February 3-6, 2016; Milan, Italy; Day #2 Highlights – Draft

Executive Highlights

Misty Milan gave way to sunshine this morning, just in time for our walk over to the Milano Congressi in advance of ATTD Day #2. Corporate updates galore headlined today’s agenda as: (i) Medtronic shared that the US pivotal trial of its MiniMed 670G is wrapping up in 25 (!) days; (ii) Abbott announced that it has obtained a CE Mark for a pediatric indication of FreeStyle Libre (down to age four – parents will be very excited); and (iii) Dexcom disclosed several positive updates on G5 and G6. Please see below for details and be sure to check out our ATTD preview for what the last day and a half have to offer.

1. An outstanding, unopposed Medtronic symposium shared big news: the US pivotal trial of the MiniMed 670G is “close to the end” – the last patient will leave the study in just 25 days, three months earlier than we expected it to end. Dr. Fran Kaufman hinted that the 670G might be available in 2017 – we’re certainly hoping so and it was terrific to learn the work was so far ahead of schedule. Ms. Annette Brüls shared new plans to launch next-gen CareLink Pro reports in the next 12 months (including pump settings optimization) and future plans to build Bluetooth directly into Medtronic’s hybrid closed loop. This company is really moving! Go, Annette …

2. Abbott announced that it has obtained a CE Mark for a pediatric indication of FreeStyle Libre, allowing it to market the system to children 4-17 years old (previously only 18+ years).

3. Dexcom CTO Jorge Valdes shared a few key pipeline updates: (i) the new G5 smaller transmitter (expected in late 2016 or early 2017, per JPM) will cut the volume in half vs. the current transmitter; (ii) G5 Android is still expected to launch this year; (iii) Dexcom will incorporate predictive hypoglycemia alerts into G6, offering a larger 15-minute prediction window. The company is carefully balancing advance notice with nuisance factor.

4. Tidepool’s Mr. Howard Look shared that the non-profit hopes to release its novel Nutshell app “within the next quarter,” enabling users to improve mealtime insulin dosing by keeping track of insulin and glycemic data from past meals. We were fans of Meal Memory when it launched, though it had some serious shortcomings we hope Nutshell will overcome.  

5. We had a chance to speak to DreaMed in the exhibit hall, where we confirmed the MD-Logic Pump Advisor’s design and learned of plans to begin the preliminary HCT-funded study this fall. The design is exactly as we had hoped – pattern recognition that identifies very clear, specific changes in pump settings like basal rates and insulin-to-carb ratio (e.g., change basal from 0.95 u/hr -> 0.8 u/hr from 12-8am due to pattern of nighttime hypoglycemia).

6. Dr. Aaron Kowalski shared enthusiasm for using time-in-range as the main measure of glycemic control, while Dr. Richard Bergenstal asserted that CGM data should be displayed in a standardized, easy-to-interpret report (AGP).

7. A very valuable CGM session offered reimbursement perspective from three of the largest markets in Europe – Germany, France, and England – along with thoughts on Medicare reimbursement of CGM (unlikely in the near term).

8. Incoming ADA President Dr. Desmond Schatz (University of Florida, Gainesville, FL) delivered a passionate talk on the need for changes to the current paradigm for type 1 diabetes cure research.

9. A T1D Exchange survey found that patients’ top-cited reasons for discontinuing pump or CGM therapy included problems with insertion/adhesive, cost/lack of insurance coverage, and sensor inaccuracy.

10. Joslin’s Dr. Lori Laffel shared promising one-year results from the CGMi trial that is evaluating the effectiveness of a “family-focused behavioral teamwork intervention” vs. standard education in the initiation of CGM in youth ages 8-17 with type 1 diabetes. Notably, discontinuation was driven by device choice more than psychosocial outcomes.

Top Ten Highlights

1. An outstanding, unopposed Medtronic symposium shared big news: the US pivotal trial of the MiniMed 670G is “close to the end” – the last patient will leave the study in just 25 days, three months earlier than we expected it to end (the ClinicalTrials.gov posting slated completion in May). Medtronic expects 100 completers. Dr. Fran Kaufman hinted that the 670G might be available in 2017 with a reference to the US election, “Maybe in the next US President’s first year of office, we can say they live in a country where hybrid closed loop is available.” As a reminder, Medtronic’s latest JPM timeline suggested commercialization by April 31, 2017, and given the early pivotal study conclusion, that timeline seems more realistic (though the FDA review would have to be around 12 months at most). Dr. Kaufman also disclosed that Medtronic is planning a major 1,000-patient post-approval outcomes study of the MiniMed 670G (the biggest study Medtronic Diabetes has ever done). The multinational trial will enroll a representative type 1 population (spectrum of A1cs and ages), randomizing patients to three groups for six months: pump alone, sensor-augmented pump (no automation), or the MiniMed 670G (IF only there was a fourth group on MDI!). The primary endpoint will be glycemic control AND hypoglycemia, with a six-month follow-up period. The study will generate excellent real-world outcomes data, and we especially hope it is powered to show changes in severe hypoglycemia and tracks healthcare costs.

• In a digital health focused talk, Ms. Annette Brüls confirmed the timing from CES to launch a hypoglycemia prediction app with IBM Watson this summer. But in new news, she shared plans to launch next-gen CareLink Pro reports in the next 12 months (including advanced analytics to optimize pump basal and bolus settings), along with future plans to build Bluetooth directly into Medtronic’s closed loop (eliminating MiniMed Connect; no timing stated). In the coming months, Medtronic will also launch some of the current CareLink Pro features into CareLink personal, countering longstanding complaints that patients should see the same thing as providers. Ms. Brüls also showed a very compelling example of insightful Big Data (from CareLink). A colored map of the US compared dangerous hypoglycemia episodes (>3 hours at <50 mg/dl) in patients with threshold suspend turned off vs. on. Patients with threshold suspend turned off experienced 6.1 such episodes per year (!) vs. just 0.8 episodes per patient per year with threshold suspend turned on. The slide suggested the difference (5.3 episodes hypoglycemia) equates to $735 in emergency services per patient per year and $9,300 in patient services per year. Now that is some compelling evidence that even mild automation can make a difference.
• IBM Watson Health’s Chief Science Officer Dr. Shahram Ebadollahi gave an outstanding presentation on the growth of unstructured data, the company’s plans with Watson, and shared new data on the Medtronic hypo prediction app from an analysis of 2,000 users’ CareLink data. See the details below. The early results from DTM are still holding, showing 80-85% hypoglycemia prediction accuracy 3-4 hours ahead of time. This was a truly outstanding talk!
• Dr. Pratik Choudhary also shared compelling real-world data on the MiniMed 640G – see full details below.

