June 5-9, 2015; Boston, MA; Full Report – Glucose Monitoring – Draft

Executive Highlights

This document contains our coverage of glucose monitoring at ADA 2015. Immediately below, we share our themes on the category, followed by detailed discussion and commentary.

Results from Dr. Irl Hirsch’s FLAT-SUGAR pilot study were very thought provoking. The study met its primary endpoint, suggesting it is possible to randomize patients to achieve different glycemic variability and the same A1c. Dr. Hirsch believes it paves the way for a larger, longer outcomes study testing whether glycemic variability (GV) matters for long-term outcomes. We look forward to seeing more data analysis on this trial, since it is a bit hard to know if the variability differences between the groups are big enough – we give Dr. Hirsch significant credit for running this trial, a very challenging one, particularly given that the sensors were older (mostly Dexcom Seven Plus) and the analysis required keeping patients at similar A1c levels.

Consistent with ADA 2014, this was a lighter meeting for new CGM sensors, perhaps a testament to more focus on connectivity and significantly improved CGM accuracy/reliability. Medtronic showed off the upcoming MiniMed Connect device (remote monitoring of pump/CGM data) in the exhibit hall, while Dexcom highlighted the new Share Receiver and accompanying iPhone/Apple Watch/Android Follow apps. In terms of next-gen CGM sensors, a Dexcom poster (955-P) showed feasibility for 10-day wear with the company’s in-development Gen 6 sensor, and a Medtronic poster suggested little benefit to orthogonally redundant sensing.

We were encouraged to see two notable abstracts relating glucose monitoring to healthcare costs and longer-term complications. A retrospective database analysis compared CGM users to non-users, suggesting greater improvements in A1c (0.5%), an impressive 42% reduction in inpatient hospital admissions, and a 17% reduction in emergency room visits. Meanwhile, a late-breaking poster suggested an increased risk of mortality due to CMS’ competitive bidding program. Though neither study was randomized or prospective, we believe the results are directionally interesting and important from advocacy and payer perspectives.

Talk titles highlighted in yellow were among our favorites from ADA 2015; those highlighted in blue are new full report additions from our daily coverage.

Themes

• Drs. Irl Hirsch and Jeffrey Probstfield presented the long-awaited results of the FLAT-SUGAR pilot study – the 26-week, 102-patient feasibility trial met its primary endpoint, showing it is possible to randomize ACCORD-like patients to two groups (exenatide+glargine+metformin vs. rapid-acting insulin+glargine+metformin) and achieve significantly different glycemic variability (coefficient of variation) with a similar A1c (7.1% vs. 7.2% in this case). Dr. Hirsch believes it paves the way for a larger, longer outcomes study testing whether glycemic variability (GV) matters for long-term outcomes. What is somewhat unclear from the results is whether the magnitude of glycemic variability reductions are large enough to show long term differences in future outcomes – these patients did not seem to have as much hypoglycemia or variability as we would have expected, which makes us wonder about study effect. It would also be interesting to see this study in type 1 diabetes or using different therapies (e.g., SGLT-2s) to help demonstrate much larger variability differences. There is still plenty of analysis to do on this data, and it’s entirely possible different perspectives on variability will emerge over time. This was also an extremely challenging study to conduct, given the need to keep A1c’s the same, and we salute Dr. Hirsch for moving the needle significantly on the field’s understanding of glycemia – we believe a study linking variability and time in range to long-term outcomes may be essential for proving the value of multiple next-gen therapies/technologies.
• Consistent with ADA 2014, this was a lighter meeting for new CGM sensors, perhaps a testament to more focus on connectivity and significantly improved CGM accuracy/reliability. Indeed, the biggest highlights in glucose monitoring were arguably in the exhibit hall, where Medtronic showed off the upcoming MiniMed Connect device (remote monitoring of pump/CGM data), and Dexcom highlighted the new Share Receiver and accompanying iPhone/Apple Watch/Android Follow apps. These are very meaningful form factor improvements for patients and harried caregivers (who in our view never receive enough acknowledgement). We expect to continue to see progression on this front with Dexcom’s Gen 5 (FDA approval and launch by end of 2015) and Medtronic’s Guardian Mobile is in an ongoing pivotal study (n=100), with expected completion in July 2015. Looking back at ADA 2014, there was no filing or launch timeline for Dexcom’s Gen 5, and Medtronic’s Guardian Mobile had not been publicly announced – things are certainly moving much faster on the digital and connectivity sides.
  
  • An intriguing Dexcom abstract (5-LB) highlighted the power of cloud-based CGM data to illuminate the real-world patient experience. The retrospective study analyzed 50,000 hypoglycemic events in 1,177 Dexcom Share users, suggesting an event rate of 0.96/day/patient. Notably (though unsurprising to us as patients) the proportion of hypoglycemic events followed by rebound hyperglycemia was 18% within 60 minutes and 26% within 90 minutes. Even more interesting would be to know what rebound rebound hypoglycemia looked like (i.e., overtreating a rebound hyperglycemic episode, resulting in a subsequent second low). We also wonder what time-in-tight-target range (70-140 mg/dl) looked liked immediately following hypoglycemia and the 24-hour period following hypoglycemia. Those sub-analyses aside, this poster was a great reminder of the illuminating data cloud-based glucose monitoring can bring. Though RCTs have tremendous value in driving scientific discussion, they often fall short of offering real-world insight on the day-to-day rollercoaster of living with diabetes.
• In terms of next-gen CGM sensors, two posters caught our eye:
  
