in the last twelve months, using the DiAs artificial pancreas system. In the first outpatient study, hypoglycemia was reduced by 50%, in the second, in a summer camp setting, overnight hypoglycemia was eliminated; in the third, post-prandial excursion due to missed meal boluses was greatly diminished. These three studies set the scene well for a more detailed description of the fourth – an outpatient study using an overnight closed loop system designed to ‘reset’ the patient to a ‘flat’ 120 mg/dl glucose by 7am each morning. Personal testimony from Kelly Close, who took part in the study and wrote an article about it in diaTribe, was that the system worked perfectly. In the last four hours of the night, closed loop control delivered 82% time in target zone (80-150 mg/dl) compared with 39% for open loop control. As might be expected, control the next day was correlated with overnight control. The technology used in this study was a modified Android smartphone controller, Accu-Check Combo pump (Roche) and Dexcom G4 Platinum CGM. Over the last year, 86 patients experienced 60 days and 208 nights in outpatient closed loop control with no serious adverse events and no system failures. Dr. Kovatchev commented, “Now it’s time to move on”.

• Dr. Kovatchev summarized the results from three outpatient closed loop studies conducted in the last 12 months with the DiAs system. The studies were carried out by groups at UVA, Padova, Pavia, Montpellier, Sansum and Stanford, and supported in most part by the JDRF, Helmsley Trust and NIH.  The technology used was the DiAs smart phone  - a modified Android smart phone acting as a controller, the Dexcom G4 Platinum CGM sensor and receiver, and the Tandem t:slim pump, communicating to the DiAs phone via low energy Bluetooth. The studies achieved the following results:

  • Study 1: in 700 hours of outpatient closed loop control, hypoglycemia was reduced by about 50%.

  • Study 2: in a summer camp for children with diabetes, over 54 closed loop nights, there were no excursions below 60 mg/dl and time in range (70-150 mg/dl) increased from 52% to roughly 70%.

  • Study 3: in a study of missed meal boluses in adolescents with type 1 diabetes, the post-prandial excursion averages roughly 50 mg/dl lower with closed loop control two hours after peak. Variability is also substantially reduced compared to open loop.

• New data was presented for a fourth study of overnight closed loop control in an outpatient setting. In this work, the goal of the controller was to ‘reset’ patients by bringing them smoothly to a stable target of 120 mg/dl by 7am each day. Kelly Close participated in this trial and Dr. Kovatchev featured a picture of her eating a high fat/high carb meal in a restaurant, together with her recent article about her experience in diaTribe. The trial was a randomized crossover design with no meal restrictions, alcohol permitted, restrictions on intensive exercise, and driving restricted to 25 miles each day. The technology for this study used the DiAs system, the Dexcom G4 Platinum and the Roche Accu-Check Combo pump. Ten people with type 1 diabetes spent five nights in a residential facility, switching to closed loop control from 11pm-7am. Time in the target range (80-150 mg/dl) was the primary outcome.

• Study results were extremely impressive – between 3am-7am, time in target was 82% for closed loop, versus 39% in open loop. Over 49 nights of closed loop control, fasting glucose was 119±23 mg/dl versus 154±60 mg/dl in open loop – note the much tighter range, as well as the improved glucose, and a ‘perfect landing’ for the algorithm. Time below 70 mg/dl was 0.6% in closed loop versus 2% for open loop. As might be expected, control the next day was correlated with overnight control.  There were no adverse events.

• Dr. Kovatchev commented that “It’s time to move on!” – since over the last year they have studied 86 people for 60 days and 208 nights of closed loop control with no serious adverse events or system failures. All the studies showed superiority to open loop control.

2013 Summer Camp Trial of a Bihormonal Bionic Pancreas in Adolescents

Ed Damiano, PhD (Boston University, MA)

Much to the ire of punctual session moderator and longtime hypoglycemia thought leader Dr. Stephanie Amiel, Dr. Ed Damiano swooped into the morning session at the very last minute for his presentation on the 32-patient, five-day (24-hours/day), randomized, crossover Summer Camp study of the bionic pancreas – as a reminder, he shared the topline glucose results for the first time on Day #2. Overall mean glucose was 142 mg/dl with the bionic pancreas, a noticeable improvement over 158 mg/dl seen under highly supervised camp care and 189 mg/dl seen at baseline – this projects to an A1c of 6.6% with the Bionic Pancreas vs. 8.2% at baseline. Notably, the strong improvement in mean glucose was coupled with a significant reduction in mean time spent <60 mg/dl: 1.3% on the Bionic Pancreas vs. 2.2% in the control group. Said Dr. Damiano, “We are profoundly improving their glycemic control both in terms of hypoglycemia and mean glycemia.” Dr. Damiano emphasized that the control group (supervised camp care) was the best possible attention and management patients could receive – even still, it took a lot of effort and was not as good as the bionic pancreas. Dr. Damiano also shared new nuanced data that delved deeper into individual patients’ results – see below. The upcoming timeline is to begin a multi-center (MGH, Stanford, UMass, UNC) 12-day study in adults (starting this spring and running until early 2015 (“The FDA is very much on board”). Notably, patients will sleep at home and go about their daily work/routine (students/employees of the centers). The MGH/BU team will also conduct another camp study this summer in 24 pre-adolescent patients (6-11 years old).

• For the first time, Dr. Damiano showed a fascinating plot charting individual patients’ mean glucose (days 2-5) in the control condition vs. with the Bionic Pancreas. Notably, thirty-one out of the 32 patients had a mean glucose <168 mg/dl (the ADA goal of <7.5%) on the bionic pancreas. Most patients experienced a striking decline in mean glucose, with some going from over 215 mg/dl to <145 mg/dl. In the handful of patients who did see a rise in mean glucose after wearing the bionic pancreas (4/32 by our count), the system reduced high baseline hypoglycemia and still brought patients to goal (<168 mg/dl).

  • One patient did have a mean glucose exceeding 168 mg/dl on the bionic pancreas, a finding Dr. Damiano attributed to the algorithm. In most patients, it takes the algorithm ~18 hours to adapt to patients (initialization only requires weight) – this is why Dr. Damiano focuses on study data for days 2-5 (i.e., day one is not representative of how the system would perform ad infinitum). In the case of this single camper, it took the algorithm two days to adapt to the patient (instead of the typical 18 hours). Indeed, the patient’s average on days 3-5 was a solid 142 mg/dl, well below goal. The team has since modified the algorithm to allow the system to begin adapting immediately once it first comes online. 

• The randomized, crossover Summer Camp study compared five days on the bionic pancreas to five days of supervised camp care. The study took place at Camp Joslin (n=16 boys) and Clara Barton (n=16 girls) in 2013. Point of care capillary blood glucose checks occurred during the day and night (no venous glucose monitoring!). The same mobile platform was used as in Beacon Hill. Study staff and camp staff provided 24-hour, round-the-clock telemetry to monitor glycemia. A total of 160 days on the bionic pancreas were accumulated. For more information and interviews with trial participants, please see our detailed Closer Look write-up after we visited the study site this past summer.

Questions and Answers

Dr. Yogish Kudva (Mayo Clinic, Rochester, MN): Congrats on two very beautiful studies. Can you talk about the initialization based on body weight? It’s so different from other studies. And can you talk about adaptation in the first 18 hours and through days 2-5?

A: From the very beginning of our inpatient studies, we felt really strongly that the device needs to be independent of standard of care metrics for initialization. It had to be robust enough to adapt to large inter- and intra-subject variability. People can undergo tremendous variations in insulin requirements – inter-current illness, puberty, fluctuations in exercise routines, and other changes. We wanted a system that could adapt to those various demands, and we worked very hard to build a system that could do this. Weight is an objective starting point to scale the doses. It settles down in about 18 hours to automatically meet the insulin requirements of each individual. The device can diverge quite significantly from initialization. I think that work is now bearing fruit – we need that kind of adaptability to cope with the inter- and intra-subject variability of people with type 1 diabetes.

Physical Activity And Closed-Loop Control: Do We Need Additional Input Signals?

Ananda Basu, MD (Mayo Clinic, Rochester, MN, USA)

A challenging future hurdle for closed loop control is the incorporation of exercise into the algorithm. In this fascinating talk, Dr. Basu outlined the past, present and future of the impact of exercise on glucose control. The first key issue is that there is noticeable intra- and inter-personal variability in exercise, even with healthy people. But in people with diabetes this is magnified significantly. Heart rate and accelerometers are two methods of detecting exercise and measuring its intensity, and it appears that for low intensity activity, accelerometers work well. In silico studies predict that a population optimum response to moderate exercise would be to lower basal rates 50% for 90 minutes before exercise and by 30% during the exercise period – but anyone with diabetes (and a CGM) will tell you that there is definitely no ‘one size fits all’. Early attempts to fit exercise into closed loop controllers have resulted a system that uses accelerometers and galvanic skin response as well as CGM. Dr. Basu presented pilot results, which seemed to show reasonable control, but there is clearly a lot more work to do.

• More than 30 years ago, in 1982, Dr. Zinman reported work on closed loop control and exercise using the Biostator. At that time, the system was able to minimize glucose excursions during moderate exercise after breakfast (it was IV insulin) and it was noted that insulin needs declined.

