February 15-18; Paris, France; Day #2; Highlights – Draft

Executive Highlights

Bonjour de France pour le jour #2 de la conferénce ATTD! Today was a whirlwind – data here, data there, data everywhere! Some of it, we’d been waiting a long time for (Dexcom DIaMonD type 2 cohort and pump extension phase, REPLACE-BG), while we were surprised by others (Dr. Irl Hirsch’s cost-effectiveness analysis of CGM in type 2, comments on the 670G pivotal). The exhibit hall also JUST opened and we took our first trip there, where we got to catch up with the Unomedical team and hear about their latest infusion set work (see more on the pipeline).

Scroll down for those highlights and more. In case you missed yesterday’s ATTD Day #1 report, see it here, and check out our preview for a look at what’s in store for Friday and Saturday!

Top Highlights

1. Type 2 data from Dexcom’s DIaMonD study testing CGM in MDIs demonstrated a 0.3% A1c advantage for CGM (n=79) over SMBG (n=79) at 24 weeks: -0.8% with CGM vs. -0.5% with SMBG (baseline: 8.5%; p=0.02). A Dexcom poster also shared the second phase of the DIaMonD study in type 1, where patients on the G4 CGM were further randomized to continue on MDI (n=38) or switch to an OmniPod (n=37); the OmniPod group had an improvement of 78 minutes per day in time-in-range vs. a drop of 17 minutes per day for the MDI group (p=0.02). We look forward to hearing more on time in range – or, as Dr. Aaron Kowalski pointed out earlier this week, it’s really time in ranges! Yes, we’d love to know not only how much time was “in” range, but also how much was above and below.

2. Unomedical demoed its all-in-one, fully disposable, hidden needle insertion set device, sharing that it will launch “after” this summer and first come to market in an exclusive partnership with a Luer Lock pump. A novel catheter (Lantern) is expected to launch in 3Q17, adding several slits along the side that allow insulin to flow out of multiple places (in case of occlusion, bending; watch this video). We’re happy to see further work in this area – there is lots of room for patients to have better experiences and we salute work from multiple organizations on this front.

3. Results from REPLACE-BG were refreshingly unremarkable: the 226-patient, 26-week randomized trial comparing use of Dexcom’s G4 CGM with and without confirmatory fingersticks showed near-identical outcomes, including no significant difference in time-in-range, hypoglycemia, hyperglycemia, A1c, mean glucose, or coefficient of variation. A win for insulin dosing off CGM, that is for sure, and for Dexcom.

4. The ever-engaging Dr. Irl Hirsch made a strong economic case for the use of intermittent RT-CGM in people with type 2 diabetes – in terms of A1c reduction, CGM is significantly less expensive than novel oral and injectable anti-hyperglycemic agents. Of course, the right technology leading to optimal therapeutic treatment including diet, exercise, and drugs sounds like the best in order of reducing long-term complications.

5. In discussing the MiniMed 670G pivotal trial, the renowned Dr. Rich Bergenstal pointed out that some participants needed a more aggressive insulin:carb ratio and shorter duration of insulin action to help with the “post-post prandial period.” We also heard about the “safe basal timeout.”

6. Medtronic’s very smart Dr. Huzefa Neemuchwala ran attendees through the Sugar.IQ with Watson app, sharing that a “larger preview” to “a larger audience” will occur on February 28. Boy can we not wait.

7. In the conference’s opening talk on “Big Data,” a fired-up Dr. Aaron Kowalski highlighted that we rarely use ANY data in diabetes – and the technology exists to solve this problem now. He shared a great “near-term” list of goals and questions in diabetes data too – we loved it.

8. In a most valuable review of predictive low glucose suspend technology, Dr. Thomas Danne shared data from the US pivotal trial of the MiniMed 640G with SmartGuard (Buckingham et al. DTT, in press) and that patients who don’t override the algorithm tend to have better glucose profiles. As a reminder, this product was skipped in the US in favor of the 670G.

9. Citing a lack of conclusive RCTs for or against pump use in type 2 patients, Dr. John Pickup (King’s College, London, UK) walked the audience through his group’s individual patient data meta-analysis (Diabetes Care, in press), which found that pumps are best and likely most cost effective in type 2 patients with elevated A1c and insulin doses even after insulin optimization.

10. Dexcom SVP of R&D Mr. Jake Leach excitedly reviewed Dexcom’s pipeline: The Android version of G5 has launched internationally, G6 is currently in a pivotal trial and will be submitted for regulatory approval upon completion (for a 2018 launch, per the company’s 3Q16 call), and factory calibration is in the works (the first G6 product with Verily, along with G7).