2. Abbott announced that it has obtained a CE Mark for a pediatric indication of FreeStyle Libre, allowing the company to market the system to children 4-17 years old (previously only 18+ years). Many pediatric patients seem to already be using the system off label in Europe, so we’re not sure if this has significant commercial implications (FreeStyle Libre is sold online without a prescription, so anyone can buy it, including parents ... though Abbott does offer a disclaimer on this point). That said, it is excellent for Abbott to get the official label and we surmise that the indication may be of very real value down the road – Abbott, after all, is pursuing reimbursement in type 1s and type 2s and a pediatric label would be critical for coverage across the age spectrum. [As a reminder, Abbott is presenting its six-month ("reimbursement" - our quotes) data in type 1s at ADA 2016.] The approval does enable Abbott sales reps to call on pediatric endocrinologists that were previously off limits and helps Abbott stay competitive with Dexcom’s G5 that is approved down to age two. Abbott has yet to disclose FreeStyle Libre sales at this point, so we won’t know how much this increases sales from the current base. As we understand it, there will no differences in the sensor or packaging except of course for the updated label; no promotions have been announced at this time, though we imagine the “No Fingersticks” will resonate soundly with the pediatric population. It’s hard to know if the news has any implication for the labeling of the US product (which we recently learned could launch “toward the end of this year”) and Abbott has remained tight-lipped on its future plans. We’re hoping we’ll learn more tomorrow when the company is sharing the full pediatric accuracy data that was submitted for the indication in an afternoon oral – Dr. Fiona Campbell in Yellow Hall at 5:50 PM. We hope to see some of you there!

3. Dexcom CTO Jorge Valdes shared a few small pipeline updates: (i) the new G5 smaller transmitter (expected in late 2016 or early 2017, per JPM) will cut the volume in half vs. the current transmitter (previously, no size reduction estimate had ever been given); (ii) G5 Android is still expected to launch this year, though Dexcom won’t be able to cover every single Android phone – it will support the major brands, given the sheer number of Android phones to validate; (iii) Dexcom will incorporate predictive hypoglycemia alerts into G6, offering a larger 15-minute prediction window – the design is balancing more advanced notice without increasing false alarms or annoying patients. The predictive alerts were the focus of many slides for the first time, which made us wonder if it was an answer to Medtronic’s hypoglycemia prediction app with IBM (which has a 2-4 hour prediction horizon, but could obviously increase alarm fatigue). The new predictive alert is clearly effective: with its addition, only 9% of 55 mg/dl BG events will give patients <15 minutes of warning vs. a much higher 42% with just a threshold alert alone (this is most key when patients are dropping rapidly, as going from a threshold of 80 mg/dl to <55 mg/dl can happen with just a five minute heads up). Overall, the new feature has a 93% predictive hypoglycemia alert detection rate (for 93% of YSI readings <55 mg/dl, a predictive hypoglycemia alert occurred between 0-30 minutes), but not at the cost of additional nuisance: for just 11% of predictive alerts, there was no YSI reading below 70 mg/dl within the next 30 minutes, making these alerts unnecessary). We’re glad to see Dexcom balancing valuable prediction with hassle factor – it doesn’t matter if an alarm is accurate if it is annoying!  

• Mr. Valdes mentioned several times that Dexcom’s G5 is accurate enough to dose insulin on (consistent with the EU label that it replaces fingersticks), though we noticed a slight step away from the longstanding emphasis to eliminate fingersticks – he mentioned that dosing insulin can be lethal, and Dexcom wants to keep calibrations in for safety. Was this an indirect comment to address the current G5 setup (two calibrations per day) vs. Abbott’s factory calibrated FreeStyle Libre, which also has a dosing claim in Europe? This also might reflect expectations for G6 to include one fingerstick calibration per day. The final slide on “Dexcom’s R&D Mission” also notably excluded eliminating fingersticks. See those priorities below. In a follow-up conversation, management shared the following helpful perspective: “We are not backing off from eliminating fingersticks. But going forward we can see multiple classes of products, some requiring calibration and some not depending on the use case. We do believe that all sensors should allow calibrations, whether required or not, in order to cover the cases when the glucose reading do not align with the way a patient feels.” That rationale makes sense to us.

4. Tidepool’s Mr. Howard Look shared that the non-profit hopes to release its novel Nutshell app “within the next quarter,” enabling users to improve mealtime insulin dosing by keeping track of insulin and glycemic data from past meals. By logging and integrating mealtime data (bolus, correction, pre- and postprandial blood glucose values), Nutshell delivers insights that can help users make more informed bolusing decisions for future meals. We LOVE this, since patients tend to eat the same meals, and it makes perfect sense to look at what happened last time and make more informed decisions. Nutshell is currently in beta testing, and will be freely available via open source after its launch. This is the first we’ve heard on a specific launch timeline for Nutshell; we first wrote about the app in January 2014 and we are thrilled to hear that Tidepool is moving forward with it in the next few months. We do wonder about entry burden, since no Tidepool-compatible pumps can send bolus insulin levels in real-time to the phone; presumably those will be manually entered. We assume Dexcom data will important directly for G4 Share and G5 users. We were fans of Meal Memory when it launched, though it had some serious shortcomings we hope Nutshell will overcome: inability to search, no location-based notes, and more listed here.

• Mr. Look also noted that Tidepool’s partnerships with Animas (Ping, Vibe) and OneTouch (VerioIQ, Ultra2, Ultra Mini) will be launching soon, building on compatibility with Dexcom, Insulet, Tandem, Abbott, and Bayer (Medtronic and Roche have still not officially authorized Tidepool to read data from their devices, though Tidepool can import data from Medtronic’s CareLink Personal). We have been continually impressed with Mr. Look and his team for their outstanding efforts in making diabetes data more accessible, more open, and more actionable, and we look forward to providing a deeper perspective on Nutshell when it launches.

5. We had a chance to speak to DreaMed in the exhibit hall, where we confirmed the MD-Logic Pump Advisor’s design and learned of plans to begin the preliminary study this fall. The algorithm’s output is designed exactly as we had hoped – pattern recognition that identifies very clear, specific changes in pump settings like basal rates and insulin-to-carb ratio (e.g., change basal from 0.95 u/hr -> 0.8 u/hr from 12-8am due to pattern of nighttime hypoglycemia). We confirmed that the initial version will send patient data from the Glooko app to the physician, who will approve the MD Logic Pump Advisor recommendations and send it back to patients. The hope is to eventually the pump settings recommendations directly to patients without HCP confirmation. The team shared with us that it hopes to start the study this fall. As we noted yesterday, the Helmsley Charitable Trust awarded $3.4 million to DreaMed to fund the Pump Advisor’s development (leveraging data from the Glooko platform).

• In a symposium on digital advisors, Dr. Moshe Phillip also presented on the general rationale behind software prescription therapy (“software as a drug”). Emphasizing the complexity of diabetes management and the shortage of endocrinologists, Dr. Phillip discussed the many challenges patients and physicians face (i.e., swamped with information, limited time and knowledge to perform optimization, etc.). He pointed to the promise of diabetes data management platforms (i.e., Diasend, Glooko), but noted that these systems do not suggest any solutions, highlighting the need for software prescription therapy. He briefly introduced the MD-Logic Pump Advisor and reviewed a few patient cases from a feasibility study – they showed less variability and fewer postprandial highs following pump settings optimization. We look forward to seeing even more data, as we really believe this could improve care.

