Advanced Technologies and Treatments for Diabetes (ATTD 2016)

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

Executive Highlights

Friday’s sessions have wrapped up at ATTD 2016 and what a day it was! Abbott shared its VERY nuanced REPLACE data (see below for data and plenty of analysis), Medtronic offered the most details ever on its fifth-generation sensor, Cellnovo announced a partnership with TypeZero, and UVA’s Dr. Boris Kovatchev shared six-month results from the five-month extension phase of JDRF’s multicenter 24/7 closed-loop trial testing the DiAs system. Check out our Top Ten Highlights and honorable mentions – especially on biosimilars - below, followed by selected full write-ups.

Top Ten Highlights

1. Dr. Thomas Haak (Diabetes Center Mergentheim, Germany) presented the highly anticipated results from Abbott’s six-month REPLACE study, comparing FreeStyle Libre to SMBG in type 2s with a high baseline A1c (8.8%). The study disappointingly missed its primary endpoint, demonstrating similar 0.3% A1c reductions in both groups. However, A1c significantly improved with FreeStyle Libre in users <65 years old (-0.5% vs. -0.2%), and FreeStyle Libre prompted highly, highly significant reductions in hypoglycemia (particularly <55 mg/dl) – great news from a reimbursement perspective in our view. The results serve as a reminder of how much hypoglycemia impacts type 2, how challenging A1c is as a primary endpoint for interventions that reduce hypoglycemia, and the nuances of trial design for diabetes technology. We expect these results to increase the “noise” level on the competitive front overall for Abbott and think they will provide a bit of breather (real or perceived – patients love this device) for Dexcom and Medtronic in the near term. Though overall, these results aren’t what we expected, Abbott will get very good marks for reducing hypoglycemia at night – and that reduces the noise factor on the absence of alarms.  

2. A Medtronic poster offered the most details ever on its fifth-generation sensor (i.e,, Enlite 4), featuring one calibration per day, 10-day wear, and a strong overall MARD of 10.9% vs. the Bayer Contour Next Link meter (n=55 sensors, 5,709 evaluation points). Accuracy data from a pre-pivotal study of the fourth-generation sensor (Enlite 3) also demonstrated an improved MARD: 11% vs. YSI based on two fingerstick calibrations per day.

3. Dr. Roman Hovorka shared an updated overview of Cambridge’s upcoming pre-pivotal studies (three longer-term trials using Medtronic devices!) and hinted at rethinking the team’s clinical study design around easier comparators (vs. their typical SAP): “I fully agree that one needs to think about study design and reimbursement. We are tempted to rethink what we have done in the past.” Dr. Hovorka tends to be a bit understated (in the best of ways) so this is an especially big deal.

4. Abbott also shared data from its 89-patient, 14-day EU pivotal trial of FreeStyle Libre in pediatrics, which demonstrated a MARD of 13.9% vs. capillary fingersticks. It’s a slight downtick from the adult data (MARD: 11.4%).

5. In an unexpected announcement that we reported moments after it was announced, earlier today, Cellnovo announced a partnership with TypeZero to use its patch pump system in the upcoming NIH-funded International Diabetes Closed Loop Trial (n=240), starting in 2H16. The trial will use multiple pump brands (2+), and others will be announced soon. Notably, the goal is for patients in this trial and other upcoming studies to have the option of choosing which pump they want to use – we love that! We wonder what “+” means?

6. UVA’s Dr. Boris Kovatchev shared six-month results from the five-month extension phase of JDRF’s multicenter 24/7 closed-loop trial testing the DiAs system (we saw one-month results at ADA 2015). The 14 participants in the five-month extension phase had sustained and dramatic reductions in hypoglycemia (~90% overnight!).

7. Dr. Hans DeVries presented full 90-day data (n=71) from the EU pivotal trial of Senseonics Eversense implantable CGM sensor, on-body transmitter, and mobile app. Overall MARD vs. YSI at ten in-clinic visits was 11.5%, nearly identical to the interim results (n=44) shown at DTM last fall.

8. Ahmad Haidar (Ecole Polytechnique de Montreal) presented a small (n=23) 60-hour randomized trial comparing sensor-augmented pump therapy vs. insulin-only closed-loop vs. insulin+glucagon closed-loop. Glucagon showed a small hypoglycemia benefit and no advantage on time-in-range. This study design and generalizability were impossible to assess. We look forward to seeing the Bionic Pancreas team’s insulin-only data tomorrow.

9. Roche’s Dr. Matthias Axel Schweitzer stated that the artificial pancreas is “clearly on our agenda.”

10. Dr. Philip Home (Newcastle University, Newcastle upon Tyne, UK) provided a comprehensive take on what prescribers should know about biosimilar insulins, focusing mainly on the limitations of the available data. In the same session, Mr. Joseph Saldanha (Julphar Diabetes, Ras al Khaimah, United Arab Emirates) spoke about the potential for biosimilars to increase access to insulin in the developing world.

Honorable Mentions

  • The esteemed Dr. John Pickup (King’s College, London, UK) discussed several reasons for why much of the “best evidence” in the diabetes technology field is misrepresented or distorted, cautioning that while meta-analysis of randomized clinical trials is the gold standard, “all that is gold does not glitter”.
  • Dr. Primoz Kotnik (University of Ljubljana, Slovenia) presented new data, demonstrating that GI Dynamics’ EndoBarrier (endoscopic duodenal jejunal bypass liner) is a feasible therapeutic option for adolescents with severe obesity.
  • Dr. Jane Seley (Weill Cornell Medical Center, New York, NY) shared the design and preliminary data of a new ongoing trial that aims to examine strategies on preventing hospital readmission in high-risk diabetes patients.
  • Day #3 of the meeting featured the celebrated ATTD Yearbook sessions – check out the publication online here. We love this. Cheers and a big shout-out to Dr. Moshe Phillip for how ambitious he makes ATTD each year. And by the way – 2017? It’s in Paris, in mid-February. 

Top Ten Highlights

1. Dr. Thomas Haak (Diabetes Center Mergentheim, Germany) presented the highly anticipated results from Abbott’s six-month REPLACE study, comparing FreeStyle Libre to SMBG in type 2s with a high baseline A1c (8.8%). The study disappointingly missed its primary endpoint, demonstrating similar 0.3% A1c reductions in both groups. However, A1c significantly improved with FreeStyle Libre in users <65 years old (-0.5% vs. -0.2%), and FreeStyle Libre prompted highly significant reductions in hypoglycemia (particularly <55 mg/dl). The results serve as a reminder of how much hypoglycemia impacts type 2, how challenging A1c is to use as a sole primary endpoint, especially for interventions that reduce hypoglycemia, and the nuances of trial design for diabetes technology. This is going to “up” the noise for Abbott, but we’re still highly confident – this kind of product enthusiasm across the board for people with diabetes (not just intensively managed patients) hasn’t been seen in years.   