11. Dr. Ralph Ziegler provided an overview of the in-progress IDS Comparative Evaluation Study (sponsored by Roche), which investigates the dosing accuracy and functional differences between various insulin pump options. Interestingly, different pump/infusion set combinations emerged as having the best relative accuracy (i.e., smallest deviation from target dose) at different bolus doses, suggesting that no one pump is absolutely the most accurate.

12. Onduo’s Head of Product and Technology Mr. Andrew DiMichele (filling in for CEO Dr. Josh Riff at a Sanofi-sponsored symposium) announced that the joint Sanofi/Verily venture should have a product to show the world in ~one year. He detailed how the newly-formed Onduo is designed to beat the common challenges blocking other digital health companies.

13. Dr. Bruce Bode presented positive retrospective glycemic data (February 2013-November 2016), as well as cost-savings, from inpatient use of the Glucommander insulin titration software. Patients (n=5,718) admitted with hyperglycemia to one of seven hospitals arrived at their prescribed target blood glucose (100-140, 120-160, or 140-180 mg/dl) in 0.8 days from a starting average of 261.7 mg/dl. Once at target, time in range was between 65%-70% and hypoglycemia was very unlikely. The financial proposition associated with Glucommander is compelling, with an estimated $3+ million in potential savings per 250-bed hospital per year.

Top Highlights

1. Dexcom’s DIaMonD Data in type 2 Diabetes (-0.3% A1c Advantage vs. SMBG) + Phase 2 Shows OmniPod Adds One More Hr/Day in range

Dr. Rich Bergenstal presented the first type 2 data from Dexcom’s DIaMonD study testing CGM in MDIs, demonstrating a statistically significant 0.3% A1c advantage for CGM (n=79) over SMBG (n=79) at 24 weeks: -0.8% with CGM vs. -0.5% with SMBG, both from a baseline of 8.5% (p=0.02). The benefit of CGM rose with a higher baseline A1c – those starting at >9.0% saw a 1.4% reduction. At 24 weeks, these type 2 CGM users were spending ~48 more minutes per day in 70-180 mg/dl (a 6% improvement from baseline), while the SMBG users spent 9 fewer minutes in range per day (a 1% decline) (p=0.01). There was no significant difference in hypoglycemia (very low in both groups), meaning the improvement came from spending less time >180 mg/dl. Similar to the type 1 data (ADA 2016, JAMA 2017), CGM adherence was very strong: 93% of the type 2 cohort was using CGM >6 days per week at six months. Patients in this cohort were very typical of the type 2 population, with a mean age of 60 years, a median type 2 diabetes duration of 17 years, a mean of three fingersticks per day, and a mean BMI of 36 kg/m². We’re glad to see this technology being tested in a broader, real-world population. Interestingly, Dr. Bergenstal noted that medications didn’t really change throughout the study, so the impact of CGM was on lifestyle and behavior. Skeptics might argue these results are underwhelming – a 0.3% A1c advantage from a high baseline – but this is a tough population and the study really worked to minimize clinical encounters. Dr. Bergenstal emphasized that there was “not a lot of hand holding” in this study (a point also shared at ADA), and he is eager to explore more coaching and giving patients more advice on adjusting insulin. We certainly agree, just as we noted with Abbott’s REPLACE study in type 2 diabetes at ATTD last year. Overall, we’re very glad to see another major study of CGM in type 2 and wonder what can be learned from this data for future studies and product development. We compare this study to the type 1 results and REPLACE in the detailed commentary below.