• Studies of post-meal activity in subjects with and without diabetes showed noticeable reduction in post-prandial excursions, with significant inter-individual variability in both groups. However, the variability in type 1 diabetes is greatly magnified, demonstrating the major challenge of incorporating exercise into an artificial pancreas algorithm. The moral is not to watch TV after dinner, but instead go for a walk! Another closed loop study also demonstrated inter- and intra-subject variability, even with a consistent exercise schedule.

• At low intensity activity, Dr. Basu discovered that accelerometers are better than heart rate as a surrogate of energy expenditure. Having established this fact, correlation with CGM is the next step.

• Dr. Cobelli performed some in silico studies of moderate exercise in (virtual) people with diabetes, and found that the optimal population approach is to lower basal rates by 50% for 90 minutes before exercise and to lower basal by 30% during the exercise period itself. This work was informed by studies of insulin sensitivity in moderate intensity physical activity. In this work, Dr. Basu gave subjects a mixed meal then 45 minutes on the treadmill, and measured insulin, glucose and glucagon. There was an average 75% increase in insulin sensitivity in healthy people. In people with diabetes there was highly variable glucose during and after exercise. Lactate levels might be a way of addressing this variability and creating a personal baseline, and this is now under investigation.

• Studies of intra-individual variability with exercise showed that typically, insulin rises in people with diabetes during the exercise period, at the same time that glucagon response is attenuated with respect to healthy comparators. Insulin falls during exercise quite dramatically in healthy people, so the fact that it rises in people with diabetes is of interest. It may be related to the sub-cutaneous depot of insulin.

• A recent pilot study of exercise and closed loop control by Dr. Cinar et al showed evidence of reasonable glucose control, taking into account inputs from CGM, an accelerometer and galvanic skin response. This approach should be the future. The study was carried out with three patients and incorporated no announcement of meals or activity. This multi-sensor approach can be modeled in silico and then should lead to more extensive clinical trials.

Home, Sweet Home

Roman Hovorka, PhD (University of Cambridge, UK)

Dr. Roman Hovorka reprised the compelling 21-day overnight closed-loop data he first presented on Day #4 of IDF. The crossover study randomized adolescents (n=16) to home use of sensor augmented pump therapy for 21 days, a 2-3 week washout, and then overnight closed-loop for 21 days (or the reverse order). Notably, the study used no remote monitoring or connections to the cloud – patients did all troubleshooting on their own (the nocturnal system is quite simple, with just a single on/off button to initialize closed loop). Overall, the overnight data (11 pm-7 am) looked excellent – closed-loop control significantly improved time-in-target (70-144 mg/dl) from 46% to 68% (p<0.001), including nearly halving the number of nights with a glucose <63 mg/dl (17 to 10 nights; p=0.01). In new news, Dr. Hovorka shared some themes from qualitative phone interviews with participants, which reinforced the profound impact (and ripple effects) overnight closed loop has on patients’ psyche and glucose control. That said, the results also underscored the need to improve device design and form factor.

• A series of qualitative phone interviews assessed the psychological impact of closed loop control. Key positive themes included: reassurance; peace of mind; confidence; safety; improved diabetes control; “Sleep!”; “Not having to think about it”; time off from diabetes demands; and better control/feeling better during the first half of the day. Key negative themes included: calibration difficulties; size of equipment and of sensors; accuracy/trust; frustration when equipment fails; alarms (both positive and negative); and discomfort. Dr. Hovorka emphasized that many of these negative concerns are not specific to closed loop – they apply to diabetes technology more generally and underscore the need for smaller and better devices.

• We again appreciated hearing Dr. Hovorka’s views on why overnight closed loop is really needed, as closed loop insulin requirements varied widely from night to night in the study. In some patients, overnight insulin doses ranged from double the pre-programmed basal rate to half the pre-programmed basal rate! This suggests that basal insulin needs change very noticeably within subjects over a matter of days. At IDF, Dr. Hovorka put it well: “If you program something on the pump, it’s unlikely to be the optimal setting on the pump for the future. The past does not predict the future.”

• The Cambridge team has wrapped up two other home studies at this point – a three-week overnight study (n=24) at three centers (Cambridge, Kings College, Sheffield), as well as a seven-day, multicenter, multinational, 24/7 closed-loop study in adults (AP@Home – Cambridge, Profil, Graz). We cannot wait to hear results, which we assume could come at ADA or EASD 2014. Coming up, the team plans to do three-month (!) multicenter overnight studies in both children and adolescents. Dr. Hovorka did not say, though we wonder if this could qualify as a pivotal study for regulatory submission.

• Dr. Hovorka also showed a next-gen, more mobile version of the Cambridge team’s system, which will run the control algorithm on an Android phone. The system will retain the Abbott Navigator CGM/transmitter/receiver, a communication translator (USB to Bluetooth), and Dana insulin pump.

JDRF/ATTD Artificial Pancreas Systems – Delivering Products to People with Diabetes

Dexcom – CGM

Jake Leach (VP of R&D, Dexcom, San Diego, CA)

Mr. Jake Leach focused his presentation on G4AP, the new updated artificial pancreas algorithm for the G4 Platinum (same sensor and transmitter). Through work with Dr. Claudio Cobelli’s team, the new G4AP algorithm has improved MARD to 11.7% from 13.2% with the original G4 Platinum. The gain in accuracy comes from new de-noising and calibration algorithms that improve day one accuracy (16.8% to 14.7%) and hypoglycemia performance (18% to 14.8%). Notably, G4AP also reduces outlier sensors and tightens the sensor distribution – for one data set, 69% of G4 Platinum sensors had a MARD <15%, a number that improved to 80% with G4AP. Notably, Dexcom is now making the G4AP algorithm available for investigational use in closed-loop research, and AP investigators can reference an FDA approved IDE. Mr. Leach did not comment on the consumer rollout of the G4AP algorithm, which was first mentioned in Dexcom’s 1Q13 call but has not been discussed since. At the time, an FDA filing was expected in late 2013/early 2014. We assume other priorities – Dexcom Share, professional use/pediatric indications, pump partnerships with Tandem and Animas, Gen 5 – have superseded the new algorithm rollout.

Questions and Answers

Dr. Aaron Kowalski (JDRF, New York, NY): The fingerstick itself may contribute to some of this error. Could we see calibration free or intelligent calibration?

A: G4AP is starting to get to more intelligent calibration. With the original system, we relied on a fingerstick. But fingersticks aren’t perfect. Within the G4AP algorithm, we don't always put all the trust into fingersticks. We’re starting to get towards smart calibration. And all companies, including Dexcom, are working towards calibration-free sensors.

Medtronic – CGM

Anu Bansal (Medtronic Diabetes, Northridge, CA)

Ms. Anu Bansal provided a clear and comprehensive update on Medtronic’s CGM pipeline. The company is driving to sensors that can run closed-loop with a MARD <13%, outliers <5%, consensus A+B >99%, and reliability >95%. To get there, Medtronic is working on many fronts: 1) predictable and stable performance through better sensor design; 2) redundant sensors to overcome local body response; 3) intelligent diagnostics to diagnose and account for anomalies; and 4) connected care solutions to keep patients and care partners informed.

• How do you reduces outliers? Why do failures happen? Ms. Bansal identified key areas that hamper sensor performance: bleeding around the sensor at implant; glucose perfusion restriction; oxygen reduction with long-term body response; and mechanical events (e.g., sensor pullout). Medtronic is working on solutions to address all of these problems.

• Medtronic is working on improving sensor predictability through better design: minimizing the size of the implanted sensor; securing the sensor more reliably (new patch and transmitter designs); using non-traumatic forms of delivery (better designed insertion tools); and new sensor designs that match better match in vitro data to in vivo data.

• Medtronic’s CGM pipeline includes several redundant sensing efforts, all towards the goal of overcoming local sensor anomalies. One strategy puts multiple electrodes on the same sensor stick, while another uses orthogonally redundant sensing (optical + electrochemical). Ms. Bansal emphasized that optical and electrochemical sensing have different failures modes, meaning what affects one sensor does not affect the other. In one example, an electrochemical sensor’s 13.6% MARD and 84% display rate improved to 13.1% and 98% when optical sensing was added.

• Medtronic is also “very actively working” on intelligent algorithms that account for sensor health. The company has a new chip that can measure sensor health (e.g., impedance). When a sensor deviates from a “healthy response” (e.g., impedance within a certain expected range), the system could revert to/combine with another nearby sensor. In a specific example, one sensor’s MARD of 20.6% improved to 12.3% with use of intelligent diagnostics and redundancy. Medtronic is also working on smarter calibration CGM, which would go beyond just blindly taking calibrations.

• Ms. Bansal briefly mentioned Medtronic’s Connected Care device – as a reminder, this would send pump/CGM data to the cloud and smartphones. No timelines or further details were shared on the device.

Roche – CGM

Heino Eikmeier PhD (Roche Diabetes Care, Basel, Switzerland)

Roche Diagnostics is developing a CGM sensor and has presented data before on its performance. MARD figures sound very impressive (9%) but this is with respect to SMBG, so it’s difficult to compare with currently commercialized sensors.  In this presentation, Dr. Eikmeier from Roche described how this sensor is being improved for artificial pancreas applications.   

