American Diabetes Association 80th Scientific Sessions

June 12-16, 2020; Virtual; Full Report – Diabetes Technology – Draft

Executive Highlights

  • The first virtual ADA Scientific Sessions were a great success, and this report brings you our full coverage of diabetes technology. 

  • On the tech side, it was a remarkable year for AID, as ADA featured real-world and pediatric data from one new AID system (Tandem’s Control-IQ), pivotal data from a second-generation AID system (Medtronic’s MiniMed 780G), and pre-pivotal data from another system to launch in 2021 (Insulet’s Omnipod 5). The results from the three systems are promising:  not only were glycemic outcomes improved but time spent in closed loop was >90% in all three systems, which indicates that AID is becoming easier to use. That said – that’s not the metric we care most about. What we care most about are the measures that enable greater time in range. That means less hyperglycemia, particularly fewer spikes above 250 mg/dL (above 14 mmol), and less hypoglycemia, or data below 54 mg/dL (below 3 mmol). We were excited at every turn to hear about time in range and far more references to AGP. Notably, we also saw data from Beta Bionics’ iLet (insulin-only pivotal enrollment “nearly complete”) and the first data from Lilly’s AID system.
  • Evidence around CGM continues to build – this year, we noticed a focus on CGM in type 2s and data around cost-savings and ROI (return on investment). A number of studies showed significant reductions in adverse events after initiating CGM, including DKA, severe hypo, and all-cause hospitalizations and we are hearing this post-ADA as well (check out this impressive Medtronic data from its partnership with United Health Care, and this is 670G, not even 780G yet). Some presenters, namely Dr. Irl Hirsch, took this data one step further, estimating the amount of total healthcare cost saved from these reductions in adverse events. Elsewhere, a poster from Abbott (78-LB) found very strong A1c reductions in non-insulin-using type 2s using FreeStyle Libre and a number of extension studies comparing CGM vs. SMBG confirmed that the advantages of CGM are maintained through several years.

This report contains our full coverage of diabetes technology from the ADA’s 80th Scientific Sessions. Talk titles highlighted in blue represent new coverage that wasn’t included in our daily highlights (day #1day #2day #3day #4, or day #5). Note that some talks may appear in multiple sections.

Our sections proceed as follows (you can navigate through by using our click-able table of contents below):

  • Themes

  • Automated Insulin Delivery, Pumps, and Pens

  • Glucose Monitoring – BGM and CGM

  • Digital Health, Telemedicine, and Decision Support

  • DiabetesMine D-Data Exchange

  • Exhibit Hall

Table of Contents 


Diabetes Technology

1. Automated Insulin Delivery: Convincing Real-World Data from Control-IQ, MiniMed 780G Pivotal Shows Strong Improvement From 670G, Encouraging Pre-Pivotal Data from Insulet

  • After Tandem’s breakout 2019 (pivotal read-out at ADA 2019, publication in NEJM, and de novo 510(k) clearance for Control-IQ as an iController), the company presented two posters with very strong real-world data in early Control-IQ adopters. As hinted by Tandem CEO John Sheridan (our ADA interview with him is still to come) in the company’s 1Q20 earnings call, the real-world data from the first ~1,000-2,000 Control-IQ adopters was even better than that of the pivotal trial. In 1,659 subjects with at least 30 days of pre- and post-Control-IQ data in t:connect, time in range was increased by 2.4 hours/day to 78% time in range and spent a remarkable 96% of time in closed loop. This is incredibly exciting to see, though perhaps not surprising – the early adopters of Control-IQ are likely some of the most engaged people with diabetes, while the pivotal for Control-IQ had broad inclusion criteria (no entry restrictions on A1c, severe hypo or DKA, or device experience). Definitely, we would expect the time in range numbers to come down as more and more people get on Control-IQ but we wouldn’t be surprised if the improvements seen with Control-IQ remain very strong.

  • MiniMed 780G dominated day 1 of ADA with results from its US pivotal, head-to-head comparison with MiniMed 670G in FLAIR, and its New Zealand study used for CE-Marking. Before ADA, Medtronic seemed to lower expectations for the MiniMed 780G pivotal data – the previously advertised goals of >99% time in closed loop and >80% time in range (which we recognized, at the time as ambitious) have not been as frequent in Medtronic’s statements and marketing as ADA approached. Still, the data from MiniMed 780G showed a very significant step-up from MiniMed 670G and approval of the 780G will be a major move ahead for most using 670G (or 770G). In the head-to-head crossover comparison of MiniMed 780G vs. 670G in teens and young adults, MiniMed 780G showed very significant improvements in usability: the number of closed loop exits per week was 5.7 with 670G vs. 1.7 with 780G. In another big step up, MiniMed 780G enables the system to be more aggressive if the user wishes it to be and Friday’s studies showed better outcomes with those more aggressive settings. In all three trials, using MiniMed 780G with a set point of 100 mg/dl (vs. 120 mg/dl) saw increased time in range ~75% with no significant tradeoff in time in hypoglycemia. Similarly, the MiniMed 780G’s active insulin time allowed for users to turn up the system’s aggressiveness; again, with shorter active insulin time settings, time in range trended higher without increasing time in hypoglycemia. We’d imagine most users will opt for the default 100 mg/dl set point with MiniMed 780G, while tuning the system’s aggressiveness using the active insulin time setting.

  • Insulet also took part in the AID-related sessions with its newly-named “Omnipod 5 powered by Horizon” closed loop system. We like this name. Dr. Bruce Buckingham presented data from adults and children in a pre-pivotal trial, highlighted by a very positive perception of usability, where Horizon scored a 90 out of 100 for parents of young children and a 93 for teenagers and an impressive 97% of time spent in closed loop. For a long time, we’ve heard Insulet talk about designing this system with simplicity and usability as a priority and on first impression, that focus seems to be paying off. For context, Apple’s iPhone scores ~79 on the same scale. Lastly, the early data from Omnipod 5 showed very promising time in hypoglycemia results; adults spent just 0.5% of time <70 mg/dl and children spent just 0.9% of time , or about 13 minutes, <70 mg/dl. Anecdotally, there is a lot of excitement for a closed loop system from Insulet and the early data will only fuel that excitement. Omnipod 5 is currently in pivotal trial with launch expected sometime in early 2021 (we feel this could easily move to “2021” just given all the uncertainty).

  • Beta Bionics shared that enrollment for its insulin-only pivotal for Gen 4 iLet is nearing completion and we saw the first-ever data from Lilly’s AID system. The insulin-only pivotal for Beta Bionics’ iLet should begin this year with launch plans as early as ~early/mid-2021 though like with Lilly, we’ll have to see how fast the FDA moves – some of this will be well beyond the control of the organization building the technology.

  • Notably, for the first time, Lilly shared some data on its AID system, which uses Dexcom CGM, an in-house pump, and an algorithm from Class AP, from a small inpatient feasibility study of the system in response to high-carb meals with no pre-meal boluses. The only time we had really discussed this was in early May 2018 when we were invited in house to see this technology. Lilly has never shared specific timing for development of its AID system, though it has broadly said that Connected Care products will launch in stages over 2019-2021; we wouldn’t be surprised to see this timeline pushed back. Lilly has such huge potential to make a lasting impression on so many patients’ lives given its broad product portfolio and its enormous understanding of patients – we hope to see the entire company come together with this introduction. We also hope it backs off the references to a “closed” system that we’ve heard over time – we’re excited it appears to continue to plan to work with Dexcom, which is known for its prioritization of open systems.

2. Quick Calculations, Detailed Analyses, and Reductions in Adverse Events All Show It’s Time to Pay for CGM

  • The evidence base for CGM continues to build – extension studies of CGM vs. SMBG in various populations headlined much of ADA 2020. The Swedish GOLD study extension (n=107), comparing Dexcom G4 with SMBG in MDI users, demonstrated significant A1c and time in range improvements out to 2.5 years. In the pediatric population, results from the 12-month extension of SENCE RCT (ages 2-7 years) showed CGM utilization remained high after one-year. In general, however, we do note – we want to see better time in range, less time in hypoglycemia and severe hypoglycemia, and less time in “high” hyperglycemia. While we are happy at this meeting to be reporting on high utilization of CGM, that’s not a ‘thing” anymore following this meeting – high utilization is just the minimum threshold of acceptability, because if it is not reported, something is wrong. Overall, SENCE results didn’t blow us away; on a less positive note, the study also demonstrated the huge work still to be done in improving outcomes for this young population: subjects in the study spent just ~40% time in range. Data from Onduo showed that their virtual diabetes clinic delivered 2x greater A1c reductions when used with CGM compared to SMBG.

  • Several important studies at ADA found massive reductions in adverse events related to CGM use. A pre-post analysis of commercial and Medicare supplemental insurance claims among ~2,500 adults with type 2 diabetes using short- or long-acting insulin using FreeStyle Libre showed a 60% reduction in acute diabetes events and 33% in all-cause hospitalizations. Impressively, the separation in acute diabetes events with and without FreeStyle Libre appeared very quickly – after six months, there was already a 60% overall reduction in the frequency of acute diabetes events. Another large observational study in France showed that rates of DKA were cut in half following FreeStyle Libre initiation. Similarly, self-reported data from Tandem’s Basal-IQ (Dexcom G6 + t:slim X2 + predictive low glucose suspend) users showed significant improvements in adverse events. With Basal-IQ, 665 type 1s reported a 45% reduction in hypo-related paramedic visits, 77% reduction in ER visits, and 75% reduction in hospital admissions.

  • Armed with data around glycemic improvements and reductions in adverse events, we saw a number of presentations calculating potential cost savings with CGM. The presentations ranged from Dr. Irl Hirsch’s “back-of-a-napkin” math to detailed budget analyses, but they all pointed to the same conclusion: CGM can reduce healthcare costs. Self-proclaiming himself as “not a healthcare economist,” Dr. Hirsch estimated that CGM could reduce DKA-related hospitalization costs in the US by nearly $5 billion. He also cited a real-world, prospective study from Belgium showing FreeStyle Libre could reduce hospitalizations, hypoglycemic comas, work absenteeism, and hospitalization days, all helping to lower overall healthcare costs. Later during ADA, healthcare economist Mr. Michael Minshall walked through his calculation that providing Dexcom G6 to all type 1 Medicaid beneficiaries (~600,000 people) would net Medicaid ~$300 million in annual savings, driven by a $1.15 billion reduction in costs (vs. ~$850 million cost to provide Dexcom G6 to all type 1 Medicaid beneficiaries).

  • CGM for all? The evidence base around CGM in type 1s is strong and the data around cost savings is convincing. At this year’s ADA, we also saw exciting studies of CGM in type 2s, including type 2s not using insulin. As noted above, one pre-post study of FreeStyle Libre in all insulin-using type 2s found that FreeStyle Libre showed a 60% reduction in acute diabetes events and 33% in all-cause hospitalizations. A poster at ADA, presented by Dr. Eden Miller (Diabetes Nation) showed significant A1c reductions in both basal-only and non-insulin using type 2s after one year of FreeStyle Libre: non-insulin using type 2s saw their A1c drop from 8.5% to 7.6% with FreeStyle Libre and basal-only type 2s saw their A1c drop from 8.5% to 7.9%. Given the sizeable A1c reductions, presumably (based on Beck/Bergenstal DCCT follow up data, we’d almost certainly see reductions in cost and long-term complications, as well, if the study were longer. In a debate on the value of CGM in type 2s, both sides agreed on the obvious clinical utility for CGM in this population and the conversation mostly focused on who, when, and how often should type 2s use CGM. This itself is a noticeable, and positive, shift in the type of debate and conversation that might have been even just a couple of years ago.

3. Medtronic Diabetes: New Leadership and $337 Million Blackstone R&D Investment – What Will It Mean for Patients and Providers?

It was the first ever ADA for Medtronic under its new leadership, both the incoming CEO (Geoff Martha following outgoing Omar Ishrak) and Head of the Diabetes Group (Sean Salmon coming in for Hooman Hakami). Mr. Salmon got rave reviews after the analyst meeting as well as diaTribe’s Musings under the Moon. There, the topic was “The Law of Inertia during a Global Pandemic: What's Here to Stay and What's Next in Diabetes Care,” and Mr. Salmon was very prescient. Medtronic held its annual Analyst Day on Friday of ADA with Mr. Salmon, Diabetes CMO Dr. Robert Vigersky, and Vice President of R&D for Medtronic Diabetes Ali Dianaty. As Mr. Dianaty walked investors through the Diabetes pipeline, we noticed the absence of anticipated timelines for regulatory submissions and launches. When we got a chance to talk to the Medtronic team on Saturday morning, Mr. Salmon told us the company doesn’t want to “overpromise” to its patients and providers. Certainly, like virtually all other organizations, there have been plans on FDA’s part that have changed, prompting Medtronic to miss on articulated timelines – though this is prudent, we missed the excitement of watching what might happen when. Ultimate of course, this is a smart decision by the new leadership.

  • Medtronic announced a $337 million R&D investment from Blackstone Life Sciences during ADA and it was much-discussed during the week among the more commercial players. The funding will be used for “four identified Diabetes R&D programs” and if successfully commercialized, Medtronic expects to pay royalties in the “low- to mid-single digit range.” There were no hints of what projects the funding would be used for, aside from “specific programs in [Medtronic’s] pump and CGM pipeline … beyond PCL [Personalized Closed Loop] and Synergy.” During the Analyst Day Q&A, several analysts asked about a potential patch pump project from Medtronic. A patch pump wouldn’t be a huge surprise from Medtronic, given the commercial marketplace and benefits of a patch pump; we’ve seen patch pump projects move in and out of Medtronic’s publicly disclosed pipelines before. Most recently, at Keystone 2019, Medtronic showed a slide with an all-in-one patch that includes a pump, CGM, and algorithm.

  • In the nearer term, Medtronic’s CGM pipeline must remain a priority. In a conversation with Mr. Salmon, he emphasized that CGM was the fastest-growing part of the Diabetes business, though it wasn’t growing as fast as the “[CGM] market, more broadly.” On the CGM front, Medtronic is facing very stiff competition from Abbott and Dexcom and Guardian Sensor 3 remains the weakest component of Medtronic’s AID systems. We were impressed with Mr. Salmon’s humility and focus on reinvigorating the CGM pipeline, but ultimately, the company will have to deliver on its Synergy sensor (fully-disposable, day 1 calibration, 7-day wear) sooner, rather than later, to stay within striking distance of the competition.

4. Telehealth and Digital Health: Increasing Access and Improving Outcomes for Rural and Underserved Patients

  • Advanced Comprehensive Diabetes Care (ACDC), a rural telehealth intervention among veterans with high baseline A1c, drove a 1.4 percentage point A1c reduction after six months. Despite the promise of telehealth, which has gained prominence since the COVID-19 pandemic (see the diaTribe Foundation's Musings 2020 and Insulet webinar), intensive interventions have rarely been implemented in standard diabetes care due to a lack of trained staff, lack of equipment, limited integration with electronic health records, and poor reimbursement. To address this implementation gap, the Department of Veterans Affairs (VA) has invested in a nationwide network of Home Telehealth (HT) nurses for the telemonitoring of diabetes and other chronic conditions. ACDC leveraged the VA’s HT network and EHR infrastructure to deliver intensive telehealth care to rural veterans with diabetes. In the initial cohort, mean A1c was reduced one percentage point at six months among patients randomized to ACDC (n=50; baseline: 10.5%). Across patients who enrolled in the program in 2017 to 2019 (n=125; 5 sites), there was a mean A1c reduction of 1.4 percentage points after six months (from a baseline A1c of 9.3%) Notably, these reductions were sustained for up to 18 months. Veterans are particularly at risk for diabetes, and we were excited to see such promising results for this underserved population.

  • Geisinger’s Fresh Food Farmacy Food Insecurity Intervention drove a 2- percentage point reduction in A1c (baseline >9%), 49% drop in hospital admissions, and 13% drop in ER visits. The Fresh Food Farmacy program was developed to meet the health needs of food insecure patients with diabetes in Pennsylvania, providing 482,219 meals since program inception. This session also highlighted the extreme food insecurity that people with diabetes often face. One in five people with A1cs between 6.5 and 9.0% are food insecure, and one in four are food insecure among those with A1c >9%. This is especially shocking considering that one in eight are food insecure in the US population more broadly. Fresh Food Farmacy was designed with five basic elements: (i) identification of food insecure individuals; (ii) food as medicine (providing only health options; i.e., fruit, vegetables, lean meats, grains, etc.) to supply 10 meals/week for the patient’s entire household; (iii) education and clinical support with meal planning, recipes, and lifestyle change support from clinicians and dieticians; (iv) care beyond health including transportation, housing, and food stamp programs; (v) and community partnerships with local grocery stores and community health assistants. In addition to improvements in A1c and hospital admissions, Fresh Food Farmacy participants also reduced fasting glucose by 27%, cholesterol by 13%, LDL by 9%, and triglycerides by 15% after one year in the program.

  • RTI International’s Mr. Simon Neuwahl presented a simulation modeling analysis indicating that a combination of risk-based (e.g., DPP) and population-based interventions (e.g., soda tax, worksite health promotion, and bike lanes) will be necessary to reduce type 2 diabetes incidence in the US by 17% within the next 10 years. The interventions are estimated to come with a price tag of ~$500 per person – $164 billion overall – and appear to fall short of the CDC’s target of a 21% incidence reduction by 2025. For context, 1.4 million people were diagnosed with type 2 diabetes in the US in 2018, so the cocktail of interventions proposed by Mr. Neuwahl would prevent ~2.4 million cases for the country at a cost of ~$68,000 per prevented case. Mr. Neuwahl emphasized that these are estimates, and more research is needed to investigate the cost and effectiveness of whole population-based diabetes prevention interventions. The main whole population-based intervention that has been studied is the soda tax, which has been shown to be effective. For more on whole-population interventions, see one of our favorite talks from WCPD 2018.

  • The Telemedicine for Reach, Education, and Treatment (TREAT) model, a team-based approach for delivering telemedicine, delivered a 2-percentage point A1c reduction at both six months and 12 months, a significantly greater reduction relative to usual care. TREAT provides an electronic platform for medical nutrition therapy, DSMES, specialty care, and primary care. The TREAT framework is centered around the team: the patient, primary care physician, Diabetes Care and Education Specialist, and endocrinologist. Patients who used shared decision-making tools in tandem with TREAT were able to set, and in many cases achieve, behavioral goals related to diet and reducing A1c and fatty food consumption. Patients also reported feeling more empowered and knowledgeable about their diabetes care, allowing them to have improved goal-setting conversations with providers. Because it uses a virtual platform, TREAT has the potential to be a particularly useful tool in expanding access to rural and underserved communities.

5. Diabetes Technology in the Era of COVID-19: Challenges, Opportunities, and Hope

  • ADA’s 80th Scientific Sessions was held virtually this year due to the COVID-19 pandemic. Beyond transforming the way many participants interact in a conference environment, COVID-19 has fundamentally altered the ways diabetes technology will be accessed and delivered to patients. Potentially the biggest change on the horizon is the implementation of real-time CGM in hospitals, which Dexcom CEO Kevin Sayer sees a market for after the pandemic. As a reminder, the FDA gave the green light to both Abbott and Dexcom for CGM use in the hospitals in April 2020 during the early stages of the pandemic. During a Dexcom-sponsored symposium, Dr. Shivani Agarwal shared a variety of takeaways from inpatient CGM, characterizing the overall experience as positive. Beyond the importance of training, Dr. Agarwal stated that obtaining the buy-in and support from hospital staff was the biggest factor for success. As a result, other hospitals interested in using CGM must create training and outreach initiatives that cater directly to that setting. While the experience at Albert Einstein was positive, additional data on the safety, accuracy, and use of CGMs in inpatient settings, both clinical and legal, are required in the future.

  • Sessions shed light on the combination of CGM and telemedicine, which has also seen significant uptake since the advent of COVID-19. Scripps Whittier’s Dr. Einhorn characterized both tools as vital in bringing about “a new era” in diabetes care, with the ability for HCPs to now follow up with patients in a matter of weeks as opposed to months, set meaningful goals, and remotely monitor data. Dr. Einhorn’s experience is far from being the minority as a number of events (see Insulet webinar)  and literature (see DT&T “Silver Lining” publication) have demonstrated meaningful benefits of diabetes telehealth ranging from improving patient autonomy to increasing time in range. Despite the opportunities, issues related to data privacy, cybersecurity, and reimbursement will be crucial for telehealth to remain widely used post-pandemic.


Automated Insulin Delivery, Pumps, and Pens

MiniMed 780G Pivotal Trial: +1.4 Hour/Day Gain on TIR and 0.5% A1c Improvement on MiniMed 780G vs. Baseline (SAP or MiniMed 670G); 95% Time in Closed Loop Through 3-Month Study

Dr. Bruce Bode (Emory University) presented extremely positive results from the ~3.5-month, single-arm (n=157, ages 14-75) study compared Medtronic’s “Advanced Hybrid Closed Loop” (AHCL), a.k.a., MiniMed 780G system, with sensor-augmented pump therapy or MiniMed 670G. During the 14-day run-in phase, about one-third of patients were using MiniMed 670G with Auto Mode (automatic basal modulation with set point of 120 mg/dl). Compared to run-in, AHCL delivered improvements on every glycemic outcome. Time in range during the three-month study period improved from 69% to ~75% (+1.4 hours/day); most of this improvement came from reductions in hyperglycemia – percent time >180 mg/dl fell from ~28% to ~23% (-1.1 hours/day). Adolescents (ages 14-21; n=39) saw a particularly large improvement in time in range, from 62% to 73% (+2.5 hours/day). For the overall group, AHCL delivered an A1c reduction of 0.5% (baseline: 7.5%) and mean sensor glucose was reduced from 153 mg/dl to 148 mg/dl. Trial participants spent an impressive 95% of time in closed loop, a significant jump from the 87% time in closed loop reported with MiniMed 670G’s pivotal trial. Participants averaged just 1.3 Auto Mode exits per week in the 780G pivotal study.

As we’ve seen in many closed loop studies now, outcomes saw the greatest improvement overnight (midnight – 6 AM). During the overnight period, time in range increased from 71% to 82% using AHCL and mean glucose fell from 150 mg/dl to 140 mg/dl. Most of the improvement came from reductions in hyperglycemia: time >180 mg/dl fell from 26% to 16% in the overnight period and time >250 mg/dl dropped from 5.1% to 2.6%.

Medtronic’s press announcement shares that a study questionnaire found 96% of participants saying the system was “easy to use.” It also notes that system requests for fingersticks fell by 46% compared to 670G. Ease of use was a major focus of MiniMed 780G and participants in the pivotal averaged ~1.3 closed loop exits/week, a huge improvement from the nearly once per day Auto Mode exits seen with 670G.

The study reported no serious adverse events (SAEs), such as DKA, severe hypoglycemia, serious adverse device effects, or unanticipated device effects. There were three non-device related SAEs, including one incident of severe hypoglycemia during run-in, one appendicitis during the study period, and one sepsis during the study.

Like the 670G pivotal, Medtronic elected for a single-arm study, comparing the same group of users before and after AHCL. This contrasts with Tandem’s Control-IQ pivotal study, which randomized users to participants into separate SAP and closed loop groups. Halfway through this trial (45 days), Medtronic also included a cross-over with half of participants changing their target glucose set points from 100 mg/dl to 120 mg/dl, and the other half changing from 120 mg/dl to 100 mg/dl.

There were no updates on regulatory timelines for MiniMed 780G, though Medtronic’s accompanying press announcement confirms that the algorithm will be part of a class III Premarket Approval (PMA) application, i.e., not a class II, interoperable automated glycemic controller (“iController”). Medtronic announced CE-Marking for its AHCL MiniMed 780G system yesterday – that report contains a detailed breakdown of MiniMed 780G’s product features.

Pictures were not allowed, but we’ve summarized the data below.


Overall Group (n=157)

Adolescents (ages 14-21; n=39)

Adults (ages 22-75; n=118)








Time in range














Time >180 mg/dl







Time >250 mg/dl







Time <70 mg/dl







Time <54 mg/dl







Mean CGM

153 mg/dl

148 mg/dl

162 mg/dl

150 mg/dl

151 mg/dl

147 mg/dl

Time in closed loop







Note: Time in closed loop during baseline is from some participants using the MiniMed 670G system during run-in.

  • Time in range improved from 69% to 75% (p<0.001), an improvement of 1.4 hours/day. During the overnight period (midnight – 6 AM), time in range jumped from 71% to 82% (p<0.001), while time in range saw a more modest improvement from 68% to 72% (p<0.001) during the daytime period.

    • At baseline, 54% of participants were meeting the consensus target for >70% time in range. With AHCL, 73% of participants were able to hit the >70% time in range goal. When looking only at AHCL with a set point of 100 mg/dl, an impressive 79% (!) of participants were able to meet this goal. This effect was particularly pronounced in adolescents: at baseline, 18% of adolescents had time in range >70%; during the study period, more than half (59%) of adolescents had time in range >70%.

  • Mean sensor glucose fell from 153 mg/dl to 148 mg/dl with AHCL (p<0.001). When using AHCL and a set point of 100 mg/dl, mean glucose fell even further to 144 mg/dl. The difference was greater in the adolescent subgroup, which saw mean glucose improve from 162 mg/dl at baseline to 150 mg/dl with AHCL and 147 mg/dl with AHCL and a 100 mg/dl set point. During the overnight period, the overall group saw a mean glucose of 135 mg/dl when using AHCL and a set point of 100 mg/dl.

Overall Group (n=157)



Study w/ 100 mg/dl set point

Mean glucose

153 mg/dl

148 mg/dl

144 mg/dl

Daytime mean glucose

155 mg/dl

151 mg/dl

148 mg/dl

Overnight mean glucose

150 mg/dl

140 mg/dl

135 mg/dl

  • Time in hypoglycemia was low at baseline and slightly reduced with AHCL. Overall, time <70 mg/dl decreased from 3.3% to 2.3% (-14 min/day) and time <54 mg/dl decreased from 0.8% to 0.5% (-4 min/day). Of note, the results shared are all well below consensus goals for time in hypoglycemia, even without automation. As seen in the table below, using the lower (default) set point of 100 mg/dl increases time <70 mg/dl, though it’s unclear whether that difference is statistically significant.

  • A1c was significantly improved from 7.5% to 7% (p<0.001). At baseline, about one-third (34%) of participants were meeting the A1c goal of <7%; during the study period, the number of participants meeting this goal was almost doubled, to 61%.

  • MiniMed 780G includes the ability to adjust active insulin time between two and eight hours. With shorter active insulin time settings, time in range trended higher without increasing time in hypoglycemia. The most common active insulin time settings were between 2-3 hours.

Time in range

Active Insulin Time

2 hours

>2-3 hours

>3-4 hours

>4 hours

100 mg/dl set point









120 mg/dl set point










Time <70 mg/dl

Active Insulin Time

2 hours

>2-3 hours

>3-4 hours

>4 hours

100 mg/dl set point









120 mg/dl set point









  • The 16-center study was limited to participants with baseline A1c <10% and >6 months of pump therapy (CGM use was not a requirement). The study enrolled 39 adolescents, ages 14-21, and 118 adults, ages 22-75 years. We believe ~155 participants completed the entire trial, though this was not officially given, nor were reasons for drop-outs. Dr. Bode reported that just two weeks into the study period of the trial, clinic centers and patients began asking Medtronic about keeping their AHCL systems, as they felt like they “can’t live without [the system].”

Baseline Characteristics

Overall (n=157)

Adolescents (n=39)

Adults (n=118)





Female (%)

86 (55%)

23 (59%)

63 (53%)





Weight, kg




BMI, kg/m2




Diabetes duration, years




  • The slides below taken from ATTD 2019 nicely summarize key CGM outcomes from the MiniMed 670G pivotal trials. The NIDDK-sponsored FLAIR trial (see highlight below) provides a more direct, randomized, cross-over comparison between the AHCL and 670G systems.

Real-World Data from Control-IQ: +2.4 Hours/Day TIR, 96% Time in Closed Loop in 1,649 Early Adopters; Improvements for Both Type 1s and Type 2s

Tandem presented two posters featuring very positive real-world data from early adopters of Control-IQ. Control-IQ was cleared in December 2019 and officially launched in January 2020. The data presented in both posters came from Tandem users who had begun using Control-IQ before March 11, 2020. As of April, Tandem shared that “more than 30,000” t:slim X2 users had updated their pumps to the hybrid closed loop algorithm. See below for a summary of both posters and a comparison to the pivotal data.

  • Through the first 30-days of Control-IQ use, time in range was increased by 2.4 hours/day (compared to pre-Control-IQ data) and users spent a remarkable 96% of time in closed loop (95-LB). This data set included participants with at least 30 days of pre- and post-Control-IQ data in t:connect and included a total of 1,659 subjects. The time in range improvement was driven by a 9.5% reduction in time >180 mg/dl (-2.3 hours/day). The reduction in time <70 mg/dl was low both pre- and post-Control-IQ (1.2% before vs. 1.1% after). This result is unsurprising as most users will likely on Tandem’s predictive low glucose suspend algorithm, Basal-IQ, before going to Control-IQ. Mean glucose fell from 161 mg/dl to 148 mg/dl and GMI fell from 7.2% to 6.9% before and after Control-IQ.