  • A Dexcom poster (955-P) provided a first look at human durability data from the company’s in-development Gen 6 sensor, suggesting that extended wear is indeed feasible over 10 days. The early findings were positive – 94% of sensors lasted to 10 days, though when Dexcom included all sensors in the analysis (i.e., including non-sensor-related CGM system failures), 79% lasted for the full period. The latter was lower than we would have expected based on our experience wearing the G4 Platinum. We continue to wonder if Dexcom will go for a 10-day wear indication for the Gen 6 sensor – management most recently indicated in April that it will move into a pre-pivotal trial this summer, with plans for a 1H17 commercial launch.
  • The latest human data from Medtronic’s orthogonally redundant CGM (combining an electrochemical + optical sensor) demonstrated very little marginal benefit relative to electrochemical alone - for the full seven-day period, accuracy improved very slightly with the redundant sensor (from 11.6% to 10.3%). The biggest advantage of redundancy came in the first 48 hours after insertion – MARD fell from 16.8% with the single electrochemical sensor to 10.8% with the orthogonally redundant sensor configuration on Day 1, and from 13.7% to 10.2% on Day 2.
• After lots of excitement at ATTD 2015 in Paris, Abbott’s FreeStyle Libre was unsurprisingly (given that it is not yet FDA approved) absent at ADA … though it still seemed to be on the minds of some international (and US) attendees. On the conference floor and in hallway conversations, we heard enthusiasm for the product’s potential, which has been on the European market for nine months now. Indeed, it was notable to see the extent to which attendees (from KOLs to clinicians to educators) were familiar with the device. The “No fingersticks” message continues to inspire awe, particularly in terms of how much insight it can bring type 2 patients who aren’t in good control, and type 1 patients not on CGM because of a real or perceived hassle factor. Adding to the “Libre Fever” (our wording), Dr. Steve Russell (MGH, Boston, MA) announced that the Bionic Pancreas team hopes to use the blinded version (Libre Pro – available in India now) in its pivotal Bionic Pancreas study. FreeStyle Libre has completed its pivotal study in the US, and we assume a regulatory submission is in the works or has possibly already happened – a big question is whether Abbott will obtain an insulin-dosing claim in the US. Of course, it may be moot, since patients will dose insulin off it anyways in the real world, assuming it is still factory calibrated.
• We were encouraged to see two notable abstracts relating glucose monitoring to healthcare costs and longer-term complications:
  
  • A retrospective database analysis compared CGM users to non-users, suggesting greater improvements in A1c (0.5%), an impressive 42% reduction in inpatient hospital admissions (all-cause), and a 17% reduction in emergency room visits. Dexcom’s Dr. Claudia Graham presented the results, which were gleaned from 14 million enrollees in Optum’s national US database. Though selection bias is certainly possible in this type of study, the data are encouraging and lend some credibility to the technology’s cost-saving potential. We continue to look forward to results from Dexcom’s DiaMond study, which should offer more scientifically robust data on the value of CGM. The 338-patient, 20-center study aims to understand CGM outcomes (including healthcare costs) in MDI users. The estimated study completion date is March 2016.
  • A late-breaking poster suggested an increased risk of mortality due to CMS’ competitive bidding program. Findings demonstrated that patients in competitive bidding test sites had more than twice as many inpatient admissions (982 admissions vs. 460) and more than double the associated costs ($10.7 million vs. $4.7 million) compared to a non-test site population, which translated into almost twice as many deaths (102 deaths vs. 60). Of course, the study does not imply causation, though the results are directionally interesting and potentially alarming. As a reminder, these data differ greatly from CMS’ April 2012 report on adverse outcomes associated with competitive bidding, which suggested that there was no disruption of access to supplies and no negative healthcare consequences associated with the program. Ultimately, the results raise red flags for what is already a heavily scrutinized program. Carryover from competitive bidding (which began in July 2013) continues to exert significant pressure on the market, which was nowhere more evident than in the exhibit hall: J&J was the only member of the Big Four BGM companies present (similar to last year, but very different from years past).
• ADA was an important meeting for diabetes data, headlined by Glooko’s partnerships with Medtronic, Dexcom, and Insulet. Tidepool was also present in the exhibit hall for the first time, demo-ing its upcoming blip software in Insulet’s booth. Meanwhile, data veteran Diasend updated attendees in its own dedicated booth, highlighting compatibility with 120 devices. The proprietary data siloes seem to be increasingly opening up, though there are plenty of remaining questions: What aggregation software will patients and providers prefer (i.e., Diasend, Glooko, Tidepool)? Will glucose monitoring reports be widely standardized to an output like AGP? What will it take to bring meaningful clinical decision support/analytics to patients and providers? Who will apply for the Helmsley Charitable Trust’s Diabetes Data Innovation Initiative? Will industry adopt the interoperability standards spearheaded by Dr. Joe Cafazzo, JDRF, and HCT? Will patients be able to easily port their data from one software platform to another?

Detailed Discussion and Commentary

ADA President’s Oral Session

Glucose Variability in Type 2 Diabetes: The Initial Results of the FLAT-SUGAR Trial (385-OR)

Irl Hirsch, MD and Jeffrey Probstfield, MD (University of Washington, Seattle, WA) on behalf of Barry R. Davis, Andrew Ahmann, Richard Bergenstal, Matthew Gilbert, Connie Kingry, Dorrine Khakpour, Dejian Lai, Sara L. Pressel, Kelley Branch, Mathew Riddle, Kevin D. O’Brien

Drs. Irl Hirsch and Jeffrey Probstfield (on behalf of the FLAT-SUGAR writing committee) presented the long-awaited results of the FLAT-SUGAR pilot study – the 26-week, 102-patient feasibility trial was successful, showing it is possible to randomize ACCORD-like patients to two groups (exenatide+glargine+metformin vs. rapid-acting insulin+glargine+metformin) and achieve significantly different glycemic variability (coefficient of variation) with a similar A1c (7.1% vs. 7.2% in this case). Dr. Hirsch believes it paves the way for a larger, longer outcomes study testing whether glycemic variability matters for long-term outcomes – a modern day DCCT! In speaking with him after the session, Dr. Hirsch characterized the study as a “home run” but not a “grand slam” – he wished the secondary endpoints measuring hypoglycemia and some of the other glycemic variability endpoints had emerged as statistically significant (these all went in the right direction, but the study was not large enough). The biggest surprising finding was a much larger-than expected difference in weight – the exenatide group lost a striking 4.8 kg in body weight (11 lbs) vs. a 0.69 kg weight gain (1 lb) in the basal/bolus insulin group. We will look forward to hearing more discussion on the magnitude of glycemic variability reductions and expectations regarding long term differences in future outcomes – the patients studied did not seem to have tremendous hypoglycemia or variability, which made us wonder whether doing this study in type 1 or choosing different therapies (e.g., SGLT-2s? Afrezza?) would help demonstrate larger variability differences – we could definitely imagine a study effect and it will be interesting to think about how big data would be used in tandem with a larger RCT to draw conclusions. Of course, there is still plenty of analysis to do on this data, and we will await those.

• The 26-week study compared glycemic variability between two randomized groups: exenatide + glargine + metformin (n=52) vs. a rapid-acting analog + glargine + metformin (n=50). The primary endpoint was change in coefficient of variation (“glycemic variability”) assessed by CGM measurements (Dexcom Seven Plus, and some using the G4 Platinum) at baseline vs. 26-weeks. Secondary endpoints included other measures of glycemic variability (SD, IQR, MAGE, CONGA, MODD), hypoglycemia, weight, other biomarkers (ALT, serum amyloid A, hs-CRP). FLAT SUGAR enrolled an ACCORD-like population of people with diabetes requiring insulin. The presentation did not show participants’ baseline characteristics.
  