  • The most compelling takeaway from this trial was the hypoglycemia data (measured via masked Libre Pro in the SMBG arm), which improved markedly with FreeStyle Libre overall, overnight, and particularly for dangerous hypoglycemia. Relative to the control group, patients using FreeStyle Libre spent ~30 minutes fewer per day <70 mg/dl (p<0.001), ~13 minutes fewer per day <55 mg/dl (p=0.001), and ~8.5 minutes fewer per day <45 mg/dl (p=0.001). For the FreeStyle Libre group, these reductions equated to major 55%, 68%, and 75% reductions in those respective zones from baseline to six months (based on the limited data given, it was not possible to calculate these percentages for the control group). All measures of nocturnal hypoglycemia were also significantly lower with FreeStyle Libre, countering the criticism that the device’s lack of alarms poses a nighttime danger (presumably, the retrospective glucose data helped identify nocturnal hypoglycemia). There were no device-related serious adverse events and nine instances of minimal adverse events (e.g., infection, allergy) from six subjects that we term not a big deal at all.
  • Ultimately, we had high expectations coming in to REPLACE – it seemed like slam dunk to improve A1c in insulin-using type 2s (baseline A1c: 8.8%) testing ~four times per day (although, we know that is far more often – 400% more often – than the average type 2). While we are disappointed that the trial missed its primary endpoint in all patients, we were pleased to see it showed profound and meaningful reductions in hypoglycemia – particularly the ~75% reduction in time spent <45 mg/dl – and patients <65 years did see a 0.3% benefit on A1c at the same time. A1c is the most devilish of outcomes for diabetes technology, as devices typically profoundly reduce hypoglycemia, often at the expense of raising average glucose.
    • Most importantly, the study design did not use a per-protocol mandatory insulin adjustment (i.e., not treat-to-target), which would have unquestionably improved the magnitude of A1c benefit. Providers and patients could decide what to do with the data at their regular visits, a deliberate decision to ensure a real-world trial. We wonder if providers were drawn to the red traffic light on AGP that identifies hypoglycemia – particularly in the older type 2 patients! – and backed off therapy accordingly. Of course, it is also much easier to fix hypoglycemia (reduce insulin) than to safely bring mean glucose down (“Is it correction or food bolus? Or is it basal? Or is it...”). We believe this also opens the door for the role of the community health worker who can help patients identify how to interpret data and change therapy, on an ongoing basis (at least for type 2 – diabetes is progressive).
    • We also wonder about the study population, as these were patients far from A1c goal and already testing four times per day. Within that framework, the full results have to be interpreted in the proper context – not a smashing success, but certainly not a failure either (e.g., this trial could have met its primary endpoint with a hypothetical 0.5% reduction in A1c for all patients, but increased hypoglycemia at the same– that would have been a failure in the context of “success”).
  • All things considered, we salute Abbott for conducting this ambitious, long-term outcomes study of FreeStyle Libre. The ultimate mark of any technology is whether patients will buy it, and with FreeStyle Libre, they are paying out of pocket and demanding it faster than Abbott can make sensors. Of course, reimbursement will open access for far more patients, and we hope this study and subsequent studies make a strong case that more frequent, actionable glucose data is beneficial. Last, it will be interesting to see how the type 1 data from IMPACT (to be shown at ADA 2016) will complement these results. We can only assume it will show similar or larger reductions in hypoglycemia, since patients in that trial have an A1c <7.5%.
  • Separately, we’ve learned that Nightscout users in Italy have added remote monitoring of FreeStyle Libre via an Android app, Glimp. Glimp has existed for some time as an unauthorized app for reading the FreeStyle Libre sensor (we wrote about it last October), though Nightscout users in Italy are now using it to remotely monitor patients on FreeStyle Libre. A major win for parents, and the app can also send readings to HCPs. The instruction manual is posted within the Nightscout Italy Facebook group (which requires permission to enter). Even with Abbott’s LibreLink Android app for reading the FreeStyle Libre sensor (limited launch in Sweden in November), it does not enable remote monitoring to our knowledge. We hope that the addition of a pediatric label will move Abbott to add remote monitoring to LibreLink in the future – parents love Dexcom Share and MiniMed Connect. It’s worth noting that Glimp is an unauthorized app with fairly good reviews (4.3/5.0, 151+ ratings), though it does reveal the downsides of patients hacking into devices themselves – the FreeStyle Libre reader makes some corrections to the raw sensor data, so Glimp does not display the exact same value as the reader (“it’s close,” notes the reviews). Of course, parents will only use it if it works, and it seems like it does based on the reviews.

2. A Medtronic poster offered the most details ever on its fifth-generation sensor (i.e,, Enlite 4), featuring one calibration per day, 10-day wear, and a strong overall MARD of 10.9% vs. the Bayer Contour Next Link meter (n=55 sensors, 5,709 evaluation points). We had not ever known this would be 10-day wear or one calibration per day, though that would exactly match Dexcom’s plans for G6. The fifth-gen CGM (!) includes a 90-minute warm up, redundancy via two sensor flexes, a proprietary fusion algorithm to combine the two outputs, and intelligent diagnostics to assist with fault detection and sensor health. The poster showed the standalone Guardian Connect setup we mentioned on day #1, featuring a Bluetooth-enabled CGM transmitter and a smartphone app display (no receiver or pump); as a reminder, the Enlite 3 version of this system is under CE Mark review in Europe and expected to launch by April 2017 in the US. This accuracy study included 25 participants with diabetes who wore up to four sensors on the abdomen or arm for 10 days. At three in-clinic session (Days 1, 7, 10), meal challenges were administered and blood glucose measurements were recorded every 15 minutes for three to four hours with the Bayer Contour Next Link meter. Participants were also asked to take 8-10 blood glucose measurements daily when at home. Overall MARD was 10.9% (11.7% on the abdomen and 9.9% on the arm), including a day #1 MARD of 12.6%. Roughly 45% of sensors had a MARD <10%, with most of the remaining sensors between 10% and 15%. Mean absolute difference (MAD) in hypoglycemia (<70 mg/dl) was 12 mg/dl, and 86% of overall points were within 20 mg/dl or 20%. Sensors were removed from analysis early due to adhesiveness or battery failures – the percentage was not specified, and both are critical question marks for Medtronic’s clamshell transmitter design (larger on the body and less secure than Dexcom and Abbott sensors). While this is still a feasibility study, this sensor shows a marked improvement from the original Enlite, helping Medtronic catch up to Abbott and Dexcom’s more accurate sensors.

  • Medtronic also displayed accuracy data from a pre-pivotal study of its fourth-generation sensor (Enlite 3), to be used with the MiniMed 670G or the Guardian Connect mobile app. It demonstrated an MARD of 11% vs. YSI based on two fingerstick calibrations per day (days 1, 3, 7 visits; YSI values recorded every 15 minutes for 12 hours; 4,805 paired points). The accuracy was much stronger on the arm (8.7%) vs. the abdomen (11.9%-12.6%) – Abbott clearly figured that out too with FreeStyle Libre. The seven-day wear Enlite 3 sensor has an improved algorithm with intelligent diagnostics that determine if it is safe to enter closed loop. The algorithm will also request a calibration when the system detects the overall performance can be improved, and data is not displayed when it detects poor sensor performance. It’s great to see these safeguards in place. In the oral presentation discussing the in-clinic US pivotal study of the MiniMed 640G, Enlite 3 demonstrated an MARD of 12.6% on the first day; further performance was not mentioned, though that was consistent with the pre-pivotal’s day #1 MARD (12.9%). In that study, the 640G was not quite as strong in avoiding hypoglycemia (60% of low limit events avoided vs. ~75-80% in other studies), though the hypoglycemia induction protocol made the study far more challenging and less real-world. Every EU patient we’ve talked to on the 640G loves it – one told us he’s had 0% time in hypoglycemia and 80% time-in-range over the past several months. As a reminder, we learned at JPM that Medtronic is likely to skip the 640G/Enlite 3 in favor of launching the 670G/Enlite 3 in the US first. The pivotal study for the latter will wrap up 25 days.

3. Dr. Roman Hovorka shared an updated overview of Cambridge’s upcoming pre-pivotal studies (three longer-term trials using Medtronic devices!) and hinted at rethinking the team’s clinical study design around easier comparators (vs. their typical sensor augmented pump – we’d love to see different comparators, since the SAP is so far from standard of care). On the pre-pivotal front, Dr. Hovorka laid out a slew of ambitious trials slated for the near future – some of which we had heard of before and some of which we hadn’t: (i) a 3-month, 24/7 closed-loop RCT in adults and children piloting the team’s new setup (Medtronic’s MiniMed 640G/Enlite 3 + an Android phone running Cambridge’s MPC); (ii) a 12-month RCT in children and adolescents [n=130; 6-18 years; 24/7 closed loop (n=65) vs. sensor-augmented pump (n=65); primary endpoint = A1c]; and (iii) a 24-month RCT in newly diagnosed children/adolescents with type 1 diabetes. What is unclear is whether the Cambridge group will use data from these studies to support a regulatory submission. We don’t believe so judging from previous commentary upon receiving an NIH grant– “facilitate pivotal studies” – suggesting that a pivotal study to support PMA application is still in the works. This group, though, continues to put together multicenter trials that are very scientifically rigorous and we salute their work in really blazing a trail of ambitious, at-home, free-living closed-loop studies. Dr. Hovorka did not comment on the team’s commercialization plans.

  • “I fully agree that one needs to think about study design and reimbursement. We are tempted to rethink what we have done in the past.” Dr. Hovorka spoke to the importance of thinking critically about comparator groups in closed-loop studies, noting that the ultimate goal is to allow the maximum the number of people to benefit from the technology. The Cambridge team has ambitiously and unfailingly used sensor-augmented pump therapy as the comparator (harder than MDI, the best alternative therapy available), though the trade-off, of course, has been diminished contrast between the intervention and control groups. From an academic perspective, we can certainly see how SAP as the control group makes the most sense since studies are scientifically testing whether the addition of a closed-loop algorithm to SAP makes a difference. At the same time, practical considerations would seem to advocate for a more real-world approach – after all, to get the penetration everyone is hoping for, AP studies need to expand beyond sensor-augmented pump users alone and enroll a broad spectrum of type 1s, including those on MDI/SMBG and pumps alone. It’s notable to hear Dr. Hovorka reconsidering this design and we heartily salute his opened-minded, patient-centric approach. We hope pivotal studies and outcomes studies enroll a broad spectrum of type 1s. Automated insulin delivery should appeal to more than just current pump+CGM users – the market will be too small otherwise.  