• Roche has a CGM sensor in development that is claimed to offer an MARD of 9% (vs. SMBG) in the 40-400 mg/dl range, and in the hypoglycemia range (<70 mg/dl) it has an MARD of 13%. Typically, MARD is expressed relative to YSI blood glucose measurement, and it’s not yet clear how the sensor performs compared to commercially available systems. For high rates of glucose increase (>3 mg/dl/min) the MARD falls to 15.9%, but for glucose decreases, even higher rates of change have MARDs in the 10% range.

• Roche is working to improve the CGM system for the artificial pancreas. Innovations include the use of two simultaneously operating sensors, the ability of the transmitter to store up to seven days worth of data (which is automatically uploaded when connected to the receiver) and connectivity via Bluetooth Low Energy (BLE).

• The artificial pancreas architecture is controller-centric. A separate universal controller device (which looks a bit like a smartphone) connects to the CGM sensor as well as a durable or patch pump. The controller handles CGM calibration, it incorporates a blood glucose meter, it contains the closed loop algorithms, data management and a bolus advisor, together with connectivity to PCs and other devices. The controller is also meant to be upgradeable in the field.

Roche – Artificial Pancreas

Sebastiaan La Bastide PhD (Roche Diabetes Care, Indianapolis, IN)

Dr. La Bastide gave us a detailed and practical outline of Roche’s plans to create an artificial pancreas. As we heard earlier, Roche are developing a CGM sensor that will connect with a durable pump or patch pump through a central ‘universal controller’ (which also contains a meter). He also mentioned a plan to ensure security and privacy both on the body and in the cloud. Roche are working through the steps in the JDRF model, implying that they intend to develop a predictive low glucose suspend product, an overnight control product and in due course a fully closed loop product. Dr. La Bastide mentioned some of the challenges to the roadmap, such as improving sensor performance, improving insulin absorption dynamics and obtaining approval for a Class III medical device. Although the presentation was short on study data, the plans seemed coherent, detailed and practical. Dr. La Bastide finished by mentioning Roche’s participation in the recent UVA overnight closed loop trials.

• Roche is taking a modular approach to constructing an artificial pancreas. Their system includes a durable pump, a patch pump, a CGM sensor and a universal controller at the center, which can communicate with smartphone apps and software in the cloud. The controller includes a blood glucose monitor and a bolus advisor and analysis tools. It’s also intended that the controller can be upgradeable in the field.

• Roche’s roadmap to the artificial pancreas is very similar to the classic JDRF plan. First they intend to implement alerts, then a predictive low glucose suspend, then an overnight control-to-range product, and finally fully closed loop with meal control.

• Dr. La Bastide had a few comments on connectivity. Roche’s approach is standards-based and will use low energy Bluetooth and Continua with a proprietary security layer that is subject to external security testing. Roche will also employ data encryption and de-personalization as part of their cyber-security and privacy plan.

• There are still many challenges to be overcome both at an individual and industry level. These include improving sensor accuracy and speed to obtain an accurate and low noise CGM, and improving insulin absorption dynamics (learnings from Accu-Chek Diaport  - an intraperitoneal insulin port – will be valuable here). The system and all components will be regulated as a Class III/PMA device with all the challenges that implies.

• Roche is now participating in a number of artificial pancreas initiatives. These include work with AP@Home, Reaction, GoCarb, the UVA DiAs system, and the JDRF Diaport project.

Animas – Artificial Pancreas

Ramakrishna Venugopalan PhD (Director Artificial Pancreas Initiative, Animas, Philadelphia, PA, USA)

In a rather corporate strategy-like presentation, Dr. Venugopalan described the approach of LifeScan/Animas in commercializing the artificial pancreas. The basic system is based on the pump at the center, which communicates directly with a Dexcom sensor, incorporates the algorithm and connects to the outside world, (such as OneTouch Reveal software). Dr. Venugopalan showed some data from feasibility studies, and informed us that these studies were now complete. Dr. Venugopalan also implied that the development of the algorithm was also essentially done. Moving into the transition phase, the design needs finalizing in a few areas – such as handling session startup/shutdown, consumable changes and potentially re-using sensors (which have a longer life).

• For Animas, a guiding principle for design is “less things to carry”. Their future system is an insulin pump that talks to a CGM sensor and is connected via an infusion set. The pump contains the algorithm and wirelessly communicates with OneTouch Reveal software. First generation technology is focused on unlocking user needs such as simplicity (too much information, too many devices), effectiveness (avoid highs and lows) and caregiver reassurance.

• Dr. Venugopalan showed data from an Animas closed-loop study of n=20 patients, using a hypoglycemia/hyperglycemia minimizer. Control appeared excellent, and the system also operated in the hypoglycemia minimizer mode. He also showed a predictive suspend example.

• The path to a commercial artificial pancreas goes from feasibility studies, to the transitional phase and then pivotal trials. Feasibility studies are now complete, using an Animas/Dexcom system. In the transitional phase, the pump will make a direct wireless connection with the sensor.

• In the transitional phase, work needs to be performed to finalize the design. Key areas include ‘de-risking’ closed loop session start up and shut down, consumable changes (infusion sets, insulin refills, CGM sensors), and the potential re-use of CGM sensors after a few days wear.

Medtronic – Artificial Pancreas

Lane Desborough, MSc (Chief Engineer, Insulin Delivery, Medtronic Diabetes, Northridge, CA)

The insightful Mr. Lane Desborough addressed what he sees as the biggest issue in artificial pancreas development – how do we take closed loop from a narrow use case (CRC, transitional studies) to an extremely broad use case (running on anyone with diabetes)? He had a compelling slide summarizing the magnitude/scale changes on the path to commercialization (see below). To get there, Medtronic is using lots of clinical/non-clinical data (e.g., CareLink), learning from other fields (e.g., aerospace, automotive, finance), and employing modeling/simulation (e.g., virtual patients). Hearing the company’s philosophy was definitely helpful and illuminating, though it was somewhat surprising not to hear timelines/details on specific products, including the upcoming MiniMed 640G predictive low glucose management pump (set to launch internationally by July 30). 

Tandem – Artificial Pancreas

Aymeric Lecanu-Fayet (Tandem Diabetes Care, San Diego, CA)

Mr. Aymeric Lecanu-Fayet’s presentation focused on the company’s near-term pipeline: a t:slim with Dexcom G4 Platinum CGM integration (to be submitted to the FDA “later this year”); the 480-unit capacity t:flex (a larger capacity cartridge; same t:slim pump); the discreet, Cellnovo-like t:sport pump (a slimmer version of the t:slim meant to be worn under the clothes with a standard infusion set and a wireless, cloud-connected handheld controller); and the t:dual hormone version of the t:slim in partnership with JDRF (no timing given). Mr. Lecanu-Fayet also highlighted Tandem’s numerous artificial pancreas partnerships (UVa, Stanford, UCSB, Oregon, University of Padova, University of Montpellier), headlined by Drs. Ed Damiano and Steven Russell’s work at Boston University/MGH (“the project we are most proud of”). Overall, we thought this presentation did a good job of laying out Tandem’s upcoming pipeline, though we had hoped to hear more ambition on the closed-loop front (e.g., a predictive low glucose suspend or overnight closed-loop project). As a new player in the pump field, we think Tandem would stand to gain a lot by investing full on in automated insulin delivery technology.

The Bionic Pancreas – A Bi-Hormonal Close-Loop Artificial Pancreas Industry Symposium (Sponsored by Tandem Diabetes Care)

The Bionic Pancreas in the City – The Beacon Hill Study

Ed Damiano, PhD (Boston University, MA)

Dr. Ed Damiano discussed the outpatient, randomized, crossover Beacon Hill study of the Bionic Pancreas (n=20), sharing new data on top of the brief results shared on Day #2 of ATTD (reviewed below). Dr. Damiano discussed individual patient data (highly valuable), which showed all patients reaching the ADA goal of a mean glucose <154 mg/dl (~7% A1c). Dr. Damiano also addressed what he believes is a major game changing feature of the bionic pancreas: nighttime control (a big theme of ATTD 2014). Pointing to the overnight, non-closed-loop control period of a 24-hour modal day chart, he said, “I see fear on this plot…there is lots of suffering coming from the nighttime. People are afraid to go to bed at night.” Indeed, the picture clearly illustrated the huge benefits of the bionic pancreas overnight: a lower mean glucose, less glycemic variability, and improved time in range (time spent <180 mg/dl: ~90% on bionic pancreas vs. ~60% with usual care; time spent <70 mg/dl: 3-4% on bionic pancreas vs. ~10% with usual care).

• “Every subject came under goal for therapy. That’s pretty compelling.” Similar to his presentation on the Summer Camp study, Dr. Damiano showed a plot charting individual patients’ mean glucose (days 2-5) in the control condition vs. with the Bionic Pancreas. Notably, all 20 Beacon Hill patients had a mean glucose <154 mg/dl (the ADA goal of <7%) on the bionic pancreas. Most patients experienced a striking decline in mean glucose, with one patient going from an average of 215 mg/dl to <120 mg/dl on the bionic pancreas. From the chart, only one patient saw a rise in mean glucose after wearing the bionic pancreas (approximately +10 mg/dl), and in that case, the system reduced a high level of hypoglycemia. Under usual care, there was also a wide dispersion in mean glucose between patients, with some having a mean under usual care of >210 mg/dl and others at <120 mg/dl. After wearing the bionic pancreas, mean glucose levels converged to 115-153 mg/dl.

