November 10-12; Bethesda, MD; Days #2-3; Highlights – Draft

Executive Highlights

Hello from Bethesda, where the 16^th annual Diabetes Technology Meeting just wrapped up. Kelly still remembers the very first one of these! Highlights from days 2-3 included the latest in CGM (most notably Dexcom’s compelling G6 and Verily updates); views on the future of the artificial pancreas; and the latest in software and apps (Glooko, Tidepool, and more). Read on for our top 15 highlights from the final two days of DTM (plus two honorable mentions), and see our highlights here from day #1 (implantable CGM, reimbursement).

Top 15 Highlights

1. Dexcom’s Peter Simpson presented highly compelling G6 pre-pivotal data (including an 8.8% MARD with zero calibrations!) and showed new impressive pictures of the second-gen sensor with Verily (smaller than an M&M in volume, non-adjunctive, zero calibration, Bluetooth-enabled). We were very impressed on both fronts and it’s clear Dexcom is pushing hard on better accuracy, no calibration, outstanding form factor, and better data. See pictures below!

2. TypeZero’s Dr. Patrick Keith-Hynes showed snappy-looking screenshots and discussed the company’s three commercial software products: (i) inControl AP to run closed loop from a smartphone or integrated into a pump; (ii) inControl Advice to give smart open-loop dosing advice for MDIs or pumpers; and (iii) inControl Remote Care Platform for HCPs.

3. Roche’s Dr. Guenther Schmelzeisen-Redeker shared positive accuracy (overall MARD = 10.6%) and time lag (mean lag post-processing: 5.5 minutes) data from the CE mark approval study of Roche’s Accu-Chek Insight CGM. A limited EU launch will occur by the end of 2016.

4. Abbott’s Mr. Scott Harper presented new data from the open access phase of REPLACE: 100% of participants (type 2s on basal/bolus therapy in poor control) opted to continue using the sensor past the 6-month study period and maintained elevated levels of testing and decreased time in hypoglycemia.

5. Medtronic’s Mr. Brian Kannard reviewed data from the pivotal study of the new “Guardian Sensor 3,” approved as part of the MiniMed 670G in September.

6. OHSU’s Dr. Jessica Castle reviewed the landscape of glucagon development, including news to us that Lilly/Locemia’s phase 3 intranasal glucagon will be submitted to FDA “by 1H18.”

7. Glooko’s CEO Rick Altinger announced that the company’s basal insulin titration product (mobile insulin dosing software [MIDS]) has undergone two pre-submissions to FDA and is currently in human factors testing. We cannot wait for this! Glooko told us a 510(k) submission is expected by the end of 2016, with hope for clearance in 1Q17. The Pump Advisor in collaboration with DreaMed will launch a multi-site clinical trial “in just a few months.”

8. Tidepool CEO Mr. Howard Look commended Dexcom for its new developer platform (he’s such a class act) and upcoming open APIs, indicated that the Nutshell app is in beta testing, displayed updated Tidepool device compatibility, and laid out his three keys for enabling a data ecosystem.

9. Dr. Joan Lee Parkes provided an update on the first series of DTS’ BGMS Surveillance Program, a post-market study of 18 BGMs on the US market. She had planned to share results at DTM, but pushed timing for completion and pushed results back again to “early 2017.”

10. FDA’s Dr. Courtney Lias highlighted some of the key changes in the newly published BGM Final Guidances and shared excitement for the upcoming post-market DTS BGM surveillance program data.

11. Arkray’s Dr. Shinjiro Sekimoto returned to DTM for the third consecutive year to share limited progress on the company’s glucose dehydrogenase-dependent DiET (direct electron transfer) CGM. Clinical trials in the US and Japan are expected between 1Q17 and 2018.

12. FDA cybersecurity expert Dr. Suzanne Schwartz commended J&J’s response to cybersecurity vulnerabilities in the OneTouch Ping insulin pump, commenting: “This is the proactive behavior the FDA has been looking to see from the medical device manufacturer and research community and demonstrates the collaborative manner in which vulnerabilities can be addressed in a way that best protects patients.”

13. King’s College’s Dr. Pratik Choudhary gave a masterful overview of the current state of diabetes apps, concluding that they can help alleviate the burden of some “mechanical” tasks (insulin dose calculation in mySugr and Roche’s Accu-Chek Connect), but behavior modification apps need to be simpler and more automated to take off.

14. University of Michigan’s Dr. Joyce Lee highlighted the Nightscout Study, a new initiative “to develop a patient-designed research collaborative innovation network for T1D.”

15. Industry vet Dr. Bill Tamborlane presented a concise summary of the future directions of closed loop: ease of use, personalization, and fully automated control.

Honorable Mentions

1. Dr. John Pickup gave a masterful overview of patch pumps, noting their advantages and disadvantages, sharing optimism for use of simple infusion devices in type 2 diabetes (CeQur, Valeritas), and touching on accuracy and precision.

2. Talks from Verily’s Dr. Howard Zisser and IBM Watson’s Dr. Shahram Ebadollahi shared nothing new on their respective diabetes partnerships.

Top 15 Highlights

1. Dexcom’s Outstanding G6 Pre-Pivotal Data, Compelling Verily Sensor

Dexcom’s Peter Simpson presented highly compelling G6 pre-pivotal data (including an 8.8% MARD with zero calibrations!) and showed new impressive pictures of the second-gen sensor with Verily (smaller than an M&M in volume, non-adjunctive, zero calibration, Bluetooth-enabled). We were highly impressed on both fronts:

• In the pre-pivotal study of G6 (n=32 adults, n=17 pediatrics), the sensor demonstrated an overall MARD of 8.1% vs. YSI at three in-clinic visits (n=993 paired points) and only one calibration per day after startup over 10 days. Outliers were minimal, with 96.1% of points within 20%/20 mg/dl, better than G5 with one fewer calibration per day and nearly 50% longer wear (see below). When Dexcom ran the raw sensor data prospectively through an optimized NO fingerstick calibration algorithm, accuracy was remarkably consistent: a MARD of 8.8% and 95.5% of points within 20%/20 mg/dl (n=1,009 vs. YSI). This exceeded the company’s expectations, and Mr. Simpson said the team has been “impressed by the G6 sensor’s stability and reproducibility in manufacturing.” Three sensor lots were used in the study (key for factory calibration) and YSI measurements occurred at the beginning, middle and end of sensor wear (no further details provided). While the initial version of G6 will be one calibration per day after startup and 10 days of wear, these results support pursuing a no calibration version of G6, and Dexcom believes the sensor will last up to 14 days. Notably, data in both calibration scenarios was equally strong on day one (MARD ~10%), and the adult-only results with one calibration per day were particularly strong overall: MARD of 7.2% and 99.4% within 20%/20 mg/dl. These are outstanding pre-pivotal data and we look forward to seeing if they are confirmed in the larger 300-person G6 pivotal trial that is now underway. As of Dexcom’s 3Q16 call, a G6 launch is expected in 2018. See the detailed G6 pre-pivotal data tables below.