4. Abbott also shared data from its 89-patient, 14-day EU pivotal trial of FreeStyle Libre in pediatrics, which demonstrated a MARD of 13.9% vs. capillary fingersticks. The study had 5,493 paired sensor-fingerstick points, and the data were impressively consistent across the board as 84% of points were in Zone A of the Consensus Error Grid and 16% in Zone B, a testament to the device’s accuracy as patients would experience it (i.e., relative to fingersticks). The data represent a definite downtick on what has been achieved in the EU pivotal trial, where Libre demonstrated an overall MARD of 11.4% vs. FreeStyle Precision BGM (n=13,195 paired points), though given the vagaries of fingersticks in pediatric patients, we would argue that the overall accuracy is highly encouraging. MARD was not broken down by glucose range, so accuracy in hypoglycemia is an unanswered question. Still, patients in the real world clearly trust the device enough to test rarely so we have no major concerns about this data. We do note that it is noticeably higher than what Dexcom has achieved with Software 505 (G4AP), which showed a MARD of 9% in adults and 10% in pediatrics. On the safety front, no severe adverse events were reported and all patients were able to see the trial to conclusion. For more details from the trial – including a look at stellar patient/caregiver and provider questionnaire results – please see our full coverage below. Most notably, 95% of girls and 98% of boys said they strongly agreed with the statement, “I would recommend [FreeStyle Libre] to someone else with diabetes.”

  • As a reminder, Abbott announced yesterday that it has received CE Mark for the pediatric indication of FreeStyle Libre. We assume this data was used in the submission.

5. In an unexpected announcement, Cellnovo announced a partnership with TypeZero to use its patch pump system in the upcoming NIH-funded International Diabetes Closed Loop Trial (n=240), starting in 2H16. The trial will use multiple pump brands (2+), and others will be announced soon. The goal is for patients in this trial and other upcoming studies to have the option of choosing which pump they want to use – we love that. We’d assume Tandem and Insulet are the most likely other pumps to be chosen for these studies. As a reminder, this trial will test a commercial version of UVA’s DiAs system (TypeZero’s inControl AP) paired with a Dexcom CGM; it is intended to support a PMA submission, and FDA has informed the trial design (see our previous coverage). Today’s announcement did have some contradictory wording – the subtitle firmly states trial will use the pump, while the first paragraph says “in consideration for use.” The latter reflects a hedge on the regulatory front (IDE approval is always uncertain, particularly for Cellnovo’s currently EU-only pump), and of course, the fact that other pumps will be used in the trial too. The algorithm integration path is still under consideration – TypeZero may integrate the algorithm into Cellnovo’s handset or keep it in a smartphone platform. The selection of Cellnovo is sensible from a design perspective, as its connected handheld offers remote monitoring and a better user interface than most traditional pumps. Additionally, one-third of IDCL sites are in Europe, where Cellnovo’s pump is already approved. The announcement is a victory for Cellnovo, who is still in the early stages of commercializing its pump (224 systems shipped since launching in 2014). The company has a seen a marked decline in valuation since its IPO last July (down to ~€62million from ~€83 million in 3Q15 and ~€139 million shortly following the IPO), and we’ve heard about some key management departures in recent months. Still, we’re encouraged to see TypeZero enabling patient choice, as usability is such a major part of closing the loop successfully.

6. Dr. Boris Kovatchev shared six-month results from the five-month extension phase of JDRF’s multicenter 24/7 closed-loop trial testing UVA’s DiAs system (we saw one-month results at ADA 2015). Overall, 14 participants were recruited to use DiAs in the five-month extension phase, with sustained and dramatic reductions in hypoglycemia from baseline vs. the last three months of the study – time <70 mg/dl declined ~68% (from 4.1% to 1.3%; p<0.001) and time <50 mg/dl declined a marked ~90% (from 1% to 0.1%; p<0.001). A1c declined non-significant 0.2% overall at six months from a well-controlled baseline of 7.2% (p=0.16); however, A1c was significantly reduced in three out of four sites (-0.5%, -0.4%, and -0.2%; p=0.02), with one site seeing a converse 0.5% increase in A1c (presumably due to the reduction in hypoglycemia). Extension participants spent an impressive 77% of the time in 70-180 mg/dl overall, with 83% time-in-range at night – DiAs really shines while patients are sleeping (it steadily targets 120 mg/dl by morning), and we expect many well-controlled patients will get a lot of benefit from only using it at night. As has been demonstrated with CGM data, outcomes were related to system use – those who wore DiAs >70% of the time experienced an outstanding 87% reduction in time <70 mg/dl (from 7% to 0.9%; p=0.01) and a 0.5% reduction in A1c (baseline: 7.2%; p=0.02). We see this as terrific efficacy in very well-controlled patients. Overall, DiAs was used in closed-loop mode about 70% of the time during the extension, encouraging utilization for a 24/7 system still in the research phase (participants had to carry an Android smartphone, Accu-Chek pump, and G4 Share receiver). The team is gearing up for the International Diabetes Closed Loop trial (n=240), which will generate the safety and efficacy data to satisfy a regulatory submission after six months. As a reminder, this trial will use a commercial version of DiAs from startup TypeZero (see our previous report). Dr. Kovatchev did not mention the partnership with Cellnovo for this trial (announced this morning; see above). More from this talk is below!

7. Academic Medical Center’s Dr. Hans DeVries presented full 90-day data (n=71) from the EU pivotal trial of Senseonics Eversense implantable CGM sensor, on-body transmitter, and mobile app. Overall MARD vs. YSI at ten in-clinic visits was 11.5%, nearly identical to the interim results (n=44) shown at DTM last fall. Accuracy still diminished in the hypoglycemic range (<75 mg/dl), where overall MARD was 20%. Still, the Clarke Error Grid showed a strong 85% of measurements in Zone A. Sensors were fairly durable, with 82% making it to 90 days, and though we have long criticized the on-body transmitter, compliance with wearing it was excellent: a median 23.5 hours of wear time per day. Patients even saw a 0.5% decline in A1c from a baseline of 7.6% (p=0.051; no control group obviously, but an encouraging result). Six-month data will be presented later this year, and as of our trip to the exhibit hall on Wednesday, Eversense is still pending a CE Mark. Senseonics has already learned from this data and improved the algorithm’s MARD to 10.5%, which is being used in the just-initiated US pivotal study. Dr. DeVries also revealed that FDA is requiring a blinded sensor in the US pivotal study. We had not previously known that and are baffled by that decision. The big question for Senseonics is whether it can expand the EU and US glucose sensor markets, where competition is fierce with Abbott’s FreeStyle Libre, Dexcom’s G5, and Medtronic’s 640G (EU) and 670G (coming). Does Eversense’s on-body transmitter negate the implantable advantage in those naïve to wearing things on the body? (This trial wasn’t a true test of that, as 32% of study participants had used a CGM and 42% were on a pump before this study.) Will cost be similar to current CGM? How hard will it be to make the sensor at scale and train providers to do the insertion? The company has clearly come a long way on a supremely challenging problem, and of course, it has to walk before it can run to next-gen systems.

8. Ahmad Haidar (Ecole Polytechnique de Montreal, Quebec, Canada) presented a small (n=23) 60-hour randomized trial comparing sensor-augmented pump therapy (SAP) vs. insulin-only closed-loop vs. insulin+glucagon closed-loop. Glucagon showed a small hypoglycemia benefit and no advantage on time-in-range. The outpatient study used a Roche pump, Dexcom CGM, and had patients carb count and meal bolus. Mr. Haidar skipped over the rest of the study design details too quickly to even take a picture of. Results showed a marginal benefit for glucagon in hypoglycemia: time <72 mg/dl was 7.9% with SAP vs. 3.9% with insulin-only closed loop vs. 3.6% with insulin+glucagon closed-loop (p=0.07 for single vs. dual-hormone). The advantage was bigger for time spent <63 mg/dl, where hypoglycemia was halved with glucagon: 3.5% vs. 1.9% vs. 0.9% (p<0.05 for single vs. dual-hormone). Surprisingly, there was no significant difference between the interventions on mean blood glucose (135 mg/dl vs. 142 mg/dl vs. 142 mg/dl). Time-in-range was also not significantly different between any of the interventions, though it trended better with more automation (64% vs. 75% vs. 79%). The Montreal team uses the same insulin algorithm during single- and dual-hormone, meaning it does not dose more aggressively with glucagon. We’ll be interested to see what results look like tomorrow from the insulin-only tests of the Bionic Pancreas! We are fans of more options for patients, and believe dual-hormone could be worth it for those desiring full automation or at serious risk of hypoglycemia. Of course, the additional cost and complexity has to be weighed against any potential clinical advantages.

