ATTD (Advanced Technologies and Treatments in Diabetes) 2021

June 2-5, 2021; Virtual; Full Report

Executive Highlights

  • Thinking about early June, novelist John Steinbeck once wrote, “the world of leaf and blade and flowers explodes, and every sunset is different” (The Winter of Discontent). That was the setting for the 14th annual ATTD meeting which took place nearly four months later than its traditional February timeslot as a result of the COVID-19 pandemic. Although the meeting didn’t quite get the April, 2021 Paris opening that had been the goal once upon a time (in 2020), with a largely successful vaccine rollout in the US, our own Close Concerns office back open again, and a deep bench of exciting learnings from ATTD, it does indeed feel in some respects like things are starting to return from a long “Winter of Discontent” – this, even though so many countries are still deep in the midst of the pandemic (India and Brazil in particular, with the former not so far from overtaking the US in total number of infections and what seems like moments from over 30 million infections – and the latter nearing the same number of deaths as the US from a country far smaller in total population).  And yet and still - from the ATTD Yearbook session to the absolutely stirring emcee work by Drs. Moshe Phillip and Tadej Battelino (we will never forget Professor Battelino’s reassuring voice each day, “hello, friends …”) to the strategically scheduled timing of the meeting (meaning we weren’t starting at 2 am each day as we were during EASD), ATTD 2021 stuck out at the top of the list for our favorite virtual conferences of the year.

  • CGM: Results of the MOBILE study were published simultaneously in JAMA as IDC’s Dr. Tom Martens read out results on day 1 of ATTD. The multi-center RCT included 175 individuals with type 2 on basal-only therapy from primary care practices. At eight months, the CGM cohort spent 3.8 more hours/day in range than did those in the BGM cohorts and the CGM cohort saw a 0.4% adjusted A1c improvement over the BGM cohort. In the pipeline, results from the PROMISE study, the 180-day pivotal study for Senseonics’ Eversense implantable CGM, revealed an overall MARD of 9.1%. Overall sensor MARD for Guardian 4 with no calibrations was 10.6% for upper arm wear and 10.8% for abdomen wear.

  • AID: We got our first look at real-world data from MiniMed 780G users, following the system’s launch in fall 2020. The results came from the first ~4,000 users of MiniMed 780G across eight countries, including the UK, Italy, Belgium, the Netherlands, Qatar, South Africa, Sweden, and Switzerland. Overall Time in Range for MiniMed 780G users in the real-world was 76%. Prof. Goran Petrovski presented very impressive results from a small study that initiated 34 children and adolescents (ages 7-17) straight from MDI to MiniMed 780G. Time in Range increased by a remarkable 8.8 hours/day after three months on MiniMed 780G. From UVA, a new AID algorithm, RocketAP, was tested against USS Virginia (a research version of Control-IQ) in a small study and performed significantly better on unannounced meals (n=18). Bruce Buckingham read out post-hoc analysis results from the Omnipod 5 pivotal trial, which highlighted the advantages of Omnipod 5 over open loop in a new way. From Tandem, Dr. Steph Habif read out interim results from the ongoing real-world Control-IQ Observational (CLIO) Study. At three months, participants saw a mean Time in Range of 71% and spent 1.6% of time <70 mg/dl and 27% of time >180 mg/dl, all improvements from baseline.

  • COVID-19: The highly-respected and well-known Dr. Will Cefalu from NIH gave an insightful presentation on recent efforts at the NIH to support research on COVID-19 and diabetes, highlighting the need for additional research on post-acute symptoms, underlying disease mechanisms, and COVID-19 related new-onset diabetes. T1D Exchange’s COVID-19 Registry found stunning differences in COVID-19 outcomes by technology usage. While the two groups were not well matched, rates of ICU admission were six times higher for non-CGM users than CGM users (18% vs. 3%).

  • Diabetes drugs: The future of type 1 treatment took center stage at this year’s ATTD, fueled by the 100th anniversary of insulin’s discovery. We left the meeting with the sense that longer acting, more stable, and faster formulations of insulin are just around the corner. During a Novo Nordisk-sponsored symposium, Dr. Chantal Mathieu (KU Leuven) spotlighted the handful of once-weekly basal insulin formulations currently in development: Fc fusion (candidates from AZ, Lilly, and Hanmi), PEGlyated (candidate from Antria/Rezolute), and acylated (insulin icodec from Novo Nordisk). Enthusiasm was also palpable for stem cell-derived beta cell replacement therapies, which could solve the ongoing shortage of donor pancreas/islet samples. ViaCyte’s then CSO Dr. Kevin D’Amour – who has since resigned to become CSO of Brooklyn ImmunoTherapeutics – presented the latest on ViaCyte’s beta cell replacement pipeline, including a first look at CGM data from a participant using PEC-Direct (showing a 22% increase in Time in Range or ~5.3 more hours/day post-implantation). As usual, Dr. Jay Skyler (University of Miami) gave his annual closing talk at ATTD on type 1 diabetes immunotherapies, in which he shared that clinical trial data from the past year on candidates like anti-TNF and anti-IL-21/liraglutide support the need for chronic and combinatorial treatment, as the vast majority of candidates only show a transient period of C-peptide improvement or maintenance.

The 14th annual ATTD conference was certainly one to remember. Immediately below, you’ll find our top themes from the meeting, followed by highlights in the following categories:

  • Continuous Glucose Monitoring

  • Automated Insulin Delivery

  • COVID-19 and Telemedicine

  • Big Picture

  • Diabetes Drugs

  • Decision Support and Digital Health

  • Time in Range and Beyond A1c

  • Posters

  • ATTD Yearbook

  • Exhibit Hall

Table of Contents 

Themes

Kudos to Drs. Moshe Phillip and Tadej Battelino! ATTD Delivers Both Learning and Community in Virtual Format

Like nearly all aspects of life during the COVID-19 pandemic, there were a lot of things that were different about ATTD 2021 compared to previous years. To start, the conference took place in June for the first time, nearly four months after its traditional February timeslot. Of course, the conference also took place in a virtual format for the first time, a departure from the buzzing crowds we’ve seen in Madrid, Berlin, and Vienna in recent years. Despite this, ATTD still brought the immense learning we’ve come to expect from the meeting, as well as the sense of a “diabetes community” we love about the tightly-run meeting.

  • Despite being thousands of miles apart, meeting organizers Drs. Moshe Phillip and Tadej Battelino appeared together in a virtual studio and didn’t miss a beat. Drs. Phillip and Battelino were upbeat and rapport was natural. We especially loved the annual ATTD Yearbook session, in which Drs. Phillip and Battelino emceed from their “studio” and brought in a long list of KOLs to talk about some of the most relevant studies published this past year. In the meeting’s final plenary session, Dr. Jay Skyler came out to give his annual update on type 1 diabetes cures. Drs. Phillip and Battelino brought Dr. Skyler onto their big screen in the virtual studio and conducted an interview. If you haven’t already, we’d highly recommend checking out the #ATTDAnywhere photo contest.

  • This year’s ATTD ran from approximately 1 PM to 9 PM Central European Summer Time. Though this might have skewed late for attendees in Europe, we commend this decision to make the conference more accessible for viewers tuning in from around the world. For context, our tech team covered the meeting from the Atlantic Time Zone, placing sessions from 9 AM to 5 PM on most days. For the rest of our team on the West Coast, sessions took place from 5 AM to 1 PM – a bit early for some, but certainly still accessible. Of course, for those in less accessible time zones (e.g., Asia and Australia), the virtual format of the meeting meant most sessions were available on-demand shortly after they took place.

Study Designs – Gotta Catch ‘Em All: Randomized Clinical Trials, Crossovers, In Silico, Real-World, and More

As usual, study result readouts were some of our biggest highlights from ATTD. Perhaps less usual, we saw a lot of different study designs. As more data-generating devices (e.g., CGMs, connected BGMs, insulin pumps, connected pumps, fitness trackers, etc.) enter the diabetes ecosystem, we’re almost assured to see more real-world studies. The FDA also continues to progress its thinking on real-world evidence and total product lifecycle-based regulation. Looking ahead, we’re curious to see how patients and providers use these newer types of studies to evaluate safety and efficacy of interventions. 

  • On the first day of ATTD, we saw results from the landmark MOBILE randomized controlled trial. The multi-center RCT included 175 individuals with type 2 on basal-only therapy from primary care practices randomized to either use the Dexcom G6 CGM (n=166) or the One Touch Verio Flex BGM (n=59) for eight months. The study results for MOBILE were published in JAMA on the same day, along with a real-world retrospective study of 41,573 people with insulin-requiring diabetes in Kaiser Permanente of Northern California’s registry. Both studies, though very different in design, supported the utility of using CGM in people with type 2 diabetes, something eloquently argued in a JAMA editorial that was published with the two studies. From our view, the two different study designs made the argument for CGM in type 2s even stronger, providing evidence that the device is effective (shown in the RCT) and that that effect is generalizable (shown in the real-world study).

  • Elsewhere, there was a lot of real-world data (we have an entire separate theme on this below). The meeting kicked off with our first look at very positive real-world data from MiniMed 780G users in Europe (n=4,120). Notably, presenter Dr. Ohad Cohen highlighted the fact that the first batch of real-world data matched well with the company’s internal in silico trials and the pivotal study read out at ADA 2020. T1D Exchange’s COVID-19 registry provided troubling disparities in COVID-19 outcomes for people with diabetes by technology usage. The registry revealed that rates of ICU admission were six times (!) higher for non-CGM users than CGM users and three times higher for non-pump users than pump users. Staying on COVID, Glooko registry data showed slight improvements in Time in Range during COVID-related lockdowns. Tandem, Insulet, and Dexcom also got in on the real-world action sharing glucose and other data from their users.

  • Though not intentionally designed, we enjoyed seeing “cross-over” results from Insulet’s Omnipod 5 pivotal trial. As a reminder, the study protocol initially aimed to gather two weeks of open loop at baseline, then have participants spend three months using the Omnipod 5 hybrid closed loop system. Instead, the trial was paused in early March 2020 and resumed in June 2020 due to a software anomaly. As a result, participants had to effectively “crossover” from closed loop back to manual mode in March and then cross back over to closed loop in June 2020 to finish out the study. Other notable crossover trials we saw at ATTD included UVA’s small, but very promising test of its RocketAP algorithm and the MOBILE study extension phase.

CGM Innovation is Not Dead: New Form Factors, Longer-Wear, and New Use Cases

At the beginning of last year, in an interview with MobiHealthNews, Dexcom CEO Kevin Sayer was asked: “Getting smaller, lasting longer, getting more accurate — is that where [the CGM] space is now? Is it almost like phones, just a bigger camera, or are there still paradigm-shifting changes to come?” If this year’s ATTD was any answer, there is still meaningful innovation happening in the CGM space.

  • Implantable CGM represents a significant departure from the on-body, adhesive-based, under-the-skin electrode CGMs from Abbott, Dexcom, and Medtronic. At ATTD, we saw data from PROMISE, Senseonics’ pivotal trial for its 180-day implantable Eversense CGM. In all, the 181 adults wore 279 sensors (96 subjects had a sensor placed in each arm and two sensors were replaced). Participants took part in ten clinic visits with glucose manipulation over the 180 days, generating a total of 49,613 paired points that were compared against YSI reference. With one calibration per day, Eversense accuracy was quite strong and would likely support approval as a non-adjunctive CGM. Of course, Senseonics’ CGM still requires an on-body transmitter to collect glucose data. For a truly “invisible” CGM experience, we had to visit the Exhibit Hall, which featured Indigo Diabetes and its two-year implantable device that does not require any on-body component.

  • The cost of CGM was another topic that come up during this year’s meeting. For people on rapid-acting insulin, where CGM is already widely reimbursed in the US, we were encouraged by a claims analysis that showed CGM was able to pay for itself within just one-year of initiation. On the other side of the equation, longer-wear for CGMs could not only deliver more convenience for patients, but also lower per-day cost of wear. In addition to the 180-day Eversense data, we saw pre-pivotal data from Dexcom’s next-gen G7 CGM. That device will launch initially with ten-day wear, but Dexcom aims to extend that out to 14-16 days in the near future. Of the major CGM systems, that would mean one implantable system that lasts 180 days, a 14-day CGM (FreeStyle Libre 2 or 3), a 14-16-day CGM (Dexcom G7), and a 7-day CGM (Medtronic Guardian 3 or 4).

  • Finally, ATTD 2021 highlighted a number of new and emerging potential use cases well-beyond every day insulin or even diabetes management. On day #2, we saw data from a small survey of Team Novo Nordisk members who showed interest in using CGM for improving their exercise performance. On the final day of the meeting, Dr. Spencer Frank (Dexcom) presented feasibility results using Dexcom G6 as a tool for diagnosing diabetes. Overall, ten days of Dexcom CGM data (i.e., one wear session) performed similarly to oral glucose tolerance tests at diagnosing diabetes. In the future, it’s possible we could see broader populations (e.g., all pregnant women, people with CVD, etc.) get a CGM during their clinic visit. Presumably, this would help catch a significant number of people with undiagnosed diabetes or prediabetes earlier in the disease progression. Use of CGM in the hospital continues to be a theme at meetings: ATTD provided an excellent overview of recent inpatient CGM studies, as well as a study of FreeStyle Libre in patients with diabetes hospitalized with COVID-19 pneumonia with positive results.

Evidence and Support Mounting for Diabetes Technology Use in Populations Outside of Tech-Savvy, Well-Controlled Type 1s

ATTD 2021 built on increasing momentum to expand diabetes technology access and use to populations outside of tech-savvy type 1s with low A1c values. Study populations ran the gamut this year with AID and pump studies on tech-naïve MDI users and those with high rates of hypoglycemia to CGM studies on primary care-treated type 2s on basal-only therapy. It was refreshing and powerful to see how effective diabetes technologies could be in supporting these broader populations in their diabetes self-management and in improving outcomes. The breadth of data was a reminder of the power that these technologies could provide once broader populations have access and coverage.

  • AID and pump therapy for patients who are on MDI or are tech-naïve was a major talking point at this year’s meeting. On Day #1, Prof. Goran Petrovski (Sidra Medicine) presented data from a small study in which 34 children and adolescents (ages 7-17) on MDI at baseline saw a +8.8 hour/day improvement in their Time in Range to 79% three months after initiating MiniMed 780G. This was achieved with only a 10-day initiation protocol and with about half of participants being both on MDI and CGM-naïve at baseline. Though a small study, the very impressive results suggest nearly everyone using rapid-acting insulin can benefit from MiniMed 780G, regardless of technology experience at baseline. On Day #2, Dr. Grazia Aleppo (Northwestern) read out real-world Omnipod DASH data showing that those with higher baseline A1c values (>9% in particular) saw the greatest improvements in A1c and Time in Range in the first 90 days. Likewise, those on MDI at baseline saw significantly greater A1c improvements than did those on other pumps at baseline. It’s encouraging to see more research on AID systems and insulin pumps in these tech naïve groups and to see such promising results.

  • We also saw data showing that AID systems are beneficial to those with “complicated” diabetes. On Day #3, Dr. Bruce Buckingham read out post-hoc analysis results from the Omnipod 5 pivotal trial that showed that Omnipod 5 is highly effective at bringing people with a lot of hypoglycemia into better glucose control. Looking ahead, Dr. Roman Hovorka shared that CamDiab has ongoing studies on the use of CamAPS FX in newly-diagnosed adolescents, pregnant women, elderly, and type 2s, which will expand the data we have on AID system use in broader, less frequently studied populations.

  • With the smart pen landscape broadening with the entrance of Bigfoot Biomedical in the US and NovoPen 6 and EchoPlus in Europe, we also saw more data on expanded support for patients on MDI. On Day #2, Dr. Ofri Mosenzon (Hassadah Medical Center) showed data indicating non-inferiority of DreaMed’s Advisor Pro insulin dosing recommendations to advice from expert providers for patients with type 2 diabetes on MDI, suggesting that the tool may be suitable for use among type 2s on MDI (we also saw similar data for type 1s on MDI). On Day #1, Mr. David Dunleavy (Medtronic Diabetes) presented a brief update on Medtronic’s pipeline, revealing a new project aiming to build a personalized dosing decision support app under the name Project Janus. Project Janus will utilize CGM, machine learning, and Medtronic’s acquisitions, like Nutrino and Klue, purchased in 2018 and 2019, to help provide bolus reminders and more detailed insulin dose decision support and/or food support. Support tools like those that DreaMed and Medtronic are building are essential to meeting patients where they are. It is unlikely that all patients will want to be on an AID system, so it is important that we expand access to tools and technologies that can support patients in their current diabetes self-management practices, ease their burden, and improve their outcomes.

  • The read-out of the MOBILE study drove much conversation surrounding CGM in type 2s. The eight-month RCT showed that primary care-treated type 2s on basal-only therapy (n=175) spent +3.8 hours/day in range (59%) on Dexcom G6 compared to BGM users. Simultaneous to the ATTD 2021 special session, the study was published in JAMA alongside a real-world retrospective study on Dexcom G6 in insulin-requiring type 1s and type 2s from Kaiser Permanente and an editorial by Drs. Monica Peek and Celeste Thomas. In a later session at ATTD 2021, Dr. Grazia Aleppo read out preliminary data from the extension phase of the MOBILE study. Together with the 2017 DIAMOND study that illustrated the glycemic benefit of CGM in type 2s on MDI, these results are incredibly exciting in the benefit they show CGM to offer type 2s on basal insulin. It is this sort of RCT data is incredibly valuable in its support for expanded access to CGM among type 2s, and we hope to see coverage decisions follow the science on this. Of course, the next step will be similar RCTs on CGM use in type 2s not on any insulin at all, which will provide particularly impactful data. However, given that in the MOBILE study, there was not a significance difference in the insulin dosing change between the CGM and BGM groups, the MOBILE study may suggest that much of the improvement in the CGM group was due to behavior changes, although at this point, that’s conjecture, and more data is needed.

  • Eventually, with expanded access to technology, patient onboarding and support and provider workflow integration become even more pressing issues. At this year’s meeting, we already saw greater attention paid to these issues with a talk from Dr. Laurel Messer (Barbara Davis) on the need for comprehensive, standardized AID education and clinical follow-up and an oral presentation on how classifying CGM profiles into a set of “motifs” with associated clinical recommendations could make CGM interpretation easier and more actionable for primary care providers. 

Real-World Data Galore: Maturing Diabetes Technology Field Moves into Real-World Data with Implications for Patients, Providers, and Payers

Real-world data stole the show at ATTD 2021 highlighting the applicability of diabetes technology across populations and demonstrating a strong association with improved outcomes. In our view, the recent bounty of real-world data has largely been driven by increased patient uptake of diabetes technologies. Additionally, as more providers become interested in using technological solutions for their own patient populations, many are likely to turn to real-world data to help inform their clinical decision-making. Additionally, as more diabetes technologies continue to receive approval both in the US and abroad, real-world data has the potential to play a key role in securing strong reimbursement by showing improved outcomes and reduced health care expenditures outside the more controlled environment of a clinical trial. The real-world data presented at ATTD this year was all the more impressive because much of it indicated similar, if not superior, outcomes to those seen in clinical trials, especially for the growing range of automated insulin delivery systems. From our perspective this speaks volumes about the quality of the systems currently on the market and helps us maintain a hopeful outlook for the future of diabetes technology use and adoption.

  • If last year was the year of AID at ATTD, this year was the year of real-world AID data. Medtronic presented real-world data from its Minimed 780G system indicating an average Time in Range of 76% (n=4,120), outperforming the pivotal trial result of 75% Time in Range (n=157). We also caught interim results from the Control-IQ Observational (CLIO) Study that demonstrated users maintained 71% Time in Range three months after initiating the system. Additionally, results from early adopters of Control-IQ showed Time in Range values ranging from 71%—75% after only six weeks on the system with the most significant improvements coming from patients with lower baseline Time in Range. This data was in-line with other reports we’ve seen showing that patients with lower baseline Time in Range or higher baseline A1c often experience the most significant benefits from adopting automated insulin delivery systems. Now with real-world data also in support of this trend, we feel this provides strong evidence in favor of expanding access to diabetes technologies and potentially re-thinking conventions regarding who will experience the largest benefits of automated insulin delivery systems.