  • Control-IQ significantly improved glycemic outcomes for both type 1 and type 2 users (126-LB). This second poster looked at 2,896 participants with type 1 and 144 participants with type 2 diabetes who had at least 14 days of pre- and post-Control-IQ data in t:connect. The data are summarized in the table below and both groups spent 96% of time in closed loop. Time in range was improved by 2.1 hours/day in the type 1 subgroup, compared to a 1.4 hour/day improvement in the type 2 subgroup, though the type 2 group had a higher baseline. After two weeks on Control-IQ, participants spent an incredible 77%-79% of time in range! Notably, total daily dose of insulin was increased in both groups, with the type 2 group seeing a sizeable 12% increase (73 U vs. 82 U).


Type 1s (n=2,896)

Type 2s (n=144)





Time in range





Time <70 mg/dl





Time >180 mg/dl





Total daily dose

46 U

48 U

73 U

82 U

Time in closed loop



  • Improvements from Control-IQ’s early adopter real-world data and the US pivotal trial are comparable, with real-world users spending even more time in closed loop (96% vs. 92%; ~1 hour/day). In the US pivotal, time in range was improved by +2.6 hours/day (59% to 71%); however, given the much higher baseline in the real-world users, the +2.4 hours/day improvement from Control-IQ in the real-world may be even more impressive (68% to 78%). In both trials, the vast majority of the time in range improvement came from reductions in hyperglycemia. Improvements in A1c/GMI and mean glucose were identical (-0.3% and -13 mg/dl, respectively), though from lower baselines in the real-world user group. It’s worth noting that these were early adopters of Control-IQ and likely are not representative of general t:slim X2 users and are certainly not representative of the general type 1 population; in contrast, the Control-IQ pivotal was notable for its broad inclusion criteria (no entry restrictions on A1c, severe hypo or DKA, or device experience). Finally, while the real-world results compare pre- and post-Control-IQ data (i.e., single-arm), the pivotal study randomized users to SAP vs. Control-IQ (i.e., double-arm.



Real-world data

SAP at Six Months (n=56)

Control-IQ at Six Months (n=112)

30 days before Control-IQ (n=1,659)

30 days after Control-IQ (n=1,659)

Time in range










Time >180 mg/dl





Mean CGM

170 mg/dl

156 mg/dl

161 mg/dl

148 mg/dl

Time <70 mg/dl





Time in closed loop





First Ever Head-to-Head Comparison of Closed Loop Systems: MiniMed 780G Delivers Improved Glycemic Outcomes vs. 670G in Teens/Young Adults; 5.7 Auto Mode Exits per Week with 670G vs. 1.7/Week for 780G

On behalf of the cleverly-named FLAIR (Fuzzy Logic Automated Insulin Regulation) study group, Dr. Rich Bergenstal (International Diabetes Center) presented results from the first-ever study of an AID system with a commercially approved AID comparator. The six-month trial showed superiority of Medtronic’s Advanced Hybrid Closed Loop (AHCL) over MiniMed 670G and baseline therapy on nearly every glycemic and usability outcome. The international, seven-center study (four sites in the US, two in Europe, and one in Israel) was sponsored by Jaeb and sponsored by NIDDK. At the beginning of the presentation, Dr. Bergenstal listed out three specific focuses for the FLAIR study: (i) daytime, post-prandial hyperglycemia; (ii) glycemic control in individuals 14-29 years old (adolescents/young adults); and (iii) broad entry criteria. The study randomized 113 participants with type 1 diabetes, ages 14-29, into MiniMed 670G and AHCL (MiniMed 780G) groups with a crossover halfway through (at ~three months). Two of the 113 study participants dropped out due to personal reasons.

  • Dr. Bergenstal called the study’s participants “probably the broadest study population done to date” in AID. One-fifth (20%) of participants were on MDI at baseline and two-fifths (38%) were CGM-naïve. Additionally, a quarter of participants had baseline A1cs between 8.6%-11% and just 19% had A1cs between 7%-7.4%. At time of randomization, mean A1c was 7.9%, though at time of screening, Dr. Bergenstal noted that mean A1c was higher, at “8.2% or so.”

  • AHCL was superior to MiniMed 670G on both primary endpoints: time >180 mg/dl during the daytime (6 AM – midnight) was improved to 34% with AHCL vs. 37% with 670G (-43 min/day; p<0.001) without increasing time <54 mg/dl (p<0.001 for non-inferiority). The 24-hour glucose profiles show the biggest benefit with AHCL during the overnight and early morning periods, though mean glucose through the entire 24-hour period is lower for AHCL. MiniMed 670G also delivered significant improvements over baseline during the overnight period, with less benefit during the daytime hours.

  • A1c at baseline was 7.9%, compared to 7.6% with 670G and 7.4% with 780G. Time in range at baseline was 57%, compared to 63% with 670G and 67% with AHCL. The entirety of the time in range benefit came from reductions in hyperglycemia, though time in hypoglycemia was not significantly increased. At baseline, just 12% of participants were meeting consensus goals of >70% time in range. This number rose to 22% of participants with 670G, increasing further to 32% with AHCL. Glucose variability remained steady from baseline to 670G to AHCL.



MiniMed 670G

AHCL (MiniMed 780G)





Time in Range




Time <54 mg/dl




Time <70 mg/dl




Time >180 mg/dl




Time >250 mg/dl








  • AHCL was superior to MiniMed 670G across age groups and baseline A1cs. In the teenage group (14-20 years), time in range was 59% at baseline, compared to 64% on 670G and 67% with AHCL. The young adult group (21-29 years) had almost identical results, with time in range improving from 57% to 63% to 67% for baseline, 670G, and 780G, respectively. Stratifying by baseline A1c, those with higher A1cs saw a slightly smaller improvement in time in range. For baseline A1c ≤8.5%, time in range rose from 58% to 65% to 68% for baseline, 670G, and 780G, respectively; for baseline ≥8.6%, time in range rose from 46% to 52% to 54%.

  • In the 14 participants who were on MDI+SMBG, time in range rose from 45% at baseline to 65% with AHCL. For these participants, initiating on MiniMed 670G was enough to bring time in range 670G. On these results, Dr. Bergenstal urged clinicians and researchers not to exclude participants who are not experienced with technology. Across all groups by insulin delivery method (MDI vs. CSII) and CGM-use at enrollment, AHCL delivered superior time in range outcomes (some differences may not have been statistically significant).

Time in range by tech use at enrollment


MiniMed 670G

AHCL (MiniMed 780G)

SMBG + MDI (n=14)




SMBG + CSII (n=29)




CGM + MDI (n=9)




CGM + CSII (n=46)




MiniMed 670G (n=15)




  • With AHCL, mean total daily dose of insulin increased to 55 units (0.75 U/kg), compared to 50 units (0.67 U/kg) on 670G. Remember, this was a crossover study, so the participants are the same. This increase in insulin use was driven by AHCL’s automatic correction boluses; for the 670G periods, insulin use was evenly split between bolus and basal insulin (49% vs. 51%), whereas bolus insulin made up 64% of the total daily dose with AHCL. Of that 64%, 36% (i.e., 13 units) came from automatic correction boluses. Dr. Bergenstal speculated that the higher reliance on automatic corrections comes from the study population being more likely to forget meal boluses, demonstrating the 780G’s greater “forgiveness” for missed or miscalculated meal boluses.

  • The number of closed loop exits per week dropped from 5.7 with 670G to 1.7 with AHCL. The 1.7 exits per week translates to approximately one exit every four days and is actually slightly higher than results from the US pivotal (see above) of 1.3 exits/week, though this could probably be attributed to the “more difficult” study population in FLAIR. The reduced number of closed loop exits translated to much higher time in closed loop, at 86% vs. 75%. With AHCL, CGM use was also 86%, meaning when CGM data was available, the system was in closed loop nearly the entire time.

  • Neither MiniMed 670G or AHCL arms reported incidents of DKA, though there was one severe hypo event in the AHCL arm requiring “mild assistance.” The 670G arm recorded two “other” serious adverse events, one relating to suicidal tendencies and one due to a ruptured appendix. The MiniMed 670G arm reported three events of hyperglycemia or ketosis related to the pump, while the 780G arm reported two of these events.

Horizon Pre-Pivotal Delivers +1.7 Hour/Day Time in Range in Adults (n=18) and +3.3 Hours/Day in Children (n=18); 97% (!) Time in Closed Loop and Very High System Usability Scores

In front a Stanford Cardinal-colored background, the eminent Dr. Bruce Buckingham (Stanford) presented very impressive study results from the first outpatient study of Insulet’s Omnipod Horizon hybrid closed loop system. The single-arm study compared 14-days of standard therapy with 14-days of hybrid closed loop therapy for 18 adults (ages 14-70) and 18 children (ages 6-13). In the closed loop phase, participants spent 3 days with a glucose target of 130 mg/dl, 3 days at 140 mg/dl, 3 days at 150 mg/dl, followed by 5 days in which participants were free to choose set points ranging from 110 mg/dl to 150 mg/dl. After the study, participants were invited to continue on closed loop therapy and participate in the pivotal trial for Horizon, which all participants elected to do.

  • In adults, time in range was improved from 66% to 73% (+1.7 hours/day) from the standard therapy to hybrid closed loop with user-specific glucose set point. Time in range was improved on both ends, with marked reductions in both time in hyperglycemia and hypoglycemia. Impressively, time below 70 mg/dl dropped from 2.6% to 0.3% (-33 min/day) and time below 54 mg/dl dropped from 0.6% to 0% (-9 min/day). After a longer period on Horizon of 4-9 weeks, it appears that time <70 mg/dl bounced back slightly to 0.9%, but still that translates to just ~13 minutes/day – wow! Mean glucose and A1c remained about the same on both standard therapy and closed loop, but glucose variability was crushed, with coefficient of variation dropping from 36% to just 29%. In the 24-hour glucose profile below, the blue line and blue-shaded regions represent the median and IQRs for glucose values on closed loop and are, very clearly, as Dr. Rich Bergenstal would say, “flat, narrow, and in-range.”

  • Time in range for pediatric trial participants was even more dramatic, jumping from 51% on standard therapy to 65% (+3.3 hours/day) on the five closed loop days with user-chosen set points. For the 11 participants who chose the lowest set point of 110 mg/dl, time in range was a striking 71%. Similar to the adults, pediatric patients saw very low levels of hypoglycemia, but also massive reductions in hyperglycemia. Time >180 mg/dl fell from 47% to 34% with closed loop (-3 hours/day) and just 27% for closed loop with a set point of 110 mg/dl. With those reductions in hyperglycemia, mean glucose decreased from 185 mg/dl to 167 mg/dl to 155 mg/dl and GMI fell from 7.7% to 7.3% to 7% from standard therapy to closed loop to closed loop with a set point of 110 mg/dl. Importantly, these improvements were maintained for longer term use of Horizon, as well.

  • As with most AID systems, there was a massive benefit in overnight glycemic control. In 6-<14 year old participants, time in range overnight (midnight – 6 AM) rose from 56% to 72% with closed loop. For closed loop with a set point of 110 mg/dl, overnight time in range was 77%. In the table below, the 0% time <70 mg/dl when using closed loop also really stands out. Lastly, overnight mean glucose improved significantly, as did glucose variability.


  • Throughout the presentation, Dr. Buckingham highlighted Horizon’s focus on reducing hypoglycemia, and the results (see above) are stunning. Dr. Buckingham shared that with Horizon taking away the fear of lows, a lot of the kids in the study were able to go on their first sleepovers. Beyond what’s reflected in the glycemic outcomes in the tables above, Dr. Buckingham focused on Horizon’s ability to remove mental and psychological burden from having diabetes or having a child with diabetes.

  • Both children and adult users spent an incredible 97% of time in closed loop and scored very high on system usability. As the Horizon algorithm is built directly onto the Omnipod pump, which can communicate directly with the Dexcom G6 sensor, users don’t need to worry about staying within range of a handset or receiver. Additionally, both Omnipod and G6 are waterproof, allowing users to stay in closed loop when showering or swimming. For parents of children ages 6-<12 years, Horizon scored a 90 (out of 100) on system usability. For teenagers, the system received a 93. During Q&A, Dr. Tim Bailey (Advanced Metabolic Care and Research) noted that Apple’s iPhone scores ~79 on the same scale. Dr. Buckingham even suggested Horizon might have a higher usability when launched, as Insulet plans on launching with smartphone control, whereas study participants had to use a separate, locked-down Samsung phone for controlling their Omnipods. User friendliness and simplicity have been a focus on the design of Horizon from the very beginning and the pre-pivotal data suggest that focus has paid off greatly.


  • There were not serious adverse events reported in the study, though there was one incident of prolonged hyperglycemia when one adult participant’s cannula was dislodged.

  • When launched, Insulet’s hybrid closed loop system will likely be marketed under the name “Omnipod 5.” For a long time, we (and others) have referred to the system as Omnipod Horizon, though Horizon may actually be the term for the algorithm component of the system. Dr. Buckingham’s presentation was titled, “Omnipod 5 Automated Glucose Control System, Powered by Horizon” and during the ADA session’s live-chat feature, we saw Insulet Medical Director Dr. Trang Ly write: “Horizon (now called Omnipod 5) is expected to be avail first half 2021 in US [sic].”

  • Just last week, Insulet announced resuming of the pivotal trial for Omnipod Horizon after a ~three month pause. We’d estimate the trial (n=240) will wrap up within the next 3-4 months, in line with plans to launch in the “first half of 2021.”

Tandem’s Basal-IQ (Predictive Low Glucose Suspend) Significantly Reduces Self-Reported Severe Hypo Paramedic Visits (-45%), ER Visits (-77%), and Hospital Admissions (-75%)

Ms. Molly McElwee-Malloy (Tandem) presented strong real-world data on reductions in severe hypoglycemia related paramedic visits, ER visits, and hospital admissions with Tandem’s predictive low glucose suspend Basal-IQ algorithm. Across all three types of adverse events, the biggest improvements were seen in type 1s who switched from MDI to Basal-IQ. The study surveyed 665 type 1s who had recently started using t:slim X2 with Basal-IQ and received a $20 gift card for their participation. Participants self-reported severe hypo-related adverse events at baseline and after six months on Basal-IQ. Participants had a mean age of 37, 15% were on MDI at baseline, and 91% used CGM at baseline.

  • The percentage of participants reporting hospital admissions related to severe hypoglycemia in the last six months was reduced from 3.7% (24/655 participants) at baseline to 0.9% (6/655 participants) with Basal-IQ. The 95 participants switching from MDI saw a dramatic reduction: at baseline, 10 participants experienced a severe hypo hospitalization in the last six months, compared to just 2 during six months of Basal-IQ.

  • The percentage of participants reporting ER visits related to severe hypoglycemia in the last six months was reduced from 5.9% (39/655 participants) at baseline to 1.4% (9/655 participants) with Basal-IQ. Once again, the 95 participants switching from MDI saw the biggest improvement: at baseline, 14 participants experienced a severe hypo ER visit in the last six months, compared to just 4 during six months of Basal-IQ.

  • Similarly, the percentage of participants reporting paramedic visits related to severe hypoglycemia in the last six months was reduced from 5.3% (35/655 participants) at baseline to 2.9% (19/655 participants) with Basal-IQ. The 95 participants switching from MDI saw the biggest improvement: at baseline, 15 participants experienced a severe hypo paramedic visit in the last six months, compared to just 6 during six months of Basal-IQ. Notably, Ms. McElwee-Malloy also shared that ~5% of all EMS calls nationally are related to hypoglycemic events. Given the very high cost of severe hypo adverse events, we loved seeing these results and would be interested in seeing a longer-term and broader cost-effectiveness analysis of Basal-IQ (and Control-IQ).

Tandem’s Control-IQ in Young Children (2-5 Years; n=12) Increases TIR by 1.8 Hours/Day, Time <70 mg/dl Reduces From 3.7% to 1.5% (-32 min/day)

Stanford’s Dr. Laya Ekhlaspour read out very positive results from a small (n=12) study of Tandem’s Control-IQ in young children (ages 2-5 years). The study involved 2-7 days of run-in, 48-hours of Control-IQ in a supervised hotel setting, followed by 3-days of home use. During run-in, participants used a study pump at home in open-loop; during the hotel phase, the children participated in 30 minutes of activity per day, but there were no restrictions on meals or snacks and boluses were delivered per parents’ routines. Compared to run-in, every glycemic outcome was improved with Control-IQ at home. The study’s primary outcome measured the percentage of subjects meeting less than 6% time <70 mg/dl and less than 40% time >180 mg/dl goals. At baseline, one-third of participants met both goals, compared to two-thirds during the hotel phase (p=0.01) and 75% in the Control-IQ at home phase (p=0.002).

  • Time in range was improved by 1.8 hours/day with Control-IQ at home vs. open-loop at home (62% vs. 68%). Most notably, Control-IQ at home delivered large reductions in hypoglycemia, reducing time <70 mg/dl by 32 min/day (3.7% vs. 1.5%; p=0.004) and time <54 mg/dl by 7 min/day (0.6% vs. 0.1%; p=0.004). Time in hyperglycemia (>180 mg/dl) was decreased by 1 hour/day (34% vs. 30%), though this difference was not statistically significant (p=0.075).

  • As we saw with Control-IQ’s adult and older children pivotals (read outs at ADA 2019 and ATTD 2020, respectively), Control-IQ delivered the biggest improvements overnight (11 PM – 7 AM). Time in range overnight was 59% at baseline, compared to 79% during the hotel phase and 76% during the at home phase. Time >180 mg/dl was reduced from 39% at baseline to 24% during at home and time <70mg/dl was reduced from 2.1% to 0%.

  • At the time of ADA, Tandem’s Control-IQ was approved for ages 14+, but comes with a specific black box warning against use in patients under 6 years old. Mean total daily dose in this study was 15U/day and a special modification from Tandem was required to carry out the study for patients this young. Pediatric indication (6+ years) has already been submitted to the FDA, but larger and longer studies will certainly be needed in this very young population.

MiniMed 780G New Zealand Data: Time in Range +3.3 Hours/Day with Advanced Hybrid Closed Loop and 100 mg/dl Set Point vs. Run-in; 96% Time in Closed Loop; CE-Marked Announced Yesterday

In the third (and final) MiniMed 780G presentation from the day, Dr. Martin de Bock (University of Otago) presented data from a randomized crossover study of AHCL vs. the predictive low-glucose suspend MiniMed 640G system. Results of this study (n=60) were used for the CE-submission for MiniMed 780G; just yesterday, Medtronic announced CE-Marking for the system. The study enrolled 60 participants (20 participants 7-15 years old). The sensor-augmented pump (SAP) run-in phase lasted 2-4 weeks before participants were randomized to MiniMed 640G or AHCL for four weeks, followed by two weeks of washout with SAP, and finally, four weeks of crossover. At enrollment, 44% of participants were on SMBG, while >6 months on pump therapy was required for inclusion. Time in range increased from 59% during run-in to 73% for AHCL with a 100 mg/dl set point (+3.3 hours/day); accordingly, time >180 mg/dl fell from 38% to 25% (-3.1 hours/day). Results with AHCL and a set point of 120 mg/dl were slightly less impressive, but still improved over run-in. During Q&A, moderator Dr. Tim Bailey commented that the results (along with those of the other two MiniMed 780G talks) really showed that MiniMed 780G is a device that is “designed to be used” with a set point of 100 mg/dl. Dr. de Bock also added that the improvements in the study were seen even with “plenty of petrol in the tank” (i.e., relatively conservative settings): the study used a longer active insulin time of three hours, kept insulin:carb ratios from baseline, and the higher set-point in pediatric participants.



MiniMed 640G

AHCL with 120 mg/dl set point

AHCL with 100 mg/dl set point

Mean glucose

168 mg/dl

171 mg/dl

162 mg/dl

149 mg/dl

Time in range





Daytime time in range





Nighttime time in range





Time > 180 mg/dl





Time <70 mg/dl





Time <54 mg/dl





  • Time in closed loop was remarkably high, at 96%. Patients experienced a mean of 1.2 Auto Mode exits/week and the number of alarms and alerts was reduced from ~13/day with MiniMed 640G to 8/day with AHCL. Additionally, Diabetes Treatment Satisfaction Questionnaires showed improvements in treatment satisfaction for AHCL vs. MiniMed 640G for adolescent and adult users.

  • The study saw one episode of mild DKA, due to a combination of a site occlusion and viral illness. There were no incidents of severe hypoglycemia. There were five “possible or probable device related adverse events”; four were skin reactions related to infusion sites and one was a skin infection.

  • Lastly, but certainly not least, Dr. de Bock also shared a few powerful quotes and impressions from the trial:

    • “…created a situation for our family that was as close to not having diabetes as we have been in the last decade. It wasn’t what it made us do that made the difference, it was what we no longer had to do.”

    • “We didn’t have to worry, we didn’t have to be fearful at night or have that thought when we opened her bedroom door in the morning that she might not be conscious.”

    • “She could think about horses and friends and Keeping up with the Kardashians like any other tweeny with a junk TV habit.”

    • “I forgot I had diabetes today.”

Five Medtronic Posters on Extended-Wear Infusion Set: 7-Day Survival Rate of 81%; Can Reduce Annual Insulin Waste by 5-10 Vials Compared to 2/3-Day Sets

We rounded up five posters on Medtronic’s Extended Wear Infusion Set (EWIS), which was CE-Marked as of ATTD 2020 and is currently in pivotal trial in the US. At ATTD, Dr. Ohad Cohen (Medtronic) shared that EWIS included a new “H-Cap Connector” to improve site performance, improved tubing to improve insulin preservative retention & stability, and an improved adhesive patch for better adherence. Three of the posters below do an excellent job outlining some of Medtronic’s work and considerations around those three improvements. A final poster also calculates ~$1,500-$3,000 in annual cost savings from reduced insulin waste with a 7-day infusion set.

  • A small (n=21 participants) study showed 7-day survival rates of ~81% for Medtronic’s EWIS (994-P). The 21 participants wore four infusion sets (82 total insertions) until set failure or seven days. There were no safety signals, including DKA, severe hypo- or hyperglycemia, device-related severe adverse events, and death. Survival of the extended-wear set at 7-days (80.5%) was actually higher than the published survival rate of Medtronic’s 3-day infusion set at 3 days (77%). When this data was first presented at ATTD 2020, Dr. Cohen also noted that survival rate increased to 85% if insertion failures were removed. 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.

  • A modified version of Medtronic’s 3-day infusion set showed 7-day survival rates of ~73%-75% (997-P). The study, presented by Dr. Bruce Buckingham (Stanford), modified the cap of the MiniMed Quick-set connector to create a chamber which contained either foam or foam with 80 units of heparin (anticoagulant). Twenty participants were randomized into foam or foam + heparin groups to start. The participants wore their modified infusion sets for a week (or until failure) before crossing over. The study lasted four weeks (four insertions) with crossover after each week. The rate of 7-day survival was not significantly affected by the addition of heparin, as both modified versions had 7-day survival rates of ~73%-75%. The connector foam, without heparin, is part of Medtronic’s EWIS, in addition to other improvements.

  • Medtronic tested seven different adhesives across three studies – the best performing variant achieved an 8-day survival rate of 100% (986-P). The studies enrolled 75 adults testing 2 adhesive variants each, totaling 150 adhesive placements. “Non-functional” Medtronic pumps were used to simulate conditions. Eight-day survival rates for the first four variants (adhesives used with current infusion sets) ranged from 63% to 89%; eight-day survival rates for three new adhesive variants ranged from 75% to 100%. Across variants, there were no significant differences around overall appearance, skin irritation, or device awareness; however, the newer adhesives trended slightly more difficult to remove and clean off. The highest-performing adhesive (100% 8-day survival rate) was chosen in Medtronic’s EWIS.

  • In-vitro and porcine tests showed infusion set wear-time and inflammatory response were significantly impacted by loss of preservatives in the insulin (1012-P). One test demonstrated pumping insulin through an infusion set lowered the preservative content, though this did not significantly impact insulin chemical stability. However, lower preservative content insulin created more aggregates when the insulin was shaken. With increased aggregates in the insulin solution, the inflammatory response was greatly increased in an in-vitro cell culture and greatly decreased survival time of an infusion set in a porcine model.

  • Assuming insulin costs of $300/vial ($3/U), Medtronic estimated annual savings of ~$1,500 to $3,000 per user with a 7-day infusion set vs. 3-day (1167-P). The estimate used a total daily dose of insulin of 35U/day, translating to 5-10 vials per year of reduced insulin waste. Savings would be lower for those using >35U/day as the pump reservoir wouldn’t hold enough volume for seven days of insulin. Medtronic is also taking a look at the environmental benefits from increasing reservoir volume and extending infusion set wear time.

Beta Bionics Gen 4 iLet Insulin-Only Pivotal Enrollment “Nearly Completed”; Home-Use Study Shows +1.9 Hours/Day TIR, Time <54 mg/dl From 0.6% to 0.2% With Bi-Hormonal vs. Insulin-Only

Dr. Jordan Sherwood (Massachusetts General Hospital) provided a detailed breakdown of positive results from Beta Bionics’ first home-use trial using liquid-stable dasiglucagon. Results from the 14-day crossover study were first announced via press release about one year ago. Towards the end of his presentation, Dr. Sherwood also shared that the insulin-only pivotal trial for Beta Bionics’ Gen 4 iLet is “currently underway,” with enrollment “nearly completed.” Assuming the trial will commence soon, it will come at a ~three-month delay from previous expectations to start in 1Q20; the goal is for insulin-only iLet to launch in ~early/mid-2021.

  • The small, home-use study randomized 10 adults with type 1 diabetes with pump and CGM experience to insulin-only and bi-hormonal iLet configurations. After one week, participants crossed over to the other configuration. The bi-hormonal configuration drove 1.9 more hours/day in-range (71% vs. 79%; p=0.002), reduced time <54 mg/dl by 6 min/day (0.6% vs. 0.2%; p=0.15), and reduced mean glucose from 149 mg/dl to 139 mg/dl, compared to the insulin-only configuration. Though not mentioned today, time <70 mg/dl was 2.4% during the bi-hormonal period and 3.6% during the insulin-only period – a difference of ~17 minutes. Notably, participants initiated therapy by entering only their body weight into the device (and no other parameters).

  • New from today, Dr. Sherwood also provided a look at subject-level differences in the insulin-only vs. bi-hormonal configurations. As shown below, nine of the ten participants saw lower mean glucose with the bi-hormonal iLet; this difference was significant (p<0.05) for eight of the ten participants. Additionally, Dr. Sherwood noted that during the bi-hormonal phase, 9/10 of participants had mean glucose below 154 mg/dl, which would correspond to an A1c ~7%; half of participants had mean glucose <154 mg/dl during the insulin-only phase. Time <54 mg/dl was low in both arms, though 8/10 participants saw improvements with the bi-hormonal configuration. With insulin-only iLet, 6/10 participants reached the consensus target for <1% time <54 mg/dl, compared to 9/10 participants with bi-hormonal iLet.

Bi-hormonal Bionic Pancreas Optimal for Mean Glucose (136 mg/dl) and TIR (81%) at 100 mg/dl Set Point, with no Increase in Hypo (Compared with 115 and 130 mg/dl)

MGH’s Dr. Marwa Tuffaha presented Bionic Pancreas (Beta Bionics) data from 2015 showing that lower glucose targets reduce mean glucose and greater time in range without additional hypoglycemia. Since this study was conducted a few years ago, the Bionic Pancreas AID system consisted of an iPhone 4s running a control algorithm, a Dexcom G4, and two Tandem t:slim pumps (one with insulin and one with glucagon). While today’s gen 4 iLet is fully-integrated, the algorithm is the same, requiring only body mass to initialize and qualitative meal announcements. 20 participants with type 1 diabetes and wide demographic ranges (e.g., A1c 6.1%-9.3%, ages 21-78 years, BMI 20-42 kg/m2, and diabetes duration 5-54 years) were enrolled in a random-order, real-world crossover study evaluating the Bionic Pancreas at three set points (130 mg/dl, 115 mg/dl, and 100 mg/dl), plus usual care. As seen in the figure below, lower targets were associated with significantly lower mean sensor glucose (lowest: 136 mg/dl @ 100 mg/dl target) and higher mean time in 70-180 mg/dl (highest: 81% @ 100 mg/dl target). These stellar outcomes did not come at the cost of more time <60 mg/dl, which is explained by a significantly higher daily dose of glucagon—8.3 ug/kg were administered daily with the 100 mg/dl set point. Interestingly, there was no difference in total daily dose of insulin across set points, though we’d bet there was significantly more “prandial” insulin delivered with the lower set points, quickly bringing users back into range, as opposed to the highest set point where meals were likely handled more conservatively leaving the basal rate to chase highs for the rest of the day. User preference was very high across all three set points, though with significantly higher satisfaction for the 100 mg/dl set point. Note that the participants were not blinded to the system setting, so one might expect higher satisfaction ratings with the lowest set point. None of the patients in the study had DKA or severe hypoglycemia, there was no difference in hypoglycemia symptoms, interventions, or carbs administered per day, and glucagon-induced nausea was not an issue. The integrated Gen 4 iLet with Zealand Pharma’s dasiglucagon is expected to enter pivotal trials in the back half of this year (see FFL 2019) — enrollment is “nearly complete” (see above).