9. Roche’s Dr. Matthias Axel Schweitzer stated that the artificial pancreas is “clearly on our agenda.” In a talk on using technology to personalize diabetes management (that featured questions from our own CES report on the opening slide), Dr. Schweitzer indicated that more work needs to be done to perfect closed-loop algorithms, but that the company is “more than well prepared” for the future. We were very encouraged to hear this, as we have heard nothing from Roche on the closed loop in recent years. Dr. Schweitzer also commented that clinical studies of Roche’s novel CGM are going well and that the product will be available soon. This is consistent with the 2016 EU launch timeline Roche provided for the product – now being called the Accu-Chek Insight CGM – in its 4Q15 update this week. We are encouraged by Roche’s recent excitement around this device and hope to learn more about how the company plans to differentiate itself from its more seasoned competitors in this area. The remainder of Dr. Schweitzer’s talk focused on health apps (promising in theory but lacking hard evidence), the benefits of automated bolus calculators and carb estimators, and Roche’s efforts to address aspects of diabetes management beyond technology (such as with its Emminens automated pattern detection software for clinicians and a new behavioral training program for diabetes nurse educators).

10. Dr. Philip Home (Newcastle University, Newcastle upon Tyne, UK) provided a comprehensive take on what prescribers should know about biosimilar insulins, focusing mainly on the limitations of the available data. He noted that prescribers do not have access to the majority of the preclinical, manufacturing, and chemical characterization data on biosimilars and suggested that most would not have the skills to interpret it even if they did. He placed the responsibility on regulatory authorities and organizers of continuing medical education programs to help disseminate this information in an understandable way. He also suggested that many providers will likely rely on the manufacturer’s reputation as a proxy for manufacturing quality, which we imagine should work to Lilly/BI’s advantage in the case of biosimilar insulin glargine. On the clinical side, Dr. Home critiqued several aspects of the typical studies used to support approval. For PK/PD studies, he argued that endpoints like area under the curve and max concentration are not sensitive enough when comparing insulin profiles and that the typical confidence intervals are too wide in a case where a dose difference of 5% is clinically significant. In phase 3 trials, Dr. Home believes change in fasting or postprandial plasma glucose is a much more sensitive and clinically relevant endpoint than change in A1c – the need to move away from an A1c-centric view of diabetes has certainly been a recurring theme this week. He also suggested than hypoglycemia, while “usually ignored in this context by regulators,” is more important to patients and providers than the typical efficacy endpoints. Furthermore, in his view, adverse event-related endpoints cannot be adequately evaluated with current studies, as most trials are too small and too short, while post-marketing surveillance is too weak.

  • In the same session, Mr. Joseph Saldanha (Julphar Diabetes, Ras al Khaimah, United Arab Emirates) spoke about the potential for biosimilars to increase access to insulin in the developing world. He opened with some striking statistics about the current state of the insulin market: three of the world’s 42 insulin manufacturers account for 93% of the revenue and 92% of the production. The plan is that the advent of biosimilars will introduce more competition into this market, thereby lowering prices and improving access, which we were glad to hear is already trending in the right direction. However, Dr. Saldanha acknowledged that the prospect of investing ~$400 million to develop and manufacture a drug that is unlikely to provide a huge short-term return is a daunting one for many companies. Interestingly, he also suggested that discounts for biosimilar insulins are unlikely to be greater than 30%, partly because some physicians have indicated that they would not trust the safety of a cheaper product. This was an interesting new perspective to us, as we have assumed the more common reaction will be disappointment that biosimilars are not discounted on the same level as small molecule generics. The diabetes market research firm dQ&A has done extensive work on this front – if you are curious to know about the reaction of patients and providers, be in touch with the great Richard Wood at richard.wood@d-qa.com.

Honorable Mentions

  • Day #3 of the meeting featured the celebrated ATTD Yearbook sessions – check out the publication online here. The comprehensive plenary reviewed the year’s progress in: (i) self-monitoring of blood glucose; (ii) new medications for the treatment of diabetes; (iii) continuous glucose monitoring; (iv) insulin pumps; (v) closing the loop; (vi) new insulins, biosimilars, and insulin therapy; (vii) using digital health technology to prevent and treat diabetes; (viii) immune intervention in type 1 diabetes; (ix) advances in exercise, physical activity, and diabetes mellitus; (x) diabetes technology and therapy in the pediatric age group; and (xi) diabetes technology and the human factor. Stay tuned for our coverage of these in our full report. We also loved the sessions’ conclusion of a mini early surprise birthday celebration for the beloved Dr. Bruce Buckingham – please join us in wishing Dr. Buckingham a very, very fantastic 70th birthday next week! ; >
  • The esteemed Dr. John Pickup (King’s College, London, UK) discussed several reasons for why much of the “best evidence” in the diabetes technology field is misrepresented or distorted, cautioning that while meta-analysis of randomized clinical trials is the gold standard, “all that is gold does not glitter”. We certainly agree with this. To that end, he asserted that short duration trials (<four to six months) and trials with first generation or outdated devices should not be included in meta-analyses as they are irrelevant and skew results – this was great to hear. In terms of study design, Dr. Pickup explained that exclusion criteria can lead to the selection of patients with no clinical problems, creating a misleading “no effect” outcome for a given intervention. For example, if patients in a pump study are well controlled at baseline and severe hypoglycemia is excluded at entry, then switching to CSII or CGM or closed-loop may show little to no effect in improving A1c. He stated that such misleading “no effect” outcomes also occur when trial participants have a response that is not measured in the study (e.g., A1c is measured but hypoglycemia is not). Dr. Pickup further explained that the convention of using mean values in data analysis is clinically “meaningless”; doctors treat patients, not means, and patient responses vary widely. Last, Dr. Pickup described his recent work checking the validity of published data in 11 RCTs using individual patient data. The results were alarming – many of the papers evaluated had major discrepancies (results entered in the wrong table columns, incorrect table legends, improver citations, etc.), and errors were not detected by the multiple authors, reviewers, and journal editors before submission – we are not too surprised with the latter since so many people in this field are being asked to do so much. Dr. Pickup concluded by advocating for stakeholders to collaborate in setting, regulating, and interpreting the diabetes technology evidence-base, highlighting the importance of establishing robust evidence to form the basis of clinical conclusions and healthcare policy. We agree that the value of strong evidence cannot be understated in diabetes technology – evidence determines regulatory approval, recommended usage in guidelines, cost-effectiveness, reimbursement policy, and patient outcomes, and is essential for positive progress in the field. And the bar is higher and more important than ever.
  • Dr. Primoz Kotnik (University of Ljubljana, Slovenia) presented new data, demonstrating that GI Dynamics’ EndoBarrier (endoscopic duodenal jejunal bypass liner) is a feasible therapeutic option for adolescents with severe obesity. The study included 15 participants, whose mean age was 17 years and mean BMI was 42 kg/m2. The results found that the majority of participants achieved meaningful weight loss, with a mean weight loss of ~16% and BMI reduction of 6 kg/m2 at 12 months. In addition, the findings demonstrated improvements in several metabolic parameters, including glucose metabolism, dyslipidemia, and blood pressure. According to a psychological evaluation at 12 months, participants also experienced some positive changes in eating behavior and other psychosocial measures (i.e. less socially withdrawn, physically aggressive). Results also demonstrated an acceptable safety profile within this patient population. As bariatric surgery remains relatively controversial in this age range, this data is promising in potentially expanding the number of treatment options for adolescents with severe obesity.
    • In addition, Dr. Kotnik briefly pointed to the liver abscess safety signal in the terminated US ENDO Trial, noting that the liver abscess rate elsewhere has remained much lower. Similar to the German Diabetes Society’s recent statement on the trial termination, he shared that the liver abscess rates globally (n=3,000) and in Germany (n=651) are 0.73% and 0.46%, respectively, compared to the US trial’s 3.2%. Regarding possible reasons as to why the trial’s rate is significantly higher, Dr. Kotnik pointed to the possible roles of the trial’s high-dose proton pump inhibitors, type 2 diabetes population, or even a difference in immune status. As a reminder, GI Dynamics reported in its 3Q15 update that the company has an ongoing review of the trial’s preliminary results – we look forward to this full data to gain a more comprehensive understanding of the device’s risk-benefit profile.
  • Dr. Jane Seley (Weill Cornell Medical Center, New York, NY) shared the design and preliminary data of a new ongoing trial that aims to examine strategies on preventing hospital readmission in high-risk diabetes patients. Regarding key factors of readmission, she pointed to low socioeconomic status, racial/ethnic minority, multiple co-morbidities, public insurance, urgent admission, and recent hospitalization. In her study (n=36), Dr. Seley provides every patient with the opportunity to receive diabetes education, which includes three key strategies (in addition to the usual care of inpatient self-management education) as part of a transitional care flow chart: (i) “Med-to-Bed” medication delivery (patients have all prescriptions filled upon returning home); (ii) a three-day follow-up phone call; and (iii) a seven-day post discharge outpatient visit. According to the preliminary data, of the 36 participants, 11 participated in none of the strategies; five participated in all strategies; 14 participated in only “Med-to-Bed”; five participated in “Med-to-Bed” and the three-day call; and one participated in “Med-to-Bed” and the seven-day follow-up. We look forward to hearing more of Dr. Seeley’s findings, especially her insights on how to motivate and engage patients to participate in diabetes education, as such use appears to remain extremely low in even the most well-resourced patients – see our coverage of CDC data on this for more.