  • Turning to CGM, data from the T1D Exchange Quality Improvement Registry indicated CGM was associated with lower average A1c among a real-world population (n=11,472). Specifically, the data indicated that among the study population, CGM users had an average A1c of 8.1% compared to patients on SMBG with an average A1c of 8.7%. We also saw real-world data from Dexcom demonstrating an 86.6% retention rate among a population (n=31,034) of G6 users. Among this population, G6 users also saw an average Time in Range increase of +57 minutes/day after two years on the system. Finally, real-world data from Glooko that was collected during the COVID-19 pandemic indicated that patients using CGM saw small but significant Time in Range increases during lockdowns from 62% in early 2020 to 63% in the late spring. Throughout the COVID-19 pandemic, Glooko and other remote monitoring services have shown the power of passive real-world data collection and we expect they will be here to stay. As we continue to see more and more real-world data from CGM users, we are also interested to see if data from emerging populations including patients with type 2 and pregnant women. While we know CGM can improve outcomes in these populations, real-world data has the potential to help scale adoption and support efforts to secure payer coverage.

Moving Beyond the Past 100 Years of Insulin: Type 1 “Cures” Space Gains Momentum with Immunotherapies, Next-Gen Insulins, and Cell-Based Therapies

  • In diabetes therapy, the future of type 1 treatment took center stage at this year’s ATTD, fueled by the 100th anniversary of insulin’s discovery. We left the meeting with the sense that longer acting, more stable, and faster formulations of insulin are just around the corner. During a Novo Nordisk-sponsored symposium, Dr. Chantal Mathieu (KU Leuven) spotlighted the handful of once-weekly basal insulin formulations currently in development: Fc fusion (candidates from AZ, Lilly, and Hanmi), PEGlyated (candidate from Antria/Rezolute), and acylated (insulin icodec from Novo Nordisk). Notably, insulin icodec is progressing through the phase 3 ONWARDS program and Lilly’s basal insulin-FC is completing a “larger” phase 2 program. At ATTD, Stanford’s Dr. Eric Appel presented the latest on his lab’s efforts to design ultra-fast insulins using excipients discovered via high-throughput screening. In our view, the advent of high-throughput screening has the potential to revolutionize many areas of diabetes drug development – by testing hundreds of unique candidates in rapid fire, Dr. Appel’s group was able to discover an excipient that delivered peak action 4x faster than Humalog in a porcine model.

  • Enthusiasm was also palpable for stem cell-derived beta cell replacement therapies, which could solve the ongoing shortage of donor pancreas/islet samples. ViaCyte’s then CSO Dr. Kevin D’Amour – who has since resigned to become CSO of Brooklyn ImmunoTherapeutics – presented the latest on ViaCyte’s beta cell replacement pipeline, including a first look at CGM data from a participant using PEC-Direct (showing a 22% increase in Time in Range or ~5.3 more hours/day post-implantation). Up next for ViaCyte (and presumably the majority of companies in this space), its team is working to tackle acute cell survival (the number of cells that survive post-implantation) in order to improve the frequency (number of people) and magnitude (increase in C-peptide) of benefit.

  • Dr. Jay Skyler (University of Miami) gave his annual closing talk at ATTD on type 1 diabetes immunotherapies, in which he shared that clinical trial data from the past year on candidates like anti-TNF and anti-IL-21/liraglutide support the need for chronic and combinatorial treatment, as the vast majority of candidates only show a transient period of C-peptide improvement or maintenance. Dr. Skyler’s DIPIT trial, designed to apply long-term anti-TNF, deliver immunomodulation via ATG or anti-CD3, and preserve beta cell health with a GLP-1 to test this hypothesis, is slated to begin in December 2021.

  • Last, but certainly not least, a symposium from Europe’s INNODIA “private-public partnership against type 1 diabetes” gave us immense hope for the future of type 1 “cure” therapy discovery. Back in 2020, INNODIA worked to develop a “master protocol,” to standardize and streamline data collection for type 1 clinical trials. As of April 2021, INNODIA has recruited more than 4,500 trial participants and four clinical trials are ongoing: (i) MELD-ATG; (ii) IMPACT; (iii) Ver-A-T1D; and (iv) CFZ533. MELD-ATG, investigating ATG in people with newly diagnosed type 1 diabetes, is expected to complete in September 2021, and we hope to hear an update by next year’s summer conferences. More broadly, we wonder if lessons learned from Europe can be applied in the US, giving ongoing struggles to enroll type 1 “cure” clinical trials.

Continuous Glucose Monitoring Highlights

Eight-Month RCT Shows Primary Care-Treated Type 2s on Basal-Only Therapy on Dexcom G6 Spend +3.8 Hours/Day in Range (59%) Compared to BGM Users; Just Published in JAMA!

In a special session via Zoom, Dr. Tom Martens (International Diabetes Center) read out results from the MOBILE study, which investigated the impact of CGM on the glycemic control of type 2s on basal-only therapy. The study, “Effect of Continuous Glucose Monitoring on Glycemic Control in Patients with Type 2 Diabetes Treated with Basal Insulin: A Randomized Clinical Trial,” was simultaneously published in JAMA this afternoon! This is a very big deal. It’s always enormously exciting, and we don’t have that many examples yet, to see diabetes studies published in notable journals like JAMA and NEJM (like the Control-IQ pivotal trial), especially this study, given its focus on CGM in non-intensive type 2s. The multi-center RCT included 175 individuals with type 2 on basal-only therapy from primary care practices. All participants were >30 years old, had their diabetes managed in primary care, used 1-2 long- or medium-acting insulin injections/day, had A1c values between 7.8% and 11.5% (average 9.1%), were CGM-naïve, and had used BGM at least three times/day for the month prior to screening. Following screening, participants were randomized to either use the Dexcom G6 CGM (n=166) or the One Touch Verio Flex BGM (n=59) for eight months. Regardless of the treatment arm, all participants had four clinical visits and three virtual visits with study clinicians at the same intervals across the eight-month period. For both groups, diabetes specialists acted as advisors to the participants’ primary care physicians, which while innovative, limits the generalizability of the results to at least some degree (though we will be eager to see what might prompt more focus on “in general” what interventions among people with type 2 diabetes work the best). A six-month extension phase has just been completed, from which preliminary data will be read out later in another (!) ATTD 2021 session. We can hardly believe the latter is true as we would’ve expected that would be another conference later this year. What a win for people with diabetes, attendees at ATTD, and the conference organizers.

Metrics at Month Eight

BGM (n=59)

CGM (n=166)

Adjusted Difference

Time in Range

43%

59%

+3.8 hours/day (16%)

Time >250 mg/dl

27%

11%

-3.6 hours/day (-15%)

A1c (change from baseline)

8.0% (-0.6%)

8.4% (-1.1%)

0.4%

  • Back to the findings! Remarkably, at eight months, the CGM cohort spent 3.8 more hours/day in range than did those in the BGM cohorts - with Time in Range values of 59% and 43%, respectively (adjusted difference of 16%; p<0.001). Clearly, while there’s significant room to spend more time in range, nearly four hours more a day or a whopping 27 hours more a week or more than a full day, with a lot of upside to go. Notably, CGM participants also spent 3.6 hours less per day with glucose values >250 mg/dl than did the BGM group (11% vs. 27%, respectively; adjusted difference of 15%). Together with the 2017 DIAMOND study that illustrated the glycemic benefit of CGM in type 2s on MDI, these results are incredibly exciting in the benefit they show CGM to offer type 2s on basal insulin.

  • The CGM cohort saw a 0.4% adjusted A1c improvement over the BGM cohort. Specifically, the CGM cohort’s average A1c declined 1.1% to 8.0% at eight months while the BGM cohort’s average A1c declined 0.6% to 8.4%. This improvement in the BGM group is notable and is a reminder that the CGM cohort was compared to optimized care for the BGM group, suggesting that potentially, the real-world improvement with CGM could almost certainly be further enhanced compared to true “normal” BGM care (even with an endo!). While this adjusted average A1c improvement with CGM is promising, one-third of participants in the CGM cohort still did not achieve an A1c <8% at month eight. Furthermore, there was not a significant treatment difference in the eight-month A1c improvements among those with baseline A1c values ≥8.5% in the Dexcom G6 and BGM groups. During his presentation, Dr. Martens suggested that this could be due to the minimal change in medication usage and dose in the study and argued that potentially, more aggressive approaches to therapy management alongside CGM would incur greater benefits to patients with higher A1c values at baseline. We certainly (!!) think this is true – most should have, in our view, gone on GLP-1 or an SGLT-2 or prandial insulin immediately (and we note that a few did, but overall a small number). Those in the CGM cohort had a 26 mg/dl adjusted lower mean glucose value than did those in the BGM cohort at 179 mg/dl and 206 mg/dl, respectively. We felt it was notable, incidentally, that even in the ravages of COVID-19, the CGM group came in under the high end of the target range of 70-180 mg/dL.

  • Participants maintained what Dr. Martens characterized as high use of the CGM over the eight-month period (average CGM use of 6.1 days/week – this is good and we look for it to get better over time as systems improve) and were highly satisfied with the system with a CGM Satisfaction Survey score of 4.1 (1 = least positive; 5 = most positive). One participant in each group reported a severe hypoglycemic event (1% of the CGM group and 2% of the BGM group), which we don’t think is surprising - while we imagine this stemmed from an over-correction or a mistake in carb counting, and while this is speculation, numbers here were low, and equal in the arms. We imagine that severe hypoglycemia (SH) would’ve been seen much more in people on MDI treated only by BGM; CDC data from 2016 shows that nearly 300,000 people per year go to the ER for SH and nearly an equal number for DKA, as we recall.

  • Notably, the trial included a diverse sample with 47% of participants identifying as non-Hispanic White, a majority of patients with less than a college degree and without private insurance, nearly an equal proportion of men and women, and a baseline mean A1c of 9.1% – we are impressed that the race/ethnicity more closely the basic make-up of the US more than most trials we see. This is particularly true given the disparities in prevalence in type 2 diabetes and outcomes that are disproportionately seen in African-American and Hispanic populations, as well as Native American/Native Alaskan, the latter of which had relative under-enrollment. Those with lower socioeconomic status also are known to have disproportionately higher rates of diabetes. The trial recruited across 15 primary care clinics around the country from Amarillo, TX to Ann Arbor, MI to Henderson, NV. As we understand it, while MOBILE trial was originally seeking 205 or more participants, it slightly under-enrolled due to timeline – specifically, we believe a decision across the centers when it was found that more participants than originally expected were moving through the trial more quickly than expected. This slightly affected power of the study, again, as we understand it.

  • Not only were the results of the study impressive, but also the session being via Zoom rather than a webinar felt a little closer to an in-person conference with our associates once again tuning into hear the first read-out of exciting data alongside KOLs like Drs. Rich Bergenstal, Stephanie Amiel, Roy Beck, Satish Garg, Moshe Phillips, and Tadej Battelino, among multiple other key opinion leaders and 140 other excited attendees.

Eversense 180-Day PROMISE Study Shows Overall MARD of 9.1% and ±20/20% of 93%; Accuracy Maintained Out to 180 Days

Dr. Satish Garg (Barbara Davis Center) presented results from the PROMISE study, the 180-day pivotal study for Senseonics’ Eversense implantable CGM. The large, multi-center trial took place across eight clinical sites and enrolled a total of 181 adults with diabetes. In all, the 181 adults wore 279 sensors (96 subjects had a sensor placed in each arm and two sensors were replaced). People with type 1 diabetes made up 70% of participants; the remainder had type 2 diabetes. Participants took part in ten clinic visits over the 180 days, generating a total of 49,613 paired points that were compared against YSI reference. During those clinic days, patients had their glucose levels manipulated to ensure enough sampling at glucose concentrations ranging from 40 to 400 mg/dl. Up until day 21, the study required participants to perform two fingerstick calibrations per day; following day 21, calibration requirements fell to “primarily” one per day. At a glance, the accuracy data for 180-day Eversense are very strong and would likely support approval as a non-adjunctive CGM (with one calibration/day). Accuracy was also maintained out to 180 days. As a reminder, we heard last month that the FDA submission for 180-day Eversense was picked up in April with approval expected sometime in the back half of this year.

YSI glucose range (mg/dl)

Number of paired points

±20/20%

Mean Absolute Relative Difference (MARD)

Overall

49,613

92.9%

9.1%

40-60

2,281

89.4%

9.4%

61-80

5,270

92.2%

8.8%

81-180

19,001

90.9%

9.0%

181-300

14,578

94.7%

7.7%

301-350

6,862

96.5%

7.1%

351-400

1,510

93.9%

8.0%

  • Overall MARD for the Eversense 180-day sensor was 9.1%. This falls in line with previous remarks from Senseonics that the accuracy data of the 180-day and 90-day sensors were comparable. (The 90-day Eversense has a MARD of about 8.5%.) Impressively, the sensor recorded a MARD below 10% at all glucose ranges, including the most difficult 40-60 mg/dl range.

    • The sensor performed well in hypoglycemia, detecting blood glucose levels ≤70 mg/dl correctly 93% of the time. For the ≤60 mg/dl threshold, hypoglycemia detection was slightly lower at 87%. On the upper end, the Eversense sensor detected glucose levels ≥180 mg/dl correctly 99% of the time.

  • Over the sensor life, MARD varied somewhat, particularly at the beginning and end of wear. At day 1, MARD was 11%, but fell consistently to just 7.7% at day 60. After day 60, MARD began to increase again, reaching 10.4% by day 180. Of note, the reduced requirement for calibrations from two per day to one per day did not seem to affect MARD.


  • Overall sensor ±15/15% and ±20/20% were 86% and 93%, respectively. Similar to the MARD measurement, Eversense 180-day’s accuracy was lowest at the beginning and end of sensor life, around 89% and 90% within ±20/20%, respectively. The sensor also saw its highest ±20/20% figure at day 60 of 96%.

  • Of the 279 sensors, 43 had a modified sacrificial boronic acid chemistry designed to reduce oxidation of the glucose-sensing chemistry (i.e., improve sensor life). The data referenced above represent data from the non-modified sensors. The sensors with modified chemistry actually performed significantly better than the non-modified sensors, with overall MARD of 8.5% and ±20/20% of 94% on 12,034 paired points. The modified sensors also had much longer survival rates out to 180 days. The non-modified sensors had a survival rate of 65% at 180 days, compared to 90% for the sacrificial boronic acid sensors. According to Senseonics’ press announcement on the study results, data from both the modified and non-modified sensors were submitted to the FDA. Finally, the PROMISE study recorded two mild skin infections, but no other insertion/removal procedure related serious adverse events.

Guardian 4 Sensor Accuracy Data (With No Calibrations): MARDs of 10.6% (Arm Wear) & 10.8% (Abdomen), 88% Within ±20/20%; 99% Consensus Error Grids Zones A+B

Just after Guardian 4’s CE-Marking with no calibrations last week, Dr. Ron Brazg (University of Washington) presented accuracy data from Medtronic’s new CGM. The 15-center study, used as part of the CE-Mark submission, enrolled 169 adult participants (~65% with type 1 diabetes) who wore the Guardian 4 (same form factor and sensor as Guardian Sensor 3, with the updated “Zeus” algorithm) on both the arm and the abdomen. Participants were randomized to conduct four visits for frequent venous draws to generate paired points against YSI reference.

MARD for sensor at various wear locations and glucose ranges.

 

<70 mg/dl

70-180 mg/dl

>180 mg/dl

Abdomen wear

14.8%

10.4%

9.4%

Arm wear

12.6%

10.2%

10.5%

  • Overall sensor MARD with no calibrations was 10.6% for upper arm wear and 10.8% for abdomen wear. For abdomen wear, a total of 18,423 paired points were generated and for the arm, a total of 20,612 paired points were generated. Generally, accuracy improved over time from day 1 to day 7 – for abdomen wear, MARD was in the ~8% range on days 6-7 for glucose values >70 mg/dl. Overall ±20/20% was 88% for both abdomen and arm wear. These figures are almost identical to the numbers on Medtronic’s Guardian Sensor 3 label, though that device requires twice-daily calibrations.

  • Consensus error grid analysis showed 91% of readings in Zone A for abdomen wear and 88% of readings in Zone A for arm wear. For both the abdomen and the arm, 99.9% of readings came in Zones A or B.


Preliminary Pre-Pivotal Data on G7 Performance (n=91 Adults) Shows Strong Accuracy with Overall MARD of 8.7%; US Pivotal is “Well Underway”

To close out the morning Dexcom symposium, Dexcom Chief Technology Officer Jake Leach read out new G7 feasibility data from studies conducted prior to launching the US pivotal, which is now “well underway” and is anticipated to complete this month, as of early May. The two studies were conducted across four clinical trial sites and included 91 adults with type 1 or type 2 ages 19-76. Participants wore four G7 devices (two on the upper arm and two on the abdomen) and attended three in-clinic sessions on Day 1 or 2, Day 4 or 7, and Day 10. During the in-clinic sessions, participants’ blood glucose levels were manipulated between 40 mg/dl and 400 mg/dl, and G7 readings were compared to YSI readings every 15 minutes. For the 91 participants, the overall MARD for G7 was 8.7% with 94% of values within ±20/20% based on 46,619 data points from arm and abdomen wear. This is in line with feasibility data (n=98) presented at DTM 2020, which showed an adult MARD of 8.7% and 93% of values within ±20/20%. In the data that Mr. Leach presented today, the MARD was expectedly higher on Days 1-2 (10.2%), but then fell to 7.7% during Days 3-7 and 8.0% on Day 10. The percentage of values within ±20/20% also improved from 91% on Days 1-2 to 96% and 95% on Days 3-7 and Day 10, respectively. Importantly, 99% of G7 sensor readings were clinically accurate and safe for treatment decisions based on the surveillance error grid (shown below).

 

DTM 2020

ATTD 2021

N= (Paired points)

128 adults and children (43,547)

91 adults (46,619)

 

MARD

%20/20

MARD

%20/20

Overall

8.7%

93%

8.7%

94%

Days 1-2

9.4%

90%

10.2%

91%

Days 3-8

8.3%

93%

7.7%

96%

Days 9-12

8.6%

92%

8.0%

95%

Days 13-16

7.9%

96%

-

-

  • While today’s presentation did not include pediatric data, the feasibility data presented at DTM 2020 (n=30) suggested a pediatric MARD of 8.4% and 93.4% of values within ±20/20%.

  • Today’s feasibility data also did not break down the MARD by the wear location. DTM 2020 data suggested that both positions achieve strong accuracy with an upper-arm-wear MARD of 9.3% with 93% of values within ±20/20% and an abdomen-wear MARD of 9.1% with 91% of values within ±20/20%.

  • As a reminder, G7 brings a number of new features to Dexcom’s CGM franchise: (i) a one-piece fully disposable wearable (integrated sensor/transmitter) that is 60% smaller; (ii) a significantly lower cost design; (iii) an applicator that is smaller, lighter, less plastic, and more convenient; (iv) a shorter 30-minute warmup period; (v) an Android/iOS app that displays real-time data with Bluetooth and offers data insights to users; and (vi) direct Bluetooth connectivity with smartwatches. Today’s presentation was the first time we heard about the shorter 30-minute warmup time and the direct Bluetooth connectivity with Apple Watch. Although G7 is accurate out to day 16, Dexcom plans to initially launch G7 with 10-day wear (see Dexcom 3Q20) because of reliability issues due to adhesive problems (among other things). As of November, we heard that G7 was “about 90%” reliable out to 10 days. However, with “adhesive improvements” and perhaps other modifications, Dexcom is still aiming to get out to 14-16-day wear. At JPM 2020, Verily appeared to confirm for the first time that the G7 would also contain a built-in accelerometer, although later reports appeared to clarify that announcement, indicating that no formalized plans had been determined.