• Mr. Simpson and Verily Technical Project Lead Dr. Will Biederman showed impressive pictures of the second-gen Dexcom/Verily product: smaller than an M&M in volume (“thinnest CGM ever”), no fingersticks, non-adjunctive labeling, and direct-to-smartphone Bluetooth connectivity. The second-gen sensor has completed initial animal studies and the team is getting approval for clinical use. The pictures below show how truly impressive this form factor will be, and both speakers commented that it will expand CGM adoption beyond early adopter type 1s and into the type 2 population. Verily’s Dr. Biederman held up a functional prototype, drawing laughter from the crowd, since no one could actually see it in his hand. (We tried to get a peek up close, but we couldn’t persuade him to let us see it!) Neither speaker commented on launch timing, but as of Dexcom’s 3Q16 call, commercialization was expected “as early as 2020” for the second-gen sensor and in 2H 2018 for the first-gen sensor. The latter was not mentioned today, but it is expected to use the G6 sensor and have a smaller form factor than Libre (though obviously larger than this compelling second-gen version). It’s worth noting that this partnership is only 15 months old (signed in August 2015), signaling remarkable progress and what seems like a great culture fit between the companies.

• See the detailed section below for more pictures and highlights from this talk.

2. TypeZero shows inControl AP, inControl Advice, and inControl Remote Care

TypeZero’s Dr. Patrick Keith-Hynes shared snappy looking screenshots and discussed the company’s three commercial software products in more detail than ever: (i) inControl AP to run closed loop from a smartphone or integrated into a pump (about to start the IDCL study with Tandem and Dexcom); (ii) inControl Advice, an app that gives smart open-loop dosing advice for MDIs or pumpers; and (iii) inControl Remote Care Platform for HCPs to optimize insulin therapy and remotely monitor patients using pens or pumps. TypeZero has taken the UVA algorithms, modified them, and brought them to what looks like a commercial-ready state. The screenshots we saw today for the first time looked very sharp (see below) and were a far cry from the less modern original DiAs research interface. We’re particularly glad to see TypeZero working on bringing its artificial pancreas algorithms into MDI users – the potential for smarter insulin dose titration is massive in our view. No timing or commercial details were shared, but we are very optimistic about the potential for smarter open loop injection therapy in both type 1 and type 2 diabetes. Notably, the Advice app has a chat-like, text-message interface, getting away from pushing buttons (similar to what Medtronic/IBM Watson are doing with Sugar.IQ). Dr. Keith-Hynes showed the smartphone version of inControl AP for the first time we’ve seen, which modulates basal insulin and can issue automated correction boluses, a key difference from basal-only algorithms like the MiniMed 670G. As a reminder, the upcoming 240-patient, NIH-funded IDCL study will support an expected late 2017 PMA submission of a Dexcom G6-integrated Tandem t:slim X2 pump with TypeZero’s inControl AP algorithm built into the pump. Patients will use the system for six months. (The study’s first pilot phase will use Dexcom’s G5, inControl AP running on a smartphone, and a Tandem pump.) See pictures and more details below.

• The inControl AP app home screen showed the current CGM value in green (120 mg/dl), with some simple text indicating predictions and system status. inControl AP includes a basal attenuation module, a basal increase module, a hyperglycemia correction module, and a user input module. The algorithm makes a new dosing decision every five minutes, and overnight, gradually treats to a target of 120 mg/dl by morning.

• See the detailed section below for more from this talk!

3. Updated Data from Roche’s Insight CGM CE Mark APproval Study

Roche’s Dr. Guenther Schmelzeisen-Redeker shared positive accuracy (overall MARD: 10.6%) and time lag (mean lag post-processing: 5.5 minutes) data from the CE Mark approval study of Roche’s Accu-Chek Insight CGM. We heard topline results from the trial at EASD in September, but Dr. Schmelzeisen-Redeker provided additional details. The study investigated performance in 36 people with diabetes, each wearing two sensors simultaneously for seven days. The study also included two days with induced glucose variations. As previously reported, overall MARD was 10.6%. The sensor performed very well in the ≤70 mg/dl range – MARD of 9.5% – with performance in the euglycemic and hyperglycemic ranges on par with the overall MARD, at 10.6% and 10.7%, respectively. As expected, accuracy improved over the seven-day wear period, as MARD fell from 12.9% on day one, to 9.1% on day seven. Dr. Schmelzeisen-Redeker also showed that MARD during induced glucose variations was on par with overall accuracy: 10.7% in aggregate, 10.4% in the hypoglycemic range, 10.1% in the euglycemic range, and 11.5% in the hyperglycemic range. This suggests that lag time may not significantly affect the sensor’s accuracy. Indeed, on the following slide, Dr. Schmelzeisen-Redeker displayed a table showing that mean raw time delay was 9.2 minutes in the trial. However, when a real-time correction algorithm is applied, mean time delay decreased to 5.5 minutes (the algorithm works by applying the population average time delay to the current figure and extrapolating from the current glucose trend). This accuracy and lag data is encouraging, especially because the Insight CGM is Roche’s first-gen CGM. Our main concerns regarding uptake pertain to the two necessary calibrations per day (whereas Abbott has eliminated them and Dexcom is headed in that direction with G6) and the sensor’s on-body size. As a reminder, the Insight CGM will be rolled out in a controlled fashion in the Netherlands, Sweden, Norway, and Denmark via specialized diabetes centers by the end of 2016.