6. A star-studded session offered a vision of a future in which time-in-range is the main measure of glycemic control and CGM data is routinely displayed in a standardized, easy-to-interpret report. JDRF’s eloquent Dr. Aaron Kowalski enthusiastically supported using time-in-range as the main measure of glycemic control and more broadly asserted that interventions must balance “diabetes health” and “diabetes happiness.” He noted many of the limitations of A1c, highlighted the intuitive advantages of time-in-range, and mentioned JDRF’s new T1D Outcomes Program, which aims to define the metrics of importance to patients, providers, researchers, industry, FDA, and payers. More broadly, Dr. Kowalski reiterated many of the elements of his Diabetes Scorecard (Diabetes Care 2015), emphasizing the need for better “value” in type 1 diabetes, which is defined differently for patients, providers, and payers – and is not just A1c, but includes device burden, time investment, fear, provider quality metrics, hospitalizations, etc.! In the same session, Dr. Richard Bergenstal (Park Nicollet International Diabetes Center, Minneapolis, MN) discussed the need for standardized CGM reports – an “ECG for glucose patterns” – that incorporate information beyond A1c. His presentation focused primarily on the Ambulatory Glucose Profile (AGP), a most-valuable one-page report with critical glycemic data at the top, followed by a simple visualization of glucose patterns. To date, Abbott is the only sensor company to adopt this and we wish all the manufacturers would because it would be far better to have standardization and consistency here than not. Dr. Bergenstal focused in particular on the need to standardize the definitions of hypo- and hyperglycemia in such reports, arguing that dividing the data into sub-categories (e.g., time below 70 mg/dl, 60 mg/dl, and 50 mg/dl) is more clinically useful than lumping them all together – we’ve been following this forever and SO want to see this happen for patients (and HCPs!). See our detailed report below for more on these informative talks.

7. A very valuable CGM session offered reimbursement perspective from three of the largest markets in Europe – Germany, France, and England – along with thoughts on Medicare reimbursement of CGM (unlikely in the near term). The purpose of the session was to provide a broad overview of current reimbursement status and processes in these countries and, consequently, many of the presentations were general in scope. That said, this was very appropriate for the international audience and certainly wasn’t without its moments. We have summarized the most notable points of discussion below.

• Dexcom’s Dr. Claudia Graham was not able to provide a firm guidance on when we might see Medicare coverage of CGM though that did not stop her from delivering a scathing overview of the “RIDICULOUS” bureaucratic red tape muddling the process. This was one of the most dynamic presentations we have heard on the topic – one that is quite nuanced and complex – and we thought she did an outstanding job breaking down the disconnect between US private payers and Medicare for the international audience. As a reminder, Medicare’s primary objection to covering CGM stems from the adjunctive labeling (i.e., that treatment decisions be based on BGM readings rather than CGM readings). Dexcom had seemed close to overcoming this objection, though its most recent financial calls and JPM update suggested FDA discussions are still ongoing. Importantly, Dr. Graham stressed that clinical data alone will not be sufficient to support coverage of CGM, noting instead the ongoing “surround sound” approach that has brought patient, professional association, and industry groups together in an advocacy campaign. Indeed, we were struck by just how challenging the political angle to this problem is (how to deal with an agency – CMS – that isn’t responding to reason?), and we salute Dr. Graham and collaborators for continuing to drive awareness in Washington. As a reminder, bipartisan support of the Medicare CGM Access Act of 2015 continues to grow following the reintroduction in the Senate and House of Representatives in late March, though judging from Dr. Graham’s commentary, it sounds like the 2016 election cycle is going to complicate this picture in a big way.
• England’s Dr. Peter Hammond informed attendees that NICE’s assessment of sensor-augmented pump (SAP) therapy – which we believe will be the final word on whether the UK will adopt national coverage for SAP – has been delayed to March 2016 (the guidance was previously slated for last week). As we understand it, the assessment comes on the heels of what was a major reimbursement victory for diabetes technology this past August, when UK clinical guidelines recommended CGM for the first time in certain adults and children with type 1 diabetes (which centered on patients with lots of hypoglycemia, which sounded very right to us). The recommendation does not guarantee a positive ruling from NICE, but is certainly better than what we might expect from the tough agency in an area that needs more clinical data. It’s a big win that NICE has implicitly acknowledged CGM’s cost-effectiveness, at least in certain patients (while we certainly know – in spades! – the value is there, we acknowledge there is not enough data). Dr. Hammond suggested that NICE’s decision is hinging on the incremental benefit of standalone CGM vs. a sensor-augmented pump – if only Dexcom’s DIaMonD data was available to answer this question (expected later this year)! We’re not sure if the reimbursement decision will have implications for the artificial pancreas in the UK, since the clinical benefit will presumably be greater than SAP alone.
• Germany’s Dr. Norbert Hermanns’ blunt opening statement summarized nicely the current status of CGM reimbursement in his country – “The situation is VERY unsatisfying.” It was one of the more patient-centered attitudes we heard on the day, a sentiment that is not often associated with the Germany’s Federal Joint Committee (G-BA). As a reminder, the G-BA has been particularly challenging in diabetes where, on the drug front, it found no added benefit with Sanofi’s Lyxumia (lixisenatide) in 2013 (prompting the drug’s withdrawal) and in 2007 ruled that the benefit of then-Lilly’s Byetta (exenatide) was not yet proven. On the device side, the decision on CGM reimbursement has proceeded excruciatingly slowly (it started in 2011!) and it sounds like the timeline is still wide-open. For context, Dr. Hermanns shared that IQWiG (the country’s independent scientific institute that provides the G-BA with reimbursement recommendations) has acknowledged a benefit to CGM and sent its report to the agency in May 2015. At the same time, he pointed out that the data summarized by IQWiG is not the most convincing as it doesn’t incorporate results from the newest devices (no surprise considering this process has been going on for the greater part of five years!) and he did not sound particularly hopeful. Ultimately, we continue to find it tough to get a grasp on how German authorities are thinking about diabetes.

8. Incoming ADA President Dr. Desmond Schatz (University of Florida, Gainesville, FL) delivered a passionate talk on the need for changes to the current paradigm for type 1 diabetes cure research. Much of his talk focused on the profound differences in the disease between children and adults: younger patients start out with less C-peptide at baseline, lose beta cell function faster, and often respond differently to therapies in clinical trials compared to adults. An excellent Consensus Conference on this topic last year covered these issues in great detail, and we hope this discussion can lead to changes in research and regulatory frameworks. Dr. Schatz also focused on the need to stage and develop treatments for early, asymptomatic stages of type 1 diabetes. This was the subject of a JDRF workshop in October 2014, which led to a new proposed classification system recently published in Diabetes Care. The framework divides type 1 diabetes into three stages: (i) multiple autoantibodies and normoglycemia; (ii) multiple autoantibodies and dysglycemia; and (iii) symptomatic disease. Dr. Schatz (and many others) hopes that this new framework can enable early clinical trials with intermediate endpoints to replace the “expensive and unwieldy” studies of the past. He pointed to TrialNet’s Abatacept Prevention Study as one promising example; the study has development of abnormal glucose tolerance as its primary endpoint and is slated to last four years (about half the duration of other recent prevention trials).