Detailed Discussion and Commentary

Oral Presentations

Use of Novel Flash Glucose-Sensing Technology to Optimize Glucose Control in Individuals with Type 2 Diabetes on Intensive Insulin Therapy (REPLACE)

Thomas Haak, MD (Diabetes Center Mergentheim, Germany)

Dr. Thomas Haak shared long-awaited results from Abbott’s FreeStyle Libre REPLACE study, a randomized six-month trial comparing Abbott’s new flash glucose monitoring (the FreeStyle Abbott Libre) to SMBG in type 2 patients on basal/bolus therapy in poor control (baseline A1c: 8.8%) on insulin therapy. The session was jam packed (they stopped letting people in), and we have all the data and Close Concerns’ own analysis below. Patients were 2:1 randomized to either use capillary blood glucose testing (n=75; FreeStyle Lite ~ 4 times per day) or real-time use of FreeStyle Libre (n=149) for self-management and review at regular clinician visits. The trial’s primary outcome was not met at six months – from a baseline of 8.8%, both the SMBG and FreeStyle Libre groups experienced a 0.3% reduction in A1c (p=0.82). A pre-specified secondary analysis did reveal an A1c advantage in the subgroup <65 years old: -0.5% vs. -0.2% (p=0.03). The opposite was true in patients ≥65 years, who actually performed far better in the control group (-0.1% vs. -0.5%; p=0.008), which we believe was probably due to HCP caution over having these patients “too” well controlled and risking hypoglycemia.

Overall, the most compelling takeaway from this trial was the hypoglycemia data (measured via masked Libre Pro in the SMBG arm) improved markedly with FreeStyle Libre overall, overnight, and particularly for dangerous hypoglycemia. Had hypoglycemia been the primary outcome, the trial would presumably easily have been a success. Relative to the control group, patients using FreeStyle Libre spent ~30 minutes fewer per day <70 mg/dl (p<0.001), ~13 minutes fewer per day <55 mg/dl (p=0.001), and ~8.5 minutes fewer per day <45 mg/dl (p=0.001) – remarkably strong in our view, especially since any time spent at 45-55 mg/dl is very dangerous and can easily result in an ambulance call or hospital visit or stay. For the FreeStyle Libre group, these reductions equated to major 55%, 68%, and 75% reductions in those respective zones from baseline to six months (based on the limited data given, it was not possible to calculate these percentages for the control group). All measures of nocturnal hypoglycemia were also significantly lower with FreeStyle Libre, countering the criticism that the device’s lack of alarms poses a nighttime danger - presumably, the retrospective glucose data helped identify nocturnal hypoglycemia. There were no device-related serious adverse events and nine instances of minimal adverse events (e.g., infection, allergy) from six subjects.

Ultimately, we had high expectations coming in to this trial – it seemed like slam dunk to improve A1c in insulin-using type 2s (baseline A1c: 8.8%) testing ~ four times per day though we note those are highly engaged patients (the average person with type 2 tests less than once a day and plenty even on mealtime insulin test less). We are extremely disappointed that the trial missed its primary endpoint in all patients. That said, it showed profound and meaningful reductions in hypoglycemia – particularly the ~75% reduction in time spent <45 mg/dl – and patients <65 years did see a 0.3% benefit on A1c while simultaneously reducing severe hypoglcyemia – a clear win! A1c is the most devilish of outcomes for diabetes technology, as devices typically profoundly reduce hypoglycemia, often at the expense of raising average glucose.

We do wonder what could have been done differently. Most importantly, the study design did not use a per-protocol mandatory insulin adjustment (i.e., not treat-to-target), which would have unquestionably, in our view (and other smart people) improved the magnitude of A1c benefit. In this trial, providers and patients could decide what to do with the data at their regular visits, a deliberate decision to ensure a real-world trial – that said, it’s been clear for some time that if people do not take action on data, A1c will not improve! We wonder if providers were drawn to the red traffic light on AGP that identifies hypoglycemia – particularly in the older type 2 patients! – and backed off therapy as a result (and too much?). Of course, it is also much easier to fix hypoglycemia (reduce insulin) than to safely bring mean glucose down (“Is it correction or food bolus? Or is it basal?”) We also wonder about the study population, as these were patients far from A1c goal and already testing four times per day. Within that framework, the full results have to be interpreted in the proper context – not a smashing success and not a success in terms of primary outcome, but clearly a major success in terms of safety. Notably, this trial could have met its primary endpoint with a hypothetical 0.5% reduction in A1c for all patients, and could’ve reduced hypoglycemia at the same – that would have been a clear success but we still consider this directionally very strong, as we think with the right advice from doctors or nurses, patients will be able to drop their A1c. All things considered, we salute Abbott for conducting this ambitious, long-term outcomes study of FreeStyle Libre. The ultimate mark of any technology is whether patients will buy it, and with FreeStyle Libre, they are paying out of pocket and demanding it faster than Abbott can make sensors. In the immortal words of highly regarded Dr. Jane Seley, whom we saw a few hours ago in Milan, “and patients will do this – they WANT this!” Of course, reimbursement will open access for far more patients, and we hope this study and subsequent studies make a strong case that more frequent, actionable glucose data is beneficial. An A1c impact will make things more clear – this is also a reminder that A1c is isolation just isn’t the best metric – and we question why hypoglcyemia wasn’t included as a primary outcome. Last, it will be extremely important to see how the type 1 data from IMPACT (to be shown at ADA 2016) will impact these results – since we know that patients acting on data can reduce A1c and make for “higher quality” A1cs, we hope we see that. Since patients in that trial have an A1c <7.5%, there may well be room for an even bigger impact on the hypoglycemia front.  

Study Design

  • REPLACE (ClinicalTrials.gov Identifier: NCT02082184) randomly assigned 224 type 2 patients on insulin therapy to six months of either capillary blood glucose testing (n=75; FreeStyle Lite) or sensor glucose data (n=149; FreeStyle Libre) for optimization of glucose levels. All patients entered the study as regular blood glucose testers (≥10 fingersticks/week, averaging to roughly four per day). During the study phase, patients in the intervention arm reviewed FreeStyle Libre Software summary reports (Ambulatory Glucose Profiles) with their clinician at regular intervals in order to make therapy adjustments, while those in the control arm reviewed diary readings with their clinician at similar intervals.
  • Notably, insulin dose adjustments were made on an intention-to-treat basis. Providers were instructed to optimize therapy as they saw fit, but there were no A1c targets or previously mandated dose adjustments.
  • Patients at baseline had a mean age of 59 years, a mean 17 year duration of diabetes, a mean A1c of 8.8%, a mean BMI of 33 kg/m2, and a mean self-reported blood glucose frequency of ~3.7 per day. Roughly 95% of patients (n=212) were on MDI. In short, this was a challenging population in which to show benefit (not at goal but testing four times per day), but also one with high potential to reduce A1c meaningfully if they were acting on data.

Results

  • Use of FreeStyle Libre was associated with significantly improved A1c in subjects <65 years but not for the entire population. Mean A1c improvement in the total population was a similar 0.3% for both groups (p=0.82), who started with baseline A1c values of 8.8% (control) and 8.7% (intervention), respectively. In the younger subgroup (<65 years old), A1c improved more in the intervention arm (-0.5% vs. -0.2%, p=0.03) though the reverse was seen in patients ≥ 65 who actually performed better in the control group (-0.1% vs. -0.5%, p=0.008). The latter could have been because HCPs are so worried about hypoglcyemia that they are backing off too much from appropriate therapy.