Medtronic’s ADA Analyst Day: $337 Million R&D Investment from Blackstone; No Publicly Shared Pipeline Timings Under New Leadership

On Friday evening, Medtronic hosted its annual one-hour Diabetes investor briefing, the first under new Diabetes Group head Sean Salmon – download the slides here and watch the webcast here. Obviously, a lot of airtime was dedicated to MiniMed 780G, which was really the star of the day on Friday (see read-outs from the US pivotal study, FLAIR trial comparing 670G with 780G, and a New Zealand study). Diabetes CMO Dr. Robert Vigersky gave most of the company’s remarks around MiniMed 780G, focusing on the system’s ability to increase time in range without increasing hypoglycemia, the system’s versatility across various populations (including teens and young adults), and ease of use (44%-46% fewer fingersticks vs. 670G, 95%-96% time in closed loop). Both Mr. Salmon and Dr. Vigersky also highlighted MiniMed 780G’s automatic correction bolus feature (which makes the system more forgiving of mistimed, miscalculated, or altogether missed meal boluses) and the ability to use a glucose set point of 100 mg/dl. Dr. Vigersky’s slides are on pages 15-22 and do a good job of highlighting key points from yesterday’s readouts – we’ve put a few down below, but you can get the full slide deck here.

  • At ADA, Medtronic Diabetes announced a $337 million investment from Blackstone Life Sciences. That $337 million figure certainly jumps off the page and the funding will be used to “pull forward specific programs in [Medtronic’s] pump and CGM pipeline … beyond PCL [Personalized Closed Loop] and Synergy.” The funding will be used for “four identified Diabetes R&D programs”; when we had a chance to talk with Medtronic’s team this morning, we didn’t get many more details. If successfully commercialized, Medtronic expects to pay royalties in the “low- to mid-single digit range.”

  • Noticeably absent from this year’s presentation were anticipated regulatory submission and launch timelines for Medtronic’s pipeline products. See last year’s slides for comparison. On our call with Medtronic this morning, Sean Salmon noted that the company doesn’t want to “overpromise” to its patient and provider users. This seems like a prudent approach, as Medtronic, and many other diabetes device companies, have often failed to deliver on publicly announced products and timelines. “Reinvigorating” the CGM pipeline remains a priority for Medtronic, first with its “Zeus” sensor, which will bring calibrations to day 1-only in the US and zero calibrations outside the US. Synergy will bring a new, slimmer and fully-disposable design, along with an easier insertion process. An extended-wear infusion set remains in a US pivotal trial; CE-Marking was announced at ATTD 2020.

First Lilly AID Data Shows System is Safe in Small, Inpatient Feasibility Study—Negligible Hypoglycemia after Pizza/Pancakes, With or Without Premeal Bolus

Lilly’s Ms. Amy Bartee presented what we believe to be the first data Lilly has shared on its AID system (read about the system, which uses Dexcom CGM, a cool DEKA-developed pump, and the MPC algorithm from Class AP). This was a small (n=10) inpatient feasibility study that demonstrated safety in response to high carbohydrate meals with and without pre-meal boluses. Specifically, participants (mean age 52 years; mean A1c 7.1%; all pump users) were given four meal challenges over a 48-hour inpatient period: (1) bolus with pizza (high-fat meal); (2) missed bolus with pizza; (3) bolus with pancakes (fast-acting carbohydrate meal); and (4) missed bolus with pancakes. As expected with the pizza challenge + bolus, there was a slow rise in glucose and extended postprandial period—the algorithm steadily increased basal insulin delivery to compensate, yielding <1% below 70 mg/dl and 66% time in range. In the pizza challenge + missed bolus, the algorithm responded with an immediate and incremental increases in basal delivery for the next four hours, delivering <1% below 70 mg/dl and 37% time in range. No correction boluses were given in either pizza challenge. In the pancake challenge + bolus, there was no hypoglycemia and 56% time in range. As expected, the pancake challenge + missed bolus was the most challenging for the algorithm, which rapidly ramped up basal insulin delivery but 60% of the next four hours were spent >250 mg/dl (no time <70 mg/dl). One patient required a correction bolus in this condition. Ultimately, the system was safe in every case, though the study emphasized the importance of bolusing for meals, particularly in the absence of automated corrections. Lilly has never shared specific timing for development of its AID system, though it has broadly said that Connected Care products will launch in stages over 2019-2021; we wouldn’t be surprised to see this timeline pushed back.

Percent time in ranges in four hours postprandial


Pizza with bolus

Pizza without bolus

Pancakes with bolus

Pancakes without bolus

% Time in 70-180 mg/dl





% time <70 mg/dl





% time >180 mg/dl





% time >250 mg/dl





Dr. de Bock Very Enthusiastic About 780G’s Impact on TIR in “Challenging” Adolescent Group

University of Otago’s Dr. Martin de Bock rehashed very positive MiniMed 780G CE registration trial (see Day #1 coverage), emphasizing that the greatest day-time time in range benefit was seen in the most challenging population: adolescents (ages 14-21). In fact, the whole population was skewed young, with a mean age of 23.5 years and 56% under 21. Said Dr. de Bock, “This is really exciting for us, especially as pediatric endocrinologists…Maybe we shouldn’t be so concerned about people who aren’t so adherent to their therapy – they are the most likely to see advanced time in range on [780G].” We would caution that the mean A1c in the study was 7.6% and all participants were already on a pump, so this population, though still adolescents, may not be the “most challenging of the challenging.” For reference, mean A1c in people ages 15-18 in the T1D Exchange Registry is 9.3%. Still, the signal is encouraging, and may due to the system’s improved user experience. There was only ~1 auto mode exit per week in this study, compared to ~1 per day as seen with the 670G. Similarly, the system significantly reduces alarms to 8 per day (vs. 13 per day with 640G). P-979 further elaborates ono improved satisfaction and sleep quality with 780G vs. 640G. Echoing his Friday comments, Dr. de Bock concluded: “If we had more time, we hypothesize we could’ve optimized settings…and seen overall time in range that was higher than reported here.”

Barbara Davis Center’s PANTHER Project Provides Targeted Education to Help Patients Onboard with MiniMed 670G and Control-IQ Hybrid Closed Loop Systems

The audience-favorite Dr. Laurel Messer (Barbara Davis Center) presented on the PANTHER Project: Practical Advanced THERapies for Diabetes, a program addressing concerns and barriers with automated insulin delivery (AID) adoption. Despite huge advances in AID in the last few years, Dr. Messer highlighted a variety of barriers that prevent adoption, most notably, issues with the infusion set. According to Dr. Messer, in nearly all AID trials, the number one reason patients experience hyperglycemia or DKA is because of failures with infusion sets. Beyond infusion set issues, other barriers vary by age, further complicating the problem. At a similar talk at ISPAD 2019, Dr. Messer shared data identifying the top barriers to diabetes tech use in adolescents: hassle of wearing devices all of the time (38%), dislike having devices on the body (33%), dislike how devices look on the body (29%), nervousness that the device won’t work (25%), and not wanting to spend more time managing diabetes (20%). During today’s presentation, Dr. Messer walked through some of the experiences and successes seen onboarding patients at BDC in the PANTHER project with MiniMed 670G and Control-IQ.

  • MiniMed 670G: Patients (n=72) interested in the MiniMed 670G were first trained on using Manual Mode. Trainings were in-person for 2-3 hours and happened in groups of families to promote peer support. Five to seven days after this training, patients in smaller groups of families were given ~1-2 hours of Auto Mode training to reinforce conventional insulin pump and CGM use on the new system. Three follow-up phone calls in the first four weeks after training were done to assess system use, make insulin adjustments, and provide targeted re-education. Overall, patient engagement in these classes was high. Following perfect retention in the introduction class and the video conference, 92%, 81%, and 53% of participants responded to follow-up calls one, two, and three. Dr. Messer shared that a variety of changes were made during these sessions, with 75% of participants increasing their insulin: carbohydrate ratio and 44% altering active insulin time. Topics ranged from pre-meal blousing (65%), insulin correction doses (48%), and addressing system alerts such as alarms (45%).

  • Control-IQ: Participants (n=107; A1c: 7.5%) underwent a different process for onboarding as individuals in this program had already downloaded Control-IQ and were ready to start using it. As a result, the BDC team decided to work with the patient to pick an ideal day for starting, scheduling a follow-up call with a RN/DCES after two weeks of starting, and download their pump data. The PANTHER team created four metrics and benchmarks to assess device success in a week-long period: (i) time using Control-IQ (>5 days per week); (ii) time using the CGM (>5 day per weeks); (iii) time in range (>60%); and (iv) time spent <70 mg/dl (<5%). Median follow-up time for the call was 18 days, slightly more than anticipated, but ~64% of participants (n=68) met all four benchmarks. Of the 18 of 39 patients that completed an addition second follow-up call, nine ended up meeting all four benchmarks, with mean time in range between calls one and two increasing a remarkable 12.5% (baseline not provided). Dr. Messer noted that these results demonstrate active engagement coupled with education can promote technology uptake and improved glycemic outcomes.

Ultra-Rapid Insulins Safe and Effective in 670G & 780G, but Apparently Not More So Than Rapid-Acting Insulins; Possible Exception: 780G with FiAsp May Blunt PPG Relative to Novolog

We enjoyed back-to-back orals investigating the benefit of ultra-fast insulins in Medtronic’s (advanced) hybrid closed loop systems. In short, the benefit appears to be marginal, at best. However, it is does seem safe and feasible.

  • Atlanta Diabetes Associate’s Dr. Bruce Bode first showed in a double-blind crossover study (n=42) that there was no difference in glycemic outcomes with MiniMed 670G when participants used Lilly’s Ultra-Rapid Lispro (URLi) or Humalog. Time in range was 77% with Humalog and 78% with URli, and time <70 mg/dl was ~2% in both cases. There was no difference in time in range between the two conditions during day or night, nor post-meal. There were five cases of infusion site pain/reaction in the URLi condition vs. one in the Humalog condition, but overall, booth conditions were very safe and there were no discontinuations. We and a number of chat participants felt that perhaps the participants were already too well-managed on the 670G at baseline to see an impact of even faster-acting insulin. Mean baseline A1c was 7.1%, mean time in range was 78%, and participants were already spending ~90% in-range. Dr. Bode additionally proposed that a hybrid closed loop system that only modulates basal insulin “may not be optimized for the potential advantages of an insulin with differential effects seen primarily during bolus delivery.” Benefits may be realized, he reasoned, if it were used in a system that gives automated boluses, such as MiniMed’s AHCL (780G) system. The next oral looked at exactly that.

  • University of Melbourne’s Dr. Melissa Lee presented a similarly-designed crossover study (n=12) comparing the use of Novo Nordisk’s FiAsp and Novolog in the MiniMed 780G. Despite Dr. Bode’s hypothesis, this study found no overall benefit of FiAsp, with the potential exception of blunted postprandial spikes after breakfast. Though smaller, the study population was similar to above, with experienced pumpers with mean baseline A1c of 7.1%. In both conditions, time in range was ~80%, hypoglycemia was <0.5%, and mean glucose was <140 mg/dl, with no statistically significant differences but trends toward FiAsp benefit. There was no difference in CGM metrics post meal when all meals were pooled, but when separated, incremental AUC over 2 hours was significantly lower with FiAsp only for breakfast. Breakfast-time may be unique for a whole host of reasons, said Dr. Lee: less insulin resistance in the morning, snacks/stress/exercise later in the day, or just simpler carbs at breakfast. While 780G was safe with both insulins, there were four individuals who experienced at least one mild infusion-site reaction using FiAsp vs. only one using Novolog. 

AID, Pumps, and Pens Posters



Details + Important Takeaways 

Control-IQ Technology in the Real World: The First 30 Days

Lars Mueller, Alexandra Constantin, Harsimran Singh, Stephanie Habif

  • Retrospective analysis of first users (n=1,659) with >30 days pre- and post-Control-IQ software update; mean age 43, 52% female, mean diabetes duration 21 years, 90% type 1, 4% type 2, 6% other/unknown; users all had ≥75% CGM use

  • Control-IQ software update led to 10% increase (p<0.001) in median sensor time in range (70-180 mg/dl) from 68% to 78%; 10% decrease in median sensor time >180mg/dl, 0.1% decrease in median sensor time <70mg/dl

  • Mean time in closed loop system was 96%

Diabeloop Closed-Loop System Allows Patients with Type 1 Diabetes (T1D) to Significantly Improve Their Glycemic Control in Real-Life Situation, without Serious Adverse Events: A 6-Month Follow-Up

Coralie Amadou, Sylvia Franc, Pierre Y. Benhamou, Sandrine Lablanche, Erik Huneker, Guillaume Charpentier, Alfred Penfornis

  • 24 patients participated in this 6-month Diabeloop Closed-Loop study; patients used Dexcom G6 sensors, Kaleido pumps, YourLoop cloud data platform

  • All patients already used insulin pumps; mean age 43, 76% female; 7-day run-in period

  • After 6 months A1c decreased from 7.9% to 7.1% (p<0.001); time in range increased from 53% to 67% (p<0.0001), and time <70 mg/dl decreased from 2.4% to 1.3% (p=0.03)

Safety and Glycemic Outcomes of the MiniMed Advanced Hybrid Closed-Loop (AHCL) System in Subjects with T1D

Anders L. Carlson, Bruce W. Bode, Ronald L. Brazg, Mark P. Christiansen, Satish K. Garg, Kevin Kaiserman, Mark Kipnes, David R. Liljenquist, Athena Philis-Tsimikas, Rodica Pop-Busui, Jennifer Sherr, Dorothy I. Shulman, Kamalpreet K. Singh, Robert H. Slover, James Thrasher, Xiaoxiao Chen, Benyamin Grosman, Scott W. Lee, Louis J. Lintereur, Margaret Liu, Neha Parikh, Andrew S. Rhinehart, Fen Peng, Anirban Roy, John Shin, Di Wu, Robert Vigersky


  • Single-arm, 16-center, in-home trial, with 157 individuals with type 1; type 1 diagnosis ≥ 2 years, A1c <10%

  • AHCL feature enabled with set point of either 100 or 120mg/dl autocorrecting every 5 minutes; patients used both set points for ~45 days

  • AHCL therapy improved mean A1c from 7.5% to 7.0%; 61% achieved A1c <7% during study compared to 34% at baseline; time below <70 mg/dl decreased from 3.3% at baseline to 2.3% during the study

  • Most significant improvements (79% time in range) when set point was 100mg/dl with active insulin time of 2 hours; resulted in no increase in hypoglycemia

  • Total number of AHCL exits during study was 1.3/week (~1 exit every 5 days)

Individual Glycemic Response to Ultra-Rapid vs. Rapid Insulin Delivered Automatically with the Ilet Bionic Pancreas System

Jordan Sherwood, Hui Zheng, Courtney A. Balliro, Laya Ekhlaspour, Bruce A. Buckingham, Firas El-Khatib, Edward Damiano, Steven J. Russell

  • Patients participated in a random-order crossover trial using the iLet bionic pancreas system with either typical insulin (aspart n=7, lispro n=27) or the ultra-fast-acting aspart 

  • Insulin algorithm was the same in both arms; insulin absorption speed set with a maximum of 65 minutes

  • Overall CGM glucose values were similar across both arms; similar numbers of patients experienced significant reductions or increases in CGM glucose values while using ultra-rapid insulin

  • Other studies are currently investigating the effect of adjusting the insulin delivery algorithm for improved results among patients

Closed-Loop Control Reduces Hypoglycemia without Increased Hyperglycemia in Subjects with Increased Prestudy Hypoglycemia: Results from the iDCL DCLP3 Randomized Trial

Carol J. Levy, Grenye O'Malley, Sue A. Brown, Gregory P. Forlenza, Yogish C. Kudva, Roy Beck, Craig Kollman, John Lum, Dan Raghinaru, Boris Kovatchev

  • Post-hoc analysis of data from iDCL DCLP3 study; patients were randomized 2:1 to Closed Loop Control (Tandem Control-IQ pump & Dexcom G6) or SAP for 6 months

  • Participants with ≥4% time <70 mg/dl (n=55) included in sub-analysis

  • Closed Loop Control with Control-IQ improved time in range (9.9% improvement over baseline), and reduced time <70 mg/dl (4.6% decrease from baseline)

Eighteen-Month Use of Closed-Loop Control (CLC): A Randomized, Controlled Trial

Sue A. Brown, Dan Raghinaru, Bruce A. Buckingham, Yogish C. Kudva, Lori M. Laffel, Carol J. Levy, Jordan E. Pinsker, R. Paul Wadwa, John Lum, Craig Kollman, Roy Beck, Boris Kovatchev

  • 18-month RCT to assess safety and efficacy of Closed-Loop Control using Tandem Control-IQ compared to SAP and PLGS; study used Tandem t:slim X2 insulin pump with either Basal-IQ or Control-IQ and the Dexcom G6 CGM

  • 164 participants randomized to 3 groups: (i) Closed-Loop Control for 6 months, then extended Closed-Loop Control to study end; (ii) Closed-Loop Control for 6 months, PLGS (Basal-IQ) for 3 months; Closed-Loop Control to study end; (iii) SAP for 6 months, Closed-Loop control to study end

  • Group 1 participants experienced and maintained significant improvements in time in range over 18 months with mean time in range of 69.9%; time in range deteriorated for Group 2 when placed on PLGS for 3 months, but restored when placed back on Closed-Loop Control; Group 3 experienced persistent improvements in time in range and A1c after transitions from SAP to Closed-Loop Control

  • Closed-Loop Control use 90% or higher among groups showing no trend toward decreased use over time

Cross-Study Comparisons Done Right: An Illustration Using Two Pivotal Trials of Closed-Loop Systems

Marc D. Breton, Roy Beck, Richard M. Bergenstal, Boris Kovatchev

  • Comparing results across studies poses challenges due to sampling differences; better to compare increments from baseline to active study time in range

  • Methods to correct baseline discrepancies in comparison: eliminate participants on tail-end of baseline distribution; resample/replace participants; use functions to match shift and match distributions

  • Comparing across studies using time in range metrics is challenging due to the variable nature of time in range outcomes

Increasing Insulin Pump Use across Five National Diabetes Centers: Results from the T1D Exchange Quality Improvement Collaborative (T1DX-QI)

Osagie Ebekozien, Nicole Rioles, Justin A. Indyk, Sarah Lyons, Shideh Majidi, Taylor Proffitt, Joyce M. Lee



  • Five T1D Exchange centers across the country implemented Quality Improvement initiatives including Plan-Do-Study-Act cycles to increase insulin pump use among patients 12-26 years old

  • Plan-Do-Study-Act interventions included (i) mobile technology classes (ii) redesigning clinic workflow to increase patient education on insulin dosing (iii) coaching patients for meal bolusing and correction (iv) removing barriers to meal-time bolusing

  • Data were shared across clinics monthly to develop best-practices

  • Over 20 months there was a 10% increase in pump use among target cohort; 3 of 5 centers had substantial improvements ranging 6-17%

Optimized Pump Utilization and Increased Pump Initiation with Zero Additional Resource Utilization

Lori A. Miysshiro, Deborah Hinnen

  • Used Certified Pump Trainers to teach patients how to best initiate and utilize insulin pumps (Medtronic n=4; Tandem n=1; Insulet n=21)

  • Certified Pump Trainers integrated into routine clinical care for patients; partnerships with Medtronic, Tandem, and Insulet doubled pump uptake among patients in under 1 year

  • Baseline A1c range from 10.1%-14.5% reduced to 5.2%-7.2%; 82% reduction in diabetes related hospitalizations

Predictive Low Glucose Suspend (PLGS) Necessitates Less Carbohydrate (CHO) Supplementation to Rescue Hypoglycemia: Need to Revisit Current Hypoglycemia Treatment Guidelines

Jordan E. Pinsker, Amy Bartee, Michelle Katz, Amy Lalonde, Richard Jones, Eyal Dassau, Howard Wolpert

  • 10 patients with type 1 (mean age 39, baseline A1c 7.2%) participated in a 20-hour inpatient study with a predictive low glucose suspend algorithm imbedded into Lilly’s investigational insulin pump paired with Dexcom G6 sensor

  • Basal insulin rates were titrated upwards to induce PLGS and carbohydrate rescues were administered if necessary

  • Only 7 of 59 PLGS suspensions required administration of carbohydrate rescue; all carbohydrate rescues took place postprandially; the majority of carbohydrate rescues required less than the standard recommended amount of 15g

  • PLGS has the potential to reduce carbohydrate intake requirements for hypoglycemia rescue

Missed Bolus Doses (MBDs) Are Associated with Reduced Time in Range (TIR): The Influence of Hypoglycemic Fear

Stephanie S. Edwards, Xuanyao He, Jennal Johnson, Eric Meadows, Wenjie Wang, Howard Wolpert, William Polonsky

  • 12-week single-arm study sponsored by Lilly in patients with type 1 (n=40) and type 2 (n=28) on insulin; patients used connected insulin pen and CGM

  • Mean age 48, 44% female, mean A1c 9.6%

  • Missed Bolus Doses were negatively correlated with time in range (p=0.02) and positively correlated with time above range (p=0.03)

  • Missed Bolus Doses among participants with type 2 were significantly correlated to hypoglycemia fear subscales of worry (p=0.004), avoidance (p=0.004), and maintaining hyperglycemia behaviors (p=0.040)

Carbohydrate Needs for Prolonged Fasted Exercise with and without Basal Rate Reductions in CSII

Sarah M. Mcgaugh, Rubin Pooni, Dessi Zaharieva, Ninoschka C. D'souza, Todd Vienneau, Trang T. Ly, Michael Riddell





  • Randomized crossover trial among type 1s on CSII (n=15) completing 3 120-minute treadmill exercise visits; mean age 36, mean A1c 6.9%, 60% female

  • 3 interventions were: (i) no basal rate reduction and dextrose intake of 0.3g/kg/hr when BG was <180 mg/dl (oral carbohydrates only); (ii) 50% basal rate reduction set 90 minutes before exercise and dextrose as needed to treat hypoglycemia; (iii) 50% basal rate reduction set at exercise start time plus dextrose intake of 0.3g/kg/hr when BG was <180 mg/dl

  • 50% basal rate reduction 90 minutes before exercise was the most effective strategy for maintaining glucose control during prolonged fasted exercise

Impact of Carbohydrate Counting on Glycaemic Control in People with Type 1 and Type 2 Diabetes on Intensified Insulin Therapy

Norbert Hermanns, Dominic Ehrmann, Bernhard Kulzer

  • Cross-sectional study (n=154) in patients with type 1 and type 2 on intensified insulin therapy; mean age 46, 87% with type 1; mean A1c 8.2%, mean time in range 47%

  • Participants wore flash glucose monitor for 14 days, at study end completed SMART questionnaire to assess carbohydrate estimation abilities

  • Level of carbohydrate estimation error correlated significantly with glucose variability (r=-0.29; p<0.001), range of glucose values (r=0.33, p<.001) and maximum glucose values (r=0.32, p<.001)

  • Underestimation of carbohydrates in a meal was more significant than overestimation in terms of glucose variability; carbohydrate estimation is associated with lower glucose variability, but does not necessarily affect glucose levels

Effect of Late Bolus Injections on Glycemic Variability Studied by Connected Pens

Johan H. Jendle, Niels V. Hartvig, Anne Kaas, Jonas Moller, Ann-Charlotte M. Mårdby, Sergiu-Bogdan Catrina

  • Real-world study with post-hoc analysis (n=96) of type 1 patients given NovoPen 6 for administration of basal and/or bolus insulin; CGM data uploaded and assessed by HCP; meals identified using validated GRID algorithm

  • NovoPen 6 led to 42% reduction in late bolus doses (taken more than 60 minutes after meal start) and 48% reduction in missed bolus doses

  • Delays in bolus insulin doses associated with increases in glycemic variability (assessed by coefficient of variation); every 10-minute delay in bolus timing associated with 0.47% increase in coefficient of variation

Evaluation of Hybrid Closed-Loop Insulin Delivery System for Patients with Type 1 Diabetes in Real-World Clinical Practice: A One-Year Qualitative Observational Study

Ahmed H. Eldib, Shaheen Tomah, Shilton E. Dhaver, Hannah Gardner, Mhd Wael Tasabehji, Marwa Albadri, Astrid Atakov-Castillo, Elena Toschi, Osama Hamdy

  • Prospective observational study over 1 year of MiniMed 670G use in real world clinical practice among 53 patients with type 1

  • Patients used MiniMed 670G in Auto Mode capturing data every 5 minutes and using a built-in algorithm to automatically adjust basal insulin delivery targeting a glucose value of 120 mg/dl

  • Mean age 47, 62% female, mean baseline A1c of 7.6%

  • Patents experienced mean A1c reduction of 0.3% (p<0.002) and percent time in range increase of 7.9% (p=0.015); fear/worry of hypoglycemia among patients was also significantly reduced (p=0.017) following 12 months of MiniMed 670G use

Alleviating Carbohydrate Counting with a FiASP-and-Pramlintide Artificial Pancreas: A Randomized Pilot Study

Michael Tsoukas, Emilie Palisaitis, Joanna Rutkowski, Julia E. Von Oettingen, Laurent Legault, Jean-François Yale, Ahmad Haidar

  • Randomized cross-over trial among 7 patients (mean age 26, 43% female) assessing the efficacy of a Fiasp-and-pramlintide AID system with simple meal announcement over a Fiasp-alone AID system requiring full carbohydrate counting and boluses

  • Time in range was similar across both treatments (86% for Fiasp alone and 87% for Fiasp-and-pramlintide) suggesting simple meal announcement is sufficient without degrading glycemic control

  • A more extensive study is currently underway to evaluate quality of life improvements from reduced carbohydrate counting

Omnipod Personalized MPC Algorithm at Target Glucose of 110mg/dl Is Safe in Adults and Adolescents without Increasing Risk of Hypoglycemia

Gregory P. Forlenza, Bruce A. Buckingham, Jennifer Sherr, R. Paul Wadwa, Alfonso Galderisi, Laya Ekhlaspour, Cari Berget, Liana Hsu, Melinda Zgorski, Joon Bok Lee, Jason B. Oconnor, Bonnie Dumais, Todd Vienneau, Lauren M. Huyett, Trang T. Ly

  • 96-hour hybrid closed loop study (n=20) in patients with type 1 using modified Omnipod patch pump, Dexcom G4 Share CGM system and the Omnipod Horizon personalized model predictive control algorithm with a lower target of 110 mg/dl compared to 120 mg/dl

  • Mean age 18, 45% female, mean baseline A1c 7.6%; participants in free-living conditions with moderate daily exercise and meal challenges (missed boluses, high fat meals)

  • Time in Range was 76% and 73% overall and 84% and 87% at night at the 110 mg/dl and 120 mg/dl targets, respectively 

  • The algorithm is currently being evaluated in a pivotal study with targets 110-150 mg/dl

Improved Technology Satisfaction and Sleep Quality with Medtronic Minimed Advanced Hybrid Closed-Loop Delivery Compared with Predictive Low Glucose Suspend in People with Type 1 Diabetes in a Randomized Crossover Trial

Olivia Collyns, Renee Meier, Zara Betts, Dennis Chan, Chris Frampton, Carla M. Frewen, Barbara Galland, Niranjala Hewapathirana, Shirley Jones, Natalie Kurtz, John Shin, Robert Vigersky, Benjamin J. Wheeler, Martin De Bock

  • Randomized two sequence cross-over study (n=59) comparing Medtronic MiniMed AHCL to SAP+PLGS in type 1 patients naïve to AID

  • Mean age 24, 58% female, baseline A1c 7.6%

  • AHCL users had higher technology satisfaction than patients on SAP+PLGS; participants >16 on AHCL experienced improved sleep quality compared to those on SAP+PLGS

First Home Evaluation of the Omnipod Horizon Automated Glucose Control System at Specific Glucose Targets in Adults with Type 1 Diabetes

Sue A. Brown, Bruce W. Bode, Carol J. Levy, Gregory P. Forlenza, Bruce A. Buckingham, Amy B. Criego, Trang T. Ly,

  • Single-arm outpatient study using Omnipod 5 insulin pod in direct communication with Dexcom G6 CGM sensor and personalized model predictive control algorithm built into pod

  • Glucose targets were set at 130 mg/dl, 140 mg/dl, and 150 mg/dl for 3 days each followed by 5 days of free choice of 110-150 mg/dl

  • Mean age 35, 72% female, mean A1c 7.1%

  • Time in range (n=18) was 75%, 68%, and 64% for the targets of 130, 140, and 150mg/dL respectively compared to 73% with the lower target of 110mg/dL (n=12) during the 5-day free choice period

  • Overnight time in range was highest with a glucose set point of 130mg/dL at 81% (p<0.01) compared to other set points

  • No serious adverse events 

Assessment of Adhesive Patches for an Extended-Wear Infusion Set

Gina Zhang, Sarnath Chattaraj, Evan Anselmo, Lance Hoffman, Michelle Tran, Shannon Bondy

  • Participants (n=75) were recruited for 3 studies to wear 2 different adhesive patches on conventional IS sites for up to 8 days; used nonfunctional Medtronic pumps to simulate conditions; 

  • Mean age 37, 52% female; adhesive patches placed (total placements = 150) on abdomen (n=139), arm (n=7), and thigh (n=4)

  • Patches currently used with IS had 7/8 day survival rates 67%/63% - 89%/89%; 7/8 day survival rates for 3 new adhesive patches (variants 5-7) using Medtronic Quick-set and Mio Advance were 95%/95%, 85%/75%, and 100%/100% respectively