• Dr. Schatz was adamant that combination therapy is the future of type 1 diabetes treatment. He urged those in the field to think outside the box, learn from other disease areas (AIDS, cancer, tuberculosis) in which combination therapy is the norm, and move away from searching for a one-time silver bullet. We have heard similar sentiments from a number of type 1 diabetes researchers, though one challenge noted by Dr. Jeffrey Bluestone (UCSF, San Francisco, CA) is that it is difficult from a regulatory perspective to conduct efficacy trials of combinations with no FDA-approved drugs – he believes achieving the first “foundational” approval will be an important step toward studies of more innovative combinations.

9. A T1D Exchange survey found that patients’ top-cited reasons for discontinuing pump or CGM therapy included problems with insertion/adhesive, cost/lack of insurance coverage, and inaccuracy (for CGM). Dr. Viral Shah (University of Colorado, Aurora, CO) presented data from an electronic survey of 2,452 adults with type 1 diabetes that aimed to identify human factors associated with pump and CGM discontinuation. In this cohort, 67% of patients were currently using a pump, 31% were not using a pump, and only 2% had discontinued within the past year. 30% of patients were current CGM users, 59% were not using CGM, and 11% had discontinued within the past year. The demographic factors significantly associated with pump discontinuation were younger age (under 26), infrequent blood glucose testing, and lower income and education level. The top-cited reasons for pump discontinuation (patients could select more than one reason) were problems with insertion/adhesive (60%), cost/lack of insurance coverage (45%), and interference with sports (42%). For CGM, lower income level was the only demographic factor significantly associated with discontinuation. Interestingly, patients who reported testing their glucose more frequently were more likely to discontinue CGM, though the association was not significant. Dr. Shah’s explanation for this was that these more engaged patients likely that did not trust the CGM measurements were accurate. This is consistent with the finding that inaccuracy/device malfunctioning was the most commonly cited reason for discontinuing CGM (cited by 71% of participants), followed by insertion/adhesive problems (61%) and cost/lack of insurance coverage (58%). When asked during Q&A, Dr. Shah noted that the CGM discontinuation rate was higher with Medtronic devices compared to Dexcom but did not provide specific numbers. With the caveat that this was a small sample from a fairly unrepresentative cohort (the Exchange is an earlier adopter population at top clinics), such human factors information should be very useful for providers trying to convince patients to initiate pump or CGM therapy, and for companies aiming to design more patient-friendly devices. 

10. Dr. Lori Laffel (Joslin Diabetes Center, Boston, MA) shared one-year results from the CGMi trial evaluating the effectiveness of a “family-focused behavioral teamwork intervention” vs. standard education in the initiation of CGM in youth ages 8-17 with type 1 diabetes. Notably, discontinuation was driven by device choice more than psychosocial outcomes. The 24-month randomized control trial aimed to assess whether psychosocial support optimization could help overcome the barriers to CGM use and improve glycemic control. One-year findings demonstrated that ~30% of patients discontinued CGM use, higher than we would have expected (especially given the above study). However, Dr. Laffel shared that the discontinuation rate was independent of study cohort group (i.e., no difference between intervention and control group) and was instead predicted by CGM device. [As a reminder, the trial enrolled in 2013 when Dexcom was still rolling out its G4 – thus, some patients spent the entire 12-month period on the Seven Plus; some transitioned halfway to through the trial to the G4; some spent the entire 12-month period on the G4 – these are quite old products so this discontinuation rate is far less relevant overall.] Apparently, choice of device (G4 vs. Seven Plus) predicted both discontinuation rate and median hours/week CGM use in both groups: “It wasn’t psychosocial aspects of CGM that predicted outcomes. It was the device itself!” The conclusion was not meant to dismiss the importance of psychosocial factors but to highlight that fundamental deficits in devices themselves, as we have long said - e.g., technical glitches, false alarms, inconvenient insertion – still represent major problems and can be quite self-defeating. But think about the upside!  

Detailed Discussion and Commentary

Transforming Diabetes Care Together Industry Symposium Supported By​ Medtronic

Greater Freedom with Advanced Technology

Fran Kaufman, MD (Chief Medical Officer, Medtronic Diabetes, Northridge, CA)

Dr. Fran Kaufman gave a masterful overview of Medtronic’s efforts to close the loop, sharing that the US pivotal trial of the MiniMed 670G is “close to the end” – the last patient will leave the study in just 25 days, three months earlier than we expected it to end (the ClinicalTrials.gov posting slated completion in May). Medtronic expects 100 completers. Dr. Kaufman hinted that the 670G might be available in 2017 with a reference to the US election, “Maybe in the next US President’s first year of office, we can say they live in a country where hybrid closed loop is available.” That reinforces the confirmations to date that the 670G will be available by April 2017. Indeed, as a reminder, Medtronic’s latest JPM timeline suggested commercialization by April 31, 2017, and given the early pivotal study conclusion, that timeline seems imminently more possible, though the FDA review would have to be around 12 months at most. Dr. Kaufman also disclosed that Medtronic is planning a major 1,000-patient post-approval outcomes study of the MiniMed 670G (the biggest study Medtronic Diabetes has ever done). The multinational trial will enroll a representative type 1 population (spectrum of A1cs and ages), randomizing patients to three groups for six months: pump alone, sensor-augmented pump (no automation), or the MiniMed 670G (IF only there was a fourth group on MDI!). The primary endpoint will be glycemic control AND hypoglycemia, with a six-month follow-up period. The study will generate excellent real-world outcomes data, and we especially hope it is powered to show changes in severe hypoglycemia and tracks healthcare costs.

• As Medtronic disclosed at DTM, Dr. Kaufman mentioned that some 670G trial participants have successfully petitioned the FDA for continued access and use of the system. She shared a striking quote from one letter for the first time, a testament to the enthusiastic patient reaction: “Dear FDA: I am currently participating in a 3-month home study of the Medtronic 670G insulin pump. This pump has changed my life and that of my whole family immensely.” We see this as a positive early sign of a system people want to use – “I don’t want to give it back” – and we’re glad to see Medtronic and the FDA making this possible. Apart from the benefit for these patients, it also suggests FDA has become more confident in the real-world safety of closed loop systems. That might result in a speedier review, though it’s always hard to predict.
  
