ATTD (Advanced Technologies and Treatments in Diabetes) 2020

February 19-22, 2020; Madrid, Spain; Day #3 Highlights - Draft

Executive Highlights

  • Our team powered through another big day at ATTD, navigating Feria de Madrid’s multiple oral presentations, parallel sessions, and corporate symposia to bring you the latest in diabetes. See below for the top 13 highlights and make sure to check out days #1 and #2.

  • People packed into the La Paz room at Feria de Madrid to hear the six-month results of the Helmsley-funded Tidepool/Jaeb fully virtual, observational study of DIY Loop. Mean Time in Range increased in new Loop users from 67% at baseline to 73% during months 1-3 and stayed at 73% during months 4-6 (p<0.001). Mean A1c was 6.8% at baseline, dropping to 6.4% at month 6 (p<0.001). Loop was shown to be safe, with one DKA event occurring (judged not related to Loop but presumably related to pump use) and no indication of increased severe hypoglycemia compared to baseline. See below for our full takeaways on the study and get the slide deck at Tidepool’s website.

  • UVA’s Dr. Sue Brown presented results from the 13-week extension phase of the iDCL3 protocol (+1.4 Hour/Day in Range, -0.3% A1c, No Hypoglycemia Difference With Control-IQ vs. Basal-IQ.) Of the 112 participants randomized to the Control-IQ group in iDCL3 (all of whom completed the study), 109 elected to continue into the extension phase. Fifty-five of the participants were randomized to the Basal-IQ (predictive low glucose suspend) group, while fifty-four remained on Control-IQ. It’s worth noting that on most glycemic outcomes, the Control-IQ extension group saw slight worsening during the extension, though it’s unclear if these differences were statistically significant.

  • In other notable tech news from the day:

    • Medtronic announced CE-Marking for its new extended wear (7-day) infusion set. The new infusion set includes several new components, including a new “H-Cap Connector” that improves infusion site performance, tubing that improves insulin preservative retention, extended wear adhesive patch, and a new inserter. The most recent study demonstrated that infusion sets were still performing >80% of the time at the end of the seven-day wear, which is equal or better to that observed for the current three-day infusion set at the end of 3-day wear. Another small study (n=21) showed a survival rate of 81% at 7 days. 

    • Dr. Thomas Danne presented data from the SWEET type 1 registry (Europe, Australia, Canada, India) showing that 62% of children use some diabetes technology (i.e., CGM and/or pump). Unsurprisingly, use of diabetes devices was associated with lower A1c (we suspect CGM prompted better outcomes than pumps but that is pure speculation).

    • Ms. Sarina Dass (Northeastern Data) showed intriguing data from an artificial intelligence study that identified pediatric type 1 patients most likely to have a diabetic ketoacidosis (DKA)-related hospitalization within 180 days. The recurrent neural network made a ranked list of 1,523 patients most likely to experience a DKA-related hospitalization; of the ten individuals the model identified as most likely at risk, all ten did indeed experience a DKA-related hospitalization. For the top 90 most at-risk, a third were hospitalized. Overall, about 6% (90 individuals) of study participants were hospitalized.

  • SGLTs in type 1 were again a focus on the therapy side, with a star-studded Friday session featuring Drs. Satish Garg, Chantal Mathieu, and Thomas Danne. Dr. Garg reviewed unmet needs in type 1, Dr. Mathieu reviewed the gauntlet of phase 3 data for the class, and Dr. Danne closed the session with a deep dive on DKA risk mitigation.

  • Friday also featured the audience favorite ATTD Yearbook session, with some of the biggest names in diabetes tech delivering rapid-fire updates on topics and trends in diabetes. See below for Klemen Dovc (Ljubljana University Medical Centre) on CGM; Rayhan Lal (Stanford) on insulin pumps; Jan Bolinder (Karolinska University) on new insulins, biosimilars, and insulin therapy; Eyal Dassau (Harvard) on decision support systems and closed loop; Neal Kaufman (Canary Health) on digital health for prevention and treatment of diabetes; Helen Murphy (King’s College London) on technology and pregnancy; David Maahs (Stanford) on diabetes technology in pediatrics; Michael Riddell (York University) on physical activity and diabetes; Laurel Messer (Barbara Davis Center) on implementation of diabetes technology; Alon Liberman (Schneider Children’s Medical Center) on diabetes technology and the “human factor”; Desmond Schatz (University of Florida) on immune intervention for type 1 diabetes; Irl Hirsch (University of Washington) on new medications for the treatment of diabetes; and finally, Kavita Garg (University of Colorado) on NAFLD/NASH.

It was another jam-packed day in Madrid, as ATTD 2020 rolled on to day #3. A late afternoon series of oral presentations headlined the day, featuring data readouts from the Tidepool/Jaeb Loop observational study and iDCL3 (Control-IQ pivotal) extension. We’ll be back soon with more from day #4.

Table of Contents 

Diabetes Technology Highlights

1. Tidepool/Jaeb’s Virtual Observational Study of Loop: Time in Range +1.4 Hours/Day (67% to 73%), A1c Drops From 6.8% to 6.4% In New Loop Users After 6 Months; No Indication of Increased Severe Hypoglycemia Compared to Baseline

The tunnel-like La Paz room at Feria de Madrid slowly filled up on Friday afternoon, as an increasing number of people packed in to hear the six-month results of the Helmsley-funded Tidepool/Jaeb fully virtual, observational study of DIY Loop. The results were limited to “new” users of Loop (n=607), comparing glycemic outcomes at baseline to months 1-6 of Loop. Overall, the highly anticipated study results show statistically significant improvements on Loop, made even more impressive since the cohort had fairly tight glycemic control at baseline. The presentation is available at Tidepool’s website.

  • Mean Time in Range increased significantly from 67% at baseline to 73% during months 1-3 and stayed at 73% during months 4-6 (p<0.001). Mean A1c was 6.8% at baseline, decreasing to 6.4% at month 6 (p<0.001). Mean time <70 mg/dl was mostly flat from 2.9% at baseline to 2.6% during months 4-6; time <54 mg/dl fell from 0.4% to 0.3%. Time in closed loop was 83% during months 1-3, dropping slightly to 79% during months 4-6. Loop was shown to be safe, with one DKA event occurring (judged not related to Loop) and no indication of increased severe hypoglycemia compared to baseline.

  • The study will continue until March 31, 2020, and results will be used to help get Tidepool Loop through the FDA. We heard that submission could come as soon as this year – we would certainly expect submission within that timeline. The study is also collecting patient-reported outcomes, such as diabetes distress, sleep quality, and hypoglycemia fear, which will be presented at a later date – these are areas that also need standardization and we look forward to hearing more on this front.

  • The study cohort is very young (23), well-educated (85% BA equivalent or higher), wealthy, and racially homogenous (91% white); unsurprisingly, the group also had fairly tight glycemic control (6.8% average A1c) at baseline. Notably, Loop showed Time in Range increases across all age brackets (see below). Participants were 56% female and 71% reported an annual household income ≥$100,000.

  • Jaeb’s John Lum estimated a worldwide Loop user base of ~9,000, based on RileyLink order history – this has risen quite a lot since our last estimate! As of February 2020, there are 2,585 Medtronic Loop users and 6,449 using Loop with Omnipod. As a reminder, Omnipod Loop compatibility came less than one year ago. At DiabetesMine’s Innovation Summit 2019, Tidepool CEO Howard Look showed a graph of new users enrolling in the Tidepool/Jaeb study, demonstrating a huge increase after the DIY Loop Omnipod code was posted in April 2019.

  • The study was conducted as a “virtual” observational study, meaning CGM and pump data flow straight from the Loop app to Apple Health to Tidepool/Jaeb. Fingerstick collection kits were mailed to study participants to collect blood for A1c data. Informed consent and demographic data were collected online and web-based questionnaires were sent at baseline, three months, six months, and twelve months. Participants were also sent a weekly ~one-minute survey to report severe hypos, DKA, or hospitalizations in the prior week. As devices and connectivity become increasingly common and easy to use, we’d expect virtual (or partially virtual) studies to become more common. Still, tremendous work has gone into this study.




Months 1-3

Months 4-6

6-month vs. baseline difference

Time in Range




+1.4 hr/day


Time <70 mg/dl




-4 min/day


Time <54 mg/dl




-1 min/day








  • Notably, with Loop, Time in Range during daytime (6 AM – 10 PM) and nighttime was identical (10 PM – 6 AM) at 73%. At baseline, Time in Range was 68% during the day, but just 65% at night. Indeed, after three months, median Time in Range was at least slightly higher at every hour of the day.