Change in A1c

Group

N

Baseline Mean

Change in A1c at six months

P-value

Control

75

8.8%

-0.3%

p=0.82

Intervention

149

8.7%

-0.3%

Change in A1c for Age Subgroups

Age (years)

Group

N

Baseline Mean

Change in A1c at six months

P-value

<65

Control

47

8.8%

-0.2%

p=0.03

Intervention

95

8.8%

-0.5%

≥65

Control

28

8.4%

-0.5%

p=0.008

Intervention

54

8.3%

-0.05%

  • Notably, all measures of hypoglycemia were significantly lower following intervention with FreeStyle Libre vs. SMBG. Specifically, patients: (i) spent ~30 minutes fewer/day < 70 mg/dl (p<0.001); spent ~13 minutes fewer/day < 55 mg/dl (p=0.001); and (iii) spent ~8.5 minutes fewer/day <45 mg/dl (p=0.001). These are very important data given the costs associate with hypoglycemia, particularly severe hypoglycemia.
    • All measures of nocturnal hypoglycemia, too, were significantly lower following intervention with FreeStyle Libre vs. SMBG. For context, patients: (i) spent ~17 minutes fewer < 70 mg/dl (p=0.0001); spent roughly seven minutes fewer < 55 mg/dl (p=0.003); and (iii) spent roughly five minutes fewer < 45 mg/dl (p=0.004). While some may question these numbers and “lower” overall – we stress that any time at all not spent in severe hypoglcyemia is a big win.

Table: Time in Hypoglycemia (hours per 24-hour day)

Glucose Level

Intervention Group Baseline (days 1-15)

Intervention Group Final (days 194-208)** ^

Difference (vs. control) in change from baseline

P-value

<70 mg/dl

1.30

0.59

-0.47

P=0.001

<55 mg/dl

0.59

0.19

-0.22

P=0.001

<45 mg/dl

0.32

0.08

-0.14

P=0.001

* Note: Abbott did not report the time in nocturnal hypoglycemia for the control group at either baseline or final study period. **Similar baseline and final results were not reported for the control group. ^ The study had a 15-day run-in with blinded CGM to determine baseline control, followed by a 180-day study period (SMBG vs. FreeStyle Libre), and then a 15-day masked CGM phase in the control group vs. 15 more days of FreeStyle Libre.

Table: Time in Nocturnal Hypoglycemia (hour between 11 PM and 6 AM)

Glucose Levels

Time in Nocturnal Hypoglycemia (hours between 11 PM and 6 AM)

P-value

Intervention Group Baseline (days 1-15)

Intervention Group Final (days 194-208)

Difference (vs. control) in change from baseline

<70 mg/dl

0.55

0.49

-0.29

P=0.0001

<55 mg/dl

0.27

0.18

-0.12

P=0.003

<45 mg/dl

0.16

0.08

-0.08

P=0.004

* Note: Abbott did not report the time in nocturnal hypoglycemia for the control group at either baseline or final study period

  • FreeStyle Libre pretty much completely replaced blood glucose testing, suggesting a high level of confidence in the factory calibrated sensor. SMBG frequency with FreeStyle Libre fell from a mean of 3.8 tests/day at baseline to 0.3 tests/day (one every three days) at six months. Patients in the FreeStyle Libre arm scanned for glucose 8.3 times per day, which is once every two waking hours (Certainly more real-time glucose data than they were getting with SMBG, but probably not as much as type 1s will scan). The trend is a testament to the real-world accuracy of FreeStyle Libre in patients on insulin therapy. As a reminder, FreeStyle Libre’s label recommends confirmatory fingersticks : (i) during times of rapidly changing glucose; (ii) when hypoglycemia or impending hypoglycemia is reported by the system; or (iii) when symptoms do not match the system readings. However, these data are a reminder that patients don’t do fingersticks with FreeStyle Libre in the real world, something we’ve hear since the system launched.
    • The control group maintained their level of blood glucose testing through the study – baseline: 4.0 test/day; six months: 3.0 tests/day.

Figure: Number of FreeStyle Libre Scans and Blood Glucose Tests Per Day

  • FreeStyle Libre improved quality of life and patient-reported outcome measures, per two separate metrics. Diabetes-Treatment-Satisfaction Questionnaire results showed an increased overall treatment satisfaction for FreeStyle Libre vs. SMBG (13.1 vs. 9.0; p<0.001), while the Diabetes Quality of Life (DQoL) survey also showed increased treatment satisfaction for FreeStyle Libre vs. SMBG (-0.2 vs. 0.0; p=0.03). We absolutely loved how Abbott reported these outcomes and would urge the field to decide together on how to report this data and move toward standardized reporting.
  • No device-related serious adverse events were reported. Overall, 520 adverse events were reported during the course of the study; of these, only nine (reported by 6 subjects) were related to the device. According to Abbott, all nine events were related to an adhesive reaction, allergy, rash, or local infection at the sensor site and quickly resolved. This corroborates some patient reports on Twitter complaining about the adhesive – this is to be expected with any technology like this, and we assume Abbott is thinking about how to improve the adhesive in next generations although there’s not really a lot they have to do as most patients are not experiencing problems.

