September 12-16, 2016; Munich, Germany; Full Report – Diabetes Technology – Draft

Executive Highlights

As in past years, EASD was relatively light on the tech side, but concurrently announced publications and diabetes data partnerships stole the show in Munich. On the penultimate day of the meeting, Dr. Rich Bergenstal revealed that Medtronic’s MiniMed 670G US pivotal trial was published as a two-page Research Letter in JAMA. In the preceding days, we learned that Abbott’s IMPACT study (demonstrating a significant reduction in hypoglycemia in type 1s using FreeStyle Libre) was published in The Lancet and that a randomized crossover trial demonstrating fewer severe hypoglycemia events with CGM vs. SMBG in type 1s had been published in The Lancet Diabetes and Endocrinology. We were glad to see three heavy-hitting papers published in prominent medical journals – a sure sign that the diabetes technology field continues to mature as CGM improves and automated insulin delivery comes to market.

Data management platforms Diasend and Glooko shook up the diabetes data world by revealing a merger during EASD. The unified company aims to deliver the “World’s Premier Data Management Platform,” and we believe the combined companies will move much faster together. . 

We also had our first look at Roche’s new Insight CGM system (seven-day wear with two fingersticks per day, preliminary data suggests MARD of 10.5%) – read on for more on form factor and how we see the product stacking up to the competition! This report also includes coverage of corporate symposia from Abbott and Dexcom.

Talk titles highlighted in yellow were some of our favorites from the meeting, reflecting what we found to be most notable. Talk titles highlighted in blue are new full report additions, and were not part of our daily highlights coverage while we were in Munich.

Themes

Major Journal Publications: JAMA (670G), Lancet (Libre), Lancet Diabetes & endocrinology (CGM & Severe Hypo)

• EASD is a small meeting for diabetes tech, but we were glad to see three notable journal publications released in tandem with the conference: (i) Medtronic’s MiniMed 670G US pivotal trial was published as a two-page Research Letter in JAMA (an oral at this EASD and originally a late-breaking ADA poster); (ii) Abbott’s FreeStyle Libre IMPACT study was published in the prominent UK medical journal The Lancet (also presented in poster form at ADA); and (iii) Lancet Diabetes and Endocrinology published a robust randomized crossover study showing CGM reduces severe hypoglycemia in type 1s with impaired awareness (also an oral at this EASD). (As an honorable mention, about a month before the meeting began, Diabetes Technology and Therapeutics published the 52-week COMISAIR study showing that CGM confers significant benefits in MDIs.) We’re glad to see the field of diabetes technology maturing to longer studies, broader populations, and publications in high-profile journals.
  
  • We previously covered most of the MiniMed 670G pivotal data from the ADA late-breaking poster, but the JAMA Letter included a few additional metrics: mean and median glucose in the trial (~150 mg/dl); time spent >300 mg/dl (a strong 26% decline), nocturnal time-in-range data (excellent), within-day coefficient of variation, and median values for all CGM metrics. We assume this could not be an “original article” as a single arm, non-randomized study, which Medtronic obviously choose for speed and getting to market quickly – we’re glad to see that was prioritized over a perfect pre-market RCT that gets published in long-form in a major journal, but takes longer to complete. To us, the big takeaway is that the 670G is safe and effective (even in a motivated patient population), and it will be a good first-generation starting point to learn and improve upon. The device was approved in the US on September 28 after a remarkably ~3-month priority review; launch is expected in Spring 2017 in the US. International approval is expected in Summer 2017.
  • Abbott’s IMPACT study of FreeStyle Libre in type 1s (a poster at ADA) was published in the prominent UK medical journal The Lancet, alongside a balanced (but mostly positive) accompanying Comment from Cambridge’s Dr. Roman Hovorka and colleagues. This publication represents a major victory for the company and the field and hopefully a positive for broad EU reimbursement. Getting the dramatic 38% hypoglycemia reduction (~74 minutes fewer per day) published in such a prestigious journal helps build the evidence base and cost-effectiveness proposition around this device, corroborates the REPLACE outcomes study in type 2 (presented at ATTD), and certainly reduces the noise factor around “no alarms.”
  • In a major outcomes victory for CGM, a randomized crossover trial in type 1s with impaired hypo awareness (published in the Lancet Diabetes and Endocrinology) demonstrated a significant 53% fewer severe hypoglycemia events with CGM vs. SMBG (14 events vs. 34 events; p=0.03). The very well conducted trial (out of the VU university hospital and Academic Medical Center, both in Amsterdam) randomly assigned 52 patients to either (i) 16 weeks of CGM followed by SMBG (n=26); or (ii) 16 weeks of SMBG followed by CGM (n=26). The study used the Medtronic Enlite CGM running on the Veo pump, but used only as a receiver (no low glucose suspend turned on). Severe hypoglycemia (requiring third-party assistance) was actually a secondary endpoint, but boy was it powerful to see the major change in this high-risk population in a fairly small study – that the SMBG phase included 34 severe hypoglycemia events in 52 patients over just 16 weeks was striking. Notably, the proportion of patients with >1 severe hypoglycemia event was 10 (CGM) vs. 18 (SMBG), translating to a compelling 0.48 odds ratio (borderline significant at p=0.06). As expected, time-in-range (70-180 mg/dl), significantly improved during the CGM phase (65% with CGM vs. 55% with SMBG; p<0.0001), translating to over two extra hours in range per day with CGM. This included a whopping 40% reduction in time spent <70 mg/dl (7% vs. 11%; p<0.0001) and 15% less time spent >180 mg/dl (28% vs. 33%; p<0.0001). A1c was not significantly different (-0.1%), meaning glycemic control was not ceded for the hypoglycemia benefits – yes! This trial further builds the case for CGM’s cost-effectiveness (even with an older CGM sensor), and we especially hope it helps convince payers that this technology saves money in the short-term and long-term.

Digital Health and Diabetes Data – the Era of Partnerships?