  • Time in Range was increased in young and middle aged users, ranging from <7 to >50 years. The greatest benefit was seen in the 25-49 years age group (n=203), with Time in Range improving from 70% to 77%. Time in range for those below 13 was either consistent at months 4-6 compared to earlier in the trial; those over 13 all saw a slight drop off from month 3 to months 4-6. The smallest benefit was seen with the >50 years age group (n=50), as Time in Range went from 74% at baseline to 76% after six months (while we are not sure if this is technically significant, an improvement of 30 minutes time in range would certainly be of interest to users). The mean age of the entire cohort was 23 years, with just 8% of users ≥50 years old and presumably very few older than 55 or 60 or 70 or older since it was not specified. DIY systems are virtually always seen as requiring more technical know-how to operate and troubleshoot; we’d be curious to see how time in closed loop varied by duration of diabetes.

  • Time <70 mg/dl dropped insignificantly from 2.9% at baseline to 2.8% during months 1-3 and 2.6% during months 4-6 (p=0.12). Time <54 mg/dl fell statistically significantly, though perhaps not meaningfully, from 0.4% at baseline to ~0.3% during months 4-6 (p=0.004). Mr. Lum noted that the rate of hypoglycemic episodes (defined as distinct excursions below 54 mg/dl) was ~0.7/week before and after Loop.

  • A1c fell from 6.8% at baseline to 6.5% after three months and 6.4% after six months (p<0.001). Given the very low A1c at baseline, a -0.4% reduction in A1c is quite impressive.

  • Time in closed loop was 83% during months 1-3, falling slightly to 79% during months 4-6. Mr. Lum did note that these were lower bounds of confidence intervals as data in the study is still being collected. In total, 32 participants (5%) in the group stopped using Loop – they will be surveyed to learn their reasons for discontinuation. While some say that given the need to carry an extra device (RileyLink), it’s not too surprising to see a reduction in time in closed loop compared to Tandem’s Control-IQ system (92% time in closed loop during pivotal, we were a bit surprised it wasn’t higher). Unsurprisingly, time using the Dexcom G6 CGM was quite high, at 96% during months 1-3 and 93% during months 4-6.

  • Loop was shown to be safe, with one recorded DKA event (not related to Loop) and no indication of increased severe hypoglycemia compared to baseline. Severe hypoglycemia was defined on the weekly questionnaire as “need[ing] the assistance of someone else to treat the low.” Two severe hypo events were judged to be related to Loop. One event was related to a large suggested insulin bolus due to an overestimated glucose prediction (this is otherwise known as a user error), while another event could not rule out a Loop malfunction as the cause.

  • One supplemental slide (on Tidepool’s website, not shared at the presentation), shows a few selected survey results. Notably, three-quarters of respondents (sample size unknown) said they were highly likely to recommend Loop. About one-third (30%) needed help starting on Loop, and of those, 71% had another Loop user’s help. Unsurprisingly, 98% of users said they read online material before starting Loop.

2. iDCL3 Extension Phase Shows +1.4 Hour/Day in Range, -0.3% A1c, No Hypoglycemia Difference With Control-IQ vs. Basal-IQ

Continuing from a strong 2019 (Control-IQ readout at ADA 2019, publication in NEJM), UVA’s Dr. Sue Brown presented results from the 13-week extension phase of the iDCL3 protocol. Of the 112 participants randomized to the Control-IQ group in iDCL3 (all of whom completed the study), 109 elected to continue into the extension phase. Fifty-five of the participants were randomized to the Basal-IQ (predictive low glucose suspend) group, while fifty-four remained on Control-IQ. Participants only had two clinic visits: one at baseline and one at the end of the trial. The Basal-IQ group received two phone calls to help with on-boarding. At baseline (i.e., the end of the main study), the groups were well matched with mean A1c of 7.1% and 7% and Time in Range of 70% and 71% in the Basal-IQ and Control-IQ groups, respectively. It’s worth noting that the Control-IQ extension group also saw slight worsening in glycemic control, though it’s unclear if these differences were statistically significant.




Adjusted Treatment Group Difference



13 Weeks


13 Weeks

Time in Range







Time >180 mg/dl







Time <70 mg/dl







Time <54 mg/dl














  • Time in Range decreased from 70% at baseline to 60% at the end of 13-weeks on Basal-IQ. In the Control-IQ group, Time in Range fell slightly from 71% to 68%. The risk-adjusted treatment group difference was +6% (p<0.001). This difference was most notable at night (midnight – 6 AM), where the treatment group difference was +10% (p<0.001). During the day (6 AM – midnight), Control-IQ has a +5% advantage over Basal-IQ on Time in Range. This was not a surprise, given that the Control-IQ algorithm performs more aggressively when Sleep Mode is activated.

  • The entire between-group Time in Range difference was driven by differences in time >180 mg/dl. Both groups spent 28% of time >180 mg/dl at baseline, increasing to 38% in the Basal-IQ group compared to 31% in the Control-IQ group (p<0.001). 

  • Time <70 mg/dl was not significantly different between the two groups. At baseline, the Basal-IQ group spent 1.3% time <70 mg/dl, rising to 1.5% after 13 weeks. The Control-IQ group spent 1.1% of time <70 mg/dl at baseline, rising to 1.4% after 13 weeks. Both algorithms seem incredibly effective at reducing time in hypoglycemia – 1.5% time <70 mg/dl is just 22 minutes/day.

  • Adjusted A1c difference was 0.3% in favor of the Control-IQ group. Both groups had a mean A1c of 7.1% at baseline, rising to 7.5% in the Basal-IQ group, compared to 7.2% in the Control-IQ group. Similarly, adjusted mean sensor glucose was 7 mg/dl in favor of Control-IQ.

  • There were no severe hypoglycemia events or DKA in either group during the extension phase. Three hyperglycemia with ketosis episodes occurred in the Basal-IQ group, while none were reported in the Control-IQ group. Neither group recorded an severe adverse events related to the devices.

3. Medtronic Extended Wear Infusion Set (7-Day) Receives CE-Mark; Small Study (n=21) Shows 81% Survival Rate at 7 Days; Launch in “Select Countries” Planned for “Early 2021”

The CE-Marking comes at the tail end of last year’s goal to launch the infusion set within “one-year.” The new infusion set includes several new components, including a new “H-Cap Connector” that improves infusion site performance, tubing that improves insulin preservative retention, extended wear adhesive patch, and a new inserter. The new inserter appears to be the all-in-one, hidden-needle Mio Advance automatic inserter device. Looking ahead, a Medtronic spokesperson later confirmed that with the CE Mark, the company is “now able to take the next steps to develop plans with [its] alliance partners, suppliers, and commercial teams.” Excitingly, the company plans to introduce the new set in “select countries” in “early 2021,” while taking time to build inventory and enhance the company's pump software in the meantime. The inserter is already available in Europe and has been FDA-cleared for over a year, but has not launched in the US. On the US front, Medtronic announced it received investigational device (IDE) approval to begin a pivotal study of its extended wear infusion set back in August. The multi-center, non-randomized, prospective, single-arm study will enroll up to 150 participants with type 1 diabetes (18-80 years), all wearing the MiniMed 670G and comparing current 2-3 day sets to the new extended wear set. We didn’t hear anything on potential US timelines today. Medtronic featured the infusion set at a special “Innovation Suite” at the ATTD Exhibit Hall; unfortunately, it was not open to the press.

  • Dr. Cohen presented data from a clinical study of 21 subjects using “various Medtronic pumps” and either Humalog or Novolog. In total, the subjects recorded 82 insertions and 78 wears. Set failure was defined as blood ketone level >0.6 mmol/l, evidence of infection at the infusion site, or blood glucose not decreasing by at least 50 mg/dl one hour after a correction bolus for a glucose level >300 mg/dl. Kaplan-Meier curves showed 81% survival rate at 7 days for the new infusion sets; the survival rate increased to 85% if insertion failures were excluded. Reasons for failure were site reaction/blood (7% of all wears), insertion failure (5%), unexplained hyperglycemia (3%), and adhesive failure (3%). Notably, the total daily dose did not increase over the seven days, indicating insulin delivery efficiency was not significantly changed over the entire period.


4. SWEET Type 1 Registry (Europe, Australia, Canada, India) Shows 62% of Children Use Technology (CGM and/or Pump), Tech Use Associated with Lower A1c

Dr. Thomas Danne (Hannover Medical School) presented some quick results from the SWEET type 1 registry, estimating that 62% of children use at least one diabetes device (i.e., CGM or pump). The results were calculated from 25,654 children (18 years or under) with type 1 diabetes in the SWEET registry, which spans 21 centers across Europe, Canada, Australia, and India. The data analyzed was collected between August 2017 and July 2019. In the cohort, mean age was 13.8 years and 39% were on MDI and SMBG. Nearly one-third (30%) used both a pump and CGM, while 17% used a pump and SMBG and the remaining 14% were on MDI and CGM. Notably, “more than 85%” of the subjects using pumps were in either Europe or North America. These distributions did not vary by a great amount across age groups. Adjusted for demographics and region, mean A1c of CGM + pump users was ~0.9% lower than that of SMBG + MDI users. Dr. Danne highlighted pump use as being associated with lower rates of severe hypoglycemia, while rates of severe hypo were highest in MDI + CGM users.




