Diabetes Technology Meeting 2013

October 31-November 2, 2013; Burlingame, CA; Day #3 Highlights - Draft

Executive Highlights

Hello from Day #3 of the 2013 Diabetes Technology Meeting. Saturday ended on a strong note, with new CGM data, a valuable session on insulin delivery, a closer look at Dexcom Share and Tidepool, and even a talk on smart insulin.

Perhaps the most buzzworthy presentation of the day came from Roche’s Dr. Matthias Axel Schweitzer, who shared new data on the company’s novel CGM sensor – accuracy continued to look strong, with the sub-10% MARD in line with data previously presented at ADA. Notably, Dr. Schweitzer also presented data from a new small head-to-head study against the Dexcom G4 Platinum. Ten type 1 patients wore two Dexcom G4 sensors and two Roche sensors simultaneously. The Roche sensor slightly edged out the Dexcom G4 on MARD: 8.6% for Roche vs. 10.9% for Dexcom. Specifics were very light on study methodology, so we look forward to more detail in the future. Also on the novel CGM front, Senseonics’ Dr. Andrew DeHennis presented new 180-day data (n=9) on the company’s implantable CGM. Overall MARD was a very solid 12.8% over the full six-month period. Accuracy was consistent over study length, with just one of the nine sensors dropping out at day 100.

On the app side, a highlight was Mr. Jorge Valdes’ (Chief Technical Officer, Dexcom, San Diego, CA) talk on Dexcom Share. He showed a great schematic of how the remote monitoring system will work, emphasized the customizability of push notifications to caregivers, and noted positive results from human factors testing. The concluding part of his talk positioned Dexcom Share as “another approach to preventing nocturnal hypoglycemia,” perhaps an indirect comparison to Medtronic’s recently approved MiniMed 530G.

Speaking of Medtronic’s new threshold suspend system, we saw valuable new subgroup data from the ASPIRE in-home study of the MiniMed 530G. Notably, the study’s overall pooled results were consistent in subgroups of patients 16-24 years, 25-50 years, and 51-70 years. Further, it appeared that the greatest benefit on hypoglycemia mitigation occurred in the 51-70 year-old age group, leading Medtronic’s Dr. Scott Lee to conclude, “This [device] is very helpful and relevant to older patients.”

On the data side, we enjoyed a presentation on blip, the first application that sits on top of Tidepool’s open-source platform. The novel app provides “intuitive, interactive visualizations of data” from multiple sources, including insulin pumps, CGMs, and activity monitors. We were particularly impressed by the web-based software, which combined lots of data into a sleek and highly responsive user interface. 

Last, Dr. Zhen Gu (North Carolina State University, Raleigh, NC) discussed the exciting and ambitious topic of smart insulin. Early rodent studies conducted with smart insulin candidates in his lab demonstrated clear glucose-dependent insulin release, and statistically significant glucose-lowering for up to ten days. His lab’s work does not seem ready for prime time yet, though we were interested to hear Dr. Gu discuss other interesting potential options as well – insulin release triggered by infrared or ultrasound, as well as smart glucagon that could be used in tandem with smart insulin.

We’re already looking forward to DTM 2014, which is slated for November 6-8 in Bethesda, Maryland.

Table of Contents 

Detailed Discussion and Commentary

Novel Technologies for Continuous Glucose Monitors

From Self-Monitoring of Blood Glucose to Continuous Glucose Monitors: Sensor Performance is Key for Successful Clinical Use

Matthias Axel Schweitzer, MD, MBA (Roche Diagnostics GmbH, Mannheim, Germany)

Dr. Matthias Schweitzer outlined the new Roche continuous glucose monitor (CGM) sensor technology and clinical results from recent trials – significantly, Dr. Schweitzer presented results from a head-to-head study against the Dexcom G4. The 10 patients with type 1 diabetes wore two Dexcom G4 sensors and two Roche sensors simultaneously. The Roche sensor slightly edged out the Dexcom G4 on MARD: 8.6% vs. 10.9% for the Dexcom G4. Roche also looked slightly better on precision as measured by PARD (differences between paired CGM readings) – 6.2% vs. 7.4% for Dexcom. We thought it was noteworthy to see a true head-to-head trial of these two sensors, a rarity in CGM studies. However, we were disappointed not to see any discussion of methodology or secondary results. We think full transparency on data collection (calibration, reference method, start-up time, source of sensors, etc.) and study details is in the absolute best interest of everyone in the diabetes ecosystem. We hope to see additional results in the future, perhaps at ATTD. In addition, we heard an update to the Roche sensor data reported earlier this year at ADA (see page 66 of our full report); the updated pooled results gave the Roche sensor an MARD of 9.4% and an average PARD of 7.8% – quite similar to data from the first 30 patients. Dr. Schweitzer also outlined how the company developed its new sensor, remarking that the team has optimized the pattern on the six-millimeter sensor.

  • Dr. Schweitzer highlighted the design of Roche’s new sensor: six millimeters long with eight spots on one side to make up a working electrode; the other side houses counterelectrodes. The working electrode runs 350 mV vs. the reference electrode.
  • To give the audience “an idea of where [Roche] is,” Dr. Schweitzer showed a slide demonstrating that the new Roche sensor had the lowest MARD among four other sensors: the Medtronic Guardian RT Enlite (16.4% MARD), the Abbott Freestyle Navigator (12.4% MARD), the Dexcom Seven Plus (16.7% MARD), and the Dexcom G4 (10.9% MARD); Roche’s new sensor slid under the others at 9.4%. We would emphasize that it’s still hard to compare the Roche results head-to-head to other sensors at this point – while it was not specified, we’d assume the Roche sensors were built at a small scale in the research setting. This is very different from high-volume manufacturing used for the other sensors in the comparison.
  • Bringing the audience back to the big picture, Dr. Schweitzer emphasized Roche’s three aims with new diabetes technology: 1) to serve diabetes management and outcomes; 2) to embed glucose information from BGM and CGM and other technologies into personalized diabetes management; and 3) to improve quality and analytical performance. Giving an example of how Roche could embed this technology, Dr. Schweitzer highlighted connectivity between CGM, blood glucose meters, a smartphone app, pumps, and data management software.