4. New data from open-access phase of Abbott’s Replace Study

Abbott’s Mr. Scott Harper reviewed findings from Abbott’s REPLACE and IMPACT studies of FreeStyle Libre, and presented new data (to us) from the open access phase of REPLACE: 100% of participants (type 2s on basal/bolus therapy in poor control) opted to continue using the sensor past the six-month study period and maintained the elevated levels of checking and decreased time in hypoglycemia. Specifically, 12-months following the start of the study, the patients using Libre scanned 7.3 times per day on average, with 0.2 SMBGs per day – compare that with the ~10 scans and ~4 SMBGs per day when the study began. In addition to showing sustained high levels of scanning, the data demonstrates a lasting trust in Libre, as SMBG’s were almost entirely eliminated in the insulin-using group. The hypoglycemia benefits seen at six months in REPLACE were also maintained at 12 months: Relative to baseline, patients using FreeStyle Libre spent ~42 minutes fewer per day <70 mg/dl (p=0.002), ~24 minutes fewer per day <55 mg/dl (p=0.002), and ~10 minutes fewer per day <45 mg/dl (p=0.0013). Not only are these improvements statistically significant, but incredibly clinically relevant. Mr. Harper did not share total time in hypoglycemia. Importantly, the significant reductions in hypoglycemia were seen at night as well, countering speculation that the lack of alarms poses a nighttime danger (presumably, patients identify nocturnal hypoglycemia trends in retrospective glucose data). This 12-month open-access data augments the clinical and real-world evidence that Libre can and will help a lot of people – 200,000+ are already using the system in Europe, according to CEO Mr. Miles White in the company’s 3Q16 financial update. The next big step for Libre, FDA approval, could occur as early as 1Q17, according to Abbott’s 2Q16 guidance; the product was submitted to the Agency in 3Q16. We’re eager to see what happens with the dosing claim and how the company prices it in the US. More on that topic here.

5. Medtronic Reviews Guardian 3 Sensor Data, 670G Enthusiasm from Pivotal Participants

Medtronic’s Mr. Brian Kannard reviewed data from the pivotal study of the new “Guardian Sensor 3,” approved as part of the MiniMed 670G in September. As we’ve covered previously, the seven-day-wear sensor (formerly called “Enlite 3” and “Guardian”) has a labeled overall MARD of 10.6% on two fingerstick calibrations per day, and a slightly improved MARD of 9.6% with 3-4 calibrations per day (n=12,090 paired YSI-CGM points). The FDA summary (see page 45 here) actually says four calibrations per day “are recommended.” Mr. Kannard shared that as part of the MiniMed 670G continued access study, the company has gathered over a year of real-world data on the new Guardian sensor – 4,000+ sensors and 71 manufacturing lots – and seen similar accuracy to the pivotal study. Notably, a “majority” of MiniMed 670G pivotal users cited sensor accuracy as a leading factor in trusting the system – it will be key for Medtronic to see this post-launch, since many early adopters will have baggage from the less accurate earlier generation Medtronic sensors. The improved sensor accuracy has come through a new electrode design (three parts), only curing the enzyme on the working electrode now (instead of over the whole sensor), and new diagnostics to monitor sensor health (built into the transmitter). In the future, the company hopes to obtain an insulin dosing claim (Guardian does not have one, as it only modulates basal), an extended lifetime to 14 days, no calibration, easy insertion/startup, and a smaller footprint/profile – no timelines were shared, and Medtronic is definitely playing catch-up on all these fronts compared to Abbott and Dexcom. Overall, we are glad to see the new Guardian sensor has made progress on accuracy and look forward to seeing what patients think once it launches with the MiniMed 670G in Spring 2017. This sensor is also under FDA review in the US as part of the Guardian Connect standalone mobile CGM (launch expected by April 2017).

• Mr. Kannard also shared positive qualitative data from the 670G continued access study (CAS), which included 80% of pivotal trial participants: 89% of surveyed CAS patients perceived they have better control while using Automode (i.e., hybrid closed loop), and 93% of patients were either satisfied or extremely satisfied with their overall quality of life while using the MiniMed 670G.
• One quote from a 670G pivotal participant was particularly encouraging from a quality of life perspective: “It is a care system that truly says, I have diabetes, but it does not have me. The biggest concern for myself as an adult, wife, mother, and daughter were my nighttime low that I would not wake up for. I have NOT had one while participating in this study. The CGM is near perfect 98% of the time. This has dramatically improved. I thoroughly love the closed loop system.”

6. Dr. Jessica Castle’s Glucagon Landscape Overview

OHSU’s Dr. Jessica Castle reviewed the landscape of glucagon development, including news to us that Lilly/Locemia’s intranasal glucagon will be submitted to FDA “by 1H18.” Lilly acquired the glucagon formulation from Locemia a little over a year ago. Last month’s 3Q16 update noted that the product was still in phase 3 testing, and this is the first we’re hearing of NDA submission timing. We’re confirming with the company if this is true, which is a bit slower than we expected (although, all of a sudden, 2018 isn’t actually that far away); phase 3 completed in pediatrics and adults last year. See the table below for updates/refreshers on Zealand, Xeris, Latitude, Lilly, and Zosano’s glucagon projects.

7. Glooko’s Basal Insulin Titration (MIDS) and Pump Advisor Move Into Studies

Glooko CEO Rick Altinger announced that the company’s basal insulin titration product (mobile insulin dosing software [MIDS]) has undergone two pre-submissions to FDA and is currently in human factors testing. Glooko told us a 510(k) submission is expected by the end of 2016, with hope for clearance in 1Q17. The Pump Advisor in collaboration with DreaMed will launch a multi-site clinical trial “in just a few months.” MIDS is currently being tested by clinicians, though no specifics were shared on the feedback. Glooko told us that MIDS will be distributed in both clinics and large health systems – we think this could make a real difference for many, and we like that it all happens for patients through the mobile app. Mr. Altinger enthusiastically discussed September’s merger with Diasend, and noted that Glooko will be supporting the Transmitter box, PC/Mac Uploader, and Tablet Kiosk for at least the next year. Following the year, the combined company will transition to a long-term solution that combines the best features of both platforms. 