• “This is diabetes without numbers.” Dr. Damiano emphasized the robustness of the bionic pancreas algorithm, which only requires weight for initialization, adapts over time, and does not mandate pre-meal priming boluses. Patients can optionally announce meals to the system, but they only enter qualitative information using a slider – is this “more,” “about the same,” or “less” than the amount of carbs that you typically eat? In Beacon Hill, patients could use the meal announcement feature if they desired, but were not reminded about it if they forgot. One trial participant went a full 40 hours without meal announcement, and mean glucose was 130 mg/dl over that time (0% of the time <60 mg/dl). Said Dr. Damiano, “This person had near normal glucose control without having to think at all about their diabetes management…It’s really encouraging.”

• The randomized, crossover Beacon Hill study compared five days on the bionic pancreas to five days of “usual care” (what a patient would normally do, though with the addition of blinded CGM). The study included 20 adult type 1 patients >21 years. The bionic pancreas mobile platform consisted of two Tandem t:slim pumps (insulin and glucagon), a Dexcom G4 Platinum sensor and transmitter, and an iPhone 4S controller. Patients had free run of a three-square mile area of the Boston peninsula. Point of care capillary blood glucose checks occurred during the day via 1:1 nursing. At night, patients slept in a hotel with venous blood glucose monitoring and 1:2 nursing. A total of 100 days on the bionic pancreas were accumulated. For more information, read our diaTribe test drive on the Beacon Hill study.

2013 Summer Camp Study

Ed Damiano, PhD (Boston University, MA)

Dr. Damiano reviewed the study design and results of his summer camp study, described elsewhere in this report under the session title, “Closing the Loop.” In this follow-up talk, Dr. Damiano emphasized that a key difference between the Summer Camp and Beacon Hill studies– the camp study had “tremendous parity” between the Bionic Pancreas and usual care arms, while Beacon Hill didn’t have as much parity (the “usual care” control arm took place in patients’ homes, with data measured via blinded CGM). Dr. Damiano emphasized that the Summer Camp study participants “had a lot of support and a lot of help at camp,” and even still the Bionic Pancreas improved glycemic control and reduce hypoglycemia. Said Dr. Damiano, “The bionic pancreas is like better than diabetes camp in your pocket.”

The Bionic Woman: Gentlemen, We Can Rebuild Her [Pancreas]

Kelly Close, MBA (The diaTribe Foundation, San Francisco, CA)

Close Concerns’ own Kelly Close is a patient who took part in the outpatient Beacon Hill trials of the Bionic Pancreas. She experienced dramatically improved glucose management as part of the trial and also expressed the major surprise of feeling the strong effects of lifting the mental burden of diabetes. She is a patient advocate and is a strong proponent of the bionic pancreas, as it has the potential to get patients much closer to normal control, thereby lowering short-term and long-term healthcare system costs stemming from severe hypoglycemia and prolonged years of hyperglycemia. Ms. Close stated that her hope is that the artificial pancreas could now move to more extended outpatient trials, since excellent automated control is now possible both at night and during the daytime. She lobbied for patients willing to take on greater risk to be able to do this, noting that there will never be a “zero-risk” device, but for some patients, the dangers of insulin are so great that they would be glad to try to move forward research and take on more risk themselves in the trials.

• Kelly Close is a patient who has tried almost every diabetes device over the last 25 years. She also is a journalist that covers diabetes and obesity at Close Concerns. She is also a Board Member of the diaTribe Foundation, which publishes diaTribe, an online resource for patient education.

• Having diabetes is like walking a tightrope between hyperglycemia and hypoglycemia, she emphasized, which in turn imposes a psychosocial burden. People with diabetes also have to make many decisions about their care, attempting to take into account food, exercise, stress and many other factors that can swing glucose one way or another.

• The goal of patients with access to good therapy and a good healthcare team is to reach target A1c, but she argued that in particular, it would be optimal if the diabetes community could focus on a ‘high quality’ A1c with maximal time in the target zone rather than just a good A1c number that might be driven by hypoglycemia. She showed data noting that it’s possible for three people to have an A1c of 7.0%, and yet, have widely varying time in range (see below). In any case, current A1c targets are not even close to normoglycemia, she said – this is backed up by T1D Exchange data showing that average A1cs hover closer to 8-8.5% in the US (at highly praised diabetes centers) rather than at 6.5-7.0% goals. The bionic pancreas offers the opportunity to get much closer to normal, she said, and she felt that systems would improve over time “but let’s get one approved”.

• The experience with the bionic pancreas in the Beacon Hill trial felt very uplifting, she emphasized. Not only was glucose control excellent, but not having the constant mental burden of the disease (the calculating, the anxiety, the changes in mood) was surprisingly moving for her, her family, her team, and friends. 

• Ms. Close argued that the artificial pancreas is ready to move on to extended outpatient trials. Excellent overnight control has been demonstrated (>90%), and during the daytime >70% time in target is possible. Algorithms are improving and devices are becoming increasingly portable.

• She also highlighted that future challenges such as form factor can be addressed over time; initial systems will be “clunky” and that’s fine – patients that don’t want them don’t have to get them, if they are not at major risk of severe hypo.

• Other challenges like regulatory hurdles, reimbursement, the need for faster insulin, and access to a stable glucagon (for dual hormone systems) will all need to be addressed – getting deeper into trials will enable that. Ms. Close emphasized that we must not make perfect the enemy of the good in terms of trials moving forward– risk is present, but many patients are happy to take it on. Hyperglycemia and hypoglycemia often cannot be seen, but they still puts a tremendous strain on patients and the healthcare system.

Panel Discussion

Dr. Aaron Kowalski (JDRF, New York, NY): Fantastic progress, congrats. Patient with diabetes use insulin correction factors. Do you generalize what a glucagon unit does in terms of raising blood glucose?

Dr. Damiano: Yesterday, Steve Edelman showed blood glucose correction with insulin based on CGM data. It’s not one answer. A three-unit correction is different if you’re 220 mg/dl and flat vs. falling with a double down arrow. Blood glucose correction with insulin is based on glucose level and rate of change. Glucagon is similar to that. If you have lots of insulin pending, the amount of glucagon needed will be greater than if insulin is long gone. It’s very difficult to say how much glucagon one would need to raise glucose 30-40 mg/dl. For a 70 kg adult with only basal insulin on board, a 20-30 microgram bolus (i.e., 2-3% of a rescue dose) could raise your blood glucose 20-30 mg/dl. On the other hand, if there is a lot of insulin pending, such a dose of glucagon might only flatten your blood glucose or slow the rate of decline.

Dr. Kowalski: There is debate in the dual hormone glucagon closed loop world – how much glucagon to use? I fall onto the side of debate that thinks this is too much. When I look at the graphs, you are giving glucagon after every meal. We know that pump sets fail, maybe 10% of the time. I’m a little concerned that we’re overdosing insulin to control postprandial glucose because of the slow insulin kinetics, and then counting on glucagon to pull us back. In terms of exercise, I understand why we’d be counting on glucagon. Can you talk about your more aggressive use of glucagon?

Dr. Damiano: Typically we use about 0.8 mg per day of glucagon – that’s 80% of a rescue dose. That’s a lot of glucagon relative to a more sedentary environment, where we use 0.4-0.5 mg per day. We used lots of glucagon because these subjects were very active. In Beacon Hill, they got lots of exercise. Only a few subjects worked. So they were more physically active than normal, and glucagon usage was higher than what people normally see. In summer camp, the kids were ridiculously active. You are seeing lots of glucagon partly because of level of activity.

Ms. Close: They really encouraged us to eat crazy food and exercise all the time. They really wanted us to stretch the system. It’s not how we would normally behave. I would assume you guys can make decisions about where to put that safety level.

Dr. Damiano: We’re going to have the ability in the system to control that. If you see that you are using too much glucagon, you will have the opportunity to back off the whole system. And it’s not only glucagon dosing, but the set point as well. We’re building in an additional preference to say, “I don’t need to have an average of 130 mg/dl. I’d be happy with a 6.8% A1c, which would use less glucagon. Patients will have that flexibility with the system. Such a feature will raise the whole algorithm’s infrastructure to a higher target glucose.

Dr. Moshe Phillip: Are repeated injections of glucagon a concern? Is it possible you might not get the same response, especially with exercise?

Dr. Damiano: We never saw tachyphylaxis of glucagon. It’s probably because we were never really depleting glycogen storage. The controller spreads that glucagon delivery over the course of the day. Even after a period of long fasting, when glycogen storage is most compromised, we saw no diminished effect of glucagon.

Mr. Brandon Arbiter (Tidepool, Palo Alto, CA): This is pretty amazing stuff. I’m really hopeful. Kelly, I understand that when you were doing Beacon Hill, you had a nurse following you. What kind of interactions did you have with the nurse? Did anything go wrong?

Ms. Close: I thought it would be quite different from what it was. Courtney and the other nurses really became good friends. Nothing went wrong. The study was only five days. That study design maybe was put together to reassure the FDA. For me, nothing went wrong.