Long-Term Clinical Feasibility Demonstration of an Implantable Flourescence -Based Continuous Glucose Monitor

Andrew DeHennis, PhD (Engineering Manager, Senseonics, Germantown, MD)

Dr. Andrew DeHennis shared the latest on Senseonics CGM system, discussing its design and reviewing six-month data from a small UK study. The system includes an implanted subcutaneous CGM sensor (3 mm x 15 mm), a body-worn transmitter (also provides power to the sensor), and a mobile smartphone app. Dr. DeHennis showed new data from a nine-patient, 180-day study (results from only four patients were shown at EASD 2013). Patients calibrated the device with two SMBGs per day and accuracy was evaluated vs. YSI in 10 clinic visits (one every two to four weeks). Overall MARD was an impressive 12.8% over the full six-month period (vs. YSI measured at in-clinic visits every two to four weeks), ranging from a low of 10.9% in one patient to a high of 15.1% in another patient. Looking at all patients, 78% of points were in Zone A of the Clarke Error Grid, 18% were in Zone B, and 4% were in Zone D (n=2,001). Notably, accuracy over time was also consistent (historically, an issue with implanted sensors) – MARD was 12.9% on days 1-30 and concluded at 11.2% for days 151-190. Eight of the nine sensors lasted 175 days or longer, with a single sensor failing at day 100. Overall, we found these results encouraging, as the accuracy looks comparable to Dexcom’s G4 Platinum and what Medtronic has reported in the label for the Enlite (accuracy has been higher in independent studies). That said, manufacturing with precision and accuracy at scale is a whole different story, and we look forward to larger studies of Senseonics’ sensor. We think a group of patients would be interested in this six-month implantable CGM product concept, assuming it is accurate and reliable, that the on-body transmitter is not too cumbersome (it doesn’t appear to be), and that the implantation/explantation procedure is as minimally invasive as advertised.

  • The Senseonics system has three main components: a subcutaneous inserted implantable sensor (3 x 15 mm), a body worn transmitter, and a mobile medical app running on a smartphone. Senseonics also supplies insertion tools and web-based data management. Once the sensor is implanted, the only thing a user deals with is the body worn transmitter and the mobile app. The app provides full access to glycemic levels, allows event logging, and permits cloud communication with care providers.
  • The on-body transmitter powers the sensor through near-field communication and relays sensor glucose information to the smartphone app through a Bluetooth LE link. The transmitter is attached via an armband or replaceable adhesive – it measures 1.4 cm x 4 cm x 4 cm (0.55 in x 1.6 in x 1.6 in). It can be taken off at any time and put back on without doing anything on the app. The body-worn transmitter takes 15 minutes to charge from fully drained to fully charged (Dr. DeHennis suggested “while showering” – the transmitter will be rain/splash proof, though Senseonics will not recommend wearing it in water). All glucose calculation and trend information is stored on the transmitter. It also includes an LED indicator, a vibration motor (alerts for hypo- and hyperglycemia), and a single button interface for powering, pairing the device to the phone, and to suppress an alarm. Notably, the transmitter has a built in three-axis accelerometer, which sends activity tracking and sleep monitoring information to the phone. Very valuable integration in our view.
  • Senseonics’ subcutaneous implanted sensor is optically based (boronic acid). Notably, it has the highest sensitivity in the hypoglycemia range. The sensor has non-volatile memory storage (calibration, UDI traceability) and uses a near-field communication interface (ISO 15693). It has no battery and thus lies dormant without the transmitter. The sensor has on-chip digitized signal conditioning that enables measurement multiplexing (e.g., monitoring of photodiode reference currents, temperature). It’s also an equilibrium binding sensor, meaning no oxygen or glucose is consumed.
  • Dr. DeHennis shared accuracy and reliability data from a small six-month UK study of the sensor in nine patients with diabetes. Sensors were inserted into the upper arm, with in-clinic visits every two weeks: clamp studies (days 14 and 42), four-hour stays (days 28, 56, 70, and 84), and eight-hour stays (days 98, 126, 144, and 182). The reference instrument was YSI. The CGM was calibrated by two SMBG measurements per day throughout the study. All data was prospectively calculated. Since this was an early feasibility study, real-time results were blinded to the user, but were calculated by the transmitter. A pivotal study is planned to provide real-time glucose readings to the user.
  • Overall, MARD was 12.8% over the six-month study (vs. YSI measured at in-clinic visits every two to four weeks), ranging from a low of 10.9% in one patient to a high of 15.1% in another patient. MAD vs. YSI was 13.2 mg/dl ranging from a low of 8.1 mg/dl to a high of 24.1 mg/dl.


Insertion Duration



Overall (n=9)

172 days (mean)


13.2 mg/dl


  • Notably, accuracy over time was quite consistent and maintained out to 180 days. Dr. DeHennis indicated that an even longer implant time might be possible.

Days Since Implant














  • Eight of the nine sensors lasted 175 days or longer, with a single sensor failing at day 100. For that sensor, in-vivo sensitivity dropped to 20%, though upon explantation its sensitivity was still a high 92% (“indicative of significant fibrotic capsule growth”). The concluding slide noted that Senseonics is clinically evaluating sensor configuration to further moderate fibrous encapsulation.


In-Vivo Sensor Sensitivity

Explanted Sensor Sensitivity

Mean/Median (Range)

68%/70% (20-98%)

82%/83% (65-97%)


Use of Low Glucose Suspend to Mitigate Hypoglycemia

Scott Lee, MD (Medtronic Diabetes, Northridge, CA)

Dr. Scott Lee reviewed the ASPIRE in-home study of Medtronic’s MiniMed 530G, first presented at ADA 2013 (page 45 here) and simultaneously published in the NEJM (Bergenstal et al., 2013). He shared new breakdowns of the efficacy data by age – importantly, the overall pooled results were consistent in subgroups of patients 16-24 years, 25-50 years, and 51-70 years. Interestingly, it appeared that the greatest benefit on hypoglycemia mitigation occurred in the 51-70 year-old age group. Dr. Lee connected this point to T1D Exchange data, reminding the audience that severe hypoglycemia in the Exchange is most prevalent in those over 50 years old (“This [device] is very helpful and relevant to older patients”). Another point of nuance on age concerned those 16-24 years old – this group had the highest baseline and three-month A1c values and the most modest reductions in mean AUC and the rate of nocturnal hypoglycemia events. Dr. Lee concluded, “This group should benefit from insulin delivery automation, but other strategies are also needed to address youth issues and the transition to young adulthood that may impact diabetes management and outcomes.”

  • “To our surprise, there was a lot more nocturnal hypoglycemia than we thought.” In explaining the design of the ASPIRE in-home study, Dr. Lee described the trial’s run-in period in detail. To be randomized into the study, enrolled participants had to have had two or more nocturnal episodes (10 pm-8am) with a sensor glucose <65 mg/dl for >20 minutes and no evidence of user-pump interaction. Given these criteria, the team was expecting a fairly high screen failure rate. However, of the 414 patients enrolled, a surprising 247 were randomized (60%).
  • “In my clinic, we have a hard time getting patients from an A1c of 7.5% into the 6.8% range without increasing hypoglycemia.” We thought this was an astute point and a reminder of how more automated insulin delivery can help those who are fairly “well-controlled” already.