  • Adhesive patch variant #7 (100% survival rate ≥7 days) was selected for use in the new Medtronic Extended Wear Infusion Set

Do-It-Yourself Artificial Pancreas Systems for Type 1 Diabetes Reduce Hyperglycemia without Increasing Hypoglycemia

Jennifer Zabinsky, Haley Howell, Alireza Ghezavati, Dana M. Lewis, Andrew Nguyen, Jenise C. Wong

  • Retrospective double cohort study between users of DIY AID systems and conventional SAP users; the majority of the DIY data came from type 1s (81%), and parents/caregivers (18%) using the Open APS (77%) and Loop (9%) DIY systems

  • DIY closed loop users had significantly higher time in range values (p<.0001) and less time in hyperglycemia (p<0.0001) without increasing time in hypoglycemia

Type 1 Diabetes: Real-World Insulin Injection Patterns

Sergiu-Bogdan Catrina, Niels V. Hartvig, Anne Kaas, Jonas Moller, Ann-Charlotte M. Mårdby, Johan H. Jendle

  • Post-hoc observational study of insulin injection patterns in adults (n=159) and children (n=47) with type 1 using NovoPen 6 with bolus and/or basal insulin; data uploaded using Glooko cloud system

  • Results showed substantial variation in bolus injection patterns across patients and between adults and children; adults demonstrated greater variation in basal injection timing than children

  • Bolus injection patterns mirrored a 3 meal typical mealtime pattern, but peaks were less well defined in adults than children

  • Numerous patients, both adults and children, administered basal doses late at night or early in the morning

High Treatment Satisfaction with 3-Day Insulin Patch Is Independent of Patient Demographics

Mark Peyrot, Don Johns, Buddug R. Rees, Robert Rook, Jay L. Warner, Brian L. Levy

  • Multi country 48 week cross-over study (n=278) for patients ages 22-75 with type 2; patients randomized 1:1 to either wear a 3-day insulin patch or use pen injections

  • Patients reported preferring the patch to the pen reporting taking meal time insulin was significantly easier (p<0.01), taking meal time insulin was painless (p<0.001), they could do things on the spur of the moment (p<0.001), and they would recommend for mealtime use (p=0.01)

  • There were no significant interactions between patient baseline demographics and preferred insulin delivery method; participants of every subgroup (gender, age, and baseline BMI) were more satisfied with the patch than the pen

Patient-Reported Outcomes for 2,335 Adults with Type 2 Diabetes Using the Omnipod Insulin Management System Show Glycemic Improvement over First 90 Days of Use

Anders L. Carlson, Lauren M. Huyett, Jay Jantz, Albert Chang, Todd Vienneau, Trang T. Ly

  • Retrospective observational study (n=3,592) in patients using Omnipod or Omnipod DASH systems; mean age 57, 54% female, mean diabetes duration 14 years, mean A1c 9.2%

  • Mean A1c significantly decreased from 9.2% to 7.9% (p<0.001) at ≥90 day follow-up; percentage patients meeting ADA-recommended A1c <7% increased from 10% at baseline to 22% at follow-up

  • Total daily insulin dose decreased from average ≥100 U/day to 71 U/day at follow-up

  • Hypoglycemic events decreased from >1/week at baseline to 0.68/week at follow-up

Clinical Study of a New Extended Wear Infusion Set

Jacob Ilany, Ohad Cohen, Noa Konvalina, Gina Zhang, Sarnath Chattaraj


  • Participants (n=26) wore 4 Medtronic Extended Wear Infusion Sets until the Infusion Set failed or 7-day use was reached

  • Mean age 45, mean baseline A1c 7.1%

  • Insulin absorption remained consistent over 7-day use of Extended Wear Infusion Sets; performance and safety of Extended Wear Infusion Sets at 7 days were equivalent to the performance and safety of a current 3-day IS when worn for 3 days

  • Extended Wear Infusion Sets received CE-Mark in February 2020 and are currently undergoing clinical evaluation in US pivotal trial

Real-World Insulin Delivery Patterns for 1,202 Adults with Type 2 Diabetes Using the Omnipod Insulin Management System with Cloud-Based Data Management

Irl B. Hirsch, Ruth S. Weinstock, Lauren M. Huyett, Jay Jantz, Albert Chang, Todd Vienneau, Trang T. Ly

  • Retrospective assessment of insulin delivery patterns and glycemic outcomes for adults with type 2 (n=1,202) using Omnipod system with an integrated BG meter and/or Dexcom G6 CGM with Glooko data management system

  • Most common bolus patterns were 2 to<3/day and 3 to <4/day

  • ≥50% of total daily insulin came from basal delivery among 73% of users and ≥70% total daily insulin from basal delivery in 26% users

  • More frequent bolusing was associated with lower mean blood glucose based on blood glucose meter data; time in range was significantly higher for those bolusing 4 to <6/day compared to the 3 to <4/day group (p<0.05)

A Randomized Study to Evaluate the Efficacy of Insulclock Pen Device in Insulin-Treated Patients with Uncontrolled Type 2 Diabetes

Clementina Ramos, Rodolfo J. Galindo, Muhammad M. Alam, Saumeth Cardona, Bonnie S. Albury, Omolade Oladejo, Francisco J. Pasquel, Priyathama Vellanki, Maya Fayfman, Alexandra Migdal, Georgia Davis, Jeehea Sonya Haw, Guillermo E. Umpierrez

  • Randomized cross-over study in patients with type 1 (n=82); patients used Insulclock system for 12 weeks either with reminders (intervention arm) or without (control arm)

  • Mean age 56, 55% female, baseline mean A1c 9.23%

  • A1c improved significantly more in the Insulclock intervention arm with a reduction of 0.9% compared to 0.7% in the control 

Testing a Novel Infusion Set for Extended Wear Duration

Bruce A. Buckingham, Tatiana Marcal, Lance Hoffman, Gianluca Musolino, Laya Ekhlaspour, Gina Zhang, Sarnath Chattaraj

  • 20 participants (mean age 29, 60% female) were randomized to double-blinded cross over study; participants wore 4 infusion sets for 7 days or until set failure

  • Modified cap of the MiniMed Quick-Set connector was modified to contain either foam (control) or foam + 80 units heparin

  • There was no difference in mean set wear length (survival length) between the foam or foam + heparin caps 

  • Both the control and Heparin Extended wear sets out performed historic studies of Quick-Sets (p=0.005); >70% Extended Wear sets were functioning at 7 days compared to 44% of standard Quick-Sets 

Six Months At-Home Hybrid Closed-Loop vs. Manual Insulin Delivery with Finger-Stick Blood Glucose Monitoring in Adults with Type 1 Diabetes: A Randomized Controlled Trial

Sybil A. Mcauley, Melissa H. Lee, Barbora Paldus, Sara Vogrin,Mary B. Abraham, Leon Bach, Morton Burt, Neale Cohen, Peter G. Colman, Elizabeth A. Davis, Christel Hendrieckx, Martin De Bock, Jane Holmes-Walker, Joey Kaye, Kavita Kumareswaran, Richard Macisaac, Roland W. Mccallum, Catriona M. Sims, Jane Speight, Stephen Stranks, Steven Trawley, Vijaya Sundararajan, Glenn Ward, Anthony C. Keech, Alicia Jenkins, Tim Jones, David N. Oneal

  • RCT study among adults with type 1 using MDI or insulin pump for ≥3 months randomized 1:1 to either use Medtronic 670G hybrid closed loop or continue current therapy for 26 weeks

  • Baseline characteristics similar across control and intervention groups, A1c of 7.5% and 7.4% respectively

  • At 26 weeks 670G users experienced 14.8% more time in range than control (p<0.001) and a mean decrease of 0.4% A1c over control (p<0.001)

  • Patients using 670G also reported a lower diabetes impact on life score (p=0.02) and higher diabetes/specific positive well-being score (p=0.005) compared to control

A Feasibility Study Assessing the Accuracy of a Simplified Meal-Time Bolus Calculation Option

Anirban Roy, Benyamin Grosman, Louis J. Lintereur, Neha Parikh, Di Wu, Dorothy I. Shulman, Mark P. Christiansen, Robert H. Slover, Patrick E. Weydt, Robert Vigersky

  • Experienced MiniMed 670G users (n=12) with type 1 participated in two phases to use a personalized meal simplification algorithm to simplify carb counting

  • Participants input traditional carb counting and bolus data (minimum 40 meal events) for meal simplification algorithm to then use data for machine learning to cluster meals based on meal size and postprandial glycemic response

  • Active algorithm allows user to choose between customized meal sizes (S/M/L/XL) and corresponding boluses 

  • No significant differences in time in range between carb counting and meal simplification algorithm phases; postprandial glycemic profile using personalized algorithm is indistinguishable from using carb counting

Do-It-Yourself (DIY) Loop Is Associated with Psychosocial Benefits

Korey K. Hood, Jessie J. Wong, Sarah Hanes, Ryan Bailey, Peter Calhoun, Roy Beck, Victoria R. Barnes-Lomen, John W. Lum, Brandon Arbiter, Diana Naranjo



  • Observational study of patients with type 1 initiating DIY Loop software integrated with commercial CGM (Medtronic Enlite; Dexcom G4, G5, G6) and commercial insulin pumps (Medtronic; Insulet Omnipod) utilizing “RileyLink” communications hardware bridge

  • Mean age 38, mean A1c 6.64%, 64% female, 70% time in range

  • At 3-month follow-up new Loop users experienced reduced diabetes distress (p<0.001), improved sleep quality (p<0.001), reduced fear of hypoglycemia (p<0.001), and increased confidence in detecting hypoglycemic (p<0.001)

  • The majority of Loop users would be very likely to recommend Loop to others

Estimating an Optimal Meal Bolus for Persons with Diabetes on Multiple Daily Injections Therapy without Carb Counting

Boyi Jiang, Yuxiang Zhong, Pratik Agrawal, Toni L. Cordero, Robert Vigersky


  • 7 adults with type 1 on MDI using the Guardian Connect CGM system participated in this study; mean age 39, baseline average time in range 51%

  • Algorithm used to calculate optimal bolus using machine learning to integrate contextual information with CGM time based data and predict the impact of the meal and adjust bolus accordingly

  • Results suggest automated bolus optimizer can increase time spent in range and that customized boluses for breakfast, lunch, and dinner could achieve positive clinical outcomes

Personalized Hybrid Closed-Loop Therapy Using a Digital Twin in Patients with Type 1 Diabetes: At-Home Data

Benyamin Grosman, Anirban Roy, Di Wu, Neha Parikh, Louis J. Lintereur, Nicole Schneider, Ronald L. Brazg, Satish K. Garg, Robert Vigersky

  • Experienced MiniMed 670G users with type 1 (n=19) completed a 2-week study with either personalized auto or manual mode based on data collected during 3-week run-in

  • Personalized AutoMode increased time in range from 74% to 76%; personalized AutoMode reduced time <70mg/dL from 2.1% to 1.9%

  • Run-in (non-personalized) AutoMode outperformed personalized Manual Mode with time in range dropping from 79% in run-in AutoMode to 68% in Personalized Manual Mode

Study of Insulin Stability Impact on Pump Therapy: Test Model Development

Sarnath Chattaraj, Gina Zhang, Evan Anselmo, Jenny (Hsi) C. Fusselman



  • Fast acting insulins (insulin aspart and lispro) were pumped though >4 infusion sets under simulated-use conditions and tested for preservative effectiveness, dynamic light scattering, and inflammatory responses to evaluate potential causes of IS failure

  • The impact of preservatives in insulin formulations was also tested in a porcine model of CSII therapy with results demonstrating lower levels of preservatives significantly shortened IS wear time from 4.9 days to 1.7

  • “microphage number, the inflammatory response, and device wear-time were all significantly impacted by loss of preservative and trace aggregates/particles 

Unannounced Meals at Home with the Medtronic Advanced Hybrid Closed-Loop

Amir Tirosh, Roy Shalit, Maya Laron Hirsh, Ohad Cohen, Natalie Kurtz, Anirban Roy, Benyamin Grosman

  • 14 adults with type 1 used the Medtronic Advanced Hybrid Closed Loop system for 6 months, 3 without announcing meals (i.e. no pre-meal bolus for meals <80g), and 3 with announcing meals; patients used setpoint of 100mg/dL, active insulin time of 2 hours

  • Baseline A1c 6.8%, baseline TIR 68%

  • 82% meals (n=2720) did not require “salvage bolus”; TIR during unannounced meals was 68%

  • Though AHCL is programmed for meal announcement the system provided safe glycemic control without increased hypoglycemia, but meal announcement is still necessary to achieve optimal targets

Effects of the T:slim X2 Insulin Pump with Basal-IQ Technology on Glycemic Control in a Pediatric Urban Academic Diabetes Practice

Alyson Weiner, Elizabeth Robinson, Rachelle Gandica

  • Patients 6-18 years old (n=47) with type 1 used the t:slim X2 insulin pump with PGLS Basal-IQ integrated with the Dexcom G6 CGM for 6 months; 45% female, mean age 14

  • At 6 months A1c was 7.5% compared to 7.6% at baseline (not significant)

  • t:slim X2 with Basal-IQ did not improve glucose variability in the study population after 6 months

  • Subjects used Basal-IQ techology for a median 88.5% and 90% of the time at 3 and 6-months respectively

Durable vs. Disposable Insulin Pumps for Type 1 Diabetes: Health Care Costs and Utilization

Mona Shah, Cyrus Zhu

  • 5-year longitudinal analysis in patients, with type 1 using either durable (n=2,013) or disposable (n=642) insulin pumps 

  • 1 year baseline, 4 year follow-up study; patients qualified if they made ≥1 pump related claim/6 month interval

  • Over 4-year follow-up durable pumps cost $6,606 less than disposable pumps (p<0.0001); durable pumps had $1,037 less in out of pocket expenses than disposable pumps

  • Significant economic advantage for patients using durable insulin pumps

CSII and Insulin: Does Extending the Wear Duration of Infusion Sets Save Expensive Insulin?

Sarnath Chattaraj, Marisa Fienup, Gina Zhang, Marie Tieck

  • Modeling and experimental studies analyzed the impact of Medtronic extended wear CSII on reduced insulin waste and associated cost

  • Modeling 7 day wear instead of 2-3 day set life reduced the amount of insulin needed by reducing wasted insulin (17-22 vials of insulin to 15 vials and 18-23 vials to 15-16 depending on tubing length of CSII)

  • Modeled reduction in insulin waste correlated to patient savings ~$1,500-$3,000/year

Patients’ Interpretation of Ambulatory Glucose Profile (AGP), CGM, and Pump Reports

Deborah M. Mullen, Richard M. Bergenstal, Mary L. Johnson

  • 21 patients with ≥6 months experience with CGM and insulin pumps completed interviews to assess Ambulatory Glucose Profile preferences

  • Participants were 100% female, 86% white, 50% 14-35 years old, 50% 35-65 years old

  • Participants overwhelmingly preferred phone/watch interfaces to computers; many also do not download and/or look at their data outside a clinic; participants reported checking their CGM values on their phone 1-4 times/hour

  • Participant responses indicate a need for standardization among CGM and insulin dose reporting and graphics; patients strongly prefer colored graphs to numerical data


Glucose Monitoring – BGM and CGM

Abbott Poster Shows FreeStyle Libre Reduces A1c by 0.9% (8.5% baseline) in Non-Insulin Type 2s (n=497) and 0.6% in Basal-Only Type 2s (n=277) After Six Months; Reductions Sustained at 12 Months

A very exciting Abbott-sponsored poster (84-LB, “HbA1c Reduction after Initiation of the FreeStyle Libre System in Type 2 Diabetes Patients on Long-Acting Insulin or Noninsulin Therapy”) presented by Dr. Eden Miller (Diabetes Nation) showed significant A1c reductions in basal-only and non-insulin-using type 2s one year after using FreeStyle Libre. The retrospective analysis looked at LibreView data from November 2017 – September 2019, A1c data from Quest Diagnostics, and medical and pharmacy claims data from Decision Resources Group. Compared to A1cs values recorded within 0-180 days before starting FreeStyle Libre, 6-month A1c values were reduced by a remarkable 0.9% (baseline: 8.5%) in non-insulin type 2s (n=497; p<0.0001). Similarly, basal-only type 2s (n=277) saw their A1cs drop from 8.5% to 7.9% (p<0.0001) after six months. Looking out to 12 months, non-insulin type 2s saw a 0.7% A1c reduction, from 8.6% to 7.9% (n=120; p<0.0001) and basal-only type 2s saw a 0.5% A1c reduction, from 8.4% to 7.9% (n=87; p=0.001). Unfortunately, data on demographics and comorbidities were not assessed. Similarly, data on non-insulin medications (e.g., SGLT-2s, GLP-1s) was not available and it’s certainly possible that some of the participants analyzed may have been initiated on a new medication and FreeStyle Libre at the same time. Still, the results provide convincing evidence that FreeStyle Libre has considerable value beyond intensive insulin users. As stated by Dr. Miller in Abbott’s press announcement of the abstract, “These data highlight how use of Abbott's continuous glucose monitor could be game-changing for people beyond intensive insulin users, translating to broader use of the technology to benefit all those living with diabetes, no matter where they are in the spectrum of care.”

  • Looking ahead, many (especially payers) will be especially interested in cost savings analyses for CGM in these populations. This was a particularly hot topic at ADA yesterday, where we saw Dr. Rich Bergenstal present data showing a 60% reduction in acute diabetes events and 33% reduction in all-cause hospitalizations for ~2,500 adults with type 2 diabetes after starting on FreeStyle Libre – this study included adults on intensive and non-intensive insulin therapy. We also saw an observational study from France linking FreeStyle Libre to a 47% reduction in DKA rates in type 2s on intensive insulin therapy. Lastly, Dr. Irl Hirsch looked at previous studies and some of his own calculations to estimate the US healthcare system could save $4.6 billion in DKA-hospitalization cost savings alone by using CGM.

Swedish GOLD Study Extension (n=107) Shows Sustained Effects with Dexcom G4 vs. SMBG Out to 2.5 Years: 0.35% A1c Reduction, +2.1 Hours/Day TIR, Satisfaction and Hypo Confidence Improvements

University of Gothenburg’s Dr. Marcus Lind presented strong results from the 12-month, SILVER extension phase of GOLD, comparing SMBG vs. CGM. GOLD was a 16-month Swedish trial, that enrolled 161 type 1s on MDI; after a 1.5-month run-in, participants were randomized to either SMBG or CGM (Dexcom G4) for six months, followed by four months of wash-out, and six-months of crossover. GOLD showed strong advantages for CGM vs. SMBG on every glycemic outcome, including a 0.4% A1c improvement and ~30 minutes per day less time <70 mg/dl (4.8% vs. 2.8%). Of the 141 GOLD trial completers, 107 were enrolled in the 12-month SILVER extension phase. In SILVER, all participants received CGM and had “brief consultations with a diabetes nurse” every three months. Compared to SMBG users at the end of the GOLD trial, switching to CGM produced a 0.35% A1c reduction (8.3% vs. 8.0%; p<0.001), 2.1 more hours/day time in range (51% vs. 43%; p<0.001), and significant reductions in hypoglycemia. As shown in the line graph below, A1c remained relatively steady through the entire 12-month extension, demonstrating the benefits of CGM were sustained beyond the initial 1.5-year trial.









Time <54 mg/dl




Time <72 mg/dl




Time in range




Time >180 mg/dl




Time >250 mg/dl




Standard deviation

4.2 mg/dl

3.6 mg/dl

-0.6 mg/dl


  • Compared to run-in, CGM showed even stronger improvements compared to SMBG. The end of the SILVER extension study represents ~2.5 years from run-in, providing compelling evidence for the long-term sustainability of CGM benefits. Compared to run-in, CGM improved A1c by 0.5% (baseline: 8.5%) and boosted time in range by 2.7 hours/day (!), from 40% to 51%. Though this is a significant improvement, 49% time “out of range” is obviously not exciting. Time in hypoglycemia was also greatly reduced, from 2.1% to 0.6% (-22 min/day; p<0.001) for time <54 mg/dl and 5.5% to 2.9% (-36 min/day; p<0.001) for time <70 mg/dl.

  • The 51% time in range, even with CGM, is strikingly low. At ATTD 2020, we saw a post-hoc analysis of the GOLD study demonstrating how difficult it is to achieve the new consensus CGM metrics guidelines, particularly in specific populations like adolescents. In the study, just 3 out of the 137 participants analyzed met both A1c <7% and time in range >70% goals when using CGM, and only one participant met both goals on SMBG. The graph below really underscores the difficulty – the green box was added by us and highlights the areas where both A1c and time below 70 mg/dl targets are met.

  • The SILVER study showed sustained benefits with Dexcom G4 vs. SMBG on nearly all behavioral and satisfaction outcomes. Diabetes Treatment Satisfaction Questionnaire scores jumped from ~25 to 31 with CGM. Well-being, measured by WHO-5’s Well-Being Index, rose 11% from 60 to 66. The Swedish-Hypoglycemia Fear Scale showed significant improvements around worry and a reduction in hypoglycemia avoidance behaviors, though this second reduction was not statistically significant (p=0.07). Hypoglycemia confidence scores rose from 3.2 on SMBG to 3.45 with CGM (p<0.001).

  • Through the 12-month extension, there were 5 incidents (between 4 patients) of severe hypoglycemia and no DKA. None of these events were deemed to be related to CGM-use, though it is suprising in our view that the CGM didn’t prevent the severe hypos. In the original GOLD trial, there was a non-significant trend towards fewer severe hypo events with CGM: there were five events of severe hypoglycemia during SMBG (0.19 per 1,000 patient-years) vs. one event with CGM (0.04 per 1,000 patient-years).

SENCE RCT (CGM in Young T1s) 12-Month Extension: Sustained CGM Utilization and Hypo Reduction, but Still Over Half of Day >180 mg/dl

Yale’s Dr. Michelle Van Name presented 12-month follow-up data from the SENCE RCT of CGM in children ages 2-7, showing sustained CGM use and hypoglycemia reduction but no/minimal improvement in time in range or hyperglycemia. Positive primary outcomes from the six-month study were presented at ADA 2019. As a reminder, the 14-site trial randomized participants to three groups: CGM + a family behavioral intervention (FBI; n=50) vs. CGM with standard education (CGM-only; n=44) vs. SMBG (n=49). FBI consisted of five visits (weeks 1, 3, 6 13, 19) with additional training on using and living with CGM, using CGM away from home, CGM burnout, and more. The CGM groups used the Dexcom G5 non-adjunctively, with access to Dexcom Share. Participants were not on CGM at enrollment, so baseline metrics were derived from blinded CGM. After the 6-month trial portion completed, 131 participants opted to continue using CGM for an additional six months—we’re not sure if G6 was employed during this follow-up phase. For the two CGM groups, the extension period consisted of CGM with no continued educational intervention; members of the SMBG control group who opted to continue were crossed over to receive CGM + FBI.

  • Positive finding #1: CGM usage remained high in both CGM + FBI and CGM-only groups at 12 months. 86% and 95% of children in the respective groups continued to wear CGM ≥6 days per week through the extension phase; compare these figures to 93% and 90% at 6 months.

  • Positive finding #2: Hypoglycemia reductions observed at 6 months in both CGM groups were sustained at 12 months. In both CGM groups, time <70 mg/dl was approximately cut in half at 12 months (CGM+FBI: 69->36 mins/day; CGM-only: 81->39 mins/day). Similarly, percent time <54 mg/dl at 12 months as reduced to ~one-third of baseline levels (CGM+FBI: 30->10 mins/day; CGM-only: 35->12 mins/day). In accordance with these findings, the mean hypoglycemic event rate per week was also cut in half in both CGM groups at 12 months (CGM+FBI: 2.3->1.2 events/week; CGM-only: 2.8->1.2 events/week).

    • Those in the SMBG group who crossed over to CGM + FBI for the 6-12 month extension phase realized similar hypoglycemia benefit. From CGM initiation at 6 months to follow-up at 12 months, mean time <70 mg/dl fell by 54 minutes per day (6.2% to 2.4%). There was no change in time in range.

  • Cautionary reminder: Even after a year of CGM use, young children with type 1 diabetes in a rigorous study treated at top clinical centers are still spending over half of their days with glucose levels >180 mg/dl and just ~40% in-range. Because CGM use did not significantly reduce time >180 mg/dl in either CGM group, time in range hovered at ~40% across all reported measurement periods. While it is encouraging that there was no compensatory increase in hyperglycemia as a result of reduced hypoglycemia exposure, given all we know (and are learning) about the risks of hyperglycemia exposure and the legacy effect, future interventions must target the high end. As Drs. Lutz Heinemann and David Klonoff pointed out late last year in the wake of T1D Exchange Registry data showing drastically increased CGM adoption but also increased A1c, “Investment into CGM usage does not result in an automatic improvement in glucose control.” They go on to make the case for better, more comprehensive training programs and comparative studies of such programs.

  • Dr. Van Nam highlighted the uniqueness of the study population (for a diabetes technology study), as ~1/3 were minority, ~1/3 were on a pump, and the mean A1c was 8.2%. More than one-third of study participants did not have private health insurance, and the majority of parents did not have a bachelor’s degree or higher.

Pre-/Post- Study Shows FreeStyle Libre is Associated with 60% Reduction in Acute Diabetes Events, 33% Reduction in All-Cause Hospitalization

In a retrospective pre-/post- study leveraging IBM Watson Health’s MarketScan database, presenter Dr. Rich Bergenstal (IDC) and co. found striking reductions in acute diabetes events (-60%) and all-cause hospitalizations (-33%) among 2,463 adults with type 2 diabetes on short- or rapid-acting insulin who began using FreeStyle Libre between November 2017-September 2018. Though not without limitations (e.g., retrospective; no A1c data available; exclusion of Medicare fee-for-service and Medicaid populations; no access to socioeconomic or educational data), Dr. Bergenstal suggested this study offers “compelling support for the use of flash CGM to both improve clinical outcomes and potentially reduce costs in this patient population.” In the live chat off to the side, this presentation incited optimism for CGM reimbursement as well as calls for researchers to take the next logical step for payers by calculating changes in expenditure. While true calculations of overall cost can be complex, Dr. Barry Ginsberg proposed simplifying by reporting cost-savings from reduced hospital days (~$8,000 each), and Dr. Robert Gabbay suggested extrapolating savings from A1c reductions. In the same chat, Dr. Irl Hirsch, a co-author, alluded to a poster (875-P) from the same group showing that FreeStyle Libre use is associated with reduced acute diabetes events regardless of prior blood test strip usage. He proclaimed, “the SMBG requirement [for Medicare CGM reimbursement] makes no sense.” Dr. Aaron Neinstein aptly responded, “Agree... the SMBG requirement is, in the parlance, bats$&$.”

  • Methods: MarketScan is a research database containing commercial and Medicare supplemental insurance claims for more than 30 million individuals in the US. Because it tracks down to the level of the individual and claims for health care facility utilization and pharmacy, the researchers were able to identify codes for inpatient/emergency outpatient utilization in the six months prior to and after the first use of FreeStyle Libre. Dr. Bergenstal commented that the database appears to hold “a reliable collection of data” due to the study participant characteristics: mean age 54 years, 90% with lipid disorder, 88% with hypertension, 60% with obesity, 48% with neuropathy, etc. 

  • As seen in the Kaplan-Meier plot below, there was clear separation and 60% overall reduction (HR=0.40, 95% CI: 0.31-0.51, p<0.001) in acute diabetes events in the six months after vs. six months prior to first FreeStyle Libre use. Acute diabetes events are defined as a combination of inpatient and outpatient emergency events including hyperglycemia, hypoglycemia, DKA, hypoglycemic coma, and hyperosmolarity. Overall number of events fell from 221 (0.18 per patient-year) in the six months pre-purchase to 84 (0.07 per patient-year) in the six months post-purchase. In sub-analyses, this reduction was significant for both genders and across age (stratified by <50 and 50+ years).

  • In the secondary outcome, there was a 33% reduction in all-cause hospitalization in the six months post-purchase vs. six months pre-purchase (HR=0.67, 95% CI: 0.58-0.77, p<0.001). Overall hospitalizations fell from 516 (0.42 per patient-year) to 331 (0.28 per patient-year). Dr. Bergenstal pointed out that, as expected, the number of endocrinology-related hospitalizations fell by 59% (6.4 per 100 patient-years to 2.6 per 100 patient-years). Also intriguing, he noted, against the backdrop of a resurgence in diabetes complications in the US, were reductions in hospitalizations for infectious disease, respiratory, renal, and hepatobiliary/pancreatic issues.

Dexcom G6 With Urgent Low Soon Alerts Reduces Rebound Hyperglycemia Events by 7% After <70 mg/dl Event, 33% After <54 mg/dl Event vs. Dexcom G5

Dexcom’s Dr. Giada Acciaroli presented real-world data from 24,518 Dexcom users who transitioned from Dexcom G5 to Dexcom G6 (with urgent low soon alerts) in 2018. Results showed significant reductions in rebound hyperglycemia events and duration with Dexcom G6 users who had their predictive low glucose alerts turned on. As a reminder, G6’s “Urgent Low Soon” alert comes 20 minutes ahead of a predicted <55 mg/dl event. Rebound hyperglycemia was defined as glucose levels >180 mg/dl occurring within 2 hours of a hypoglycemic event (glucose value <70 mg/dl or <55 mg/dl). Following a hypoglycemic event <70 mg/dl, the number of rebound hyperglycemia events per week was reduced from 1.83/week to 1.7/week (p<0.001). Additionally, mean duration of these events was reduced from 214 minutes to 197 minutes (-8%; p<0.001). Following hypoglycemic events <55 mg/dl, the number of rebound hyperglycemia events per week was reduced by 33% (from 0.78/week to 0.52/week; p<0.001) and duration was reduced by 22% (from 219 min to 171 min; p<0.001).