  • We’ll be interested to see what the three-month pivotal trial outcomes look like – what kind of time-in-range, A1c improvements, and reductions in hypoglycemia should be expected? We’d note that the pivotal study is single-arm and will only have 100 completers, so it won’t be ideal for comparing the system’s efficacy to other interventions. We believe Medtronic opted for the fastest, most efficient study to get the 670G to market, and the larger post-approval outcomes study will generate the comparative efficacy data everyone is keen to see.
• Dr. Kaufman alluded to Medtronic’s next-gen closed-loop plans in broad strokes: (i) “Advanced hybrid closed loop” combining the 670G PID algorithm with the licensed DreaMed algorithm; and (ii) “towards personalized closed-loop” – improving interface and meal announcement; pattern recognition; additional sensor inputs (activity, food, heart rate, sleep, free fatty acids); and detecting sensor or infusion set failure. We assume the latter will leverage the partnership with IBM Watson. There has never been launch timing associated with these, but we assume they are 2018+.
• Dr. Kaufman showed a schematic to explain how the MiniMed 670G algorithm works, the most detailed we’ve ever seen. It begins with patient specific parameter estimation and daily adaptation. At the start of hybrid closed loop, there is a sensor accuracy check, along with a glycemic target adjustment for a smooth transition to closed-loop. The 670G will revert to open loop if the sensor is inaccurate. Medtronic is using an ePID controller, which has “normal mode” and “exercise mode.” Dr. Kaufman mentioned targets of 120 and 150 mg/dl, respectively, though it was not clear if these are set or can be adjusted. The device has a max insulin limit, and it will switch to safe mode or the pre-programmed basal rate in cases like sensor failure.
  
  • At this point, a remarkable 21 investigator initiated studies (including eight 670G studies) have inform the design of the MiniMed 670G system. Dr. Kaufman admitted that the company has not been as aggressive in publishing data on the system. The system has been stressed with exercise, unannounced meals, sensor failure calibrations, lost transmission, maximal insulin delivery, and in pediatrics. This put into perspective the resource challenges other companies are probably facing – building a closed-loop device is seriously expensive and time consuming.
• Dr. Kaufman showed an automated insulin delivery landscape slide, listing competitors with Medtronic’s next-gen DreaMed system in the following order: Bigfoot, University of Cambridge, University of Virginia/TypeZero, Boston University, and Animas. The slide specified the algorithm types (proprietary, MPC) and listed all as 24-hour systems; there was no timing attached to any product. This was a unique gesture we have not ever seen in a closed-loop presentation, and it was interesting that Medtronic views these systems as competitors to its next-gen system after the 670G – though the 670G will come to market earlier than these systems, we had assumed at least some of these competitors could launch a year or two later. The slide was missing Tandem and Insulet, who also have active artificial pancreas programs.
  
  • As of November 2015, Medtronic estimates five artificial pancreas systems are in “early development”, five are pre-clinical, 22 are in clinical trials, one is in the approval process (the 670G), and one is “inactive” (we’re not sure who that refers to). The slide said this combines “looking online, gathering data, and trying to understand the landscape.” It is certainly far more systems than we are aware of from the traditional players, but perhaps it is counting academic closed-loop systems.

Further Integrating Care Together

Annette Brüls (VP and President, Diabetes Service and Solutions, Medtronic)

In a most valuable talk, Ms. Annette Brüls confirmed the timing from CES to launch a hypoglycemia prediction app with IBM Watson this summer. In new news, she shared plans to launch next-gen CareLink Pro reports in the next 12 months (including advanced analytics to optimize pump basal and bolus settings), along with future plans to build Bluetooth directly into Medtronic’s hybrid closed loop (eliminating MiniMed Connect; no timing stated). In the coming months, Medtronic will also launch some of the current CareLink Pro features into CareLink personal, countering longstanding complaints that patients should see the same thing as providers. It is great to see the company responding to this feedback. Throughout her talk, Ms. Brüls emphasized the “journey outside of devices” and towards “holistic solutions” that use real-time data from personal Medtronic devices, the Internet of Things, and Big Data to reduce the burden of diabetes. Of course, all these things intertwine, as Ms. Brüls showed in a compelling example of insightful Big Data (from CareLink). A colored map of the US compared dangerous hypoglycemia episodes (>3 hours at <50 mg/dl) in patients with threshold suspend turned off vs. on. Patients with threshold suspend turned off experienced 6.1 such episodes per year vs. just 0.8 episodes per patient per year with threshold suspend turned on. The slide suggested the difference (5.3 episodes hypoglycemia) equates to $735 in emergency services per patient per year and $9,300 in patient services per year. This is the most valuable kind of analysis, which should persuade payers of the value of technology in the real-world. The company’s investment in connectivity, apps, and digital health was crystal clear after this talk, and it’s remarkable to reflect on how different Medtronic Diabetes is from a few years ago – we wouldn’t have thought it was possible.

• The next-gen CareLink provider reports launching in the EU in the next 12 months will identify optimal basal and bolus wizard changes, show glycemic trends in a new way, give providers a clinic dashboard, and modify the report design to structure conversation and uncover problem areas. The pump settings optimization is something that HCT just funded DreaMed to build, and it’s impressive that Medtronic is already ready to launch this.
• Ms. Brüls was clearly excited about the IBM partnership and suggested that future apps could predict hyperglycemia or show glycemic profiles around specific meals – e.g., helping identify patients’ best meals or worst meals, giving recommendation in a particular situation based on previous behavior, or suggesting what happened in similar patients.
• A compelling colored map of the US compared dangerous hypoglycemia episodes (>3 hours at <50 mg/dl) in patients with threshold suspend turned off vs. on. Patients with threshold suspend turned off experienced 6.1 such episodes per year vs. just 0.8 episodes per patient per year with threshold suspend turned on. The slide suggested the difference (5.3 episodes hypoglycemia) equates to $735 in emergency services per patient per year and $9,300 in patient services per year. We love this sort of analysis, which we hope can persuade payers of the value of technology in the real-world.

Transforming Health: Leveraging Data, Analytics, and Cognitive Technologies

Shahram Ebadollahi, PhD (Chief Science Officer, IBM Watson Health)

Following Mr. Brüls introduction, IBM Watson Health’s Chief Science Officer Dr. Shahram Ebadollahi gave an outstanding, most valauble presentation on the growth of unstructured data, the company’s plans with Watson, and new data on the Medtronic hypoglycemia prediction app from an analysis of 2,000 users’ CareLink data. The early results from DTM are still holding, showing 80-85% hypoglycemia prediction accuracy three to four hours ahead of time.   

• Dr. Ebadollahi shared a new analysis from the development of the Medtronic hypoglycemia prediction app – this time with 2,000 users’ CareLink data (the DTM analysis was in 100 users). Watson analyzed data from MiniMed 530G and CareLink users with three to nine months of pump and CGM data. Watson takes into account long-term behavior (pump settings, frequency of excursions in the last month), short-term trends (current trend, sensor glucose, current carb size, bolus size), and demographic features (age, time since diagnosed, time on insulin) to figure out what is most indicative for prediction of hypoglycemia. Data was split into 80%-20% train-test ratio (80% historical patient data, more than three to nine months old) was used to train the classifier).
  