Direct Electron Transfer (DiET) Technology: The 3rd Generation of Continuous Glucose Monitor Sensor

Koji Sode, PhD (Graduate School of Engineering, Tokyo, Japan)

Dr. Koji Sode introduced the audience to direct electron transfer (DiET) glucose sensor technology, the “3rd generation” of CGM sensors. As background, most current CGMs use the “1st generation” of sensors, which depend on glucose oxidase and a redox reaction with hydrogen peroxide to generate an electrical signal. Dr. Sode noted that Abbott appears to have successfully developed 2nd generation sensors, which can use glucose oxidase or glucose dehydrogenase as the redox intermediate. What distinguishes 3rd generation sensors is that they involve no intermediate — rather, they involve the direct electron transfer from the FAD enzyme (which oxidizes glucose) to the sensor, meaning that the process is not dependent on oxygen levels. A major challenge in the development of DiET sensors is that the redox center of FAD is buried deep within the protein, but a variant of the protein (FADGDH) has been isolated that is conducive to direct electron transfer. The current DiET electrode being studied is 0.25 mm wide, and has demonstrated stable and sensitive operation for up to seven days. The data Dr. Sode presented from a very early-stage in-vivo human study (in a healthy volunteer) was somewhat disappointing. Even with once-daily calibration, sensor readings deviated from reference measurements; specifically, the sensor had a tough time tracking the full magnitude of postprandial glucose spikes. Sensor delay was approximately eight to nine minutes. While more work is clearly needed to optimize this technology’s performance, we look forward to monitoring the progress of this new generation of sensors, especially given its more direct and oxygen-independent mechanism. Dr. Sode noted that his lab has presented data to a potentially interested SMBG company in Japan (we assume he was referring to ARKRAY).


Panel Discussion

Q: When we publish papers, we always have to look at studies’ weaknesses. What is the main weakness of each of your technologies? This may be a tough one to discuss in public.

Dr. Scott Lee (Medtronic Diabetes, Northridge, California): When I showed that data on recovery from hypoglycemia, what was interesting was that sensor-augmented pump therapy or threshold suspend isn’t therapy you can just slap on a patient. You still need lots of patient education and training. If patients over-treat their hypoglycemia they can still run into problems, particularly if insulin is suspended. We’re not to the point where you can press a button and let it go. We need to say that these technologies require education.

Dr. Matthias Schweitzer (Roche Diagnostics GmbH, Mannheim, Germany): I would say that the weakness in this system is more or less the same as the weaknesses of other sensors available.

Dr. Koji Sode (Graduate School of Engineering, Tokyo, Japan): The weakness in my data is that it is a new technology and new principle. We have introduced a technology that has never yet been extensive clinical studies. Experience and exposure to the real world is an advantage we don’t yet have.

Dr. Andrew DeHennis (Senseonics, Germantown, MD): Co-location of the sensor and the transmitter is required with our device.

Q: For Dr. Lee, could you be more specific on when and why the threshold sensitivity system allowed hypoglycemia? Was it a rapid range of change in the sugar? Was it errors in the sensor?

Dr. Lee: We’re doing another analysis on this, but most of the hypoglycemia incidents we saw were from boluses that were taken later at night. It looks like people who were coming down rapidly were the ones who were the hardest to mitigate the hypoglycemia for. People who came in slower with less of a rapid rate of change we were generally able to correct. For people with moderate rates of change, the threshold suspend would go off but you would still see some hypoglycemia in 60 mg/dl range, which would stabilize and rise. If a patient then woke up and self-corrected, they would often overcorrect. But I wouldn’t call that problem specific to this system.

Q: Dr. Lee, do you have data on whether the marked difference between groups in your study was because the alarms were not optimally set in the control group? The alarms on the Medtronic device are notoriously difficult to hear.

Dr. Lee: We’re in the process of analyzing data on that, and we hope to submit and share that data soon. We know that people in both groups did set their own alarms. It was really a hands-off study, but the investigators did have the opportunity to suggest changes if they felt patients were having problems.

Dr. Howard Wolpert (Joslin Diabetes Center, Boston, MA): When you compare this data with data from studies on low glucose suspend, and studies with the Dexcom G4 sensor, which is more accurate, it’s like comparing apples and oranges.

Q: With the eight working electrodes on the Roche sensor, did you average the eight signals as a straight average or did you weight them? Generally, I’m curious about variability between electrodes.

Dr. Schweitzer: The current that we use to detect glucose is the sum of the electrodes.

Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): I see how we’ve progressed from a single sensor with glucose oxidase all the way to third generation sensors, multiple sensors for better accuracy, low glucose suspend, or new methods like the one used by Senseonics. For Dr. Sode, it’s important that you document that you have less interference with your third generation sensor, since that is the main advantage. Once you document that, you can show that your system is better than first generation sensors. To Dr. Scheister and Dr. DeHennis, you both said that your sensors are better with hypoglycemia. Do you have specific data on that?

Dr. DeHennis: We’re seeing a MARD around 13%, cumulatively, with the entire data set.

Apps for Diabetes

SHARE: A Remote Monitoring Accessory for the Existing G4 PLATINUM CGM System

Jorge Valdes, MBA (Chief Technical Officer, Dexcom, San Diego, CA)

Mr. Jorge Valdes provided a valuable overview of Dexcom Share, which is currently under review by the FDA (the PMA supplement was submitted in July). He showed a great schematic of how the system will work – 1) the G4 Platinum receiver will plug into the Share cradle; 2) data will be sent from the cradle via Bluetooth to the Share app on a nearby iPhone/iPod touch (e.g., at a bedside); 3) data will be sent to the cloud via Wi-Fi or cellular; and 4) up to five caregivers will be able to remotely receive the CGM data on the Dexcom “Follow” app (see a picture on twitter). (Dexcom is starting with Apple devices, though we assume Android is in the pipeline.) Notably, push notifications and alerts to caregivers will be customizable (e.g., a caregiver could only receive low alerts). Sensor glucose and trend information will also be accessible on demand on the Follower app. Mr. Valdes highlighted that results from human factors testing “were positive” and got “high marks,” though he did not share any quantitative specifics. The concluding part of his talk positioned Dexcom Share as “another approach to preventing nocturnal hypoglycemia,” perhaps an indirect comparison to Medtronic’s recently approved MiniMed 530G. Mr. Valdes’ explained how the Share system alerted a patient’s wife to his low blood glucose while on a business trip; she ended up calling the hotel when her husband did not pick up her phone calls (“He is first in line waiting for the FDA to approve Share”). Mr. Valdes’ concluding slide summed up what we think is one of the best parts about this emerging area: “Remote monitoring has the potential to extend the utility of CGM.” We expect to see especially strong enthusiasm among parents. What will be most interesting is to see if Share expands the CGM market, or merely offers greater piece of mind to those already using CGM.