Dr. Damiano: We did have connectivity issues – Bluetooth problems between the iPhone and Tandem pumps. The CGM side was more reliable. We think we know what those Bluetooth problems were. On the other hand, we’re not going to have a Bluetooth connection to the pumps in the final device. It will all be integrated into one system. With the current iPhone system, we lost sight of the insulin or glucagon pump ~7% of the time  – that’s a lot. That wouldn’t happen in a platform where it’s all built in. That’s an artifact of this rubber band/pin hook system. Mostly wireless connectivity issues were the big challenges.

Towards Product Development of the Artificial Pancreas

Nocturnal Control Only vs. 24 Hours

Moshe Phillip, MD (Schneider Children's Medical Center, Petah Tikvah, Israel)

Dr. Moshe Phillip discussed the pros and cons of nocturnal vs. 24-hour closed-loop control, making a strong case that nocturnal closed-loop control is clearly the way to go right now. He showed examples of 24-hour closed-loop studies, noting that most of the efficacy benefit actually came from the overnight period (and in the case of Dr. Hovorka’s 21-day home study presented on Day #3, overnight closed-loop alone actually improved overall 24-hour control). Further, Dr. Phillip contended that severe hypoglycemia is still a big problem (Katz et al., Diabet Med 2012), one that can be helped through nocturnal closed-loop control (most severe event occur at night). He also used lots of data from the DREAM consortium to support the benefits of nocturnal control: reduced hypoglycemia, tighter glycemic control, lower mean glucose, less glycemic variability, better morning blood glucose values, and perhaps an overall effect on glycemic control. Following an overnight system (“GlucoSitter”), the DREAM plan is to add a daytime “pump advisor” (DREAM 5) to help patients “cope with different challenges faced during the day.” The ultimate goal – 24-hour closed-loop control – would come after that.

Questions and Answers

Dr. Yogish Kudva (Mayo Clinic, Rochester, MN): Have you had a chance to look at the various cohorts and the effect of nighttime control on daytime control?

A: From the two published studies, and also Dr. Hovorka yesterday, our impression is that 24-hour control is better even in patients that had closed-loop only at night. We didn’t analyze the subtypes – these are very recent results.

Dr. Roman Hovorka (University of Cambridge, UK): The severe hypoglycemia data is really old – DCCT. Things have moved since then. Are you aware of newer data that looks into severe hypoglycemia?

A: The data is not very old. The Boston and Houston data is very fresh data (Katz et al., Diabet Med 2012). There were an amazing number of severe hypoglycemia episodes. Personally, I was extremely surprised. That was a prospective study.

Dr. Hans DeVries (Academic Medical Center, Amsterdam, Netherlands): Starting from today and envisioning a product on the market, what is the largest barrier?

Dr. Phillip: I don’t think there are too many barriers. I think that we and other groups are moving fast. It’s the commitment of hardware companies that is needed. I think that with their commitment, we can close the regulatory process and bring a safe product to market to help patients and parents get through the night safely. The phrase “GlucoSitter” tries to grasp that notion. I don’t know one family that sleeps quietly during the night. Those activities that are done around the world right now are getting closer to the first product, and I believe the first product will be night.

AP Product Development (the cutting edge) vs. AP basic research (the bleeding edge)

Howard Zisser, MD (Medical Director, Insulet, Bedford, MA)

Dr. Howard Zisser, Insulet’s new Medical Director, gave a practical, thought-provoking, and philosophical talk on the balance between AP product development (“the cutting edge”) vs. AP Basic research (“the bleeding edge”). Overall, he seemed to conclude that both are needed – basic research is necessary to push the boundaries and see what is possible, while product development is ultimately needed to get products to patients. He shared his views on where current efforts fall, calling overnight control something the field could get to “fairly quickly” (this has unquestionably been a major theme of ATTD 2014). Conversely, Dr. Zisser was less optimistic on dual-hormone control and fully closed loop in the near-term, given the significant challenges each faces (e.g., stable glucagon, delayed insulin action). He noted that “the artificial pancreas is not one thing,” a point researchers readily recognize but consumers may not. Instead, Dr. Zisser believes “everyone with diabetes probably has a different conception” of what an artificial pancreas would be. Certainly, this sort of terminology will be critical in setting patient expectations as devices come to market.

Questions and Answers

Dr. Yogish Kudva (Mayo Clinic, Rochester, MN): On the bleeding edge side, you showed two engineering approaches and three biologic approaches. Can you talk about the translation of those to product development, as well as the effect sizes?

A: For effect size, I would defer to my local mathematician. These are things that get started as little ideas. You start down the road and hit up against the wall – how can we do better? These help us leap over the wall and do things in a carefully constructed, safe environment. A lot of these things aren’t on the market yet. Until they are available, they remain a research project.

Dr. Roman Hovorka (University of Cambridge, UK): Do you think we should involve more patient views to help us decide where to go?

A: I think so. I think it’s always helpful when designing any product to get input from the end users. You run into all kinds of problems when you just develop something and expect people to like it.

Priorities for AP development: From CRC and Home Prototypes to Products

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

Dr. Hovorka, one of the most experienced researchers in the artificial pancreas field, gave his candid perspective on the right way to approach the development of the closed loop, from researchers entering the field to future commercialization of the technology. ‘Manual’ closed loop (transferring data by hand to/from devices and computer) has many advantages for feasibility studies, since it has a much lower regulatory and technical burden and is much more flexible. Building or modifying devices is not efficient, although a reliance on manufacturers can cause delays, as happened with his Abbott Florence system. The question of whether to put the algorithm in the pump or on a separate handheld device seems to fall in favor of a separate device, probably in the form of tightly coupled consumer electronics. The best intended use of the system is for overnight closed loop control – a theme we have heard many times at this conference. Adding faster acting insulins would be of high benefit, although also adding risk. The incremental benefit of adding glucagon is not yet clear to Dr. Hovorka. Adding accelerometers or a heating pad were low risk/low benefit add-ons. Finally, remote monitoring was considered not beneficial to mitigate safety concerns, because of the effort required to monitor the data and patient pushback. However, connectivity is valuable for storing and managing data.

• ‘Manual’ closed loop (having a nurse transfer data to/from devices and computer) is really beneficial for feasibility studies. The alternative, automating and connecting the hardware, causes a tremendous amount of work – on the technical side and with regulatory. It also takes time – a change to the algorithm can cause a 60 day delay in the UK. The ‘manual’ method is only considered decision support and has a much lower regulatory burden. Automated methods also need a technician to be available during studies and possess liability concerns.

• In developing prototype devices for outpatient studies, the best option is to use unmodified CGM and pumps, although they have some practical issues and they impose dependencies on the manufacturers. Dr. Hovorka showed the Florence family of home prototypes, which incorporates the Abbott Navigator. He took the decision to use unmodified CGM and pumps, which made life a lot easier. However he cautioned that Bluetooth connectivity can be unreliable, and providing power to the devices can also cause headaches. He also noted that third party dependencies can be an issue – he experienced problems with long lead times from companies and with discontinued devices.

• The algorithm can be integrated with the pump or placed on a handheld device – the best option seems to be the handheld, since you can innovate faster and incorporate connectivity. If the algorithm is placed on the pump, we have improved on-body connectivity and a smaller footprint (and one device less). But a handheld device allows faster innovation, better convenience, more privacy and remote connectivity (pumps don’t typically connect to the internet). Handheld devices can be tightly coupled medical devices (like the OmniPod PDM), tightly coupled consumer electronics, or an unrestricted use smartphone.

• Overnight closed loop control is currently the lowest risk pathway. Low glucose suspend is already available, predicted suspend is coming, and treat to range is the next stage in the JDRF roadmap. However Dr. Hovorka supports moving next to overnight control, which is getting good results, is encouraged by the FDA and captures a lot of value. The 24 hour fully closed loop is not yet ready for primetime.

• In terms of ‘add-ons’, the highest benefit is the use of faster analog insulins, even though they add high risk. Adding glucagon is high risk and high hassle, but the incremental benefit over the insulin-only artificial pancreas has not been clearly demonstrated. Adding a single port (for CGM and insulin infusion) is also high risk (although it results in higher patient acceptability). Adding features like accelerometers or a heating pad is considered to be low risk/low benefit.

• Finally, remote monitoring is good for data capture, but not for safety mitigation. As a product feature, parents/caregivers are in favor, but the patients want privacy. As a safety feature, it’s very difficult to enforce continuously because of the labor required to monitor. But for data capture it could have benefits for the physician. A good example is the Cellnovo product, which has just reached the market in the UK.

Questions and Answers

Dr. Hans De Vries (Academic Medical Center, Amsterdam, Netherlands): The biggest issue seems to be getting the commitment of the hardware companies. What do you think we can do about this?

A: We need a manufacturer with a willingness to lock the system and say ‘This is it, we are going ahead.’ Somebody needs to make the call and say “We are doing it now!”

Modular Architecture Of Closed-Loop Control – The Blueprint For Sequential Product Development Of The Artificial Pancreas

Boris Kovatchev, PhD (University of Virginia, Charlottesville, VA)

Dr. Kovatchev’s ‘blueprint’ for the artificial pancreas consisted of several recommendations, as with Dr. Hovorka’s presentation earlier. On hardware configuration, he recommended thinking of the artificial pancreas as a ‘mobile medical network’ and using consumer electronics running a medical grade operating system to create the ‘AP Hub’. Dr. Kovatchev recommended a modular approach to closed loop control. This allows the system to gradually add functionality, getting approval at each stage, building on aspects that have already been shown to be effective and safe. Of course, the foundation layer is safety. Finally the system should have a multi-function scalable operation, being able to add functionality depending on what hardware is connected and ‘degrade gracefully’ if components fail. Dr. Kovatchev asserted that all these design elements were already embodied in the DiAs system, which is currently being used in closed-loop outpatient trials.