• In major news during EASD, Diasend and Glooko merged into one joint company under the Glooko name and will create a unified data management platform in the next year. The company will be led by current Glooko CEO Rick Altinger, while Diasend CEO Anders Sonesson will manage Global Operations. Combined, the two companies cover 95% of devices and are currently deployed at 4,000+ health systems in 23 countries and 15 languages – wow! We spoke to reps from both sides at the Glooko booth, who were eager to share details on the merger aimed at creating the “World’s Premier Diabetes Management Platform.” We see tremendous synergy in the combined company – Diasend brings existing integration with 140+ devices, a sizable global footprint (23 countries), and reimbursement expertise in a wide variety of markets, while Glooko brings strong user interface, great data analytics and reporting, population tracking, and pipeline of insulin dose titration algorithms. We see this as excellent news for the field, given the upside from a go-to global diabetes data repository that works with every device, enables easy zero-hassle upload (in clinic, at home, in real-time), and gives patients and providers quick, actionable analysis. We believe both companies – and diabetes data in general – will move faster by combining efforts. See our detailed thoughts on the merger and an interview with the Glooko and Diasend teams.
• A new integration agreement between Abbott and mySugr will allow Libre users to automatically and passively upload data to their mySugr accounts via the LibreLink Android app. We’re glad to see Abbott now enabling a broader data ecosystem around Libre, as the company has arguably made fewer moves than Medtronic and Dexcom on this front. This news is also a win for the widely loved mySugr app, which has nearly 800,000 users globally (!) and a partnership with Roche’s Accu-Chek Connect BGM. The Abbott team also told us about a new partnership with Social Diabetes, a popular Android app in the EU that enables users to log their food and insulin and get dosing recommendations from the app. We have not ever heard of this app, but it has a notable 50,000+ downloads on Android and very positive reviews. With the hot-off-the-press FDA approval of Libre Pro two weeks ago and the 3Q16 submission of the consumer version (approval expected as early as 1Q17, per the 2Q16 financial update, which occurred before submission), these data integration agreements could be relevant for American customers very soon.

The CGM vs. FGM Debate – When Should Each Be Used?

• A small, head-to-head, investigator-initiated trial compared the hypoglycemia impact of Dexcom’s G5 (n=15) vs. Abbott’s FreeStyle Libre (n=17) in type 1s with very impaired hypoglycemia awareness. Dexcom’s G5 bested Libre on hypoglycemia, showing a significant reduction in time spent <60 mg/dl (3.3 mmol/l): from 6.1% at baseline to 3.3% in the last four weeks of the study (p=0.02). By contrast, the FreeStyle Libre group saw no significant change in hypoglycemia (from 7.4% to 8.2%; p=0.5), for a between-group difference of 3.6% (p=0.03). This translated to a clinically significant 52 minutes per day less time in hypoglycemia (<60 mg/dl) with G5 vs. Libre in a very high-risk group, and all other hypoglycemia thresholds were also in favor of G5 and statistically significant. Notably, there were no significant between-group differences in time-in-range or time in hyperglycemia – both groups saw strong and nearly equal improvements, which was a positive for Abbott in our view (despite alarms, time in range was the same as with Dexcom). Dr. Nick Oliver emphasized that the alarms really drove the greater hypoglycemia improvement in the G5 group, a key device feature for patients with a high degree of impaired hypoglycemia awareness. He characterized the FreeStyle Libre as “reflective” and the G5 as “reactive,” and noted that hypoglycemia measurement (frequency, severity) and awareness should be considered before commencing CGM – presumably to stratify patients to see if they need alarms. The study is not yet complete and will report more data in early 2017.
  
  • The temptation is to say “G5 > Libre,” but this misses the more interesting point: when should one glucose sensing technology be used vs. another? All else equal (cost, size on the body, hassle, patient preference), high-risk patients with significant hypoglycemia unawareness SHOULD have the benefit of CGM alarms. Of course, access and patient preferences are rarely equal between two options, which is why both of these technologies exist in the first place – offering a broader spectrum of device form factors, cost, and features to target different patient populations. This is not an “either-or” but a “both-and” – how can we bring glucose sensing to the masses in the most cost-effective way? CGM has not been accessible to nearly enough people to date, which is what motivated Abbott to rethink the design with FreeStyle Libre. We’re glad to see Dexcom/Verily doing the same and we hope both companies can expand the market significantly beyond those studied in this trial (who will most certainly benefit from traditional CGM, and for whom reimbursement should be a no-brainer).

Detailed Discussion and Commentary

Oral Presentations: Devices and Desires

Hybrid closed-loop (HCL) pivotal trial in type 1 diabetes

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

JAMA published Medtronic’s MiniMed 670G US pivotal trial as a two-page Research Letter, concurrently presented as an oral presentation by Dr. Richard Bergenstal this afternoon. We saw the data in a late-breaking ADA poster in June, but it was excellent to see the visibility in JAMA (even as a “research letter”), and we were glad to see several metrics we did not see in the ADA poster: mean and median glucose in the trial (~150 mg/dl); time spent >300 mg/dl (a strong 26% decline), nocturnal time-in-range data (excellent), within-day coefficient of variation, and median values for all CGM metrics. The data is otherwise identical to what we saw at ADA and reinforces the topline takeaways: (i) a strong safety profile; (ii) a solid 0.5% reduction in A1c from a low baseline (7.4%); (iii) excellent hypoglycemia avoidance (time <70 mg/dl declined 44%; (iv) outstanding efficacy overnight (75% time in 70-180); and (v) much-improved accuracy for Enlite 3 (MARD: 10.3%). A group of serious luminaries in the field authored the JAMA letter – Drs. Rich Bergenstal, Satish Garg, Stu Weinzimer, Bruce Buckingham, Bruce Bode, Bill Tamborlane, and Fran Kaufman. We assume this could not be an “original article” as a single arm, non-randomized study, which Medtronic obviously choose for speed and getting to market quickly – we’re glad to see that was prioritized over a perfect pre-market RCT that gets published in long-form in a major journal, but takes longer to complete. As would be expected, the JAMA Research Letter places less focus on efficacy: “Safety” is the first word in the piece’s title, “Table 1” shows device related adverse events (“Table 2” shows the study outcomes), and no p-values are shown comparing the two-week baseline to the intervention period. We hope readers can infer the efficacy from the Research Letter’s table, though it will be interesting to see if closed loop shows better “real-world” efficacy than in trials. Managing expectations will of course be critical as this comes to market, something we heard over and over again at ADA. The 670G was submitted to FDA in June, and as of the 2Q16 call earlier this month, was “on track” for approval by April 2017. 