  • Study flow: After an 8-12-week run-in period of stabilization on basal-bolus insulin with metformin, 102 participants were continued on basal insulin and metformin, and subsequently randomized to prandial therapy with either exenatide or a rapid-acting insulin analog. Masked CGM and metabolic markers of cardiovascular risk were assessed at baseline, 13- and 26-weeks in 92 participants. Each study phase have very narrow A1c targets to progress to the next stage – 255 patients were screened (A1c 7.5-8.5%), 144 completed the run-in, 102 were randomized (A1c: 6.7-8.0%), and 92 were analyzed at week 26. (So few were analyzed at week 26 because participants had to maintain a super tight A1c to progress through to each phase of the trial; otherwise, they wouldn’t have achieved the same A1c in both groups by study end. Ninety-six patients completed the trial and four were excluded because they did not have complete CGM/Holter monitor data.)
  • Participants had to maintain an A1c of 6.7-7.3% during the trial. In both groups, glargine titration was mostly dependent on fasting blood glucose. Rapid-acting insulin (in the basal/bolus group) was titrated to optimize pre-lunch/dinner and bedtime glucose values. Exenatide was taken before at least two, and up to three meals per day; the maximal dose was 20 mcg/day.
• At 26 weeks, A1c’s were nearly identical: 7.2% in the basal/bolus group and 7.1% in the exenatide group (both declined from a baseline A1c of 7.9%). This was of course a major goal of the study – achieve the same A1c but different variability in the two groups.
• Primary endpoint: The exenatide group experienced a significant reduction in glycemic variability (coefficient of variation) at 26 weeks vs. the basal/bolus group (p=0.024). Estimating from the graph, the on-trial change in mean CV was +0.5 in the basal/bolus group vs. -2.5 in the exenatide group.
  
  • Secondary endpoints (glycemic variability): The change in MAGE was significantly improved in the exenatide group vs. the basal/bolus group (-5 vs. + 3 – these are estimates, as none of the graphs gave the absolute values; p=0.049). MAGE was the only secondary glycemic variability endpoint that achieved statistical significance, though all the other endpoints trended towards reduced variability in the exenatide group: SD (-2 vs. 0; p=0.3), IQR (-1 vs. 0; p=0.6), CONGA (-3 vs. +1; p=0.2), MODD (-2 vs. +1) – note, these numerical values are our estimates, as the slide did not provide them.
  • We imagine it is challenging for those assessing the results to benchmark these reductions in variability or to even conjecture if they are clinically meaningful enough to change long-term outcomes – that would be, of course, the point of the longer-term trial. This was only a demonstration pilot – we would be interested to see what differences in variability would emerge with more accurate CGM (the Dexcom Seven Plus was primarily used, and some used the G4 Platinum) and, of course, in a bigger study that was more “real life”-like or with further analyses.
• Dr. Hirsch showed 24-hour CGM profiles of the two groups, which suggested that the variability advantage for exenatide really came at breakfast. Overall, however, the reduction in variability was not highly apparent in the 24-hour CGM traces. The after-breakfast glycemic excursion in the exenatide group was definitely lower at 26 weeks relative to baseline, while the basal/bolus group had similar breakfast excursions at baseline and 26 weeks. There did not appear to be material differences in variability for the exenatide group at lunch and dinner, but the timing of meals was not standardized, so it’s hard to know. Overall, this slide was probably not the best indicator of how variability changed on an individual or aggregate level, and we imagine further prandial analyses will drill down on this.  
• Hypoglycemia trended lower in the exenatide group, though there was very little overall – this was a bit surprising for this population, though the fact that they received significant HCP attention at top diabetes centers may have played a role. We imagine a longer, larger trial would use a more diverse patient population at a broader number of centers. See the table below for our estimates of the values shown on the slide. There was no severe hypoglycemia in either group. We were a bit surprised that hypoglycemia rates were this low, given that one group was on Basal/Bolus therapy, but they were receiving fairly consistent care. The CGM (Dexcom Seven Plus, and some using the G4 Platinum) was only worn during 7-10 day stints at study start and end, and we wonder if 24/7 CGM would have bigger unearthed variability differences. 

• The exenatide group lost a striking 4.8 kg in body weight (11 lbs) vs. a 0.69 kg weight gain (1 lb) in the basal/bolus insulin group (p<0.001). This was characterized as an “odd finding” and a much larger-than expected difference in weight. Dr. Hirsch did not have a definitive explanation why the weight loss was so striking. It might be that compliance in this study was better with exenatide than has been seen in previous studies of the medication.
• The exenatide group experienced significant reductions in ALT (-10 vs. +1; p <0.01) and serum amyloid A (p=0.02). There were no between group differences in IL6, hs-CRP, albuminuria, or urinary isoprostanes.
• Dr. Hirsch believes the pilot study provides evidence supporting the design of a larger, longer, outcomes-based trial. The complex pilot trial design proved safe and feasible at decreasing glycemic variability while maintaining nearly equivalent Ac values. The study also showed a positive biomarker signals (ALT, serum amyloid A).
• Dr. Hirsch expressed potential interest in using Sanofi’s lixisenatide in the future, given its once daily dosing and short-acting PK/PD (cutting the postprandial spikes is essential, necessitating a short-acting GLP-1) – we think this is a good idea since even in a clinical trial, patients may not be as adherent to two or three times post-mail daily shots - .
• FLAT SUGAR’s premise: diabetes control is clearly more complex than a treatment paradigm based only on A1c. Dr. Probstfield highlighted the importance of glycemic variability, which can vary considerably between two patients but result in the same A1c. He described glycemic variability in two components: hyperglycemic spikes and hypoglycemia troughs. Glycemic variability as been shown to release inflammatory cytokines, suggesting they may promote long-term complications. We point readers to Drs. Hirsch and Brownlee’s 2010 JAMA article, which reminded readers that only 11% of the variation in retinopathy risk in the DCCT was explained by A1c and duration of diabetes.
• FLAT SUGAR = FLuctuATion Reduction With inSULin and Glp-1 Added together. See the ClinicalTrials.gov post here.

Questions and Answers

Q: Did you have enough power to detect differences? And do you anticipate a larger, longer trial – after weight loss plateaus – might get a different result?

Dr. Probstfield: On the power question, we borrowed information from one of JDRF trials. I believe that our original calculation was about 90% power if we completed 100-110 patients. We completed 92 patients, which gave us about 85% power to discern about a four-unit difference in variation.