  • Watson analyzed the data, identified 10 clusters of patients, and developed specific hypoglycemia prediction models for each of them. An example of three of the clusters of patients: (i) young patients, early onset age, <5 years on insulin; (ii) early 30s, early onset (<10 years), >19 years on insulin; (iii) elderly, late onset, and long-term insulin users (>20 years).
  • “The goal is to get to an n of 1. Can we get to a prediction machine for every single patient?” Dr. Ebadollahi emphasized the importance of having group-specific prediction models, which are much more accurate than blanket prediction models for the whole population. It is hugely impressive that Watson can do this just by learning from the data.
  • Watson demonstrated 80%- 85% hypoglycemia prediction accuracy three hours ahead of time, and a remarkable 79-83% accuracy with a four-hour prediction window. The slide listed the prediction accuracy for all ten groups, and the lowest on the slide was an impressive 78.6%. 
• A list of IBM Watson partners notably excluded Novo Nordisk, who announced a partnership with the group in December. The slide included Medtronic Diabetes, CVS Health (adherence), J&J (orthopedics), Teva (real world evidence), Under Armor (exercise), and Apple HealthKit and Research Kit.
• Unstructured data is growing exponentially and projected to reach 44 zettabytes by the year 2020: we learned from Dr. Ebadollahi if grain of rice is a byte, a zettabyte will fill up the Pacific Ocean – that means 44 zettabytes is 44 pacific oceans. Dr. Ebadollahi emphasized that growing data volume and complexity demands a new approach. Structured data was the previous generation of data, though the predominant new types of data are unstructured: sensors and devices, medical images, images/multimedia, natural language, enterprise data. Every five years, he said, the available knowledge in medicine doubles.
• A vast amount of untapped data could have a great impact on our health – yet it exists outside medical systems. As evidence, Dr. Ebadollahi pointed to an oft-quoted study that health is comprised of 10% clinical factors (0.4 TB per lifetime), 30% genetics (6 terabytes of data per lifetime), and 60% exogenous factors (1000 terabytes per lifetime).

​SmartGuard: Let the MiniMed 640G do the work

Pratik Choudhary, MD (King’s College Hospital, London, UK)

Dr. Pratik Choudhary shared new real world data from 4,818 EU users of Medtronic’s MiniMed 640G/Enlite Enhanced system with SmartGuard. Based on data collected from CareLink personal uploads (this is a remarkably valuable dataset), SmartGuard performed remarkably well over the yearlong study period – in 693,623 total SmartGuard events, 75% occurred without sensor glucose levels ever reaching the pre-set low limit (3.6 mmol/l, 65 mg/dl on average). Further, the mean sensor glucose value at suspend was 5.2 mmol/l (94 mg/dl), glucose returned to a normal range within two hours in 90% of SmartGuard events, and users spent ~70% time “in range” – between 3.9 and 10 mmol/l (70-180 mg/dl). The number was not compared to baseline, so it’s hard to know if this was an improvement; we assume it was. Notably, the Enlite Enhanced sensor had a MARD of 11.9% in this study, suggesting improved accuracy over the original Enlite (MARD of 13-14%). While the system achieved impressive hypoglycemia prevention, it did not appear to increase glycemic variability, and insulin delivery auto-restarted to prevent rebound hyperglycemia in two-thirds of SmartGuard events. Dr. Choudhary repeatedly emphasized that SmartGuard resulted in less rebound hyperglycemia compared to low glucose suspend systems. Dr. Choudhary repeatedly highlighted the provider and patient learning curve with the 640G – the challenge is learning to trust the system and not take action when it suspends in the target range to prevent a low. In a workshop yesterday, he highlighted that the team now turns off the predicted low alarm, as the 640G takes care of lows in background better than users do when the alarms go off. 

• Dr. Choudhary shared data collected from CareLink personal uploads from 4,818 EU patients using Medtronic’s MiniMed 640G/Enlite Enhanced system with SmartGuard. Patient ages ranged from 0 to 56+, with the greatest number under the age of 15. Perhaps the earliest adopters of the 640G have been younger patients, or perhaps parents of children with diabetes are far more likely to upload to CareLink than adults with diabetes.
• Users spent an impressive ~70% time “in range” – between 3.9 and 10 mmol/l (70-180 mg/dl), and <2% of the time ≤ 2.7 mmol/l (50 mg/dl). The proportion of time spent between 2.8 and 3.9 mmol/l (50-70 mg/dl) and ≥ 16.7 mmol/l (300 mg/dl) was also low (~4% and ~7%, respectively). Average time spent in the 10-16.6 mmol/l (180-300) range was larger at ~30% of a 24-hour period. The number was not compared to a baseline value; we assume it did improve, but are not sure of the magnitude. We love moving toward this kind of analysis that goes so far beyond A1c.
• The SmartGuard system perform particularly well at night, with sensor glucose levels reaching the pre-set low limit in only 24% of events. By comparison, 26% 0f SmartGuard events that occurred during the day reached the pre-set low glucose limits. This is excellent real-world efficacy in our view – three out of four times, SmartGuard completely avoids a low.
• Dr. Choudhary also shared results from a within patient comparison of patients who switched from Veo to the 640G (n=851), demonstrating that patients had 0.26 less hypoglycemic events per day when using SmartGuard. Notably, this occurred without a coinciding increase in hyperglycemia.
• Mr. Choudhary positioned these results as corroborating those from the 40-patient user evaluation study that we reported on at last year’s ATTD, though we would point out some subtle differences: the real-world MARD was higher than that demonstrated by the user evaluation (11.9% vs. 9.8%), and the percentage of SmartGuard events that reached the low limited was substantially higher than we wrote last year (25% vs. 3%). However, it is tough to compare the two given the very different sample sizes (40 vs. almost 5,000 participants), and the study biases inherent to the user evaluation study (patients knew they were being studied and possibly performed better as a result) did not exist for this very real world data analysis.
• In terms of future data analysis, we would be curious about how users’ A1cs may have changed during the study period (did A1c increase as a result of reduced hypoglycemia?), as well as how the SmartGuard system might affect behavior – does a behavioral response to predicted low glucose (consuming carbohydrates) combined with a system response (shutting off insulin) cause rebound hyperglycemia?