Using Dexcom G6 as a Diagnostic Tool for Diabetes Performs Comparably to Oral Glucose Tolerance Test

In a very interesting presentation, Dr. Spencer Frank, algorithm engineer at Dexcom, presenting feasibility results using Dexcom G6 as a tool for diagnosing diabetes. Of course, the traditional method for diagnosing diabetes is the A1c test with varying thresholds defined as normoglycemia, prediabetes, and type 2 diabetes. However, Dr. Frank noted that oral glucose tolerance tests (OGTTs) are often used to diagnose diabetes, particularly gestational diabetes, and OGTT results do not always agree with A1c. By looking at both mean glucose and glycemic variability using CGM data, Dr. Frank hypothesized that an algorithm would be able to classify patients’ profiles as normoglycemia, prediabetes, or type 2 diabetes. Aggregated Dexcom CGM data was taken from several previous studies (n=716 total sensor sessions), fed to the classification algorithm, and compared against the A1c measurement.

  • Overall, Dexcom CGM + the classifier algorithm performed similarly to oral glucose tolerance tests at diagnosing diabetes. Based on a single wear session (i.e., 10 days of data), the algorithm was able to diagnose diabetes with a sensitivity of 0.71 and specificity of 0.93. For reference, OGTT performs with a sensitivity and specificity of 0.77 and 0.99, respectively. The positive predictive value (i.e., the probability that someone diagnosed with diabetes actually has diabetes) was relatively low at 0.71; however, Dr. Frank believes this number can be improved by further algorithm enhancements. Importantly, the diagnostic algorithm performed well for those who clearly had or clearly did not have diabetes: sensitivity on patients with A1c ≥7% was 1 (i.e., the algorithm correctly identified all of the patients with A1c ≥7% as people with diabetes) and the specificity on patients with A1c <5.7% was 0.98 (i.e., the algorithm correctly identified 98% of patients with A1c <5.7% as normoglycemic).

  • Use of Dexcom CGM as a diagnostic tool could fit nicely within Dexcom’s broader CGM initiatives. Over the past few years, Dexcom has made significant investments in penetrating the newer type 2 non-intensive market with Dexcom Hello, an expanded sales team, and significant ad spend all in the last year. In the future, it’s possible we could see broader populations (e.g., all pregnant women, people with CVD, etc.) get a CGM during their clinic visit. Presumably, this would help catch a significant number of people with undiagnosed diabetes or prediabetes earlier in the disease progression – and the value, of course, that would come from the additional education for patients and HCPs is massive, as is, of course, the energy that would come to both groups as well as the authority (also for both!)

MOBILE Study Extension Phase Sneak Peak: CGM Discontinuation Associated With 2.9 Hour/Day Decline in Time in Range and 0.3% A1c Increase (n=53)

As promised two days ago, a morning Dexcom-sponsored symposium brought a sneak peak of results from the extension phase of the MOBILE study. As a reminder, results from the MOBILE study were presented on Wednesday at ATTD with the publication going live simultaneously on JAMA. The multi-center RCT included 175 individuals with type 2 on basal-only therapy from primary care practices. All participants were >30 years old, had their diabetes managed in primary care, used 1-2 long- or medium-acting insulin injections/day, had A1c values between 7.8% and 11.5% (average 9.1%), were CGM-naïve, and had used BGM at least three times/day for the month prior to screening. Following screening, participants were randomized to either use the Dexcom G6 CGM (n=166) or the One Touch Verio Flex BGM (n=59) for eight months. At eight months, the CGM cohort spent 3.8 more hours/day in range than did those in the BGM cohorts - with Time in Range values of 59% and 43%, respectively (adjusted difference of 16%; p<0.001). The CGM cohort saw a 0.4% adjusted A1c improvement over the BGM cohort. Specifically, the CGM cohort’s average A1c declined 1.1% to 8.0% at eight months while the BGM cohort’s average A1c declined 0.6% to 8.4%. Make sure to check out our full coverage of the primary study results here.

During the extension phase, half of the 106 completers in the CGM group continued on CGM, while the other half discontinued CGM and returned to SMBG. Of the 57 trial participants randomized to BGM initially, 55 of them continued into the six-month extension phase where they remained on BGM.

  • Time in Range in the group that discontinued CGM quickly fell from 62% at the end of the primary study to 50% after four months of CGM discontinuation. That represents a difference of about 3 hours/day. As shown in the graph below, the group that continued on CGM maintained their Time in Range seen at the end of the primary study and in fact, appear to have increased their Time in Range slightly at the end of the 14 months (see picture below).

  • A1c in the discontinue CGM group rose from 7.9% at the end of the primary study to 8.2% four months after discontinuing CGM. Note that this difference slightly missed out on statistical significance (p=0.06). In comparison, the group that continued on CGM had an A1c of 8.2% at the end of the primary study and an A1c of 8.1% at the end of the extension phase. Not too surprisingly, the BGM group had a consistent A1c at the end of the primary and extension phases with mean A1c coming in at 8.4% and 8.5%, respectively.

  • Generally, the CGM discontinuers had glycemic metrics closer to the BGM group than the CGM continuers. This is best demonstrated by the graphs below, which show the blue and green lines closer together, with the red line averaging higher Time in Range and lower mean glucose. This result is not necessarily surprising (people using BGM have similar results to people using BGM); however, the results also demonstrate, convincingly, that the benefits of CGM disappear when CGM is taken away.

Dr. Ilias Spanakis Provides Excellent Overview of New Research on CGM in Hospital Settings

During a morning Dexcom-sponsored symposium, Dr. Ilias Spanakis (University of Maryland) reviewed some of the newest literature evaluating CGM systems in inpatient settings. Interest in the topic was recently reignited, of course, by the COVID-19 pandemic and FDA’s temporary authorization for Dexcom G6 and Abbott’s FreeStyle Libre in the hospital over a year ago. As a result, Dr. Spanakis focused his presentation on studies performed in recent years, including many done in patients with diabetes and COVID-19.

  • Two recently published studies demonstrated that CGM could reduce hypoglycemia and hyperglycemia in general wards. The first, published by Singh et al. in Diabetes Care, was a randomized trial that randomized 122 people with type 2 diabetes to usual care and another 122 people with type 2 diabetes to be monitored using a CGM-based telemetry system. This system used the Dexcom G4 sensor transmitting data to a centralized dashboard on an iPad at the nurse’s station. The CGM group saw significant reductions in hypoglycemic events (0.67/patient vs. 1.7/patient) and decreased prolonged hypoglycemic events (>120 minutes below 70 mg/dl) to zero for the CGM group. Similarly encouraging results were seen in a study by Dr. Fortmann et al. published in Diabetes Care.

  • Two recent studies estimated MARD for Dexcom G6 in the hospital at 9.4% and 12.8%. The first study was published in Diabetes Care last year and enrolled ten adult patients with diabetes. In that study, 89% of paired glucose values were within the “no-risk” zone of the risk surveillance error grid. The second study, not yet published, was significantly larger, pooling data from three inpatient clinical trials across four sites (Emory University, Grady Hospital, Baltimore VA Medical Center, and University of Maryland). That study included data from 218 people with diabetes and calculated a MARD of 12.8%. Uses the consensus error grid, 81% of values fell into Zone A and 17.8% came in Zone B.

  • Looking at studies of CGM in people with diabetes and hospitalized with COVID-19, Dr. Spanakis pointed at a recent JDST publication by Sadhu et al. That study enrolled eleven patients, six wearing Medtronic’s Guardian Connect and five wearing Dexcom G6. Against point-of-care BGM reference (Roche Accu-Chek II), MARD for Medtronic Guardian Connect was 13.1% and MARD for Dexcom G6 was 11.1%. By consensus error grid, Medtronic recorded 74% of values within Zone A and 26% in Zone B. Dexcom recorded 88% of values in Zone A and 12% in Zone B. Outside of accuracy, Dr. Spanakis noted two studies (one to be published in JDST and Agarwal et al., Diabetes Care) demonstrating that CGM was valuable in helping to reduce point-of-care testing and personal protective equipment utilization. Finally, Dr. Spanakis noted a few special situations in which CGM did not appear to work properly: compression lows, hypoperfusion, and cardiac arrest (Davis et al., Diabetes Care).

Real-World Study of Dexcom G6 Users (n=31,034) Shows 86.6% Retention and +57 Minutes/Day of Time in Range Two Years After Transitioning from G5 to G6

During a Day #1 afternoon Dexcom Symposium, Dexcom’s Dr. Joost Van Der Linden presented data on the two-year retention and glycemic outcomes of Dexcom G6 users that switched from Dexcom G5 to G6 between June and December 2018. The study included tracked a cohort of 31,034 patients pre-switch, post-switch and at two follow-up periods (one year later in September 2019 and two years later in September 2020). The researchers evaluated the glycemic outcomes for a randomly sub-sampled group of 25,000 participants who had available data at all four timepoints.

  • Retention rates were found to be relatively high among participants with 90% of participants still using G6 after twelve months and 86% of patients continuing to use G6 at year two. Additionally, the study revealed improved and sustained CGM utilization, which was measured based on data sufficiency (the fraction of uploaded glucose values/the total number of possible glucose values during the time period) and persistence (the fraction of days that the patient uploaded at least one glucose value over total number of days in the study). The percentage of participants achieving data sufficiency >90% increased significantly by twelve months and was sustained out to two years at 64% of participants, 79% of participants, and 81% of participants at baseline, one year, and two years, respectively. Likewise, the percentage of patients achieving >90% persistence increased significantly after switching to G6 from 78% at baseline to 83% at twelve months and 84% at two years.

  • Time in Range improved 57 minutes/day from 57% at baseline to 61% at two years among the sub-cohort of 25,000 participants. Users also saw a small yet statistically significant 12 minute/day reduction in time in hypoglycemia (<70 mg/dl) after two years, falling to 2.5%. Time in hyperglycemia (>180 mg/dl) improved by 32 minutes day by year two, declining to 37%, which Dr. Van Der Linden noted could be attributed to the COVID-19 pandemic during which there was an improvement in average glycemic outcomes across the US.

Variability in Skin Temperature and Tonic Skin Conductance Positively Associated with Established Hypoglycemia (OR=1.38 and 1.22, Respectively)

Dr. Chukwuma Uduku (Imperial College London) presented data from their work indicating that data obtained from wearable physiologic sensors can be integrated in tandem with CGM data to “improve identification and prediction of hypoglycemia.” Specifically, Dr. Uduku gathered data from 12 adults with type 1 diabetes with a median age of 40 all of whom wore Dexcom G6 CGMs for the duration of the six-week longitudinal study. Participants also wore the Empatica E4 physiological sensor to collect data on heart rate, skin conductance, and skin temperature. According to Dr. Uduku’s data, variability in skin temperature and tonic skin conductance level were both positively associated with established hypoglycemia with odds ratios of 1.38 (p<0.01) and 1.22 respectively (p=0.04). However, minimum differential skin temperature – the difference between the smallest two adjacent points – and the variability in electrodermal activity were both negatively associated with established hypoglycemia with odds ratios of 0.47 (p<0.01) and 0.84 (p=0.04) respectively. Dr. Uduku’s group also saw significant associations between hypoglycemia and mean heart rate, heart rate variability, mean skin temperature, max phasic skin conductance responses, mean physical activity, and variation in physical activity (p<0.01 for all). However, the odds ratios for these associations were extremely small ranging from 0.96 to 1.04 leading Dr. Uduku to conclude they are unlikely to be clinically significant associations. That said, Dr. Uduku’s research presents evidence that physiological indicators have the potential to be used to identify impending hypoglycemia providing patients and clinicians advance warning allowing for earlier treatment and possible prevention of hypoglycemia.

Automated Insulin Delivery Highlights

First MiniMed 780G Real-World Data Shows Mean Time in Range of 76% (n=4,120), GMI of 6.8%; Pre-Post Analysis Shows +2.9 Hours/Day Following 780G Initiation (n=812)

Dr. Ohad Cohen (Medtronic Diabetes) presented our first look at real-world data from MiniMed 780G users, following the system’s launch in fall 2020. The results came from the first ~5,000 users of MiniMed 780G across eight countries, including the UK, Italy, Belgium, the Netherlands, Qatar, South Africa, Sweden, and Switzerland. The de-identified data was uploaded to CareLink between August 2020 and March 2021. Dr. Cohen presented results from 4,120 users with at least ten days of data using MiniMed 780G.

 

MiniMed 780G (n=4,120)

Time in Auto Mode

94%

Mean glucose

144 mg/dl

GMI

6.8%

Time in Range

76%

Time <70 mg/dl

2.5%

Time <54 mg/dl

0.5%

Time >180 mg/dl

21%

Time >250 mg/dl

4%

  • Overall Time in Range for MiniMed 780G users in the real-world was 76%. Accordingly, both time in hyper- and hypoglycemia were quite low, at least relatively speaking. Time <70 mg/dl was just 2.5%, which translates to 36 minutes a day and time <54 mg/dl was what one attendee described at “nearly non-existent” at 0.5%.  While we’d say, at 7 minutes a day or 50 minutes a week or 43 hours a year, nearly two days, that those with their glucose below 54 for that entire (interminable!) time would love to have that time back (as would their employers love for them to have it back) – but for people with diabetes on insulin, we’ll take that 0.5% for sure. On the hyperglycemia end, time >180 mg/dl was 21% and time >250 mg/dl was 4%. As we’ve seen with many closed loop system, Time in Range was particularly high at night (midnight – 6 AM), at 83%, compared to 74% during the day (6 AM – midnight). At 25% “time above range,” that’s pretty close to where guidelines want people with diabetes not to be above – but at seven hours a day, or a quarter of their lives – this tech has to be a path to a way to do better. And, 25% of time “high” truly is better, relatively speaking, for many people with diabetes.

  • Mean glucose for users was just 144 mg/dl, corresponding to a GMI of 6.8%. With 780G, a remarkable 79% of users had a GMI <7%. A slightly smaller proportion, 77%, had a Time in Range >70%. Finally, nearly three-quarters of users (74%) met both GMI <7% and TIR > 70% glycemic targets – wow! While we’d certainly expect these percentages to be the same, since that’s how the standards were designed, the fact that so many are at the standard (assuming hypoglycemia and severe hypoglycemia are together 5% or less of the time) is positive, since we so often hear that so many people with diabetes are not at target levels. Still, of course, 26% are still moving toward target – and the key part is that the closed loop is an easier way to make this possible for both people with diabetes and their healthcare teams, on average.

  • Users spent an average of 94% of time in Auto Mode. On average, the system recorded 0.9 Auto Mode exits per week (~one exit every eight days), with about half of those exits initiated by the user. This compares to the ~1.3 closed loop exits reported per week in the 780G pivotal trial and a massive improvement from the nearly once per day exits from Auto Mode seen with 670G. On average, it was said that users performed 3.4 fingersticks/day using the 780G system – we’re surprised RWD is so high.

  • Not too surprisingly, real-world outcomes with 780G were pretty uniform across countries. Interestingly, on average, Belgian users reported the highest Time in Range at 78%, while Finnish users reported the lowest at 74%. That’s about an hour more time in range for Belgium – for now, CGM isn’t widely reimbursed in Belgium and we wonder whether this made a difference and PWD perhaps used it with greater zeal or even more carefully according to their HCP! Across all eight countries, mean glucose came between 140-147 mg/dl and GMIs were all either 6.7% or 6.8%.

  • No wonder! We were very pleased to learn that in the real-world, users spent about half of the time with the most aggressive glucose set point of 100 mg/dl. We say we were pleased to see this because 1) we hear from a lot of users that they want to use more aggressive targets than some systems allow [there are still multiple safety parameters set that presumably will go down after time]; and 2) it seems like the changes to relax these standards could come sooner since so much safety data has now been amassed. As a reminder, MiniMed 780G allows users to choose between four glucose targets for basal rate modulation: 100, 110, 120, and 150 mg/dl. Users spent about a quarter of the time on the 110 and 120 mg/dl set points, with very few users, understandably, choosing the most conservative 150 mg/dl setting. The most popular active insulin time setting was 2.5-3 hours, the second most aggressive setting allowed by the MiniMed 780G system.

  • Notably, real-world Time in Range from the first ~4,000 MiniMed 780G users is actually slightly higher than that seen in the pivotal trial! Mean Time in Range in the MiniMed 780G pivotal trial was 75%, compared to 76% in the real-world data. For another comparison, real-world data from Europe on MiniMed 670G showed mean Time in Range around 71% after using Auto Mode on 670G.

  • In a separate analysis, Dr. Cohen presented results from 812 users who had at least ten days of data both before and after 780G initiation. At baseline, these users were using the MiniMed 780G system in open loop. Mean Time in Range for these users improved substantially, from 63% to 76% (+2.9 hours/day) with most of this improvement coming from reductions in hyperglycemia. Mean time >180 mg/dl declined from 34% to 22% with Auto Mode. Time in hypoglycemia was comparable (2% vs. 1.7%), unsurprisingly, since users were already on open loop. Mean glucose improved from 162 mg/dl to 146 mg/dl and correspondingly, mean GMI improved from 7.2% to 6.8%. While we are not yet at the point where GMI is seen as “synonymous” or “nearly synonymous” with A1c, many HCPs are seeing it as very close, with clinical utility trumping any accuracy issues. While 6.8% GMI won’t be the same always in everyone’s minds as 6.8% A1c, 100% of the time, we are getting the impression that a 6.8% GMI isn’t seen as likely to be significantly off a 6.8% A1c, with the advantage being that week to week improvement is very easy to identify once 14 days has passed.

  • After Auto Mode initiation, twice as many users achieved a GMI <7%. At baseline, 38% of users had a GMI <7%, compared to 75% of users after starting on closed loop. Similarly, the percentage of users meeting a >70% Time in Range target jumped from 35% to 75% after initiating closed loop, also known as Auto Mode.

  • Interestingly, Auto Mode initiation was associated with a 4.8 unit/day increase in total daily dose of insulin. As a reminder, MiniMed 780G not only includes basal rate automation, but can also deliver automatic correction boluses. After starting on closed loop, users delivered slightly less insulin manually (20 U/day to 18 U/day), but this was offset by 6 U/day of additional correction insulin given by the system. That means not only “correcting” highs, as we understand it, but just the system correcting “effective” basal rates.

New University of Virginia Algorithm, RocketAP, Delivers +30% Time in Range Over Control-IQ During Six Hour Period Following Unannounced Dinner (n=18)

Dr. Jose Garcia-Tirado (University of Virginia) presented promising results from the latest AID algorithm developed at UVA, RocketAP. Similar to the most well-known AID algorithm to come out of UVA, Control-IQ, the RocketAP algorithm is a zone-based model predictive controller. Most importantly, the RocketAP algorithm includes a new bolus priming system module that is designed to detect unannounced meals quickly and deliver bolus insulin before a lengthy hyperglycemia episode begins. The bolus priming system uses the last 30-minutes of CGM data and a logistic regression classifier (trained on previous datasets) to estimate the probability that a meal has occurred or is occurring. The actual amount of bolus delivered is dependent on the system’s confidence that a meal has actually been detected.

To assess the efficacy and safety of the RocketAP system, Dr. Garcia-Tirado’s team conducted a randomized crossover clinical trial of adolescent type 1s (ages 12-20). The short hotel study randomized participants to either the RocketAP or USS Virginia (the research version of Control-IQ) AID algorithms. The study used Tandem pumps and Dexcom G6 CGMs. Following two weeks of COVID-19-related protocols, the participants were started on either USS Virginia or RocketAP. On the first day, users bolused for meals; on the second day, users did not announce meals. Following crossover, users once again bolused for meals on day 3 and didn’t announce meals on day 4. Meals were kept identical.

  • For unannounced meals, RocketAP delivered +30% Time in Range during the six-hour post-dinner period (6 PM – midnight) compared to USS Virginia (i.e., Control-IQ). As shown in the graphs below, the two systems performed comparably when meals were announced; however, the RocketAP system (shown in blue) was clearly able to reduce the post-prandial excursion for unannounced meals.

    • Time in Range during the post-dinner period for unannounced meals was 83% for RocketAP, compared to 53% for USS Virginia. Achieving 83% Time in Range in the post-prandial period after an unannounced meal is a pretty stunning result, even given the small hotel setting of well-controlled type 1s. For announced meals, the RocketAP periods reported 100% Time in Range in the post-dinner period, compared to 93% for the USS Virginia periods (not statistically significant). Finally, the RocketAP periods recorded overall Time in Range of 87% and overnight Time in Range of 99%, compared to 80% and 92% for the USS Virginia period, respectively.