Dr. Hirsch: In terms of plateauing of weight, it was a 26 week study. I cannot comment after 26 weeks. I’m not aware of any study that showed such a large delta in weight from a GLP-1. To me, it’s pretty interesting that there was almost a 5.5 kg weight difference.

Dr. Larry Hirsch (BD Diabetes Care, Franklin Lakes, NJ): Two questions. First, I don’t really understand a 10-11 lb difference in GLP/insulin trial – what do you think explained it? Second is the $100 million dollar question. Are these differences, which are pretty modest in variability and biomarkers, really sufficient to request funding for large outcomes study from NIH? If you were on the review committee, would you approve that request?

Dr. Hirsch: I cannot explain why there was such a profound difference in weight. We are looking at absolute kilos. We started at over 100 kilos. We need to look at all the different studies of GLP-1 and insulin and see how the baseline weights look.

We learned so much from this trial, and will do other analysis in future. Most important was how to do this study in a multi-center way. Although I cannot say this for sure, next time we do this, we’ll do it more efficiently and better. I am convinced we can show larger variation in glucose in future trials. This was a pilot first attempt, and it was successful.

Q: On the choice of exenatide, could you get similar results with a longer acting GLP-1?

Dr. Hirsch: We picked exenatide not by accident. It was based on a study by Matt Riddle five years ago. It showed a flat glycemic fingerprint on top of glargine and metformin. It has been my impression that a longer-acting GLP-1 has a less definitive impact on post-prandials. Future potential studies should use a short-acting GLP-1 receptor agonist, either exenatide or potentially lixisenatide, which is not on the market yet.

Q: Congratulations on the successful pilot. Did the same team titrate in both the exenatide and insulin groups? Second, how do you separate out the GLP-1 effect in a long-term study looking at outcomes?

A: Yes, we had an open label design, so the same investigators were titrating both insulin in the basal/bolus group and exenatide in the other group. It was up to three time a day exenatide. But it was not randomized – if glucose levels were not controlled on twice a day, they went to three times a day. The investigators were only allowed to use 20 mcg day dosing.

How do you separate GLP-1 and what we have learned on variability? We can’t. It’s a fundamental flaw of the study. I’ve gotten more philosophical about it. If we get better outcomes for patients, I’m not sure I really care.

Close Concerns Questions

Q: Is the magnitude in variability change enough to be clinically meaningful? To show a change in long-term outcomes?

Q: Would it be possible to do this study in type 1 diabetes, since there is more variability and hypoglycemia?

Q: Would switching to once-daily lixisenatide increase compliance with GLP-1, improve postprandial glucose levels, and influence the results in a larger study? 

Q: Would it be possible to run a similar trial using different therapies to tease out a larger difference in variability (e.g., comparing an SGLT-2 and liraglutide in one group vs. basal/bolus insulin in the other group)?

Q: Could the significant difference in weight have influenced the difference in variability? One could imagine the 10 lb weight loss made people more insulin sensitive?

Oral Presentations: The Clinical Impact of Advances in Continuous Glucose Monitoring

Continuous Glucose Monitoring (CGM) Use in Type 1 Diabetes: Database Analysis Shows Meaningful Improvements in A1c

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

Dr. Claudia Graham presented new database data on the efficacy of continuous glucose monitoring in improving glycemic control and reducing healthcare costs. The retrospective pre-post, matched case-control study analyzed insurance claims and lab results from 14 million enrollees in Optum’s national US database. The study compared patients with type 1 diabetes who were naïve to CGM and started Dexcom G4 for a nine-month period vs. individuals who used SMBG during the same period. Findings indicated that patients using CGM saw a significantly greater reduction in A1c (-0.48% vs. -0.24%, p = 0.03) relative to those in the SMBG arm. Notably, sub-analysis of the CGM group stratified by A1c showed that CGM-MDI users lowered their A1c significantly more than patients on CGM/pump (-0.57% vs. -0.34%, p=0.22). While Dr. Graham suggested that benefits of CGM may be similar for insulin injectors relative to pump users, we would caution that the difference in baseline A1c (injectors: 8.5%; pumpers: 7.9%) could be responsible for the trend as well. On the financial front, Dr. Graham noted that the inpatient hospital admission rate (all-cause) and emergency room visit rate fell a striking 42% and 17%, respectively, in the CGM arm vs. the control arm. These are significant potential cost savings associated with CGM and are consistent with the economics of CGM that Dr. Graham discussed at DTM 2014.

• Detailed analysis of CGM utilization stratified by A1c demonstrated that each CGM subgroup had greater reductions in A1c relative to the corresponding control group. The benefit was most apparent for those with an A1c >9.0%> See Table 1 below.

Table 1: Stratified A1c Results

Questions and Answers

Q: Can you talk about what the A1c change looked like over time? Did patients A1c drop immediately and then creep back up?

A: No, the findings were actually really counterintuitive. We thought people would have dropped immediately, but they did not.

Q: What was the baseline A1c in the MDI vs. pump groups? Those seem to be different patients. Patients who have not accepted technology vs. who are already on pumps.

A: The A1c in the MDI group was ~8.6%. The A1c in the pump group was ~8.0%.

Q: What do you think is driving change? The CGM or the patient using the CGM? In other words, is CGM a marker for more engaged patients?

A: Yes. There could be a bias in there.

Q: When you talk about hospitalization event rates, was that the change in rate or the actual rate?

A: It was the hospitalization rate a year before beginning CGM relative to the hospitalization rate a year after.

The Performance of PixoTest: A New Non-meter Blood Glucose Measurement System

Jerry Chieh-Hsiao Chen, MD (China Medical University, Beigang, Taiwan)

Dr. Jerry Chen provided an introduction to a new meter from iXensor – the PixoTest – that aims to minimize the burden of carrying cumbersome meters. The system consists of a small single-use smartphone attachment with an integrated lancet device – see a very cool video of the system here. Users prick their finger using one component of the device, and transfer the blood to a chamber on the device that sits over the phone’s front-facing camera. The accompanying mobile app uses the camera and a “colorimetric analysis” to determine the concentration of glucose with the aid of a reagent in the device. The chemistry is still glucose-oxidase-based. Dr. Chen summarized the results of two clinical trials (n=100 and n=118) that have demonstrated “excellent” accuracy - 97% of results within 15% of capillary YSI. The results were used to attain a CE Mark in August 2014, though the meter does not seem to be available on the EU market. The PixoTest does need roughly six times the blood volume (2 µL) of other US meters (~0.3 µL). The all-in-one form factor looks very encouraging, and Dr. Chen reported that feasibility testing thus far has been positive. As usual, questions abound for the glucose monitoring startup related to cost and manufacturing strips at scale.