Pump + SMBG

Pump + CGM











Severe hypos





5. Successful 10-week Onboarding from MDI to MiniMed 670G System in Qatar: Time in Range Improves from 47% to 76%, A1c Decline From 8.2% to 6.7%, Median 89% of Time in Auto Mode After Three Months

Dr. Goran Petrovski from Sidra Medicine in Doha, Qatar kicked off a series of oral presentations with impressive results from a 10-day initiation protocol study (n=30) designed to onboard pediatric patients (ages 7-18) straight from MDI to the MiniMed 670G hybrid closed-loop (HCL) system. Using a four-step approach, mean Time in Range increased from 47% at baseline to 72% on day 10 and 76% on day 57 (p<0.001). Mean A1c levels declined significantly from 8.2% baseline to 6.7% after 12 weeks of MiniMed 670G (p = 0.017). The number of Auto Mode exits declined significantly from 8.4 in the first two weeks to 4.2 by the third month. Lack of calibrations and high sensor glucose levels made up most of the Auto Mode exits. Notably, participants spent a median of 89% of their time in Auto Mode and 92% using the Guardian Sensor 3 and there were no reported cases of severe hypoglycemia or ketoacidosis. For context, in MiniMed 670G pivotal studies, Time in Range values increased from 56% to 65% in ages 7-13. During Q&A, Barbara Davis’ Greg Forlenza congratulated Dr. Petrovski on the impressive results and confirmed that the Qatari patients were initiated using the newer 670G transmitter that reduced Auto Mode exits.

  • The Sidra protocol encompassed a four-step structured approach that was quite intensive. Step 1 introduced the participants to the MiniMed 670G system through group sessions and discussed responsibilities and expectations. Step 2 trained users on hybrid closed loop with participants attending group education sessions with instructors for five consecutive days. In Step 3, participants began to use the system in manual mode, collecting insulin and CGM data for a period of three days. In the final Step 4, Auto Mode was turned on and the group was measured for 12 weeks.

6. Artificial Intelligence Accurately Identifies Pediatric Type 1 Patients at Higher Risk for DKA Hospitalizations

Ms. Sarina Dass (Northeastern University) shared impressive data from a study using artificial intelligence to identify pediatric type 1 patients more likely to be admitted to the hospital due to diabetic ketoacidosis (DKA) within the next 180 days. Ms. Dass’ approach used a recurrent neural network looking at over 500 features over two years (e.g., demographics, structured EHR data (CPT codes, lab results, vitals, medications, etc.), A1c trajectories, free-text clinical documents, clinical intake forms, etc.). With data from 1,523 patients ages 8-18 across multiple diabetes centers in the Midwest, the neural network generated a list of all the individuals ranked by likelihood of DKA-related hospitalization in the next 180 days. The occurrence of DKA-related hospitalizations in the dataset within 180 days was ~6% (~90 hospitalizations). The dataset for training and validating the model were the same, an important study limitation noted by UVA’s Dr. Marc Breton during Q&A. In the ten individuals the model identified as most likely to be at risk for hospitalization, all ten did indeed experience a DKA-related hospitalization. For the top 25 most at-risk individuals, about half (12) experienced a hospitalization and for the top 90 most at-risk, a third of them experienced a hospitalization. In other words, the 90 most at-risk individuals identified by the model were indeed at a five-times higher risk for DKA-related hospitalization than the larger dataset. Presumably, the model could help clinics identify patients at higher risk for hospitalization and introduce specific interventions to help reduce likelihood of DKA.

7. New Differential Equation Model for A1c and Time in Range Produces Correlation Coefficient of 0.93, Discussions with FDA Underway

To a fully packed room, Dr. Chiara Fabris (University of Virginia) proposed a new model for relating A1c with Time in Range. The correlation coefficient for the model’s estimated A1c and lab-measured A1c was an impressive 0.93; For context, the traditionally used linear models correlating TIR with A1c from Drs. Roy Beck (2019) and Robert Vigersky (2019) have correlation coefficients around 0.6-0.7. The new model, explained at DTM 2019 by UVA’s Dr. Boris Kovatchev, uses one personalized, glycation rate parameter that is unique for each individual. In other words, the model requires one set of Time in Range and lab A1c values to be “calibrated” to that specific individual. The model also includes three fixed population parameters (i.e., biological constants that are the same for all humans). Using the dataset from iDCL 1 (mobile-enabled inControl, Dexcom G4/G5, Roche Spirit Combo; n=120), the three fixed population constants were calculated. Then, individuals’ A1c values from month three of iDCL 3 (Control-IQ embedded in t:slim X2, Dexcom G6; n=168) were used to calibrate the individual glycation rate variable for each study participant. The equation was used to estimate participants’ A1c daily and then compared to reference lab A1c at months 6 and 9 of iDCL 3.

  • In addition to the impressive 0.93 correlation coefficient between model-estimated A1c and lab-measured A1c, variability in eA1c and A1c was also had a positive correlation coefficient of 0.8. The impressive correlation is perhaps best illustrated by a slide shown at DTM 2019 (see below). The blue dots represent estimated A1c and lab A1c scatter using a linear correlation, while the red dots represent the differential model. Note that the red dots are bunched much more closely around the perfectly correlated eA1c = lab A1c line.

  • Towards the end of presentation, Dr. Fabris noted that the team has discussed this model with the FDA and further discussions are underway. We are interested in hearing more details about the discussion and what this new metric could mean in the near future from a regulatory standpoint. The FDA has generally maintained A1c as the primary efficacy endpoint for regulatory consideration, but it shouldn’t be surprising that Time in Range and A1c are intricately linked, given that they both reflect blood glucose.

8. Dr. Guido Freckmann Calls for Standardization of CGM Accuracy Measurements, Criticizes MARD, and Questions Time in Range Consensus Goals

Dr. Guido Freckmann (Institute for Diabetes-Technology at Ulm University) gave a provocative talk on the need for standardization of CGM accuracy measurements similar to what was done years ago with A1c. He discussed the importance of a traceable and unbroken chain of calibrations to adequately compare the CGM accuracy to a reference method. Indeed, we heard similar concerns about CGM accuracy, particularly CGM point values, from FDA’s Dr. Courtney Lias at DTM 2019. While regulatory submissions for CGM generally include studies comparing CGM values with YSI reference, different CGMs have different performance levels at various glucose concentrations. One comparison that comes to mind is FreeStyle Libre vs. Dexcom – FreeStyle Libre does seem to report more time in hypoglycemia than G5; see poster at EASD 2019. In particular, Dr. Freckmann recommends following the  ISO Standard for metrological traceability (ISO 17511:2003) for CGMs, similarly to requirements in the ISO standard for blood glucose monitoring systems for self-testing (ISO 15197:2013). In 2016, we saw the FDA tighten accuracy standards for BGM, requiring 95% of results to be within 15 mg/dl/15% of reference values. Of course, with the creation of special controls for iCGM in the US, these tighter accuracy requirements could become table stakes if iCGM becomes the de facto standard for CGM in the US.

  • Dr. Freckmann suggests that although a MARD of less than 10% has become a de facto standard for non-adjunctive labeling, the metric is subject to variability depending on the study design and the population tested (e.g., type 1 vs. type 2 diabetes). MARD numbers can also be misleading based on the number of paired points in various glycemic ranges. Most experts in the field have long recognized the limitations of MARD as a catch-all metric for CGM accuracy, and indeed the FDA special controls for iCGM don’t even specify a MARD requirement, instead focusing on ±15 mg/dl/15% and ±20 mg/dl/20% measures.

  • Notably, Dr. Freckmann also questioned why the 2019 consensus CGM metric goals did not include accuracy of CGM devices as a “prerequisite.” Indeed, without a guarantee for the accuracy of a CGM in hypoglycemia, is the target for <4% time <70 mg/dl measured by a CGM actually meaningful? Additionally, without a standard for CGM accuracy, there is also no reason to believe that the cutoffs set by the consensus are clinically meaningful. Dr. Freckmann’s fellow countryman Dr. Lutz Heinemann expressed a similarly skeptical viewpoint at DTM 2019 and in a paper published in JDST.