    • On average, the RocketAP’s bolus priming system was able to detect an unannounced meal ~10-15 minutes after the commencement of the meal. Impressively, in the small, controlled study, the algorithm did not accidentally deliver any boluses based on false positives.

  • In the past, Tandem has referenced ambitions for full closed loop control (i.e., an AID system with no meal announcements required). Given Tandem’s and Dexcom’s previous work with the University of Virginia on Control-IQ, it wouldn’t be surprising to see some version of RocketAP making its way to Tandem pumps eventually. For now, Tandem has focused on more incremental improvements to its popular Control-IQ system. As of 1Q21, the company had submitted two enhancements to the FDA to allow for “greater personalization in certain rates and ranges” and to add an indication for Sanofi’s biosimilar rapid-acting insulin Admelog (insulin lispro).

Initiating Children and Adolescents Straight From MDI to MiniMed 780G: Time in Range +8.8 Hours/Day to 79% (n=34); Half of Participants Were CGM-Naïve

During Medtronic’s sponsored symposium, Prof. Goran Petrovski (Sidra Medicine) presented very impressive results from a small study that initiated 34 children and adolescents (ages 7-17) straight from MDI to MiniMed 780G. The single-center, three-month study involved a 10-day protocol to train participants on insulin pumps and the MiniMed 780G system. At baseline, 18% of participants were using real-time CGM, 29% were using FreeStyle Libre, and over half (53%) were not using CGM at all. At baseline, mean A1c was 8.6% - this baseline is notable, in our view, as it shows that it takes more than tech – while many were on CGM already, and while their baselines certainly may have improved from the previous baseline, an overall baseline A1c of 8.6%, by any definition, is weaker than any system would like.

 

MDI+CGM (7 days)

MiniMed 780G (3 months)

Mean glucose

198 mg/dl

138 mg/dl

A1c

8.6%

6.5%

Time in Range

42%

79%

Time <70 mg/dl

3.2%

2.8%

Time <54 mg/dl

0.8%

0.5%

Time >180 mg/dl

55%

18%

Time >250 mg/dl

27%

5%

  • Time in Range increased by a remarkable 8.8 hours/day after three months on MiniMed 780G. Time in Range at baseline was 42% compared to 79% after three months of 780G. While it’s fashionable now to say, “this improvement was driven almost entirely by reductions in time in hyperglycemia” we also point out that even 0.5% time below 54% is severe hypoglycemia, and any time there for anyone with diabetes is likely too much for them – as such, for the system to improve 0.3% per week in severe hypoglycemia may not sound like a lot, that’s about 30 minutes a week, which is half an hour on average that anyone with diabetes would likely love to reclaim. Still, the hyperglycemia is impressive, also by virtually any definition: Time >180 mg/dl was cut in a third from 55% to 18%, down a whopping 8.7 hours/day. Time in severe hyperglycemia, which is over 250 mg/dl, was cut from an astounding 27% to 5%, down a whopping 5.2 hours/day. Forgive the adjectives on our part – this is just dramatic change.

  • After three months on MiniMed 780G, mean A1c was cut by over 2 percentage points. At baseline, mean A1c was 8.6%; after three months on 780G, mean A1c was 6.5%. With 780G, nearly four-fifths (79%) of participants met the A1c <7% target. Similarly, mean sensor glucose was reduced from 198 mg/dl to 138 mg/dl using the MiniMed 780G system. We are so happy for all these patients.

  • The improvements in Time in Range appeared quickly, with Time in Range near 70% just one week after 780G initiation. Starting from baseline (MDI+CGM), Time in Range increased from 42% to 54% with the MiniMed 780G in open loop. Following Auto Mode initiation, Time in Range increased further to 67% during days 1-3 and hit 70% by the end of the first week. Mean Time in Range continued to increase slowly, tapering around 80% Time in Range during the last few weeks of the study.

  • The participants in the study were brought onto the MiniMed 780G system following a ~10-day long initiation protocol. On day 1, participants were introduced to the system through a 90-minute session which included a discussion of expectations and responsibilities with the system. That session was followed by four days of two-hour daily group sessions of training on the MiniMed 780G system. Participants were initiated on CGM at this point to gather baseline data. After a two-day break (which, we wonder if this was to keep the average TIR on the lower side – that is speculation), on day 7, participants began using the MiniMed 780G pump in open loop with CGM review to help set insulin-carb ratios, active insulin time, and glucose set point settings. Finally, on day 10, participants began to use MiniMed 780G in Auto Mode, which they continued on for the next three months of the study.

  • MiniMed 780G also drove significant increases in Diabetes Treatment Satisfaction Questionnaire scores (DTSQ) scores for both patients and parents. DTSQ increased from 3.6 to 4.6 for patients and 3.5 to 4.8 for parents. This seems like impressive movement for both. On specific topics, treatment satisfaction for parents increased from 3.6 to 5.2; effect on parents’ lives improved from 3.5 to 5.1 – wow! (6.0 is the highest score.) Perceived frequency of hyperglycemia fell from 4.1 to 3; and perceived frequency of hypoglycemia grew slightly from 2.4 to 2.6, suggesting the hypoglycemia unawareness was starting to fall for the group at least a bit. At three months, mean “time in Auto Mode” (which just means that the system is “on” and “working” – not the Time in Range) was 92%, suggesting users learned to use the system well and were wearing it most of the time. The mean number of Auto Mode exits per week was just 0.5. Finally, the mean number of fingerstick calibrations performed per day fell from 5.5 during the first two weeks of 780G to 3.4 after three months. That’s still a lot in our view.

  • Though a small study, the very impressive results suggest nearly everyone using rapid-acting insulin can benefit from MiniMed 780G, regardless of technology experience at baseline. At ADA 2020, the FLAIR trial, which also looked at MiniMed 780G, enrolled a similarly broad study population and saw strong results. It’s encouraging to see more research on AID systems in these tech naïve groups. While some said the ten-day initiation protocol used in this study seemed fairly intensive, the improvement in glycemic outcomes is very striking – we’re excited to see this confirmed in larger numbers of patients.

Post-Hoc Analysis of Insulet’s Omnipod 5 Pivotal Trial Shows Omnipod 5 is Highly Effective at Reducing Hypoglycemia in Patients with a Lot of Hypoglycemia

The busy Dr. Bruce Buckingham (Stanford) read out post-hoc analysis results from the Omnipod 5 pivotal trial, which highlighted the advantages of Omnipod 5 over open loop in a new way. The full results for the Omnipod 5 pivotal trial were read out at ENDO 2021 with closed loop delivering an additional 2.2 hours/day Time in Range (74% vs. 65%). This post-hoc analysis looked at a subset of participants with type 1 diabetes (n=83 adults and 89 children) who had three months of data using Omnipod 5 in manual mode (i.e., open loop) in between phases of closed loop usage. This crossover-style study design was actually not intentional, as the study protocol initially aimed to gather two weeks of open loop at baseline, then have participants spend three months using the Omnipod 5 hybrid closed loop system. You might recall that the trial was paused in early March 2020 and resumed in June 2020 due to a software anomaly. As a result, participants had to effectively “crossover” from closed loop back to manual mode in March and then cross back over to closed loop in June 2020 to finish out the study.

  • For children, Time in Range increased from 52% to 69% in the first AID phase, declined back to 56% during the manual mode period, and jumped back to 68% during the second AID phase. The data for adults was similar with Time in Range rising from 63% to 74% from baseline to the first AID phase, declining to 66% in manual mode, and jumping back to 74% during the second AID phase.

  • The Omnipod 5 system did an excellent job of bringing down outliers’ time in hypoglycemia in both children and adults. As seen in the charts below, mean rates of hypoglycemia in the study were quite low to begin with. A full 87% of pediatric participants and 72% of adult participants were meeting the consensus target of <4% time below 70 mg/dl at baseline. However, the charts below also show a number of outliers with higher time in hypoglycemia in the standard therapy phase. One adult participant reported nearly 20% time <70 mg/dl in the standard therapy phase. These outliers saw their time in hypoglycemia drop significantly when in closed loop, at or below ~5%, but the outliers appeared again during the manual mode phase of the study. These results suggest that Omnipod 5 is highly effective at bringing people with a lot of hypoglycemia into better glucose control. 

Data for pediatric participants, ages 6-<14 (n=89)

 

Baseline

First closed loop phase

Manual mode phase

Second closed loop phase

Mean number of days

14

46

106

48

Mean glucose

184 mg/dl

158 mg/dl

177 mg/dl

161 mg/dl

Time in Range

52%

69%

56%

68%

Time <54 mg/dl

0.4%

0.3%

0.4%

0.3%

Time <70 mg/dl

2.1%

1.8%

2.3%

1.6%

Time >180 mg/dl

46%

29%

42%

31%

Data for adult participants, ages 14-70 (n=83)

 

Baseline

First closed loop phase

Manual mode phase

Second closed loop phase

Mean number of days

14

42

107

51

Mean glucose

164 mg/dl

153 mg/dl

160 mg/dl

154 mg/dl

Time in Range

63%

74%

66%

74%

Time <54 mg/dl

0.7%

0.3%

0.4%

0.2%

Time <70 mg/dl

3%

1.3%

2.5%

1.3%

Time >180 mg/dl

34%

24%

32%

25%

Control-IQ Associated with +9% Time in Range Among Earliest Adopters (First 4 Weeks After Launch) and +11% Time in Range Among “Late” Early Adopters (3-6 Months After Launch)

Dr. Lars Mueller (UC San Diego) presented data from early adopters of Control-IQ highlighting differences in glycemic trends between adopters in the first month (n=4,122) of Control-IQ availability and those who initiated the system 3-4 (n=1,617) and 5-6 (n=494) months after launch. Data from 6,233 Control-IQ users were included in this analysis and all participants were prior Tandem pump users with at least 30 days of pre- and post-Control-IQ pump data. Notably, adopters within the first month of Control-IQ launch had higher baseline Time in Range at 64% compared to adopters in the 3-4- and 5-6-month periods (both had an average baseline Time in Range of 59%).

  • After using Control-IQ for six weeks, users across all groups showed significant Time in Range increases. Adopters from the 3-4- and 5-6-month periods saw the most significant increase at +11% Time in Range; first-month adopters saw an average Time in Range increase of 9%. That said, it is still important to note that the earliest adopters of Control-IQ (within the first month) still had the highest Time in Range after 6 weeks on the system at 75%. The other two groups finished with similar Time in Ranges at six weeks of ~71%-72%. The Time in Range gap between the earliest and later adopters decreased slightly from 5% at baseline to 4% after six weeks on Control-IQ. Of course, early adopters for Control-IQ represented the most highly engaged patients. While adopters from the 3-4- and 5-6-month time periods may more accurately represent the broader population of Control-IQ users, we would be interested to see data from users initiating the system over six months after its launch as we feel 3-6 months post launch still represents a relatively early phase in the adoption of a novel automated insulin delivery system.

Medtronic Pipeline Includes Smartphone Bolus Feature for MiniMed 780G and Project Janus Smartphone App for Further MDI Decision Support

Mr. David Dunleavy (Medtronic Diabetes) presented a brief update on Medtronic’s pipeline, revealing a couple new projects we hadn’t heard about before. As a reminder, in the near term, Medtronic just received CE Marking for its Guardian 4 sensor, which requires no fingersticks for calibration or diabetes treatment decisions, and InPen connected pen in Europe last week. On the US end, MiniMed 780G and Guardian 4 remain under FDA review.

  • Medtronic aims to build smartphone “remote” bolus capability for its MiniMed 780G system. As a reminder, Medtronic’s 700 series pumps are the first with built-in Bluetooth capability and with the launch of MiniMed 780G in Europe and 770G in the US, Medtronic also launched the MiniMed Mobile smartphone app. Currently, that app allows for secondary display, remote monitoring, and data uploading and presumably, the future remote bolus capability would also be built in – we certainly expect all systems to have this eventually. While smartphone pump control could quickly become a common feature across insulin pumps, as Roche launched its mySugr Pump Control feature in Europe last month and Tandem and Insulet both have smartphone pump control features under FDA review, we’ll see how fast this can happen in the EU, where the regulatory environment is becoming more burdensome as we understand it. 

  • In addition to its next-gen Synergy CGM, Medtronic is also building a personalized dosing decision support app under the name Project Janus. Of course, Medtronic acquired Companion Medical back in August, which gives the company an extremely strong start in building out a full dose decisioning system for MDI + CGM users and a very smart management team as the insulin delivery goals with smart pens can expand substantially. Though not described in detail, the Project Janus app will utilize CGM, machine learning, and Medtronic’s acquisitions, like Nutrino and Klue, purchased in 2018 and 2019, to help provide bolus reminders and more detailed insulin dose decision support and/or food support. The CGM + MDI world has also become an area to watch over recent months with the recent FDA clearance of Bigfoot’s Unity system, as well as recent moves from Lilly, Novo Nordisk, and others.

CamAPS FX AID System Updates: Clinical Studies to Be Read Out at ADA 2021 and EASD 2021; Ongoing Studies on Use in New-Diagnosed Adolescents, Pregnant Women, the Elderly, and Type 2s

During a morning Dexcom Symposium, Dr. Roman Hovorka (University of Cambridge) discussed the CamAPS FX AID system, recent study results, and plans for future clinical studies. We haven’t received an update from CamDiab in quite a while, and we were very glad to hear there would be an opportunity to receive an update from Dr. Hovorka during the Dexcom Symposium. In fact, we last heard about CamAPS FX at DTM 2020 in November when we heard not from the company, but from a CamAPS FX user on their experience during a panel of patients using CamAPS FX, Tandem’s Control-IQ, Medtronic’s MiniMed 780G, and Insulet’s Omnipod 5. Today, we learned much more about the company’s recent efforts and plans for future studies. Dr. Hovorka shared that five studies have recently been completed, including studies on: (i) four-month use in children ages two to seven; (ii) six-month use in children and adolescents; (iii) use in children and adolescents ages six to eighteen; (iv) tw0-month use in adults comparing Fiasp vs. Aspart; and (v) two-week use of the fully-closed loop system (no bolusing) in adults with type 2 on dialysis. CamDiab intends to read out the data from the study on use in children and adolescents (ages 6-18) and the two-week study of the fully-closed loop system (no bolusing) in adults with type 2 on dialysis at ADA 2021. We’ll also hear the results of the four-month study on use in children ages 2-7 at EASD 2021 in September.

We are particularly excited about four ongoing studies:

  • four-year use in newly-diagnosed 10- to 17-year-olds; along with most of the rest of those watching diabetes closely, we are very eager to see data with such a the long-term focus and are very curious to hear more about study design, particularly what all the endpoints will be;

  • four-month use in adults ages 60+, an understudied population in AID (we are also very interested to learn more about the 80+ age group, where we are hearing there are still lots of newly diagnosed people with diabetes);

  • a study on use in pregnant women throughout pregnancy led by world pregnancy and diabetes expert Kings College’s Dr. Helen Murphy; and

  • a two-month study on a “fully closed loop” (i.e., no bolus) CamAPS FX in type 2s. While the timeline for these ongoing studies was not disclosed, we simply cannot wait to hear the results!

  • Dr. Hovorka briefly discussed data from an eight-week crossover RCT comparing CamAPS FX with Aspart to CamAPS FX with Fiasp, which was recently published in Diabetes, Obesity, and Metabolism. The multicenter study included 25 adults with type 1 who had an average A1c of 7.4% at baseline. While the researchers did not observe a difference in Time in Range between Fiasp and Aspart, Fiasp was associated with four fewer minutes/day in hypoglycemia compared to Aspart use. All of this benefit was observed in the severe hypoglycemia range (<54 mg/dl), which Dr. Hovorka argued makes the finding “clinically significant.” With both insulins, participants were in Auto Mode 95-96% of the time and had a GMI of 6.8% after eight weeks, a significant 0.6% reduction from baseline. At the minimum, this suggests that CamAPS FX used with ultra-rapid-acting Fiasp is noninferior to CamAPS FX used with rapid-acting Aspart, somewhat important given that CamAPS FX is the only AID system approved for use with ultra-rapid-acting insulins like Fiasp.

  • As a reminder, CamDiab’s CamAPS FX was CE-Marked for use in type 1s in March 2020. Dr. Hovorka argued that CamAPS FX has one of the widest usage approvals of an AID system with a wide range of approved total daily doses (5 U/day – 350 U/day) and a wide range of approved weights (10 kg – 300 kg). With this CE-Marking, CamAPS FX also became the first AID system approved for use with smartphone control and the only AID system available for pregnant women and children down to one year old. The system launched in the UK following its CE-Marking and was available to patients in 11 NHS clinics, where certified trainers are available, and had not yet secured reimbursement. We haven’t received an update on its availability in the UK since then, nor have we learned more about a broader European rollout or any plans to submit the system to the FDA.

  • The CamAPS FX system uses an Android app (including smartphone pump control), Dexcom G6, and either the SOOIL Dana Diabecare RS pump or the Dana-i pump. The algorithm modulates insulin delivery every 8-12 minutes, requires a “simple setup” of body weight and total daily dose, and is “highly adaptive” based on daily insulin needs, diurnal insulin needs, postprandial insulin needs independent of pump basal, insulin-to-carb ratio, and correction factor settings. Notably, the algorithm is housed on the app, which means that unlike Omnipod 5 or Control-IQ, CamAPS FX users need to keep the app in range of the system to stay in closed loop. On the app, users can set a personalized glucose target and turn on “boost” mode (i.e., more insulin) and “ease-off” mode (i.e., less insulin during exercise). CamDiab is working to add a “add meal” mode that advises the algorithm of snacking/grazing, a hypo treatment, or a slowly absorbed meal. The app also includes a bolus calculator, sends glucose value alerts, and allows users to control insulin delivery from their phone. Data can be sent automatically from the app to Glooko and to Diasend, allowing for efficient data sharing with providers and caregivers. Based on today’s presentation, it seems that integration into Dexcom Clarity is expected this year.

ATTD Tech Fair: AMF Medical’s Sigi Patch Pump, Capillary Bio’s SteadiSet Infusion Set, and Diabeloop’s DBLG-1 and DBL-hu AID algorithms

As always, ATTD’s Tech Fair brought together a number of smaller startup companies to give short ~ten-minute presentations on their technologies. Our highlights from the Tech Fair were AMF Medical, Capillary Bio, and Diabeloop, companies that are in their early stages, but already have some name recognition (certainly, we’ve covered all three companies in the past). All three companies have exciting technology in their pipeline with implications for future automated insulin delivery systems.

AMF Medical

Swiss-based AMF Medical is working on a new tubeless patch pump called Sigi. The pump consists of a durable portion that lasts for four years and is clipped into a disposable patch. Notably, Sigi is compatible with standard, pre-filled insulin cartridges, eliminating all reservoir filling manipulations and making pump changes easier. During his presentation, Product Development Director Dr. Pierre Fridez focused on three topics, Sigi’s pumping mechanism, its fast occlusion detection, and high accuracy and precision. AMF Medical came out of stealth mode back in February and as of our last update, the company aims to complete product development by late 2021/early 2022 and start testing the device with people in 2022.

  • Dr. Fridez described Sigi as a “double pump,” with a traditional piston pump connected to a custom micro-dosing pump mechanism. The first pump works like a traditional piston pump, though it has a flexible curved piston design to help minimize the patch pump’s footprint. The piston pump pushes insulin to the micro-dosing pump, which is responsible for actually delivering the insulin. The micro-dosing pump consists of two valves and a small 0.25 uL chamber. The valves and chamber operate in a canal-like fashion to deliver insulin in small, 0.025 IU pulses (for U100 insulin). Dr. Fridez highlighted that the micro-dosing pump consists of two stainless steel needles, one stretch of fluoropolymer tubing, and a cannula – all standard materials that have been well-tested with humans.

  • As a result of the micro-dosing design in the Sigi pump, Sigi is able to deliver insulin with more precision and accuracy and quickly detect occlusions. On occlusion detection, a magnet attached to the micro-dosing pump is able to detect abnormal pulsing very quickly, sending a signal back to the pump controller when an anomaly has occurred, shown with a video demo detecting an occlusion in less than 5 seconds. Additionally, Dr. Fridez showed a trumpet curve comparing the accuracy of various insulin pumps at delivering variable pump volumes. Naturally, the patch pumps had challenges in delivery precision for very small insulin volumes; however, with the micro-dosing design in the Sigi pump, Dr. Fridez noted that Sigi performs closer to tubed pumps like MiniMed 780G and t:slim X2 in these small volume ranges and significantly better than Omnipod or Accu-Chek Solo.