• As we noted at ATTD, a clinician will select and/or configure long-acting insulin-dosing titration instructions with MIDS on the web, with the option of selecting standard titration schemes (e.g., AACE geriatric dosing template). Clinicians enter a patient’s basic information, type of insulin, glucose range, and time periods. A patient’s mobile device then receives the personalized dosing configuration, with reminders to check glucose and to take insulin doses which are recalculated based on the clinicians configuration (e.g., “It’s checkup time! Let’s see if your insulin dose needs to be adjusted.”). The system also sends safety alerts (e.g., “We’ve detected you have a reading under 70 mg/dl. Contact your physician….”) Patients dose will change on a regular basis based on the glucose readings collected, which will ideally be automated as Bluetooth-enabled meters talk to Glooko.  
• The exciting Helmsley Charitable Trust-funded collaboration with DreaMed to develop the MD Logic Pump Advisor was also a focus of Mr. Altinger’s talk (also introduced at ATTD). The feature imports pump and CGM data, determines an appropriate care plan, sends insulin pump settings adjustment recommendations to the physician (e.g., for pump users, change basal from 0.95 u/hr -> 0.8 u/hr from 12-8am due to pattern of nighttime hypoglycemia), and records it in the patient’s EHR. The physician approves the recommendation, and the patient receives the recommendation through the app and either confirms or rejects it to “close the loop.” No details on the study were provided, but we believe clinicians and patients will be ecstatic to use this product. We also wonder how it might evolve in an era of automated insulin delivery, since many pump + CGM users will likely move to AID in the coming five years.

8. Tidepool’s Howard Look on Vibrant Data Ecosystem

Tidepool CEO Mr. Howard Look commended Dexcom for its new developer platform and open APIs, shared that the Nutshell app is in beta testing, displayed updated Tidepool device compatibility, and laid out his three keys for enabling a data ecosystem. Dexcom’s big move to open APIs (retrospective data initially) and accelerate an open diabetes data ecosystem was first announced at the DiabetesMine D-Data Exchange two weeks ago, and Mr. Look’s praise today was definitely noteworthy: “Great job Dexcom for taking us into the modern age and enabling a vibrant ecosystem.” As a reminder, the site has launched for developer sign-ups and will be fully available in “early 2017.” We enjoyed hearing Mr. Look share his three requirements for a vibrant data ecosystem at the end of his talk: (i) OAuth2, which allows consumers to access without inputting passwords (the “Sign in with Facebook” or “Sign in with Google” buttons that occasionally appear when logging into a platform); (ii) JSON, a lightweight data interchange format that can be easily shared and decoded; and (iii) RESTful URLs and APIs, which can be cut and pasted and return the operator to the same page/data every time. Mr. Look was so pleased with Dexcom’s new developer platform because it is based on these three key principles. And he added at the end of his talk, “Ask your developer if your company is using these three things. If not, fire him or her.” In his final slide, Mr. Look pointed out that Tidepool's software is broadly available to end users and clinics for free via www.tidepool.org/signup - we love this commitment to access.

• In Tidepool news, Mr. Look mentioned that the mealtime app, Nutshell, remains in beta testing. He didn’t share official launch timing, though at ATTD in February mentioned that he hoped launch would occur “within the next quarter”; the small non-profit is clearly taking its time to get this right, and Tidepool is currently focused on the clinical and end user experience with the Tidepool Uploader and web experience at blip.tidepool.org. We believe many will love the ability to easily track past meals and improve mealtime dosing. (We loved Meal Memory, which has unfortunately been removed from the app store.) Mr. Look also put up a slide illustrating Tidepool’s updated device compatibility, which includes all of the popular CGMs and pumps on the market, but admittedly not a large number of BGMs; “go to Glooko if you need a large assortment of BGMs – there’s a good solution there,” he added.

9. Update on DTS’ Post-Market BGM Surveillance Program: Testing Protocol Revealed and Data to Come in Early 2017

Dr. Joan Lee Parkes (Joan Lee Parkes Consulting, Bristol, Indiana) provided an update on the first series of DTS’ BGMS Surveillance Program, a post-market accuracy study of 18 BGMs on the US market. She had planned to share results at DTM, but pushed timing for completion and results back to “early 2017.” The program, which started in June and was funded by a grant from Abbott (it is a red flag that the company moving most away from SMBG is the only funder), seeks to address two concerns: (i) a lack of reporting standards and testing of BGMs and strips once they are approved; and (ii) poorly performing BGM systems populating the market. DTS has created a protocol to test the accuracy of 18 popular meters, covering the vast majority of the market (see table below). According to Dr. Parkes, assaying glucose from tiny quantities of blood is not standard practice for most commercial labs, so new protocols had to be developed and validated – hence the delay in timing. Testing of the off-the-shelf meters – acquired from retail pharmacies and online retailers from different parts of the country – is now ongoing at three clinical sites (Rainier Clinical Research Center, Diablo Clinical Research, and AMCR Institute), plus one lab site (PPD Laboratories) for further analysis. Each study requires ~110 subjects in order to amass 100 natural blood samples and 20 glycolyzed samples (for assessment of accuracy in the hypoglycemia range). Each study is repeated twice, for a total of three replicates, and each clinical site assays all 18 BGMs. The program will determine if the accuracy of the tested meters meets pre-specified standards, which are based on ISO 15197’s 2013 guidance (+/- 15%), “but more lenient.” A meter will classified into three zones: (i) “clear acceptance” if it has 95%+ probability of meeting ISO 15197; (ii) “indeterminate performance” with 5%-95% probability; or (iii) “clear rejection” if it has less than 5% probability. Surprisingly, a meter will “pass” if it is in either of the first two zones, which is a very low bar – a meter with only a 6% probability of meeting ISO 15197 will “pass,” which is very surprising given the reasons this was established in the first place. DTS will give its “Seal of Approval,” which will not mean much if every meter clears the low bar. Dr. Parkes envisions that manufacturers will approach DTS in the future and request studies, and she hopes that all devices will be tested “at least once every two years.” We’ll be interested to see what the data shows and how seriously the field and FDA takes it (see below).  It seems like this could basically be set up as a profit center.