Close Concerns’ Analysis

  • The A1c results are underwhelming in terms of the primary outcome of A1c, given the 8.8% baseline and given the enormous patient and HCP enthusiasm associated with the device. We had been hoping for population-wide improvement and it’s disappointing that patients ≥65 years old saw improved glycemic control on SMBG, though as noted – unless patients are working with HCPs to respond appropriately to date, of course their A1cs will not necessarily improve –that is not to detract at all from the importance of hypoglycemia reduction ! The small incremental A1c advantage (0.3%) in FreeStyle Libre patients <65 years was also lower than we would have expected overall and raises the case whether Libre might be more apt to focus patients on avoiding hypoglycemia rather than hyperglycemia – that is addressable without a doubt in our view.
    • The study design did not use a per-protocol mandatory insulin adjustment (i.e., not treat-to-target), which would have unquestionably improved the magnitude of A1c benefit. Providers and patients could decide what to do with the data at their regular visits, a deliberate decision to ensure a “real-world trial”. That was a very tough call, and given the A1c endpoint, we believe it would have been smart to include a treat-to-target component.
    • As we have been doing for multiple years, we point out the limitations of A1c as a yardstick for real-world value. A1c is of course an incomplete metric and one that is obscured by changes in hypoglycemia. Glucose control for many patients is not defined just in terms of this measurement but in terms of time-in-range, averages as well as standard deviation glucose data, quality of life, fear, hospitalizations, etc. The most successful therapies will improve value across the board and FreeStyle Libre is definitely passing the ultimate test of a product: demand has exceeded supply! Indeed, last year’s capacity constraint speaks to the way Libre is absolutely helping patients, and we imagine we are likely not seeing the real-world efficacy of Libre in the RCT setting. As a sidenote – we’ve heard here in Milan that patients who have Libre can now order up to six sensors at once and the elation we’ve heard associated with this change has been quite compelling.
    • Medtronic’s OpT2mise trial of pumps in type 2 is an interesting comparator on the A1c and hypoglycemia fronts. That study enrolled a population with a baseline A1c of 9.0% and doing 2.5 SMBG tests per day (slightly worse than this trial, but not by much). The trial showed an A1c reduction of 1.1% with an insulin pump vs. 0.4% in the MDI group (p<0.001) after six months, though hypoglycemia was not improved in the pump group.
    • What would the A1c outcome have looked like in REPLACE if hypoglycemia was unchanged? Put differently, what is a bigger success? A 1% decline in A1c and no change in hypoglycemia, or no change in A1c but a 50% reduction in hypoglycemia? For payers, arguably the latter has better short-term payoff.
    • Was hypoglycemia in older patients + provider bias at work? We wonder whether clinicians were discovering a lot of undocumented hypoglycemia with FreeStyle Libre in the elderly and backing off treatment. We would point out that Abbott’s Ambulatory Glucose Profile does draw attention with red traffic lights to problem areas, and it’s certainly easier to improve hypoglycemia (back off insulin, eat more) than to improve hyperglycemia. Were providers drawn to the hypoglycemia? Did they prioritize fixing that over reducing mean blood glucose? That certainly seems like a likely explanation to explain the lack of an A1c improvement in the older group – and we see this as addressable.
    • Did older patient scan less frequently? This was our first thought, but the opposite was actually true: older patients scanned more often than those in the younger cohort (8.4 times/day vs. 8.0 times/day). We do wonder whether patients’ perception of success is more associated with avoiding hypoglcyemia than hyperglycemia – probably!
    • To better understand REPLACE’s A1c data, we would love to know what proportion of patients improved their A1c by 0.5% or more? BY 1% or more? By 2% or more? What percentage of patients improved their A1c by 0.5% or more or reduced hypoglycemia by 30% or more? Were there “responders” and “non-responders” to FreeStyle Libre? Or was the small magnitude of A1c improvement consistent across the board?
  • Despite the underwhelming A1c findings, the hypoglycemia improvement lived up to its billing. All measures of hypoglycemia (day+night and nocturnal) were significantly lower with FreeStyle Libre, and it’s quite evident that these results are clinically meaningful – minutes of hypoglycemia saved daily translate to YEARS of healthy complication-free life later, particularly in the <45 mg/dl zone. For context, we the results seem somewhat comparable to those of the ASPIRE in-home study of the MiniMed 530G – and that device was taking action by suspending insulin! That trial showed a 32% reduction in nocturnal hypoglycemic events and a 38% reduction in mean area under the curve of nocturnal hypoglycemia events without an increase in A1c levels.
    • We found the significant reduction in time spent < 45 mg/dl particularly striking given how dangerous that range is. Even changes in a small number of minutes spent in that low range there could mean many healthcare dollars saved annually – we have to imagine that that is one of the most compelling takeaways for Abbott and would form the crux of any payer pitch.
    • We were impressed, too, to see the improvement in overnight hypoglycemia improvement – even without alarms with FreeStyle Libre! The finding implies that patients don’t have to have alerts to improve the overnight period … they just need the data, and then they can adjust therapy to create success.  
  • In summary, Abbott’s full results strike us as quite solid – not a smashing success but certainly not a failure. Our gut reaction was disappointment (as we’re sure it was for many) but some perspective is needed; the data could have been worse (imagine if Abbott had hit its primary endpoint with a 0.4% reduction in A1c but increased hypoglycemia?) and could have been better (imagine if Abbott had picked hypoglycemia as the primary endpoint?). In the big scheme of things, today’s results likely fall somewhere in the middle of the road – we think it’s more likely that the long-term implications will be defined by the questions Abbott asked and lessons the company (and others) learn rather than missing the primary outcome. While we very much would’ve loved to have seen a 1% A1c reduction or at least the 0.5% A1c reduction that usually implies success, this is a more complicated story and given the enormous enthusiasm surrounding the product, we believe those are the design front will be diving in
    • Speaking of the long-term implications, we are unclear what the results mean for future reimbursement of FreeStyle Libre in type 2. What would payers say about this data? Certainly, there is a high EU bar for cost-effectiveness and added benefit, though CGM is getting some reimbursement in Europe (and FreeStyle Libre is cheaper than CGM – about half the price). How will Abbott proceed with this data? When combined with the results fro IMPACT in type 1, could it form the basis for reimbursement in some major markets? Or will the company need to pursue additional studies to supplement these findings? And speaking of additional studies, we can’t help but wonder what FreeStyle Libre might look like in the real world, complemented by the right drugs and the right behavior. It could do so, so much!
    • We also have to think that the reasonably positive findings here mean that Abbott’s six-month study in type 1s (primary outcome: hypoglycemia) is even more likely to be a smashing success. That FreeStyle Libre was able to achieve a hypoglycemia improvement and partial A1c improvement in the much tougher type 2 population bodes well for type 1. As a reminder, Abbott will present findings from IMPACT at ADA 2016.
  • Ultimately, we salute Abbott for ambitiously pursuing an outcomes study and testing these uncharted waters. A1c remains the metric-of-choice in the eyes of payers and despite the surface-level disappointing results, it’s not clear that less-ambitious-but-more-impressive results would have delivered a no-brainer decision. Instead, we think the results serve as an important lesson and valuable learning experience as companies begin thinking about study design and about engaging with reimbursement bodies. The ultimate goal is to increase the increase the number of people who would benefit from this technology, and moving forward, it’s absolutely critical to think about what is best for different populations.

Close Concerns’ Questions

  • Were there “responders” and “non-responders”? What proportion of patients improved their A1c by at least 0.5% OR reduced time in hypoglycemia by 30%+? Was there anyone in the trial that saw no benefit on A1c or hypoglycemia?
  • What would the A1c outcome have looked like in REPLACE if hypoglycemia was unchanged? What is a bigger success from a payer perspective? A 1% decline in A1c and no change in hypoglycemia, or no change in A1c but a 50% reduction in hypoglycemia? Which would a payer prefer? Which would a provider prefer? Which would a patient prefer?
  • How would these results have looked with FreeStyle Libre Pro used intermittently instead of real-time data?
  • In retrospect, what would Abbott change about the study design? Should sites have been mandated to make insulin adjustments based on the data? Should the study have enrolled patients who weren’t already checking their blood glucose so regularly?
  • What are the implications for reimbursement? How will payers receive the data? Will the magnitude of the benefit be enough to show cost-benefit? What will Abbott do going forward?
  • What are the implications for IMPACT in type 1? Should these results raise our expectations for what we’ll see in terms of hypoglycemia reduction in type 1 patients?
  • What can be learned from this trial for future studies? How should the diabetes technology field think about designing outcomes studies for reimbursement? What should closed-loop investigators take away from this trial?

Nightscout - FreeStyle Libre Update

  • Separately, we’ve learned that Nightscout users in Italy have added remote monitoring of FreeStyle Libre via an Android app, Glimp. Glimp has existed for some time as an unauthorized app for reading the FreeStyle Libre sensor (we wrote about it last October), though Nightscout users in Italy are now using it to remotely monitor patients on FreeStyle Libre. A major win for parents, and the app can also send readings to HCPs. The instruction manual is posted within the Nightscout Italy Facebook group (which requires permission to enter). Even with Abbott’s LibreLink Android app for reading the FreeStyle Libre sensor (limited launch in Sweden in November), it does not enable remote monitoring to our knowledge. We hope that the addition of a pediatric label will move Abbott to add remote monitoring to LibreLink in the future – parents love Dexcom Share and MiniMed Connect. It’s worth noting that Glimp is an unauthorized app with fairly good reviews (4.3/5.0, 151+ ratings), though it does reveal the downsides of patients hacking into devices themselves – the FreeStyle Libre reader makes some corrections to the raw sensor data, so Glimp does not display the exact same value as the reader (“it’s close,” notes the reviews). Of course, parents will only use it if it works, and it seems like it does based on the reviews.

Clinical Accuracy Evaluation of Freestyle Libre Flash Glucose Monitoring System When Used By Children and Young People with Diabetes

Fiona Campbell, MD (Leeds Teaching Hospitals Trust, UK)

Dr. Fiona Campbell shared data from Abbott’s 89-patient, 14-day EU pivotal trial of FreeStyle Libre in pediatrics, which demonstrated an MARD of 13.9% vs. capillary fingersticks (the study had 5,493 paired sensor-fingerstick points). The data were impressively consistent across the board as 84% of points were in Zone A of the Consensus Error Grid and 16% in Zone B, a testament to the device’s accuracy as patients would experience it (i.e., relative to fingersticks). The data represent a slight downtick on what has been achieved in the EU pivotal trial where Libre demonstrated an overall MARD of 11.4% vs. FreeStyle Precision BGM (n=13,195 paired points) though given the vagaries of fingersticks in pediatrics patients, we would argue that the overall accuracy is highly encouraging. Abbott will need to maintain those results at scale – we have little doubt on that front at this stage, but factory calibration is of course not easy.