Capillary Bio

Capillary Bio was on hand at the Tech Fair to demonstrate its SteadiSet insulin infusion set with the SteadiFlow cannula. The SteadiFlow cannula is made of a softer material than traditional cannulas and also includes a coil inside that is designed to reinforce the cannula and prevent kinking. Additionally, the cannula has three side holes allowing delivery of insulin to continue even when the tip of the cannula is occluded. An image of this is shown below, in which the tip of the cannula is occluded, but three green blooms are visible, demonstrating that insulin is still being delivered from the cannula’s side holes. Studies in pigs have also suggested that SteadiFlow’s angled insertion significantly reduces the amount of inflammation compared to traditional 90-degree insertion cannulas. This could be particularly important for maintaining consistent insulin delivery and absorption for longer periods of wear (e.g., seven days). Capillary Bio has now completed all of its animal studies and we’ll see results from a feasibility study of SteadiSet with 20 participants at ADA 2021. Another pilot study for SteadiSet is currently enrolling and plans for a pivotal trial are in the works.

Diabeloop

Diabeloop CEO and Founder Erik Huneker gave the company’s Tech Fair update and we especially enjoyed the last half, which Mr. Huneker dedicated to the recently CE-Marked DBL-hu algorithm for people with highly unstable diabetes (a.k.a. “brittle diabetes”). As a reminder, Diabeloop is also just off a fresh new partnership with Roche and its flagship DBLG1 AID algorithm is now available for Dexcom and Roche Accu-Chek Insight pump users. People with highly unstable diabetes are estimated to make up just ~1% of all type 1s, but as Mr. Huneker noted, these people often have the most difficulty managing their diabetes. The DBL-hu algorithm, the first AID algorithm approved for use in this population, is built on the DBLG1 AID algorithm with a few modifications related to customizability, aggressiveness, and hypoglycemia minimization. For example, the standard insulin action speed setting in DBLG1 is 55 minutes; in DBL-hu, this setting can be customized from 20-100 minutes and is set at 75 minutes by default. Other newly adjustable settings in the DBL-hu algorithm include post-meal aggressiveness, digestion speed predictability, and hypoglycemia prediction horizon. With the addition of all these new settings, Mr. Huneker did note that initiating a person with highly unstable diabetes on DBL-hu requires much more work and closer collaboration with the Diabeloop team than the normal DBLG1 system. Still, most of these settings are largely fine-tuned within ~2 weeks and the early results have been quite impressive. In the small, seven-person study used to support CE-Marking, the DBL-hu system drove a massive +7.2 hour/day improvement in Time in Range (from 43% at baseline to 73%) during the trial. Importantly, %CV was also greatly improved from 32% to 27%.

Type 1 Omnipod DASH Users See 0.9% (Adult, n=2,416) and 0.8% (Pediatric, n=1,020) A1c Reductions in First 90 Days of Use, Even More Significant Decline for Those on MDI at Baseline or with High Baseline A1cs

Dr. Grazia Aleppo (Northwestern) read out the results of a real-world retrospective study on the first 90 days of Omnipod DASH use in pediatric and adult type 1s in the US (n=1,020 pediatric; n=2,416 adult). Pediatric participants saw a 0.8% decline in their A1c to 7.8% at day 90 while adults saw a 0.9% decline to 7.6% (p<0.0001 for both). The proportion of participants achieving an A1c <7% increased from 20% at baseline to 31% at day 90, and while the proportion with A1c values >9% fell from 32% to 14%.  MDI patients, who made up the majority of the study sample (63% of adults and 80% of pediatric participants), saw even greater A1c improvements. Pediatric patients on MDI at baseline saw a 0.9% A1c drop from 8.6% to 7.7% while those who had previously been on pump therapy only saw a nonsignificant 0.3% A1c decline from a baseline of 8.5%. Likewise, adult patients who had previously been on MDI saw a 1.0% A1c drop from 8.6% to 7.6% while those who had previously been on pump therapy saw a 0.6% A1c decline from 8.0% to 7.4% (p<0.0001).

  • Across pediatric and adult participants, those with higher A1c values at baseline saw significantly greater improvements in A1c. Participants with baseline A1c values >9% (n=1,110) saw a 2.3% decline in A1c to 8.5% after 90 days. This is compared to a 0.5% decline to 7.8% among those with baseline A1c’s 8% - 9% (n=735), a 0.2% A1c drop to 7.2% among those with baseline A1c values 7% - 8% (n=901), an a 0.1% A1c rise in those with A1c values <7% at baseline (n=690), all of which were statistically significant changes.

  • Self-reported hypoglycemic events fell both among pediatric patients and adults with pediatric hypoglycemic events falling 1.4 episodes/week to 1.4 episodes/week, and adult hypoglycemic events falling 1.6 episodes/week to 1.4 episodes/week. When stratified by MDI vs. pump at baseline, all sub-groups saw significant declines in hypoglycemic events per week, except for pediatric patients on pump therapy at baseline.

  • Adult participants also saw a significant decline in their total daily dose, which dropped 13 U/day to 51 U/day (p<0.0001). However, pediatric participants’ average total daily dose did not change significantly from a baseline of 32.4 U/day to an average of 31.5 U/day at day 90. Notably, when stratified by MDI vs. pump at baseline, only adults on MDI at baseline saw a significant decline in total daily dose, dropping 15 U/day to 50 U/day at day 90. The changes in total daily dose were not significant among pediatric participants on MDI or pump at baseline nor among adults on pump therapy at baseline.

A1c Outcomes

 

Pediatric (n=1,020)

Adult (n=2,416)

Baseline

Change

Baseline

Change

Total cohort

8.6%

0.8%*

8.5%

0.9%*

MDI at baseline

8.6%

0.9%*

8.6

1.0%*

Pump at baseline

8.5

0.3%

8.0

0.6%*

*Significant change from baseline (p<0.0001)

Control-IQ Observational (CLIO) Study: Control-IQ Users (n=700) See Time in Range of 71%, Time <70 mg/dl of 1.6%, Time >180 mg/dl of 27%, and Significant Quality of Life Improvements at Three Months

In an afternoon oral presentation, Dr. Steph Habif (Tandem) read out interim results from the ongoing real-world Control-IQ Observational (CLIO) Study. The readout included data from 700 type 1s on Control-IQ (59% female, 87% White, age 39±17) who had 21+ days of pump data on t:connect and ≥75% CGM use during the first three months of the study. At baseline, 89% of participants were already CGM users and 80% were already pump users. At three months, participants saw a mean Time in Range of 71% and spent 1.6% of time <70 mg/dl and 27% of time >180 mg/dl, all improvements from baseline (unfortunately, Dr. Habif did not present baseline values to get a better understanding of the degree of improvement – we have in a request to see if this is possible to get). This is terrific and in line with the results from the Control-IQ pivotal (ages 14-71) that was read out at ADA 2019 and with one-year real-world data that was published in DT&T. In the Control-IQ pivotal, participants (n=168) had a Time in Range of 70%, a time <70 mg/dl of 1.9%, and a time >180 mg/dl of 27% at six months, improvements that were achieved during the first month and sustained over six months. In the one-year, real-world study, Control-IQ users (n=9,451) saw a +2.4 hour/day increase in Time in Range to 74% and a -1.2 hour/day drop in time >180 mg/dl, and maintained a 1% time below 70 mg/dl.  These CLIO Study findings provide further evidence that Control-IQ is a highly effective system in real-world settings and further validate the system’s consistent value across various populations and time frames. The only notable difference between this data set and that of the one-year, real-world study and the pivotal study is the time in Auto Mode, which was only 86% in this dataset compared to 94% and 92% in the real-world and pivotal trials, respectively.

 

Three-Month, Real-World CLIO Study

One-Year, Real-World Study

Six-Month iDCL Pivotal Study

At three months

At one year

Change from baseline

At six months

Change from baseline

Time in Range

71%

74%

+2.5 hours/day

70%

+2.6 hours/day

Time <70 mg/dl

1.6%

1%

+0 hours/day

1.9%

-13 min/day

Time <54 mg/dl

0.3%

0.15%

+0 min/day

0.21%

-1 min/day

Time >180 mg/dl

27%

19%

-1.2 hours/day

27%

-2.4 hours/day

Time >250 mg/dl

7%

6%

-49 hours/day

-

-

  • The CLIO Study also shows significant improvements in participants’ quality of life after three months on Control-IQ. Participants reported greater satisfaction with their insulin delivery device and its impact on their diabetes management (7.06 vs. 8.77, baseline and three months respectively, p<0.001) and a significant reduction in the perceived negative impact of diabetes across all dimensions (4.79 vs. 3.26; p<0.001). Likewise, nearly all participants (96%) reported improved sleep quality after three months of Control-IQ use with the percentage of participants reporting good or very good sleep increasing from ~20% at baseline to ~55% at three months.

Dr. Laurel Messer Argues Current AID System Training is “NOT Standardized, NOT Consistent, and NOT Sufficient,” Should Include Comprehensive Pre-AID Education and Strategic Early Clinical Follow-Up

Barbara Davis’ Dr. Laurel Messer argued that current AID system training is not adequate to meet patients’ needs when initiating AID system use: “It is NOT standardized, NOT consistent, and NOT sufficient.” During her Day #1 presentation, she suggested that what is needed is comprehensive pre-AID assessment and education that precedes AID training, along with strategic post-AID early clinical follow-up. During pre-AID education, clinicians should assess and build upon previous knowledge, offer refreshers on the basics, support patients in choosing the best system for them, and help patients set reasonable expectations for their AID system. Dr. Messer argued that clinical follow-up is hugely important for two practical reasons: (i) device discontinuation often occurs within the first three to six months of use; and (ii) suboptimal device use results in suboptimal glycemic control. These two issues are also related, as discontinuers have worse glycemic and device use outcomes at one month than non-discontinuers based on this study from Messer et al., suggesting that improving early experiences and outcomes with AID systems might prevent discontinuation down the line.

  • During pre-AID foundations training, education should build on previous education and should establish expectations for what AID systems can and cannot do. As Dr. Messer put it, “it’s not revolutionary, but it bears repeating. Core diabetes knowledge needs to be assessed and implemented.” For those new to insulin pumps and CGMs (and those who need a refresher), pre-AID education should covers the important basics of these system components. Prior to device training, providers should also support patients in choosing the best system for them, in understanding what to expect with AID systems and how AID systems differ from open-loop insulin therapy, and in articulating the importance of early follow-up with the clinical team.

    • On helping people with diabetes chose the best system for them, Dr. Messer argued that this is a core responsibility of a diabetes clinician and should not be left to industry representatives. She underscored the importance of providers taking time to present the options, ask questions, share patient experiences, and assess the system that best matches patients’ needs and preferences prior to ordering a device. Per Dr. Messer, clinicians should discuss and consider five factors in supporting patients’ AID system choice: (i) lifestyle, which impacts hardware and software component choices; (ii) education to ensure that a patient’s literacy and learning processes are matched to the system; (iii) current and previous device experience; (iv) support resources available from the clinic, industry representatives, etc.; and (v) insurance/payer coverage and cost. 

    • Dr. Messer spent a large portion of her presentation arguing that establishing expectations is core to pre-training education. Specifically, she argued that both clinicians and industry members should set expectations for patients around: (i) the learning curve; (ii) the workload and actions needed to keep the system working; (iii) giving up some personal control to the system; (iv) expected glycemic outcomes; and (v) the need to maintain basic diabetes self-management skills. On the learning curve, patients should understand that the system takes 1-2 weeks to adapt the algorithm to their total daily insulin needs, and users take 3-4 weeks to “get the hang of” using AID features, so patients should consider the first few weeks a “getting to know you” phase before forming a full opinion on the system. Per Dr. Messer, these conversations should take place before, during, and after AID initiation.

  • During early follow-up sessions, clinicians should assess and address patients’ risk factors for discontinuation. Per Dr. Messer, these sessions should be held in the first two to six weeks of AID use via any modality (phone, video conference, in-person), but should always include data review. She argued that the goal of these sessions should be device optimization by assessing system use, optimizing insulin dosing and behavior, providing troubleshooting and education, and reinforcing expectations. Below summarizes the risk factors for discontinuation that were identified by Messer et al. and how to address the issues. 

  • Dr. Messer offered her own clinic’s model as a case study for how strategic early follow-up can be implemented. At the Barbara Davis Center, 199 youth were onboarded onto Control-IQ between February and May 2020. Educators reviewed their data and called new users between one and three times during the first six weeks of Control-IQ use depending on whether the user met benchmarks for early success based on a clinical follow-up tool available on the program’s website. Those who did not meet the benchmarks after the first follow-up had a second follow-up to further optimize their AID system use, after which patients were highly satisfied and saw improved glycemic results.

User Interaction with CamAPS FX App Varies by Cohort with Caregivers of Young Children Spending 81 min/day In-App, Compared to Adults with T1D at 16 min/day and Adolescents with T1D at 10 min/day

Data drawn from six ongoing trials over 11-weeks among CamAPS FX hybrid closed-loop users and their caregivers (n=134) demonstrated significant differences in user interaction with the system by demographics. The demographics studied included: (i) caregivers of very young children between the ages of 1-7 with type 1; (ii) children and adolescents with type 1; (iii) adolescents ages 10-17 using closed-loop therapy from diagnosis with type 1; and (iv) adults; (v) pregnant women; and (vi) older adults over the age of 60. Researchers assessed user interactions with the CamAPS FX system based on minutes per day spent in the system’s connected app. As a reminder, the CamAPS FX system utilizes Dexcom G6 CGM, SOOIL Dana Diabecare RS pump or the Dana-i pump, and an Android app which hosts the algorithm. Within the app, participants could have been initiating prandial insulin boluses, starting/stopping a glucose sensor, reviewing data, updating app settings, responding to alerts, or troubleshooting other issues. Across all demographic groups, adjusted mean time spent “in-app” was 36 minutes per day. At 36 minutes/day, the results clearly show that even with an “automated” system, people are still spending a significant amount of time managing their diabetes. This varied dramatically with adolescents using CamAPS FX from diagnosis spending the least amount of time per day in-app at 10 minutes, followed by general adult and adolescent populations with established type 1 spending 16 minutes per day in-app. These values contrast with the pregnant population who spent “double the time in-app” compared to non-pregnant adults at 32 minutes per day. Finally, caregivers of young children and older adults had the highest diabetes burden, measured by time in app, spending an average 81 minutes and 63 minutes per day in-app, respectively.

“Catching Some ZZZ’s”: Real-world data (n=61,653 observations) reveals Control-IQ sleep activity associated with +7% overnight Time in Range without increased hypoglycemia

During a Tandem sponsored symposium, UVA’s Dr. Marc Breton made the compelling proposition that Control-IQ’s “sleep activity” function could lead the charge in transitioning overnight periods from a time of fear for patients with diabetes to an “opportunity for tighter glycemic control.” As a reminder, when Control-IQ’s sleep activity is enabled, the algorithm’s basal target range is tightened and narrowed to achieve blood glucose values between 112.5 –  120 mg/dl by morning. In addition to existing clinical trial evidence that this functionality improves overnight Time in Range, Dr. Breton today presented new real-world evidence using data from Tandem’s t:connect web application. In total, 72,248 users (5,149,243 days of data) met the study’s selection criteria (>6 years old, Control-IQ users, data available form January 11th to April 28th, 2021), of which the vast majority were adults (80% adults vs. 20% pediatric) with type 1 diabetes (93.5% type 1, 5% type 2, and 1.5% unreported).

  • Sleep activity use was characterized by a bimodal distribution, with large groups of participants either always using the feature or never using it. More specifically, 18.6% of users did not use sleep activity, while 44.2% of users used sleep activity every day. Excluding users who did not use sleep activity, school-aged children used the function slightly more on average than adults at 9 to 9.5 and 8 hours per day respectively.

  • When using sleep activity, participants showed a statistically and clinically significant increase of Time in Range. From 61,653 total observations, sleep activity was associated with a 7% increase in Time in Range, largely driven by decreases in time above 180 mg/dl and 250 mg/dl. Reassuringly, hypoglycemia was unaffected. Furthermore, overnight glycemic variability was reduced with sleep activity, with a trend toward decreasing variability as duration of sleep activity was increased.

Terumo Symposium Showcases Study Demonstrating Medisafe With’s Superior Low-Dose Bolus and 1.0 Unit Basal Rate Accuracy to Omnipod

Terumo’s symposium discussed the advantages of its tubeless, detachable insulin patch pump Medisafe With, specifically focusing on its comfort benefits and its high dosing accuracy. Dr. Tomoyuki Kawamura (Osaka City University) discussed the need for improvement in insulin pump therapy due to low use in Japan, estimating that as low as 4-5% of the Japanese population with type 1 diabetes used an insulin pump. He presented past survey data indicating that the most common reasons for discontinuing pump therapy among Japanese patients with type 1 diabetes were discomfort and infusion difficulties, and he stressed that comfort in the “daily handling of the pump is the key to continuing pump therapy.” As a result, Dr. Kawamura argued Medisafe With may be a more attractive alternative for patients due to its lack of tubing and its simple infusion operation as a detachable patch pump.

  • Following Dr. Kawamura, Dr. Guido Freckmann (Institut für Diabetes-Technologie) presented results from a study he conducted comparing Medisafe With, Minimed 640G, and Omnipod. Study methods followed guidance published by Kamecke et al. in 2019. The basal rate for the three pumps was 1.0 unit/hour, and bolus amounts of 0.2 unit, 1.0 unit, and 7.0 units were tested. 25 consecutive boluses for 0.2 unit and 1.0 unit were delivered, while 24 consecutive boluses for 7.0 units were delivered. Each test was performed with 3 devices per pump model and 3 repetitions, yielding a total data set of 9 repetitions for each scenario and pump model.

  • Results indicated Medisafe With and Minimed 640G had comparable bolus accuracy for all three bolus sizes, and Medisafe With had superior bolus accuracy to Omnipod for the 0.2-unit and 1.0-unit bolus size. For the 0.2-unit bolus, 95% of the data for Medisafe With and Minimed 640G was within the ±15% deviation limit (unlike Omnipod). For the 1.0-unit bolus, 50% of the data for Medisafe With and Minimed 640G fell within the ±5% deviation limit, and all of the data was within the ±15% deviation limit – thresholds Omnipod did not achieve. However, all three pumps were comparable for the 7.0-unit bolus with 95% of their data within the ±5% deviation limit. For basal rate accuracy, Medisafe With and Minimed 640G remained comparable, while Omnipod was significantly more inaccurate for low-dose boluses. Dr. Freckmann remarked Medisafe With showed “good accuracy” relative to the traditional durable pump and called for additional research into the degree of dosing deviation that is clinically relevant for different users. See the detailed data in the images below.

  • Dr. Dario Pitocco (Agostino Gemelli University Policlinic) highlighted survey data demonstrating an increase in insulin pump use in Italy within the last decade, referring to Medisafe With as a beneficial option for those beginning pump therapy. He referred to Medisafe With as a “doctor-, patient-, and eco-friendly” alternative due to its high accuracy dosing, easy insertion and operation, and its detachable design.

COVID-19 and Telemedicine Highlights

NIDDK to Support Research on Mechanistic Underpinnings of Severe COVID-19 Among Patients with Diabetes and/or Obesity; COVID-19 Likely Drove ~200,000 New-Onset Cases of Diabetes in Last Year

Dr. William Cefalu (Director, Division of Diabetes, Endocrinology, and Metabolic Diseases, NIDDK) gave an insightful presentation on recent efforts at the NIH to support research on COVID-19 and diabetes, highlighting the need for additional research on post-acute symptoms, underlying disease mechanisms, and COVID-19 related new-onset diabetes. Looking first at the disproportionate effect of COVID-19 on people with diabetes and/or obesity, Dr. Cefalu recognized the need for increased research on the mechanisms underlying COVID-19 comorbidities. Additionally, Dr. Cefalu discussed the significant observational data from COVID-19 patients demonstrating an association between hyperglycemia and poor outcomes, especially among hospitalized patients with and without diabetes as a key area of research moving forward. Additionally, across the organization, the NIH has implemented a “strategic plan for COVID-19 Research” to address  areas of inquiry across five priorities: (i) improving fundamental knowledge of SARS-CoV-2; (ii) advancing research to improve virus detection; (iii) supporting research to advance treatment; (iv) accelerating research to improve prevention; and (v) preventing and redressing poor COVID-19 outcomes, demonstrating a significant commitment of publicly funded research to better understanding, treating, and preventing COVID-19.