• Dr. Parkes emphasized that the study is testing accuracy, not usability. This is a key distinction, as meter accuracy is only worth so much (read: nothing) if the operator can’t use the device properly. A composite score describing both accuracy and usability might be interesting for consumers and providers, though that would massively expand cost and scope and this program has moved slow enough as it is.

10. FDA’s Dr. Courtney Lias Discusses New BGM Guidances

FDA’s Dr. Courtney Lias highlighted some of the key changes in the newly published BGM Final Guidances and shared excitement for the upcoming post-market DTS BGM surveillance program data. She noted that the Agency does not recognize ISO 2013 standards, and the new FDA guidances are intended to address some key issues left out of ISO: non-OTC BGM use in healthcare facilities, meters’ design features, performance by intended users, and improved performance in the hypoglycemia range. Studies that will fulfill the new FDA guidance will also fulfill ISO requirements, though the reverse is not always true. As she did at DiabetesMine a few weeks ago, Dr. Lias also pointed out some of the key updates in the guidances: (i) larger accuracy evaluations in at least 350 patients; a focus on usability in intended users; OTC accuracy requirements that use percent only for both hypoglycemia and non-hypoglycemia (patients previously got confused with both absolute mg/dl and %); additional testing for low (<80 mg/dl) and high (>300 mg/dl) samples (may be contrived); subjecting test strips to typical shipping and handling conditions prior to use in the study; and FDA review of lot release criteria (“We think this will go a long way” towards ensuring real-world performance mimics labeled performance – we hope that this will be done fairly). She is “extremely excited” for the upcoming DTS BGM Surveillance Program testing results (expected in early 2017, according to Dr. Joan Parkes). As she has in the past, Dr. Lias said these data can help inform directed FDA inspections – “This is meant for us as a signal and another tool to prioritize where quality needs to be investigated.” The agency is also working on a BGM-specific MDR guidance to better standardize adverse event reporting, which varies widely between companies. Finally, Dr. Lias said the Agency is “open to suggestions” on how accuracy should be conveyed on the outside of strip boxes; the team was not sold on the proposed labeling included in the final guidance, but no further suggestions were received.

11. Arkray to Conduct clinical trials in US/Japan starting in 1Q17

Arkray’s Dr. Shinjiro Sekimoto returned to DTM for the third consecutive year to share limited progress on the company’s glucose dehydrogenase-dependent DiET (direct electron transfer) CGM. Clinical trials are expected in the US and Japan between 1Q17 and 2018. This year, he presented an internal feasibility study (n=10, and 8 without diabetes) presenting an overall MARD of 9.9% with one retrospective calibration per day. This sort of study says nothing about the sensor’s potential accuracy with prospective calibration and used in people with diabetes. (Last year, Dr. Sekimoto presented accuracy data from another small sample of people without diabetes, in which the device demonstrated a mean MARD of 10.9%, ranging from a low of 5.4% to a high of 20.3%. No data on the hyperglycemic and hypoglycemic ranges was included.) The company has moved slowly in the past three years on this product (at least from a public progress point of view) and we are eager to see how the novel sensor performs in the upcoming clinical trials. We hope the studies are robust and consist of a large number of people with diabetes and prospective calibration. There is a long way to go here.

• Arkray’s system involves two components: an on-body component (sensor + transmitter) worn on the abdomen and a mobile medical app running on a smartphone. The app stores glucose readings and provides real-time glucose trend information. The sensor electrode itself measures 0.25 mm in diameter (with an estimated 3 cm length) and is designed to penetrate the skin at an angle (~30 degrees from perpendicular).
• The key element in Arkray’s implantable CGM is the company’s novel sensor chemistry that employs a novel Direct Electron Transfer (DiET) technology. The sensor works at electric potentials of 0.15V vs. Ag/AgCl – compared to the traditional CGM potential difference of 0.6V vs Ag/AgCl – reportedly reducing interference from common culprits such as acetaminophen. Arkray’s approach uses the enzyme glucose dehydrogenase, which it hopes will improve the reliability and sensitivity of glucose sensing, and enable a simpler manufacturing process.

12. FDA Commends J&J for Response to Security Vulnerabilities

FDA cybersecurity expert Dr. Suzanne Schwartz commended J&J’s response to cybersecurity vulnerabilities in the OneTouch Ping insulin pump, commenting, “This is the proactive behavior the FDA has been looking to see from the medical device manufacturer and research community and demonstrates the collaborative manner in which vulnerabilities can be addressed in a way that best protects patients.” Diabetes Technology Society Founder Dr. David Klonoff echoed this praise, calling J&J’s coordinated disclosure “the right approach.” As we reported in our J&J 3Q16 report, Animas sent a letter to customers informing them that the OneTouch Ping insulin pump may be accessed wirelessly to deliver extra insulin (via the unencrypted RF communication that connects the meter and pump). The likelihood of unauthorized access is extremely low, and the letter suggests helpful mitigations (very strong in our view). Behind the scenes, Mr. Jay Radcliffe, the security researcher at Rapid7 who exposed the weakness, shared the vulnerabilities to J&J. J&J then put together a quality test protocol, replicated the process internally, and maintained an ongoing dialogue with Rapid7, FDA, and DHS. This process allowed thorough examination of the situation and proper notification of consumers, along with proper mitigating controls. We see cybersecurity as a “must-have” for products that infuse insulin, but we also hope the absolute risk is kept in perspective. See more in our J&J 3Q16 report.