• Meanwhile, a Dexcom poster shared the second phase of the DIaMonD study in type 1 diabetes, where patients on the G4 CGM were further randomized to continue on MDI (n=38) or switch to an OmniPod (n=37). The OmniPod group won handily on the primary endpoint of time spent in 70-180 mg/dl (weeks 5-28 pooled): an improvement of 78 minutes per day from baseline vs. a drop of 17 minutes per day for the MDI group (p=0.02). Time in hyperglycemia (>180) was also highly in favor of the OmniPod group: an improvement of -47 minutes per day vs. +59 minutes per day when continuing on MDI (p=0.009). On the other hand, time in hypoglycemia (<70 mg/dl) actually favored the MDI group, who spent 9 fewer minutes per day low compared to 15 more minutes per day in the OmniPod group (p<0.001). The difference in A1c was not statistically significant: the pump group saw a +0.3% change in A1c vs. +0.1% in the MDI group (baseline: 7.6%; p=0.32). The poster emphasizes that the A1c results from the six-month phase 1 of DIaMonD were sustained out to one year in this extension, a very positive finding indeed. As expected, bolus frequency increased in the pump group by +0.6/day vs. -0.1/day in MDI, which the authors tie to the increase in hypoglycemia in the pump group – it’s an aside in the poster’s conclusion, but a definite possibility (since pumps makes bolusing so easy, stacking insulin is also easy, particularly when on CGM). CGM adherence remained excellent in the study, with 96% using it >6 days per week at six months. We see these data as a win for CGM (sustained positive outcomes at one year) and encouraging evidence that adding a pump on top of MDI+CGM brings some further value (over an hour more per day in range). Automated insulin delivery could be a killer app for both technologies, though it’s hard to say who will benefit the most and whether the ROI will be there vs. MDI+CGM.
• Compared to the type 2 DIaMonD results, the type 1 cohort of DIaMonD saw larger improvements with CGM: a 0.6% A1c advantage (-1% vs. -0.4%), 76 more minutes per day in range (a 12% improvement from baseline), and 22 fewer minutes per day <70 mg/dl (a 34% improvement).
• The type 2 results also tell a different story than Abbott’s six-month REPLACE study comparing FreeStyle Libre to SMBG in type 2s with a baseline A1c of 8.8% (ATTD 2016). That study disappointingly missed its primary endpoint – similar 0.3% A1c reductions with both SMBG and Libre. The most compelling takeaway in that trial was actually the hypoglycemia data, which improved markedly with FreeStyle Libre overall, overnight, and particularly for dangerous hypoglycemia (<55 mg/dl). 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. Taken together, we think REPLACE and DIaMonD show continuous glucose monitoring (either traditional or Flash) can drive meaningful improvements in hypoglycemia, time-in-range, and hyperglycemia in type 2s on insulin.

2. Unomedical’s New All-In-One, Hidden Needle Inserter To Launch After This Summer in Exclusive Luer Lock Partnership; New Catheter to Launch in 3Q17, Include Side Slits For Insulin Flow

In the exhibit hall today, Unomedical demoed its all-in-one, fully disposable, hidden needle insertion set device, sharing that it will launch “after” this summer and first come to market in an exclusive partnership with a Luer Lock pump. The company also has a novel catheter (Lantern) expected to launch in 3Q17, which includes several slits along the side that allow insulin to flow out of multiple places in the case of occlusion or bending (watch the cool one-minute video here). We got to see the all-in-one inserter device up close in the hall and were very, very impressed – inserting a set only requires peeling the tape, pulling the plastic tab at the top, pressing the button, and lifting the device off the body (see picture below; this red version is for Parkinson’s, and the pump version just has a different color). The fully disposable serter does not require cocking or loading, and insertion occurs in a microsecond without seeing the needle before or after – a big improvement over BD’s FlowSmart set insertion in our experience (though Adam didn’t get to try Unomedical’s on his body, so we can’t comment on pain). The exclusive partnership news was a surprise to us (Unomedical supplies all the tubed pump companies), and perhaps a response to BD’s exclusive partnership with Medtronic for FlowSmart. Unomedical told us it may expand to sell the all-in-one inserter to other companies, but for an initial period of time it will be an exclusive deal. We loved the form factor of the new inserter, though the team seemed far more excited about the new catheter, called “Lantern” – the video is worth a watch (picture below), showing how the slits expand to flare out and allow the insulin to flow out of many places. The team hopes it may even enable faster insulin absorption from a wider subcutaneous insulin depot, though this has to be studied. Lantern also seems like a response to BD’s two-hole FlowSmart catheter, though Unomedical has added wider side openings for more insulin flow. We are surprised Unomedical has not received any funding from JDRF for set innovation, as both of these novel approaches seem in line with the non-profit’s goals and recent funding of BD and Capillary Biomedical. The principal investigator of a Lantern study will be in Unomedical’s booth tomorrow at 2:30 pm. We’ll be interested to see real-world human data on this, especially whether kinking is more likely with the additional slits.

3. REPLACE-BG: Non-Adjunctive Use of Dexcom G4 = No difference in Time-in-Range, Hypo/Hyperglycemia, A1c Vs. Adjunctive Use

Jaeb’s Katrina Ruedy presented long-awaited, positive results from REPLACE-BG, a 226-patient randomized T1D Exchange trial comparing use of Dexcom’s G4 CGM with and without confirmatory fingersticks (n=77 for CGM+BGM vs. n=149 for CGM-only) over 26 weeks. As we expected, the groups had near-identical outcomes by study end, with no significant difference in time-in-range, hypoglycemia, hyperglycemia, A1c, mean glucose, or coefficient of variation – see the table below. The outcomes were also not significantly different in subgroups based on age, type 1 diabetes duration, and education level, a good sign that non-adjunctive CGM use is safe across the board. Protocol adherence was excellent, with 91% of the CGM-only group and 95% of the CGM+BGM group wearing the G4 for >6 days/week (mean: 6.7 and 6.8 days/week). Participants used the very accurate Contour Next BGM, with fingersticks totaling just 2.8/day in the CGM-only group (two for calibration plus an additional one here and there) vs. 5.4/day in the CGM+BGM group. There was one severe hypoglycemia event in the CGM+BGM group and zero in the CGM-only group. Ultimately, these RCT results show clear non-inferiority and confirm the simulations Dexcom presented at FDA last July prior to the non-adjunctive FDA approval in December. We include below the training materials that informed non-adjunctive use in this study, and we’ll be interested to see how Dexcom rolls this claim out in the US. The company’s webpage dexcom.com/fingersticks already shares some of these recommendations in very clear, succinct text and pictures. We salute the T1D Exchange for conducting this impressively large and rigorous study, and we wonder if other companies’ hopes to get non-adjunctive label claims (e.g., Abbott, Senseonics) may benefit from this data.