• Dr. Kovatchev organized his ‘blueprint’ for artificial pancreas product design into a set of decisions:

  • Decision #1: Hardware configuration. The artificial pancreas can be thought of as a kind of a mobile medical network. At the center is an ‘AP Hub’ which should be deployed on consumer electronics and contains a graphical user interface and the control algorithm. Other aspects of the system, such as safety, can be distributed across components, requiring devices that can talk to each other well.

  • Decision #2: System Claim. As we move from “adjunct” to a ‘replacement’ claim for the artificial pancreas, the system should be modular so it can be constructed sequentially, getting regulatory approval at each stage as functionality is expanded. The bottom layer should be safety (working continuously to prevent hypoglycemia, to improve safety of the pump and the accuracy of the sensor). The safety part has an “adjunct” claim – it works in addition to normal basal-bolus therapy, intervening only if risk for hypoglycemia is detected. Adding a post-meal correction module results in control within range, which also has an adjunct claim, minimizing hypo- and hyperglycemia. A fine-tuning basal rate module and control-to-target would have a “replacement claim” – the basal therapy would be completely replaced by the closed loop.

  • Decision #3: Algorithm Scalability and Product Sequence. The system should be a multi-function device that controls various actions based on the peripherals available. If just a CGM is available, it can enable trending and remote monitoring, if just a pump were available it handles safety and hypoglycemia prevention. In a more complex artificial pancreas system, this would be the failure mode if a sensor stops working (i.e. it degrades gracefully).

• The DiAs device is designed with this modular approach, to activate a body sensor network, to be scalable, to have local/global modes of operation and to connect to a remote a server through a WiFi or cellular network. It runs on a modified smartphone, and is being used in outpatient closed loop trials.

Questions and Answers

Dr. Hans De Vries: Do you think that a product with some kind of closed loop control, offered as a replacement device, will be launched first in the EU or the USA?

A: Experience says it will be first in Europe.

Artificial Pancreas Using Intra-peritoneal Insulin Delivery: Why Should it Be Developed and How to Move Toward Market Approval

Eric Renard, MD, PhD (Montpellier University, France)

Dr. Eric Renard optimistically discussed intraperitoneal insulin delivery, asserting that it is safer and more effective than subcutaneous insulin delivery (much faster insulin action, restoration of glucagon response to hypoglycemia, better reproducibility, reduced risk of catheter failure, etc.). His final point argued that intraperitoneal insulin delivery is “easier to use in everyday life” (fewer devices to carry around), though his slides did not make a very persuasive case – indeed, the graphic abdominal view of an implanted pump was not very appealing from a patient perspective and no mention was made of cost or related hassles with surgeries, problems that could come up that would require additional surgeries, etc. Currently, the implantable pump is limited to 412 patients in Europe (315 in France), though an expansion is planned in Germany. Overall, Dr. Renard believes intraperitoneal insulin delivery is an “underestimated way of closing the loop,” and more research and development is needed on this front. He pointed to Roche and Medtronic, who each have potential to make closed-loop intraperitoneal systems.  

Artificial Pancreas Data Club Open Forum

Acceleration of Insulin Action by Hyaluronidase During Closed-Loop Therapy

Eda Cengiz, MD (Yale University, New Haven, CT)

Dr. Eda Cengiz presented data from the first patient in Yale’s highly anticipated closed-loop trial testing Halozyme’s PH20 preadministration vs. a PH-20/insulin co-formulation vs. insulin only. The randomized crossover design includes three inpatient study days, one testing each approach. For the first patient, both PH20 approaches showed a big improvement in postprandial glucose excursions after breakfast and dinner (see below). Lunch control was worse with PH20, a finding attributed to algorithm parameters not optimized for ultra-fast insulin. For upcoming patients, the plan is to fine tune the algorithm to better reflect the faster PK/PD of insulin when PH20 is used. Importantly, the first patient tolerated both PH20 approaches without any site reactions. Dr. Cengiz called the early data “promising.” The team is targeting enrollment of 20 patients – on ClinicalTrials.gov, this study has an expected completion date of December 2015. We look forward to the full results, along with data from the 400-patient phase 4 CONSISTENT 1 trial testing Hylenex preadministration (expected at a medical meeting this year, which we assume will be ADA in June).

Questions and Answers

Q: Will you retune the algorithm before all 20 patients are completed?

A: Yes, I think we will retune the algorithm before we complete the study.

Q: What is the issue with the lunch? Was it a different in what they are? Or was it that lunch came after breakfast?

A: Very good question. I just received the meal content for the subjects. We need to look into the differences in fat and carbohydrate. It almost looked like it was the insulin on board. You saw this beautiful drop at breakfast. The system was shutting down and not delivering enough insulin before lunch. I have a feeling it might just be the insulin delivery.

Oral Presentations

Overnight Closed-Loop Control With A Proportional-Integral-Derivative Based Algorithm In Children And Adolescents With Type 1 Diabetes At Diabetes Camp

Trang Ly, MBBS, DCH (Stanford University, Palo Alto, CA)

Dr. Trang Ly presented results from a pediatric inpatient (n=16) and diabetes camp (n=21) study testing an overnight closed loop system (OCL). The OCL consisted of a Medtronic Enlite CGM, MiniMed 530G, and a proportional-integral-derivative algorithm with insulin feedback (PID-IFB) operating on an Android smartphone. In the camp study, participants (mean A1c = 7.9%) were randomized to OCL or SAP on the first night and then alternated therapy each night. Patients spent significantly more time in range (70-150 mg/dl) when on OCL (66%) than when on SAP (46%; p<0.005). When OCL lasted at least six hours the results for hypoglycemia were some of the best we can ever recall seeing – significantly less time spent <70 mg/dl (~1% vs. ~20%), <60 mg/dl (<1% vs. ~10%), and 50 mg/dl (~0% vs. ~5%). Participants’ mean glucose trended higher on OCL than on SAP (128 mg/dl vs. 110 mg/dl), though this was statistically insignificant (p=0.129). Turning to the inpatient study, participants were admitted to the hospital for a 22-hour overnight period and performed two periods of exercise to simulate camp conditions. The OCL was activated in 13 of 16 nights (81%), and the percent of time in range was 63%. Dr. Ly characterized the Enlite sensor as performing “reasonably well” in the inpatient study with a MARD of 14.1% (n=247); however, the Enlite did “poorer” in the camp study with a MARD of 19.2% (n=798).

• The diabetes camp study was conducted over two, one-week week sessions at Camp Conrad-Chinnock in California. Participants were randomized to either OCL or SAP on the study’s first night and alternated systems each subsequent night. In order for the OCL to be initiated, a person’s blood glucose level had to be between 70 and 350 mg/dl, and the sensor error had to be <20%. Meter checks were conducted at midnight, 3 AM, and 6 AM; calibration was conducted if sensor error was >20%. An intervention was performed if a person’s blood glucose was <70 mg/dl or >250 mg/dl with at least 0.6 mmol/l of ketones. For the SAP, Alarms were set at 70 mg/dl and 250 mg/dl, and participants were under the care of the camp’s medical staff. When a person was on SAP, meter glucose checks were performed at midnight and 3 AM as required, and no remote monitoring was conducted.

• OCL was initiated on 50 out of 55 possible nights. On the 50 nights that the OCL did start, it operated for at least six hours on 37 nights (74% of nights). It turned on for fewer than six hours on the other 13 nights because of either a sensor error >20%, a loss of connectivity, an infusion set failure, or a sensor calibration failure (further breakdown of these issues was not reported). 

• The 21 campers who participated in the study (10 of whom were male) had a mean age of 14.7 years and mean A1c of 7.9%. The participants’ mean duration of diabetes was 7.9 years.

• Campers spent significantly more time in range (70-150 mg/dl) when on OCL (66%) than when on SAP (46%; p<0.005). When OCL lasted at least six hours, OCL was associated with significantly less time spent <70 mg/dl (~1% vs. ~20%), <60 mg/dl (<1% vs. ~10%), and <50 mg/dl (~0% vs. ~5%). On the hyperglycemia end, patients on OCL trended (insignificantly) towards spending less time >150 mg/dl (~25% vs. ~30%) and >180 mg/dl (~8% vs. ~18%). Participants spent significantly less time >250 mg/dl when on OCL vs. SAP (~1% vs. ~4%). Mean glucose was higher on OCL than on SAP (128 mg/dl vs. 110 mg/dl;), though the finding was statistically (and we would say clinically) insignificant (p=0.129). Additionally, while glucose variability appeared to gradually decline following OCL’s initiation, variability remained relatively consistent when a person was on SAP.

• Turning to the inpatient study, participants were admitted to the hospital for a 22-hour overnight period. They performed two periods of exercise to simulate camp conditions. Researchers started OCL within 12 hours of sensor insertion, and YSI was performed every 30 minutes. OCL was activated in 13 of 16 nights (81%). The three nights OCL was not initiated was because of either a failed sensor calibration at initiation or a sensor error of >20%.