  • Within the NIH, the National Institute of Diabetes and Digestive and Kidney Disease (NIDDK) launched its own initiative to provide funding for “projects leading to rapid translation and impact in the COVID-19 emergency.” These areas of research interest include: the collection of bio-samples to inform the pathogenesis of COVID-19 related “kidney, gastrointestinal, or metabolic/endocrine diseases”; studies to better understand differential outcomes experienced by patients with COVID-19 endocrine, digestive, or kidney diseases; and studies to identify risk factors related the NIDDK relevant diseases and COVID-19 that require therapy modifications. Within these larger research areas, the NIDDK is currently supporting research into “artificial pancreas” systems for improved glucose management in diabetes and assessing the impact of hyperglycemia in hospitalized patients with COVID-19 on clinical outcomes as well as having an interest in the underlying “mechanism of hyperglycemia in critically ill COVID-19 patients.” Additionally, the NIDDK is also supporting research and have interests in  cytokines that “most strongly predict the clinical course” of COVID-19 in patients with type 2 while also investigating the impact of COVID-19 infection on human pancreatic endocrine cell function with the desired goal of protecting pancreatic cells from COVID-19 infection and potential subsequent damage.

  • Building on the NIDDK supported research, Dr. Cefalu also emphasized the need for research, based on recent studies,  to evaluate new onset diabetes related to COVID-19. Citing data from Al-Aly et al. published in Nature earlier this year, Dr. Cefalu shared that from the recent studies COVID-19  appeared to increase the number of cases of diabetes when assessed months from diagnosis and appeared to be related to the initial severity (non-hospitalized, hospitalized, and ICU admit). The data are certainly concerning indicating that there were an estimated 200,000 additional cases of new-onset diabetes driven by COVID-19 in the last year representing a significant additional burden to the healthcare system and patients. Additionally, for patients hospitalized with COVID-19, onset of multi-organ impairment, including diabetes, was seen following discharge hinting at the long-term effects of COVID-19 infection.

  • There is growing evidence of long-term complications from COVID-19 termed “post-acute COVID syndrome” which the NIH intends to research by establishing a “SARS-CoV-2 Recovery Cohort.” This cohort may include  over 10, 000  children, adults, and pregnant women to conduct longitudinal observations of COVID-19 related complications including assessment of  multi-organ effects . The NIH also plans to leverage electronic health record data in conjunction with health system and real-world data to conduct retrospective analyses of the incidence and severity of post-acute COVID syndrome across populations and will also be conducting autopsy studies of “brain/other tissue for injury due to SARS-COV-2 infection and/or its sequelae” that contribute to post-acute COVID syndrome. Via this last approach, we are curious to learn more about the mechanism of new-onset diabetes from COVID-19 via channels of insulin resistance or potential islet cell damage.

T1D Exchange COVID-19 Registry Shows Massive Differences in COVID-19 Outcomes by CGM and Pump Usage; 42% of non-CGM Users Ended up in Hospital or ICU vs. 10% of CGM Users

On Day #1 and Day #2, T1D Exchange’s Associate Director of Population Health Dr. Nudrat Noor and VP of Quality Improvement and Population Health Dr. Osagie Ebekozien presented extremely troubling results from the T1D Exchange COVID-19 registry by CGM and pump usage, respectively. The analysis was performed using EHR data from April to September 2020 across 56 endocrinology clinics. Of the 794 patients included in the registry, 63% were CGM users (n=501) and 37% were non-CGM users (n=293); 47% were pump users (n=376) and 53% (n=418) were non-pump users. Not surprisingly, the CGM and pump users had significantly lower median A1cs at baseline compared to the non-CGM and non-pump users at 7.9% and 9.5%, respectively for CGM users and non-CGM users, and 7.8% and 9.1%, respectively for pump users and non-pump users. CGM users were also significantly more likely to be privately insured (67% vs. 38%), as were pump users (72% vs. 42%). Pump users were also significantly more likely to be non-Hispanic White than were non-pump users (77% of pump users vs. 49% of non-pump users; p<0.001).

Highest level of care

CGM users (n=501)

Non-CGM users (n=293)

p-value

Intensive Care Unit

3%

18%

p<0.001

Inpatient/Hospitalization

7%

24%

p<0.001

  • The registry revealed that rates of ICU admission were six times (!) higher for non-CGM users than CGM users. In the registry, nearly one-fifth (18%) of non-CGM users who tested positive for COVID-19 ended up in the ICU, compared to 3% of CGM users. Down the care cascade, another quarter of non-CGM users were hospitalized with COVID-19, compared to 7% of CGM users. In sum, 42% of non-CGM users with diabetes who tested positive for COVID-19 ended up in the hospital or ICU; 10% of CGM users with diabetes who tested positive for COVID-19 ended up in the hospital or ICU.

  • While the study did not have the statistical power to examine differences in mortality, deaths were higher in the non-CGM group. In the non-CGM group, four patients died during the period analyzed, representing 13.6 deaths per thousand patients. In the CGM group, one patient died during the period analyzed, representing 2 deaths per thousand patients. While this analysis does not account for confounding variables (e.g., median A1c, insurance status, etc.), those numbers tell a very stark picture.

Highest level of care

Pump users (n=376)

Non-pump users (n=418)

p-value

ICU Admission

3%

14%

p<0.001

Inpatient/Hospitalization

7%

19%

p<0.001

DKA

5%

20%

p<0.001

  • The registry data also showed that hospitalization/ICU admission rates were more than three times higher for non-pump users than for pump users. Specifically, in the registry, those on MDI were significantly more likely to be hospitalized (19% vs. 7%) or to be admitted to the ICU (14% vs. 3%) than those using insulin pumps. While using a pump, of course, is undoubtedly in many cases a “marker” for a patient having other advantages (race, socioeconomic, other), it also, of course, no doubt denotes greater education and greater focus by people with diabetes and/or their families. When segmented by race, the disparity in hospitalization/ICU admission was even greater in non-White populations. Non-Hispanic Black non-pump users were nine times more likely to be hospitalized or admitted to the ICU than their pump-using counterparts (27% vs. 3%) – while we are not sure if this controls for income, these statistics are discouraging. On the other hand, while the data showing that Hispanic non-pump users were nearly seven times more likely to be hospitalized or admitted to the ICU than Hispanic pump users (20% vs. 3%), we also believe that this data provides several important “jumping off points” where awareness and education could be considered in different lights – there is no one who will say they are anything close to acceptable. Indeed, and while this is true, we also wonder how potential options will be introduced and put forward; we salute Drs. Noor and Ebekozien  and the T1D Exchange for preparing and highlighting this data and we are very eager to learn about what next steps will be. ATTD for these are alarming within-group disparities, whose underlying factors must be addressed. Also notable, in the overall participant population, those using pumps also saw significantly fewer events of DKA while positive for COVID-19 than did non-pump users (5% vs. 20%; p<0.001).

  • The speakers did not report the percentage of pump users who use CGM (and vice versa) nor did they report the percentage of pump users on AID systems, all of which would be helpful in determining the impact of CGM and pump use separately and together. Together, these results reinforce the association between diabetes technology use and avoiding adverse events in type 1s with COVID-19. 

UK Adult FreeStyle Libre Data Shows that All Age Groups Saw Improvements in Time in Range at Onset of Lockdown and Maintained Improvement through June; Time Below Range Did Not Change

On Friday afternoon, University of Leicester’s Dr. Pratik Choudhary read out early pandemic data from 8,914 adults in the UK who used FreeStyle Libre between January 2020 and June 2020. Although other studies have investigated how the COVID-19 pandemic impacted clinical outcomes for people with diabetes (see here for a DT&T retrospective US study with Dexcom and here for preliminary data from Italy), this is the first study that assessed the impact of the pandemic by age group to the best of our knowledge. The study split 8,914 FreeStyle Libre users who used LibreView and had ≥5 days of sensor readings for each month from January to June 2020 into four age groups: 18-25 (n=736), 26-49 (n=3,446), 50-64 (n=2,873), and 65+ (n=1,859). Overall, the study found that all age groups saw significant improvements in Time in Range when lockdown began in March and maintained these improvements through June. The largest improvement in Time in Range was seen among seniors (ages 65+; +42 minutes/day to 54%), followed closely by young adults (+45 minutes/day to 61%),, the two age groups that had the highest and lowest Time in Range at baseline, respectively. The vast majority of these Time in Range improvements were due to time above range improvements, as time below range did not change significantly in any age group. The young adult cohort saw the largest proportion of its members increase their Time in Range by >5%, which is a hugely exciting finding given that this age group historically faces the most challenges with glycemic control.

  • Across all age groups, the percentage of participants achieving the Time in Range consensus target of >70% increased from January 2020 to June 2020. The greatest improvements were seen among seniors (ages 65+) and young adults: the proportion of seniors and young adults achieving >70% Time in Range increased from 24% to 31% and from 15% to 20%, respectively. The proportion of participants achieving the time below range target of <4% remained consistent from January to June 2020 and ranged from 48% among young adults to 56% among seniors. However, when combined, the proportion of participants achieving both the Time in Range and time below range targets increased significantly across all age groups. At baseline, the proportion achieving both goals ranged from 6% among young adults to 12% among seniors, which increased to a range from 9% among young adults to 16% among seniors. Again, seniors saw the largest increase with a +4% improvement in the proportion achieving both goals, followed by young adults, adults ages 50-64, and adults ages 26-49.

  • These improvements in Time in Range and time above range were not due to increased scanning rates, as might have been expected. Daily scanning rates remained relatively consistent for all age groups across the six-month period, although the young adult cohort saw a slight increase in scans per day. Across all age groups, participants scanned their Libre ~14 times/day.

Glooko Real-World Data Shows Improvements in Glycemia Following Shelter-in-Place Orders During COVID-19 Pandemic for Both CGM and BGM Users; TIR from 62% to 63% for CGM Users

On Saturday morning, Senior Data Scientist Sarine Babikian (Glooko) presented results form a large set of Glooko users in the US and Europe showing slight improvements in glycemia during the early part of the COVID-19 pandemic. The results presented by Dr. Babikian built on the data Glooko shared back in June 2020 via blog post. Dr. Babikian’s analysis included ~62,000 CGM users and ~130,000 BGM users across the US and Europe using Glooko. CGM users were much more likely to have type 1 diabetes (92% type 1s vs. 45%) and significantly younger (28 years vs. 64) compared to BGM users.

 

Jan – mid-March 2020

mid-March – May 2020

US CGM users

 

 

Time in Range

62%

63%

Mean glucose

166 mg/dl

165 mg/dl

US BGM users

 

 

Mean glucose

169 mg/dl

168 mg/dl

European CGM users

 

 

Time in Range

82%

83%

Mean glucose

162 mg/dl

160 mg/dl

  • In general, Time in Range and mean glucose improved slightly following shelter-in-place orders. For CGM users in both the US and Europe, the Time in Range increase was ~1 percentage point (~15 minutes/day); for European users, Time in Range increased from a much higher baseline of 82% vs. 62%. Correspondingly, mean glucose declined by 1-2 mg/dl following shelter-in-place orders. While Time in Range was not available for BGM users, mean glucose declined by a similar 1 mg/dl, suggesting the glycemic improvements from the pandemic were similar.

  • Looking at the data graphically, another trend is obvious: the weekly fluctuations in glycemia were flattened out during the early part of the pandemic. In particular, glucose control is typically best on Thursday of any given week – the chart below uses Time in Range as an example. Glucose control is typically worst on the weekends. From mid-March to May 2020, this day-to-day variability was largely flattened as traditional work-/school-week patterns were erased. Of course, the chart below also shows clearly that Time in Range was on average higher on any given day of the shelter-in-place than before.

  • For US users with activity data, the average number of daily steps decreased from 2,300 to 1,300. This is a pretty drastic decline, though it matches well with data we saw from Evidation Health in early 2020.

Joslin Diabetes Center Data Shows Older Adults (≥65 Years) More Likely to Use Phone Visits vs. Video Visits Compared to Younger Adults (40-64 Years)

Dr. Elena Toschi (Joslin) presented results from a retrospective analysis of Joslin patients over the past couple years, in particular focusing on differences in telemedicine utilization by age. The data were all taken from adults with type 1 diabetes, broken into two groups: adults aged 40-64 years old and adults 65 years or older. At Joslin, between March 13 and mid-May 2020, all clinic visits were done via phone. Between mid-May and June 5, 2020, all clinic visits were done either via phone or video. After June 5, 2020, some in-person visits began to take place again. In total, Dr. Toschi’s analysis included clinic visit information from 3,585 patients from 2019 and 3,406 patients from 2020 – what a dataset!

  • For both groups of adults, average number of visits per patient increased in 2020 compared to 2019. In 2019, essentially all of Joslin’s visits were conducted in person with both groups of adults averaging ~two visits per person. In 2020, the average number of visits per person rose to ~2.5 – while, of course, potential explanations for this could include the additional guidance and advice needed during the COVID-19 pandemic, that could easily also be ease of use of telemedicine and less downside coming from competing priorities. The 2.5 visits per person in 2020 broke down into ~1 in-person visit and ~1.5 telemedicine visits.


  • On average, older adults were more likely to use phone-based visits over video-based visits. As shown in the charts below, when both options were available, ~30%-40% of older adults chose to have phone-based videos, compared to ~15% of the younger adults. Similarly, 35%-40% of older adults chose to have their visits over video, compared to 60% of younger adults. Presumably, the remainder chose to conduct their visits in-person. Not too surprisingly, we’ve heard multiple providers over the past year generally express a preference for conducting telemedicine visits over video; see here, for example, for more learning in an excellent “Beyond Telemedicine” Virta mid-pandemic webinar. Of course, multiple KOLs also point out that they want what works well for people with diabetes, some of whom don’t have good video access. That said, on the video “plus” side, in particular, the screen-sharing feature has been highlighted as being valuable for allowing patients and providers to review glucose or insulin data together. While that is possible, it goes on what we’ve begun calling our “diabetes SLUDGE list” in  honor of Dr. Cass Sunstein, celebrated Harvard scholar and famous author of Pulitzer prizewinning book Nudge: Improving Decisions About Health, Wealth, and Happiness. Sludge is, according to Dr. Sunstein, “frictions – paperwork burdens or administrative obstacles – that hinder meeting the targets for the individuals …” What would you add to the SLUDGE list in diabetes? On this study, in the future, we’d love to see more analysis for the drivers of older adults’ preferences for phone-based visits – specifically, what is modifiable and what’s not. Finally, for both in-person and video-based visits, older adults had lower rates of missing appointments, with comparable levels between in-person and video-based (see graphic below).


Two Studies from Colombia and the Netherlands Highlight Successes with Virtual Training for Medtronic MiniMed 670G and MiniMed 780G During the COVID-19 Pandemic

Closing out one of this afternoon’s oral presentation sessions, Dr. Ana Maria Gomez Medina (Hospital Universitario San Ignacio) and Dr. Henk-Jan Aanstoot (Diabeter Rotterdam) presented positive results from their clinics’ transitions to virtual-based pump training during the COVID-19 pandemic. The studies’ results are encouraging and seem to confirm various remarks we heard throughout the pandemic from Tandem, Insulet, and Medtronic about success with virtual pump onboarding (see Tandem 2Q20 and Insulet 2Q20). Medtronic, in particular, published a paper in DT&T in July showing that virtual trainings for MiniMed 670G had comparable glycemic results as in-person training, as well as better patient satisfaction.

  • In Bogota, Colombia, Dr. Medina’s clinic trained 91 people with type 1 diabetes on MiniMed 670G between March and July 2020. Previously, patients were using MDI or sensor-augmented pump therapy (SAP). Depending on patients’ baseline therapy regimen, the training consisted of at least seven sessions.  The virtual training utilized Zoom and Medtronic CareLink. Mean Time in Range in manual mode was 77%, which improved to 82% after two weeks in Auto Mode (+1 hour/day). Time below range was also significantly improved from 2.7% to 1.8% (-13 minutes/day). Dr. Medina described these results as “similar to face-to-face training.”

  • In Rotterdam, Netherlands, Dr. Aanstoot’s clinic trained 35 people with type 1 diabetes on MiniMed 780G. At baseline, all participants had been using MiniMed 670G for at least three months. All participants were able to start using MiniMed 780G in Auto Mode within 48 hours of training. Mean Time in Range during the last week of 670G was 76%, which improved to 84% after five weeks on 780G (+1.9 hours/day). The improvement was entirely driven by reduced time above range, which fell from 21% to 13%. Time <70 mg/dl remained the same at 2.6%. Mean glucose fell from 166 mg/dl to 142 mg/dl after upgrading from 670G to 780G. Finally, the time spent in Auto Mode was significantly improved from 74% to 96%.

FreeStyle Libre Detects 3x Hyperglycemic Events and 2x Hypoglycemic Events Compared to BGM in Patients with Diabetes Hospitalized with COVID-19 Pneumonia

On Saturday morning, Italy’s Dr. David Brancato (Hospital of Partinico) presented the results of an observational, retrospective study that assessed the clinical utility of FreeStyle Libre for inpatients with COVID-19 pneumonia and diabetes (n=31; baseline A1c of 7.9%), prediabetes (n=11; baseline A1c of 6.1%), or incidental hyperglycemia (fasting glycemia > 125 mg/dl; n=9). All participants were treated with the steroid dexamethasone and underwent parallel glucose assessment via BGM (specifically MyStar Extra or Glucomen Areo 2K) and via FreeStyle Libre CGM. The study found that FreeStyle Libre detected three times more episodes of hyperglycemia and twice as many hypoglycemic events than did routine BGM assessments in patients with diabetes (14 vs. 4 episodes/day, p<0.0001; 3.4 vs. 1.7 episodes/day, p<0.0001, respectively). FreeStyle Libre also revealed many hyperglycemic and hypoglycemic events in patients with prediabetes and with incidental hyperglycemia (7.9 episodes/day and 3.7 episodes/day, respectively), suggesting that these patients should receive similar routine management to those with diabetes and that it is important to assess for incidental hyperglycemia in patients with COVID-19 pneumonia. With CGM, episodes of hyperglycemia were defined as >15 minutes >180 mg/dl and episodes of hypoglycemia were defined as >15 minutes <70 mg/dl.

  • Patients with diabetes and COVID-19 pneumonia had a Time in Range of 58% and spent 22% of time >180 mg/dl and 3.2% of time <70 mg/dl. Notably, participants with diabetes spent 15% of time >250 mg/dl, which is significantly higher than the consensus goal for <5% of time >250 mg/dl. It’s worth noting that this is a group of patients who had reasonably low A1c values at baseline (mean 7.9%), and yet they were still spending a significant amount of time in severe hyperglycemia while hospitalized with COVID-19 pneumonia. Patients with prediabetes or incidental hyperglycemia had an average Time in Range of 79% and spent only 14% of time >180 mg/dl and 1.5% of time <70 mg/dl.

  • The study found that there were high levels of variability in the mean glucose value throughout a 24-hour period among both patients with diabetes and patients with prediabetes/incidental hyperglycemia. Among patients with diabetes, mean glucose values were significantly above target throughout the day during five intervals (hours 03-08, 08-12, 12-18, 18-22, and 22-02) with a majority of participants with diabetes having a mean glucose above the goal between 8 am and 10 pm. Per Dr. Brancato, this study “is the first study to show such a pattern in COVID-19 disease.”

Diabetes Patients’ Perceptions of Telemedicine during the COVID-19 Pandemic

A few oral presentations discussed survey findings analyzing patients’ perceptions of telemedicine. Intriguingly, it seems a significant portion of patients across all ages tend to perceive telemedicine as useful, yet variations in perception exist when differentiating by A1c levels. However, while many patients may find telemedicine useful in theory, substantial improvements in data collection and appointments through telemedicine are still required to provide adequate and smooth care to diabetes patients.