9. A Rural Chilean Study of Pediatric Type 1s Highlights the Potential of Telemedicine: Mean A1c Drops from 8.5% to 7.9% in Two Years, 93% of Telemedicine Contacts Performed via Smartphone

Ms. Julie Pelicand (Universidad del Valparaiso) presented results from her study following 21 families of pediatric patients with type 1 diabetes over two years. These patients lived in rural, remote areas of the Aconcagua Valley and used telemedicine for their diabetes care and follow-ups beginning in 2016. This study aimed to explore opportunities for advancement of telemedicine in rural communities and to assess satisfaction, utility, and challenges of this form of diabetes care. In order to do so, the number and content of telemedicine contacts between families and physicians was tracked, and satisfaction was assessed using surveys and semi-structured qualitative interviews. Perhaps the most impressive result from the study, mean A1c in the particularly challenging pediatric and teenage group (mean age 11 years) fell from 8.5% in 2016 to 7.9% in 2018. It’s incredibly powerful to see such positive data in populations and regions that are often forgotten and we give credit to ATTD for highlighting Ms. Pelicand’s excellent work.

  • The vast majority of telemedicine contacts were made on a smartphone. Of the 716 total telemedicine contacts made over the duration of the study, 93% were made via smartphone, with an average of 12 smartphone contacts per family per year. The remaining 7% were made via email, with an average of 5 (±4.35) email contacts per family per year. In Chile, while 71% of people have computer internet connections, 94% own smartphones. For these reasons, Ms. Pelicand emphasized smartphone-compatibility as a critical component of telemedicine platforms and an important tool to increase access. It’s certainly a testament to the transformative power that smartphones have had and their immense potential to reach traditionally hard-to-reach groups via telemedicine.

  • Use of telemedicine for administrative purposes, like scheduling hospital visits or accessing supplies was the reason behind 35% of telemedicine contacts. Sharing glycemic data in order to adjust insulin dosing made up nearly a quarter (23%) of contacts, while support managing emergencies like hypoglycemia and diabetic ketoacidosis made up another ~21%. Psychosocial support for patients and their families regarding social, economic, and psychological situations was the reason behind 11% of contacts, health education on topics like nutrition and physical activity made up 5% of contacts, and finally, technical support and treatment adaption of insulin therapy composed 5% of telemedicine consultations.

  • In addition to improving diabetes management, familial health literacy, and health access, telemedicine use was found to strengthen relationships between patients and their healthcare providers. Insights from patients and their families suggest that telemedicine use may increase confidence in healthcare providers and improve perceptions of their empathy and availability.

  • Ms. Pelicand shared the key improvements needed in this emerging field, first highlighting the importance of expanding telehealth communication to include all members of a healthcare team. The researcher also recommended the inclusion of institutions like clinics, hospitals, and even schools, on telemedicine online platforms to facilitate better administrative management and a unified set of health guidelines. She also proposed integrating telemedicine with social platforms to allow peer-to-peer support and sharing of experiences. Finally, she emphasized the importance of improving the sharing of data between patients and healthcare professionals via patient portal.

10. Dexcom-Sponsored Symposium: “Alert Settings Should be Individualized for the Patient”; Time in Range Improvement by ~5% with Increased CGM User Engagement

Dexcom’s Friday-morning symposium featured retrospective data, highlighting Time in Range, A1c, and glycemic variability improvements in patients with higher daily Dexcom G6 feature use (e.g., high alarms, low alarms, Clarity, Share, and Urgent Low Suspend). Dexcom examined data from over 25,000 new patients who started using the G6 between January-July 2019 and measured each patient’s “mean engagement” over three months from August to November. About 9,000 users were categorized as “low” engagement, meaning they used less than 3 features; ~11,000 were “medium” engagement (used 3-4 features), and ~7,000 users were categorized as “high” engagement (>4 features). Patients found with “low engagement” (had a Time in Range of 56%) were not so different from those with a higher Time in Range – e.g., 61% in those with “high engagement. Patients with the highest levels of engagement also observed 38% reduction in time <54 mg/dl and had reduced glycemic variability: 64% of high utilization users achieved a %CV <36% vs. 57% in the low utilization group. Dexcom has shown similar data in the past and while they are interesting correlations, but they are not very surprising and certainly not suggestive of any causative effect: users who utilize more features are likely more engaged in their care and have better glycemic control. We would be curious to learn what sorts of factors drive some to use more features than others.

  • In the same cohort of 26,184 patients, mean Time in Range was 43% for those only using CGM (i.e., no features). For users who utilized Clarity, mean Time in Range increased to 62% and increased again slightly to 64% for those who used Clarity and received at least one push notification. Of course, it’s no surprise to us that those who use Clarity to look at and review their retrospective data have higher mean Time in Range. In fact, much of the power of CGM lies in the ability to look back and identify behaviors that  may be helpful (“Bright Spots”) and harmful (“Landmines”). We wonder how Time in Range might differ in a randomized comparison (i.e., random assignment to use Clarity vs. not use Clarity). At AADE, we’d heard that the open rate for Dexcom Clarity app’s weekly summary report notification is 80% - a very impressive rate.

  • Dexcom also shared that its Share/Follow remote monitoring feature was commonly used in pediatric and adolescent populations, peaking at~95% at around age 16 and declined as age increased. In three different cohorts of pediatric patients (<6 years, n=1,054; 6-12 years, n=7,014; 13-18 years, n=7,308), Time in Range steadily improved from ~45% to ~55% for all groups as the number of remote followers increased from zero to four or more.

  • Improved control were also seen across those who chose to customize their alarm settings. For example, in a sample of 19,540 patients, those who set their hyperglycemia threshold below the default (i.e., more likely to trigger alarms) spent ~2.5 hours/day in hyperglycemia (>240 mg/dl) compared to ~4.5 hours/day in those who had their alarm thresholds set above default (i.e., less likely to trigger alarms). The same trend was observed for hypoglycemia, with individuals customizing their alarms to be more sensitive spending ~15 minutes/day <70 mg/dl compared to ~37-40 minutes/day in those who did not. These decisions would, of course, come at the increased risk of alarm fatigue and the ideal thresholds balancing alarm frequency with glycemic control vary greatly by person. On the same topic, a recent article in the Journal of the Endocrine Society estimated ideal thresholds for hypoglycemia and hyperglycemia alarms as 75 mg/dl and 170 mg/dl.


11. Brian Levine Highlights Digital Health as a Potential Solution to the “Last Mile” Problem of Healthcare Access

Onduo’s and Close Concerns alum Brian Levine defined and taxonomized the current connected diabetes care space, before giving us a glimpse into its future. Mr. Levine began by highlighting the need for greater healthcare access in the US and the critical link between access and outcomes. Connected care, he suggests, may fill the gap between resources and those who need them. A connected care platform was defined as having remote coaching, often delivered through a smartphone app, that may receive data from connected care devices. An excellent paper reviewing the diabetes digital health landscape authored by Mr. Levine, Kelly Close, and Dr. Bob Gabbay was published in DT&T  in October.

  • Connected care platforms were broken into five main categories: (i) quantified self, (ii) portals for brick-and-mortar HCPs, (iii) AI coach, (iv) non-physician HCP-driven platforms, and (v) virtual clinics.

    • Quantified-self platforms provide connected devices like activity trackers or connected blood glucose meters to members and may use virtual incentives or gamification. These are often used by employers, payers, and employee wellness benefit managers in order to encourage improvements to health behaviors and outcomes. 

    • Portals for brick-and-mortar healthcare providers are software programs that can be licensed by healthcare clinics or providers to more effectively manage patients. Features often include risk-stratification of patient panels to better allocate care and two-way secure communication between members of the care team as well as between the care team and patient.

    • Artificial intelligence (AI) coaching platforms are completely based on user-generated data uploaded either passively using connected devices or through manual data entry. These platforms may include automated coaches that advise using pattern recognition.

    • Non-physician healthcare provider driven (NPHD) platforms provide remote human coaching in addition to AI coaching.

    • Virtual clinic platforms, have features of NPHD platforms but also include physicians capable of remotely prescribing and managing medications and connected devices. While these platforms have the highest level of clinical rigor, they are also the most expensive. For this reason, three virtual clinic platforms are often partnered with telemedicine companies.

  • Mr. Levine proposes that these categories exist on a spectrum of clinical rigor, and that the ultimate goal of connected care is to use platforms together in a risk stratification pyramid. This would optimize care by allowing patients to escalate or deescalate the spectrum depending on their individual needs. Levine predicts that connected care will increase access to healthcare and act as a democratizing agent of CGM use. In doing so, connected diabetes care platforms have the potential to decrease comorbidities as well as burden on both patients and payers. Mr. Levine emphasized the significant need for more rigorous investigation of connected diabetes care platforms, with the hope of increasing reimbursement and thus access. 