Decision Analytic Model: Cost Implications of RT-CGM Use in Insulin Requiring Patients with Hypoglycemic Unawareness

Claudia Graham, PhD, MPH (Dexcom, San Diego, CA)

Dexcom’s Dr. Claudia Graham presented a very interesting model designed as a “starting point” to support the cost-effectiveness benefits of CGM for patients with hypoglycemia unawareness. Results showed that for a US third party payer with 10 million members, giving CGM to all hypoglycemia unaware patients would save between $59 million and $321 million per year (factoring in the cost of CGM). The simple model had a lot of assumptions baked in (see below – all were cited from published literature), but we found the results fairly compelling and highly instructive. Dr. Graham concluded that we need a lot more data in the future: studies of CGM to examine reductions in severe hypoglycemia, more RCTs in the MDI population, more studies in type 2 diabetes, and regional costs of hypoglycemia and hospitalizations. We hope things are moving head on many of these fronts, especially because much of the reimbursement evidence for CGM dates back to early generation systems that were less accurate and reliable (e.g., the JDRF CGM trial).

Questions and Answers

Q: What you’re missing in here is reduction in SMBG tests per day…

A: Yes, that’s correct. I didn’t look at that. We’re assuming using CGM per the labeling – you still need to do a fingerstick. We do know that once people start using CGM and gain trust in it, they start cutting back on strips.

First Evaluation of an Orthogonally Redundant Glucose Sensor System in People with Type 1 Diabetes

Sybil McAuley, PhD (St. Vincent’s Hospital, University of Melbourne, Australia)

Dr. Sybil McAuley presented the first human study (n=18 total) of Medtronic’s orthogonally redundant sensor (electrochemical + optical), a project supported by JDRF and the Helmsley Charitable Trust. The seven-day study compared the redundant sensor to a single electrochemical sensor (we assume Enlite); involved in-clinic (vs. YSI) and home portions (vs. SMBG); calibrated sensors four times per day; analyzed data after the study, but using a prospective algorithm; and evaluated several iterations of the redundant sensor over the course of six months (four patients wore the latest configuration). The system is clearly still a work in progress and the results were solid but not impressive. The biggest advantage of redundancy came in greater reliability – display time rose from 94.6% with the single electrochemical sensor (n=18) to 98.7% with the latest redundant sensor configuration (n=4). Dr. McAuley emphasized that this translates to one hour per day of additional display time. Overall accuracy improved marginally with the redundant sensor (from 13.9% to 12.3%), though day six accuracy improved markedly (16.2% to 10.6%). In-clinic MARD was 9.6%, an incremental improvement over 11.0% with the single electrochemical sensor. Based on these data, overall accuracy/reliability seems roughly on par with Dexcom’s G4 Platinum (MARD: 13.2%; 97% data capture). At this point, the major downsides of the redundant system are the need for four calibrations per day and the size of the on-body device (it appeared to be the size of a second-gen Insulet OmniPod). Future work will focus on continued development of the optical sensor, as well as use of multiple electrochemical sensors combined with the optical sensor.

• This seven-day study included 18 patients with type 1 diabetes. Patients had three study visits over the course of the week – insertion occurred on day one, an in-clinic meal test (vs. YSI) happened on day four, and sensors were removed on day eight. Patients were asked to perform eight or more SMBGs per day. The redundant electrochemical/optical sensors were inserted together at a single site. A recorder was worn on top of the sensor (it appeared about the size of a second-gen Insulet OmniPod). At the end of the study, the recorder was connected to a docking station, and data was uploaded from the recorder.

Use of Closed Loop In The Hospital

Closing The Loop – Beyond Type 1 Diabetes

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

Controlling glucose in the critically ill is an opportunity to save lives. In the past, hospitals have not been able to deliver lower average glucose without hypoglycemia, which has resulted in increased mortality. Dr. Hovorka described a very exciting study in which he used a sub-cutaneous CGM sensor (Abbott Navigator) in a closed loop system to tightly control glucose levels over a 48-hour period. The key is to calibrate the sensor very frequently. In a study of 24 patients, results were excellent – time in target zone (108-144 mg/dl) was 54% compared to 18% for the local protocol, and hypoglycemia was zero – not an easy feat given the very high variability of critically ill patients. Dr. Hovorka commented that ‘protocol choice has a greater influence on glucose outcomes than glucose measurement method’. Another exciting inpatient study he described was for type 2 inpatients in which a similar approach yielded 41% time in target (70-144 mg/dl) versus 24% for usual care, again with no hypoglycemia. This pilot will transfer to the general ward of the hospital.

• Glucose abnormalities are common among the critically ill, and hypoglycemia, hyperglycemia and glycemic variability are all independently associated with increased mortality. Trials attempting to lower glucose have had mixed results (such as NICE SUGAR, which unfortunately increased mortality). Dr. Hovorka believes that tight glycemic control leads to reduced mean glucose (good), reduced variability (good) and increased hypoglycemia (bad). In some studies the bad outweighs the good and vice versa.

• Dr. Hovorka described an automated closed loop system in the hospital intensive care unit (ICU) using a Navigator sensor (Abbott), a bedside computer and IV pumps for glucose and insulin. The sensor was calibrated hourly with arterial blood and patients were given standard nutrition.

• In this trial of 24 patients for 48 hours, glucose was tightly controlled in the range 6-8 mmol/l (108-144 mg/dl) - which is not easy because of the high variability of critically ill patients. Compared to local protocol, time in the target zone was increased from 18% to 54%, and both approaches spent no time below 4 mmol/l (72 mg/dl). The chart of median interquartile range showed dramatically better performance, particularly on inter-patient variability. The closed loop system also appeared to use less insulin, although the result was not statistically significant.

• In the UK, people with diabetes occupy 18% of the inpatient beds, yet good glycemic control remains poor and errors are common. A study with a randomized crossover design was designed to test closed-loop control over a 24 hour period against the patient’s usual control. Insulin lispro (Humalog, Lilly) was infused using an Animas pump and the CGM was the Abbott FreeStyle Navigator. Patients were insulin naïve, primarily so that they didn’t have to washout usual therapy before the hospital visits.

• Time in the target zone (70-144 mg/dl) was 41% versus 24% on usual therapy and time below 70 mg/dl was zero in both arms. Clever insulin assays demonstrated that during closed loop, the pancreas was rested – patients produced less endogenous insulin than in open loop. This work will move shortly to the General ward.

The Journey for a Viable Artificial Pancreas: The Pieces of the Puzzle are Coming Together

Eyal Dassau PhD (Sansum Diabetes Research Institute and UCSB, Santa Barbara, CA)

Dr. Dassau outlined the journey to the artificial pancreas from the Sansum perspective, describing the highs and the lows (pun intended) and the learnings along the way. In 2007, the first patient was switched to closed loop control, tethered to a laptop. Six years later, the same patient was walking around Santa Barbara wearing an artificial pancreas. Along the way, the team at Sansum investigated control to target, control to range, inhaled insulin and intra-peritoneal delivery, while also creating hardware and software that have been used by many groups around the world. Dr. Dassau concluded by thanking his many colleagues past and present, and by asserting that the work is transferable to the hospital where it will undoubtedly save lives.

• Dr. Dassau outlined the story of the artificial pancreas development at Sansum. Here are some of the highlights he mentioned:

  • 2007 – the Sansum team created the Artificial Pancreas System (APS), a hardware platform that connected to pumps and CGM sensors and hosted various control algorithms. The APS first supported the OmniPod and Dexcom Seven, and then later the Navigator, Animas and Roche pumps.

  • April 2007 – the first patient used fully automated closed loop control (with the APS). This patient was tethered to a laptop. After a few adventures, safety constraints for insulin on board were added in July 2007 and the study repeated with good results.

  • 2009-2011 – MPC control to target studies with unannounced meals. The algorithm was cleverly condensed into the form of a lookup table to save computing power.

  • 2009 – Control to zone studies – it was felt that set point control was not exactly what was needed. This study had both unannounced meals and exercise. Time in zone (70-180 mg/dl) was 81% with no time spent under 60 mg/dl.

  • 2012 – Semi-automated artificial pancreas with Technosphere insulin (Afrezza, MannKind).  This very rapid acting inhalable insulin tries to mimic first phase insulin response.

  • 2012 – Closed loop intra-peritoneal (IP) delivery (with Dr. Renard) using the Diaport 2 (Roche) and the Dexcom Seven Plus. Results were good, although breakfast was still a challenge. Interestingly, when you suspend an IP pump, you really suspend it – insulin action stops very quickly, unlike with sub-cutaneous infusion. After a very short time the patient moves out of hypoglycemia.

  • 2013 – Ambulatory artificial pancreas. This study merged aspects of psychology, physiology, behavior and control design, predicting glucose 45 minutes ahead. Exercise was unannounced, but the algorithm noticed the drop in glucose and reacted accordingly.  The same patient that was wired to a laptop in 2007 was in the study, walking freely around the streets of Santa Barbara on closed loop control.

• There are many learnings from this history. In particular, MPC is very flexible and can handle multiple variables, so it can be transferred to and tailored for the hospital setting. The MPC algorithm can be compared to playing chess – you think several moves ahead, but react each time to your opponent’s move by modifying your plan. The technology was developed for process control in the 1970’s and is very flexible. For this reason, MPC is the prime candidate for multi-variable glucose control in the hospital, where it will certainly improve patient outcomes.