• To us, the big takeaway is that the 670G is safe and effective (even in a motivated patient population), and it will be a good first-generation starting point to learn and improve upon. Patients in the pivotal clearly loved the system (80% chose to continue using it), a sign that these first-gen products will appeal to current pumpers. Though we’ve heard the 670G requires a fair amount of setup and the user interface needs improvement, it should reduce burden from what many are currently doing every day, and it certainly gives amazing peace of mind overnight and a great morning glucose to start the day.

Real-time continuous glucose monitoring improves time spent in euglycaemia and prevents severe hypoglycaemia in type 1 diabetes mellitus patients with impaired awareness of hypoglycaemia

Koen van Beers, MD (VU University Medical Center, Amsterdam, Netherlands)

In a huge outcomes victory for CGM, a randomized crossover trial in type 1s with impaired hypo awareness demonstrated a significant 53% fewer severe hypoglycemia events with CGM vs. SMBG (14 events vs. 34 events; p=0.03). The study was concurrently published in the Lancet Diabetes and Endocrinology. The very well conducted trial (out of Dr. Hans DeVries’ group at Academic Medical Center) randomly assigned 52 patients to either (i) 16 weeks of CGM followed by SMBG (n=26); or (ii) 16 weeks of SMBG followed by CGM (n=26). A 12-week washout came in between the periods, and the study enrolled type 1 adults with impaired hypoglycemia awareness, a baseline A1c of 7.5%, a mean age of 49 years, 31 years of diabetes duration, and 56% MDI users. The CGM study phase used the Medtronic Enlite CGM running on the Veo pump, but used only as a receiver (no low glucose suspend turned on). A blinded iPro2 device collected CGM data during the SMBG phase. The primary endpoint, time-in-range (70-180 mg/dl), significantly improved during the CGM phase: 65% with CGM vs. 55% with SMBG (p<0.0001), translating to 2.3 extra hours in range per day with CGM. This included a 40% reduction in time spent <70 mg/dl with CGM (6.8% vs. 11.4%; p<0.0001) and 15% less time spent >180 mg/dl (28.2% vs. 33.2%; p<0.0001). Severe hypoglycemia (requiring third-party assistance) was actually a secondary endpoint, but it was still powerful to see the huge change in this high-risk population – that the SMBG phase included 34 severe hypoglycemia events in 52 patients over 16 weeks was striking. Notably, the proportion of patients with >1 severe hypoglycemia event was 10 (CGM) vs. 18 (SMBG), translating to a compelling 0.48 odds ratio (borderline significant at p=0.06). A1c was not significantly different (-0.1%), meaning glycemic control was not ceded for the hypoglycemia benefits – yes! CGM surprisingly did not significantly improve awareness of hypoglycemia (as measured by Gold Score) after 16 weeks, though perhaps a longer trial would be needed. Interestingly, Lilly and Sanofi funded the trial. We think the results would be even better with a more recent device.

• This trial further builds the case for CGM’s cost-effectiveness (even with an older CGM sensor), and we especially hope it helps with European reimbursement. The publication neatly summarized the study implications: “In earlier trials, CGM did not live up to the expectations of the diabetes community regarding its ability to reduce severe hypoglycemia. However, our findings here support the benefit of CGM, both with and without combining it with continuous subcutaneous insulin infusion, for improving glycemic control and diminishing severe hypoglycemia in adult patients with type 1 diabetes and impaired awareness of hypoglycemia, who are at highest risk of severe hypoglycemia.”
• There was no carryover effect, meaning the time-in-range improvement from CGM came solely from wearing it and using the data real-time (not a learning effect that persisted). See below for the time-in-range outcome for the CGM-SMBG sequence (red line) and SMBG-CGM sequence (blue line). CGM immediately benefited time-in-range, and as soon as it stopped, patients returned to their baseline. There was also no interaction between insulin treatment modality (MDI vs. pump), confirming Dexcom’s recent studies (DIaMonD study, COMISAIR) showing the benefits of CGM in MDI users.

Corporate Symposium: Dexcom Continuous Glucose Monitoring: Leading the Way as the Standard of Care, Clinical Outcomes and Market Access

Impact On Hypoglycemia Awareness Of Real-Time CGM And Intermittent Continuous Glucose Data (I HART CGM)

Nick Oliver, MD (Imperial College, London, UK)

Dr. Nick Oliver presented preliminary data from a fascinating investigator-initiated head-to-head study comparing the hypoglycemia impact of Dexcom’s G5 (n=15) vs. Abbott’s FreeStyle Libre (n=17) in type 1s with very impaired hypoglycemia awareness (one severe event in the past 12 months or a Gold Score >4). Following a two-week baseline run-in with blinded Dexcom G4, CGM-naïve patients on MDI (baseline A1c: 7.5%) were randomized to use either G5 or FreeStyle Libre for eight weeks. Dexcom’s G5 won handily on the primary hypoglycemia endpoint, showing a significant reduction in time spent <60 mg/dl (3.3 mmol): from 6.1% at baseline to 3.3% in the last four weeks of the study (p=0.02). By contrast, the FreeStyle Libre group saw no significant change in hypoglycemia (from 7.4% to 8.2%; p=0.5), for a between-group difference of 3.6% (p=0.03). This translated to a clinically significant 52 minutes per day less time in hypoglycemia (<60 mg/dl) with G5 in a very high-risk group. All other hypoglycemia thresholds were also in favor of G5 (<70, <63, <50 mg/dl) and statistically significant. Notably, there were no significant between-group differences in time-in-range or time in hyperglycemia – both groups saw strong and nearly equal improvements. Dr. Oliver emphasized that the alarms really drove the greater hypoglycemia improvement in the G5 group, a key device feature for patients with a high degree of impaired hypoglycemia awareness. He called FreeStyle Libre “reflective” and G5 “reactive,” and noted that hypoglycemia measurement (frequency, severity) and awareness should be considered before commencing CGM – presumably to stratify patients to see if they need alarms. The study is not yet complete and will report more data in early 2017. Dr. Oliver emphasized that these are preliminary results and will hopefully be published at some point in the future.