• Participants in the CGM-only (non-adjunctive) group were instructed to dose insulin and make management decisions based on the CGM glucose except in the following circumstances: for 12 hours after insertion of a new sensor; on a sick day; for four hours after taking acetaminophen; symptoms present suggestive of hypoglycemia but CGM glucose not low; 20 minutes after treating low CGM glucose if CGM glucose not rising; prior to giving insulin bolus when CGM glucose >250 mg/dl; fasting CGM glucose >300 mg/dl or >300 mg/dl for one hour.

4. Dr. Irl Hirsch Makes Cost-Effectiveness Case for Intermittent RT-CGM in Type 2

The ever-engaging Dr. Irl Hirsch made a strong economic case for the use of intermittent RT-CGM in people with type 2 diabetes. His arguments and calculations mostly stemmed from the landmark 2012 Vigersky et al. Diabetes Care paper which showed that, after 12 weeks (2 weeks on, one week off, repeat) of CGM use, A1c was 0.5% lower than in an SMBG group, and, 4o weeks later, the difference was maintained at 0.6% – an effect Dr. Hirsch dubs “CGM-behavioral memory.” So how cost-effective is CGM in these cases? Dr. Hirsch outlined two theoretical scenarios: In the “base case,” the assumption is that patients don’t use CGM after year one, so the effect of the treatment in Vigersky et al. are only applied for one year. In the “refresher scenario,” the assumption is that the patient uses CGM again the next year, so the glycemic benefits are maintained for an additional year. A later study from the Vigersky group (Fonda et al.) calculated that, in the base case of CGM use, the incremental cost effectiveness per life-year gained is $6,293, while the incremental cost effectiveness per QALY gained is $8,893. In the refresher scenario, these numbers come out to $9,319 and $13,030, respectively. These numbers are very, very strong given that a benchmark is often $50,000 per QALY – CGM QALY’s in this case are so cheap because the device is only used for eight weeks, but its impact lasts at least the year (assuming Vigersky’s data is generalizable).

• How does CGM stack up to other diabetes therapies? Dr. Hirsch calculated that eight weeks of CGM therapy spread out over a year in the Vigersky study would cost ~$1,721. Compare this to sitagliptin ($4,560/year; 0.6% A1c reduction), liraglutide ($9,563/year; 1.1% A1c reduction), exenatide ER ($7,728/year; 0.9% A1c reduction), canagliflozin ($4,920/year; 0.9% A1c reduction), and pioglitazone ($168/year; 1.6% A1c reduction, but at high baseline of 10.3%). To take it a step further, Dr. Hirsch calculated the cost per month per 1% drop in A1c, finding that CGM fell in between the expensive drugs (sitagliptin, liraglutide, and canagliflozin; ~$400-$640/month/1% A1c drop) and the cheaper drugs (glyburide, metformin, and pioglitazone; ~$3-$6/month/1% A1c drop), at $239/month/1% A1c drop. From an A1c-centric perspective, the cheaper drugs may be the most short-term, cost-effective way to achieve glycemic control. However, they can put patients at elevated risk of complications (particularly hypoglycemia) or be maxed out such that another therapy is required. Intermittent CGM can both protect against hypoglycemia and reduce A1c, and at a lower price point than novel orals and injectables. 
• As Dr. Hirsch pointed out, his presentation only included A1c as a measure, and didn’t consider the cardiovascular benefits seen in EMPA-REG and LEADER. If these were considered, we imagine that the value proposition of SGLT-2s inhibitors and GLP-1 agonists would look considerably different for type 2 patients in this analysis. While CGM can help protect against hypoglycemia, hyperglycemia, and glycemic variability, the long-term benefits of cardiovascular protection plus glycemic control may shift the cost/month/QALY in favor of novel agents. It’s hard to say how payers view this and what horizons they consider valuable – on the other hand, CGM might have the edge for reducing severe hypoglycemia in a 1-2 year time frame, which may be more financially attractive to payers.
• In Q&A, Yale’s Dr. William Tamborlane and Dr. Hirsch debated the significance of improving A1c by “0.3-0.6%” in older patients with long-standing diabetes – according to Dr. Tamborlane, the lack of data suggesting that tight control at this age prolongs life or improves CV health means that the QALYs gained may be zero, so providing CGM would be a waste. Dr. Hirsch responded that close to one-third of US adults will have diabetes within the next five to ten years, and most of them will be older people with type 2. “If one develops type 2 in his 50s, 60s, or 70s, as opposed to 20 years ago, he’ll be around for at least 10-20 years, so if we don’t treat aggressively, we’ll be looking at more retinopathy, more nephropathy, more neuropathy, etc. It’s a good point about life span, but we still have to worry about complications. I don’t think many would agree that it’s ok to keep A1c at double digits in anyone.”
• Dr. Hirsch also touched on the literature behind professional CGM, flash glucose monitoring, and CGM in type 2s:
  