Questions and Answers

Q: You used YSI as the reference. Why?

A: The regulatory policy asks for fingerstick testing vs. YSI.

Q: I applaud the interest in making meters usable for patients. One issue for lancets is the varying needs of patients. Does the lancet have any customization in terms of depth of penetration?

A: The lancet is a bit longer than multi-use lancet device options, but uses the smallest needles. We found that the reported pain score from users was slightly less than it was for other devices on the market.

Q: Was there ever a failure to get enough blood? How often did that happen?

A: If there’s not enough blood, the meter will show an error. It didn’t happen very often; less than 10% of the time for inexperienced users.

Q: Most sensors are chemical, not optical, and use glucose dehydrogenase, not glucose oxidase. What is the volume being measured? 

A: The volume is 2 microliters. It is glucose oxidase. We can also test cholesterol in arteries.

Q: That’s 10x more volume than a typical strip requires. That’s got to be painful.

A: Although the volume is greater, it’s still relatively small. We saw a higher adoption rate since users did not have to buy a meter.

Comment: I hope you continue work on some of these issues being raised from the audience. My pediatric patients express extreme frustration that this type of product doesn’t exist. They never lose their phones, but they lose their meters.

Dexcom G4 Platinum and Medtronic Enlite Glucose Sensors Perform Equally Well during Exercise in Patients with Type 1 Diabetes

Nadine Taleb, MD (McGill University, Montréal, Canada)

Dr. Nadine Taleb summarized new data showing that the Dexcom G4 Platinum and Medtronic Enlite continuous glucose monitors perform equally well during exercise in patients with type 1 diabetes (n=12). The very small study sought to assess the effect of lag on accuracy by comparing sensors both prior and during times of rapidly changing glucose. Findings indicated that there was no difference in sensor bias or median absolute relative differences between the two sensors at rest or during exercise – see table 1 below. Detailed analyses, however, did reveal statistically significant changes in accuracy during exercise compared to rest within each sensor individually, as median ARD fell from 16.5% to 9.7% for the G4 Platinum and from 14.2% to 10.6% for the Enlite (both p < 0.001). Questioners criticized the use of median ARD, which obscures outliers – we would agree that the metric is misleading relative to mean ARD. Dr. Taleb concluded that both sensors had comparable accuracy during times of rapidly changing glucose, but were less accurate during exercise compared to rest. We’re always glad to see head-to-head data, but would note that two larger and more rigorous studies (Russell, ADA 2013; Kropff, ATTD 2014) gave the accuracy edge to the G4 Platinum.

• The study assessed the accuracy of the Dexcom G4 Platinum and Medtronic Enlite prior and during exercise in 12 subjects (age 39; A1c 8.0%). Sensors were inserted simultaneously 24 hours before the study began and were calibrated to the same glucose values three times a day. Plasma glucose values measured with a YSI analyzer served as a reference. Measurements were taken every 30 minutes at rest for 150 minutes and every ten minutes during exercise (60 minutes) and recovery (30 minutes). There were 186 paired points during rest and 245 paired points during exercise.
• Findings indicated that there was no difference in sensor bias, median absolute relative differences, or the percentage meeting the 2013 ISO criteria between the two sensors at rest or during exercise – see table 1 below.

Table 1: G4 vs. Enlite Accuracy During Exercise and At Rest

• Unsurprisingly, each sensor’s individual accuracy worsened during exercise. However, the number of points falling in Zones A and B of the Clarke Error Grid did not change, suggesting that the inaccuracy comes from points moving from Zone A to Zone B. Dr. Taleb suggested that this difference is not clinically accurate, though during Q&A, this conclusion came up as a point of debate. Dr. Barry Ginsburg argued that, “zone B accuracy was acceptable in 1998 - I don’t think that’s acceptable now.” We would agree that Zone B is a misleading metric and does not reflect the true promise of CGM – moving forward, we would prefer to see data separated out between Zones A + B.

Questions and Answers

Dr. Howard Wolpert: In practice, because of problems with lag, many patients rely on the rate of change alert on CGMs. Was there a difference in how quickly the alerts went off? Which one went off first?

A: We did not look at that.

Dr. Barry Ginsburg: The idea that that number - 95% - is acceptable comes from 1998. Zone B accuracy was acceptable in 1998. I don’t think that’s acceptable now. In Zone B, you could be at a glucose of 71 mg/dl and you’re sensor could read 171 mg/dl.

A: I totally agree. It definitely affects long-term control.

Q: Where was the insertion site?

A: We inserted the sensors into the abdomen and the back. We kept the two sensors at a distance from one another and we always inserted them at the same time.

Q: Do believe you had enough glucose variability outside of the euglycemic range to challenge the sensors during the study?

A: They had a big range of glucoses especially during exercise.

Q: You have reported median ARD, but that is resistant to outliers. That’s like if you have you’re head in the freezer and feet in the oven, then with the median your belt buckle would feel fine. However, the mean takes outliers into account.

A: The data was not normally distributed. That’s why we didn’t use mean average relative difference.

Comment: That’s actually exactly why you want to use a mean. Those outliers have a clinical impact.

The Glycemia Fluctuation Index (GFI) and Coefficient of Fluctuation (CF): New Indices of Glucose Variability in Diabetic Patients

Jean-Pierre Le Floch, MD (Clinique de Villecresnes, Strasbourg, France)

Dr. Jean-Pierre Le Floch summarized the findings of a recent study that aimed to assess the relative value of various indices of glycemic variability: standard deviation (SD), coefficient of variation (CV), interquartile range (IQR), mean amplitude of glycemic excursions (MAGE), mean of absolute glucose change (MAG), and glycemia fluctuation index (GFI). Dr. Le Floch highlighted that relying solely on A1c is insufficient, particularly as next-gen technologies come to market that improve time-in-range but do not impact A1c. Dr. Le Floch seemed particularly excited about GFI, a newer metric that is similar to MAG but places more weight on consecutive glucose excursions that trend in opposite directions (e.g., a high preceded by a low). Results bore out his confidence in the metric, suggesting that GFI was, indeed, one of the most useful metrics of glycemic variability when considered in the context of the ratio of GFI to mean glucose. We continue to wonder if the diabetes community will agree on a glycemic variability metric –the merits of different statistical measures have been debated for years, though there does not seem to be consensus on which is best to use. We applaud research into more complicated metrics, but hope researchers also think about integrating glycemic variability feedback into clinical practice. Ultimately, the goal is to accurately characterize variability in a way that is understandable for patients and helps diagnose problem areas. While time-in-range does not necessarily quantify variability, we believe it is a useful metric for characterizing glycemic control at a more nuanced level vs. A1c.     