  • Dr. Acciaroli also presented real-world Dexcom data demonstrating the correlation between rebound hyperglycemia events and glucose variability. Looking at the quartiles of Dexcom users with lowest glucose variability (%CV<31.2%) and highest glucose variability (%CV>39.1%), the unstable cohort saw 7.4x more rebound hyperglycemia events per week, 0.41 events vs. 3.03 events/week (p<0.001). This translates to a difference of one event every ~17 days in the low variability quartile vs. one event every ~2.3 days in the high variability quartile.

Use of Dexcom G6 Delivers Two-Times Greater A1c Improvements vs. Connected BGM Across All Baseline A1c Groups in Onduo’s Virtual Diabetes Clinic

Onduo’s Head of Clinical Affairs Dr. Ron Dixon presented data from the company’s Virtual Diabetes Clinic, demonstrating ~2x greater A1c reductions in participants who used CGM (Dexcom G6) vs. those using a connected BGM. The retrospective analysis looked at 612 Onduo participants from February 2018 – April 2019, comparing 213 participants who were initiated on Dexcom G6 with 399 who were not. Participants were well matched at baseline with a mean ages of 53 and 54 years, mean BMIs of 35 and 36 kg/m2, and baseline A1cs of 7.8% and 7.7% for the CGM and non-CGM groups, respectively.

  • For the highest baseline A1c group (>9%), the CGM group saw a mean A1c reduction of 3.3% (!), compared to 1.7% for the non-CGM group after six months (p<0.001 for between-group difference). Though less pronounced, the CGM group saw greater A1c reductions vs. non-CGM in every other baseline A1c grouping. For baseline A1c of 8%-9%, A1c reductions were 1.3% and 0.6% for CGM and non-CGM groups, respectively (p=0.004). For baseline A1c of 7%-8%, these reductions were 0.4% and 0.1% (p=0.02). Lastly, for those with baseline A1cs <7%, A1c rose by 0.1% in the CGM group and 0.2% in the non-CGM group.

  • As seen in the graphs below, using CGM as part of Onduo’s virtual clinic had a “flattening out” effect on participants’ A1cs across the initial A1c cohorts. While initial A1c in the groups ranged from <7% all the way to ~11%, after six months, mean A1c across all groups was between ~6%-8%. This reminded us of Dr. Rich Bergenstal’s (International Diabetes Center) presentation yesterday in which participants on MDI+SMBG were brought onto Medtronic’s Advanced Hybrid Closed Loop and saw their time in range shoot up from 45% to 65% (+4.8 hours/day). So often, clinicians and researchers seem to exclude participants who are CGM- and pump-naïve, and those with poor baseline A1cs, when in fact, they may have the most to gain from using technology like CGM.

Dexcom-Sponsored Symposium Sheds Light on Success, Failures, and Opportunities for Inpatient CGM Implementation During the COVID-19 Pandemic

During a Dexcom-sponsored symposium, Dr. Shivani Agarwal (Albert Einstein College of Medicine) presented takeaways from implementing inpatient CGM during the COVID-19 pandemic at Montefiore Medical Center, a three-hospital system in the Bronx. The FDA first authorized use of CGM in the hospital setting in April in the early stages of the pandemic, giving both Abbott and Dexcom the greenlight. Dr. Agarwal characterized the overall experience as positive, sharing key insights for early implementation, logistics, training, challenges, lessons, and future opportunities. For context, Dr. Agarwal noted that some of these experiences were unique to her hospital and patient population and that naturally, other hospital systems attempting to bridge the implementation gap should consider their unique circumstances. Regardless of the conditions, Dr. Agarwal shared the key to success in any setting is obtaining buy-in and support all around from hospital staff.

  • Early Implementation: Identifying eligibility criteria for patients requiring CGM on top of maintaining a positive attitude were two critical components of success during implementation. At her hospital system, type 1 patients, individuals with documented hypoglycemia, those requiring insulin drips and one-hour fingerstick measurements were eligible. Exceptional circumstances per nurse request or where self-monitoring was not possible also existed. To reiterate, creating a positive environment to encourage hospital buy-in was key.

  • Logistics: Inpatient diabetes nurse practitioners were responsible for inserting sensors in the arm. Receivers were placed on the door of the patient room facing outwards, within 20 feet of the patient, and reused after sanitization.

  • Training: The inpatient diabetes team assembled dedicated trainings to help hospital nurses and medical staff learn how to interpret CGM data and make treatment decisions. Paper instructions were placed on patient doors, and 1:1 conversation was regularly utilized between the diabetes team and the hospital staff. Alarm triggers were used for low and high blood sugars, and for when signal disconnects occurred, indicating the need for a point-of-care fingerstick test. The hospital validated CGM readings with the common “20/20” rule, where CGM readings must be within 20 mg/dl or 20% of the fingerstick value. Interestingly, not all data was documented into the hospital’s electronic medical records system. Dr. Agarwal noted that an ongoing point of clarification is determining how much CGM data is necessary to collect and store from patients.

  • Successes: To-date, Dr. Agarwal shared that CGM has been implemented on twenty patients within their hospital system. The reduced need for personal protective equipment and hospital staff contact led to massive hospital buy-in. CGM use also shifted how care teams within the inpatient ward operated and increased appeal for additional medical device usage in inpatient settings. Today, hospital physicians partner with inpatient nurses and care teams to provide educational lessons about technology used in the hospital setting, increasing overall enthusiasm for future pilots.

  • Future: Dr. Agarwal shared that additional information on integrating CGM into data records and identifying patient characteristics most amenable to CGM are critical. Additional randomized control trials to determine the safety, accuracy, and utility of CGMs in hospitals are needed, along with securing legal approval to use cellphones as receivers.

At an Abbott-sponsored symposium, Dr. Irl Hirsch (University of Washington) shared his quick math estimating that in the US alone, the healthcare system could see an astounding $4.6 billion in DKA-related cost savings by using CGM in all people with diabetes.  Note that the following calculations come with Dr. Hirsch’s disclosure that he is not a healthcare economist. Dr. Hirsch’s math hinges on two recent studies from the UK (one from ADA 2019 and another published in Diabetologia in 2019) showing that the use of Freestyle Libre reduces DKA events and hospitalizations by 80%. Using an average cost of $26,556 per DKA-related hospitalization and 188,965 DKA-related hospitalizations recorded in 2014 (Diabetes Care, 2018), Dr. Hirsch calculated the cost of DKA in 2014 at ~$5.0 billion. Then turning to a medical inflation calculator, Dr. Hirsch estimated the cost of DKA hospitalizations in the US in 2019 at $5.73 billion. Using the 80% reduction in DKA hospitalizations using FreeStyle Libre from above, the total potential hospitalization cost savings comes out at $4.6 billion. This simple back-of-a-napkin calculation from Dr. Hirsch is pretty compelling evidence for, at the very least, expanding access to CGM to people with diabetes and cost-savings analyses will become increasingly important as the field continues to move into new populations, particularly basal-only or non-insulin type 2s.

  • Dr. Hirsch’s estimate for annual DKA-related hospital spending (~$6 billion) is a shocking number, but less than 2% of the CDC’s most recent estimate of $327 billion for the total cost of diagnosed diabetes in the US. Additionally, undiagnosed diabetes, prediabetes, and gestational diabetes has been estimated to cost another $77 billion. In classic Dr. Hirsch fashion, he tried to put some of these massive numbers into context, this time using three examples: (i) one billion seconds is equal to ~31.5 years; (ii) one billion pennies stacked on top of each other would make an 870-mile high tower; and (iii) if you saved $100,000 every year, it would take you 10,000 years to save $1 billion.

  • Dr. Hirsch also highlighted findings from a retrospective, observational analysis which showed immediate reductions of adverse events among patients with type 2 diabetes after initiating FreeStyle Libre. Additionally, there was a less significant, but still notable drop in all-cause inpatient hospitalizations with FreeStyle Libre. This analysis was presented as an oral presentation this morning and is highlighted in more detail in Diabetes Technology highlight 3 above.


  • The real-world, prospective FUTURE study from Belgium showed reductions in adverse diabetes event hospitalizations, hypoglycemic comas, work absenteeism, and in days spent in the hospital due to diabetes events following nationwide reimbursement for FreeStyle Libre in type 1s. Dr. Hirsch pointed to the reduction in work absenteeism as being particularly notable, because it is an often-forgotten economic benefit from using continuous glucose monitoring systems.

  • Lastly, cost-effectiveness analysis of the 2017 2017 DIaMonD study (6-months of Dexcom G4 in type 1s) projected a reduced risk of long-term complications and a quality-adjusted life-year (QALY) increase of 0.54. Incremental cost-effectiveness ratio (ICER) for Dexcom G4 was $98,108/QALY with a 7-day sensor and $35,459 with a 10-day sensor. As Dr. Hirsch noted, US payers are open to paying for interventions <$100,000/QALY and most payers will cover interventions lower than <$50,000/QALY. UK’s NICE generally considers products conferring ~$30,000-$40,000 per QALY worth covering. We would anticipate the newer and improved CGMs would show even more promising long-term benefits. Dr. Hirsch focused these studies as providing strong evidence that CGM is a powerful, and underutilized, tool for bending the massive and quickly rising cost curve of diabetes. Many other clinicians have expressed similar sentiments, perhaps most eloquently written in a powerful commentary published last year in DT&T, calling current eligibility requirements for CGM insurance coverage “clinically irresponsible” and “penny wise and pound foolish.”

Analysis Estimates ~$300 Million in Annual Cost Savings if CGM Were Given to All Type 1 Medicaid Beneficiaries; $1.15 Billion in Reduced Costs vs. $850 Million to Cover Dexcom G6 for All Type 1 Beneficiaries

Mr. Michael Minshall (Certara Evidence & Access) presented a budget impact analysis estimating ~$300 million in net annual cost savings if CGM were given to all type 1 Medicaid beneficiaries. The analysis calculated ~$1.15 billion in total reduced costs with real-time CGM, compared to ~$850 million to cover Dexcom G6 (in lieu of SMBG) for all ~600,000 Medicaid beneficiaries with type 1 diabetes. A walkthrough for these calculations is provided in the bullets below and a summary of analysis’ findings are in the table below.

  • Mr. Minshall estimated a total of 593,378 Medicaid beneficiaries with type 1 in 2020. This came from ~72 million Medicaid and CHIP enrollees in 2019, multiplied by 13.9%, the reported prevalence of diabetes in the Medicaid population. Of those ~10 million Medicaid beneficiaries with diabetes, Mr. Minshall estimated ~585,000 with type 1 based on the prevalence of type 1 vs. type 2 diabetes in the general US population. Lastly, Mr. Minshall added on ~9,000 new type 1s in the Medicaid population, in line with historical trends.

  • Mr. Minshall estimated $396-738 million in cost savings related to A1c reductions from CGM vs. SMBG. Based on data from the 2017 DIaMonD trial (Dexcom G4 vs. SMBG in type 1 MDI adults), Mr. Minshall assumed CGM could deliver a 0.6% A1c reduction vs. SMBG. Mr. Minshall then referred to two studies estimating annual savings from reductions in A1c. One study (Gilmer et al., 2005) estimated $1,111 in savings from a 1% A1c reduction, while another (Wagner et al, 2001) estimated $2,073 in savings for the same A1c reduction. These cost savings, multiplied by 0.6 (from the 0.6% A1c reduction with CGM), formed the low and high ends of ranges used in Mr. Minshall’s cost analysis.

  • Mr. Minshall estimated $210 million in savings from reduced severe hypo hospitalizations and $207 million from reduced DKA hospitalizations. The estimated reductions in severe hypoglycemia and DKA-related hospitalizations came from the RESCUE study in Belgium, which found 73% reductions in severe hypoglycemia hospitalizations and 80% reductions in DKA hospitalizations after real-time CGM was reimbursed for type 1s. Notably, Dr. Irl Hirsch cited similar statistics in his talk on Saturday where he calculated a potential $4.6 billion in potential cost savings for the entire US around DKA-related hospitalizations. The cost per severe hypo hospitalization was estimated at $4,068 (Liu et al., 2018) and the cost per DKA hospitalization was estimated at $9,733 (Tieder et al., 2013). Mr. Minshall noted that both of these numbers were on the more conservative side of estimates.

  • Lastly, according to Mr. Minshall, moving all type 1 Medicaid beneficiaries from SMBG to real-time CGM would add ~$849 million in added annual cost. These calculations estimated total SMBG cost at $1,091/year per user ($0.39/test strip, $0.04/lancet, 7 fingersticks/day). Total cost of Dexcom G5/G6 were based on CMS’ Fee Schedule for US state Medicaid plans for 4 transmitters and 37 sensors. 

  • Limitations of this budget impact analysis include indirect costs associated with diabetes care and treatment. Mr. Minshall was careful to note the limitations of this analysis stating they only investigated the direct costs associated with A1c and hypoglycemia or diabetic ketoacidosis and therefore could be missing hidden costs. Furthermore, this budget impact analysis did not investigate indirect benefits of CGM such as increased productivity and quality of life metrics that could be important in future assessments.

Dr. Richard Bergenstal and Emma Wilmot Face off on Whether Technology Alone Can Prevent Severe Hypoglycemia

In a fascinating debate, Drs. Richard Bergenstal (International Diabetes Center) and Emma Wilmot (University Hospitals of Derby and Burton) hit on whether or not technology alone can prevent hypoglycemia and its complications in patients. Dr. Bergenstal argued in favor of technology alone while Dr. Wilmot took the opposite stance for the purposes of the debate. Overall, the discussion touched on a number of valuable topics, highlighting the distinctions between A1c and time in range, the value of structured education in diabetes therapy, and questions of access and reimbursement. See below for a table summary of the key points from both sides.

Dr. Bergenstal’s arguments FOR “Technology Alone Can Prevent Severe Hypoglycemia”

  • Flash glucose monitoring is sufficient to improve glycemic awareness. For example, a European analysis of over 60 million FreeStyle Libre scans found a direct correlation with the number of scans per day and reduction in hours spent in hyperglycemia. For example, in the UK, as the number of scans increased from 0 to 50/day, the hours per day spent above 180 mg/dl decreased on average from ~11 hrs to ~6 hrs. Regardless of external circumstances, simple technology has the power to reduce time spent in hypoglycemia and increase time in range.

  • Dr. Bergenstal showed one-day ambulatory glucose profiles from three patients enrolled in the HypoDE study who all had the same A1c of 6.7%. However, the time spent in hypoglycemia significantly varied by patient, with data suggesting high-quality technology could reduce time spent in hypoglycemia. For example, one patient on MDI therapy spent 9% of time (130 min/day) in hypoglycemia while a patient using an insulin pump with CGM spent just 6% of time below range (86 min/day). Finally, the patient using a hybrid closed loop system spent just 1% of time in hypoglycemia (15 min/day).  

  • The benefits of diabetes technology are not limited to just the type 1 population. Dr. Bergenstal presented data from the REPLACE trial, showcasing blood glucose management among the type 2 population (n=224) on flash glucose monitoring relative to those on SMBG. After three months, patients on flash CGM saw hypoglycemia rates decrease 43% compared to those on SMBG. These findings illustrate that regardless of patient subgroup, patients who receive advanced technology such as CGM naturally improve.

  • Continuous Glucose Monitoring as a Matter of Justice – Dr. Bergesntal showcased a new paper that argues that improved glycemic control resulting from technology such as CGM has further downstream benefits, including decreased stigmatization, improved autonomy, stronger interpersonal relationships, and higher productivity. The reductions seen in hypoglycemia, simply through technology onboarding, in addition to the potential social and systemic benefits are a case for its sole usage in diabetes treatment and management.

Dr. Wilmot’s arguments AGAINST “Technology Alone Can Prevent Severe Hypoglycemia”

  • According to Dr. Wilmot, no technology has been able to demonstrate a significant and sustained improvement in hypoglycemia awareness, a strong risk factor for severe hypoglycemia. Dr. Wilmot presented data from 135 CGM users in Utah, 33% of which had impaired awareness of hypoglycemia. Despite CGM usage, 25% of patients had at least one episode of severe hypoglycemia within the last six to twelve months. However, if you look at patients who receive structured education, outcomes change. A 2005 study among 9,683 type 1 patients in Germany found a 50% reduction in severe hypoglycemia after 20 hours of structured inpatient training on insulin therapy. While technology certainly confers benefits, enhancing it with educational and behavioral health support yields significantly improved outcomes.

  • Access to technology is highly limited because of gaps in national policy and reimbursement. Based on 2018 data from national registries, pump use is highly fragmented by age group and region in the UK. For example, only ~37%-38% of young diabetes patients (<18 years old) in Wales, Scotland, and England use an insulin pump compared to 70% in Denmark. With such large fractions of society disconnected, it is unreasonable to expect technology alone to fill gaps in care.

  • Regardless of the technology, there are always subgroups of patients who struggle with onboarding and use. Dr. Wilmot presented data from Pediatric Diabetes (n=92 youth) which found that 30% discontinued using the MiniMed 670G system. Skin reactions, data overload, and unwanted alarms are a few of the many reasons why certain patient populations, regardless of the device, do not bear the benefits of technology. A separate study on adults published in Diabetes Care (n=84) found that after 12 months, 1/3rd of individuals had also stopped using the 670G mostly because of sensor issues (62%) and challenges in securing supplies (12%).

  • Even with technology and education, some patients may still not benefit. Dr. Wilmot presented a case study of one patient, a type 1 woman since 1999, who started facing impaired awareness in 2009. Despite receiving structured education in 2011 and an insulin pump in 2012, she still experienced three severe hypoglycemic episodes per week. The case became so difficult that she was banned from a local McDonalds because staff mistook her hypoglycemia for inebriated behavior, and her husband had to stop working to observe her. However, after undergoing islet cell transplantation, she began to see improvements in time in range, reducing her severe hypoglycemia by ~90%. Furthermore, in extreme cases, additional intervention beyond technology and education are required to support diabetes patients.

Dr. DeSalvo’s Clinical Pearls on Shared Goal-Setting, CGM Alerts and Arrows, Hybrid Closed Loop, and Nutrition

We enjoyed a pragmatic talk chock full of clinical pearls for pediatric type 1 diabetes management from Texas Children Hospital’s Dr. Daniel DeSalvo. Read on for “Dr. Dan’s” tips for goal-setting, setting sensor glucose alerts, dosing off CGM trend arrows, adjust hybrid closed loop settings, and even medical nutrition.

  • Dr. DeSalvo is on board with the aspirational consensus goal of 70% time in range, but emphasized the importance of personalizing goals. This requires considering family circumstances, including their access to technology, risk/fear of hypoglycemia, the patient’s diabetes support system, and diabetes distress/burnout/depression. He recommended that personalized goals should be achievable­, offering the example of a person with severe diabetes distress and burnout with a suboptimal diabetes support system and time in range of 20% whose goal might be set at 25% in-range. Finally, he called for using supportive language when speaking with patients and praising their successes.

    • Prior to arrival of consensus time in range targets, Dr. DeSalvo used to teach his fellows “Dr. Dan’s Rule of 7-5 for Pediatrics.”

      • A1c <7.5%

      • Average glucose <175 mg/dl

      • Less than 7.5% <70 mg/dl

      • Less than 0.75% <54 mg/dl

      • Bonus: Over 75% time in range (70-180 mg/dl)

  • Dr. DeSalvo preaches the “Three As” for CGM alerts, saying they should be: Actionable; Avoid alert fatigue; and Adjustable. When starting a new pediatric patient on CGM—by the way, >70% of commercially-insured patients in his clinic are on CGM within 3 months of diagnosis and Medicaid numbers should soon climb—he just sets the low alert, generally around 70 mg/dl. He sets the high alert at a subsequent meeting, possibly at 300 or 350 mg/dl depending on the baseline level of glycemic control. Over time, he tightens the alerts to make them more actionable. Finally, he adds fall and (in some cases) rise alerts. We love this stepwise approach!

  • Dr. DeSalvo’s 30-60-90 rule for adjusting insulin off Dexcom CGM arrows. We’ll let the figure do the talking, but want to point out that the Diabetes Spectrum paper it came from (published February 2020; co-written by Dr. DeSalvo and Dr. Sarah Corathers) is a fantastic read about therapeutic inertia in pediatric diabetes. This paper is also the first time we were exposed to “Vision Zero,” Dr. DeSalvo’s “personal statement” which asserts an ethical refusal to accept diabetes related complications.


  • Dr. DeSalvo provided a number of clinical pearls to enhance patient success with hybrid closed loop:

    • Know how the device works! He recommended following Dr. Laurel Messer et al.’s CARES paradigm.

    • Set expectations—help the user understand that it might take a little time for them to achieve in-range glucose values.

    • Make sure pump settings are accurate prior to initiating closed loop

    • Hybrid closed loop systems can usually tolerate a higher ICR (insulin:carb ratio) to dampen post-meal spikes (and the basal rate will automatically attenuate on the back end to prevent hypoglycemia).

    • Sensitivity/correction factors may need to be tightened by 10%-20% across the board. For example, he has found that he often needs to decrease the five-hour active insulin time on Tandem’s Control-IQ in his patients to allow for better control.

    • Offer users tips for success. E.g., (1) Bolus before meals using the bolus calculator and (2) you may need fewer carbs to treat lows. Instead of the rule of 15s, Dr. DeSalvo recommends the rule of 4, 8, or 12 grams of carb based on glucose, insulin on board, and trend arrow.

  • Dr. DeSalvo and peers recently published a review on the medical and psychological considerations for carb-restricted diets in youth with type 1. The paper concluded that there is limited data to suggest improved glycemic control with such a diet, which may contribute to positive psychological outcomes. However, there are also medical (dyslipidemia, poor growth) and psychological (disordered eating, burnout, “treat insecurity”) concerns with this approach. Dr. DeSalvo did not advocate for a carb-restricted diet, but did recommend against “carb stacking” (continuous snacking; see Seckold et al. 2019) as it leads to persistent hyperglycemia and endorsed whole foods, less refined carbs, lean proteins, and unsaturated fats. “Telling patients they can eat whatever they want and cover the carbs feels good to say, but the outcomes will differ.” Finally, he gave a shoutout to Dr. Carmel Smart of John Hunter Children’s Hospital in Newcastle, Australia, whose center has achieved better A1cs, likely in part due to a focus on nutrition. At ADA 2019, Dr. Smart outlined four key behaviors for best glucose management in pediatric type 1 diabetes: (1) Observing a routine meal pattern; (2) dosing insulin before meals; (3) counting carbs, with a focus on carb quality and estimation (not absolute accuracy); and (4) using CGM to identify and target postprandial spikes.

Drs. Philis-Tsimikas and Huang Debate Value of CGM in T2D; Clear Clinical Utility, but Case for Cost-Effectiveness Needs More Research and to Take the Who, What, and Where into Account

Drs. Athena Philis-Tsimikas (Scripps Research Translational Institute) and Elbert Huang (University of Chicago) debated the value of CGM in type 2 diabetes. Dr. Philis-Tsimikas bolstered the “pro” argument while Dr. Huang—admitting that he was assigned the more challenging side to defend—argued the “con.” Even as Dr. Huang made a strong economic-based case against the use of CGM in type 2 diabetes, he was primarily arguing that CGM shouldn’t be worn by all type 2s all of the time: “Should it be used for the whole population or are there subpopulations or moments for using CGM? As a technology for the entire type 2 diabetes population, the answer is probably that, no it’s not valuable to be deployed widely.” Both speakers were intrigued by the concept of “rental” (intermittent) CGM use and noted that the expected decline in cost in the coming years will tip the scale in the direction of value.

  • Wearing her clinician hat, Dr. Philis-Tsimikas wondered aloud: “Do we have to wait until someone who has type 2 diabetes for many years is on a complex medical insulin regimen and potentially has significant complications before providing them with a CGM to manage their disease?” She cited a handful of the early evidence of CGM benefit in non-MDI/CSII-treated type 2 diabetes: Erhardt et al (JDST 2011); Vigersky et al. (Diabetes Care 2012); Yoo et al. (Diabetes Res Clin Prac 2008); Lensing et al. (Diabetes Spectrum 2019); Allen et al. (Diabetes Res Clin Prac 2008); Cox et al. (Diabetes Care 2016); and Bailey et al. (DT&T 2016). These studies are all encouraging and range from prediabetes to basal-insulin-treated diabetes, but we’re in the early innings of building the evidence base for CGM in this population as many of them were small and short-term. Still, there is certainly enough evidence that both patients and providers can use CGM to improve clinical and behavioral outcomes to warrant further investigation. Dr. Philis-Tsimikas proposed that the value of can be improved (i.e., made more cost efficient) by integrating it into DSMES, leveraging AI tools to aid in data interpretation, using it intermittently and integrating with coaching, and by reducing/removing medications due to improvements from CGM.

  • Dr. Huang wore his economist hat (he’s also a physician), pointing out that not everything in health care can be “valuable” and there are budget constraints in the US health system. How then, given budget constraints, do we decide between spending money on CGM and spending money on other cost-effective interventions? While that question remains in the air, he did propose that CGM could be more valuable if it were used intermittently, the price were to decline, it were incorporated into the DPP without increasing programmatic costs (we’re not sure how this could work), improved glucose control while reducing need for medicines, and—most provocative—if we could show that it produces clinical benefits beyond glucose control, such as allowing people to social distance during the COVID-19 pandemic. We love this last idea of identifying positive externalities of CGM use and presenting them as opportunities for reducing cost! Dr. Huang also made the case that CGM use in type 2 diabetes hospitalizations and the long-term care setting is likely cost-effective, though that has not yet been studied.

    • The US is “famous” for spending >$10,000 per capita on healthcare annually (the highest in the world), for healthcare share of GDP that approaches 18%, for high obesity/diabetes rates and subsequent costs, and for high prices of services and products (insulin, for example). In this setting—exacerbated by COVID-19 and the likely coming depression—Dr. Huang explored whether CGM in type 2 diabetes is where we should be spending our marginal dollar. In type 1 diabetes, a lifetime of Dexcom CGM use was determined to be cost-effective in the DIaMonD trial, with an incremental cost-effectiveness ratio (ICER) of ~$98,000/quality-adjusted life-year (QALY). CGM is costly, but it’s A1c reductions are modeled to lower incidence of amputations, stroke, heart failure, renal disease, and other costly complications, resulting in a substantial QALY increase of ~0.5 years. Notably, with real-world use (e.g., extended sensor wear), ICER was reduced to ~$34,000/QALY—this shows that ICER for CGM in type 1 diabetes is highly sensitive to cost. In cost-effectiveness for type 2 diabetes, Dr. Huang pointed to two studies that have “wildly” different results: (1) Garcia-Lorenzo et al (J eval clin pract 2018) calculated an ICER of $198,453/QALY based on a meta-analysis of 5 studies; and (2) Fonda et al. (JDST 2016) calculated an ICER of $13,030/QALY. [Editor’s note: The cost of CGM use in Fonda et al. was significantly lower—$631.72—because it was used intermittently over the span of 3 months.] Fonda et al. point to the possibility of very cost-effective CGM, but there are many cost-effective interventions in diabetes, e.g., lifestyle, statin use for secondary prevention of CVD. How do we decide where to spend the marginal dollar? “Should a patient spend money on CGM over fresh food or medications?”, Dr. Huang posited. He also put forth that more data might not be better for mental health, and people might not share their data with their doctor, both of which could decrease the utility of CGM.

The “Frontline of Diabetes Care”: Dr. Richard Bergenstal Demonstrates Practical CGM Use in Type 2 and Hints at Further Uptake with COVID-19 Telehealth

“Is there anyone more grounding than Dr. Richard Bergenstal?” is a question we often ask ourselves while watching presentations from the Executive Director of Park Nicolett’s International Diabetes Center – today’s fantastic session on practical ways to personalize and optimize care using CGM was no exception. During the session, which came as part of Abbott’s industry symposium entitled “Translating Clinical Evidence for Sensor-Based Glucose Monitoring and Technological Innovations to the Front Line of Diabetes Practice,” Dr. Bergenstal treated us to one of his favorite rules-of-thumb: treat patients by both “thinking fast and slow” – “Fast” referring to in-the-moment therapeutic or behavioral changes based on real-time CGM numbers or trends, and “slow” meaning deliberative analyses of AGPs (many may also know the famous and Nobel-prize winning book by this name). To exemplify “slow” thinking, Dr. Bergenstal walked the virtual audience through a number of clinical examples, in which CGM led to practical, actionable changes in care.

  • For Jean, a 72-year-old woman with type 2, Dr. Bergenstal began by recommending HCPs ask the patient themselves to look at their ambulatory glucose profile (AGP) and self-identify areas that could be improved – although seemingly obvious, we found this tip to be quite useful in a time when both HCPs and patients may be rushed through in-person appointments or still trying to familiarize themselves with telehealth platforms. In Jean’s case, Dr. Bergenstal honed-in on her history of heart disease, and with a time in range of just 65%, recommended a long-acting GLP-1 to provide further cardio-protection and glucose management. Removal of Jean’s sulfonylurea further improved care by reducing time below 70 mg/dl.