13. A perspective on diabetes apps from Dr. Pratik Choudhary

King’s College’s Dr. Pratik Choudhary gave a masterful overview of the current state of diabetes apps, concluding that they can alleviate some of the burden of “mechanical” tasks (e.g., bolus insulin dose calculation), but behavior modification apps need to be more simple and automated in order to take off. Dr. Choudhary specifically referenced mySugr and Roche’s Accu-Chek Connect as apps he routinely recommends. On the other hand, he believes that there needs to be more research to enhance behavioral modification apps: they generally only work for patients who are already in the “determination” and “action” phases of the “stages of change model” – leaving the large segment of patients in the “pre-contemplation” and “contemplation” phases behind. To make matters worse, patients frequently start an app, but then relapse and leave the cycle. Another model, Michie et al.’s COM-B model, posits that, to instigate behavior change, a patient needs capability, opportunity, and motivation. Apps provide opportunity for the 64% of adults in the US (and those internationally) who have a smartphone, but the motivation to change hasn’t proven to be sufficient. A recently published DT&T paper found that just one in five adolescents (the go-to population segment for digital media) with type 1 diabetes uses mobile apps to support self-management. In response to the primary author’s Tweet detailing the paper, various Twitter users replied with comments along the lines of “yep. Classic. Look nice but no utility. Utility trumps everything :)” and “So many apps, most cheap/incomplete.” This is the crux of app design – added hassle (e.g., inputting data) must be outweighed by added value. Dr. Choudhary then turned to activity monitoring apps, which passively collect data, eliminating hassle. He noted that these apps don’t often translate to action, citing his own experience with Run Keeper: “Pratik, you haven’t run in four days.” This type of feature may motivate some people, but it didn’t work for Dr. Choudhary, nor the 404 participants in Project SMART – a two-year RCT that found no benefit of app use on weight loss in overweight and obese young adults. Dr. Choudhary sees potential for big data – the entry of Google and IBM Watson into the field – to improve data and insights, though he’s not positive that better analytics will change motivation.

• At least three speakers at DTM 2016 called for improved quality control and regulation of mHealth apps. Drs. Choudhary, Signe Schmidt (Hvidovre University Hospital, Denmark), and David Kerr (William Sansum Diabetes Center, Santa Barbara, California) all referenced an oft-discussed 2015 paper (Huckvale et al.) that analyzed 46 meal insulin calculation apps and found most to be questionable, at the very least. Over half (27) said “not to be used for medical purposes,” only 21 made their formulae available, only 9 had insulin on board compensation, only 10 adjusted for exercise level, 8 didn’t adjust for current blood glucose value … the list goes on, culminating in a statement that 31 had a risk of inappropriate dosing (of one of the world’s most dangerous drugs!). The difficulty here is that these apps are regulated devices, but have not sought FDA clearance; FDA doesn’t have the resources to police app stores, so it falls upon HCPs, professional associations, and patients to separate the signal from the noise – and that is very difficult. Presumably, the apps would continue to improve.

14. Nightscout Study: Patient-Designed Research Initiative

University of Michigan’s Dr. Joyce Lee spoke about Nightscout Study, a new initiative “to develop a patient-designed research collaborative innovation network for T1D. The preliminary charter lists four main goals of the initiative: (i) To promote the participation of patients and caregivers throughout the research life cycle; (ii) To utilize the tools of mobile technology and social media for conducting research; (iii) To accelerate the research lifecycle; and (iv) To promote the principles of open science. Some of the most popular thus far include: How does lack of sleep affect the mental well-being of people with type 1 diabetes and their caregivers? How accurate and consistent are different types of blood glucose meters? How do certain foods affect blood glucose/management? What is the ideal timing of insulin for meals and snacks? How does blood glucose respond to various diets? What kind of glycemia do people commonly experience? [We believe these were all crowdsourced and upvoted by way of Facebook]. See Nightscoutstudy.info for more information – we look forward to seeing where this initiative goes and the impacts it may have in diabetes and other disease areas!

15. Dr. Bill Tamborlane on the Future of Closed loop

Industry vet Dr. Bill Tamborlane presented a concise summary of the future directions for closed loop, highlighting ease of use, personalization, and fully-automated insulin delivery. Before looking to the future, Dr. Tamborlane acknowledged the tremendous progress achieved in the past decade; it was just 10 years ago that Dr. Gary Steil et al. completed the first feasibility study of subcutaneous CGM-insulin delivery in type 1 diabetes. And less than two months ago, FDA historically approved Medtronic’s hybrid closed loop system – a “breakthrough” for the Agency, said Dr. Tamborlane. Yet there is still ample opportunity for advancements within hybrid closed loop and toward fully-automated closed loop. The next steps for hybrid closed loop, according to Dr. Tamborlane are to simplify the user interface and administration of correction doses, something we have heard from some MiniMed 670G pivotal participants. Dr. Tamborlane also suggested that hybrid closed loop systems would benefit from individualized and adjustable target glucose levels, a point we heard repeatedly at ADA 2016. Earlier in the session, Harvard’s Dr. Ravi Gondhalekar argued that MPC algorithms circumvent the need for personalization, but the common opinion shared by others in the session was that inter-subject variability should be addressed. Dr. Tamborlane continued, mentioning that the amount of insulin that can be infused automatically should be “liberalized” so as to reduce patient burden, and clinical strategies have to be developed to prevent hypoglycemia during exercise (the most obvious, of course, is proactive snacking). To achieve fully closed loop, Dr. Tamborlane outlined the need for optimized algorithms, a consideration of single- vs. dual-hormone systems, and potential assistance from adjunctive therapies, such as SGLT-2 inhibitors or pramlintide. We wonder about co-formulated insulin-pramlintide in particular, which has not moved forward very quickly.   

Honorable Mentions

1. Patch Pump Overview from Dr. John Pickup

Dr. John Pickup gave a masterful overview of patch pumps, noting their advantages and disadvantages, sharing optimism for use of simple infusion devices in type 2 diabetes (CeQur, Valeritas), and touched on accuracy and precision. He covered the two available full-featured patch pumps for type 1, OmniPod and Cellnovo, and noted several others “in development”: Debiotech’s Jewel Pump, Kaleido’s pump (CE marked as of EASD; “The company makes a lot of noise about the different colors”); Medtrum’s TouchCare patch pump (new to us from China; includes PLGS and CGM connectivity); SFC Fluidics low-cost pump (few details available; see our previous coverage); and Tandem’s t:sport (PMA submission in 2018, per the 2Q16 call). Turning to type 2 diabetes, Dr. Pickup said there is now “enthusiasm” on the heels of positive results from Medtronic’s OpT2mise trial. He asserted that the future of CSII in type 2 diabetes is in smaller, easier to operate pumps (e.g., V-Go, CeQur), as 80% of type 2s need only one basal rate (Parker et al., Diabetes Obes Metab 2008), and a bolus calculator probably isn’t needed (59% of the pump group in OpT2mise used the bolus calculator for <25% of time). Dr. Pickup mentioned other patch devices for type 2 in development, including Unilife’s Imperium (see our coverage); Debiotech’s JewelPump2; and Medtrum’s P6 Easy Patch (new to us from China; three-day disposable, mobile app controlled). He did not mention J&J’s upcoming OneTouch Via (previously Calibra Finesse), perhaps because it is bolus only. More details below.