  • Mr. Valdes noted that the concept of Share was inspired in part by the results of a study of remote monitoring done by Stanford’s Dr. Bruce Buckingham in the summer of 2012. A slide displayed a picture of Dr. Buckingham watching glucose in real-time from a central cabin in the middle of night. Mr. Valdes joked that it wasn’t the intention to have Dr. Buckingham do the monitoring!
  • Dexcom Share will have two modes for remote monitoring: an automated event-based mode (push notifications) and a mode for manual on-demand viewing (current CGM value and trend). Mr. Valdes emphasized that the push notification mode will be programmable and customizable. Patients can have up to five followers, a number that Dexcom picked after doing research. However, “there is technically no limit” on the number of followers, so it could be expanded in the future.
  • Results from Share’s human factors testing were “positive” and users found that Share was “easy to use.” Test participants gave Share “high marks for its potential clinical utility.” Several use scenarios were explored, including: a young adult with type 1 diabetes at college; an adult with type 1 diabetes on travel for work; a parent of a child with diabetes at home; and a parent of a child with diabetes at a sleepover.
  • A slide noted that initial use of Share will be for “home (bedtime) remote monitoring” – we assume this refers to its official labeled indication. That said, we imagine patients and caregivers could (and probably will) use it in a variety of settings (at work, at school, etc.).
  • Share will initially be available for Apple devices, though we assume Android is in the pipeline. From a development perspective, we have heard that building apps for Apple’s iOS is easier, since there are far fewer phones to consider in testing. Android is the dominant smartphone platform in the overall US market (52% market share, per data from Glooko), though we wonder if Apple’s iOS is more widely used in the early-adopter, tech-savvy Dexcom customer base.
  • As a reminder, Dexcom management has positioned Share as a baby step towards the Gen 5 mobile platform (see our Dexcom 2Q13 report). Share has been deliberately designed to make sure the G4 Platinum receiver is still the primary medical device, with the phone(s) as a secondary display(s). Still, Dexcom “did submit a rather large filing” to the FDA that contains all the software validation and verification (e.g., that the CGM signal gets through, that the warnings are what they claim to be, etc.). From a product pipeline and development perspective, the incremental approach is very smart. It would be a big step for the FDA to go straight from Gen 4 (standalone receiver) to Gen 5 (smartphone as the receiver) – Share is a great hybrid in the interim. 
  • As we understand it, CGM data will not be stored in the cloud when Share comes out – the goal for now is remote monitoring. Dexcom’s cloud-based SweetSpot data management platform is in the company’s product pipeline, though we have not heard a timeline update on this product in some time. We expect it will be out once the Gen 5 mobile platform is completely on the market – as of Dexcom 2Q13, the goal was to roll Gen 5 out “in several stages over the next two years.”


“blip”: A Novel App for Diabetes

Jenise Wong, MD, PhD (University of California, San Francisco, San Francisco, California); Howard Look (Tidepool, Palo Alto, CA)

Dr. Jenise Wong introduced the audience to blip, the first application that sits on top of Tidepool‘s open-source platform. blip provides “intuitive, interactive visualizations of data” from multiple sources, including insulin pumps, CGMs, and activity monitors. The app facilitates ongoing conversations between health care providers and their diabetes patients using Facebook and other messaging systems. HCPs, family members, and anyone else one chooses can communicate via Facebook and social media about a logged event. For example, parents can use Facebook to ask a child why their last blood glucose reading was so low, to discuss the event with a healthcare provider, to look at the pump/CGM/BGM/activity data associated with the low, and annotate what happened (e.g., went for a walk and over-bolused for lunch). Blood glucose values are tracked on a main-screen grid that has weekdays across the top and hours down the sides. The points are color-coded such that green means in range, yellow means high, and red means low, and specific days can be expanded to show more detail, such as meals, boluses, and additional comments. blip is made in a collaboration with Tidepool, a non-profit whose goal is to build an infrastructure for a new generation of smart diabetes management apps. We were very impressed by the software, which combined lots of data into a sleek and highly responsive user interface. 

  • blip is a single-page app built on backbone, d3, etc. and (currently) uses the Facebook graph API for a social networking mashup and authentication. The feed from a private Facebook group is overlaid onto diabetes data displayed with d3. It provides the data-driven awareness and lets patients use Facebook as their “log.” The backend is node/express.
  • Tidepool is a secure, open-sourced, cloud-based platform that will collect data from a number of areas, including CGMs, BGMs, social media, and even the Fitbit. In the words of the organization, “We are here to create a world where diabetes data is: acquired seamlessly and in real time; stored and managed in a cloud-based platform; made available for providers and patients to look at anytime; and presented in a standardized, intuitive way across all devices.”
    • We think this very ambitious goal is fantastic for patients and HCPs – having such an integrated tool would help add context to diabetes data, critical for making therapeutic changes.
    • Mr. Howard Look leads the non-profit. See the organization’s website for a good description, or read its strategy and architectural vision.
    • Looking to integrate further, Tidepool hopes to pull data from Medtronic, Animas, Dexcom, Tandem, and Insulet devices.
  • Dr. Wong highlighted that the University of California has been analyzing the frequency of data downloads – the stats are discouragingly low. Dr. Wong pointed out that the data is often overwhelming and difficult to understand. To add insult to injury, there is no common software among different devices.
    • The majority of CGM users (64%) never download their data outside of the provider’s office (Wong, Neinstein, and Adi, unpublished data). Similarly, 52% of insulin pump users and 67% of glucose meter users never download their data outside of the office.
    • Only about 10% of CGM users download their device weekly and 35% never download it at home (Wong et al. 2013 in preparation). Only 11% of insulin pump users and 14% of glucose meter users download once a week. A more striking statistic from 2012 shows that <3% of patients download data from their glucose meter and review it once a week.
    • Even when patients download their data, they are not always looking at it. Of those who downloaded their data, only 40% actually reviewed the pump/CGM/SMBG data.