Questions and Answers

Q: How might this metric be applied in clinical practice?

A: This could be used for patients on CGM and even meters. A1c is not sufficient as a metric; we should be using indices of variability as well.

Q: MAG assumes that larger fluctuations are worse than smaller fluctuations and thus, you do some squaring. However, how can we be sure large fluctuations are worse than smaller ones?

A: It’s not clear, and it is difficult to prove one way or the other. We’ll need to wait for clinical data.

Q: Patients feel better with improvements in variability, so I think it’s important. Did you compare GFI with CONGA [continuous overall net glycemic action]?

A: CONGA wasn’t used in this study, but it could be used. This study was not really trying to compare indices, but trying to tease out when and how to use different indices.

Q: Indices like standard deviation are used for measuring overall variability whereas indices like MAG and GFI look at high frequency variability. You’re really looking at high frequency versus low/medium frequency fluctuations.

A: We think it’s important to describe the drivers of quick variability.

Comment: If we’re evaluating measures of variability, we need to have utility functions that take into account when variability is clinically significant. For example, a 50 mg/dl variance at 100 mg/dl is clinically significant whereas a 50 mg/dl difference at 400 mg/dl is not. If a patient is at 70 mg/dl and going up, that is very different from 70 mg/dl and dropping.

Comment: If you go back 15-20 years, we talked all the time about beta cell oscillations. If you lost an oscillation, it was highly predictive of developing type 2 diabetes. We were talking about them as positive, but now variability is considered universally bad. I don’t think we fully understand what is bad and good with variability. A high A1c is bad, but I’m not sure we have the same evidence for glycemic variability.

Comment: I take issue with that last comment. Glycemic variability impacts quality of life unequivocally. This is important to measure, because A1c doesn’t capture variability at all.

Using Technology for Better Outcomes

Making Sense of Technology

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

"How can we use technology to drive better outcomes?” Our very own Ms. Kelly Close tackled this compelling question with a twofold answer: (i) rethinking how studies are done; and (ii) changing how we design and deploy technology. Below, we summarize the major points of her presentation. See Kelly’s slides here.

• “A three-month average cannot tell the full story.” Kelly stressed that there is a profound irony in the way outcomes are measured in diabetes. She emphasized that our community relies on a metric (A1c) that tells us nothing about hypoglycemia. Of even more concern, she pointed out that when a therapy actually reduces hypoglycemia, A1c goes up. Thus, to understand if technologies are actually working, she stressed that we need to get beyond our A1c-centric approach to evaluating therapies.
  
  • Kelly noted that the use of 24/7 CGM in trials would help accurately characterize the benefit of technologies. For context, she reminded the audience that a patient with low variability and perfect time-in-range can have the same A1c was a patient with medium (or high) variability with prolonged hypoglycemia and hyperglycemia. Given the accuracy and convenience of currently available technologies – particularly Dexcom’s G4 Platinum CGM and Abbott’s FreeStyle Libre in the EU – we can get truly meaningful and accurate 24/7 glycemic data that can help the FDA and other regulatory agencies make more information decisions.
• Kelly also suggested that studies should enroll broader, real-world populations of patients. She highlighted the paradox of trial inclusion criteria, in which the patients most likely to benefit from an intervention (e.g., those with severe hypoglycemia, diabetes distress) are often excluded from the trial. Said Ms. Close, “We are underselling the outcomes and value of available technologies.”
• The diabetes community needs to move beyond a drug-based randomized-control model for testing diabetes devices. Ms. Close pointed out that this model applies a far-too-simple pharmacological rationale to test the efficacy of a complex treatment: a single treatment is delivered to subjects and it either works or does not work with respect to a well-defined outcome (usually A1c). In her view, the problem with devices is that – unlike medications – they are not uniform interventions administered in an identical manner across all patients and settings. The utility varies considerably by the clinical question addressed, the recommended frequency and timing of use, the expertise of patients regarding its use, and the involvement and knowledge of clinicians in interpreting and responding to SMBG data.
  
  • In Kelly’s view, trials need to be geared to answer more real-world questions: What is the optimal patient/provider education? What does it cost to deliver the intervention? How does it fit into clinical workflow? Can it scale? What type of patient is this intervention best suited for? These questions tackle larger and more important issues than just whether presence or absence of a single device changes A1c.
• In the second half of Kelly’s presentation, she shared her vision for how a combination of population tracking, positive feedback, optimized training, and automated software will lead to better outcomes in the very near future:
  
  • Population tracking: Kelly suggested that there is huge value to tracking patients at a population level. Such tracking can help avoid some of the catastrophic events that drive healthcare costs - such as visits to the emergency room for severe hypoglycemia – but also reduce the burden on patients who do not need to be seen as often. That sad, Kelly acknowledged that the conversation around population tracking is particularly nuanced - after all, providers have limited time and resources to care with patients with diabetes. Acknowledging the value here is important, but Kelly stressed that automated systems that help providers sift through big data is what the field still desperately needs.
  • Positive Feedback - “Let’s celebrate who is doing well!” Kelly stressed that so much of the way diabetes is presented feels defeating – e.g., an out-of-range number staring you in the face. In thinking about technology design, Kelly emphasized that there is huge potential to think about the use of more encouraging, educational, and empowering language. We have summarized some of her examples in the table below:
  • Optimized training: Kelly acknowledged that there is much complexity to device use to begin with – from counting carbs to setting insulin sensitivity factors to uncontrollable variables (e.g., is the pump cannula blocked?) – that we practically set patients and providers up to fail in a lot of ways. In this sense, she stressed that there is so much more the community can do to better train patients and providers to set them up for success. One particular avenue to explore is looking at “bright spots” – deeply understanding expert users to inform best use and education practices.
  • Help with insulin dosing: While collecting data is valuable, Kelly suggested that translating that data into actionable decision-making is what the field still needs. After all, patients with diabetes are asked every day to self-administer a potentially deadly drug that is affected by a huge number of variables – thus, the challenge is providing patients with automated insulin-dosing algorithms that can reduce this burden and uncertainty.

Questions and Answers

Q: Is there a big role for patient advocacy in driving outcomes?