• Dr. Pickup also covered “the somewhat controversial” topic of accuracy and precision of patch pumps. He said that results may be method dependent, and accuracy varies between pumps, at least short-term and in vitro. Results are good when manufacturers perform or sponsor studies, but there is little independent research in this area. Insulet’s OmniPod has “somewhat worse short-term accuracy in several in vitro studies, but the clinical significance of varying pump accuracy is unclear. It might be important in children, insulin sensitive individuals, and with faster-acting insulin.”

2. No Updates in Talks from Verily and IBM Watson

Talks from Verily’s Dr. Howard Zisser and IBM Watson’s Dr. Shahram Ebadollahi shared nothing new on their respective diabetes partnerships with Dexcom, Sanofi, Novartis (Verily) and Medtronic, Novo Nordisk, and ADA (IBM Watson). Dr. Zisser emphasized one point above all others: “sensors are going to be ubiquitous in our society, smaller, faster, cheaper, and used in ways we cannot predict.” Dr. Ebadollahi mostly reiterated his ATTD 2016 presentation and showed screenshots of the Medtronic Sugar.IQ mobile app (first demoed at Health 2.0 in September). Dr. Ebadollahi reiterated the ~85% hypoglycemia prediction accuracy 2-3 hours in advance, which will be incorporated in a future version of Sugar.IQ. See our detailed coverage of the app here, which is available now in a limited launch (100 MiniMed Connect users) and set for a broader launch by the end of this year.

Detailed Discussion and Commentary

New Developments in Needle Sensor CGMs

Dexcom: Pushing the Boundaries of CGM Performance

Peter Simpson (Dexcom, San Diego, CA)

Dexcom’s Peter Simpson presented highly compelling G6 pre-pivotal data (including an 8.8% MARD with zero calibrations!) and showed new impressive pictures of the second-gen sensor with Verily (smaller than an M&M in volume, non-adjunctive, zero calibration, Bluetooth-enabled). We were highly impressed on both fronts and are glad to see Dexcom pushing hard to get to a much better form factor, even stronger accuracy, no calibrations, lower cost, greater ease of use, and far more insightful data. Accuracy will always be critical for new products, though we believe these other factors will truly drive the field’s expansion. Dexcom is in a strong pipeline position based on what we saw today, and the pressure is on with Abbott launching FreeStyle Libre in the US likely before G6 and the Verily product.

• Accuracy exceeded expectations in the pre-pivotal study of the G6 sensor (n=32 adults, n=17 pediatrics): an overall MARD of 8.1% vs. YSI at three in-clinic visits (n=993 paired points) and only one calibration per day after startup over 10 days. Outliers were minimal, with 96.1% of points within 20%/20 mg/dl, better than G5 with one fewer calibration per day and nearly 50% longer wear (see below). When Dexcom ran the raw sensor data prospectively through an optimized NO fingerstick calibration algorithm, accuracy was remarkably consistent: a MARD of 8.8% and 95.5% of points within 20%/20 mg/dl (n=1,009 vs. YSI). This exceeded the company’s expectations, and Mr. Simpson said the team has been “impressed by the G6 sensor’s stability and reproducibility in manufacturing.” Three sensor lots were used in the study (key for factory calibration) and YSI measurements occurred at the beginning, middle and end of sensor wear (no further details provided). While the initial version of G6 will be one calibration per day after startup and 10 days of wear, these results support pursuing a no calibration version of G6, and Dexcom believes the sensor will last up to 14 days. Notably, data in both calibration scenarios was equally strong on day one (MARD ~10%), and the adult-only results with one calibration per day were particularly strong overall: MARD of 7.2% and 99.4% within 20%/20 mg/dl. These are outstanding pre-pivotal data and we look forward to seeing if they are confirmed in the larger 300-person G6 pivotal trial that is now underway. As of Dexcom’s 3Q16 call, a G6 launch is expected in 2018. See the detailed G6 pre-pivotal data tables below.

• Mr. Simpson and Verily Technical Project Lead Dr. Will Biederman showed impressive pictures of the second-gen Dexcom/Verily product: smaller than an M&M in volume (“thinnest CGM ever”), no fingersticks, non-adjunctive labeling, and direct-to-smartphone Bluetooth connectivity. The second-gen sensor has completed initial animal studies and the team is getting approval for clinical use. The pictures below show how truly impressive this form factor will be, and both speakers commented that it will expand CGM adoption beyond early adopter type 1s and into the type 2 population. Verily’s Dr. Biederman held up a functional prototype, drawing laughter from the crowd, since no one could actually see it in his hand. (We tried to get a peek up close, but we couldn’t persuade him to let us see it!) Neither speaker commented on launch timing, but as of Dexcom’s 3Q16 call, commercialization was expected “as early as 2020” for the second-gen sensor and in 2H 2018 for the first-gen sensor. The latter was not mentioned today, but it is expected to use the G6 sensor and have a smaller form factor than Libre (though obviously larger than this compelling second-gen version). It’s worth noting that this partnership is only 15 months old (signed in August 2015), signaling remarkable progress and what seems like a great culture fit between the companies.

• Dr. Biederman said the companies have produced purpose-built microchips for CGM that include sensing, data processing/storage, and the wireless radio on a chip smaller than a grain of sand (barely visible with the naked eye). Verily clearly brings tremendous semi-conductor manufacturing and miniaturized electronics experience to the table.
• Though the G6 pre-pivotal study showed data from 10-day wear, an earlier slide noted the sensor duration will last “up to 14 days.” This is the first time we have heard that and assume Dexcom will eventually pursue 14-day wear after the initial 10-day wear is approved. G6 will also have the new 30% thinner transmitter and the simple push button applicator (currently under FDA review), along with acetaminophen blocking. See picture below.