12. Dr. Irl Hirsch on the Future of SMBG in the Next Five Years: CGMs Exploding in the US and Beyond, but SMBG Access is Still Abysmal in Many Regions

Dr. Irl Hirsch (University of Washington) began his talk by polling the audience for a show of hands of how many think SBGM will go away within the next five years. Less than half of the room raised their hand – we certainly would not have. His presentation began with a stroll down memory lane, starting with a fun history of clinical glucose monitoring from 1841 (urine testing) to the present. The first US patent for a glucose monitor was filed in 1968, followed by the Ames Eyetone meter hitting the market in 1972. Over time, blood glucose monitors became smaller, faster and chock full of features. Dr. Hirsch went on to question why he was asked to give a talk on SMBG, when the era of CGM has “exploded” with 80-95% of his patients with type 1 diabetes and up to 30% of his patients with type 2 diabetes on CGM in his clinic in Seattle, WA. The difference in the U.S. CGM penetration from 2015 to 2019 shows a whopping 432% increase for type 2 diabetes and a 52% increase in type 1 diabetes – see dQ&A’s data for more. Still, although US coverage for CGM is expanding, both commercial and Medicaid coverage remains highly variable across the nation. This means that SMBG may be the only option for many US patients, especially those that don’t use insulin. Looking outside the US, in many low- and middle-income countries, SMBG still has a long way to go. The table below illustrates that the availability of glucose test strips is worse than that of insulin in most countries. At ISPAD 2019, Australia’s Dr. Graham Ogle shared powerful data showing that simply moving the state of diabetes care in low-resource countries from “minimal” (fingerstick testing only at clinic, human premixed insulin only, minimal education) to “intermediate” (2-3 fingersticks/day, human insulin MDI, diabetes education) could greatly improve A1c, reduce complications, and mortality.

  • Dr. Hirsch showed a slide from Inkwood Research, projecting the CGM market will grow to $15.2 billion by 2027 (32.1% compound annual growth rate). During the same time, they estimated the SMBG market will grow by $22.3 billion (5.2% CAGR).

Diabetes Therapy Highlights

1. Star-Studded Session Highlights Unmet Needs in Type 1 Diabetes and the Potential of SGLT Inhibitors for Cardio- and Renal- Protection (with Appropriate DKA Risk Mitigation)

Friday’s agenda kicked off with a star-studded session on the ever-relevant issue of SGLT inhibitors in type 1 diabetes, featuring Drs. Satish Garg, Chantal Mathieu, and Thomas Danne. Dr. Garg (Barbara Davis Center, Aurora, CO) set the stage with an overview of unmet needs in type 1 diabetes care. He underscored that despite vast improvements in insulin formulations, technology for insulin delivery, and glucose monitoring over the past decade, population-level outcomes show only a small fraction of the type 1 population meeting A1c targets, and an increasingly large fraction developing overweight and obesity. Furthermore, the total lifetime costs of type 1 diabetes (factoring in medication costs and direct and indirect costs of hypoglycemia and diabetes complications) totals ~$0.5 million per person per year, or about $1 trillion in lifetime costs for the US diabetes population of ~1.25 million. The bottom line, he emphasized, is that more treatment options for type 1 diabetes are sorely needed – particularly in light of the burden of overweight/obesity and diabetes complications such as cardiovascular and renal disease, all of which SGLT inhibitors are well-poised to address.

Turning to efficacy, Dr. Mathieu (KU Leuven, Belgium) ran through the phase 3 data for SGLT-2 inhibitors dapagliflozin (DEPICT) and empagliflozin (EASE) and SGLT-1/2 inhibitor sotagliflozin (inTandem) in type 1 diabetes, remarking on the within-class consistency in improving A1c, weight, blood pressure, and time-in-range. She noted, “in my hands as a clinician, it is one of the most effective drug classes, and you can see the effects from day 1.” On the negative side, SGLT inhibitors are also associated with increased risk of genital mycotic infections and, more concerningly, DKA. Dr. Mathieu underscored that risk mitigation strategies are essential to manage the increased risk of DKA, and for this reason she only recommends SGLT inhibitors to a select group of patients with strong adherence, high engagement in their diabetes self-management, relatively high insulin requirements, and no signs of underweight. In her clinic, this translates to 58 patients out of 1,400, only one of whom experienced DKA (in relation to an episode of gastroenteritis). The EMA has approved sotagliflozin and dapagliflozin for type 1 diabetes, but stateside there has been more resistance to the idea of SGLT inhibitors as adjunctive therapy for type 1 diabetes. Recently the FDA released a CRL for sotagliflozin (March 2019) and dapagliflozin (July 2019) and a November Advisory Committee Meeting on empagliflozin culminated in a 14-2 vote against approval for type 1 diabetes – all largely due to concerns over DKA risk.

On this note, Dr. Danne closed the session with a deep dive on risk mitigation strategies for DKA. He emphasized the need for much more comprehensive patient and provider education about DKA, as well as the importance of patient selection for SGLT therapy (the ideal patient for SGLT therapy, in terms of DKA risk, is highly engaged, not underweight, has relatively high insulin requirements, and is not on a low carb or ketogenic diet).

  • Echoing his commentary from day 1’s debate session on adjunct SGLT inhibitor therapy in type 1 diabetes, Dr. Danne underscored that while DKA risk is a concern to be addressed, we shouldn’t downplay the potentially game-changing benefits of SGLT inhibitors in terms of cardio- and renal protection. These are well-established for the type 2 diabetes population, and with the approval of dapagliflozin for heart failure we know that these benefits extend also to the diabetes-free population at high risk of CV disease. Dr. Danne argued that people with type 1 diabetes are at even greater risk of CV disease than this population, and arguably even the type 2 diabetes population: “It is very sobering to read that if you develop diabetes prior to age 10 it leads to 17 years shorter life expectancy for girls and 14 years for boys according to Swedish registry data.” Dr. Danne further argued that SGLT inhibitors should be studied in the pediatric type 1 diabetes population – “I think we shouldn’t wait until these children develop CVD to offer cardioprotective therapies.”

2. Locemia’s Co-Founder Mr. Robert Oringer Details Personal Narrative Regarding Development of Nasal Glucagon

Locemia co-founder Mr. Robert Oringer provided a heartfelt, detailed personal narrative regarding the development of nasal glucagon, subsequent acquisition by Lilly, and commercialization as Baqsimi. As a reminder, Lilly acquired worldwide rights to nasal glucagon from Locemia in October 2015 after successful phase 3 studies of the candidate had been completed in both adults (see coverage from ATTD 2015) and pediatrics (see coverage from EASD 2015). Mr. Oringer noted that his presence on the stage today was “highly unusual,” seeing as that normally when a large pharma company acquires a small company’s product, the founder of the small company “usually disappears into the sunset.” However, Mr. Orginer’s personal connection to the development is quite noteworthy and fully captured the audience at this Baqsimi symposium.

  • Mr. Oringer began his story 30 years ago, explaining that he was a young entrepreneur just beginning his relationship with diabetes management by selling thinner lancets and glucose tablets. In doing this, he began to gain an understanding of the specific aspects of diabetes care, especially in hypoglycemia. This relationship deepened when over the course of one month, both of his sons (one at age three and the other at nine months) were diagnosed with type 1 diabetes. Mr. Oringer soon realized just how deeply the fear and anxiety related to hypoglycemia impacted daily life for his entire family – he and his wife were constantly thinking about “who had our kids’ back” in emergency situations. His experiences having kids with type 1 deepened his understanding of how fear of hypoglycemia leads to situational avoidance, constant compromises, and loss of confidence: “I thought I had empathy during the early years when I was marketing diabetes treatments, but I realized that I didn’t have real empathy until I raised kids with type 1.” 

  • Mr. Oringer explained that he sat on this feeling for ten years after the diagnosis of his children, and decided that he wanted to do something so that his children “could feel that people had their backs” regarding the treatment of hypoglycemia. Mr. Oringer had no previous pharma experience, however, so he partnered with Dr. Claude Piche to tackle this problem. He and Dr. Piche soon realized in an epiphany moment that what is truly unique about glucagon is that it is a drug that is prescribed to patient, but intended for someone else to actually use for that patient. That third-party user is most likely unfamiliar with injections as well. The two partners therefore keyed in on four developmental goals for their glucagon candidate: it would have to be non-invasive, easy to use, easy to teach, and easy to carry.

  • Jumping off of this, they legally formed Locemia Solutions, a mission driven company focused on making hypoglycemia rescue treatment simpler (fun fact: Locemia comes from low blood sugar hypoglycemia). Locemia intended from the start to pursue the nasal route of glucagon delivery (past research had shown the viability of glucagon absorption in the nose), but soon realized that a liquid nasal glucagon approach was not feasible due to stability concerns. Therefore, a powder formulation was necessary. The next step was to create a device for nasal delivery (Mr. Oringer joked that the only such device known at the time was a rolled up $100 bill!). Through a series of steps, the device was finally made, and the nasal glucagon candidate successfully progressed through human factor studies as well.