  • On professional CGM in type 2s, Dr. Hirsch said that data is sparse, but promising. There are nine studies published in peer-reviewed journals (three are RCTs, n=158 total). Each demonstrates significant A1c reductions, but more data needs to be collected before a health economic argument can be put forward.
  • For flash glucose monitoring and prolonged CGM in type 2, Dr. Hirsch pointed to Abbott’s REPLACE study and Dexcom’s DIaMonD (type 2 cohort). We covered the form in detail at ATTD 2016 (and it was just e-published in Diabetes Therapy in December), and see the DIaMonD write-up above. Dr. Hirsch concluded that the initial data for FGM in those using prandial insulin is encouraging for the hypoglycemia prevention, but more studies are required. Dr. Hirsch left the audience with questions: What is the benefit of FGM in those not on prandial insulin? What is the role of CGM in those using prandial insulin – can patients be more aggressive than seen in the DIaMonD study?

5. MiniMed 670G Learning: Adding Daytime Aggressiveness via Insulin:Carb Ratio, Shorter Insulin Duration

Today’s agenda was packed with cuts of data from the MiniMed 670G pivotal trial, though our learning was more thematic:

• Dr. Rich Bergenstal pointed out that some 670G pivotal participants needed a more aggressive insulin:carb ratio and shorter duration of insulin action to make the system more aggressive during the day – especially in the “post-post prandial period.” He further noted two other things we’ve seen in our own experience with automated insulin delivery: (i) treating lows requires less carbs (since basal is suspended already – “let the algorithm do the work”); and (ii) stacking correction boluses on top of automated basal insulin can result in serious lows. (And even a subsequent rollercoaster if insulin suspension is again combined with eating correction carbs.) Adam has experienced both of these in his test drive of the DIY Loop system, reaffirming what we’ve heard going back to the 640G – there is a behavior learning curve with these systems and old habits need to be relearned.
• The MiniMed 670G will actually revert from Auto Mode back to Manual mode during prolonged periods of hyperglycemia, prolonged delivery of maximum or zero insulin, and sensor/self-diagnostic issues. This is called “safe basal timeout.” We weren’t aware of the prolonged hyperglycemia and insulin delivery pieces of safe basal timeout, which were somewhat surprising to hear – aren’t these exactly the times where automation might be most useful? However, this does confirm what we’ve heard from certain users – the system can be finicky and does get kicked out of closed loop. Barbara Davis Center’s Laurel Messer shared her center’s 670G pivotal trial experience, noting that the percent of time adolescents at BDC (n=12) spent in Auto Mode significantly declined from 86% in the first two weeks of the study to 72% at the end of the three-month trial period (p<0.05); roughly 70% of the time, getting kicked out of Auto Mode was due to the safe basal timeout. She said attention will need to focus on optimizing time in Auto Mode.
• Many asked, “What is the max insulin delivery rate the 670G can deliver per hour?”, to which the most common response was, “It’s complicated to explain.” The algorithm does adapt every day in each person, so there is not a one-size fits-all answer to this question. (It sounded like audience members wanted to know what multiple of the normal basal rate the algorithm can deliver – e.g., 2x, 4x, etc.) The 670G algorithm needs at least 48 hours in Manual Mode before Auto Mode can be initiated, and it uses a patient’s actual insulin data to individualize insulin delivery for Auto Mode. At midnight, the algorithm reassesses up to six days of the most recent pump data to update: (i) the feedback controller gain (aggressiveness); (ii) the maximum auto-basal that can be delivered by the system; and (iii) other system parameters.