  • Next, a man with type 2 diabetes exhibited what Dr. Bergenstal referred to as a “classic stair-step” AGP, in which glucose levels rise after each meal. The patient had previously tried and failed to tolerate a GLP-1 and had already worked up to 70U of insulin glargine at bedtime, with only 51% time in range and a worrisome amount of time in hypoglycemia . Dr. Bergenstal noted that he would recommend starting meal-time insulin, subsequently reducing glargine, and dropping the sulfonylurea. Dr. Bergenstal surmised that even one dose of prandial insulin at breakfast could improve the entire day’s outlook by preventing time above 180 mg/dL straight away.

  • In terms of “fast thinking,” Dr. Bergenstal shared a classic example of using trend arrows to better calculate insulin dose at mealtime. Patients can require vastly different doses depending on current glucose trends. Here, a patient with an “up” trend arrow, would need to add 3.5 units, while the same patient with a “down” arrow would need to reduce by 3.5 units, despite using the same insulin-to-carb ratio and correction factor calculation in each example.

  • Overall, Dr. Bergenstal’s very easy-to-understand clinical applications in patients with type 2 further highlight Abbott’s commitment to expanding into this patient population. During the company’s 1Q20 update, we estimated ~one million Freestyle Libre users with type 2, demonstrating Abbott’s undeniable headway into the field. Of course, Abbott’s low-cost CGM, in addition to its accessibility through the pharmacy, make it a very appealing option for payers, patients, and providers alike.

  • Closing out his presentation, Dr. Bergenstal emphasized that the process of slowly working with patients through their AGPs “visit after visit, phone call after phone call” does seem to be translating well to telehealth during current stay-at-home advisories. As we imagine a healthcare system “after” COVID-19, we wonder if HCPs who have had to rely on CGM data during the pandemic will continue to prioritize the technology even when in-person visits are more of a possibility, and the CMS likely rolls back temporary exceptions for the treatment.

Auf der Bult’s Prof. Thomas Danne opened a debate with University of Florida’s Dr. William Winter about the value of time in range vs. A1c by establishing a true north for diabetes medicine: “Who would’ve thought 10 to 15 years ago, that even on the most remote areas of this globe, that people would have cellphones allowing them to speak to each other without any lines. I’m pretty sure that if we as a community push for the availability of CGM, then very soon, the discussion about A1c will quiet down considerably.” Meanwhile, Dr. Winter laid out a very cogent case for the utility of A1c: It’s useful for screening for diabetes, diagnosing diabetes, and assessing long-term glycemic control; it’s easily accessible for almost everyone; and it’s a robust and reproducible assay. He acknowledged, however, myriad limitations of A1c: It’s a weighted average that’s biased toward the last month; it is not a good measure of hypoglycemia or variability; it doesn’t instruct clinicians or patients how improved glycemic control can be achieved; a desirable A1c does not equate to zero risk of complications; and various analytical or medical conditions can produce biased results. He also conceded that CGM-derived metrics (among others like DKA) are necessary to track in order to achieve the end game of reduced A1c and twice mentioned that there’s room for time in range and A1c to share the stage as both are valuable. His chief reservations against adopting time in range all-in, well-summarized in the following quote, were cost/access and lack of evidence: “It could be that time in range or other parameters with CGM do predict complications with equal power as A1c and time in range can tell you about highs and lows. But from what I found, the challenges to CGM are cost, patient acceptability, and insurance, where A1c, for better or worse, is really robust and accessible to all patients. What about our patients who aren’t on CGM? Maybe all of our patients should be on CGM.” Given the current CGM reimbursement constraints, A1c may still be the primary glycemic metric, but we appreciated Dr. Danne’s point that CGM will soon be much cheaper and much more widely accessible. Regarding Dr. Winter’s argument on a lack of evidence underlying time in range, Dr. Danne emphasized that a DCCT reboot with CGM is likely not feasible but both speakers pointed to Dr. Roy Beck’s DCCT post-hoc analysis as promising. As discussed at a diaTribe time in range coalition meeting, there are other relatively untapped (non-RCT) paths toward validating time in range. In the meantime, Prof. Danne explained how time in range is “clearly preferred” as a patient-centered outcome because it empowers patients and their families to monitor diabetes closer and make more tweaks on the fly.

DKA Rates Among PWDs on Pump/MDI Cut in Half Following FreeStyle Libre Initiation in Large French Observational Study

Bichat Hospital’s Dr. Ronan Roussel presented compelling data from a nationwide cohort study (n=74,076 people with diabetes), showing that DKA rates were cut in half in the year following vs. preceding initiation of FreeStyle Libre. The analysis was performed using the SNDS database (the French reimbursement claims database), including type 1s and 2s on pump therapy or ≥3 injections/day (French reimbursement criteria) who had at least a year of follow-up. This cohort is made of 45% type 1s and 55% type 2s. As seen in the chart below, hospitalization rates fell by 52% for type 1 and 47% for type 2s (albeit from a lower base). Improvements were observed in both pumpers and MDIers. And, while the improvement was seen across baseline strip usage, the most marked reduction was in the group who was registered as using zero strips per day. We’re noticing a greater push to generate evidence that baseline strip usage has little to no bearing on outcomes with CGM (see 875-P)—Medicare requires records of 4+ daily fingersticks before it will cover.

  • Remarkably, pharmacy claims show that ~1/4 of type 2s and type 1s on pump/MDI are using zero strips per day. Only roughly have are using at least 4 strips per day. It’s possible that they are obtaining strips through some other method (e.g., subscription bundle, black/gray market), but on its face, it is very striking that people who are dosing insulin multiple times per day are performing fingersticks so infrequently. Then again, this is one of the powerful forces underlying CGM adoption in this population.

  • Belgium has also reported very strong results from nationwide reimbursement of FreeStyle Libre. Their 1-year follow-up report documents higher treatment satisfaction, less severe hypoglycemia (including coma), maintained A1c, and less work absenteeism.

Cleveland Clinic’s CGM Shared Medical Appointments Reduce A1c by ~0.8% (Baseline: 8.4%-9.1%); New DATAA (Download, Assess, Time in Range, Ares to Improve, Action Plan) Model for CGM Interpretation

The Cleveland Clinic’s Dr. Diana Isaacs outlined her practice’s approach to shared medical appointments with CGM, based on a group-based care delivery model that encourages peer support, provider oversight, and discussion when interpreting CGM results. The shared medical appointments consist of two parts with 4-6 patients in up to ten classes per month. Diabetes care and education specialists (DCESs, f.k.a. CDEs), nurses, dietitians, and pharmacists are all involved in promoting conversation and discussion around glucose patterns and trends, identifying behavioral health changes, and making medication dose adjustments. In the first part of the appointment, which involves 60 minutes of class times, Dr. Isaacs shared that the “do’s and don’ts” of CGM are discussed. Discussions with patients are centered around how to wear the device and maintain in addition to logging food, activity, and medications. Patients collaborate to set blood glucose range goals, identify ideal system alerts, and determine ways of avoiding hypoglycemia and hyperglycemia. In part two of the appointment, which can go up to 90 minutes, patients download devices and remove sensors while also discussing helpful (“Bright Spots”) and harmful (“Landmines”) in their ambulatory glucose profile. Dr. Isaacs shared that this phase of the appointment is where the majority of the learning curve is established as patients assess how expectations of their CGM were met, determine how specific behavioral patterns correlates to blood glucose trends, how to interpret metrics (e.g., time in range, %CV, time <70 mg/dl). During these sessions, significant focus is placed on the day during which time in range is the highest to positively reinforce activities and behavior which appear to have the highest impact on blood sugar. The valued peer support CGM shared medical appointments provide at the Cleveland Clinic is simultaneously rooted in positive clinical outcomes. While the specific timeframe was not mentioned, both types 1 (n=45; baseline A1c: 9.1%) and 2 patients (n=119; A1c: 8.6%) on average have seen 0.8% reductions in their A1c levels. Qualitative data also illustrates that participants touch on a variety of diabetes care practices during these discussions. Among a subset of the same participants (n=132), 57% reported changing nutrition goals,24% self-management behaviors, 23% medication adjustments, and 20% on increasing physical activity.

  • In addition to CGM SMAs, Dr. Isaacs explained “DATAA (Download, Assess, Time in Range, Ares to Improve, Action Plan),” a new model currently in press for publication on how to most effectively review CGM data. As a whole, the five-step model encourages HCPs to help patients understand specific metrics, components of the ambulatory glucose profile, time in range, opportunities for improvement using self-care practices, and a concrete action plan that patients actively implement. Dr. Isaacs emphasized that the overarching goal of this model is to put a positive spin on data, highlighting that the numbers exist not for judgement but for growth and learning.

  • To illustrate this in action, Dr. Isaacs presented the case of Lisa, a 68-year old type 2 patient (A1c: 12.3%), who had a time in range of 1% and regular episodes of hypoglycemia. During a 1:1 appointment with Lisa following CGM onboarding, several clear patterns emerged through the DATAA model. For example, while “assessing safety,” Dr. Isaacs learned that Lisa had stopped using metformin because of her hypoglycemic events and would also go low normally after a single 30-unit injection of insulin lispro despite a full meal. During the “areas to improve” phase of the model, Lisa mentioned that she had actually stopped monitoring her glucose at home because she did not enjoy seeing her numbers and that she felt “like a failure.” To best assist her, Dr. Isaacs provided a range of resources as part of a new action plan, including counseling on metformin, referral to a diabetes care and education specialist, decreasing lispro doses to 15 units with meals, and advising Lisa to check her glucose levels at least 4x/day. Six months later, Lisa had decreased her A1c to 7.2% and improved her time in range up to 72%, had rare episodes of hypoglycemia, and reduced her BMI from 34 kg/m2 to 32 kg/m2. She continued to meet once monthly with her CDCES and to increase physical activity, joined silver-sneakers and played pickle ball 4 days/week. 

Flash CGM Scanning Frequency Indicates that Time Until Performing a Scan After Dropping <54 mg/dl Might be Strongest Predictor for Impaired Awareness Assessment

Dr. Othmar Moser (Medical University of Graz) presented some of the first data investigating whether type 1 patients with impaired awareness of hypoglycemia and flash glucose monitoring demonstrated different “scan” behavior compared to those with regular awareness of hypoglycemia. Participants (n=92; baseline A1c: 7.3%) who had used flash CGM for at least three months along with those with Gold-, Clarke-, and Pedersen-Bjergaard scale scores indicating impaired awareness were included. Notably, the distribution of participants with normal and impaired awareness of hypoglycemia varied by scale used (see table below). Using Pederson-Bjergaard scores, the split of participants between impaired awareness and non-impaired awareness was exactly 50/50. 

  • GOLD Scale: When comparing data on participants classified via the GOLD Scale (n=18 impaired; n=74 non-impaired), significant differences were observed in scan time for those with level 1 hypoglycemia (54-69 mg/dl) and nighttime level 2 hypoglycemia (<54 mg/dl). Participants with impaired awareness took 78 minutes to perform a scan after reaching a hypoglycemia compared to 63 minutes for those with impaired awareness for daytime level 1 hypoglycemia. Similarly, those with impaired awareness took 140 minutes to make a scan relative to 96 minutes.

  • Clarke Scale: No statistical differences were seen for any hypoglycemia levels, night or day.

  • Pedersen-Bjergaard Scale: Statistical differences (n=46 impaired; n=46 non-impaired) were observed for level 1 hypoglycemia, nighttime level 1 hypoglycemia, and nighttime level 2 hypoglycemia. Participants with impaired awareness respectively took 76 minutes, 132 minutes, and 134 minutes compared to 54, 89, and 80 minutes for those with regular awareness.

  • Receiver operating characteristic (ROC) curve analysis for the time until performing a scan after reaching level 2 nocturnal hypoglycemia was done and resulted in an area under the curve of 0.79 (p<0.0001) along with a sensitivity and specificity of 73% each. While technical explanations were not provided, Dr. Moser mentioned that if an episode of level 2 hypoglycemia (<54 mg/dl) occurs at night and a patient performed a scan after 135 minutes, then this was sufficient to diagnose the individual with impaired awareness of hypoglycemia.

Minimally-Invasive Biolinq CGM Early Feasibility Studies (n=15 and n=10): Strong Correlation with YSI Values Through 7-Days; Clinical Studies “Later in 2020”

In a poster, San Diego-based CGM company Biolinq shared some of the first data on its minimally-invasive CGM (69-LB). The company ran two small, feasibility studies: the first was a 2-day study with 15 subjects (5 without diabetes, 10 with diabetes) and the second was a 5-7 day study with 10 subjects (all with diabetes). Types of diabetes were not specified. In addition to the small study sizes, most of the data was shared graphically, making it difficult to draw meaningful conclusions. Still, the readings from the Biolinq CGM sensor showed strong correlations with YSI (R2=0.9938). Performance of the sensor was relatively consistent across three different insertion sites (right and left forearm and left upper arm) and maintained through seven-days of wear. Of note, Biolinq is preparing a “second-generation version of its microarray CGM for clinical studies later in 2020.”

  • Biolinq’s CGM sensor is small and minimally invasive, using an array of microneedles inserted into the skin. The device is also designed to be low-cost. We got a look at the sensor at JDRF’s Mission Summit (Biolinq has investment from JDRF’s T1D Fund) and the device was very attractive with a slim form factor (about the size of a quarter and not much thicker). Biolinq is aiming to get their CGM cleared as an iCGM.

German “SPECTRUM” CGM Training Program Delivers A1c Reduction of ~0.1% from Baseline 7.7% and Improves Patient-Reported Outcomes

Dr. Guido Freckmann (Institut für Diabetes-Technologie Forschung) presented on SPECTRUM, a six-month manufacturer-independent training and treatment program for real-time CGM, available to all age groups. The program’s training modules consist of educational materials which teach users about the ins and outs of CGM, ranging from how to set alarms and calibrate the system to interpreting data and making treatment decisions. Secondary measures including A1c improvements and severe hypoglycemic events were also collected. Following an initial visit, where participants (n=120 type 1s; baseline A1c: 7.7%) performed a baseline knowledge check test using the CGM-Profi-Check exam, patients spent two months learning about CGM through six training modules, each 90 minutes. During a second in-person visit, patients were given the same knowledge test provided at baseline to assess changes but also a test assessing their proficiency in using the device and their satisfaction with the device. Six months following baseline, patients were given the same tests again to assess changes in scores in addition A1c and severe hypoglycemia checks.

  • Based on data from the Profi-Check exam, the metric used to assess CGM knowledge, participants increased scores from 21.2 to 30.4 within two months following the first training, representing a 43% improvement. Notably, these improvements stayed consistent after six months relative to baseline. While Dr. Freckmann did not delve into the specifics of what constitutes “knowledge,” these indications presumably involve metric interpretations (e.g. time in range, %CV, time <70 mg/dl), performing insertions/removals, and downloading data.

  • Secondary endpoints including training satisfaction, CGM handling, satisfaction with the CGM, and acceptance of the system were assessed. Notably, over 95% of participants report some satisfaction with the program, with over 60% indicating “very good.” Measures of satisfaction were pooled through assessing if the program was understandable, whether or not important questions were answered, and if the individual felt prepared after the training to use CGM at home. On the “handling test,” which assessed patient ability to setting the time/date, recharge receiver, change alarm settings, turn off vibrations, performing analysis, and calibrate the system, over 80% of participants at the second visit (2 months) and third visit (6 months) reported success on these measures without any external help. Lastly, with respect to satisfaction with the CGM system itself, which compiled information on personal safety, comfortability, system connection, trust, alarm fatigue, and information accessibility, ~50% of participants reported “high satisfaction.”

  • Among glycemic outcomes, overall A1c decreased from baseline of 7.7% to 7.6% after six months. Before training, two out of 120 subjects had reported three severe hypoglycemic events, but after six-month training, only one event was reported out of 108 subjects still remaining. Notably, the subject with initially two hypoglycemic events dropped out during their study, and of the remaining 108 who had completed, the number of hypoglycemic events did not change. Overall, Dr. Freckmann encouraged to see the program has having an effect on A1c without a statistically significant change in hypoglycemic events.





Glucose Monitoring Posters



Details + Takeaways

Clinical Cost Offset Analysis Comparing Real-Time CGM (RTCGM) with SMBG Based on the COMISAIR 3-Year Follow-Up Study

An Soupal, John J. Isitt, George Grunberger, Martin Prazny, Christopher Parkin, Michael E. Minshall, Peter M. Lynch

  • Used COMISAIR 3-year study results to assess cost offsets due to A1c reductions and avoided hospitalizations for rtCGM vs. SMBG in MDI and CSII type 1s

  • For both MDI and CSII groups, using rtCGM results in: (i) A1c reductions and (ii) reduced hypo and DKA hospitalizations that offset significant healthcare costs

  • For MDI patients, the total incremental cost offsets (read: savings) of rtCGM over SMBG was $5,777-$8,549/person

  • For CSII patients, the total incremental cost offsets of rtCGM over SMBG was $2,732-$4,753/person

Use of Real-Time CGM Is Associated with Fewer Hospitalizations Compared with SMBG in the Insulin-Treated Medicare Population

Gary Puckrein, David G. Marrero, Christopher Parkin, Gregory J. Norman, Liou Xu, Peter M. Lynch, Bruce T. Taylor

  • 12-month, retrospective analysis that used CMS data to assess impact of CGM use (vs. SMBG use) in patients using insulin with a record of acquiring a CGM in the first 6 months of their CMS coverage

  • Lower percentage of Black CGM users vs. Black SMBG users (2.9% vs. 9.4%)

  • Higher percentage of SMBG users with comorbidity risk vs. CGM users (44.2% vs. 33.7%)

  • Average per patient rates of in-patient hospitalizations were significantly lower (p=0.04) for CGM users than for SMBG users (34% vs. 29%) 

  • CGM users’ all-cause in-patient hospitalization costs (~$2,400/6 months) were significantly lower than that of SMBG users (~$3,600/6 months) and slightly decreased over time (p=0.06)

HbA1c Reduction Associated with a FreeStyle Libre System in People with Type 2 Diabetes Not on Bolus Insulin Therapy

Eugene Wright, Jr., Matthew S.D. Kerr, Ignacio J. Reyes, Yelena Nabutovsky, Eden Miller

  • Real-world, retrospective, observational study to evaluate changes in A1c after FreeStyle Libre prescription for non-MDI type 2’s using IBM Explorys database (n=1,034, 51% male)

  • 1.48% A1c reduction after FreeStyle Libre prescription (p<0.001)

  • Baseline A1c impacts degree of A1c reduction: 8-10% A1c’s had 0.53% reductions while those with A1c’s >12% saw a 3.70% reduction

  • Insulin-users at baseline had a lower reduction in A1c compared to non-insulin users (-1.15% vs. -1.62%)

Daily Tracking of Hemoglobin A1C through Personalized Glycation Model

Chiara Fabris, Roy Beck, Boris Kovatchev

  • Uses DIaMonD type 1 and type 2 datasets to develop an intermediate metric which reconciles TIR and A1c to estimate A1c

  • Correlation between estimated A1c and actual A1c was 0.91 (type 1) and 0.86 (type 2), significantly better than a direct correlation between TIR and A1c

  • Percent of readings within 5% of laboratory A1c was 74.5% and 72.6% for type 1 and type 2 respectively

  • Percent of readings within 15% of lab A1c was 100% and 99% for type 1 and type 2 respectively

Including Continuous Glucose Monitoring (CGM) to Provide Personalized Glycemic Profiles as Part of a Pilot Worksite Health Screening

Ilene J. Klein, Margaret Crawford, Gregory J. Norman, Kazanna C. Hames, Jordan I. Justus, Sabrina Treadwell, Tiffany M. Combee, Susan Kinzler, John Welsh, Matthew Johnson

  • Dexcom & Healthstat blinded study to assess the feasibility and acceptability of worksite diabetes CGM screening and to determine rate of undiagnosed and under-diagnosed diabetes (n=176, ~35 years, ~79% female, >75% White)

  • 42% of the population had no diabetes, 20% pre-diabetes, 35% type 2, 3% type 1

  • Point-of-care A1c identified 7 participants with undiagnosed type 2 and 24 participants with undiagnosed prediabetes

  • 85% of participants provided at least 4 days of CGM data

  • 100% of respondents would recommend worksite CGM health screening to family and colleagues

FreeStyle Libre System Use Associated with Reduction in Acute Diabetes Events and All-Cause Hospitalizations in Patients with Type 2 Diabetes without Bolus Insulin

Eden Miller, Matthew S.D. Kerr, Gregory J. Roberts, Diana Souto, Yelena Nabutovsky, Eugene Wright, Jr

  • Funded by Abbott, retrospective observational analysis evaluates effects of acquiring FreeStyle Libre flash CGM on acute diabetes-related and all-cause inpatient hospitalizations for type 2s not on intensive insulin therapy (n=7,167, 53.3 years, 51.5% male)

  • Acute diabetes event rates decreased 30% from 0.71 to 0.52 events/patient-year over 180 period 

  • 13% reduction in all-cause hospitalizations from 0.180 to 0.161 events/pt-yr

Technology Use by Age and Region in Adults with Type 1 Diabetes (T1D) in the SAGE Study

Steven Edelman, Daniela Bruttomesso, Kelly L. Close, Andre G. Vianna, Felipe Lauand, Sr., Sandrine Brette, Eric Renard

  • Study of Adults’ GlycEmia (SAGE) in T1D, a multinational, cross-sectional, observational study outside the U.S., assessed the diabetes technology use in Asia, Eastern Europe, Western Europe, Latin American, and the Middle East in several age groups with type 1 (n=3,858)

  • Of technologies included in questionnaire, finger-stick BGMs were most frequently used (92.0%)

  • Overall, CGMs were used by only 23.3% of people; highest use in Western Europe (46.4%), lowest in Middle East (2.5%)

  • Insulin pumps used by 19.5% of people; highest in Western Europe (42.3%)

  • Use of blood ketone meters (11.1%) and use of dosing support apps (4.6%) were overall low even though in most cases, an app was recommended by a provider (89.2%)

A1C Reductions and Improved Patient-Reported Outcomes following CGM Initiation in Insulin-Managed T2D

Adam Noar, John Welsh, David A. Price

  • Real-world data from 39 patients with type 2 initiating CGM with the Dexcom G6 sensor was collected using validated surveys at 12 weeks following CGM initiation

  • Patients completed the Diabetes Distress Scale, the hypoglycemic attitudes and behavior scale, and the diabetes quality of life short inventory

  • 12-week data showed a mean A1c reduction of 1.5%, diabetes distress scores improved by 31%, hypoglycemic attitudes and behavior scores improved by 22%, and diabetes quality of life scores improved by 24%

  • 95% participants found the G6 system satisfactory or very satisfactory and 85% of participants found the system easy to use

Utilization of Continuous Glucose Monitoring (CGM) and Its Impact on the Care of Veterans

Gayane Barsamyan, Alexis Da Silva, Loren Whyte, Morolake Amole, Hans Ghayee, Julio A. Leey

  • Retrospective chart review of VA hospital patients (A1c >7%) on insulin therapy who received blinded FreeStyle Libre Pro (n=208, 91% male, 86% type 2, mean A1c 8.4%)

  • With FreeStyle Libre Pro, time in range was 45%, 26% of time was spent above range, and >5% time in hypoglycemia was seen in 32% of patients

  • 65% of patients had a change in treatment regimen after CGM use

  • For those who had insulin treatment changes after CGM use, A1c significantly decreased from 8.2% to 7.5% 

Using CGM as Feedback Tool in Diabetes Mellitus Patients Management: A Retrospective Real-World Evidence Study Conducted in India

Mudit Sabharwal, Srivani Palukuri, Saurav Deka, Garima Chanana, Rohit Kumar

  • Retrospective, 2-week CGM reports were collected and used to change diabetes medications after first week of CGM (India, n=214, 45% female, 74% <10 years of diabetes)

  • After the first week, patients who had frequent hyperglycemia saw significant improvement with medication changes (-9% time >180 mg/dl) although there were episodes of hypo during the night

  • Estimates of A1c after the 2 weeks suggest an increase in proportion of participants w/A1c <7% compared to baseline (75% vs. 30%)

  • Glycemic variability (%CV) reduced from 47% to 38% in 14 days

  • Suggests that clinicians could use CGM after the first week to make quick improvements to diabetes treatment rather than waiting full 2 weeks 

A1C Reductions and Improved Patient-Reported Outcomes following CGM Initiation in Insulin-Managed T2D

Adam Noar, John Welsh, David A. Price

  • Preliminary results from ongoing Dexcom study following cohort of type 2 patients treated with intensive insulin therapy for 1st 12 weeks on Dexcom G6 (n=39)

  • Significant reduction of A1c after 12 weeks of using Dexcom G6 from 8.7% to 7.1%; A1c values decreased for 85% of participants

  • Mean percentage of time in range almost met the consensus target (70-180 mg/dl 69%); time above range (>180 mg/dl, 22%) and below range (<70 mg/dl, 0.9%) met targets

  • Improvements in quality of life scores and high device ratings were consistent across all starting A1c ranges 

Improved HbA1c Estimation Using CGM Data

Joshua Grossman, Andrew T. Ward, David M. Maahs, Priya Prahalad, David Scheinker

  • Accuracy of A1c estimation can be improved compared to standard Glucose Management Indicator (GMI) by including additional patient information and using machine learning

  • Created A1c estimation models via machine learning using CGM, A1c, and demographic data (n=4,212 A1c measurements from 1,182 participants)

  • Initially, best performance from random forest-based model (20% lower estimation error than GMI, p<0.0001)

  • After retraining with prior A1c as a covariate, LASSO-based model performed best w/27% lower estimation error than GMI (p<0.0001); ordinary least squares model had an estimation error rate only 0.3% higher than LASSO

Flash Glucose Monitoring: Effect on Glycemic Control, Hypoglycemia, Diabetes-Related Distress, and Resource Utilization: A Nationwide Study

Harshal Deshmukh, Emma G. Wilmot, Jane Patmore, Christopher Walton, Roselle Herring, Robert E. Ryder, Thozhukat Sathyapalan

  • UK FreeStyle Libre audit in 114 centers to determine FreeStyle Libre’s impact on A1c, hypo awareness, resource utilization, and diabetes distress (n=10,370, 38 years, 52% female, 97% type 1)

  • 0.5% A1c reduction after FreeStyle Libre; for those with, baseline A1c >8.5%, there was a 1.1% A1c reduction

  • 9% of patients reported a reversal of hypo unawareness

  • Possible reductions in hospitalizations due to DKA/hypo (awaiting 1-year follow-up data)

  • Mean diabetes distress scores significantly reduced

Variation of Measured HbA1c from Predicted HbA1c on Freestyle Libre

Jayasri Muthabathula, Lydia Grixti, Tamsin Fletcher-Salt, Ananth U. Nayak

  • Compares FreeStyle Libre A1c estimates to laboratory A1c values (n=144)

  • Significant correlation between measured and predicted A1c values (r=0.86, p<0.001); however, the difference between measured and predicted A1c was 0.2%, and there was a wide range of differences (-3.2% to +1.1%)

  • Confirms that variation in measured and FreeStyle Libre predicted A1c values could be clinically significant for ~1/5 (18.4%) of individuals

Utilization of Continuous Glucose Monitors Is Associated with Reduction in Inpatient and Outpatient Emergency Acute Diabetes Events Regardless of Prior Blood Test Strip Usage

Irl B. Hirsch, Matthew S.D. Kerr, Gregory J. Roberts, Diana Souto, Yelena Nabutovsky, Richard M. Bergenstal

  • Funded by Abbott, retrospective, observational analysis explored if reduction in acute diabetes events after CGM purchase depends on level of prior test strip usage (n=17,003 MDI CGM users)

  • Regardless of test strip usage, users saw an immediate and significant reduction in acute diabetes events 

  • Hazard ratio was 0.50 and 0.57 for low-test strip usage and high-test strip usage respectively

  • Of particular interest because currently Medicare and other payers restrict access to CGM devices in part based on SMBG testing frequency

Real-World Data from U.S. Patients Using a Long-Term Implantable Continuous Glucose Monitoring (CGM) System: Age Effect on Glycemic Control

Katherine Tweden, Samanwoy Ghosh-Dastidar, Andrew D. Dehennis, Francine Kaufman

  • Senseonics study uses Eversense data management system to characterize sensor glucose parameters by age (n=1,656 users, 60% type 1)

  • For all targets, percent of users achieving targets increased with age 

  • Median wear time was >80% for all age groups except 18-24.9 years (60% wear time)

  • 2 oldest cohorts (45-65, >65) met the 7.0% GMI (estimated A1c) target

  • All groups met or almost met the time in hypoglycemia target of <4.0% 

  • Even for the youngest cohort, CGM use resulted in 7.5% predicted A1c and acceptable hypo risk

Positive Impact of Use of Continuous Glucose Monitoring on Glycemic Outcomes in Young Adults with Type 1 Diabetes, in Adult Clinical Setting, Independent of Insulin Administration Method

Elena Toschi, Robert A. Gabbay, Austin Clift, Madeline Bennetti, Astrid Atakov-Castillo

  • EHR data from type 1s ages 18-30 years (n=891)