• Mr. Simpson began and ended the presentation with a true story of his dad, who used orals to manage his type 2 diabetes (sulfonylurea and metformin), and after going on CGM, dropped his A1c by ~1%, lost 25 lbs, eats better, and gets more exercise. Just as critical, Mr. Simpson’s dad was experiencing dangerous lows on an SU, including one while driving (Mr. Simpson noticed it remotely with Dexcom Share). As we’ve long said, CGM is not a technology for just insulin users! Mr. Simpson concluded with a great quote from his dad: “Diabetes has been a series of losing battles for 20 years. CGM has given me my first win.”
• We were glad to see Verily on stage discussing its broader diabetes strategy too: Big Data analytics and machine learning; a scalable platform for wireless sensing; and miniaturized electronics and flexible biocompatible encapsulation/substrates. The company sees itself at the intersection of Huge Problems, Breakthrough Technology, and Radical Solutions – we like the sound of that. Dr. Biederman mentioned “ubiquitous sensing” and extremely small, low cost, and burden-free devices that are cloud connected to provide better data insights. No other details were shared on the Sanofi (Onduo) or Novartis diabetes partnerships.
• Mr. Simpson showed a slide of Dexcom’s “Diabetes Management” partners: Animas, Insulet, Tandem, Bigfoot, TypeZero, and Beta Bionics. The latter three have not been routinely shown in pump partner slides to date and we were glad to see this formalized, since Dexcom has devoted less air time to AID products in recent quarterly updates. (Of course, the timing and submission of these products is a pump partner issue.) The inclusion of TypeZero quickly follows last week’s news that the NIH-funded IDCL study is starting up and will serve as the pivotal trial for Tandem’s second-gen automated insulin delivery device.

Artificial Pancreas – Recent Innovations in Algorithms

inControl: a Closed Loop Control System for Type 1 Diabetes

Patrick Keith-Hynes, PhD (TypeZero Technologies, Charlottesville, VA)

TypeZero’s Dr. Patrick Keith-Hynes shared snappy looking screenshots and discussed the company’s three commercial software products in more detail than ever: (i) inControl AP to run closed loop from a smartphone or integrated into a pump (about to start the IDCL study with Tandem and Dexcom); (ii) inControl Advice, an app that gives smart open-loop dosing advice for MDIs or pumpers; and (iii) inControl Remote Care Platform for HCPs to optimize insulin therapy and remotely monitor patients on pens or pumps. TypeZero has taken the UVA algorithms, modified them, and brought them to what looks like a commercial-ready state. The screenshots we saw today for the first time looked very sharp (see below) and were a far cry from the less modern original DiAs interface. We’re particularly glad to see TypeZero working on bringing its artificial pancreas algorithms into MDI users with type 1 and type 2 – the potential for smarter insulin dose titration is massive in our view. No timing or commercial details were shared, but we are very optimistic about the potential for smarter open loop therapy. Notably, the Advice app has a chat-like, text-message interface, getting away from pushing buttons (similar to what Medtronic/IBM Watson are doing with Sugar.IQ). Dr. Keith-Hynes showed the smartphone version of inControl AP for the first time we’ve seen, which modulates basal insulin and can issue automated correction boluses, a key difference from basal-only algorithms like the MiniMed 670G. As a reminder, the upcoming 240-patient, NIH-funded IDCL study will support an expected late 2017 PMA submission of a Dexcom G6-integrated Tandem t:slim X2 pump with TypeZero’s inControl AP algorithm built into the pump. Patients will use the system for six months. (The study’s first pilot phase will use Dexcom’s G5, inControl AP running on a smartphone, and a Tandem pump.) See pictures and more details below.

• The inControl AP app home screen showed the current CGM value in green (120 mg/dl), with some simple text indicating predictions and system status. inControl AP includes a basal attenuation module, a basal increase module, a hyperglycemia correction module, and a user input module. The algorithm makes a new dosing decision every five minutes, and overnight, gradually treats to a target of 120 mg/dl by morning.

• The baseline DiAs algorithm has been tested in over 200,000 hours of clinical trials and the commercial version (inControl AP) is in two ongoing studies: Project Nightlight (n=84, 11 months of use) and Safety at Home (n=22, five months of use). The IDCL study, which will support a Tandem FDA submission, will start this month and support a late 2017 FDA submission.
• InControl Advice is a smartphone app that provides open-loop insulin dosing advice and pattern recognition for those on MDI or pumps and CGM. The app includes a smart bolus calculator based on TypeZero’s artificial pancreas algorithms, physical activity advice, interactive risk monitoring, and clinician messaging. The user interface is chat-based (like a text message), a move away from buttons – yes! Dr. Keith-Hynes positioned it as a “Smart assistant” that is “context aware,” similar to how Medtronic/IBM Watson are positioning Sugar.IQ. The app will integrate with pumps, connected insulin pens/caps (once they are available), and CGM, closing the loop on whether an insulin dose recommendation was actually carried out. No FDA submission timing or commercialization plans were shared, but we see this as very exciting for MDI users in particular – leveraging the intelligence of artificial pancreas algorithms, but without requiring a pump.

• inControl Remote Care Platform is a tool for clinicians to optimize insulin pump or MDI therapy. The platform conducts risk-based therapy optimization using 2-4 weeks of CGM, BGM, and meal data to propose changes in insulin therapy, including basal, carb ratio, and correction factor. inControl Remote Care Platform calculates a patient-specific model, identifies glycemic risk zones, and proposes adjustments to insulin therapy profiles. The screenshot (see below) looked very clean.

• Interim results from a small pump-based trial of TypeZero’s decision support system (n=11) were shared at ADA 2016 (see picture below); full results are pending, and pen-based results are ongoing. Dr. Keith-Hynes characterized the effect size as “modest” in the early results from this small initial study: time in range (70-180 mg/dl) improved from ~52% to ~65%, while time <70 mg/dl declined from ~3% to 2%. Metrics of glucose variability (risk index, ADRR) also declined. The RCT tested what sounded like a patient-facing decision support system, including automated treatment optimization, pre-exercise advice, and a CGM-based bolus calculator.

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