6. Medtronic’s Sugar.IQ app with Watson to Undergo larger Preview to a larger Audience on February 28

Medtronic’s very smart Dr. Huzefa Neemuchwala (Head of Innovation, Diabetes Service and Solutions) ran attendees through the Sugar.IQ with Watson app, sharing that a “larger preview” to “a larger audience” will occur on February 28. As of CEO Omar Ishrak’s presentation at JPM last month, a full launch of Sugar.IQ was expected in May-October. This upcoming launch will make sure the infrastructure holds up and get more feedback on the patient experience. The app was first demoed last fall and rolled out to 100 MiniMed Connect users, and we imagine a lot of valuable feedback came in. Dr. Neemuchwala’s demo on the app built on those we’ve seen at recent conferences, reminding us of its compelling potential to tease actionable insights out of diabetes data, change behavior, and stimulate discovery and teaching conversations with HCPs. He showed a few new Watson insights that we do not recall seeing before (see below), all focused on pattern recognition around mealtime choices or hypoglycemia/hyperglycemia behaviors. This presentation emphasized that Sugar.IQ will pair with Medtronic’s Guardian Connect (standalone mobile CGM for MDIs), perhaps another reason why the full US launch has been delayed – Guardian Connect is still under FDA review, slated for a May-October launch. Dr. Neemuchwala shared a vision to make a “dynamic Wikipedia of diabetes knowledge” with model to drive patient engagement (curiosity), build habits (trigger, action, variable reward, investment), and offer personal guidance: How many carbs are in this slice of bread? Am I at risk for hypoglycemia tonight? If I eat this burger, what will happen to my blood sugar? I did not know that I experience morning highs? We love the concept, but recognize the value rests on engagement – will patients keep using the app and logging meals, insulin doses, etc.?

• “I notice that you tend to go low after meals with >20 grams of protein.”
• “I see that you often go low between 12-3pm on Saturday. I will keep an eye on the patterns during this time as I get more data. Be aware as you plan your day.”
• “Way to go! Great! I noticed that you had only one nighttime low in the last month. Whatever you’re doing seems to be working very well.”
• “I see that between 6am-9am, your glucose often goes high (300+ mg/dl) after taking an insulin injection. This trend might be worth discussing with your physicians or dietitian during your next visit.”
• “After your glucose is high for more than 120 minutes, you then tend to go low. Be especially careful with corrections from hyperglycemia.”

7. Dr. Aaron Kowalski On Diabetes Data – We Need Universal Use of CGM, Cloud-Connected Devices, Insulin Dose Titration Software

In the conference’s opening talk on “Big Data,” a fired-up Dr. Aaron Kowalski (JDRF) highlighted that we rarely use ANY data in diabetes – and the technology exists to solve this problem now. He shared a great “near-term” list of goals and questions in diabetes data (see tables below), highlighting the importance of “universal” CGM use in type 1 (24/7) and type 2 (intermittent); computer based dosing recommendations for pumps and MDI; capturing all insulin dosing data; real-time data flowing into the cloud directly from devices/phones; and better use of device data in population registries. We loved seeing him emphasize CGM in type 2, algorithms to titrate insulin, and pushing the entire room to solve these problems now. He showed a few pictures from the DIY community, who has already solved many of these issues – e.g., auto-tuning basal rates with OpenAPS and bolusing from an iPhone/Apple Watch with Loop. “People sitting in their garage are doing this. I know some companies will say, ‘It’s the FDA and regulations,” but this can be done and we have to work together to get over this hump.”

8. 640g US Pivotal Data, Plus How to Succeed with PLGS

In an excellent review of predictive low glucose suspend technology, Dr. Thomas Danne shared data from the US pivotal trial of the MiniMed 640G with SmartGuard (Buckingham et al. DTT, in press) and that patients who don’t override the algorithm tend to have better glucose profiles. The trial (n=80) consisted of hypoglycemia induction by basal escalation with the system’s “suspend before low” feature set to 65 mg/dl (similar to the 530G pivotal). Whereas the control group had hypoglycemia in 93% percent of cases following basal escalation, blood glucose levels below 65 mg/dl were avoided 60% of the time in the 640G group. Further, blood glucoses below 60 mg/dl and 50 mg/dl were avoided 68% and 81% of the time, respectively, in those on 640G. As a reminder, Medtronic is not releasing the 640G as a dedicated product in the US – the 630G started shipping in September, and the 670G hybrid closed-loop system is now projected for a full launch in May-October of this year (per JPM 2017) – but this data may still be useful in supporting user uptake/reimbursement of the 640G in Europe. Dr. Danne continued to discuss the recently e-published German MiniMed 640G user evaluation, a six-week trial during which 24 type 1 patients were assigned to either SAP plus SmartGuard (with a threshold of 70 mg/dl) or SAP alone. The paper found that SmartGuard didn’t significantly alter mean glucose, but measures of hypoglycemia (time under 70 mg/dl, area under curve, number of excursions) were all improved as expected. We continue to see this in studies of CGM and diabetes tech, where the average doesn’t change but lows and/or highs do improve. Investigators also looked at the way that patients interacted with the technology – patients had a tendency to manually resume insulin and/or take a carbohydrate after suspension, which was “not such a good idea in every case, resulting in elevated blood glucose levels.” Dr. Danne’s main message was to trust the algorithm,” and let it do the work. Dr. Rich Bergenstal echoed this sentiment in the following talk; he observed that patients who were the least active blood glucose-managers did the best on 670G in the pivotal trial because they simply trusted the algorithm.