A: Earlier this year, patients were invited to a manufacturer to talk about their preclinical work. We were asked a lot of questions and were able to bring a lot of perspective. I think that’s one really positive way patient voices can be incorporated into the process. As clinicians and educators, I also think you all need to lobby for better reimbursement. You guys need to own that you deserve better resources for working with us. This is having an impact on younger doctors too; they are coming out of school and have to pay off their debt. I don’t know who’s going to want to be an endocrinologist. Better reimbursement would really help though.

Q: Do you have any advice for providers trying to help patients decide which pump to use, aside from obviously individualizing the device?

A: I would advocate using social media to listen to what other people are saying and doing. Also, take time to look at the pumps. It should not be an incredibly fast decision necessarily. (A point added afterward by Ms. Close - Sometimes formularies will choose a pump for a patient!)

Q: My impression is that diabetes is more of a psychological disease than it is a medical disease. One of the biggest problems is getting better reimbursement for mental care. What do you think about that?

A: Mental health is so incredibly underfunded in our country – we should be paying far more attention to this. You’re standing up and saying this, and you’re probably not even thinking about the amount of time you spend working on this. I have a child who has special needs and we recently took him to a dentist for children with significant special needs. The dentist sat him down and then looked at us and said, “Tell me your story.” I was shocked. I nearly wanted to say, “You don’t have the time! You’re not getting reimbursed for this!” But it was so nice. That’s just an example. Things are backwards in society if we can’t reimburse HCPs to spend more time on behavioral needs and just asking patients how they are. Mental health is underfunded and this absolutely must be changed.

Posters

CMS Competitive Bidding Program Disrupted Access to Diabetes Supplies with Resultant Increased Mortality

G Puckrein, F Zangeneh, G Nunlee-Bland, L Xu, C Parkin, J Davidson

This late-breaking poster found an increase in mortality associated with CMS’ competitive bidding program due to reduced access to SMBG diabetes testing supplies. The study investigated CMS data from 2009-2012, looking at testing supply access and mortality in all Medicare beneficiaries with an insulin prescription within the nine test markets (n=43,939) and all non-test markets (n=485,688). Findings demonstrated that the acquisition of SMBG supplies was disrupted in the test site populations – specifically, the percentage of test vs. non-test beneficiaries with no SMBG record increased (17% vs. 1%). Further analyses also showed that patients in test sites had more than twice as many inpatient admissions (982 admissions vs. 460) and more than double the associated costs ($10.7 million vs. $4.7 million) compared to the non-test site population, which translated into almost twice as many deaths (102 deaths vs. 60). As a reminder, these results differ greatly from CMS’ April 2012 report on adverse outcomes associated with competitive bidding, which suggested that there was no disruption of access to supplies and no negative healthcare consequences associated with the program. Ultimately, the results raise red flags for what is already a heavily scrutinized program. With another round of competitive bidding underway, we believe the findings underscore the need for more careful monitoring of the program moving forward.

Incidence of Hypoglycemia Overtreatment in the SHARE Real Life Use Population

K Narakmura, T Walker, J Leach, L Bohnett, J Valdes, Andrew B

A late-breaking poster from Dexcom has demonstrated that the real-world incidence of hypoglycemia (glucose < 70mg/dl) is frequent and that consequent overtreatment (glucose > 180 mg/dl) is not at all uncommon. The retrospective study analyzed 50,000 hypoglycemic events identified in 1,177 users of the company’s SHARE platform and found a striking hypoglycemia rate of 0.96 event/day/patient. Of more concern, the proportion of hypoglycemic events followed by rebound hyperglycemia was 18% within 60 minutes and 26% within 90 minutes. Talk about glycemic variability! It certainly made us look forward to the results of FLAT-SUGAR (see above!) to know that at least one-fourth of hypoglycemic events in this real-world population were followed by hyperglycemia. Notably, rebound hyperglycemia occurred more frequently during the day than the night (p < 0.001), which may be partially related to meals. Ultimately, the findings are a strong reminder of the value of cloud-based data platforms in facilitating population tracking and broader learnings about real-world diabetes management.

The Performance of an Orthogonally Redundant Glucose Sensor Compared with a Simply Redundant Electrochemical Glucose Sensor in Adults with Type 1 Diabetes

S McAuley, TT Dang, JC Horsburgh, A Bansal, GM Ward, AJ Jenkins, RJ MAcisaac, RV Shah, DN O’Neal

The latest human data from Medtronic's orthogonally redundant sensor (electrochemical + optical) demonstrated very little marginal benefit relative to electrochemical alone (n=21). The seven-day study compared the redundant sensor to a single electrochemical sensor (“Enlite” – it didn’t specify the generation) and involved both in-clinic (vs. YSI) and home portions (vs. SMBG). For the full seven-day period, accuracy improved very marginally with the redundant sensor (from 11.6% to 10.3%). The biggest advantage of redundancy came in the first 48 hours after insertion – mean average relative difference (MARD) fell from 16.8% with the single electrochemical sensor to 10.8% with the orthogonally redundant sensor configuration on Day 1 and from 13.7% to 10.2% on Day 2. The accuracy was also generated on four fingerstick calibrations per day, and the sensors were “post-processed prospectively.” On a slight positive note, there was a trend toward increase durability with mean redundant sensor lifetime increasing from 5.4 to 6.5 days and mean display time increasing from 74% to 87%. We have followed this project for some time, and have never been impressed with the marginal accuracy/reliability benefit of the optical + electrochemical redundancy, especially when cost and form factor are bundled in. Future work will return to animal models and focus on the continued development of the optical sensor. As a reminder, we saw the first human data from Medtronic’s orthogonally redundant sensor at ATTD 2015.

• This seven-day study included 12 women and nine men with type 1 diabetes (baseline A1c = 7.3%). Patients had three study visits over the course of the week – insertion occurred on day one, an in-clinic meal test (vs. YSI) happened on day four, and sensors were removed on day eight. Patients were asked to perform four or more SMBGs per day. The redundant electrochemical/optical sensors were inserted together at a single site and a recorder was worn on top of the sensor, which appeared about the size of a second-gen Insulet OmniPod. Notably, the redundant sensor was inserted at a 45° angle and the electrochemical sensor was inserted at a 90° degree angle.
• Redundant sensor accuracy showed a significant insertion benefit immediately following sensor insertion though this benefit waned over the course of the week – see table 1 below. The early results are solid, though the system is still a work in progress.