• Though these were preliminary results from a small trial, we were glad to see a very strong study design (randomized, parallel group, strong parity, latest devices), an investigator initiated study, a test of these two great technologies head-to-head, and an important study in a high-risk patient group. These comparative effectiveness studies must happen to give providers, payers, and patients better data – who is most likely to benefit from a particular technology?
• The temptation is to say “G5 > Libre,” but this misses the more interesting point: when should one glucose sensing technology be used vs. another? High-risk patients with significant hypoglycemia unawareness SHOULD have the benefit of CGM alarms – this small study provides clear data on that front, assuming equal access and patient preferences. Of course, access and patient preferences are rarely equal between two options, which is why both of these technologies exist in the first place – offering a broader spectrum of device form factors, cost, and features to target different patient populations. This is not an “either-or” but a “both-and” – how can we bring glucose sensing to the masses in the most cost-effective way? CGM has not been accessible to nearly enough people to date, which is what motivated Abbott to rethink the design with FreeStyle Libre. We’re glad to see Dexcom/Verily doing the same and we hope both companies can expand the market significantly beyond those studied in this trial (who will most certainly benefit from traditional CGM, and for whom reimbursement should be a no-brainer).
• As might be expected, the sensor failure rate was higher in the FreeStyle Libre arm (7/60 sensors, 11.7%) vs. the G5 arm (1/120 sensors, 0.8%) – each group had only one sensor with unacceptable accuracy, though the Libre group had six out of the 60 sensors fall out before the 14-day life. We have heard some complaints about this aspect of FreeStyle Libre – particularly for very active patients – and we wonder if Abbott will work on improving the adhesive or providing patients with over-tape.

 HYPOGLYCEMIA RESULTS

• Dexcom’s G5 won handily on the primary hypoglycemia endpoint, showing a significant reduction in time spent <60 mg/dl (3.3 mmol): from 6.1% at baseline to 3.3% in the last four weeks of the study (p=0.02). By contrast, the FreeStyle Libre group saw no significant change in hypoglycemia (from 7.4% to 8.2%; p=0.5), for a between-group difference of 3.6% (p=0.03). This translated to a clinically significant 52 minutes per day less time in hypoglycemia (<60 mg/dl) with G5 in a very high-risk group.
• The results were highly consistent for time <50 mg/dl and <70 mg/dl, suggesting across-the-board reduction in hypoglycemia with G5.
  
  • Time <50 mg/dl was 3.7% with G5 at baseline, which was more than halved in the last four weeks of the study to 1.4% (p=0.02). Libre users saw no significant change in time <50 mg/dl (4.6% to 5.2%; p=0.49), translating to a between group difference of 2.9% in favor of G5 (p=0.02), or 42 minutes fewer per day under 50 mg/dl.
  • Time <70 mg/dl was 10.1% in the G5 group at baseline, which improved to 7.1% in the last four weeks of the study (p=0.04). By contrast, the FreeStyle Libre group saw no significant change in time <70 mg/dl (from 11.1% to 12.9%; p=0.33), for a between-group difference of 4.9% in favor of G5 (p=0.04), or 71 minutes per day fewer under 70 mg/dl with G5.

TIME-IN-RANGE RESULTS

• Both G5 and FreeStyle Libre improved time-in-range from baseline, but there were no significant differences between the groups.
  
  • Time in 70-180 mg/dl improved a whopping +9.6%-point with G5 (57.2% to 66.8%; p=0.03) and +6.6%-points with Libre (52.2% to 58.8%; p=0.004) from baseline to the last four weeks of the study. This equated to over two more hours per day in range with G5 and 90 minutes more per day in range with Libre. There was no statistical difference between the groups (p=0.5).
  • Time in 70-140 mg/dl was 39.8% with G5 at baseline and improved to 45.4% in the final four weeks of the study (+5.6%; p=0.23). Libre users saw a similar +4.6%-point boost in time in range (35.1% to 39.8%; p=0.06). Put simply, both groups saw roughly one more hour per day in range by the end of the study, and no statistically significant difference between the groups (p=0.86).

HYPERGLYCEMIA RESULTS

• Both G5 and FreeStyle Libre improved hyperglycemia from baseline, but there were no significant differences between the groups.
  
  • Time >180 mg/dl improved a solid -6.5%-point with G5 (32.6% to 26.1%; p=0.11) and -8.4%-points with Libre (36.7% to 28.3%; p=0.01) from baseline to the last four weeks of the study. This equated to 90 minutes less per day in hyperglycemia with G5 and two hours less per day in hyperglycemia with Libre. There was no statistical difference between the groups (p=0.69).
  • Time >270 mg/dl was cut in both study groups: from 10.8% to 5.5% in the G5 group (-5.3%; p=0.007) and 10.7% to 6.1% in the Libre group (-4.6%; p=0.018). This translates to an hour less per day in each group spent at dangerously high blood glucose levels over 270 mg/dl. There was no statistical difference between the groups (p=0.77).

STUDY DESIGN DETAILS

• At baseline, patients had a mean A1c of 7.5%, a mean age of ~50 years, a mean duration of diabetes of ~29 years, and a mean Gold Score of ~4.8. There were no significant differences between the groups. The study enrolled patients that experienced severe hypoglycemia in the last 12 months requiring third party assistance, or those with a gold score of >4. Patients had to have been on MDI for >6 months and type 1 diabetes for greater than three years.
• Notably, the study hypothesized that FreeStyle Libre Libre would have an equivalent impact on hypoglycemia as the Dexcom G5 CGM. The primary outcome was the difference in % time spent <60 mg/dl from baseline to the end of the eight week study. Glucose data from baseline came from a two week run-in period measured via blinded Dexcom G4. Patients were then randomized to use FreeStyle Libre or Dexcom G5 for eight weeks. Glucose outcomes were calculated from the last 30 days of each treatment period. Analysis was intention to treat. A telephone visit occurred at week 2 and a clinic visit occurred at week 4.
• The data today reported results from 32 out of a planned 45 CGM naïve, MDI-using adults (18+ years) with type 1 diabetes. Patients also had the opportunity to continue on Dexcom G5 for 8 weeks following the main study period, though that data was not presented today. A total of 34 patients have been randomized to date, and two are currently ongoing (one in each group), leaving 32 presented today.