  • As a reminder, Abbott announced yesterday that it has received CE Marking for the pediatric indication of FreeStyle Libre. We assume this data was used in the submission.
  • The pediatric study was conducted a bit differently relative to the previous US and EU adults studies. The trial enrolled patients at nine centers across the EU in type 1 patients on insulin therapy. Patients wore two sensors on the back of their arm for 14 days and were asked to: (i) perform four capillary blood glucose tests daily; and (ii) scan the sensor following each test. [Note: They were not asked to attend in-clinic YSI sessions.] Notably, Dr. Campbell shared that patients had a mean baseline A1c = 7.7%, ranging from a low of 5.6% to a high of 10.4% for a nice mix of well-controlled and out-of-control patients.
    • Dr. Campbell noted that the glycemic variability seen in the study was comparable to the variability seen in other pediatric CGM studies, stressing that the sensor was tested across the full range of reasonable sensor values.
  • MARD was not broken down by glucose range, so accuracy in hypoglycemia is an unanswered question. The product label has not changed other than the pediatric indication, so we assume it will still recommend a confirmatory fingerstick when patients are hypoglycemic. Still, patients in the real world clearly trust the device enough to test rarely, so we have little concern about the hypoglycemia data in this study.
  • Dr. Campbell shared positive data from user/caregiver experience questionnaires of FreeStyle Libre in the study. There was no specifics on how these questions were asked though answers could vary between: Strongly Agree /Agree /Neutral /Disagree/ Strongly Disagree. The data certainly point to why European patient uptake has been so strong in these early days, especially in those that have avoided current CGM due to comfort/wearability. The data below were for the “Strongly Agree” answer choice.
    • It did not hurt when the sensor was put on. Girls, 84% (Strongly Agree) - Boys, 84% (Strongly Agree)
    • It was easy to put the sensor on. Girls, 91% - Boys, 93%
    • I did not mind wearing the sensor on my arm. Girls, 93% - Boys, 98%
    • It was comfortable to wear the sensor. Girls, 88% - Boys, 66%
    • It was easy to scan the sensor. Girls, 100% - Boys, 100%
    • It was more comfortable [than BGM]. Girls, 93% - Boys, 95%
    • It was less painful [than BGM]. Girls, 91% - Boys, 85%
    • It was more private [than BGM]. Girls, 83% - Boys, 88%
    • It was quicker to check my blood glucose. Girls, 100% - Boys, 95%
    • It was easier to use. Girls, 98% - Boys, 98%
    • It did not get in the way of my daily activities. Girls, 88% - Boys, 85%
    • I liked how my glucose readings were shown on the screen. Girls, 88% - Boys, 90%
    • It gave me more information than my current glucose meter to take care of my diabetes. Girls, 71% - Boys, 68%
    • It helped me understand how my daily activities changed my glucose levels. Girls, 67% - Boys, 70%
    • It made me feel more interested in taking care of my diabetes. Girls, 74% - Boys, 73%
    • I would recommend it to someone else with diabetes. Girls, 95% - Boys, 98%
  • Dr. Campbell also shared positive data from provider experience questionnaires of FreeStyle Libre in the study.
    • 100% agreed that the report’s visual presentation means I would be able to effectively share information with my patients/caregivers.
    • 100% agreed that the reports present information in such a way that I would be able to engage patients/caregivers with the information.
    • 100% agreed that the reports help me identify glucose trends that are not visible with BGM data and could potentially support informed therapy decision.
    • 100% agreed that the reports help me to assess the effectiveness of my patient’s current therapy.
    • 78% agreed that the reports help me identify hypoglycemic risk.
    • 100% agreed that the reports allow me to easily determine how much time my patients’ glucose level is within their target range.
    • 89% agreed that the reports help me to easily identify post-prandial trends.
    • 89% agreed that the reports provide insight to frequency, duration, and pattern of hypoglycemic events.
    • 100% agreed that the reports provide insight to frequency, duration, and pattern of hyperglycemic events.
  • Very few adverse events were reported among patients in the study – 44 subjects reported any sort of discomfort related to the sensor insertion and all reports were consistent with what would be expected following insertion of a sensor into the skin. There were no severe adverse events reported and all patients were able to see the trial to conclusion.
  • Unsurprisingly, Dr. Campbell shared sky-high enthusiasm for FreeStyle Libre. Asked in Q&A which patients she would recommend Libre to, Dr. Campbell did not skip a beat: “ALL of them!”

Questions and Answers

Q: Can you talk about the data in hypoglycemia and hyperglycemia?

A: I haven’t shown this here. There is still a lot of data to come out of this. This was preliminary data that I showed today. There will be more to do to evaluate all of that.

Q: What percentage of pediatric type 1 patients would you recommend Libre for?

A: All of them!

Q: Do you think FreeStyle Libre should replace CGM?

A: I’ve had a lot of people asking me that. CGM is acceptable for some but not all people and we know that there are difficulties getting young people to comply with alarms. Libre is straightforward and has no alarms. I would say that Libre should replace SMBG. However, it’s a real discussion when thinking about the patients that would succeed on CGM vs. FGM. I think about it as yet another addition to our diabetes armamentarium.

Plenary: Closing The Loop

JDRF Multi-Center 6-Month Trial Of 24/7 Closed-Loop Control

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

Dr. Boris Kovatchev shared six-month results from the five-month extension phase of JDRF’s multicenter 24/7 closed-loop trial testing UVA’s DiAs system (we saw one-month results at ADA 2015). Overall, 14 participants were recruited to use DiAs in the five-month extension phase, with sustained and dramatic reductions in hypoglycemia from baseline vs. the last three months of the study – time <70 mg/dl declined ~68% (from 4.1% to 1.3%; p<0.001) and time <50 mg/dl declined a marked ~90% (from 1% to 0.1%; p<0.001). A1c declined non-significant 0.2% overall at six months from a well-controlled baseline of 7.2% (p=0.16); however, A1c was significantly reduced in three out of four sites (-0.5%, -0.4%, and -0.2%; p=0.02), with one site seeing a converse 0.5% increase in A1c (presumably due to the reduction in hypoglycemia). Extension participants spent an impressive 77% of the time in 70-180 mg/dl overall, with 83% time-in-range at night – DiAs really shines while patients are sleeping (it steadily targets 120 mg/dl by morning), and we expect many well-controlled patients will get a lot of benefit from only using it at night. As has been demonstrated with CGM data, outcomes were related to system use – those who wore DiAs >70% of the time experienced an outstanding 87% reduction in time <70 mg/dl (from 7% to 0.9%; p=0.01) and a 0.5% reduction in A1c (baseline: 7.2%; p=0.02). We see this as terrific efficacy in very well-controlled patients. Overall, DiAs was used in closed-loop mode about 70% of the time during the extension, encouraging utilization for a 24/7 system still in the research phase (participants had to carry an Android smartphone, Accu-Chek pump, and G4 Share receiver). The team is gearing up for the International Diabetes Closed Loop trial (n=240), which will generate the safety and efficacy data to satisfy a regulatory submission after six months. As a reminder, this trial will use a commercial version of DiAs from startup TypeZero (see our previous report). Dr. Kovatchev did not mention the partnership with Cellnovo for this trial (announced this morning; see below).

  • Dr. Kovatchev shared a striking testimonial from the extension phase, a clear statement of how much automated insulin delivery will help many with type 1: “When the study started my A1c was 7.7%, half way through it had dropped to 7.1%, and on the morning before we turned in the equipment, I had my regular, quarterly appointment with my endocrinologist and my A1c was.....drum roll please ..... 6.6. Is that awesome or what?!?! That is my all-time lowest A1c ever! Along with the drop in my A1c, I have lost 22 pounds...”
  • For the first time, Dr. Kovatchev revealed that the International Diabetes Closed Loop trial (n=240) will have both hypoglycemia and A1c outcomes, depending on baseline characteristics. For those with an A1c >8%, the team aims to see reduction in A1c without increasing hypoglycemia. For those below 8%, the team wants to see the reverse – a significant reduction in hypoglycemia without an increase in A1c. We like the approach!
  • Phase 1 of the trial (one month long) was presented at ADA 2015 and demonstrated: The primary endpoint of hypoglycemia (time spent <70 mg/dl) was halved overall (4% during open loop vs. ~2% during closed loop; p<0.05), with a marked improvement at night (improved by more than two-thirds, 3% to ~0-1%; p<0.05). Overall 24-hour time-in-range (70-180 mg/dl) improved a bit (66% in open-loop vs. ~73% in the two closed-loop phases of the study), as did overnight time-in-range (62% to ~71-74%).
    • You can read Adam and Kelly’s personal experience in phase 1 of this study here – they found that the overnight algorithm was excellent (it treated to a target of 120 mg/dl by 7 AM), and the daytime algorithm erred on the side of conservative in this feasibility study. They both loved the overnight system and were glad to see the daytime version being tested to identify and work out bugs.
  • Looking ahead, Dr. Basu at the Mayo Clinic is notably NIH-funded to test a triple-hormone artificial pancreas system (insulin+glucagon+amylin). The slide indicated a timeline of 2015-2020, so we’re not sure when the study will actually begin. The goal, of course, is to replace the three missing hormones in type 1 diabetes –this is exciting from a full automation perspective, though cost and real world usability are key concerns. UVA’s Dr. Marc Breton is also NIH-funded (2015-2019) to test a multi-signal artificial pancreas system to deal with exercise (informed by a body sensor array: blood glucose, heart rate, physical activity).

 

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