  • 8% of participants were on sliding scale insulin; 38% on MDI; and 51% on pump

  • CGM use was 32% for those on sliding scale; 46% for MDI users; and 67% for pump users

  • CGM users in each insulin delivery subgroup had lower A1c than non-insulin users

Resolution of Hypoglycemia Using CGM Data

Lisa M. Norlander, Keyuree Satam, Sarah Hanes, Bruce A. Buckingham

  • CGM data from 50 children and adolescents at Northern California diabetes camp; 44 on Dexcom and 6 on Medtronic

  • Campers treated with 10-20 g of carbs depending on weight and severity of hypoglycemia

  • When glucose was <70 mg/dl but >54 mg/dl, 89% of hypoglycemia episodes were resolved within 15 minutes

  • When glucose dropped below 54 mg/dl, less than half of episodes were resolved by 15 minutes

Discrepancies in Key Metrics of Glycemic Control between Retrospective Continuous Glucose Monitoring System and Flash Glucose Monitoring in Adult Patients with Type 1 Diabetes

Yongwen Zhou, Hongrong Deng, Daizhi Yang, Jinhua Yan, Wen Xu, Hongxia Liu, Lu Gan, Sihui Luo, Xueying Zheng, Hua Liang, Bin Yao, Jianping Weng

  • Head-to-head comparison of Abbott’s FreeStyle Libre with Medtronic’s iPro 2 in 27 adult type 1s (31,287 total paired points)

  • Mean glucose, standard deviation, %CV, estimated A1c (GMI), and time in hyperglycemia were statistically identical

  • FreeStyle Libre reported 7% time <70 mg/dl compared to just 4.5% for iPro 2; FreeStyle Libre reported 2.6% time <54 mg/dl compared to just 1.4% time <54 mg/dl

Durability of Continuous Glucose Monitoring (CGM) Use in Young Children, Teens, and Young Adults with Type 1 Diabetes (T1D)

Daniel Desalvo, Lauren Kanapka, Colleen Bauza, Cicilyn Xie, Linda Dimeglio, Lori M. Laffel, Kellee Miller,

  • Use of CGM after one year in the CITY (14-<25 years) and SENCE (2-<8 years) studies

  • CGM use maintained at one-year in SENCE (92% at six months and twelve months); CGM use increased at one-year in CITY (79% at six months and 86% at twelve months)

  • At 12 months, 88% of SENCE and 97% of CITY participants used Dexcom G5 to dose insulin non-adjunctively

  • At 12 months, 54% of SENCE and 55% of CITY participants used remote monitoringfeatures

Comparison of Dexcom G6 CGM with Self-Monitoring Blood Glucose in Young Adults with Type 1 Diabetes: The Millennial Study

Hood Thabit, Womba M. Mubita, Catherine Fullwood, Shazli Azmi, Andrea Urwin, Ian M. Doughty, Lalantha Leelarathna, Joshi Paul-Prabhu


  • 30 young adults and adolescent (16-24 years) type 1s in Manchester, UK 

  • Randomized, crossover trial comparing Dexcom G6 with SMBG; 2 week run-in, 8 weeks after randomization, four weeks washout, and eight weeks after crossover

  • Dexcom G6 showed time in range improvement of 11% (36% vs. 25%); time >180 mg/dl improved from 74% to 62%

  • Trend towards increased time <70 mg/dl with Dexcom G6 (1.4% vs. 0.5%; p=0.06)

  • A1c reduced by 0.5% with CGM vs. a 0.2% increase in SMBG group

Benefit of Continuous Glucose Monitoring (CGM) in Reducing Hemoglobin A1c Is Sustained through 12 Months of Use among Adolescents and Young Adults with Type 1 Diabetes (T1D)

Kellee Miller, Lauren Kanapka, Colleen Bauza, Lori M. Laffel


  • CITY study: 153 participants adolescents and young adults (14-<25 years) with type 1 diabetes

  • RCT comparing Dexcom G5 vs. BGM during study; all participants used Dexcom G6 during 6-month extension

  • CGM use was very high during extension phase (86% in CGM group, 91% in BGM group)

  • BGM group reduced A1c from 8.9% to 8.5% during extension; CGM group reduced A1c from 8.5% to 8.3%

  • BGM group increased time in range from 33& to 38% during extension; CGM group time in range steady from 42% to 41%

  • BGM group reduced time <70 mg/dl from 62 min/day to 30 min/day during extension; CGM group reduced time <70 mg/dl from 46 min to 33 min/day

The Association of British Clinical Diabetologists Audit of FreeStyle Libre (FSL) in Diabetes in United Kingdom: Determinants of Time-in-Target Range

Emma G. Wilmot, Harshal Deshmukh, Jane Patmore, Thozhukat Sathyapalan, Christopher Walton, Roselle Herring, Robert E. Ryder


  • Clinical data collected from 2,191 FreeStyle Libre users at 101 UK hospitals (99% type 1s)

  • Median time in range was 43%; 15% of users met the 70% TIR goal

  • Predictors of greater TIR were older age (beta=0.13), lower baseline A1c (beta=-0.45), and higher scan frequency (beta=0.22)

Higher Glucose Thresholds for Hypoglycemia Alarms on Continuous Glucose Monitoring Systems (CGMs) Are Associated with Less Time in Hypoglycemia in Patients with Impaired Awareness of Hypoglycemia

Yu Kuei Lin, Danielle Groat, Owen Chan, Rodica Pop-Busui, Michael Varner, Simon Fisher


  • Observational study of 81 adult type 1s using CGM, 42 with impaired hypoglycemia awareness (IAH) as measured by a Clarke score

  • For IAH patients, higher hypoglycemia alarm thresholds (cut-off at 73 mg/dl) were associated with 58% less time <70 mg/dl and 75% less time <54 mg/dl

  • Compared to normal hypo awareness patients, IAH patients were older, longer diabetes duration, and higher glucose variability

Effectiveness of FreeStyle Libre Flash Glucose Monitoring System on the Diabetes Distress among Type 1 Diabetes: A Prospective Study

Mohammed Aldawish


  • Prospective study of FreeStyle Libre in 187 young type 1s in Saudi Arabia

  • At baseline, 38% of participants had diabetes distress scores of “a somewhat serious problem,” “a serious problem,” or “a very serious problem”; after 12 weeks with FreeStyle Libre, 24% of participants fell in these groups

  • At baseline, 12 (out of 187) participants saw diabetes as “not a problem”; after 12 weeks with FreeStyle Libre, this doubled to 23 participants

  • A1c dropped from 8.2% to 7.9% after 12 weeks

  • Frequency of fingersticks was 2/day at baseline; with FreeStyle Libre, users scanned ~7 times/day 

Continuous Glucose Monitor Utilization and Adherence in Children with Type 1 Diabetes

Emily L. Montgomery, Kahir S. Jawad, Stephany Eubanks, Kupper A. Wintergerst, Heather M. Rush, Sara Watson

  • Retrospective chart review of pediatric type 1s using CGM at University of Louisville outpatient clinic 2017-2019 (n=253)

  • Median patient used CGM for 20 months before discontinuing; 70% of patients used CGM >1 year

  • CGM was utilized 79% of the time (~5.5 days/week)

  • Every 10% increase in CGM utilization time was correlated with 0.2% reduction in A1c, 7 mg/dl reduction in mean glucose, and 3% higher time in range

Digital Health, Telemedicine, and Decision Support

One Drop to Launch Long-Term (1-6 Month) Outcomes Forecasts For 30-day Average Glucose, Blood Pressure, and Weight “Within the Year”; Overnight Hypo Prediction Model Achieves AUC of 0.82 for CGM Users

This morning, One Drop announced plans to launch long-term outcomes forecasts for “diabetes-related biomarkers” and overnight hypoglycemia predictions for CGM users. The prediction capabilities are expected to launch “within the year” and are part of One Drop’s efforts to shift health management from “reactive to prospective.” The predictions for CGM users will be based on real-time CGM data – the regulatory classification for this “depends on a bunch of things,” so we’d imagine the timeline could change quite a bit. One Drop also told us that the long-term outcomes forecasts may be presented directly to users, but the primary use would be to “provide personalized self-care guidance.”

  • One Drop’s supervised learning models for predicting blood pressure, weight, and 30-day average glucose were significantly more accurate than a “naïve persistence” model (assuming no change over time) on all prediction horizons (38-LB). The study included data from ~55,000 One Drop app users and generated over 200,000 test-set predictions across blood pressure, weight, and average glucose. Root mean square error (RMSE) for systolic blood pressure was 9.4 mmHg on a 1-2 month prediction horizon (~17% better than persistence model), rising to 11.4 mmHg on a 4-6 month prediction horizon (14% better than persistence). Predicted weight RMSE was 2.1 kg on a 1-2 month horizon (6% better than persistence) and 3.9 kg on a 4-6 month horizon (7% better than persistence). RMSE for 30-day average glucose in BGM users was 34 mg/dl on a 1-2 month horizon and 44 mg/dl on a 4-6 month horizon (22% and 18% better than persistence, respectively. Predictions for CGM users were more accurate: RMSE was 14 mg/dl on a 1-2 month horizon (26% better than persistence) and 19 mg/dl (13% better than persistence).

  • Another poster showed One Drop’s model for predicting overnight hypoglycemia (<70 mg/dl) in CGM users achieved an AUC of 0.82 (14-LB). The machine learning-based model was trained on ~360,000 nights of data and tested on ~200,000 nights of data from “over 3,000” One Drop users with CGM. 86% of users in the dataset had type 1 or LADA, 8% had type 2, and 6% were unreported. The model achieved an AUC of 0.82 and appeared well-calibrated (see table below): in users with a predicted probability of hypoglycemia of 90%-100%, the actual frequency of overnight hypoglycemia was 97.5%. One Drop identified certain combinations of glucose variability, activity, food, and heart rate data that drove better predictions – these combinations were available in ~30% of the total dataset and these “high-confidence predictions” achieved an AUC of 0.87.

Users Randomized to Glooko’s MIDS Basal Titration System and Control Group (Paper-Based Titration Tool) Have Similar A1c Reductions at 16 Weeks

In a large prospective RCT, use of Glooko’s FDA-cleared Mobile Insulin Dosing System (MIDS) for basal insulin titration in type 2 diabetes resulted in similar A1c reductions as standard of care (paper titration + enhanced CDCES support) at 16 weeks. The study enrolled 242 adults with type 2 diabetes who were on (89%) or initiating (11%) basal insulin and were not on or planning to start short-acting insulin. Both groups used Tresiba pens (Degludec U-200) and the Degludec Step-Wise titration algorithm which is based on fasting SMBG readings and personalized treatment plans to adjust insulin doses in increments of two units. The difference was the algorithm was configured by a provider and baked into the patient-facing Glooko app (see a deeper dive on this in the bullets), or done manually by the patient on a worksheet, with the guidance of an educator. At week 16, median A1c had improved significantly in both groups (-1.3% in MIDS; -1.2% in control), but the between-group difference was not significant. Similarly, median daily insulin doses increased significantly from baseline in both MIDS (+8 units) and control (+10 units) groups, but there was no between-group difference. While there were unfortunately no CGM outcomes to share (even periodic blinded CGM wear would’ve gone a long way), SMBG outcomes were presented: notably, proportion of readings <70 mg/dl were similar between MIDS and control (0.9% vs. 1.6%), but the MIDS group had significantly higher percent of readings in 70-180 mg/dl (77% vs. 70%) and fewer percent of readings >250 mg/dl (5% vs. 9%). While the study was planned (and failed) to show superiority of MIDS on the primary outcome of A1c reduction, we consider this study to be a major victory for Glooko and the field of digital health. Scripps Health’s Dr. Athena Philis-Tsimikas, the presenter, does as well:We’ve always known that having the help of an educator starting someone on insulin can encourage and allow them to better manage their diabetes. It’s impressive that you could do this equally with a digital tool that didn’t have as much contact with patients and you still see effect on A1c.” Indeed, MIDS is far more scalable and user-friendly than paper-based titration–assuming a base level of technological literacy and access—and there was no sacrifice in quality of outcome. Furthermore, a clinic that uses MIDS will free up its educators to do other, higher-impact work. We are hopeful that this rigorous study will propel MIDS and the handful of other basal titration apps (many also FDA-cleared) to increased adoption.

  • Dr. Philis-Tsimikas provided a glimpse at the clinician and patient interfaces. The clinician interface—which is presumably built right into Glooko—allows the clinician to enter insulin type and starting dose, titration period/end date, SMBG data sufficiency requirements, and to configure fasting target range and adjustment paradigms. On the other end, SMBG readings are synced remotely via the Glooko mobile app, which reminds the patient to take fasting blood glucose readings and inject insulin daily. Every three days, a “Dose Adjustment Check” will determine, based on the clinician-configured plan, whether a dosage change is needed. Both interfaces look simple and intuitive. Last we heard, MIDS was piloting at a small number of US clinics—we wonder if this has since expanded.



Group Lifestyle Intervention for Weight Loss and Glycemic Control Among Adults Delivers 5% Weight Loss at 12-Months, Regardless of In-Person vs. Telephone Program Delivery

A group lifestyle intervention for adults with type 2 diabetes shows a 6-month 5.6% weight-loss and a 12-month 4.6% weight loss, significantly greater reductions than those of Medical Nutrition Therapy (MNT), the current standard of care. Excitingly, almost 50% of lifestyle participants (as compared to 15% for MNT) achieved a 5% weight loss at both 6 and 12 months for both the in-person (IP) and telephone conference call (TCC) versions. It is worth noting that the lifestyle intervention cost considerably more than MNT (MNT: $591, in-person: $1,380, telephone: $1,814). Ms. Linda Delahanty (Massachusetts General Hospital Diabetes Center) shared these results from REAL HEALTH-Diabetes, a large effectiveness trial comparing a lifestyle group program to MNT. In the 2-year program, sessions focus on skills to apply to nutrition, activity, and behavioral topics, using materials adapted from DPP and Look AHEAD. Although DPP, MNT, and Look AHEAD have all shown their efficacy in improving health outcomes, as Ms. Delahanty argues, effectiveness (feasibility, utilization, etc.) – not efficacy – determines real-world outcomes in real-world clinical settings.

  • It is notable that REAL HEALTH-Diabetes found that both the in-person and the telephone-conference-call lifestyle group programs were effective. There were no significant differences in their impacts on weight loss and glycemic control. Not only was the telephone version effective, but also it was equally feasible: for both versions, ~70% of participants attended over 70% of sessions. It is particularly valuable that the telephone version of the program was effective not only given the context of COVID, but also because it allows for greater scalability and increases accessibility. A similar study from Kaiser Permanente published in June also found similar outcomes from DPP delivered in-person and virtually.

  • Previously, Look AHEAD, an 11.5-year randomized multicenter efficacy trial, found that a 19-week group lifestyle intervention improved weight, fitness, and A1c – along with the extensive list below – for people with type 2 diabetes who are overweight. There was no significant improvement in cardiovascular morbidity and mortality attributable to the lifestyle intervention.

Q: In-person and telephone results were fairly similar, so what do you see as the key strengths or weaknesses of in-person vs. telephone? I think that that is a really important question right now in light of our COVID-19 environment, and from what I am hearing, so much care is being delivered via telephone, and I think that this experience will open the floodgates hopefully for more telephonic communication or telehealth.

A: We surveyed participants before the trial started to find out if they had a preference for one of the treatment assignments. Of interest, 24% had no preference, 20% preferred MNT, 43% preferred in-person group, and 14% preferred telephone conference call group. We think that that might be because they had no relative experience to know what that might be like. When we did qualitative interviews of those participants to drill down further why they answered the way they did, we found out that there were four themes. People really no matter what program they’re going for, they’re looking at a way for accountability, they’re looking for something that matches their learning styles, that is convenient to their schedule, and that gives them the kind of support that they need. For some people, MNT served those goals – they wanted personalized attention; they felt that made them more accountable, and for others, the group support served that role. Those people who weren’t sure about telephone conference call were saying things like “I don’t know if I’ll be too distracted and I’ll multitask…” but I really feel like it comes down to they don’t know what they don’t know. At the end, we’ve also polled these participants, and just anecdotally, I think most all of the [telephone-based] participants were surprised at how well they liked the convenience of the [telephone] group and how effective their weight loss results were.

Onduo Virtual Diabetes Clinic Drives Significant Reduction in Diabetes Distress Among Type 2s After 3 Months

The esteemed Dr. William Polonsky (Behavioral Diabetes Institute) presented results from a retrospective pre-post analysis of diabetes distress among type 2 enrolled in Onduo’s Virtual Diabetes Clinic. Results showed significant reductions in overall diabetes distress among participants (n=228) after three months (p<0.001) and saw significant reductions across all four categories of diabetes distress: regimen-related (p<0.001); emotional (p<0.001); interpersonal (p=0.002); and physician-related (p=0.006). Additionally, patients using CGM experienced greater reductions in diabetes distress than non CGM-users, “driven by differential improvement in regimen distress” (it is important to note that a higher percentage of type 2s on CGM completed the follow-up survey). On ADA Day #2, Onduo’s Dr. Ron Dixon presented results showing CGM-using Onduo participants also saw significantly greater A1c reductions compared to BGM-using Onduo participants. Onduo’s Virtual Diabetes Clinic combines supported self-care (i.e. health coaches, connected BGM, and A1c home test kit options), with expert care from DCESs (f.k.a. CDEs) and CGM while also providing patients with the opportunity for live video communication with endocrinologists. Providers are able to manage medication and remotely prescribe and ship CGM devices to all 50 states in the U.S. Inclusion criteria for this study included at least moderate diabetes distress at baseline assessment (Diabetes Distress Scale (DD17) score ≥2). Dr. Polonsky made sure to emphasize the importance of reducing diabetes distress in helping patients achieve health goals and indicated personalized telehealth options such as Onduo Virtual Diabetes Clinic may be an important tool in addressing diabetes distress among the type 2 population.

Algorithm-Guided Basal-Bolus Therapy Shows Six-Month Average Blood Glucose and Time in Range Improvements in Type 2 Patients

Dr. Daniel Hochfellner (Medical University of Graz) highlighted how electronic decision support automating insulin dosing could support type 2 patients in the hospital. Dr. Hochfellner’s team implemented an Electronic Decision Support System (eDSS), a platform providing automatic basal-bolus insulin dosing decisions via an algorithm, among type 2 patients (n=71; baseline A1c: 8.6%) at the Medical University of Graz, and compared glycemic outcomes to standard insulin titration treatment from hospital providers. Beyond algorithmic support, the system also included an electronic blood glucose chart, was fully integrable in the hospital’s electronic medical records system, and provided data access on mobile phone and PC. Data was analyzed six months after eDSS implementation and included 535 treatment days, featuring 274 days with standard care therapy and 261 with algorithm-supported basal-bolus insulin treatment from eDSS.

  • Patients who received algorithm-guided basal-bolus insulin therapy vs. standard care saw significantly improve outcomes with respect to average blood glucose (157 mg/dl vs. 172 mg/dl), total daily insulin dosage (42.2 IU/24 h vs. 33.7 IU/24 h), time in range (70.5% vs. 61.2%), time spent in hypoglycemia (1.6% vs. 1.9%), and time spent in severe hypoglycemia (0.2% vs. 0.6%). Overall, Dr. Hochfellner attributed improvements seen in patients receiving algorithm-guided therapy the algorithm considering insulin administration based on both the meal and blood glucose value at the time instead of only blood glucose data.


Electronic Health Record-Based Real-Time Glycemic “Excursion Event Alerts” at Yale Medical Center Improve Inpatient Time in Range by ~4%

Yale’s Dr. Leigh Bak presented on a fascinating electronic health record-based system alerting clinicians of glycemic excursions in the inpatient setting. Dr. Bak’s team used Yale’s Epic-based EHR system to create glucose “Excursion Event Alerts” (EEAs) which would notify providers when either two blood glucose readings were >300 mg/dl within 48 hours or <70 mg/dl in 24 hours. Alerts were coded to not fire within the first 12 hours of patient admission or for the same patient for the same reason unless specific criteria were met again. When responding to an alert, a provider could respond on the EHR system by starting basal-bolus insulin therapy, ordering an endocrinologist consultation, adjusting other medication, or stating “no action needed.” Additionally, if an alert went to an incorrect provider, the clinician also had the option for selecting “wrong provider” when responding to the notification to avoid alarm and annoyance fatigue. To assess how alerts could create visibility on hypoglycemic and hyperglycemic events, Yale implemented a pilot study in 14 medical wards assessing performance compared to 12 control wards that received no intervention. The study utilized a metric called the “Quality Hyperglycemia Score-2” (QHS2), a scale scored from 0 to 100 developed in 2011 at Yale, comprised of five components: (i) percentage of days patient spends in 70-180 mg/dl range; (ii) percentage of days patient has a blood glucose >300 mg/dl; (iii) percentage of days patient spends with glucose <70 mg/dl; (iv) percentage of days the patient spends with glucose <40 mg/dl; and percentage of hypoglycemic events (<70 mg/dl) where blood glucose is rechecked within 60 minutes and is >70 mg/dl. A higher score is indicative of “stronger” performance (e.g. more days spent in range, fewer hypoglycemic events, less hyperglycemic events).

  • Overall, QHS2 scores dramatically improved from baseline 67.7 to 77.1 (+13.9%) for wards (n=14) that received excursion alerts over the study period of three months. Patient days spent in range (70-180 mg/dl) increased 4.3% (baseline not provided) with no changes in the number days spent in hypoglycemia or severe hypoglycemia. Dr. Bak mentioned a “trend” towards a reduction of days spent in hyperglycemia although data was not provided. Notably, no statistically significant changes in QHS2 scores were found in the control group.

  • During the three months, hospital provider received 1,685 total EEAs, 398 of which were for hypoglycemia and 1,287 for hyperglycemia. Provider responses to both types of EEAs were nearly identical, with 47% of both hypoglycemia and hyperglycemia alerts requiring no response. Dr. Bak explained that with their cost-effectiveness and simplicity, EEAs could improve glycemic control through increasing provider awareness of excursions.

Predicting In-Patient Hypoglycemia with Machine Learning: Model’s Top 10% Highest-Risk Patients Make up Half of All Hypo Events

Dr. Michael Fralick (University of Toronto) presented promising data around using machine learning techniques to predict hypoglycemia events in the hospital. Dr. Fralick began his talk by outlining the need for such a tool: at St. Michael’s Hospital in Toronto, prevalence of severe hypoglycemia was ~10% in critical care patients, ~5% in internal medicine wards, and 2%-10% in surgical wards. Dr. Fralick’s models were trained on data from 2013-2017, tested on data from 2017-2018, and validated using data from 2018-2019. Three machine learning techniques, LASSO regression, gradient boosted trees, and recurrent neural network, were fed with data from medical records (e.g., drugs, orders, MD and nurse notes, patient characteristics, past glucose data, etc.). The dataset included ~8,500 internal medicine inpatients and ~8,000 cardiovascular surgery inpatients with a mean age of 68 years old and 80% prevalence of diagnosed diabetes. The dataset included ~2,400 total hypoglycemic events, which, on average, occurred during day 3 of the hospital stay.

  • All three machine learnings saw comparable performance, with areas under the curve (AUCs) in the 0.75-0.83 range for both hypo (<70 mg/dl) and severe hypo (<54 mg/dl) events. We found Dr. Fralick’s “cumulative gain chart” more intuitive than the AUC measure: as seen in the figure below, successful interventions with the top 10% of highest-risk patients (as predicted by the model) could reduce <70 mg/dl events by ~half.

  • This was a single-center study, limiting its generalizability. With AUCs of ~0.8, the models perform similar to those we’ve seen before – at Hospital Diabetes Meeting in April, Dr. Nestoras Mathioudakis (Johns Hopkins) presented results from his logistic regression model, which achieved an AUC of 0.87 with internal validation and 0.82-0.85 with external validation.

Rural Diabetes Telehealth Intervention Delivers 1.4% A1c Reduction After Six Months among Veterans with High Baseline A1cs

Ms. Elizabeth Kobe, a third-year medical student at Duke University, presented results on Advanced Comprehensive Diabetes Care (ACDC), a six-month novel telehealth intervention providing diabetes care to rural veterans. Despite the promise of telehealth, which has gained prominence since the COVID-19 pandemic (see the diaTribe Foundation's Musings 2020 and Insulet webinar) intensive interventions have rarely been implemented in standard diabetes care due to a lack of trained staff, availability of equipment, limited integration with electronic health records, and poor reimbursement. To address this implementation gap, the Department of Veterans Affairs (VA) has invested in a nationwide network of Home Telehealth (HT) nurses, currently used for telemonitoring of diabetes and other chronic conditions. ACDC leveraged the VA’s HT network and EHR infrastructure to deliver intensive telehealth care to rural veterans with diabetes. Overall, the intervention involved four steps: (i) nurses held 30-minute phone calls every two weeks for six months where they would review glucose data, reconcile medication usage, and analyze medication adherence; (ii) patients received self-management support through eight modules; (iii) a clinical report was compiled; and (iv) a pharmacy manager would review the data and recommend care changes to the HT. In the initial cohort, mean A1c (n=50; baseline: 10.5%) among patients randomized to ACDC (vs. usual care) in the program saw a reduction of 1.0% at six months.

  • Following the pilot, ACDC began its formal implementation phase in 2017, expanding to seven total VA sites across North Carolina, Eastern Colorado, New Mexico, and Montana. Across patients who have enrolled in the program 2017-2019 (n=125; 5 sites), participants have seen a mean A1c of reduction of 1.4% (baseline 9.3%) after six months – notably, these reductions were sustained up to 18 months (-1.1%). Participant engagement has also been satisfactory, with an average of 8-10 of 12 scheduled intervention calls completed per patient at each site. These findings continue to demonstrate how telehealth can improve access to care, promote diabetes self-management, and improve health literacy.

  • Veterans are particularly at-risk for diabetes and we’re excited to see such promising results for this underserved population. About a quarter of veterans in the US have diabetes a rate more than double the national average, and 1.5+ million veterans with type 2 diabetes currently receive diabetes care from the VA. In 2017, the VA estimated annual spend on diabetes to be $1.5 billion, which sounds low to us.

Maryland’s 3-month Telehealth Diabetes Bootcamp for High-Risk Adult Type 2s Costs $1,471/Participant, 1.6% A1c Reduction After Three Months

Dr. Alex Montero (Medstar Georgetown University Hospital) shared that the total cost of care (net cost) associated with their Diabetes Bootcamp was $1,471 per person, which we feel is fairly expensive for a bootcamp program. However, some of that cost comes from increased pharmacy spending (~$460), which is presumably for medicine which participants should have been taking before the intervention. The analysis of Bootcamp outcomes showed that in-patient spending decreased ~$350, which is in line with their previous results indicating decreased hospitalizations with the Diabetes Bootcamp. The Bootcamp is a 3-month telehealth diabetes care management intervention for high-risk (A1c>9%) adults with type 2 diabetes in Maryland. Previously, an RCT validated the efficacy of Diabetes Bootcamp, showing that at 90 days, the intervention arm saw a 3.1% A1c reduction, a 1.6% greater reduction relative to the control group (p<0.001). Given the sizeable A1c reduction, we’d imagine that the Diabetes Bootcamp could deliver long-term savings around reduced complications and improved quality of life; of course, this would require longer-term studies of the program to determine how well A1c reductions are sustained.

  • The original EHR-only evaluation of total cost missed out-of-network events and underestimated the cost of Diabetes Bootcamp relative to Medicaid claim-based analysis. In their previous efficacy validation study, the researchers estimated a potential annual cost savings of $3,086 per Bootcamp patient based on EHR hospitalization data, a drastically different finding than the $1,471 annual net cost per patient found with Medicaid-based analysis. Dr. Montero argues that Medicaid data offers a better cost estimate than does EHR because it includes both in-network and out-of-network costs.

  • Pharmacy spending went up $462.77/participant with the intervention, presumably because patients were taking the meditation that they needed but were previously lacking. Specifically, spending on GLP-1 rose by ~$440 in the DBC intervention group while it only rose ~$80 in the control group. Interestingly, insulin spending increased about equally in both groups.

Project ECHO T1D Tackles Complexity of Diabetes Cases in Primary Care

Dr. Nicolas Cuttriss presented the model and details of Project ECHO (Extension for Community Healthcare Outcomes) and shared the program’s six-month data, highlighting the complexity of diabetes managed by primary care providers (PCPs) and the need for greater innovation in supporting PCPs. He opened by stating that patients with type 1 diabetes have faced a “system failure,” with an inadequate supply of specialists and one-third of patients with A1c levels over 9% as of 2018. After describing barriers for endocrinology care in type 1 diabetes (including financial, supply-demand mismatch, and distance), Dr. Cuttriss described ECHO as an innovative way to overcome these barriers. He shared that ECHO connects providers with specialists through ongoing, interactive, tele-mentoring sessions, as it serves as a low-dose, high frequency workforce development hub-spoke model connecting multidisciplinary specialists (“hub”) with PCPs (“spokes”). Specifically, its goal is to increase capacity of PCPs to manage underserved patients with type 1 diabetes who do not receive routine specialty care. The model held weekly teleECHO sessions for the first six months, followed by alternating weekly sessions after six months, with each session consisting of two case presentations (uploaded in de-identified form by PCPs). Dr. Cuttriss presented data, showing that over the first six months, 54 and 44 cases were presented in California and Florida, respectively, with frequent cases of middle age adults (33.9 +/- 17.4 years, 20% pediatric) with poorly controlled A1c. Dr. Cuttriss also emphasized the complexity of care PCPs were managing, noting that many cases had hospitalizations (28% for California, 15% for Florida) and ER visits (33% for California, 28% for Florida) in the past 12 months. Additionally, common barriers to care reported by PCPs included social support (36% for California, 29% for Florida), financial (30% for California, 26% for Florida), and transportation (23% for California, 3% for Florida). Dr. Cuttriss also noted the high prevalence of behavioral health comorbidities (anxiety disorder and depression) in the presented cases. He shared that ECHO T1D revealed that while the majority of type 1 diabetes patients rely on – and prefer – their PCPs for management, PCPs lack confidence in type 1 diabetes management, specifically in diabetes technology and insulin management. Thus, Dr. Cuttriss emphasized the primary care setting as an opportune setting to improve diabetes outcomes on the population level and shared the strong interest and response that ECHO T1D has already received. In closing, he also shared the program’s virtual resource of weekly series, “ECHO Diabetes – In the Time of COVID-19” – more on this can be found here.