Questions and Answers

Dr. Pratik, Choudhary (London, UK): This is a really important debate. In the Dexcom group, did you have predictive alarms set up to avoid hypos? And the algorithms in Dexcom and Libre are very different. Is there a problem in interpreting the hypoglycemia data?

A: We call this hypoglycemia, but it’s not. It’s interstitial fluid measurement, either calibrated in the factory or fingersticks to the blood value. It’s not true hypoglycemia. These are different algorithms. It may be if we had YSI, we would see different results. But these are the devices we have, these are how they work in real time, and these are the outcomes we measure and things we are empowering and asking patients to respond and react to. For the technical aspects, there may be differences we are not pulling out, but this is the real life hypoglycemia point of view, and it’s important.

For predictive alarms, apart from setting it at 3.9 mmol/l (70 mg/dl), we let participants do what they wanted with receiver.

Q: I assume you don’t have A1c for an eight-week study, but do you have mean sugar?

A: I’m not sure it adds a great deal. This is a high risk group of people with severe hypoglycemia. In my clinical practice, if they had an A1c that went up, I wouldn’t be too bothered – that’s an acceptable outcome to reduce the risk of severe hypoglycemia. In both groups A1c fell, and fell significantly, but there were no between group differences.

Q: What about nighttime?

A: I’ll tell you in January. We will look at nocturnal and 24-hour data.

Q: Did you have any questionnaires on fear of hypoglycemia and fear of hypoglycemia at night?

A: No, we don’t have a night-specific one.

Q: Which system is more closely to real blood glucose?

A: Within this study, I don’t know. This is not an accuracy study. Both technologies in the UK are licensed for non-adjunctive use. FreeStyle Libre has no fingersticks. We are unable to make any accuracy comparison. There are other data, not in a hypoglycemia impaired population, that compare accuracy. With G5, it’s about 9%-10% absolute difference from blood glucose.

CGM Reimbursement Update and Cost Effectiveness in 2016

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

Dr. Claudia Graham gave a persuasive and highly encouraging update on CGM reimbursement, arguing that CGMs are highly cost effective (evidence: Germany’s positive reimbursement decision), should absolutely come before pumps (“If you have only $5,000 to spend per year, where would you spend that money?”), and now have some reimbursement (albeit limited) in 11 EU countries (see slide below). She shared a fascinating cost-effectiveness analysis used as an example in Germany, which gave Dexcom CGM a compelling incremental cost-effectiveness ratio (ICER) of 12,533 euros/QALY, far below NICE’s threshold of 20,000-30,000 euros/QALY, and even better than atorvastatin’s 40,069 euros/QALY. This assumed 2.3 fingersticks per day (calibration plus some extra), no receiver cost, and a 50% reduction in severe hypoglycemia events – the latter is of course not a proven fact with CGM, but is “very doable” in Dr. Graham’s view. She highlighted the DiaMonD study results from ADA 2016 and noted the study has been submitted for publication and will hopefully be out soon. In her slide on US reimbursement, Dr. Graham encouragingly said that 25% of US payers have CGM coverage policies for type 2s on insulin, more than we would have guessed. Overall, her talk echoed the one she gave at EASD 2015, strongly urging CGM first, given the high cost of pumps. In the UK, Dexcom CGM costs 3,465 pounds per year (direct pricing from Dexcom), less than half of the 7,196 pounds per year for sensor-augmented pump therapy as listed in the NICE sensor augmented pump assessment in 2015 (slide below). “Let’s give CGM that first start,” she said. Dexcom is focused on driving cost-effectiveness in three areas: (i) appropriate patient selection (e.g., hypoglycemia prone, high A1c patients, pediatrics); (ii) lower-cost technology (e.g., remove cost of receiver, reduce transmitter costs, reduce/eliminate fingersticks); and (iii) simplifying healthcare instructions (e.g., online training).

• “Why has it taken so long for CGM to gain access in Europe?” Dr. Graham highlighted the lengthy assessment processes (70 months on average in Germany, five years in the UK) and individual country-by-country systems. “It takes a long time in Europe to convince the authorities, to have the right data, and to have support of the experts and position statements. But it’s still taken too long for CGM. I’m disappointed about that.” She noted three particular challenges:
  
  • Misperceptions that CGM had to be used with a pump, or follow a pump, or be a last resort. Pumps came to market first, and have been linked to the artificial pancreas quest for a long time. There has also been a belief that MDIs are not “tech savvy” (“it’s too much for my patients”) – DIaMonD and COMISAIR (see below) disprove that. “We lost sight that CGM and pumps can be used alone. Most of Dexcom patients are trained online in how to use the device in under an hour, and they get it. It’s surprisingly intuitive on how to use this.”
  • Rapid product iterations make it difficult to assess effectiveness. Outdated publications are used in meta-analyses for health technology assessment, and these include much older technologies that are less effective.
  • Outcome publications and models have primarily been done on sensor-augmented pumps. These are “confusing and costly,” she noted.

CGM and Multiple Daily Injection Patients: A1c Outcomes and Clinical Benefit Demonstrated by the COMISAIR Study

Martin Prazny, MD (Charles University, Prague, Czech Republic)

Dr. Prazny covered the newly published 52-week (!) COMISAIR trial (n=65; Soupal et al., DT&T 2016), offering a highly positive takeaway for Dexcom: adding CGM to MDI drove a similar A1c reduction as adding CGM and a pump (-1.2%; baseline: 8.3%). Further, adding a pump alone (no CGM) only drove a 0.5% reduction in A1c (baseline: 8.4%). Though the study was very small (n=18 on MDI + SMBG, n=12 on MDI+CGM, n=20 on pump + SMBG, and n=15 on pump + CGM), it was great to see the long length and four-group design. We first heard about this study in the AADE 2016 exhibit hall, and the results confirm the preliminary findings from the DIaMonD study (presented at ADA) – CGM offers significant benefits to MDIs.