Table 1: Performance of redundant sensor vs. electrochemical sensor

• Orthogonal sensor durability tended to show improvement over the electrochemical sensor - mean sensor lifetime increased from 5.4 to 6.5 days and mean sensor display time increased 74% to 87%. However, we wonder whether the different may relate to the difference angles of insertion or, as the researchers note, to the potential of the optical filament to act as a physical support for the electrochemical filament and enhance its performance.

Extended Wear of a Next Generation of CGM Sensor

L Bohnett, K Nakamura, L Jepson, J Leach

A Dexcom poster provided our first look at human durability data from the company’s Gen 6 sensor, suggesting that extended wear is indeed feasible over 10 days. The two-week trial enrolled 84 patients (79 with type 1 diabetes; 7 with type 2 diabetes) and compared CGM responses to SMBG measurements in real-time. The early findings are positive, though the sensor survival to day 10 was 79% in the patient population that completed the study – this is lower than we would have expected based on our experience wearing the G4 Platinum. Only one subject presented with a device-related adverse event (mild erythema and local infection), suggesting that the extended wear is feasible from a safety perspective, too. Of course, the key will be accuracy data, and this poster did not share any. As a reminder, we saw accuracy data on Gen 6 at JPM 2015 - a MARD of 12% over 10 days. Dexcom expects to conduct a pre-pivotal trial of Gen 6 this summer, with plans for a 1H17 commercial launch.

Examining the Role of Continuous Glucose Monitoring (CGM) in Noninsulin Treated Type 2 Diabetes

L Young, M Duclos, A Marquis, Y Teng, S Davis, B Bode, J Buse

This two-center pilot study examined the effect of periodic masked continuous glucose monitoring in patients with non-insulin treated (NIT) type 2 diabetes. The study randomized 35 NIT patients in sub-optimal control (A1c 7.5-9.0%) to either SMBG 1-3 times/day or blinded CGM for six months. Results were somewhat disappointing, indicating no significant difference between the arms – those on CGM exhibited a 0.8% reduction in A1c from a 7.7% baseline vs. a 0.7% reduction for those on SMBG from an 8.2% baseline. We wonder how much of this was attributable to the study itself: blinded CGM benefits providers with more comprehensive data, though we imagine the marginal benefit of that data vs. SMBG was fairly minimal in the non-insulin-treated population. We continue to hope for further assessment of CGM in type 2 patients, though we imagine the big clinical impact in non-insulin users will come from real-time CGM. Harkening back to Dr. Bob Vigersky’s trial of CGM in non-insulin-using type 2s, episodic short term use of real-time CGM did drive a significant and lasting A1c benefit.

• The study randomized 35 NIT type 2 patients in sub-optimal control (A1c 7.5-9.0%) to either SMBG 1-3 times/day or CGM for six months. Patients met with study staff at six-week intervals to review SMBG or CGM results from the prior six days. Therapy changes were based upon standardized treatment algorithms.
• Within each group, glycemic control improved, though differences between the groups were not statistically significant. In the CGM group, A1c improved from 7.7% at baseline to 6.9% at 90 days; this decline was maintained at 180 days. The SMBG group saw a reduction in A1c from 8.2% to 7.2% at 90 days, though there was a bit of a rebound at 180 days (A1c: 7.5%). None of these differences were statistically significant.

Company Updates

Glooko Partners with Insulet, Dexcom, and Medtronic

• Glooko announced pump and CGM integration partnerships with Dexcom and Insulet. The next generation Glooko platform (iPhone/Android apps and a web dashboard) will roll out this summer (super fast!) and include access to both Dexcom CGM and Insulet OmniPod pump data. The news was not a major surprise, as the March news of Glooko’s $16.5 million in funding first announced CGM/pump integration plans (as a reminder, Medtronic was a first-time investor in the Series B round). Still, it is great from a patient and clinician perspective, since downloading continues to be a hassle for many. Our detailed report has some screenshots of the new dashboards, which look excellent. Patients will be able to download OmniPod and Dexcom data to Glooko via a micro-to-micro USB OTG cable on Android devices, and providers will be able to download in the clinic using a Glooko kiosk; Dexcom Share receiver users will also be able to send their data straight to Glooko’s platform seamlessly via Apple’s HealthKit.
• As expected following its investment in March, Medtronic and Glooko also announced a data partnership during ADA. The move is a notable one, as Medtronic data has historically sat in its proprietary CareLink system. There is no timing on when Medtronic data will be able to download to Glooko. We’re glad to see Medtronic opening up to sharing its data, something naysayers in the past didn’t think was possible. For Medtronic, this news continues a slew of new partnerships in the new Service and Solutions business (led by powerhouse Annette Bruls), including the acquisition of Diabeter (April) and a partnership with IBM’s Watson Health (April).
• These announcements add to Glooko’s impressive device compatibility with 30+ BGMs, popular activity trackers, and blood pressure and weight measurement devices.

Medtronic Announces MiniMed Connect for Remote Monitoring Pump/CGM Data

Medtronic announced FDA Clearance of MiniMed Connect, a keychain device that sends pump/CGM data via Bluetooth to a smartphone app and enables remote monitoring for caregivers – see a picture here. Launch is slated for this fall for MiniMed 530G and Revel users; it will be $199 (cash pay), and interested users can sign up to be notified at www.medtronicdiabetes.com/minimed-connect (strong marketing out of the gate: “Pump and CGM data on your mobile device?”). Initially, the MiniMed Connect paired smartphone app will only be available on Apple’s iOS, though Medtronic announced a separate partnership with Samsung (see below) to build out an Android version and apps optimized for Samsung devices. MiniMed Connect allows caregivers to remotely monitor patients on any product with an internet connection via a web display, a nice device-agnostic approach (similar to the Nightscout setup). Caregivers can also receive text messages for uncleared pump alarms or when sensor glucose levels are too high or too low – we imagine parents will absolutely love this.

• The MiniMed Connect form factor looked excellent in the Exhibit Hall – the tiny device is about the size of a car key fob, rechargeable via USB, and will receive the data from the pump via RF and send it to the nearby phone via Bluetooth.
• Data will also go to CareLink automatically (CGM data every five minutes, pump data daily), saving HCPs downloading hassle.
• While Medtronic has recently lagged behind Dexcom on the connectivity front, this gives the company a more competitive offering to Dexcom’s Share Receiver and associated Share system. [However, Dexcom will again move ahead of Medtronic by the end of year, if Gen 5 is approved as expected. Medtronic’s similar Guardian Mobile system is still in a pivotal trial, expected to wrap up in July.]

-- by Adam Brown, Varun Iyengar, Dana Lewis, and Kelly Close