GDm-Health: Remote Monitoring and Treatment of Gestational Diabetes

Oliver Gibson (University of Oxford, Oxford, UK)

Mr. Oliver Gibson discussed Oxford’s mHealth monitoring system for gestational diabetes (GDM). He noted that GDM is a time and resource-intensive disease for both patients and providers – for instance, patients in the UK consult with providers every two to four weeks. The system designed by Mr. Gibson and the team at Oxford utilizes a LifeScan OneTouch UltraEasy BGM connected to a Polymap Bluetooth meter accessory that transmits results to an Android phone with a custom application. The app, in turn, transmits the data to the patient’s care team, which can respond to the patient via text message. The application also allows the user to give each reading a meal tag and to compare day-to-day results for a given tag (as well as view a chart with full-day readings). Dr. Gibson acknowledged that not every patient may view this data, but stated that the graphs could help more proactive patients manage their own care to a greater extent. [This is one of the very key questions in diabetes data – how do you build a system that engages and encourages disinterested patients to use it? We think the key is hassle free upload and actionable therapy recommendations that are auto-generated.] He noted that the system has been well received by patients and providers, who appreciate that the system improves efficiency by eliminating the need for the patient to read results back to the provider. So far, 49 patients have used the system, resulting in 621 patient-weeks of data and over 17,000 blood glucose readings. An ongoing randomized control trial enrolling 175 patients will examine the system’s impact on A1c and mean blood glucose. Of course, getting the data is only one piece of the puzzle – using it to drive therapy improvements is a whole different and more challenging ball game. 


Panel Discussion

Dr. Barry Ginsberg (Diabetes Technology Consultants, Wyckoff, NJ): I’d like to congratulate you, Dr. Wong, and the rest of the Tide Pool group for using open source. Nobody does that anymore. Everyone tries to keep their product proprietary to gain market share and nobody does.

Q: For Mr. Gibson, I have a very pragmatic, low-tech question from the perspective of a clinician. It’s great that you’re focusing on an area that doesn’t get enough attention. A lot of those women will be feeling quite nauseated. I was amazed that you had those prandial markers so well controlled. Might your patients be lying to you or are they different than the ones that we see elsewhere? Also regarding languages, GDM is a multiethnic language and clinics often struggle with language barriers.

Mr. Oliver Gibson (University of Oxford, Oxford, UK): The proof of the pudding is going to be in the randomized control trial that is underway. We look forward to showing that we can provide equivalent care with half the visits. Hopefully we can do better than conventional care. As a side project, we’re going to look at glycemic variability and other issues associated with GDM. You’re right about languages; our app could be switched quickly between languages.

Q: Could providers be held liable for not looking at the data?

Mr. Valdes: [Dexcom Share] was never an intention to put providers on the spot and put them in a position where they would be liable. We generally tested it with family members. Also, you have the choice to not receive alerts, and you will only see results when you want on demand. Regarding power demands, the battery can handle hundreds of charges and not lose capacity. I believe that in testing, after hundreds of cycles, there was only a 5% loss in capacity. Running the system overnight does put a strain on the iPhone, so you will need to plug that in overnight.

I do think that his area is very exciting. There’s going to be huge changes in the way people receive their treatment. I think you’ll start seeing a separation between device companies and software companies out to add further value. When we look at the future of Dexcom, we’re looking at it from an open point of view. We will share our data with those who need it. You’ll have device companies with data, and hundreds of companies tapping into that data. [applause]

Q: Can Tidepool read Dexcom data?

Mr. Howard Look (Tidepool, Palo Alto, CA): Yes, it can.

Q: I’m concerned about data overload. We’re compiling all this information, and we need to develop a way to automatically sort and assess it.

Dr. Wong: I agree. The first step is getting all the data, and now the next step is the more sophisticated and complicated step of picking up the things that the physician needs to know in a safe and legally covered way.

Mr. Valdes: The doctor doesn’t have time to sit there and look over pages of data. We have something called Portrait that will take the mobile data and add value. You want to provide data that you can view in thirty seconds you know what is going on with the patient. If a patient has problems regularly around 3:00 PM in the afternoon, you want to communicate that clearly to the provider. That is far preferable to the provider looking over chart after chart after chart.

Q: What sort of metrics do you foresee as being useful in determining patient engagement?

Dr. Wong: I think the validated scales of self-management and efficacy are useful. We’re planning to work with behavioral psychologists, who have more expertise in measuring and evaluating engagement. For me as a pediatrician, engagement is especially important.

Mr. Look: From a software platform perspective, the way Facebook and Zynga track engagement is by how long a cohort sticks with the service, and how long people use the site each day. In a similar way, we will have answers to those questions.


New Insulin Delivery Technologies

Intraperitoneal Insulin Delivery

Howard Zisser, MD (Sansum Diabetes Research Institute, Santa Barbara, CA)

Dr. Howard Zisser, who is now “bi-coastal” due to his recent appointment as Insulet’s medical director, gave a valuable overview of intraperitoneal insulin delivery. He started with historical background, noting that the absorption benefits of the peritoneum were outlined in literature as early as the 1890s. After a brief discussion of implantable pumps and the EVADIAC study (Lancet 1994; Diabetes 1995), he shifted focus to Sansum’s recent closed-loop work using Roche’s DiaPort. The ten-patient study compared subcutaneous closed-loop delivery to intraperitoneal (IP) closed-loop delivery. The data were first presented at ADA 2013. Dr. Zisser emphasized that the trial really pushed the system, as the study’s three meals were quite large (40-70 g carbohydrates) and the system was fully automated with no meal announcement. Average blood glucose was 190 mg/dl following subcutaneous delivery vs. 150 mg/dl with IP delivery. Overall, the postprandial glucose peak was 58 mg/dl lower with IP delivery vs. subcutaneous delivery. Additionally, patients on IP delivery spent 25% more time in range (70-180 mg/dl) during the postprandial period. IP delivery also allowed faster recovery from hypoglycemia after pump suspension. Positive results aside, Dr. Zisser was confident that the system “can do much better with meal announcement.” We certainly agree.

  • Sansum’s closed-loop study used Roche’s DiaPort with an Accu-Check Spirit Combo pump and a Dexcom Seven Plus (the G4 Platinum was not available for this study). The system runs on Sansum’s Artificial Pancreas System. The pump catheter sits in the intraperitoneal space. Within seven or eight minutes of delivery, there is almost 100% uptake into liver (“Where we want it to act”).
  • “Almost any control algorithm can do overnight control quite well.” Dr. Zisser noted that overnight control between IP and subcutaneous delivery was not clinically different.
  • A fully automated closed-loop is a very commendable and ambitious goal, though this study illustrates how challenging that might be in practice. Even with lightning fast IP insulin delivery, normoglycemia was not achieved in the postprandial period. We think meal announcement in the closed-loop setting is a fairly small burden and price to pay in the near-term as we await faster-acting insulin.
  • That said, MannKind’s Afrezza would certainly be an interesting add to the closed loop, should it be approved by the FDA (PDUFA date: April 15, 2013). As a reminder, Sansum/UCSB/JDRF are currently evaluating mealtime use of Afrezza on top of closed-loop insulin delivery – early results have looked very, very encouraging given the product’s 12-15 minute peak. We look forward to the full results, which according to MannKind CEO Mr. Al Mann, should be published in the next few months.