Corporate Symposium: Redefining Diabetes Management with Metrics Beyond HbA1c (Sponsored by Abbott)

Data from Abbott’s IMPACT study (presented in poster form at ADA) was published in the prominent UK medical journal The Lancet today, alongside a very balanced (but mostly positive) accompanying Comment from Cambridge’s Dr. Roman Hovorka and colleagues. This publication was mentioned only briefly in today’s symposium, and represents a major victory for the company and hopefully a positive for eventual reimbursement. There were not major new details in the paper vs. the ADA poster, but getting the dramatic 38% hypoglycemia reduction (~74 minutes fewer per day) published in such a prestigious journal helps build the evidence base and reduces the noise factor around “no alarms.” We also learned today that a new integration agreement between Abbott and mySugr will allow Libre users to automatically and passively upload data to their mySugr accounts via the LibreLink Android app. This is outstanding news for Abbott to begin leveraging data from Libre in the widely loved mySugr app, which also has a partnership with Roche’s Accu-Chek Connect BGM. The Abbott team also told us about a new partnership with Social Diabetes, a popular Android app in the EU that enables users to log their food and insulin and get dosing recommendations from the app. We have not ever heard of this app, but it has a notable 50,000+ downloads on Android and very positive reviews. Abbott’s actual symposium did not share new data or pipeline updates, but the session – “FLASH GLUCOSE MONITORING: Redefining Diabetes Management with Metrics beyond HbA1c” – featured an established slate of experts (Drs. Irl Hirsch, Thomas Danne) who spoke about recent clinical trials of the Libre and reasons why A1c is a flawed, or at least, incomplete outcome metric.

• The Lancet publication is largely comprised of the information we covered at ADA (Poster 868-P). The accompanying commentary from Dr. Hovorka et al. is balanced but mostly positive, noting that the Libre’s lack of alarms “has surprisingly little effect on user acceptability” and that the device could alleviate many of the barriers to wider adoption of CGM, such as the need for calibrations, alarm fatigue, frequent sensor changes, and cost. However, they do caution that “studies in a more generalizable clinical population, in which aspects of behavior modification induced by use of this technology can be ascertained, are needed to provide further guidance to healthcare providers and funders, as well as comparisons with emerging automated insulin-delivery systems.” 
  
  • The Commentary notes that the convenience of FGM trumps that of SMBG (after 90 days, Libre users scanned an average of 15 times/day, compared with the 6 fingersticks/day of the SMBG group), which fosters rapid behavioral adaptation. The downside is that this requires “unrelenting round-the-clock attention to diabetes self-care,” which many patients may not commit to. The study did not include details on how the patients altered their management in response to the additional data (and Dr. Bolinder indicated during Q&A that it is unknown at this point). As a reminder, the study population was intentionally composed of well-controlled patients with lots of hypoglycemia at baseline (mean A1c=6.7%), though the commentary wonders how well the improvements would translate to other groups who may have different interactions with the technology. This is the classic commentary about any study, and overall, we think Abbott should be highly commended for running IMPACT in type 1s and REPLACE in poorly controlled type 2s.
• Abbott’s integration with mySugr is a win for both companies and particularly for Libre users, who can now view glucose levels alongside the rest of their food and medication data in the motivating mySugr app. In June, Abbott struck a similar data integration agreement with Diasend’s mobile app. It is excellent to see Abbott making it easier for patients to view and extract insights from their data, especially in a mobile app. Abbott has lagged behind Medtronic and Dexcom on the data front and we’re glad to see it now making moves to maximize the value of its sensor.
• Highlights from the symposium included: Dr. Thomas Danne (Auf der Bult Hospital for Children and Adolescents, Hannover, Germany) on FGM in the pediatric population, Dr. Jan Bolinder (Karolinsak University Hospital Huddinge, Stockholm, Sweden) on IMPACT results, and Dr. Irl Hirsch (University of Washington Medical School, Seattle, WA) on the limitations of A1c.
  
  • Dr. Danne championed the use of FGM in pediatric populations. Children are notorious for their unpredictable lifestyles, which give rise to higher glycemic variability and risk of hypoglycemia. Due to the legacy effect (or “metabolic memory”), it is crucial that both elements of poor glucose control are minimized as soon and as much as possible. Dr. Danne asserted that FGM can lead to better control and higher quality of life of the child and family – reducing painful fingersticks, allowing parents to check a child’s glucose at night without disturbing him/her, etc. Libre is currently approved for children as young as four years in Europe; we saw the accuracy data in an oral at ATTD 2016).
  • Dr. Bolinder’s talk focused on the IMPACT trial (presented in poster form at ADA), which found that, relative to SMBG, well-controlled (baseline A1c: 6.7%) type 1 patients using FreeStyle Libre spent ~74 minutes fewer per day <70 mg/dl (a 38% reduction; p<0.001). There was not a significant difference in A1c between the groups by study end, though both saw a non-significant 0.15% increase from baseline to six months (6.7% to 6.9%). The quality of the A1c was unquestionably better in the Libre group. In addition, quality of life metrics were almost entirely in favor of Libre. See our detailed coverage here and the Lancet publication here.
  • Dr. Irl Hirsch on how we manage type 2 diabetes: “Wrong!” Dr. Hirsch walked attendees through a litany of primary literature to prove his point: A1c does not reliably correlate with an average glucose level on an individual basis, ESAs (erythropoietin stimulating agents) cause a decrease in A1c, iron causes an increase in A1c, aortic disease causes reduced red blood cell survival and subsequent decrease in A1c, the list goes on. Yet the standard of care is to escalate or deescalate a type 2 patients’ therapy based on an A1c, which may or may not accurately reflect the patient’s glycemia, measured every 3-4 months. As only Dr. Hirsch could put it, “It is ok to be glucocentric while treating glucose.” Well said!

Selected Questions and Answers

Q: What do patients do on Libre to improve metabolic control and achieve benefits?