  • After this, Lilly discovered the candidate, and Mr. Oringer fondly noted that “they embraced us and our innovation and really adopted our baby. Lilly has shown a deep commitment to fulfilling the promise of the product, and taking it through the necessary regulatory and commercial steps, which was something we could have never done on our own.” In a very nice closing touch, Mr. Oringer also explained the origin of the brand name “Baqsimi” – it’s based off of the idea that “having someone’s back” is the root of treatment for hypoglycemia, and ties in perfectly with Mr. Oringer’s personal narrative of the drug’s development. Mr. Oringer closed his touching remarks by expressing that his “real hope is that nobody will ever have to use Baqsimi.”

ATTD Yearbook

1. Continuous and Intermittent Glucose Monitoring

  • Slovenia’s Dr. Klemen Dovč highlighted three studies during his presentation on this year’s advances in CGM, flash glucose monitoring, and self-monitoring of blood glucose. ATTD has now moved to combining both intermittent and continuous glucose monitoring into one chapter, which contained a total of 14 publications for 2019. Of note, last year’s chapter on CGM (excluding IGM) was the second most read chapter in the entire yearbook. To start, Dr. Dovč featured the SMILE study, assessing the effect of Medtronic’s MiniMed 640G on hypoglycemia-prone adults with type 1 diabetes. Impressively, use of MiniMed 640G reduced the average number of hypoglycemia (<55 mg/dL) events per week from 4.4 to just 1.1. Dr. Dovč then pivoted to real-world data (n=900) from Scotland on NHS-funded flash glucose monitoring. A1c reductions were substantial, especially for those with a high A1c at baseline: -0.6% (95% CI: -1.2 to 0.1%, p<0.001) for those with an A1c ≥ 7.5% at baseline. Lastly, we were treated to one of our favorite studies on the cost-saving benefits of real-time CGM, using data from the CONCEPTT trial on pregnancy complicated with type 1 diabetes. There, analyses showed that implementation of real time CGM could save the NHS a substantial €9.6 million annually, driven by reductions in NICU admissions and shorter duration of NICU stay.

2. Insulin Pumps

  • Stanford’s Dr. Rayhan Lal presented the insulin pumps section of this year’s ATTD Yearbook, choosing to focus on structured education for pumps, dosing accuracy of pumps, and infusion sets. Dr. Lal presented Dr. Dominic Ehrmann’s paper in Diabetes Care on the effectiveness of education for pump therapy, Dr. Lal underscored the point that structured education around pump features is required in order to see the A1c improvement from switching from MDI to pumps, calling it “crucially important” to teach skills around the diabetes devices. Next, Dr. Lal showed Alderisio’s paper demonstrating weight gain comparable to MDI over 10 years, despite improved glycemic control and reduced insulin doses with CSII. Around this, Dr. Lal also called infusion sets the “weakest link in the chain.” Indeed, innovation in this space has been somewhat stalled for several years, though Medtronic recently received CE-Marking for a 7-day, extended wear infusion set. To finish, Dr. Lal showed a study from Dr. Guido Freckmann evaluating bolus dosing accuracy across several popular insulin pumps. While most pumps performed relatively well, the spread in delivery accuracy was noticeably poorer for Omnipod (MO below).

3. New Insulins, Biosimilars, and Insulin Therapy

  • On basal insulins, Dr. Bolinder reviewed results of the BRIGHT trial, which pit the two currently available next-gen basals on the market against each other (Sanofi’s Toujeo and Novo Nordisk’s Tresiba). See our past coverage of the trial here. Dr. Bolinder concluded from these results that both insulins provided “essentially similar” glycemic control and “comparable risk” of hypoglycemia. Looking ahead, he underscored that all trials of these next-gen basals in head-to-head settings have been in type 2 patients (see the similar CONCLUDE trial in type 2s as well) and that there should be similar trials performed in type 1 as well.

  • Dr. Bolinder touched on oral insulin and highlighted recent results from a phase 2 proof-of-concept trial for Novo Nordisk’s oral basal insulin I338. I338 is a long-acting basal insulin analog co-formulated in a tablet with an absorption enhancer, with key amino acid modifications to prevent gut degradation. The trial was done over eight weeks in insulin-naïve type 2 patients and showed similar effects on fasting plasma glucose levels (the primary endpoint) to insulin glargine. However, bioavailability was (predictably) a large concern, with end of trial doses of I338 >50 times that of the corresponding dose for the comparator insulin glargine. Further development of this candidate has been terminated by Novo Nordisk due to this issue. Still, Dr. Bolinder was relatively optimistic about the future of oral insulins in general, noting that the “concept has definitely been proven and is likely to see further exploration.”

4. Decision Support

  • Harvard’s Dr. Eyal Dassau took the Decision Support section of the Yearbook, focusing his chapter on AID algorithms and clinical decision support systems – both hot topics at this year’s ATTD. Dr. Dassau began with a valuable graph showing the Time in Range and time <70 mg/dl results for AID studies from 2014-2019. Of course, Control-IQ’s pivotal trial was the biggest AID study of 2019, published in NEJM in October. Dr. Dassau then shifted to clinical decision support systems, focusing on insulin dosing systems. In particular, Dr. Dassau pointed at Dr. Rich Bergenstal’s publication in The Lancet, which showed a full 1% A1c drop for the group using Hygieia’s d-Nav insulin dosing guidance (baseline: 8.7%) with HCP support compared to 0.3% for a control group with HCP support alone (baseline: 8.5%).

5. Technology and Pregnancy

  • It was a busy ATTD for Dr. Helen Murphy (King’s College), who presented the technology and pregnancy section of the Yearbook. Most notably, Dr. Murphy began with a “plea [for her] US colleagues” to notice the “excellent” evidence base and encourage use of CGM in pregnancy. Indeed, Dr. Murphy showed T1D Exchange data showing just one-third of pregnant women using CGM; nearly three-quarters used pumps, even though the evidence base for pumps in pregnancy is “much more limited.” Dr. Murphy looked at a real-world Swedish study demonstrating that mean Time in Range was ~40% at the beginning of pregnancy, increasing to ~50% by week 10, before plateauing again until the third trimester. Citing the consensus CGM goals for pregnancy, Dr. Murphy emphasized the need for the healthcare system to do a better job taking care of pregnant women, so that they are achieving the targets through the entire pregnancy: “We aren’t aiming for 70% Time in Range right at the end [of pregnancy], but sooner in the gestational age.”

6. Diabetes Technology and Therapy in the Pediatric Age Group

  • Dr. Maahs broadly tackled the question of whether these technologies and therapies are different in pediatrics, noting that “yes they are” and pointing to new pieces of data in the field. He began with a paper comparing international data from the T1D Exchange and DPV, showing that CGM usage in pediatrics has increased in both the US (3% to 22%) and Germany (4% to 19%) between 2011 and 2016, and that this use has been associated with lower A1c’s. He then touched on the PROLOG trial of Basal-IQ in pediatrics, which demonstrated a 31% reduction in hypoglycemia (<70 mg/dl), with no increase in mean glucose vs. SAP. Next, he pointed to results of 12-week results of closed loop in type 1 (nearly half of participants were <21 years old), showing 11% time in range improvements, 0.4% A1c drop, and less time in hypoglycemia vs. control. He explained that results from this study and others support adoption of closed loop tech in clinical practice across all age groups – a common theme of data presented at ATTD 2020 more broadly. Regarding progress of diabetes therapies in pediatrics, Dr. Maahs highlighted Victoza’s recent approval in adolescent type 2 diabetes as a welcome addition to the treatment armamentarium in this population.

7. Advances in Exercise Physical Activity and Diabetes Mellitus

  • Canada’s Dr. Michael Riddell (York University) presented the Exercise chapter of the Yearbook, noting that at least 8 papers were published in 2018/2019 around the topic of exercise in closed loop systems. Dr. Riddell also noted that more research needed in the area, particularly around how time of day of exercise affects glucose profiles. In the FIT study, Aronson et al. showed that morning (fasted) high intensity exercise in type 1s generally led to all-day hyperglycemia unless insulin correction was taken. In a study on timing of exercise in type 2s, Savikj et al. found that afternoon was superior to morning exercise for glucose control.

8. Practical Implementation of Diabetes Technology

  • The seemingly omnipresent Dr. Laurel Messer (Barbara Davis Center) presented “Practical Implementation of Diabetes Technology,” choosing to focus on three themes: (i) access and cost effectiveness; (ii) tech education; and (iii) automated decision support. On the first theme, Dr. Messer showed a map of CGM reimbursement in Europe (see below): on the left, representing 2009, the map is full of red and yellow, demonstrating poor access; on the right, representing 2018, the map is mostly green, with a few red countries. On the second theme, Dr. Messer showed T1D Exchange data demonstrating an increase in CGM use, but worsening glycemic control across the population. The answer, according to Dr. Messer, is more structured education. Backing her answer, Dr. Messer highlighted results from the INPUT study for insulin pumps and FLASH trial for FreeStyle Libre. Lastly, Dr. Messer noted the need for clinical decision support systems as patients and providers are increasingly dealing with diabetes devices and the data generated by these devices. On this front, she specifically highlighted DreaMed’s Advisor Pro system (see ATTD Day #2 highlight) and Hygeia’s d-Nav insulin dosing guidance system.