Dexcom Continuous Glucose Monitoring - Leadership And Innovation In The Category Of CGM (Sponsored by Dexcom)

Dexcom CGM Beyond 2016

Jorge Valdes (CTO, Dexcom, San Diego, CA)

Dexcom CTO Jorge Valdes shared a few small pipeline updates: (i) the new G5 smaller transmitter (expected in late 2016 or early 2017, per JPM) will cut the volume in half vs. the current transmitter (previously, no size reduction estimate had ever been given); (ii) G5 Android is still expected to launch this year, though Dexcom won’t be able to cover every single Android phone – it will support the major brands, given the sheer number of Android phones; (iii) Dexcom will incorporate predictive hypoglycemia alerts into G6, offering a 15-minute prediction window – the design is balancing more advanced notice without increasing false alarms or annoying patients. The predictive alerts were the focus of many slides for the first time, which made us wonder if it was an answer to Medtronic’s hypoglycemia prediction app with IBM (which has a 2-4 hour prediction horizon, but could obviously increase alarm fatigue). The new predictive alert is clearly effective: with its addition, only 9% of 55 mg/dl BG events would give patients <15 minutes of warning vs. a much higher 42% with just a threshold alert alone. Overall, the new feature has a 93% predictive hypoglycemia alert detection rate (for 93% of YSI readings <55 mg/dl, a predictive alert occurred between 0-30 minutes), but not at the cost of additional nuisance: for just 11% of predictive alerts, there was no YSI reading below 70 mg/dl within the next 30 minutes, making these alerts unnecessary). We’re glad to see Dexcom balancing valuable prediction with hassle factor – it doesn’t matter if an alarm is accurate if it is annoying!   

• Mr. Valdes mentioned several times that Dexcom’s G5 is accurate enough to dose insulin on (consistent with the EU label that it replaces fingersticks), though we noticed a slight step away from the company’s longstanding emphasis to eliminate fingersticks – he mentioned that dosing insulin can be lethal, and Dexcom wants to keep calibrations in for safety. Was this an indirect comment to address Abbott’s factory calibrated FreeStyle Libre, which also has a dosing claim in Europe but requires no calibrations? It also might reflect expectations for G6 to include one fingerstick calibration per day. The final slide on “Dexcom’s R&D Mission” also excluded eliminating fingersticks (though we loved the clear goals):
  
  • Dexcom R&D mission. “Best in class CGM experience: simple, welcoming out-of-box experience; reliable, actionable, and helpful CGM; automated, friendly troubleshooting and support. An ecosystem of tools for patients to make life simple and improve outcomes: Integration with insulin systems, meal, activity tracking; retrospective patterns and insights; decision support.”
  • In a follow-up conversation, management shared the following helpful perspective: “We are not backing off from eliminating fingersticks. But going forward we can see multiple classes of products, some requiring calibration and some not depending on the use case. We do believe that all sensors should allow calibrations, whether required or not, in order to cover the cases when the glucose reading do not align with the way a patient feels.”

T1D Exchange Data - The Largest Analysis Of CGM Use Informs The Future

David Price, MD (VP Medical Affairs, Dexcom, San Diego, CA)

Dr. David Price shared CGM data from the T1D Exchange, highlighting increasing adoption (now at 15%; total n=16,853) and a strong preference for Dexcom (63%) over Medtronic (27%).   Nearly 80% of Dexcom users use CGM >25 days per month vs. ~45% of Medtronic users. The Exchange is not truly representative of the US type 1 population – its pump penetration is at 60% – though we think it provides an instructive look at what leading centers are doing. Dr. Price also discussed Exchange data comparing A1cs in MDI, pump, MDI+CGM, and MDI+pump, showing that in every age group, the same pattern holds – lower A1cs with CGM regardless of insulin delivery method (~0.6%-1.3% lower), and similar A1cs for CGM users on MDI or a pump. Dexcom has shown this in several exhibit halls, though we agree it is a compelling to counteract that prevailing belief that a CGM should come after a pump.

Is It Time To Move To Time In Range As The Main Glycemic Control Measure?

IT IS TIME TO MOVE TO TIME IN RANGE(S) AS the Main* MEASURE OF GLYCEMIC CONTROL

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

Dr. Aaron Kowalski enthusiastically supported using time-in-range as the main measure of glycemic control, and more broadly, asserted that interventions must balance “diabetes health” and “diabetes happiness.” His closing slide showed him running a marathon with type 1 diabetes, summarizing the presentation aptly: “A1c is not what I’m thinking about when I’m at the starting line.” Dr. Kowalski’s talk noted many of the limitations of A1c, highlighted the intuitive advantages of time-in-range (though some challenges to iron out), and mentioned JDRF’s new T1D Outcomes Program, which aims to define the metrics of importance to people with type 1 diabetes, providers, researchers, industry, FDA, and payers – the steering committee includes all the major professional associations, and we sincerely hope this can move the needle on defining outcomes beyond A1c, and more importantly, getting them accepted. Following the advocacy success of getting time-in-range into the FDA artificial pancreas guidance, JDRF is now working on a “similar pathway for drugs and biologics.” Now that could be impactful, particularly for non-insulin interventions for type 1 like SGLT-2 inhibitors or ultra-rapid insulins. More broadly, Dr. Kowalski’s talk reiterated many of the elements of his Diabetes Scorecard (Diabetes Care 2015), emphasizing the need for better “value” in type 1 diabetes, which is defined differently for patients, providers, and payers – and is not just A1c, but includes device burden, time investment, fear, provider quality metrics, hospitalizations, etc.! The most successful therapies will improve value for all three stakeholders; we wonder if automated insulin delivery can meet that bar.

• Dr. Kowalski noted the limitations of A1c, despite its status as a validated surrogate for long-term outcomes: (i) long-term measurement – not reflective of day to day; (ii) not very actionable; (iii) not reflective of hypoglycemia; (iv) not reflective of variability; and (vi) not informative regarding glucose perturbations cause and effects.
• Dr. Kowalski called time-in-range much more intuitive, as it provides: (i) both hyperglycemia and hypoglycemia transparency; (ii) direct visibility to both short-term and long-term diabetes risks; (iii) a more physiologic target for improved glycemic control; and (iv) the best representation of diabetes glycemic health.
  
  • Still, Dr. Kowalski acknowledged that there are a few nuances to time-in-range that must be ironed out: (i) What is the ideal range? (is it 70-180 mg/dl, which is generally accepted in the artificial pancreas world and by FDA? Or is it 70-105 mg/dl, the non-diabetes physiological range?); (ii) What is the goal – how much time should be spent in each range?; and (iii) time-in-range is CGM-dependent to a large degree; what do we do for those using BGMs alone?
  • In Q&A, Dr. Irl Hirsch pressed the panel for specific recommendations on time-on-range, to which Dr. Kowalski admitted it’s will be “very hard” – it might be better for clinicians to think in terms of improvement. For instance, it will be hard to come to a consensus on what threshold for time-in-range is optimal – 75%? 100%? Instead, one could look at simply improving time-in-range, or striving to reduce time-out-of-range. This seems a most valuable recommendation, since it is actionable, realistic, and could align much more with the value-based healthcare system toward which things are moving.
• Dr. Kowalski ran through several examples of old, burdensome devices that may have improved glycemic control, but weren’t realistic for patients to wear in the real world (e.g., the infamous backpack dual-hormone artificial pancreas; the bedside biostator; the “blue brick” auto-syringe insulin pump). For these devices, the ROI for patients wasn’t there to encourage wide adoption. By contrast, he noted that modern CGM devices (the slide showed Dexcom’s G5, Medtronic’s MiniMed 530G, and Abbott’s FreeStyle Libre) have dramatically improved the ROI for patients – and thus, will pave the way for greater effectiveness and greater adoption.