  • Dr. Sam Scott (University of Bern) shared results from a survey conducted among people with type 1 diabetes from March 24 to May 5, 2020, suggesting those with higher A1c levels tend to perceive telemedicine as less useful. The survey received 7,477 responses across 89 countries, and results seemed “against the general perception” that older people are less inclined to telemedicine as nearly 80% of respondents ages 65+ perceived telemedicine as “useful,” suggesting telemedicine is still viable for older patients. However, individuals with higher A1cs (> 9%) tended to find telemedicine less useful than other respondents, which Dr. Scott speculated may be due to frustration with results or a lack of support. Therefore, he emphasized patients should be assessed on an individual basis to determine their perception of telemedicine and whether it would be useful for their diabetes management rather than resorting to demographical assumptions to determine for whom it may be beneficial.

  • Caterina Florissi highlighted the main findings from a dQ&A survey on telemedicine during the COVID-19 pandemic, which revealed shortcomings in patient education and telemedicine platforms. Ms. Florissi spoke on the potential benefits and growth of telemedicine during the pandemic, citing a CDC study that showed a 50% increase in telemedicine visits from January-March 2020 compared to the same time frame in 2019. In order to evaluate how providers collected patient data and what platforms they used for telemedicine, dQ&A conducted a mixed method study in June and July 2020 with an initial qualitative phase consisting of video-conference interviews with providers (n=24) followed by a quantitative phase with an online survey (n=318). Results revealed that accessing patients through telemedicine was a significant challenge for providers, due to the fact that many patients did not own connective devices that would allow them to send data in digital form to providers. Moreover, even when patients did have connective devices, many of them relied on in-person visits where the providers downloaded the data and could not figure out how to do so on their own. On average, healthcare providers used three different methods gathering data – 72% reported using verbal descriptions of data and just under half reported using faxes or photos of paper logs, which are older forms of data exchange considered suboptimal for telemedicine. Additionally, healthcare providers on average used two platforms to communicate with their patients, with Zoom, audio phone calls, and FaceTime being the most common. Lack of integration between these conferencing tools and electronic medical records was a source of frustration for both providers and patients, as providers must continuously switch between screens during appointments. Ms. Florissi concluded with a call to action for the implementation of modern and integrated platforms in telemedicine with three targets: (i) patient training; (ii) a centralized platform for data uploads; and (iii) an integrated platform for video conferencing and data management.

The Beneficial Impact of Emerging Telemedicine Platforms on Diabetes Management and Research

Due to the shortcomings in healthcare accessibility that were exacerbated by the COVID-19 pandemic, telemedicine assumed a central focus in several oral presentations. We heard presentations on several new telemedicine strategies and features seeking to improve patients’ glycemic control and accelerate research enrollment. Across the presentations, it seems there was universal consensus on the need for more patient-first telemedicine services, and we’re glad to see there is continued attention to innovating and improving these services. See our highlights below.

  • Dr. Giada Acciaroli (Dexcom) discussed the benefits of Siri integration within the Dexcom G6 app on glycemic control, presenting results from two recent studies. First, she highlighted legally blind patients with diabetes as a population whose glycemic control could significantly benefit by using the G6/Siri integration feature, showing results from a 2021 study that demonstrated 12 months of use of this feature resulted in a significant reduction in A1c and an increase in Time in Range within this population. Second, Dr. Acciaroli showcased a much larger study in US-based G6 users that analyzed whether routine use of Siri integration influenced glycemic control. Researchers defined non-users (n=6,847) as those who did not invoke the feature in 1H20, and they defined routine users (n=2,282) as those who averaged at least one feature invocation per day in 1H20. Results revealed routine users had a higher Time in Range compared to non-users (62% vs. 57%, p<0.001), as well as lower mean glucose levels (169 mg/dl vs. 177 mg/dl, p<0.001). Although Dr. Acciaroli stated these results indicate routine use of voice integration may help users better manage their diabetes, she acknowledged more contextual information around use of the feature is needed to specifically determine the benefits on glycemic control. The feature is only available to US Apple users, but Dr. Acciaroli mentioned Dexcom is looking to roll out the feature to Android in the future.

  • Dr. Tiffany Jeanson (myDiabby Healthcare) highlighted the benefits of her company’s software on A1c levels for both people with type 1 and type 2 diabetes. The myDiabby software allows patients to input glucose readings and other measurements, which their healthcare team can access and communicate with the patient. Dr. Jeanson shared preliminary results from the ETAPES trial, an ongoing four-year clinical trial in France evaluating the impact of telemonitoring for people with diabetes. Participants (n=151) in the trial received weekly remote monitoring and monthly therapeutic coaching through the myDiabby software, and results indicated the telemonitoring model resulted in a 2% decrease in A1c for people with type 1 diabetes and a 3% decrease for those with type 2 diabetes after three months – both of which were sustained after six months and were regardless of gender and age. Although these results indicate this telemonitoring model has potential, data on the nine-month and twelve-month follow-ups is still needed to determine more long-term benefit, and it must be noted the median A1c for participants was very high (10.12% ± 0.11%).

  • Dr. Peter Senior (Diabetes Action Canada) shifted focus away from telemedicine’s possible benefits for glycemic control to discuss how digital registries can accelerate type 1 diabetes research, presenting the recently launched Connect1d – a Canadian registry that seeks to connect type 1 diabetes patients with clinical researchers. Patients can register by providing their baseline demographics, which are stored by the database, and an algorithm matches people to studies posted by researchers for which they may be eligible. Patients can allow researchers to directly contact them about study enrollment, and any additional inclusion/exclusion criteria required is further stored in the registry to finetune the matching algorithm for each individual patient. After launching in January 2021, approximately 130 patients and 45 researchers have engaged with the registry. Dr. Senior shared the developers’ vision for the registry, stating they desire to post trusted and individualized educational materials to the platform and integrate data from individuals’ diabetes devices and EMR data. Currently, the platform is only limited to adults, but pediatric and caregiver interfaces are in long-term development plans. Dr. Senior noted Connect1d is an example of making people with diabetes the “key drivers” in scoping problems and designing solutions, emphasizing teleservices should prioritize this patient-first model since they are “most beneficial” when people living with diabetes are given significant control.

Big Picture Highlights

SWEET Pediatric Diabetes Registry Data Shows A1c Reductions Across All Age Groups Between 2008-2010 and 2016-2019; COVID-19 Didn’t Impact Glycemic Control, but DKA Events Associated With Higher Countrywide COVID-19 Mortality

Dr. Thomas Danne (Hannover Medical School), Chairman of the SWEET Project, presented 10 years of data from pediatric patients with type 1 diabetes across 22 SWEET centers. Most notably, notable reductions in average A1c were seen at 21 of the 22 sites over the ten-year period. The analysis covered the time period from 2008-2018: 4,930 patients were in the registry in 2008, growing substantially to 13,654 patients by 2018. Across the centers, patients had a median age of 12 years old, 51% were male, and had an average diabetes duration of 4 years. Not surprisingly, adolescents and teens had the highest A1c values on average, but all age cohorts saw mean A1c declines over the 10-year period. This data is especially notable given data from a similar pediatric cohort from T1D Exchange that indicated A1c values had actually gotten worse for children and adolescents from 2010-2012 and 2016-2018.

  • Dr. Danne highlighted diabetes technology as a potential key driver for improved outcomes among the SWEET population. Specifically, Dr. Danne discussed a recent publication in Diabetes Care which found an association between glycemic outcomes and insulin pump and CGM use. Data from over 25,000 pediatric patients with type 1 indicated that A1c was lower “in all categories of participants who used a pump and/or sensor compared with the injections-no sensor treatment method (p<0.001).” The lowest A1c values were seen in children using both insulin pumps and CGMs (n=7,787) with an average A1c of 7.8%, followed by children using insulin pumps without CGMs (n=4,418) who had an average A1c of 8.1%. In children using MDI therapy, A1c was 8.3% for those also using a CGM (n=3,843) and 8.7% for those not using CGM (n=9,606). In addition to technology use, Dr. Danne also discussed the role of clinical targets in glycemic outcomes among the pediatric SWEET population. Across the SWEET clinics, the majority comply with ISPAD guidelines targeting A1c values of 7% or lower. Another 27% of clinics follow ADA guidelines for an A1c target of ≤7.5%. However, a number of clinics in the SWEET registry also promote stricter A1c targets than those guided for by either ISPAD or the ADA with 3% of clinics targeting A1c ≤6.7%, 17% targeting A1c ≤6.5% and an additional 3% targeting A1c ≤6% among their pediatric populations. By analyzing the glycemic outcomes of pediatric patients affiliated with these various centers, Dr. Danne expressed that lower clinical A1c targets are associated with better metabolic control independent of educational time and the composition of diabetes care teams.

  • In light of the COVID-19 pandemic for the last 18 months, Dr. Danne and the SWEET registry conducted a retrospective analysis on diabetes management in 22,820 children during the first wave of the COVID-19 pandemic (May-June 2020). Children were stratified into quartiles based on the background rate of COVID-19 mortality in their respective countries and data were compared to the same time period from 2019. Additionally, Dr. Danne also analyzed data from August-September 2020 to represent the period between COVID-19 waves when some regions re-opened. Across the quartiles, A1c saw no significant changes before and during the first wave of the pandemic, indicating glycemic control was not disrupted in the SWEET population. However, looking at DKA event rates, the data showed an increase in DKA events among pediatric patients living in countries with the higher burdens of COVID-19 (e.g., Belgium, Chile, England, France, Ireland, Italy, Netherlands, Spain, Sweden, USA). DKA rates did drop back down to pre-pandemic levels in August-September of 2020, indicating the increase seen in May-June was likely driven by the pandemic, but not indicative of longer trends. Turning to diabetes technology use, Dr. Danne reported an increase in CGM use among patients in the SWEET registry in countries with higher COVID-19 mortality. That trend was especially notable in the August-September 2020 period, suggesting patients may have transitioned to CGM during the pandemic out of the need for remote data monitoring. However, similar increases were not seen for insulin pump use.

  • For the year 2020, the median A1c for pediatric patients in the SWEET registry was an encouraging 7.7% - this is much better than has been seen historically. While there is certainly room for improvement, this value is much better than the almost 9% average A1c seen in US pediatric populations. Of course, like the T1D Exchange data, the SWEET registry also represents patients being seen at some of the top diabetes centers in their respective countries (that said, such centers hardly have a monopoly on patients with very high TIR and we certainly also know that despite excellent care, it’s very common to have people with high A1cs and low TIR). Still – on average, of course we’d expect these centers to have people with diabetes that do better on average.  Given this wealth of data in the SWEET registry, Dr. Danne also discussed clinics’ ability to measure improvements across other metrics including retinopathy, nephropathy, severe hypoglycemia, DKA, blood pressure, BMI and lipids – as readers might imagine, we were incredibly excited to hear more about this. Clinics also have the ability to compare their outcomes with those of other clinics and learn from the work of their peers via onsite visits to share best practices and mechanisms for achieving improved glycemic control. Interestingly, within the registry, clinic size is not associated with A1c outcomes demonstrating that strong glycemic control can be achieved across a variety of circumstances.

T1D Exchange Quality Improvement Database Shows Real-Time CGM Associated with 0.7% A1c Reduction Compared to SMBG, Lower DKA and Hypoglycemia

T1D Exchange’s Associate Director of Population Health Dr. Nudrat Noor shared real-world findings from the T1D Exchange’s Quality Improvement Database showing superiority of real-time CGM over SMBG on several fronts. The analysis included 11,472 people with type 1 diabetes across eight endocrinology clinics in T1D Exchange’s Quality Improvement initiative. The analysis included EMR data from patients who had at least one clinic visit between July 2017 and February 2020. Overall, 48% of people were using real-time CGM and 52% were using SMBG. (To our knowledge, data from FreeStyle Libre users were not included in this study – we are confirming this.) For reference, T1D Exchange’s most recent State of T1D Outcomes reported CGM adoption at 30% in 2016-2018. Of course, the two figures are not directly comparable, as the clinics included are slightly different. In the eight clinics used in this analysis, there were also significant differences in CGM utilization by race. In data that continues to be troubling, for white people with type 1 diabetes, CGM adoption was 49%, compared to 38% for Hispanics and just 18% for non-Hispanic Black people. Similarly, coverage by private insurance was significantly associated with higher CGM utilization. Over half (57%) of those on private insurance were using real-time CGM, compared to 33% of those on public insurance. These inequities in diabetes technology utilization have been widely discussed over the past few years (see Keystone 2019, ISPAD 2020, etc.), and we are looking to learn more about action that can change the outcomes.

 

Real-time CGM (n=5,486)

SMBG (n=5,986)

Mean A1c

8.1%

8.7%

Patients with DKA events

2%

7%

Patients with severe hypo events

8%

10%

  • Mean A1c for real-time CGM users in the database was 0.8% lower than that of SMBG users (8.1% vs. 8.9%; p<0.001). This is largely in line with findings over the past few years, such as those from the GOLD and DIaMoND trials, which found A1c advantages of ~0.5% over SMBG. In more specific populations, CGM was associated with a 0.3% difference in A1c for older adults (WISDM) and a 0.4% A1c difference for adolescents (CITY). As a reminder, these studies were randomized control trials, whereas the data presented by Dr. Noor were real-world results with real-time CGM and SMBG groups that were not always well-matched (see racial and insurance differences mentioned above).

  • Real-time CGM users reported significantly lower rates of DKA and severe hypoglycemia events. DKA events, in particular, were significantly correlated with real-time CGM usage – a whopping 2% of those on real-time CGM had DKA compared to 7% on SMBG! How does that translate? Across the CGM cohort, 16 DKA events were reported per thousand patients. In the SMBG cohort, 26 DKA events were reported per thousand patients. This effect was said to be even more pronounced for severe hypo events. Rates of severe hypoglycemia were 32 per thousand patients in the SMBG group, eight times that of the CGM group (four events/thousand patients) – from our view, from the table above, since it’s 8% of patients vs 10%, we assume the 10% was happening much more frequently than for the 8% - we’re checking on this.

  • CGM users were over two times more likely to be insulin pump users – this makes sense since CGM drives awareness, particularly of weaker glucose health. This is why we often call CGM the “killer app” for pumps, or more broadly, those seeking better insulin delivery. Over half of the patients in the CGM group were insulin pump users (68%), which compared to just 32% in the SMBG group. Surprisingly, however, broken out by mode of insulin delivery, CGM users reported a mean A1c of 7.8%, regardless of whether they were on MDI or insulin pump. We wonder if this is an artifact of something – we’re checking into this since in most studies, the opportunity to pump following MDI shows improved outcomes. For SMBG users, those on SMBG + insulin pumps had an A1c of 8.3%, compared to 8.6% for those on SMBG + MDI. We’d love to see Time in Range for this group and we’ll be asking the researchers.


  • Results from the COMISAIR study suggest most of the benefits of advanced diabetes technologies (e.g., CGM, insulin pumps) are driven by the CGM, and these T1D Exchange data seem to support that result. We don’t completely understand this – CGM is a diagnostic and insulin is a therapy that is presumably delivered in more accurate doses using CGM data. Given the advancements in hybrid closed loop technology in the T1D Exchange study period (i.e., Basal-IQ, Control-IQ, MiniMed 770G), we are very curious to see a separate breakout for CGM + pump users using an AID system.

De-Identified Insurance Claims Analysis Shows CGM Pays For Itself Within One Year of Initiation with Mean Adjusted Annual Cost Savings of $2,038

Dr. Carlos Vallarino (Lilly) presented results from a real-world analysis of people with diabetes using rapid-acting insulin, finding reductions in medical cost from CGM initiation after just one year. The results were calculated using data from IBM’s MarketScan insurance claims database between 2015 and 2017. The database only includes claims for commercial insurers. A total of 7,700 patients were included the analysis, all over the age of 40 years old and using rapid-acting insulin. Of the sample, 966 initiated CGM and 1,579 initiated on pump therapy; the remainder served as the BGM-using control group. Patients using both CGM and pump were excluded from the analysis.

  • Both CGM and pump initiation were associated with significantly reduced rates of hospitalizations. For the BGM cohort, the total rate of hospitalization was 26%, compared to 15% for the CGM group and 16% for pump users. These hospitalization rates corresponded to average annual medical costs of $25,410 for the BGM group, $17,461 for the CGM group, and $17,585 for the pump group. Note that these numbers are unadjusted for differences between the groups (e.g., the CGM and pump groups are made up of many more type 1s than the BGM group). Similarly, CGM and pump initiation were associated with lower rates of follow-up hospitalizations. The BGM cohort had a 21% of follow-up hospitalization, compared to 10% for CGM users and 18% for pump users.

  • Adjusting for differences in the groups, the results showed a $2,038 reduction in medical costs after just one year of CGM initiation. The BGM cohort served as reference and the unadjusted difference was a $6,208 mean reduction in cost. As shown in the chart below, the vast majority of this reduction in medical costs was driven by reduced hospitalization costs. Dr. Vallarino did note that this analysis didn’t include the device cost of CGM and that the cost of CGM would likely erase cost savings at one-year. After one year, pump initiation was associated with significant medical cost savings ($3,927/year), but after adjusting for multiple variables, pump initiation was actually associated with a slight increase in total medical costs ($1,786/year).


  • The cost-savings delivered by CGM are especially encouraging as some people have traditionally viewed CGM as a longer-term investment. The data from IBM’s MarketScan suggests that, at least for people using rapid-acting insulin, CGM pays for itself very quickly (~one year). Given CGM also drives significant A1c reductions, and thus rates of complications, medical cost savings would likely increase over time. In a longer-term analysis, we would also be curious how quickly insulin pump initiation pays for itself – the significant reductions in hospitalization rates (though unadjusted), seems to suggest the device eventually would have a positive return. We would also be curious to see the sources of medical cost savings (e.g., reductions in hypo-related hospitalizations, total medication costs, etc.), as well as an updated analysis with newer diabetes device technologies, namely more recently launched CGM systems, hybrid closed loop systems, and other closed loop systems such as Tandem’s Control IQ and Medtronic’s 780G and in the future, Insulet’s Omnipod 5. 

Team Novo Nordisk Survey Finds 90% of Respondents Are Interested in “Using Technology for Improved Exercise Performance” with 89% Interested in Diabetes Technology for “Glucose Management”

Dr. Sam Scott, Head of Research for Team Novo Nordisk presented results from a questionnaire study (n=15) indicating that 90% of Team Novo Nordisk cyclists who responded “showed an interest in using technology for improved exercise performance” and 89% were also interested in diabetes technology for “glucose management.” However, 28% of respondents also identified as feeling “unaware of newer technologies,” indicating an area for improved education and outreach. In this survey study, participants had a mean age of 27 with an average A1c of 6.8%, demonstrating good glycemic control. As a reminder, Team Novo Nordisk is a professional cycling team co-founded and led by Phil Southerland and comprised entirely of athletes with Type 1 diabetes. From conversations with Mr. Southerland, we understand that all Team Novo Nordisk cyclists utilize CGM technology, especially when training, but that very few cyclists use AID or pump therapy, instead opting for multiple daily injections. We are curious if this preference is due to the extreme endurance training of Team Novo Nordisk athletes and the sometimes-limited flexibility in dosing that can come with an AID system. Additionally, as diabetes technology continues to move toward connected solutions, we are curious if this has contributed to the 28% of respondents who may feel less knowledgeable about emerging technologies. Outside of diabetes technology, the Team Novo Nordisk cyclists surveyed identified sleep quality and nutrition as key factors in performance (80% and 93%, respectively) and we know both of these factors can also play a significant role in diabetes management (see more on diaTribe’s 42 Factors that affect blood glucose here). Additionally, among participants, 36% reported losing 1 night of sleep per week due to hypoglycemia while 21% reported losing 2 nights/week.

Ms. Kelly Close on the Power of Social Media and Peer Support

In her talk on the utility of social media and the diabetes online community (DOC), Ms. Kelly Close (The diaTribe Foundation) encouraged members of the audience who are researchers or clinicians to become more involved in social media to either learn more about research or spread the word about their own or to help support others – giving “social media” prescriptions is definitely gaining in popularity and we like the idea of this, especially as it might encourage those could themselves benefit who are HCPs or researchers to try it!