Insulin Delivery to the Skin Using Microneedles

Mark Prausnitz, PhD (Georgia Institute of Technology, Atlanta, GA)

Dr. Mark Prausnitz provided an overview of microneedle (MN) insulin delivery technology; his presentation focused on work done in the academic setting at Georgia Tech and Emory University, and did not cover BD’s program. He began by delineating four categories of MNs: (1) solid MNs that puncture the skin, allowing for subsequent drug penetration; (2) MNs coated with the drug; (3) dissolving MNs that contain the drug; and (4) hollow MNs that provide a channel for the drug. Soluble or coated MN patches provide the easiest administration, but can only administer unit doses, making them slightly less applicable for insulin administration. Dr. Prausnitz next discussed the result of an insulin MN study conducted in 12 pediatric type 1 diabetes patients: MNs were associated with significantly less pain during insertion, but seemed to be more painful for patients during infusion. Dr. Prausnitz argued that despite this pain (which can be reduced by reducing infusion volume), MNs can still help compliance because the lack of a visible needle is a comforting factor for patients – this is an underappreciated benefit of microneedles in our view. In terms of efficacy, MN insulin administration had a quicker onset and smaller tail. The presentation ended with a fairly attention-grabbing set of pictures: a patch of MNs is less than half the width of a nickel, and dye studies show that soluble needles are mostly dissolved in one minute and fully dissolved in five. While MN application in diabetes has largely been focused on insulin administration, we wonder whether other injectable diabetes therapies such as GLP-1 agonists could be administered in the form of a single-dose MN patch. 

  • We last saw data on BD’s microneedle program in a poster at EASD 2013 – see page 21 here. Encouragingly, the poster concluded that “additional work with intradermal sets for insulin infusion is justified and three-day extended duration studies are underway.” We look forward to seeing longer-term data, especially with PK/PD results.

Insuline Products

Andreas Pfützner, MD, PhD (IKFE GmbH, Mainz, Germany)

Dr. Andreas Pfutzner provided an overview of the completed studies on Insuline’s InsuPatch (for pump users) and the InsuPad (for MDI users) – as a reminder, both devices heat the skin around the injection site to stimulate microcirculation and accelerate insulin absorption. Studies have shown that use of the devices improves glycemic control with fewer hypoglycemic events with an ~20% decrease in insulin requirements. Dr. Pfützner first reviewed results from a study of the InsuPatch in adolescents, highlighting that patients using the device reached peak absorption levels significantly faster than those not using the InsuPatch (p=0.002). Dr. Pfützner focused on the most recent data on the InsuPad (presented at EASD), which heats the skin in three, ten-minute heat cycles over 50 minutes after an insulin injection. Notably, patients using the InsuPad experienced 45% less hypoglycemia and needed 28% less prandial insulin. This study is especially unique because InsuLine worked with BARMER-Ersatzkasse (a German payer) to design clinical outcomes from the beginning. This is very key going forward as new medical technologies seek to collect data for reimbursement, especially in Europe where payers tend to be stricter than here in the US.

  • In Germany where “saving money is more important than taking care of patients,” InsuLine partnered with BARMER-Ersatzkasse to demonstrate that InsuPad. The strategic approach worked the system backwards – instead of getting the product approved first and seeking reimbursement later, InsuLine collected the proper clinical data to gain reimbursement from the get-go. The purpose of the Barmer study was to show that patients needed less prandial insulin when using the InsuPad, while maintaining similar glycemic control.
  • The InsuPad has been launched for one week in Germany, and reimbursement discussions are in the final stages. Dr. Pfützner expects discussions to be completed within six months.
  • As a reminder, the InsuPad includes a disposable adhesive pad (replaced every day) and a rechargeable heating unit. The whole on-body portion is quite slim and a bit smaller than the second-generation OmniPod. We appreciated a full demonstration of the device in the EASD 2013 Exhibit Hall, which struck us as quite user friendly and unobtrusive.
    • Patients apply the disposable pad once per day, which has a plastic hinged door that holds the heating unit. When it’s time to inject insulin, patients open the InsuPad’s door and inject their insulin into exposed skin through an oval-shaped opening. Upon closing the door, the heating unit warms the area to 40 degrees Celsius (104 degrees Fahrenheit) for 50 minutes. Though the temperature sounded quite hot to us, the device only felt warm to the touch. The heating unit needs to be recharged every 24 hours (based on four injections per day).
    • For countries outside of Germany, we understand that Insuline will have an online store – per comments we heard at the company’s EASD exhibit hall booth, the starter kit will cost 98 euros and include two heating units and five disposable pads. A 30-day supply of disposable pads will cost 49 euros (a nice recurring revenue stream for the company). Patients will be able to buy the device online only after obtaining a prescription – according to the rep we spoke to at EASD, this includes the US, though we were somewhat skeptical this will be possible. We look forward to hearing real-world experiences with the device, as the market need for faster-acting insulin is so unmet.

Panel Discussion

Dr. Lutz Heinemann (Science & Co, San Diego, CA): Can you talk about patient perceptions of the InsuPad? This is a device that will be used in daily life. It’s critical that patients like the device and use it. My perception is that patients aren’t that keen on knowing about PK and PD.

Dr. Andreas Pfützner (KFE GmbH, Mainz, Germany): When the device was introduced to me, I thought, ‘Who the heck needs that?’ – another hassle, another technology. And when you talk to patients who have not used it, you get a similar result. But there are thrilling effects on patients’ wellbeing when they use the device. It’s not very complicated. It operates by closing it after injection. There is no pushing a button and it happens automatically; it’s a no brainer operation, which is good for my type 2 patients. At least for German patients, when wearing it, they can now be better controlled with less insulin – another psychological positive. The most important impact is that they don’t even realize. When they start to use it, they have better blood glucose and don’t realize it might be different. But if they take a break one day, what they realize immediately, is that their glucose deteriorates to a much larger extent than expected. And they have more hypoglycemia. For patients stopping device, we’ve had them come back to the study site and ask for more. This is now launched in Germany and can be acquired through the regular channels.

Q: It is really exciting to hear new information about the DiaPort because I’ve been waiting to get a little more experience with this technology. Are there any clinical studies underway that look at insulin resistance or peripheral insulin levels?

Dr. Howard Zisser (Sansum Diabetes Research Institute, Santa Barbara, CA): That is the next study underway. There will be a PK/PD (pharmacokinetic/pharmacodynamics) study next year.

Q: How long are the ports implanted for?

Dr. Zisser: They are changed, I believe, every nine or 12 months; however, the change is an office procedure. Sometimes there are occlusions before this time, in which case you can just change out the catheter.