9. Diabetes Technologies and the Human Factor

  • Dr. Alon Liberman urged audience members not to forget the importance of “the human factor” when considering the many recent wins in diabetes technology. As he put it, “technologies themselves are rarely sufficient to deliver sufficient outcomes,” and researchers and clinicians must remember that technology is only one piece of the puzzle that will enable patients to achieve better outcomes. As an example, Dr. Liberman featured the 2018 DT&T publication, “Psychosocial and Human Factors During a Trial of a Hybrid Closed Loop System for Type 1 Diabetes Management,” which found decreased levels of diabetes distress and improved attitudes towards diabetes technology using the MiniMed 670G system. Dr. Liberman also discussed an intriguing paper published in Pediatric Diabetes on the relationship between hope (ability to set goals, sustain motivation toward goals, and recruit supportive resources toward goals), depression, and glycemic control in youth with type 1 diabetes. Results from the study suggest that hope may be a key factor in helping offset the impact of depressive symptoms associated with glycemic control, and research that works to foster hope in patients could therefore have significant benefit.

10. Immune Interventions for Type 1 Diabetes

  • Dr. Desmond Schatz touched on two recent breakthrough studies in the type 1 immune therapy space – the landmark teplizumab prevention trial and a trial of low-dose anti-thymocyte globulin, both with results presented initially at ADA 2019. Dr. Schatz noted that over the past ten years, hundreds of studies have been completed or are in process in terms of type 1 immunotherapies, but these two trials represent some of the most promising results to date. Of course, the teplizumab trial has been the source of much optimism, as it represents the first trial to show that immune therapies can actually be used to delay type 1 diabetes. Excitingly, the drug was recently granted Fast Track designation by the FDA. For more, see our full type 1 cure/prevention/treatment landscape here.

11. New Medications for Treatment of Diabetes

  • Dr. Irl Hirsch highlighted four therapies in the pipeline that he and his team are particularly excited about. First, he directed attention to tofogliflozin, which data show to be the most selective SGLT-2 inhibitor available (it’s only approved in Japan). Its selectivity for SGLT-2 is comparable to empagliflozin, twice as selective as dapagliflozin, and “much more” selective than canagliflozin. Some researchers have posited that this higher selectivity may lead to better outcomes on CV and renal outcomes, although this is not yet supported by data. As a reminder, Roche/Chugai had been pursuing development of tofogliflozin but discontinued further investment in 2Q13 for unspecified reasons.

    • Dr. Hirsch then moved to licogliflozin, a dual SGLT-1/2 inhibitor that displays “really interesting and dramatic” reduction in glucose excursions and significant weight loss. However, side effects are a major concern, with 91% of patients with obesity in a 12-week weight loss trial of the candidate having diarrhea. Dr. Hirsch believes that licogliflozin could be an attractive strategy in obesity. Currently, Novartis is studying the candidate in NASH after discontinuing development in obesity in 2Q18 because of efficacy concerns.

    • Empagliflozin in NAFLD was also discussed: results from small trials show promising results in terms of liver fat decreases, leading Dr. Hirsch to wonder whether SGLT-2 inhibitors (potentially in combination with pioglitazone) could work in NAFLD, regardless of type 2 status.

    • Finally, Dr. Hirsch reviewed landmark phase 2b data of Lilly’s dual GLP-1/GIP agonist tirzepatide. This data was first presented at EASD 2018 and was staggering in terms of its effects on A1c, weight loss, and adverse events. Dr. Hirsch provided some context on the effects of GIP agonism, noting that GIP is secreted from enteroendocrine K cells, and like GLP-1, is a potent stimulator of glucose-dependent insulin secretion. One thought concerning how GIP delivers its effects is that it might increase metabolic flexibility by enabling increased fat utilization in the fasting state and reduce fat availability in the postprandial state. Dr. Hirsch closed his remarks with this: “what type of drug will this be? My guess is an expensive one!” Indeed, if efficacy holds up in Lilly’s ongoing phase 3 SURPASS program for tirzepatide (which excitingly also includes a head to head CVOT vs. Trulicity), then interest in tirzepatide should be very significant.

12. NAFLD/NASH and Diabetes

  • Dr. Kavita Garg presented on the newest addition to the ATTD Yearbook’s collection of chapters – NASH and NAFLD. Out of the 400 relevant articles, 19 were selected to be in this year’s chapter. We've certainly noticed growing interest on the subject at recent conferences, hwich is no surprise, considering the 35-40% prevalence in people with diabetes. Dr. Garg broke down the 19 featured articles into four broad categories: those discussing (i) global prevalence rates of NASH/NAFLD; (ii) preventative strategies to optimize type 2 and NASH/NAFLD management; (iii) non-invasive biomarkers to diagnosis the conditions and avoid the necessity of liver biopsy; and (iv) novel pharmacotherapies. On the third point, highlighted biomarkers included elevated ALT, triglycerides, and BMI, as well as improved imaging with US and MR elastography and CT. We were inspired by Dr. Garg’s quote that “Ten years ago, we used to see steatohepatitis on MRI/CT, and we would just ignore it because we didn’t understand it…Now, we report it as accurately as we can.” To end her presentation, Dr. Garg posed the interesting question of whether individuals with diabetes should be screened for NAFLD/NASH in the same fashion as diabetic retinopathy or nephropathy – once it is easier to screen, that may make more sense. Looking to the future, we hope to see diagnostic tools that can make this thought a reality.

13. Using Digital Health Technology to Prevent and Treat Diabetes

  • McGill University’s Dr. Tal Oron stepped in for chapter editors Drs. Neal Kaufman and Eran Mel to give a thematic presentation on advances in digital health technology. According to Dr. Oron, while there are many publications on digital health in diabetes, few have been able to deliver meaningful results. This fact reflects how relatively new a field digital therapeutics is, as well as the high cost and time barriers required of impactful studies. Dr. Oron focused the majority of his remarks on the importance of self-efficacy as a goal for digital health interventions. Individuals with higher self-efficacy look at challenging tasks as tasks to be mastered, have a deeper interest in and commitment to activities, and recover quickly from setbacks. This self-efficacy translates into accelerated positive health behavior change and reduced depression and stress-related disorders for patients. To illustrate this type of intervention, the editors chose to feature a cost-saving analysis on Livongo’s remote diabetes management program. Program access was associated with a cost reduction of $88 saving per member per month at one year on top of the self-management tools provided by Livongo.


Editor's note: see below for our original, abridged ATTD 2020 Day #3 report. 

It was another jam-packed day in Madrid, as ATTD 2020 rolled on to day #3. A late afternoon series of oral presentations headlined the day, featuring data readouts from the Tidepool/Jaeb Loop observational study and iDCL3 (Control-IQ pivotal) extension. We’ll be back soon enough with our full day #3 report – in the meantime, make sure you see our reports from days #1 and #2, as well as a few key summaries from day #3 below.

ATTD Day #1 HighlightsDr. Battelino: put DCCT in “history part” of your brain, Dr. Bergenstal: FNIR; Medtronic cost-savings data with iPro 2 CGM; opening ceremony

ATTD Day #2 Highlights – 18 highlights across CGM, AID, decision support, drugs, and big picture: WISDM extension data, Dexcom G6 in pregnancy, Control-IQ 6-14 years, Advisor Pro study + more

1. Results from the Loop observational study show that “new” Loop users saw TIR improvements 

Jaeb’s Dr. John Lum read out results from the Loop observational study, showing that “new” Loop users (n=607) saw a Time in Range improvement from 67% at baseline to 73% during months 4-6 (p<0.001). Over the same time period, A1c fell to 6.4% from 6.8% at baseline (p<0.001), definitely a meaningful improvement – 86 minutes more per day Time in Range or nearly 1.5 hours! The improvement in Time in Range was delivered primarily by reductions in hyperglycemia, while time below 70 mg/dl stayed steady from 2.9% at baseline to 2.6% during months 4-6 (four minutes a day less in hypoglcyemia). Notably, median time spent in closed loop dropped slightly from 83% during months 1-3 to 79% in months 4-6 – we are not sure why time out of closed loop would’ve dropped below 20%, as the numbers certainly show the value of staying in closed loop. The study also showed Loop was safe to use, as the entire cohort recorded one DKA event (not related to Loop) and rates of severe hypo events fell on Loop (it was not specified to what and presumably was not significant [nor would it have been powered as such]). Notably, Dr. Lum did note that Loop showed statistically significant improvements in patient-reported outcomes, such as diabetes distress, sleep quality, and hypo fear – this data was not shown today and we’ll be very eager to see it. Data from the study will be used to help Tidepool get Loop as a regulated app on the App Store, with FDA submission planned for later this year.