Dr. Bolinder: We don’t know exactly. My personal guess is that they make proactive decisions rather than retrospective analyses. It’s not the one and only value that you measure that’s important, it’s the trend. If you look at where your glucose is heading within the next hour or two, you can make proactive changes. I think this because of the SWITCH study, where patients used a sensor for a period and were switched over to SMBG again. All improvements from using the sensor were damaged when they switched back SMBG.

Q: Do patients/parents need education to use FGM properly because of all  of the data generated?

Dr. Danne: Yes, I think that’s imperative. The first prescription should be made by a diabetes team experienced with the system, one that understands the trends and arrows. It’s not a BGM, but it measures interstitial glucose.

Q: How are reductions in nocturnal hypoglycemia explained without alarms?

Dr. Ramzi Ajjan (Leeds University, UK): Because we review the patterns of glucose. That way we can find hypoglycemias and adjust accordingly. You don’t necessarily need the alarm.

Dr. Bolinder: I’m not an entirely AGP fan, because I think it’s the proactive manner that you make the changes. The highest frequency of scanning [in IMPACT] occurred in the evening before going to bed, then patients can make the adjustments to minimize risk of hypoglycemia during the night. In addition, please remember that these patients also monitored on average 1-2 times every night. That is something that we perhaps had not been aware of. Night contains things that we are not quite aware of that the patients do.

Q: Why was IMPACT restricted to type 1 diabetes with A1c higher than 7.5%?

Dr. Bolinder: Please remember that the aim of the study was to evaluate effectiveness in preventing hypoglycemia. So we selected patients in which you’d expect increased risk of hypoglycemia. We clearly stated that future studies are needed to evaluate the device in other patient groups. The first that comes to mind is patients that don’t achieve targets.

Improving Diabetes Management with Advances in Health Technology (Sponsored by Roche)

Characteristics of the new Accu-Chek Insight CGM system

Günther Schmelzeisen-Redeker (Roche, Mannheim, Germany)

Roche’s Günther Schmelzeisen-Redeker presented the most details ever on its Accu-Chek Insight CGM, which is currently in CE Mark approval trials and on track to launch by the end of the year – we learned today that the initial launch will be limited to “specialized diabetes centers” in the Netherlands, Norway, Denmark, and Sweden. Preliminary pivotal data in 36 patients wearing two sensors each (abdomen) suggests an overall mean ARD of 10.5% over seven days with two fingerstick calibrations per day (n=6,403 paired fingerstick BGM-CGM points). The accuracy was similar across the different glucose ranges (see below), though we’d note very few (3.7%) points in the <70 mg/dl range. Consistent with other CGMs, accuracy was slightly worse on day one (13%) and improved by day seven (9%). The Insight’s accuracy was similar during induced glycemic excursions (10.9%), and a section of the talk highlighted the short five-minute CGM-blood glucose time delay – we assume that is equal to or within a few minutes of other systems and are not sure if it is a clinically meaningful advantage. The sensor has a two-hour warm-up time and seven-day wear.

We also saw the first pictures of the commercial system today (see below), which has a round adhesive, a fairly large on-body transmitter (roughly two thumb drives in size; see second picture for a sense of thickness), and a slightly intimidating syringe-like insertion device. The Bluetooth-enabled transmitter sends data directly to an Android smartphone app, and it appeared that no receiver will be included – this is great to see at launch and key for competition. The app has the standard Roche utilitarian design – nothing flashy and roughly similar-looking to the Accu-Chek Connect BGM app.

• How does the Insight CGM stack up against the competition? This accuracy – if observed in real life –puts Roche right in the ballpark of competitor systems: Abbott’s FreeStyle Libre (11.8% factory cal), Dexcom’s G5 (9.0%, two cals per day), Medtronic’s Enlite 3 (10.5%, two cals per day), and Senseonics’ Eversense (8.8%-11.6%, two cals per day). Of course, it is always difficult to understand how systems compare based on individual trials. Competitively speaking, just as important for the Insight CGM are the cost, insertion process, calibration hassle, and on-body form factor, which appear to lag behind competitors based on what we saw today. It’s excellent to see Bluetooth compatibility and Android at launch.
  
  • Still, we’re glad to see a major player like Roche investing significantly CGM, which should help build the highly underpenetrated EU market. Given the number of patients in need, we believe many companies can be successful and help make CGM standard of care. Of course, Roche also has diversified its portfolio with the Senseonics distribution partnership, giving it two different CGM products to market.

*SMBG >80 mg/dl – relative differences; SMBG <80 mg/dl – absolute differences

• Study design details: CGM accuracy was compared to BGM (presumably Roche’s own). The study is ongoing, but data presented today was from 36 patients (n=27 type 1s, n=9 type 2s) wearing two CGM systems simultaneously for seven days. On two of the days patients had induced glucose variations, which were not further detailed (e.g., were they on days 4 and 7?). The frequency of SMBG measurement was one per hour during the daytime, and increased to one every 15 minutes during the induced excursions
  
  • Study Design Limitations: The study (i) is smaller compared to competitor studies (e.g., Enlite 3’s pivotal had a remarkable 23,709 total paired CGM-YSI points); (ii) had very few points (3.7%) in hypoglycemia (e.g., Dexcom’s G4 and Software 505 pivotal had 14%-15% of points <80 mg/dl); (iii) included 9 of the 36 patients as type 2s (we’re not sure if they were all on insulin); and (iv) only include two in-clinic days (relative to the typical three in competitor studies of seven-day CGMs). We note the trial did not use YSI, which is a limitation in one sense (doesn’t show the true accuracy of the system), but a strength in another (accuracy vs. BGM is more real-world from a patient perspective).

Questions and Answers

Dr. Boris Kovatchev (UVA, Charlottesville, VA): Do you have data on the frequency of large sensor deviations beyond 20%?

A: I don’t have it with me. We’re in the middle of data evaluation.

Dr. Kovatchev: That would be good to add.

Q: What about newer sensors with no need for calibration?

A: We need calibration – two per day over seven days. We need this for our performance data.