Q: Dr. Robert Vigersky (Walter Reed National Military Medical Center Bethesda, MD): Were the patients with type 2 diabetes on the InsuPad insulin naïve or were they already taking insulin?

Dr. Pfützner: Patients in the study had been on insulin for sometime; in fact, it was required that they be on insulin at least a year a before.

Dr. Vigersky: The patients you used in your study were in excellent control [A1c= 7.2%]. Do you have experience using your InsuPad in patients with poor control, and can you show similarly good results with less insulin?

Dr. Pfützner: In the real world study, where many people have A1cs over 7.2%, you don’t necessarily save insulin. We have another study to look at that. Another advantage of the InsuPad is that there is faster absorption, and you can get better control without changing the insulin dose. Hopefully we will share more data at the next conference.

Dr. David Klonoff (Mills-Peninsula Health Services, San Mateo, CA): Can each of you talk about the potential risks of your treatments? Dr. Pfutzner, what is the risk of the InsuPad’s heat damaging the insulin? Dr. Prausnitz, what’s the risk of skin damage with intradermal insulin? Dr. Zisser, what’s the risk of infection with intraperitoneal delivery? What about using these methods with an experimental ultra-fast insulin analog?

Dr. Pfützner: The InsuPad device warms the skin to 39 degrees Celsius. There is no risk for that. It’s comparable to a hot warm shower.

Dr. Prausnitz: For intradermal insulin delivery, there is not much experience. There are unknowns. BD has a few published studies. The safety profile looked good. At other conferences, we tell people how wonderful intradermal delivery is for vaccination, but then I come here and tell you there is no immune response. It’s an open question that has to be addressed.

Dr. Howard Zisser (Sansum Diabetes Research Institute, Santa Barbara, CA): The risk of using the DiaPort is the ring right around the implantation. There’s a Dacron cuff that is designed to have the skin grow in. Peritonitis has not been problem. For implantable pumps, there were more issues with infections. I don’t know the rates on that. Implantable pumps use U-400 insulin and it’s in the pump for 8-12 weeks.

Q: Can you share any information you have on use of the InsuPad with insulin resistance? There are some patients using huge doses of insulin and U-500 insulin. I was wondering if you see the same effects in all patients?

Dr. Pfützner: As you may have seen, the mean insulin dose was around 112 units; however, some of the doses went up to almost 200 units. The percentage reduction you see is the same for all dose levels. In the beginning, I wasn’t sure if this was the right observation of the device, but now this is one of the key indications in Germany for those who have already applied to be part of a real-world study. Every doctor has a patient being treated with a huge insulin dose that he or she thinks will be a good patient for this product.

Q: From the closed loop algorithm side, is the PK profile different from one patient to another?

Dr. Zisser: For intraperitoneal delivery, it’s old data, but the PK is very predictable. For each individual and throughout day, you must customize the insulin to carb ratio.

Dr. Pfützner: For those studies, we haven’t seen any problem adjusting to those differences. There are many in the room working on algorithms required for that. Algorithms can be individualized. I don’t think it’s major issue.

Dr. Prausnitz: If you have a more rapidly responsive system and a good algorithm, you should be in better shape to control it.

Q: What kind of regulatory approval does this device need? Does it need a CE mark?

Dr. Pfützner: Yes, this device needs a CE mark, and, in fact, it currently has one. Instead of launching the product and then performing a trial to gain reimbursement, we decided to do the process backwards and do a pharmacoeconomic study before the launch. That way, our product had reimbursement from the start. Both the InsuPatch and the InsuPad have a CE mark in Europe.

Dr. Gerold Grodsky (UCSF, San Francisco, CA): It seems like you have two things going on with microneedles. You are in the dermis and increase absorption of insulin. Also, the distributed insulin doses over a wider area. You had both going on. Which is contributing to faster kinetics?

Dr. Prausnitz: It’s not clear the second point you made is true in what we did. In the human study, it was a hollow microneedle – it was a similar area in the skin to that in subcutaneous injection. The patch is about a square centimeter in size. But if you make an injection, the fluid spreads. It’s probably of comparable size. It’s not clear the area being augmented is the key. There is high vascular content, especially in the superficial dermis.

Q: How well do the hollow microneedles stay in? It seems like if the needle is less than a millimeter in length, it could be pulled out pretty easily.

Dr. Mark Prausnitz (Georgia Institute of Technology, Atlanta, Georgia): I assume you are referring to the use of hollow needles as part of a pump system, in which case the needles would be in all day. We currently don’t have a study examining that. Our previous study was a bolus study during which the needles were in only one minute. While there are some adhesives that are particularly strong, I agree that it may be difficult for these needles to stay in. Only time will tell.

Dr. Gerold Grodsky (University of California, San Francisco, CA): In this session, we haven’t been able to talk about every insulin delivery technique, and another method that is being investigated rapidly is changing the milieu in which insulin is injected or stored. There are companies that are using different additives in the insulin, which is a different approach than we have talked about in this session; that needed to be mentioned at this meeting.

Dr. Zisser: How long does the InsuPad warming last?

Dr. Pfützner: The total warming cycle is 50 minutes.

Dr. Zisser: Did you put any sensors within the window to see if decreases lag time?

Dr. Pfützner: Not yet, but that’s a good idea.

Dr. Heinemann: I would like to compliment Roche on the development of the DiaPort. This is a good example of improving technology if we work on it. Dr. Zisser, do you expect better performance? Could another insulin be applied?

Dr. Zisser: I think we challenged it pretty hard with fully automated closed loop and 70 g meals. I would like to see better results. But the difference between subcutaneous and intraperitoneal delivery is exactly what we anticipated.

Nanotechnology for Diabetes Devices

Smart Insulin

Zhen Gu, PhD (North Carolina State University, Raleigh, NC)

Dr. Zhen Gu discussed smart insulins, one of the most exciting (and, to be fair, a bit hyped) classes of diabetes products in development. The goal for such insulins is to sense ambient glucose levels and release an appropriate amount of insulin. Dr. Gu’s laboratory is investigating three categories of candidates: (1) Microgels that encapsulate insulin in gel microspheres that expand to release insulin in the presence of glucose; (2) Nano-networks that use glucose-mediated solubility changes to stimulate insulin release; and (3) Peptide gel-based approaches in which glucose converts hydrophobic PBA to a hydrophilic form, allowing the attached insulin to solubilize. Studies of some of the lab’s candidates in mouse models demonstrate clear glucose-dependent insulin release that (in the case of nano-networks) can lead to statistically significant and substantial reductions in plasma glucose for up to ten days after a single subcutaneous injection. Dr. Gu also noted that his lab is investigating other systems that could be triggered by ultrasound or infrared light to release insulin with spatio-temporal control, as well as smart insulin vesicles that could mimic the function of endogenous insulin vesicles in beta cells. To emphasize the excitement regarding smart insulins, Dr. Gu mentioned that he has received communications from patients asking to be in future clinical trials, and noted that Merck purchased Smart Cells for half a billion dollars in 2010 (Merck has not provided recent updates on the status of that program, which we assume is still preclinical; see our Merck 3Q13 report). 