9. which Type 2 Patients Should Be On Pumps?

Citing a lack of conclusive RCTs for or against pump use in type 2 patients, Dr. John Pickup (King’s College, London, UK) walked the audience through his group’s individual patient data meta-analysis (Diabetes Care, in press), which found that pumps are best and likely most cost effective in type 2 patients with elevated A1c and insulin doses even after insulin optimization. Dr. Pickup and his team obtained all patient data from five trials (Raskin et al., Herman et al., Wainstein et al., Berthe et al., and Reznik et al. (OpT2mise)) and created a single large data set (n=590 participants; roughly half on MDI and half on CSII). Data on the individual participants of the selected trials was requested from the research teams, at which point Dr. Pickup could perform a one-step meta regression of covariates on outcomes to determine which characteristics affect the efficacy of pump therapy in type 2s to the greatest degree. Unsurprisingly, A1c reduction in pump vs. MDI was greatest in those with elevated baseline A1c, and reduction in insulin requirements with pump vs. MDI was greatest in those with highest baseline insulin requirements. The A1c effect size was markedly larger in the OpT2mise trial than in the overall data set (-1.1% in the pump group vs. -0.4% in the MDI group), perhaps due to the higher baseline A1c of 9.0%. This mirrors Dr. Pickup’s work in CGM, where patients also see more benefit the higher their A1c. The analysis raises a few clinical questions, assuming pumps get better coverage one day: will patients who see large benefits on pump or CGM be able to keep their devices indefinitely, particularly in more cost-conscious health systems? Or will they be asked to stop using pump/CGM therapy as soon as they are at an acceptable A1c or time in range? If they do stop using the devices, would their control return to its original level? 

10. Dexcom Pipeline: G5 Mobile Has Launched in EU and South Africa

Dexcom SVP of R&D Mr. Jake Leach excitedly reviewed aspects of Dexcom’s pipeline: The Android version of G5 has launched in Europe and South Africa, G6 is currently in a pivotal trial and will be submitted for regulatory approval upon completion (for an expected 2018 launch, per the 3Q16 call), and factory calibration is in the works (the first G6 product with Verily, launching in 2H18 and then G7 to follow). We’ll be interested to see how the international launch of G5 mobile for Android is going OUS, along with the move to no longer require the receiver component – revenue shot up 55% YOY in 4Q15 when G5 launched for iOS, and Android-compatible smartphones claim up to 60% of market share in some major European markets. Initially, G6 will be 10-day wear and one calibration per day after startup, though Mr. Leach noted at least four times throughout his talk that Dexcom “is on a quest” or “very close” to eliminating fingersticks. To back up this claim, he showed the 10-day pre-pivotal (n=49) data first presented at DTM in November: 8.1% MARD with one calibration and 8.8% MARD with zero calibrations.

11. Intro to a Roche-Sponsored Study of Pump Accuracy and Function

Dr. Ralph Ziegler (Diabetes Clinic for Children and Adolescents, Münster, Germany) provided an overview of the in-progress IDS Comparative Evaluation Study, which investigates the dosing accuracy and functional differences between various insulin pump options. Sponsored by Roche, the study evaluates pumps (both durable and patch) from different manufacturers, assessing the following parameters: (i) insulin dosing, for both basal and bolus; (ii) speed of insulin delivery; and (iii) timeline of the occlusion alarm.