  • ECHO was piloted in California and Florida over a six-month period. Specifically, California had 11 spoke sites (nine FQHCs and two non-FQHCs) and 37 clinic sites, with over 1,000 patients. Florida’s program had 12 spoke sites (5 FQHCs and 7 non-FQHCs) and 68 clinic sites, with over 1,000 patients.  

  • Dr. Cuttriss presented data on the participants’ credentials, noting that the majority were MD/DO/MBBS (41%), nurse practitioner (18%) and RN (18%).

Precision Medicine in Type 2 Diabetes – Using Individualized Prediction Models to Optimize Selection of Treatment

In this presentation, Dr. John Dennis (University of Exeter) continued the enthusiasm for precision medicine by discussing the use of simple clinical measures in developing individualized prediction models to serve as point-of-care clinical decision aids in treatment selection. He first opened by sharing the rationale for precision medicine in type 2 diabetes, emphasizing the lack of guidance on the best drug after metformin and showing the vast heterogeneity in prescribing practices within the UK. In the search for robust clinical predictors of differences in treatment response of various drugs, Dr. Dennis introduced the novel research framework of using “discovery analysis” in routine clinical data to test for clinical markers associated with differential drug response to then conduct “replication analysis” in clinical trial data to test candidate markers as pre-specified hypotheses. With this approach, he highlighted several notable data including: older age is associated with a greater A1c response with SFUs and TZDs but not a differential response; through analysis of the ADOPT trial data, non-obese males have greater responses to SFUs while obese females have greater responses to TZDs. In terms of the approach in using these data to inform selection of optimal treatment, he showed that using an individualized prediction approach (which uses continuous clinical features to order drugs by likely A1c response) outperforms the clusters approach (such as Dr. Ahlqvist’s five clusters classified by pathophysiology). In discussing the implementation of this work, Dr. Dennis illustrated the treatment selection model which inputs clinical features (A1c, age, duration of diabetes, sex, BMI, eGFR, HDL, ALT) (n=72,504) into the prediction model which examines 12-month A1c data of various drugs, and with external validation, predicts the “best” drug for an individual patient. Notably, Dr. Dennis presented some early results, which showed that DPP-4 inhibitors were not the optimal drug for A1c response for any individual, while there was an equal split for SGLT-2 inhibitors, SFUs, and TZDs. He also showed that external validation using the SCI-Diabetes Scotland cohort demonstrated that patients who were prescribed the model’s “best” drug have indeed had greater A1c responses (-3.7 mmol/mol) compared to those who did not use the recommended drug. In closing, Dr. Dennis highlighted the potential of simple clinical measures in the use of precision medicine but also stressed that key challenges including the need to consider non-glycemic drug outcomes and risk of complications remain.

Q: Did you find that adding more clinical parameters refined things more? Were you likely to get higher concordant treatments?

A: It didn’t affect concordance. But it is helpful to stratify patients more. In particular, kidney function is a huge factor in SGLT2 response.

Q: A lot of times in prescribing, we not only use the potential benefits but we also look at the concern for side effects. When you talk about non-glycemic measures, are you also talking about side effects?

A: Yes, we can also use the same approach for side effects. We started with benefits because there is very little data on what is best for glucose lowering right now. We do already somewhat know the average side effect risk for different drugs. But I think there’s a massive scope in personalizing treatment selection using this approach.

Dietician-Led Preventative Interventions Improve Health Outcomes for Patients at Risk of Heart Disease and Type 2 Diabetes; RDN-led Interventions Improve Compliance with Optimal Care Goals Among Type 2s

In a session on the role of dieticians in diabetes care, Ms. Gretchen Benson (Minneapolis Heart Institute Foundation) presented promising data on expanded use of dieticians in medication protocols driving improved outcomes in patients at risk for developing heart disease and diabetes. Ms. Benson presented results from a systematic review indicating that RDN and RN led medication therapy protocols have become more widespread and successful in the last two decades, especially when they supplement primary care visits and utilize telemedicine. Based on this systematic review, Ms. Benson and her team implemented the HeartBeat Connections Program in Minnesota to expand the role of RDNs to identify community members who were at the highest risk for developing heart disease and diabetes and provide preventative care via coaching, and medication protocols (i.e., prescribing rights) for cholesterol and blood pressure management. 1 in 3 people eligible for HeartBeat Connections joined the program and over the course of their enrollment, the smoking rate among participants dropped by 50%, there was a 30% increase in participants meeting recommended weekly exercise goals, and 70% of participants improved their cholesterol. Ms. Benson highlighted the importance of prescribing rights for RDNs and cooperation with primary care providers as the cornerstones of the HeartBeat Connection Program’s success.

  • Ms. Benson also presented data from the ENHANCED study, demonstrating RDN-led intervention significantly increases the ability of type 2 patients to reach the Minnesota optimal care goals. These goals are known as “D5 for Diabetes” and include (i) managing blood pressure (<140/90 mmHg), (ii) maintaining blood glucose (A1c<8%), (iii) be tobacco-free, (iv) take aspirin as recommended, and (v) take statin as recommended. In the ENHANCED study patients met with an RDN to review optimal diabetes measures and were given a Type 2 Diabetes basics booklet. Participants were then randomized with those in the intervention arm (n=60) receiving an average of 10 telemedicine calls with RDNs over the course of one year to provide individualized medical nutrition therapy and provider-approved medication protocols with RDN prescribing rights. Follow-up results indicated a significant improvement in compliance with optimal care goals increasing from 3.1 out of 5 at baseline to 3.7 out of 5 at follow-up. This was significant over the control group (n=58) who experienced an improvement of 2.9 out of 5 to 3.2 (p=0.017). RDN-led intervention also drove significant improvements in the secondary outcomes of increased fruit intake (p=0.011), increased whole grain intake (p=0.005), and taking medications as prescribed (p=0.014).

  • Ms. Benson credited the success of both the HeartBeat Connections Program and the ENHANCED study to collaboration between RDNs and primary care providers. Specifically, Ms. Benson emphasized the importance of RND-led interventions outside the clinic to supplement Primary Care visits for patients and support continuous treatment through prescribing rights for RDNs allowing them to more freely adjust medication to meet the needs of patients. Ms. Benson closed her presentation urging other clinics to evaluate the potential for expanding the role of RDNs and provided the graphic below on how such expansion could be successfully implemented.

DPP + Three Whole-Population Interventions Modeled to Reduce 10-Year T2D Incidence by 17% at Cost of ~$68k/case prevented; Neat Tool Helps Policymakers Determine Prevention Path

RTI International’s Mr. Simon Neuwahl presented a simulation modeling analysis indicating that a combination of risk-based (e.g., DPP) and three whole population-based interventions (e.g., soda tax, worksite health promotion, and bike lanes) will be necessary to reduce US type 2 diabetes incidence by 17% within the next 10 years. The interventions are estimated to come with a price tag of ~$500/person – $164 billion overall – and appear to fall short of the CDC’s target of a 21% incidence reduction by 2025. For context, 1.4 million people were diagnosed with type 2 diabetes in the US in 2018, so the cocktail of interventions proposed by Mr. Neuwahl and co. would prevent ~2.4 million cases for the country at a cost of ~$68,000 per prevented case.

  • Regarding methodology, the authors used the CDC/RTI Diabetes Cost-Effectiveness Model to estimate the costs and efficacy of the National DPP and the three whole-population interventions. From these estimates, they mapped the cost of an intervention combination against the efficacy of the intervention(s) in reducing new case of type 2 diabetes, creating a useful tool for local, state, and national prevention work. According to Mr. Neuwahl, the chart of cost per person vs. percentage of diabetes cases prevented is an incredibly useful tool for determining the lowest cost intervention at all levels of 10-year diabetes prevention goals. Researchers and public health workers can start with the prevention goal they aim to achieve and find the intervention combinations that would roughly achieve that goal; alternatively, they can approximate what prevention goal is achievable within a given budget. Since the y-axis presents cost per person (not per at-risk person), the tool can be scaled for local, state or national level goals.

  • Mr. Neuwhl emphasized that these are estimates and more research is needed to investigate the cost and effectiveness of whole population-based diabetes prevention interventions. There is limited research on the effectiveness of whole population-based interventions, which hindered Dr. Neuwalk and co-researchers’ ability to predict the cost and effectiveness of whole population interventions. The main whole population-based intervention that has been study is the soda tax, which has been shown to be effective. For more on whole-population interventions, see one of our favorite talks from WCPD 2018.

  • Mr. Neuwahl added that the DPP has been shown to be an effective population-specific intervention, and DPP access could be expanded by delivering the intervention virtually. However, improvements in the virtual DPP program’s engagement and long-term impacts might be needed. A recent Kaiser Permanente study shows that at 12 months, those using virtual DPP do not maintain their weight loss (while those who engaged in the in-person program do) and that only 46% of those in the virtual arm completed at least 4 of the 16 sessions.

Team-Based, Patient-Centered Telemedical Model Expands Rural Access, Shows Promising Glycemic Improvements, and Facilitates Patient Engagement and Diabetes Self-Management

Dr. Jodi Krall (University of Pittsburg) presented promising data on the Telemedicine for Reach, Education, and Treatment (TREAT) model, a team-based approach for delivering telemedicine, delivered a 2% A1c reduction at both six months and 12 months, a significantly greater reduction relative to usual care. The program also showed significant improvements in psychosocial and behavioral outcomes, including statistically significant reduced distress, improved empowerment, and increased self-care behaviors. TREAT is centered around the team: the patient, primary care physician, Diabetes Care and Education Specialist (DCES, f.k.a. CDE), and endocrinologist. Because it uses a virtual platform, Dr. Krall highlighted TREAT as a particularly useful tool in expanding access to rural and underserved communities.

  • TREAT is a great platform for both medical nutrition therapy and DSMES and is useful for both specialty care and primary care. Dr. Kroll identified the major challenge in using TREAT in primary care is convincing PCP that DCES are beneficial to their practice. Evidence of the services and benefits that DCESs could provide PCPs – including placing medication orders and improved glycemic outcomes – is essential to increasing the use of DCES’s in primary care.

  • Using shared decision-making tools in tandem with TREAT enables patients to set specific behavioral goals related to their diet and reduces their A1C and consumption of high fat foods. The 2016 study also found that patients felt empowered by the tool and more knowledgeable about their diabetes, and providers and educators found that the tool facilitated conversation and behavior goal setting.

  • Telehealth has been shown to be particularly effective in reaching underserved and rural communities. As Dr. Kroll shares, rural populations often have high rates of diabetes and unmet needs for diabetes care, education, and support. Rural residents have limited access to specialty care (like endocrinology), and the PCPs that serve these communities often struggle to provide all the diabetes care that is needed given lack of resources and time.

Hope Warshaw: We know the vast majority of diabetes care, particularly type 2 diabetes, goes on in primary care, and you’re addressing how to integrate diabetes care and education specialists into primary care. … What are your 2 key recommendations from your work to DCES to “sell” themselves or their program to primary care?

Dr. Jodi Krall: Thanks, Hope, for that question. First, I would say we’re really excited about the new title for the DCES, that that alone will be a great way to sell the service because it changes the focus from education, which is very limited when you think about the scope of the service that is provided, and really, it puts a new perspective and spin on what someone is capable of doing. I showed you some of the things very briefly of what we have done in our system, but I think whenever you are able to focus on how are you able to help practices meet their goals, then they understand the value. I don’t mean just financial, but that is an important consideration because by meeting the clinical goals, they help meet the financial goals. So being able to do that is the challenge but also what has to be accomplished.

Utilizing Personal Health Data – How Much Do We Need, and What Should We Do with It?

Dr. David Ahn (Mary & Dick Allen Diabetes Center) addressed how to handle the rapid growth of patient data being collected by various sensors and apps when providers have less and less face-to-face time with patients during regular visits. According to Dr. Ahn, diabetes care is ideally suited for virtual care because it is centered on self-management, data can be easily generated and shared through personal devices, and screen-sharing can facilitate education through web resources. However, telemedicine visits can be complicated because they tend to require more patient responsibility in terms of preparation, as patients must download data from all of their devices and have lab testing done beforehand. In terms of data overload, Dr. Ahn supported the notion that more data is not always worse – it can even be better if it is archived, shared, and visualized appropriately to make the data actionable. Dr. Ahn recommended the use of passively-collected data obtained from sensors, as opposed to actively-collected data that is manually recorded by patients, as it is more reliable, more sustainable over time, and can be used to create consolidated reports that improve patient-provider interaction and data visualization to make informed clinical decisions. Moving forward, Dr. Ahn relayed the potential benefits of open, bi-directional data transfer between patients and providers in simplifying data integration with EMR systems, helping providers identify patients who are more in need of care, and assisting providers in making informed treatment decisions using artificial intelligence.

Dr. Einhorn Describes the “New Era” of Diabetes Care Enabled by CGM and Telemedicine

During an Abbott product theater, Scripps Whittier’s Dr. Daniel Einhorn spoke excitedly about how the potent combination of CGM and telemedicine is bringing about “a new era” in diabetes care. On the CGM side, Dr. Einhorn provided a slate of Beyond-A1c aphorisms, such as “It’s the difference between having a synopsis vs. the whole short story.” He conveyed that adding telemedicine into the mix helps providers and patients get the most out of CGM: “In the old days we waited months between visits, and it would take years before people would get A1c lower. So those years are now being compressed to days and weeks. The idea of waiting three months between visits becomes silly because you can make decisions much faster [now with CGM]. We’re actually going to change the diabetes guidelines to reflect this possibility. The three-month adjustment cadence is artificial, based on A1c maturation. At the beginning I’ll see a patient a lot; when that person is stable, then I’ll see them less and less and less…If you need me every day, you got me every day. If you need me every six months, you got me every six months.” Dr. Einhorn’s team has the process down: Staff sets up the appointment through a telemedicine platform; the staff then emails/prints the preferred CGM reports for the provider; Dr. Einhorn then has a 25-30 minute discussion with the patient; Finally, the patient and doctor decide on a follow-up plan and set a time to follow up (“Historically, a month, two, or three later, but now, maybe a week or two later. It’s easier now”). This is undoubtedly a much-improved approach for diabetes care, but Scripps Whittier is a well-funded, diabetes-focused center. Even with the availability of codes for remote monitoring and CGM interpretation, reimbursement for more high-touch care (for those who need it, when they need it) is insufficient. CGM and telemedicine provide the technology, but there will likely need to be significant payment reform to encourage reorganization of care teams and delivery models before most clinics are able to offer the Scripps Whittier level of service. This is even more so the case at the primary care level, where the vast majority of people with diabetes are treated. But we’d love to be proven wrong!

Diabetes Apps – To Recommend, or Not to Recommend?

New York-Presbyterian’s Rachel Stahl discussed the potential role of mobile health apps in clinical diabetes care while providing insight for how they can be implemented in practice. With the rapid advancement of diabetes technology and the increasing accessibility of mobile health apps during the last decade, Stahl emphasized the importance of providers staying up to date with the apps that are available and how they can aid patients in self-management. In terms of their strengths, Stahl noted that most diabetes apps are accessible and easy to use, available for no or low cost, able to provide accurate transfer of diabetes data and reports, and ultimately reduce patient burden while increasing engagement through advanced analytics and decision support. Some potential challenges addressed with using these apps in practice include technological issues, limitations in terms of health literacy, health numeracy, and dexterity, a lack of quality, evidence-based research to support the proposed benefits, and concerns regarding data privacy and security. Moving forward, Stahl recommended that providers try the apps out themselves before implementing them into practice, individualize app selection to best suit patient needs, and support sustained app use by providing patients with ongoing training and education. Furthermore, Stahl suggested providers stay informed through professional interest groups formed by the ADA and EASD and ensure their recommended apps meet the security standards established by the Diabetes Technology Society.

Dr. David Wagner’s Recommendations for Diabetes Care Text Messages for Adolescents and Teens: Emojis, Memes, and Humor

Oregon Health & Science University’s Dr. David Wagner addressed the feasibility of using text messages as a personalized intervention to reach high-need groups of patients with diabetes who are socially or medically vulnerable and struggling to meet their self-management goals. Based on programs through REACH, TEXT-MED, MyT1Hero, and others, interventions via text have been shown to be preferred among adolescents and adults alike and to result in high levels of satisfaction and improvements in A1c for those who engage. Through the NICH, Dr. Wagner investigated these findings further by engaging a group of adolescents (n=26) and their parents in highly personalized, 24/7, text-based interventions focused on encouraging and reinforcing healthy behaviors. The results indicated correlations between reduced A1c and more overall texts and between fewer admissions and more texts to caregivers. Dr. Wagner recommended that, when implementing interventions via text, providers should treat individuals with diabetes as people first by supplementing content to make it more meaningful (e.g., include emojis, memes, sense of humor, references to patient interests, etc.), provide increased access and frequent interaction to maintain engagement, empathize with and attempt to normalize patient struggles, and reinforce healthy behaviors with praise. Although text-based interventions may be complicated to implement and require dedication outside of business hours, Dr. Wagner’s work demonstrated the potential of using technology to bridge the gap between providers and high need groups of patients with diabetes.

Digital Health Posters



Details + Takeaways

Machine Learning in Predicting Poor Glycemic Outcomes


Hammam Alquadan, Linda M. Siminerio, Jodi T. Krall, Jason Ng


  • Utilizes clinical data from EMR and medical/socioeconomic profiles of 38,173 patients w/T2D or A1c >6.5% to predict future glycemic control

  • Uses SHapley Additive exPlanations (SHAP), an a-theoretic approach, to explain machine learning outputs

  • Highest predictors for future poor glycemic control are prior elevated A1c, younger age at presentation, previous insulin therapy provided

  • Important for risk-stratification to determine resource allocation

Digital Diabetes Education Program Improves Knowledge, Self-Efficacy, and Health Behaviors among People with Type 2 Diabetes

Zachary White, Ryan Woolley, Sheila Amanat, Kelly Mueller


  • Digital ADA program for people with diabetes to access digital diabetes info

  • Program impact was assessed by comparing survey results from registration to those at end of 12-month program (n=897)

  • Of participants, 58% had increased knowledge of diabetes, 46% had increased confidence in self-management, and 34% increased participation in self-education classes (p<0.001)

Phone-Based Diabetes Education Improves Diabetes Knowledge and Patient-Provider Conversations

Ryan Woolley, Zachary White, Shamera Robinson, Jo Mandelson, Kelly Mueller


  • Unique form of DSMES: phone-based, national group program connects PWD to a new diabetes expert each month to listen & ask Qs

  • Phone-based surveys at several time pts after events for 1 year (n= 509/event, 2,700 unique participants, 72% 65+, 78% female, 69% post-2ndary education)

  • Diabetes knowledge increased 12% after event, 14% more had a conversation with HCP about diabetes/heart disease relationship, 10% more enrolled in DSMES

  • 91% of participants felt confident that can reduce risk of heart disease and stroke 1 month after event

Clinical Video Telehealth: An Effective Approach to Comprehensive Group Diabetes Self-Management Education and Support for Rural and Disabled Veterans

Monica Dinardo, Ada O. Youk, Nicole Beyer, Janice N. Beattie, D. Scott Obrosky


  • Compare proximal (3-6 month) and distal (12-18 month) metabolic outcomes (A1c, lipids, BMI, BP) for in-person (IP, n=359) vs. clinical video telehealth (CVT, n=197) delivery for veterans with diabetes

  • Demographics: mean 63.3 years, 76,3% white, 96.6% male, mean BMI = 33.8, mean 8.6% A1c 

  • Only A1c significantly improved at 6 months (IP: 7.6, CVT: 7.9) & was sustained for 18 months (IP: 7.4, CVT: 8.1) in both groups 

  • Although CVT improves access to DSMES and is effective; IP delivery was more effective than was CVT 

Dulce Digital-Me: An Adaptive M-Health Intervention for Underserved Hispanics with Diabetes

Addie L. Fortmann, Athena Philis-Tsimikas, Samantha R.S. Bagsic, Daniela G. Vital, Jennifer A. Jones, Haley Sandoval, Kimberly L. Savin, Taylor Clark, Kimberly Luu, Job G. Godino, Linda Gallo


  • Dulce Digital-Me (DD-Me) is a mobile health intervention of educational text messages w/personalized goal setting & feedback

  • Compares 6-month change in lifestyle changes, adherence, BG monitoring for 3 groups: texts + feedback via (1) algorithm messages or (2) medical assistant Health Coach to (3) only educational text messages (control)

  • Demographics: Hispanic adults w/T2D & A1c >8% at Federally Qualified Health Center in CA

  • Preliminary results show higher BG monitoring frequency for feedback via Health Coach than algorithm feedback or control

Do Diabetes Apps Deal with Blood Pressure; do Blood Pressure Apps Deal with Diabetes?

Gloria Wu, Jonathan Wong, Chap-Kay K. Lau, Donia Momen, Rebecca Rottenborn


  • Assessed free, popular Android & Apple apps for DSMES criteria and ACC/AHA blood pressure guidelines

  • Diabetes apps follow more DSMES criteria than BP apps follow ACC/AHA guidelines (6.7/11 vs. 4/20)

  • No diabetes apps fit any ACC/AHA guidelines while BP apps fit 3.6/11 DSMES criteria

  • Indicates the need for apps that address diabetes with comorbidities

Recruitment and Retention of Minority Women for Women in Control 2.0: A Virtual Diabetes Self-Management Education Study

Alexa Bragg, Paula Gardiner, Shakiyla B. Woods, Jessica M. Howard, Suzanne Mitchell


  • Examines relationship between Women in Control (WIC) recruitment strategies, enrollment rates, & retention of minority women in DSMES programs

  • Targeted recruitment methods (letters, phone calls) result in most screens, enrollments (975 screened, 207 eligible, 195 enrolled, 75% attended 5/8 DSMES sessions)

  • No relationship between recruitment method and retention in intervention

Telephonic Coaching is Associated with Improved Glycemic Control in Many Individuals with Type 1 and Type 2 Diabetes

William C. Biggs, Ann Buskirk, Lena Borsa, Maureen R. Lyden, Christopher Parkin, Bettina Petersen


  • 12-month, observational, self-controlled, multisite study using retrospective & prospective data of efficacy of virtual diabetes educator tool (DHP)

  • Overall, 0.5% A1c reduction; among responders (>0.5% A1c reduction group), mean 1.2% A1c reduction

  • Differences between those with vs. without A1c reduction may be attributed to (i) baseline A1c, (ii) baseline body weight, (iii) change in body weight, (iv) focus on diet/nutrition counseling

  • Participation associated w/improved outcomes & reduced costs for cohort of elderly, high-risk pts (mean 71 years, 8.4% A1c)

Telephonic Coaching is Associated with Improved Patient Activation in Diabetes Self-Management

Ann Buskirk, William C. Biggs, Lena Borsa, Maureen R. Lyden, Christopher Parkin


  • Assessed impact of participation in Diabetes Health Partnership (DHP) on patient activation using Patient Activation Measure (PAM), a validated, 100-Q questionnaire to assess diabetes self-management

  • Mean PAM scores increased from 61.8 to 72 (p=0.0005), suggesting patient activation through DMP

  • Notable difference in PAM score, confidence, and self-management for both responders & non-responders, but not associated with A1c improvements

Diabetes Prevention Care Preferences among People with Severe Mental Illness Taking Antipsychotic Medications

Esti Iturralde, Ashley L. Jones, Natalie Slama, Andrea H. Kline-Simon, Stacy Sterling, Julie Schmittdiel


  • Semi-structured interviews of adults w/severe mental illness w/out diabetes who take antipsychotic medications (mean 42 years, 72% female, 42% racial minorities)

  • ~90% rated weight or diabetes risk as a top health concern, ~30% have pre-diabetes, ~70% have obesity 

  • Many pts actively interested in diabetes prevention; 40% have attempted to attend a lifestyle change program/service in past

  • Many not aware of metformin to reduce diabetes risk and medication-related weight gain & are interested

  • Want support from providers to overcome barriers (cost, emotional, motivational)

Impact of Telemedicine Clinics in the Management of Type 1 Diabetes

Leslie A. Eiland, Mohammad Siahpush, Padmaja Akkireddy, Lisa Kuechenmeister

  • Retrospective study of regular, structured, synchronous video conferencing office visits w/endocrinologist over a long duration (n=139, 58% female, 97% non-Hispanic white, mean 44.5 years, mean 8.3% A1c)

  • Each year in telemedicine was associated w/a 0.13% A1c decline

  • No difference in A1c based on gender, age, prior provider, treatment change

  • For those with A1c >9%: each year of telemedicine associated w/ -0.47% A1c, larger improvement among those previously managed by PCP

“TechQuity” in Diabetes: Does Digital Health Technology Improve Equity?


Kelly Jean Craig, Kyu B. Rhee


  • Systematic scoping review evaluated digital health interventions (DHIs) which address global health disparities in cardiovascular related diseases/disorders with subgroup focus on diabetes

  • Preventative DHIs focus on weight loss, diabetes screening, glycemic control, and overmedicalization

  • Identified DHIs target specific disparities: 29% target Black populations, 24% Hispanic, 62% low SES, 52% rural

  • Preventative DHIs improved health (70%) and healthcare delivery (40%) outcomes in global diabetes-specific disparities

Prediction of Type 2 Diabetes Occurrence Using Machine Learning Model

Henock M. Deberneh, Intaek Kim, Jae Hyun Park, Eunseok Cha, Kyong Hye Joung, Jong Seon Lee, Dong Seok Lim


  • EHR data set screened based on continuity of care, lack of diagnosis, complete set of features (n=80,692)

  • Each instance had 1,444 features/variables which were narrowed down to final 12 features: FPG, A1c, triglycerides, BMI, gamma-GTP, gender, age, uric acid, smoking, drinking, physical activity, family history

  • Prediction accuracy was 73% using 12 features, which is superior to just 5 traditional features

  • r-GTP, uric acid, triglycerides, and lifestyle factors should be included in type 2 diabetes predictions

Targeting People with Unknown Diabetes in Health Insurance Databases through Artificial Intelligence

Sonsoles Fuentes, Rok Hrzic, Sofiane Kab, Romana Haneef, Sandrine Fosse-Edorh, Emmanuel Cosson


  • Applied machine learning to the French National Health Data System (SNDS, health insurance database) 

  • Among the 3,471 variables coded, the 5 most important discriminating factors for unknown diabetes cases were age, sex, number of out-of-hospital reimbursements for blood lipid profile tests in last year, general practitioner consultations, blood glucose tests

  • Algorithm is specific (70%), sensitive (71%), and precise (69%)

Attitudes of Physicians to Chances, Risks, and Future Options Regarding Digitalization and New Technologies in Diabetes


Bernhard Kulzer, Norbert Hermanns, Dominic Ehrmann, Lutz Heinemann


  • German online survey with 56 questions targeting diabetologists (n=326)

  • 75.8% of participants both had positive views of digitization and perceived a positive potential of digitization in diabetes care

  • Identified advantages of digitalization include better communication with patient (71.6%), more support for treatment decisions (66.8%), more patient empowerment, (62.3%), and better treatment quality (62.3%)

  • Only the lack of reimbursement (80.0%) and legal issues (50.3%) were identified as disadvantages of digitization by a majority of diabetologists  

  • Diabetologists expect digitization to increase in next few years, especially AID systems

Different Attitudes of Physicians, Parents of Children with Diabetes, and People with Diabetes towards Digitization

Bernhard Kulzer, Norbert Hermanns, Dominic Ehrmann, Lutz Heinemann

  • Online survey about attitudes toward digitization completed by PWD (n=3,427, PWDT1: 65.6%, ParentsT1: 8.1%, PWDT2: 25.5%) and HCPs (n=326)

  • Among all PWD groups and HCPs, >74% of participants had positive views of digitization

  • Software for analyzing glucose data was more important to HCPs than PWD (72% HCPs rated it as most important vs. ~50% PWD)

  • Online education was most relevant to PWDT2 (41.7%) whereas closed loop systems were rated very important by 63.3% of ParentsT1 and 65.6% of PWDT1; most HCPs did not see either as very important (13.8%, 41.7%)

Users with Type 2 Diabetes Using a Digital Platform Experienced Sustained Improvement in Blood Glucose Levels

Yifat Hershcovitz, Sharon Dar, Eitan Feniger

  • <