Non Commercial Symposium: New Technologies and Therapies in Diabetes Improving Patient Quality of Life (Sponsored by IDF Europe)

Monitoring Technologies and Apps

Mr. Jaivir Pall (National Health Services, London, UK)

NHS’s Mr. Jaivir Pall shared his takeaways from interviewing people with diabetes about current technology and their biggest needs. He said patients predominantly value three things in devices and apps: (i) support networks; (ii) convenience; and (iii) data access. Moving forward, he said, patients should be more involved in the development of the tools that are designed for them – nothing new there, but we hope to see this implemented far more rigorously and far earlier in the diabetes device design process. Mr. Pall said that only 1.2% of people with diabetes use apps, meaning there is a disconnect between what’s available and what’s desired – we’d note that data is from a 2014 report (Research2Guidance) and has likely improved since that time (e.g., connected CGM, connected BGM, etc.). Still, the point is well taken that apps are not nearly useful enough for a broad population of people with diabetes (not just those interested in data or already doing pretty well). Much of Mr. Pall’s talk shared patients’ most common sentiments about apps and devices, which we found valuable for guiding future developments in the field:

• “I want to connect to people/I want a support network.” Mr. Pall suggested that clinicians can facilitate the formation of such networks, a lot of times, by stepping back and allowing patients to solve problems. His number one advice to the clinician in this case is to “be brave.” Patients frequently mentioned that they would like Facebook and Twitter to help with network building.
  
  • We do wonder what percentage of people with diabetes globally are involved in the diabetes online community – Is it 15 million patients globally? 50 million? We have no idea, but it would be an interesting analysis to see how penetrated the diabetes online community is.
• “Don’t give me something that tries to meet every requirement.” Many patients become frustrated when they have to set up new accounts on new platforms despite already having software that supports the same function. Developers should strive to integrate their function with an existing one, if possible. Mr. Pall gave the example of Skype: if an app offers video chatting, funnel it through Skype so that the user doesn’t have to sign up for an additional account on a new platform. We do agree with the account overload problem but aren’t sure this is going to improve – most legitimate apps require creating an account to store data in the cloud and presumably to secure it.
• “Give me the ability to share my diabetes metrics with whomever I want, but make it easy.” Patients want access to free and actionable information. Mr. Pall described a Facebook group that is just open to patients and automatically populated with data from their Diasend accounts. He claimed that such group-monitoring and management removes stigma, helping people feel comfortable in the community. In fact, in a recent study, he said that patients opting to share their data saw a drop in A1c relative to non-sharers. Mr. Pall couldn’t comment on the size of the effect because it is not yet published. This is not surprising to us – human accountability works wonders for behavior change. The challenge is probably getting the majority of people with diabetes to do this, particularly those who feel guilty about not doing well.

Questions and Answers

Q: It’s excellent to see how patients can, and want to, self-manage. Do you have experience in how healthcare providers think of the sharing of data and networks?

Mr. Pall: Depends on the particular network. At my clinic, providers are very interested in data. They want it all, qualitative and quantitative. It comes down to the person that’s treating you – do they value it? What do they do with it after? I won’t send my information if it’s not being used. Many patients are disengaged because they send the data to providers and it’s not used. There is no parity and the power is all with the clinician. In my mind, the best clinicians respond to all data, whether it’s useful or not. It’s as simple as saying “I saw your results, are you ok? It looks like you go low at such and such time every day. Have you tried this?” It’s a balance of not overworking and burning out, but also having valuable communication with patients. We’re getting into a study where patients are opting to share CGM data with us, and we’re seeing a drop in A1c, improved patient satisfaction, and improved ability to concentrate at work relative to an SMBG cohort. These outcomes are beneficial at the local level and could lead to reimbursement.

Q: The patients you’re talking about are mainly younger, it seems. What age are they? Do you see the way you are collaborating with them as a way to collaborate with older groups as well?

Mr. Pall: On Facebook, they are all of our patients, both young and old. When trying to design a service, we have to get a full spectrum of opinions. On type 2 diabetes, we have a relationship with Diabetes UK which allows us to tap into type 2 diabetes patients across the country. Whenever we devise a project, we go there first to make sure we’re addressing the correct problems. Interestingly, functionality requests across different conditions were often the same – MS patients wanted connectivity as well. Even older patients are asking to interact via Facebook, which surprised me.

Q: How many people are engaging in your forums?

Mr. Pall: Depends on the project. We have 900 on Facebook. Engaged in that population? 10% post daily, 50% would say they interact on those forums. When you look at diabetes overall, only about 1.2% of patients use apps, and that number is expected to climb to 7.8% by 2018 (Editor’s Note: this is a 2014 report from Research2Guidance). Are we designing for the wrong market? We need to leverage what’s already there, not develop new apps. [Editor’s Note: We would debate this point – if what is out there is not being used, it could also be that the apps are not useful enough.]

Q: How easy is it for clients to get in touch with providers through Facebook or Skype? Is it just available at certain hours? Or 24 hours?

Mr. Pall: None of our services are 24 hours. I answer questions whenever I can. In terms of Skype, we’re exploring the way we use it. A lot of UK talk is about sharing information over an unsecured platform, but the risk is fine for me. Patients actually don’t want 24-hour service with Skype. What they really want is to be able to schedule a set time to talk.

Q: It’s interesting that you saw a drop in A1c from those who shared data – can you quantify that drop?

Mr. Pall: The drop is not published so I can’t talk too much. We just asked patients if they wanted to share sensor readouts – no goal for reimbursement, just looking for trends and algorithms across patients.

Q: Who is taking responsibility at the end of the day? What if, through advice coming from networks, through Skype, a patient dies?

Mr. Pall: At a personal level, I’d say I’m responsible. On a systematic level – that is a very good question. A quick anecdote: A woman has been giving advice on Twitter and she is now becoming a certified nurse – does she lose her indemnity? The alternative to this sort of networking is that people don’t get access to the information, so they are actually receiving more harm through inaction. We define inaction as an action. We do give them permission to go to social media, and empower them to validate the information themselves.

Comment: What if the clinician can see hypoglycemia if he’s on duty. Is he obligated to intervene?

Mr. Pall: In the UK, if a doctor drives by a car accident and doesn’t stop, he can be brought before the medical council. So yes, I’d imagine it’d be the same in this case. But I don’t think regulatory should be a barrier to going out and trying all of this.

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