2. Dr. Sue Brown presented data from the 13-week extension phase of iDCL3

Dr. Sue Brown (University of Virginia) presented data from the 13-week extension phase of iDCL3 (Control-IQ’s pivotal trial, read out at ADA 2019). Of the 112 participants in the closed loop arm of the original trial, virtually all (n=109) elected to continue into the extension phase. Fifty-five were randomized to the predictive low glucose suspend (Basal-IQ) while 54 remained on closed loop (Control-IQ). The two groups were well-matched at baseline (i.e., the end of the main trial): Time in Range and A1c in the Basal-IQ and Control-IQ groups were 70% and 7.1% (Basal-IQ) and 71% and 7% (Control-IQ), respectively. After 13-weeks, the Basal-IQ group saw Time in Range drop from 70% to 60%, while the Control-IQ group saw a far smaller drop from 71% to 68%. Notably, the vast majority of this difference was in time >180 mg/dl: the Basal-IQ group saw an increase from 28% to 38%, while the Control-IQ group increased from 28% to 31%, reinforcing the power that Control-IQ has on hyperglycemia in particular, compared to Basal-IQ. A1c in the Basal-IQ group increased from 7.1% at baseline to 7.5% after 13 weeks; A1c in the Control-IQ group stayed virtually steady, increasing only from 7.1% to 7.2% (this was a bit surprising given that hypoglcyemia was fairly constant while hyperglycemia dropped – we hypothesize that this may have been just accuracy of A1c, and that in fact there was little change in A1c at all). Neither group recorded any severe adverse events, though there were three hyperglycemia events with ketoacidosis in the Basal-IQ group vs. none in the Control-IQ group. We are not sure if any required a visit to the hospital.

3. A star-studded session on the ever-relevant issue of SGLT inhibitors in type 1 diabetes featured Dr. Satish Garg, Professor Chantal Mathieu, and Professor Thomas Danne

Headlining the drug side, we watched a star-studded session on the ever-relevant issue of SGLT inhibitors in type 1 diabetes, featuring Dr. Satish Garg, Professor Chantal Mathieu, and Professor Thomas Danne. Dr. Garg set the stage with a discussion of unmet needs in type 1 diabetes care. He emphasized that more therapies for type 1 diabetes are sorely needed, particularly in light of the growing burden of overweight/obesity and diabetes complications such as cardiovascular and renal disease, all of which SGLT inhibitors are well-poised to address. Although there is no published data on CV and renal risk reduction stemming from SGLT-2 inhibitor use in type 1 patients, there is expert opinion – Dr. John Buse was quoted at FDA last year noting that there is a 95% chance that SGLT2i use in people with type 1 diabetes with CKD or heart failure would reduce heart failure hospitalizations and CKD progression in patients. Turning to efficacy, Dr. Mathieu ran through the phase 3 data for SGLT-2 inhibitors dapagliflozin (DEPICT) and empagliflozin (EASE) and SGLT-1/2 inhibitor sotagliflozin (inTandem) in type 1 diabetes, remarking on the within-class consistency in improving A1c, weight, blood pressure, and Time in Range on the positive side, but increasing the risk of DKA on the negative side. The EMA has approved sotagliflozin and dapagliflozin for type 1 diabetes, but stateside there has been more resistance to the idea of SGLT inhibitors as adjunctive therapy for type 1 diabetes and without a partner, we imagine sotagliflozin will certainly not be commercialized. The FDA sent CRLs for sotagliflozin to Lexicon in March 2019 and for dapagliflozin to AZ in July 2019. More recently, last November, an Advisory Committee Meeting on BI/Lilly’s empagliflozin culminated in a 14-2 vote against approval for type 1 diabetes – all largely due to concerns over DKA risk (as well as, in BI/Lilly’s case, the size of the trial, which we were surprised had not been conveyed to them before the meeting). Regarding risk management, Dr. Danne closed the session with a deep dive on risk mitigation strategies for DKA. He emphasized the need for better patient and provider education about DKA, as well as the importance of patient selection for SGLT therapy. He emphasized, in closing, that the ideal patient for SGLT therapy, in terms of DKA risk, is highly engaged, not underweight, has relatively high insulin requirements, and is not on a low carb or ketogenic diet.

4. A heartfelt, personal narrative of the development of nasal glucagon

At a Lilly symposium, Locemia co-founder Mr. Robert Oringer provided a heartfelt, detailed personal narrative regarding the development of nasal glucagon, subsequent acquisition by Lilly, and commercialization as Baqsimi. Mr. Oringer also explained the origin of the brand name “Baqsimi” – it’s based off of the idea that “having someone’s back” is the root of treatment for hypoglycemia, and ties in perfectly with Mr. Oringer’s personal narrative of the drug’s development.

5. Dr. Nick Oliver confirms that PRO Solo study results likely to be shared at EASD 2020

At a Roche symposium, Dr. Nick Oliver (Imperial College London) confirmed that the PRO Solo study, a three-arm trial in Austria, Germany, Poland, and UK comparing treatment satisfaction with the Accu-Chek Solo “micropump" system vs. MDI and Solo vs. Omnipod (n=181, pump-naïve type 1s), has completed recruitment and screening, with results expected to be presented at EASD 2020. The study is expected to complete in February 2020, seven months after the original July 2019 goal. During his presentation, Dr. Oliver presented some additional data comparing the accuracy of Solo vs. Omnipod, finding that with single 1 U bolus doses (225 repetitions), Solo had 99% of bolus doses within ±15% relative error compared to Omnipod’s 65% and a flow rate accuracy (basal rate 1 U/hr, 1-hour windows over 72 hours, 9 repetitions) of 98% within the same relative error zone vs. 81% in Omnipod. Accu-Chek Solo most recently launched in Australia, Spain, Argentina, and Italy, following a pilot among ~200 people in Austria, Switzerland, Poland, and the UK.

6. Fascinating retrospective Dexcom G6 Data (TIR, A1c, glycemic variability, and more)

Dexcom’s symposium featured some fascinating retrospective data, highlighting Time in Range, A1c, and glycemic variability improvements among patients with higher daily G6 feature-use including Clarity, Share/Follow, and Urgent-Low-Suspend (ULS). Patients deemed as “low” engagement (defined by tabulated daily use of these features) had a Time in Range of 56% compared to 61% in those with “high” engagement – while this may not sound like a big difference, it represents well over an hour’s worth of time per day – 72 minutes, to be exact, which we’re sure any patient would take any day, from any baseline, all else equal. Notably, patients with the highest level of engagement also observed 38% of time less in hypoglycemia (<54 mg/dl) and had reduced glycemic variability, posting a %CV less than or equal to 36% of 64% vs. 57%. Wow! We do note that there continue to be some discussion in the field on “what CV should be” – more to follow on this.

7. A rural Chilean case study shows that importance of telemedicine

Ms. Julie Pelicand (Universidad de Valparaíso) shared insights on the perception and impact of integrated telemedicine among families living in rural Chile. Her study followed 21 families of type 1 children who used telemedicine for their diabetes care and follow-ups over two years. In addition to improved health outcomes, telemedicine use contributed to greater patient empowerment, better relationships with healthcare teams, and more efficient use of resources. These results emphasize telemedicine’s potential role in increasing health access and education in areas of high need. We look forward to returning with full data on this front.

8. Brian Levine on the future of connected care 

OnDuo’s Brian Levine provided clarity on the emerging connected diabetes care arena. The impressive young healthcare leader defined a connected care platform as one with remote coaching, often delivered through a smartphone app, that may receive data from connected care devices. After classifying current platforms, Mr. Levine gave us a look into the future of connected care, suggesting that the ultimate goal is a spectrum of virtual care that increases health access and decreases burden on both the patient and the payer. Notably, he was first author on a comprehensive paper in Diabetes Technology & Therapeutics earlier this year, “Reviewing U.S. Connected Diabetes Care: The Newest Member of the Team.”

9. Dr. Serena Dass's application of artificial intelligence on predicting pediatric DKA admissions

Dr. Serena Dass shared her application of artificial intelligence on predicting pediatric DKA admissions. By using a positive predictive value model on two years of data from the study cohort, her team was able to predict if, when, and how likely a patient was to be hospitalized for DKA. Through this technique, they were able to target those best suited for intervention. However, by using the same cohort to train and test their model, this study’s exciting findings may not be generalizable. More to follow on this.

We will be back with so much more very soon!


--by Jane Seeley, John Close, Sahaj Shah, Sara Suhl, Ani Gururaj, Abigail Dove, Rhea Teng, Martin Kurian, Albert Cai, and Kelly Close