• The study is not yet completed, but Dr. Ziegler whet the audience’s appetite with preliminary data on the bolus dosing accuracy of six pump models with different infusion sets. Twenty-five successive boluses were delivered by each pump and individually weighed on a microgravimetric scale to determine the average dose and dose variability. This was repeated for boluses of 10, 1, and 0.1 units of insulin, with nine rounds of testing per each combination of insulin pump and infusion. Across all pump/infusion set combinations, dosing accuracy was fairly strong for the 10 IU bolus, with a maximum deviation of only 8% from the target dose. Dosing accuracy was notably worse for the 1 IU bolus (maximum deviation = 42%), and particularly weak for the 0.1 IU bolus (maximum deviation = 64%), signaling a need for improved dosing capability at these small bolus amounts.
• Despite the fact that all devices were more accurate at higher bolus amounts, different pump/infusion set combinations emerged as having the best relative accuracy (i.e., smallest deviation from target dose) at each of the different bolus doses assessed. These preliminary results are blinded until the full study is completed so Dr. Ziegler could not share which manufacturer’s pumps were most accurate at each of the different bolus doses. Nevertheless, this analysis suggests that there is no one “most accurate” pump on the market but instead a spectrum of pumps that are more or less suitable depending on a patient’s particular insulin requirements. For instance, the best pump for a child with very low basal insulin requirements is not the same as the best pump for an adult.
• We remain mindful of Dr. John Pickup’s critique of comparative pump studies from DTM last fall. Foremost, he warned that pump accuracy and precision study results may be method dependent – meaning that results obtained through the IDS Comparative Evaluation Study’s technique of measuring insulin by its microgravimetric mass may not agree with results obtained from an alternative method, such as measuring insulin coming out of the pump by volume. Dr. Pickup also argued that the clinical significance of varying pump accuracy is unclear, save for the case of children and insulin sensitive individuals with lower insulin needs vs. the general population of pump users. Indeed, given the current fragility of the pump field, the accuracy and precision of different pumps may not be the highest priority item on the list of barriers to adoption.

12. Onduo: “Product to Show the World” Expected in ~One Year

Onduo’s Head of Product and Technology Mr. Andrew DiMichele (filling in for CEO Dr. Josh Riff at a Sanofi-sponsored symposium) announced that the joint Sanofi/Verily venture should have a product to show the world in ~one year. While he didn’t share much else in terms of product details, he discussed how the newly-formed Onduo is designed to beat the common challenges blocking other digital health companies. While Silicon Valley-based digital health companies boast tremendous entrepreneurial spirit, Mr. DiMichele argued that they often lack meaningful experience in healthcare. To address this, Onduo brings together experts in technology (from Verily) with experts in healthcare (from Sanofi). Capital constraints are another hurdle for many digital health companies, but Onduo was launched with a nearly $500 million combined investment from Sanofi and Verily. Lastly, Mr. DiMichele explained how “point solutions” are an attractive goal for digital health companies, promising a seemingly quick and easy path to profitability. A successful commercial digital health product, however, must be integrated within the overall healthcare system. As such, the Onduo team is keeping all important players in mind – patients, HCPs, diabetes educators, regulators, payers, etc. – in creating something that will empower patients to make better decisions day-to-day. Mr. DiMichele confirmed that the initial focus for Onduo will be type 2 diabetes, though the company aims to later expand into type 1 diabetes management.

• If you recognize Mr. DiMichele’s name, it’s because he was one of three co-founders and former CTO of Omada Health. With this incredible background, he brings strong diabetes-related digital health experience and experience with alternative outcomes-based pricing models as well as prevention expertise in general.  

13. New Glytec Glucommander Inpatient data: <1 Day time to target, ~68% Readings in Range, Low hypo Frequency

Dr. Bruce Bode presented positive retrospective glycemic data (February 2013-November 2016), as well as cost-savings, from inpatient use of the Glucommander electronic glucose management system. Patients (n=5,718) admitted with hyperglycemia to one of seven hospitals arrived at their prescribed target blood glucose (100-140, 120-160, or 140-180 mg/dl) in 0.8 days from a starting average of 261.7 mg/dl. Once at target, 67.9% of blood glucose readings and 68.5% of patient day values remained between 70-180 mg/dl. Hypoglycemia was very unlikely once target had been reached and in the next 24 hours, with just 0.0011% of time below 40 mg/dl and 0.0130% of time below 70 mg/dl, respectively. In addition, the financial proposition associated with Glucommander is compelling, with over $3 million in potential savings per 250-bed hospital per year due to shorter patient stays, reductions in point of care testing, reduction in cost of treating DKA, and reduction in cost of treating CABG (coronary artery bypass surgery) patients. This in-patient data, combined with impressive out-patient data presented in Glytec’s ADA 2016 poster and previous publications, lend credence to the idea that clinical and personal automated decision support are low-hanging fruits that can help patients stay in normoglycemic ranges and reduce costs. To illustrate the scalability of the system, Dr. Bode claimed eGMS Glytec software was used to help manage 45,000 patients in 2016, and this number will likely double in the coming year (he clarified "year or two")! In this transitioning healthcare climate, where hospitals are responsible for covering costs of within-30-day readmissions, incentives align to help guide patient care, both in the hospital and upon release. We look forward to hearing more Glucommander data (outpatient) on Friday.

--by Adam Brown, Abigail Dove, Brian Levine, Payal Marathe, and Kelly Close