  • Dr. Gu began by outlining the weakness of current insulin therapy: it requires frequent monitoring of blood glucose levels, can require multiple daily injections, and a major overdose of insulin can be harmful or even fatal. He compared this profile to the endogenous “smart” action of beta cells in nondiabetic individuals, and argued that this sort of smart insulin release (based on glucose levels) would drastically improve insulin therapy by tightening control and (theoretically) nearly eliminating hypoglycemia. He outlined two key mechanisms needed to develop smart insulin: a factor that can sense ambient glucose levels, and a material that can be activated by this factor to release insulin.
  • Most of Dr. Gu’s presentation was devoted to an overview of the three major categories of smart insulins his group is working on: microgels, nano-networks, and peptide gels.
    • Microgels encapsulate insulin and the glucose oxidase enzyme into a gel sphere. Higher glucose levels will increase the activity of the glucose oxidase enzyme, which in turn will increase the pH of the space within the gel and cause it to expand and release insulin. Dr. Gu played a rather striking video comparing gels placed in solution with normoglycemic (100 mg/dl glucose) and hyperglycemic (400 mg/dl glucose) solutions: the spheres remained relatively unchanged in the former solution, but rapidly expanded in the latter. Data from a mouse model of diabetes show that the microgel reduced glucose levels sustainably, substantially, and significantly for approximately 12 hours. Advantages of the microgel approach include its simple formulation, uniform size, and high insulin loading capacity.
    • Nano-networks utilize changes in solubility (triggered by glucose levels) to release insulin. Advantages of this approach, according to Dr. Gu, include simple preparation and injectability. Notably, in a mouse model study, one subcutaneous injection of a nano-network compound led to substantial sustained reductions in blood glucose for over ten days.
    • Peptide gel-based approaches boast straightforward preparation and tailorable structures. The gel contains a hydrophobic PBA-based element that keeps the insulin from entering solution. In the presence of glucose, the PBA becomes charged, allowing for dissociation.  
  • Dr. Gu noted that his group is exploring other smart insulin formats, including synthetic smart vesicles that mimic the function of insulin vesicles in beta cells. He also mentioned systems that can be triggered by ultrasound or infrared light to release insulin with spatio-temporal control. These strategies are intriguing, but appear to be in earlier stages of development than the three categories mentioned previously. During the subsequent panel discussion, Dr. Gu also discussed the possibility of smart glucagon.

Questions and Answers

Q: How do you improve the response speed of a smart insulin?

A: That is a big challenge. Once we engineer more sensitive materials, we can improve response times. We could also modify the insulin itself to enhance response speed.

Q: In your in vivo experiments, one wonders whether it was a slow leak of insulin that was occurring rather than acute changes during meals, as you didn’t show any oral glucose tolerance tests.

A: We performed some tolerance tests. We can show that once we inject glucose into the solution, we see a sharp rise of the glucose level. But it can be hard to distinguish this even from the basal concentration of the insulin. We have been trying to use an ELISA assay to measure immediate release, but have been having difficulty. We want to further enhance the response speed, and one day we hope to be able to prove a peak in insulin release after an oral glucose tolerance test.


Live Demonstrations

Lens Autofluorescence and Diabetes Detection

Craig Misrach (Freedom Meditech, San Diego, CA)

Mr. Craig Misrach introduced (and demonstrated) the DS-120 ocular lens autofluorescence scanner, an FDA approved non-invasive diabetes detection instrument. Mr. Misrach highlighted that the DS-120 test lasts six seconds, requires no blood draw or pupillary dilation, provides immediate results, and comes with a user-friendly touchscreen display. The device measures lens autofluorescence and compares the result to a reference curve based on a database to determine the patient’s chance of having diabetes. Mr. Misrach noted that a 250-subject study, to be published in the Journal of Diabetes Science and Technology in December, showed that the DS-120 had higher sensitivity than either fasting plasma glucose or A1c in correctly distinguishing if a person has diabetes. During the live demonstration, the device correctly placed Mr. Misrach in the green zone (diabetes unlikely) and placed an audience volunteer with longstanding type 2 diabetes into the red zone (diabetes highly likely). The test was quite fast (the measurement itself was six seconds, with an ~30 seconds of set-up), seemed easy to setup (the patient sits in a chair, the technician positions him or her, and the device scans the eye), and the results were intuitive. See a two-minute YouTube video demonstration here.

  • Turning to the principles underlying autofluorescence, Mr. Misrach cited evidence that lens fluorescence is greater in diabetes patients compared to nondiabetic individuals (within a given age group). The fluorescence is theorized to result from accumulation of advanced glycosylated end-products (AGEs), and serves as a long-term index of glycemic control (even longer-term than A1c). The technology dates back to 1995 (when it was developed to study cataracts), but was not made available for diabetes until recently because of costs as well as the past need for injecting fluorescein.

Questions and Answers

Dr. John Pickup (King’s College, London, UK): So you’re seeing this as a screening test rather than a concrete diagnosis tool, right? This is in everyone’s mind because of the controversy over diagnosing by A1c.

A: It’s information that’s valuable for clinicians, who can use this information at their discretion There is no one single test that can diagnose diabetes by itself, but we hope that with a multicenter study with multiple KOLs participating in the protocol, we can garner endorsement from one to two highly influential diabetes organizations indicating that the DS-120’s performance is at least as effective as conventional blood draw.

Q: Tell me about factors like vitamin C and aspirin. Do they affect lens auto-fluorescence?

A: We did not see any effects from any medications or vitamins. There are a couple obvious limitations, such as patients with intraocular lenses.

Q: Glycosylation is important for long-term diabetes. How does lens autofluorescence correlate with complications?

A:  That’s the whole point: to use this tool to reduce complications. I will say that in terms of what we know now, the data is all cloud-based and we don’t get patient-specific information. We’re trying to work with clinician partners to get more information on complications.

Q: If you wear a contact lens, do you have to remove it?

A: We actually see more precise data when patients have contact lenses in.

Q: Is the DS-120 being developed instead of the I-SugarX?

A: We are still developing the I-SugarX, which works by measuring real-time glucose levels in the aqueous humor of the eye.


--by Adam Brown, Hannah Martin, Manu Venkat, and Kelly Close