We Need Standard Definitions and Standard Glycemic Reports

Richard Bergenstal, MD (Park Nicollet International Diabetes Center, Minneapolis, MN)

Dr. Richard Bergenstal discussed the need for standardized CGM data reports – an “ECG for glucose patterns” – that incorporate information beyond A1c. He stressed that he does not advocate throwing out A1c entirely but suggested that, at least to start, clinicians should aim to optimize A1c while minimizing hypoglycemia. This could eventually be replaced by evaluating time in range and time in hypoglycemia. In a slide that would have been at home in a self-help book, he promised clinicians “four steps to transform your practice”: cut the cords (on old devices), aggregate blood glucose data from multiple devices, generate a standard report, and agree on an action plan (key word: agree). The rest of his presentation focused primarily on an example of a standard report, his Ambulatory Glucose Profile (AGP) developed with feedback from an expert panel he convened on the issue in 2012. The AGP is a one-page report with relevant glycemic numbers at the top, followed by a simple visualization of glucose patterns. In this talk, he focused mainly on which numbers should be included in the report, noting that is crucial to standardize the definitions of hypo- and hyperglycemia. His report includes percent time below 70 mg/dl, 60 mg/dl, and 50 mg/dl, which he suggested corresponds clinically to “low,” “very low,” and “dangerously low.” Similarly, he argued that reporting percent time above 180 mg/dl, 240 mg/dl, and 300 mg/dl is more useful for patients than lumping all the values together as hyperglycemia. He suggested that in a research setting, more complex measures like area under the curve that incorporate both the duration and severity of hypoglycemia could be even more useful. His main point was that this data needs to be standardized in some way, not that he is wedded to his particular method.

ESCAPING THE A1C-CENTRIC ROLE OF ASSESSING GLYCEMIC CONTROL IN DIABETES

Robert Vigersky, MD (Medical Director, Non-Intensive Therapies, Medtronic Diabetes, Washington, DC)

Medtronic Diabetes’ Dr. Bob Vigersky reiterated highlights from the outstanding talk he gave at DTM 2015 on novel visual and numerical representations for capturing composite diabetes outcomes (A1c, hypoglycemia, weight). He argued that composite scores are beneficial because they allow for the comparison of efficacy across different types of interventions – pharmacologic technologic, and psycho-educational. To illustrate this point, Dr. Vigersky reviewed an impressive range of visual and numerical representations of composite outcomes from a variety of studies. His discussion included (i) the glucose pentagon (DT&T 2009 and JDST 2012), a single graph and number combining five elements of glycemia (A1c; SD; time >160 mg/dl; AUC > 160 mg/dl; and mean glucose); (ii) the Q-score (BMC Endocrinologist Diab 2015), a single numerical value combining five primary factors that determine CGM profiles (central tendency, hyperglycemia, hypoglycemia, intra- and inter-daily variations); and (iii) his own novel approach (published last year in JDST) combining A1c, hypoglycemia, and weight change in a single score out of 100. Echoing his concluding commentary from DTM 2015, Dr. Vigersky advocated for “escaping” the A1c-centric world in evaluating diabetes interventions and working with regulatory bodies, clinicians, and industry to agree on a composite metric to better describe overall glycemic control. We agree that this is particularly essential for many next-gen therapies such as CGM and closed loop devices, which may not show improvements in A1c but do reduce hypoglycemia and improve time-in-range.

Panel Discussion

Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): We’ve heard A1c by itself is not enough. But with a composite outcome score, would we miss out on the individual components?

Dr. Vigersky: I think A1c could be replaced by time-in-range. It makes sense. What I didn’t show was that with the Q-score, you can select out areas of concern that can be addressed with that score in a graphic way. Bar graphs can be generated for the most important elements, like standard deviation or hypoglycemia. These can be easily generated from the data that exists.

Dr. Bergenstal: I am all for an aggregate score, and I thought this symposium was amazing. On an individual basis, however, I would plea that we never get away from having conversations around the actual profile, a person’s life picture.

Dr. Vigersky: I wouldn’t argue with that at all. We have a hard time explaining A1c to patients. We’ll have an equally hard time explaining whatever metric we choose. For studies, payers, and regulators, however, we can’t show a picture – we have to show them numbers.

Dr. Rodbard: You have high, low, in target range, average, and variability. I’m an advocate of time-in-range. If you have a high percentage in the target range, you need to have a mean glucose in the center of the range, and you need to have small variability. Conversely, if you have a mean in the center and small variability, you have a small percentage of low and highs. But you want to know, “Where is my biggest problem – lows, high, or variability?”

Q: One of the things that struck me was that you are striving hard to come up with composite measures for glucose. However, Aaron showed that on a patient basis, one of the most useful things to have is a temporal glucose profile to show trends throughout the day, and particularly what’s going on at night. Composite measures lose some of that information. A new technique that has come out recently is the functional data analysis technique, which allows temporal glucose profiles to be analyzed at a population level where you can determine the significant differences between interventions and determine where these differences are occurring. We published this in Diabetes Care last year, but it might be useful to know about.

Dr. Kowalski: It comes back to simplicity. This audience is self-selected for early adopters and technophiles, and the fact that we still have to explain A1c is telling. There’s something to be said for simplicity. In other audiences, such as regulatory, there’s a place for deeper complexity. But in the clinic, simple is better.

Dr. Irl Hirsch (University of Washington, Seattle, WA): As we’ve been sitting here, we all recognize that A1c does not tell us enough. It’s actually a little more complex. In the ADAG study, we saw that if you have an A1c of 8%, and I have an A1c of 7%, it’s quite possible that your mean blood glucose is lower than me, because we glycate hemoglobin differently. And there are 14% of people for whatever reason in which A1c doesn’t work – things like anemia. A1c has all kinds of problems. I think time-in-range is the way to go. But we’ve been talking about it in very broad strokes. If we’re going to talk about individualization of patient care, we need specifics. We need to look at time-in-range like we have a target A1c <7%. What is the target for time-in-range, amount of time in hypoglycemia, overnight, during day, or climbing a mountain. We need specifics as opposed to just continuing to talk about getting something more than A1c.

Dr. Kowalski: It’s really hard. If I think about it, it’s going to be very hard to set a euglycemic target. The tools do not allow people to get there. I think about better. So if you have hypoglycemia issues or hyperglycemia issues, can we reduce those.

Bergenstal: Coming to consensus on time-in-range will be tough. For instance, 3% hypoglycemia is fine, but is that under 70 mg/dl or under 50 mg/dl. It gets a little more complicated fast.

Brandon Arbiter (Tidepool): Time in range is a much more representative measure than A1c. Is having TIR dependent on wearing a CGM sensor? If only 10-15% of people with type 1 diabetes in the US are wearing it, what are the implications for trying to adopt time in range as a metric?

Dr. Kowalski: Everyone with type 1 should be on a CGM.

-- by Melissa An, Adam Brown, Varun Iyengar, Emily Regier, Ava Runge, and Kelly Close