Technology & Innovation in Clinical Practice — Improving Patients’ Lives (Sponsored by Medtronic)

Managing Hypoglycemia and Protecting Patients

Tadej Battelino (University Children’s Hospital, Ljubljana, Slovenia)

Dr. Tadej Battelino discussed the potential of diabetes technologies such as CGM (with and without low glucose suspend functionality) and closed loop systems to reduce hypoglycemia. Dr. Battelino reviewed data from the DREAM studies testing the MD Logic AP. His enthusiastic takeaway message is that the system almost completely eliminates lows below 60 mg/dl, especially nocturnally. Dr. Battelino also did not frame hypoglycemia as an abstract danger, but rather took the time to connect hypoglycemia with concrete adverse outcomes. He cited a growing body of evidence from neuroimaging studies that multiple episodes of severe hypoglycemia have a permanent adverse effect on white matter microstructure in the brain (white matter is responsible for conducting neural signals). Notably, during Q&A, Dr. Battelino expressed doubt that double hormone closed loop systems are the best options to pursue at the moment, and argued that insulin-only systems are sufficient for most patient situations.

Questions and Answers:

Q: Are we at a point in the closed loop studies that we can shut off the alarms?

A: At the moment, we still use surveillance in these trials. All patients on closed loop are monitored both visually and through computer systems. In all those nights we observed, there was not a major adverse event, but I wouldn’t shut off the alarms now at this moment.

Q: Can you visualize dual hormone technology in the near future, whether it is glucagon or any other hormone to reduce hyperglycemia excursions?

A: It’s a very important question. The MD Logic algorithm allows for dual infusion, which could be glucagon infusion. But we believe that here some problems with glucagon, especially the problem that it doesn’t keep its function is you use it a lot. At the moment, I don’t think we’re at the point for a double hormonal system. I think we will have to stay with an insulin-only system. The only big danger with such a system is if the patient injects himself or herself with far too much insulin, the insulin-only closed loop cannot do anything in that case, and glucagon would be needed. But for the moment, I think we can stay with the insulin-only system.

Challenges of Diabetes Management in Children and Teenagers

Panel Discussion

Q: You quickly went through the differences between the ISPAD and ADA goals. In the US, the A1c is 1% higher in the six-year-old age group compared with Sweden. The case is the same when you compare with the German registry. I think that has to do with the goals. Is this a modifiable target or not?

Dr. Lori Laffel (Joslin Diabetes Center, Boston, MA): I wish that everything was as simple as “a stricter goal makes A1cs better.” As with everything, it’s multifactorial. Adults have a goal of 7%, and most adults aren’t under 7%. We have to be realistic, and saying that changing a goal is enough to change results is not realistic. The new ADA guidelines that were released in January 2014 say that you can pursue a lower goal if it can be done without hypoglycemia.

Comment: I still think the challenge is that a number is a number, and if you set a lower goal, you would achieve better results.

Dr. Bruce Buckingham (Stanford University School of Medicine, CA): First of all, there isn’t a lot of good data showing that hyperglycemia prior to puberty results in long-term complications. If you look at studies from the Berlin group and a number of Scandinavian countries, the pre-pubertal years don’t count. The changes could be due to the fact that, post-puberty, your turnover of collagen is much slower, which could also be responsible for metabolic memory. I’m not convinced that your goal of an A1c of 7% in a young kid is necessary. Taking the other half, looking at the brain, the grey matter is affected by hypoglycemia, and white matter by hyperglycemia. We’ve done studies serially, and hyperglycemia shows a clear correlation with demyelination. So neuropathy might be a rationale for keeping blood sugars low, but probably not retinopathy or nephropathy.

Q: For the US panelists, do you consider cultural differences as a possibility for explaining the differences in A1c between the US and Europe? Is it the targets or is it cultural differences?

Dr. Laffel: These are multifactorial processes, some of which are grounded in biology and some of which are grounded in genetic factors we haven’t teased out. Some may be environmental. I think that if you ask Americans to come in fasting for lab work, they’ll come in saying that they ate breakfast. By comparison, Europeans know what fasting means, so culture could be a factor. Mark, even though you have an accent, you’re American. What do you think?

Dr. Mark Sperling (Children’s Hospital of Pittsburg, Pittsburgh, PA): The expectations cannot be the same, and the expectation is not necessarily what sets the response rate. I would say that there are cultural differences.

Q: Why do we consider glucose to be the only important factor in the development of complications? Should we look at other factors such as inflammation?

Dr. Kim Donaghue (University of Sydney, Sydney, Australia): I’m sure there are a number of factors at play that we don’t know how to measure or control, but glucose is going to be fundamental. It’s surprising that we can’t get the numbers that we would expect. It could very well be the variability, but it’s hard to measure variability well enough to associate it with outcomes.

Q: What is the explanation for the downward trend we are seeing in the age of diagnosis for type 1 diabetes?

Dr. Laffel: There has been, over the last few decades, a shift to a younger age of onset. However, more recent data, particularly from Scandinavia, have shown a leveling off of the incidence of type 1 diabetes, and in fact the youngest age group is not showing an increase anymore. That’s my take from the literature.

Q: Why is the incidence of type 1 diabetes increasing around the world? Is it food, or vitamin D?

Dr. Buckingham: I don’t think we know, but we can always postulate. It’s probably not the vitamin D theory. In one of the longitudinal tests where they did serial blood samples, the kids who got diabetes actually had higher vitamin D levels. Maybe its because we’re not throwing enough at our immune system — we could try using porcine whipworm like some studies are.

Dr. Sperling: Could the increase in obesity be a cause?

Dr. Donaghue: We have certainly found that the younger children who come in are more overweight at presentation. The ones who get it at the more typical age of puberty were closer to the normal BMI. It’s possible that the younger ones are being overfed, and that prevention should start during pregnancy.

Dr. Laffel: We understand that type 1 diabetes is an autoimmune disease today, but with genetic susceptibility that could be present in as much as 40% of Caucasians, so there is a huge pool of susceptible individuals. I suspect that there are different susceptibilities in different individuals that interact with different environmental exposures, and we don’t know what the environmental exposures are. We have important longitudinal studies going on now, and smart people looking at this. If this was something simple, we would know what it is. While our talented colleagues work on genetics and the basic science, we must today continue pushing forward for better treatments today, such as CGM and the artificial pancreas, until we get a cure tomorrow.

New Devices and Technologies that Facilitate Improved Health and Lifestyle

Exploiting Information Technology, Guided by Evidence-Based Models of Health Behavior Change, to Strengthen Self-Monitoring of Blood Glucose and Self-Management Action

William Fisher, PhD (University of Western Ontario, London, Canada)

Dr. William Fisher gave an energetic talk on ways to improve patient’ use of SMBG and their diabetes self-management through information technology. SMBG is not just a tool, he emphasized, but a behavior that requires a combination of information, motivation, and behavioral skills on the part of the patient. Data from a patient questionnaire (Fisher et al., Diabetes Educator 2011) indicates that with current technology, patients are often uninformed (~75% of type 1 and type 2 diabetes patients believed that their bodies told them if their blood sugar was high or low), unmotivated (~65% of patients thought that testing their blood glucose as often as recommended would be painful), and unskilled (20-25% of patients thought that testing their blood sugar discreetly is difficult or impossible). Adherence counseling is time and labor intensive if conducted directly by providers, but information technology provides an alternative way to improve patient self-monitoring and self-management. Dr. Fisher shared data from his group’s recent study of a cellphone-based remote patient reporting and automated feedback system, which provided positive feedback (“keep up the good work”) as well as actionable recommendations (“walking daily can increase your insulin sensitivity”). Over a mean of ten months, intervention patients saw a mean A1c reduction of 0.4% and weight reduction of 4 lbs (2 kg). As a take-home message, Dr. Fisher argued that diabetes technology has advanced substantially, but that the strategies for the application of these technologies (especially information technology) must be based on well-established evidence rather than brainstorms and best guesses.

Questions and Answers

Q: What is more powerful, positive or negative reinforcement?

A: Generations of research with rats and human beings tell us that positive reinforcement through reforming and reshaping positive behaviors is the way to go. There is also a positive role for cognitive change. If many individuals simply regard SMBG as a punishing and discouraging activity, we can try to reframe that as a way to learn about gaining strength and incrementally improving control. Using research we’ve done with patients on very toxic regimens in other areas, we’ve helped patients to visualize testing as positive rather than negative. At core, there is substantial literature on how to motivate people; it’s time we start looking more at that knowledge base.

Q: Knowing that the people who have had diabetes for a long time can have persistent false beliefs about SMBG, do you have any thoughts on changing their long-term behavior?

A: The longest-term data we have on health behavior changes suggests that the combination of changing the person and the technology are needed. In a recent paper in Clinical Diabetes, we use a technique called motivational interviewing. The provider will ask how important it is for the patient to do something, say for example paired testing. The patient will answer and begin developing positive ideas about testing. The provider will also ask about what it will take for them to ratchet up the frequency of their testing, and will work with that.

--by Adam Brown, Hannah Deming, Hannah Martin, Manu Venkat, and John and Kelly Close