Specifically, as we understand it, the DOC can simultaneously educate, engage, and empower people with diabetes to actively participate in their healthcare. HCPs are also integral to the DOC, both in publishing reputable information and resources for patients and directing patients to safe communities and platforms. For HCPs who do not want to post, consistently browsing diabetes channels on social media sites like Twitter and Facebook can provide necessary insight to patients’ thoughts and actions to improve care delivery – it can also teach them a lot about diabetes if they are following colleagues like Dr. Dan Drucker or Dr. Diana Isaacs or Ms. Hope Warshaw or Ms. Natalie Belleni or Ms. Cherise Shockley or Mr. Mike Warshaw and the list goes on and on and on! There was lots of encouragement for the audience to cover organizations like Diabetes UK and ADA and Beyond Type 1 and others. And, notably, this has been studied and there is good support for it! We’d love to see this in Standards of Care and will be asking what sort of study would have to happen for this! In fact, an impressive 2019 study by Litchman et al. found that for every one-point increase on a five-point DOC engagement scale, there was a 34% reduced chance of having A1c above 7%. DOC platforms are relatively widely used – data from diabetes market research organization dQ&A also showed that 38% of people with T1D and T2D were active online. However, Close also emphasized when utilizing social media, it is important to retain patient privacy and act ethically when posting or interacting with DOC members. Still, the majority of people with diabetes look to HCPs as their primary sources of information according to data from dQ&A. According to Ms. Close, social media is a complement to face-to-face HCP-patient interaction, with patients spending more time online than in clinic, and an opportunity to lead with curated sources of factual information and support from different people’s personal stories.

Diabetes Drugs Highlights

Dr. Jay Skyler’s 2021 Update on Type 1 Immunotherapies Spotlights Continued Lack of “Aggressive Combination Therapies,” Imatinib as the Latest Candidate to Only Show a Transient Benefit

In typical ATTD fashion, this year’s final plenary session was dedicated to type 1 diabetes cures, or as ATTD co-chair Dr. Tadej Batellino aptly put it, “what we really dream about” in the diabetes community. As at last year’s ATTD, Dr. Jay Skyler (University of Miami) gave the closing talk of the symposium, presenting on the latest in immunotherapies. While much of Dr. Skyler’s review echoed similar themes from last year, we picked up on two key advancements:

  • First, growing support for teplizumab as the first drug shown to delay the onset of type 1 diabetes. As a reminder, original data presented at ADA 2019 showed a median two-year delay to time of diagnosis, and follow-up data presented at ADA 2020 showed an extended delay of up to three years. Dr. Skyler emphasized efficacy differences between teplizumab subgroups (greater likelihood of response from individuals who are ZnT8 negative, absent of HLA-DR3, or present of HLA-DR4), which he believes may be useful for stratifying patients in the future. While not mentioned in Saturday’s talk, we note that teplizumab is currently under FDA review after a positive AdComm vote in May 2021.

  • Second, Dr. Skyler highlighted new (currently in press at Lancet Diabetes & Endocrinology) follow-up data on imatinib (tyrosine kinase receptor inhibitor) in adults recently diagnosed with type 1 diabetes – a compound that has flown relatively under the radar since its original data readout at ADA 2017. New follow-up data shows that although imatinib conferred benefits on C-peptide and beta cell glucose sensitivity during the six months of treatment, these differences were eliminated by two years’ time.

  • Imatinib is the latest example that supports Dr. Skyler’s theory that short-term treatments may not be able to deliver the long-term benefits sought after in type 1 diabetes (low-dose ATG, anti-TNF, and anti-IL-21 and liraglutide being a few other examples). In the case of anti-TNF alpha, Dr. Skyler pointed out that patients with rheumatoid arthritis use the drug for “years and years,” and people with type 1 diabetes may need to do the same. As such, Dr. Skyler continues to support an “aggressive combination therapy” delivered over a longer period of time; Dr. Skyler himself has long been working to initiate the Diabetes Islet Preservation Immune Treatment Trial (DIPIT), designed to apply long-term anti-TNF, deliver immunomodulation via ATG or anti-CD3, and preserve beta cell health with a GLP-1. Unfortunately, companies have not been willing to supply anti-TNF or the IL-2 drugs – airing concerns that other drugs will be used alongside their drugs (“My drug is going to get blamed if something goes wrong”). Back at EASD 2018, Dr. Skyler stated that it has been “very difficult” to navigate this, and we commend his longstanding commitment to this important trial. According to ClinicalTrials.gov, the trial is now set to initiate in December 2021 (pushed from December 2018), with a primary completion date of June 2024 (pushed from January 2020).

INNODIA Symposium Provides Updates on Biomarker Research and Clinical Trials: >4,500 Participants Recruited for T1D “Cure” Trials

In a session discussing prevention and early intervention therapies for type 1 diabetes, we received updates on INNODIA from Dr. Chantal Mathieu (KU Leuven), Dr. David Dunger (University of Cambridge), Dr. Anke Schulte (Sanofi), and Dr. Thomas Danne (Children’s Hospital Auf der Bult). As a reminder, INNODIA is a public-private collaboration encompassing the United Kingdom and the EU seeking to innovate clinical trial design and accelerate drug development in type 1 diabetes (see our EASD 2020 coverage for more detail). The collaboration established infrastructure to uniformly identify, enroll, collect samples, and phenotype newly diagnosed patients and unaffected first-degree relatives, including a central database to store all data. A master protocol, specifying the above factors of data collection to speed up and streamline trials, standardizes all clinical trials, including the four ongoing trials: (i) MELD-ATG; (ii) IMPACT; (iii) Ver-A-T1D; and (iv) CFZ533. Following the announcement of INNODIA’s collaboration with Dexcom to incorporate G6 CGM in these trials, Dr. Dunger confirmed CGM is being used to “define fluctuations in blood glucose,” particularly in those with presymptomatic dysglycemia who tested positive for one or more autoantibodies. See below for details on recruitment, biomarker research, and the clinical trials.

  • Dr. Mathieu provided updates on recruitment. As of April 2021, 4,661 participants have been recruited, including 595 newly diagnosed and 4,066 unaffected first-degree relatives (310 of which tested positive for at least one autoantibody, with 242 participating in follow-up). According to Dr. Mathieu, the trials have returned to “pre-COVID recruitment levels,” and they remain “on track” for their enrollment goals.

  • Dr. Schulte discussed INNODIA’s novel biomarker discovery research for new onset type 1 diabetes. She presented data on measured C-peptide levels from the 100 ND and Next100 cohorts, which include newly diagnosed individuals positive for at least one autoantibody at baseline. In both cohorts (which had comparable baseline characteristics except gender), researchers identified two subclusters of “slow” and “fast” beta cell functionality decline clusters. Dr. Schulte stated the “exact” profile fit between the two cohorts has encouraged researchers to attempt to associate the 100 ND’s  ‘multi-omics’ data (e.g. lipidomics, immunomics, genomics, etc.) with these progression clusters. According to Dr. Schulte, the next steps involve prioritizing the ‘omics’ features according to their correlational strength with the clusters, defining the list of the most credible features, and validating this list with the Next100 cohort. She also shared investigators plan to initiate a similar approach in unaffected first-degree relatives who test positive for one or more autoantibodies. We imagine that a ‘multi-omics’ profile of individuals with “slow” and “fast” beta cell functionality decline could be used to both better screen for clinical trial participants and potentially identify “super responders,” depending on a therapeutic’s mechanism of action.

  • Dr. Danne surveyed the status of the four clinical trials and unveiled a tentative design for a verapamil combination therapy study. Although specific details were sparse, he did state that each study has randomized its first set of patients. Excitingly, Dr. Danne provided insight into future clinical trial development, sharing planning details on a study called VERA Plus. VERA Plus will examine the efficacy of a combination therapy involving verapamil and several immunomodulators, including anti-CD3 (teplizumab) and anti-TNF – both which have shown positive (albeit somewhat short lived) results as standalone drugs. Like other prominent KOLS in the field such as Dr. Jay Sykler, Dr. Danne believes combination therapies are the future of type 1 diabetes research and treatment as researchers hope to “personalize it and find the right combination for each patient,” elevating VERA Plus as the next logical step for INNODIA.

The Future is Faster: Stanford’s Dr. Eric Appel Spotlights Ultra-Fast and Ultra-Stable Insulin Excipient Discovered Via High-Throughput Screenin, Fuels Newly-Launched Start-Up Surf Bio

Stanford University’s Dr. Eric Appel took to the virtual stage to discuss his lab’s ongoing work to develop ultra-fast and ultra-stable insulin formulations. More specifically, Dr. Appel’s research – which has thus far progressed through diabetic pig models – aims to create monomeric formulations of insulin. Traditionally, rapid-acting insulins have largely used hexameric formulations, which are limited by their speed of dissociation to the monomeric form.


  • The challenge is that monomeric insulin is notoriously unstable and tends to aggregate, largely due to insulin adsorption at the air-water interface. Insulin is drawn to the surface of the liquid, where it partially unfolds and aggregates into inactive amyloid fibrils. To counter this, Dr. Appel’s group sought to find a copolymer that could be used as an excipient (an inactive substance that serves as the vehicle or medium for a drug) to prevent this insulin adsorption, thus enhancing stability at the monomeric form. To do so, the group used high-throughput screening of hundreds of unique copolymers to find a candidate. The top-performing excipient candidate was then selected for further testing and added to Humalog (fast-acting insulin lispro) to create “ultrafast-absorbing insulin lispro” or “UFAL.” Data in porcine models, published in Science Translational Medicine, showed that UFAL exhibited peak action at 9 ± 4 min – a marked improvement to commercial Humalog, which has a peak action at 25 ± 10 min. When the porcine data was extrapolated to validated human models, UFAL was shown to be more than 4x faster than Humalog.  

  • On top of novel ultra-fast formulations, Dr. Appel’s lab is examining whether the excipient can be added to existing insulins to create “ultra-stable” formulations that address challenges posed by the global insulin cold chain (i.e., many countries do not have the cold storage capabilities needed to safely transport insulin at this time). Currently unpublished data shows that adding the excipient (MoNi) to Humulin increased time of stability from a week to six months. Further testing also showed that the PK/PD profile of Humulin was not changed after six months when formulated with MoNi in diabetic mouse models.  

  • Notably, Dr. Appel’s work on ultra-rapid insulins has received significant media attention. The data in porcine models was covered by an impressive spread of outlets, including US News & World Report, National Academy of Engineering, Futurity, SciTech Daily, EurekaAlert!, Consumer Health Day, Fierce Biotech, Medical Press, Weird News, New Atlas, Drug Target Review, News Break, MedPage Today, Industry Leaders Mag, Home Health Choices, Medical Xpress, CSIRO RAMP, Tech News Lit, X-mol, The Healthy Fold, and GEN News. Many of Dr. Appel’s research goals reminded us of Thermalin, another company dedicated to ultra-fast and ultra-stable formulations. Neither Thermalin nor Dr. Appel has advanced their candidates to clinical trials, and while we hope to see momentum pick up in this important area of innovation soon, we also note that pricing pressure in insulin globally won’t likely help, nor will the relative challenges as well as perceived challenges of using MDI. On the other hand, progress in insulin delivery should help enormously.

    • During his disclosures, Dr. Appel shared that he recently (January 2021) helped launch a company called Surf Bio, which has an agreement with Stanford to license the intellectual property. While we could not yet locate a dedicated website for Surf Bio, given its impressive leadership with Mr. Bryan Mazlish (former President of Bigfoot Biomedical) signed on as CEO, Dr. Jennifer Schneider (Founder and former CEO of Mode AGC, developers of the AID algorithm and software in Insulet’s Omnipod Horizon 5) as President & CMO, and Dr. Appel as Chief Scientific Advisor, we expect to hear more about this company and how it plans to advance these next-gen insulin formulations soon. 


ViaCyte CSO Dr. Kevin D’Amour provides update on stem cell-derived islet pipeline: early evidence of TIR benefit and acute cell survival as major barrier

During Thursday’s JDRF symposium, ViaCyte’s CSO Dr. Kevin D’Amour presented a fascinating update on the company’s efforts to develop stem cell-derived islets as a functional cure for type 1 diabetes. At present, two of ViaCyte’s three products candidates are in phase 2 clinical trials – (i) PEC-Direct, ViaCyte’s “open device” allowing direct vascularization; and (ii) PEC-Encap, ViaCyte’s “Encaptra device” in development with material science company W.L. Gore & Associates. As a reminder, while the two products vary in delivery method, both use the company’s proprietary PEC-01 pluripotent stem cell line, which differentiates into pancreatic endoderm cells once implanted.

  • PEC-Direct: PEC-Direct is the system that allows direct vascularization of the grafted islet cells in the device and therefore also requires continuous immunosuppression (intended for people with high-risk type 1 diabetes). PEC-Direct has shown some early evidence of clinical benefit, but Dr. D’Amour emphasized that addressing issues of acute cell survival is the company’s current priority, as only 30% of the currently enrolled trial participants exhibited de novo C-peptide production (ranging from ~0.98 to 0.30 ng/mL). When questioned on why the majority of participants failed to demonstrate measurable C-peptide, Dr. D’Amour attributed the finding to “low cute cell survival of transplanted cells…in patients that don’t see C-peptide.” This was also observed in pre-clinical models, where weak C-peptide production in rat models was “almost exclusively tie[ed] back to the quantitative level of cells that survive in the four to six weeks post-transplant.” Today, we also caught a first glimpse of CGM data produced by PEC-Direct trial participants: an “exemplary subject” demonstrated a 22% increase in Time in Range (~5.3 hours per day) and 5% decrease in time <70 mg/dL. Albeit from n=1, this was promising evidence that –with improved engraftment to increase both the frequency and magnitude of benefit – PEC-Direct could meaningfully impact glucose measures beyond C-peptide. Given that ViaCyte expanded its partnership with Gore in March 2021 to include PEC-Direct, we wonder if next steps will involve updating the open device’s membrane to see if engraftment improves. ViaCyte is also in discussions with FDA regarding possible Regenerative Medicine Advanced Therapy (RMAT) designation for PEC-Direct, which would expedite clinical development.

  • PEC-Encap: PEC-Encap is the company’s system that involves fully enclosed islet cells in an implantable, semi-permeable membrane (meant for all people with type 1 diabetes). As a reminder, PEC-Encap development was paused in early 2017 after several patients exhibited a foreign body response in a phase 1/2 trial which impeded effective engraftment. After teaming up with Gore to refine the device, however, PEC-Encap showed promising pre-clinical results (see ADA 2018), which have since been confirmed by “initial results” from PEC-Encap’s current phase 1/2 trial, slated to complete in October 2022.

  • PEC-QT: PEC-QT is ViaCyte’s newest, preclinical program of gene-edited, immune-evasive islet cells within an open device – in development with CRISPR Therapeutics. Dr. D’Amour clarified that ViaCyte aims to produce PEC-QT cells by performing a “one time gene editing event” on CyT49 stem cells to knock in and knock out certain key genes to achieve the immune evasive property. Once this is achieved, Dr. D’Amour highlighted that the immune-evasive cell line could be used to create a wide variety of cell types (not just pancreatic cells), potentially broadening ViaCyte’s focus to other disease areas. For diabetes in particular, ViaCyte also plans to apply PEC-QT to type 2 diabetes, given the device’s potential for higher dosing (more cells in device) and non-necessity of immunosuppression. Given his expertise, we’d be curious to hear Dr. D’Amour’s thoughts on the number of gene targets that would have to be altered to achieve this immune evasive state; in a recent interview, Dr. Doug Melton (Founder of Semma Therapeutics, now acquired by Vertex) estimated between five and 50 targets.

Rybelsus Data Reinforces Oral GLP-1 Efficacy across Type 2 Diabetes Continuum; Potential for Oral GLP-1s in T1D Emerges During Q&A

A star-studded Novo Nordisk symposium explored the reasons why oral formulations of GLP-1, like first-to-market Rybelsus (oral semaglutide), may benefit a diverse population of people with type 2 diabetes. To start, symposium moderator Dr. Richard Pratley explained the medical rationale for oral semaglutide and reviewed glycemic (-1.0 to 1.4% at the 14 mg dose) and weight loss data (-3.4 to 4.4 kg at the 14 mg dose) from the phase 3 PIONEER program. While the benefits of oral semaglutide do not differ substantially from injectable formulations, Dr. Pratley emphasized that its oral formulation may lead to initiation of GLP-1 treatment earlier in the continuum of the disease and may improve acceptance and adherence for some patients, who would prefer to avoid injectables. Notably, Dr. Pratley also emphasized that oral semaglutide is an “easy medication for primary care providers to prescribe” – a theme that came up throughout today’s session. 

  • Dr. Alice Cheng (University of Toronto) took to the virtual stage to discuss early initiation of an oral GLP-1. To do so, Dr. Cheng used the case example of “James,” a 42-year-old male recently diagnosed with type 2 diabetes; James is currently treated with metformin, but his A1c continues to be above target. Using the ADA’s 2021 guidelines for glucose-lowering medications in type 2 diabetes, Dr. Cheng shared that there indeed would be support for James to initiate an oral GLP-1. For patients without established ASCVD, CKD, HF, or risk factors (who would strongly be recommended to initiate a GLP-1 or SGLT-2 inhibitor), the ADA recommends using individualized goals to reach the patient’s target A1c. GLP-1s are recommended both for patients who (i) need to minimize hypoglycemia; or (ii) need to minimize weight gain/promote weight loss. Compared to other potential options like SGLT-2 inhibitors and DPP-4 inhibitors, data from PIONEER 2 showed that oral semaglutide had a greater benefit on A1c vs. empagliflozin (comparable weight loss), and data from PIONEER 3 showed that oral semaglutide had stronger A1c declines and weight loss vs. sitagliptin. As such, Dr. Cheng would strongly consider an oral GLP-1 for James due to its variety of metabolic benefits at his stage in the disease continuum.

  • Next, Dr. Esteban Jódar Gimeno (Universitary Hospital Quironsalud Madrid) presented applications of oral GLP-1 at the other end of the diabetes progression spectrum, using the case study of “Alice,” a 65-year-old woman who has had type 2 diabetes for 18 years; Alice is currently treated with metformin and insulin, has experienced two hypoglycemic episodes in the last month, and her A1c has gradually increased over the past year to 9.5%. In order to determine whether oral semaglutide would be a viable next step for Alice, Dr. Gimeno presented data from PIONEER 8, which examined the drug as an add-on to insulin. At both 26 and 52 weeks of treatment, oral semaglutide conferred statistically significant improvements to A1c and body weight vs. placebo. Importantly, these benefits were seen without an increase in hypoglycemia, a key concern for this patient population. Furthermore, across all PIONEER program trials, no interaction was seen between A1c benefits and either duration of diabetes or baseline age. Therefore, Dr. Gimeno believes oral semaglutide would be a “good option” for Alice due to its aforementioned metabolic benefits, CV safety as proven by PIONEER 6, and low risk of hypoglycemia.

Select Q&A

Dr. Pratley: From the audience, Dr. Cheng, what are your thoughts on using oral semaglutide in adolescents?

Dr. Cheng: It’s an interesting prospect because we know the benefits of GLP-1 RAs from a weight perspective as well, not just the metabolic effects on A1c. There is a clinical trial underway specifically looking at the effects of oral semaglutide in adolescents with type 2 diabetes [PIONEER TEEN], so I look forward to the results from an efficacy and safety perspective as well in that group. In that group, I think oral medications may be more acceptable than a once-weekly injection, but again, I think patient preference is something that needs to be discussed. 

Dr. Pratley: We have seen efficacy of GLP-1 RAs in adolescents with type 2 diabetes, for example with liraglutide. I expect the efficacy with oral semaglutide would be very similar to what we see in adults, but of course, we need to wait for the trial and updated label for use in this population.

Dr. Pratley: Also from the audience, are there any studies of oral semaglutide in patients with type 1 diabetes? There is no study in patients with type 1 diabetes. A related question, however, is “could this be used in patients with adult-onset autoimmune diabetes?” That’s maybe not the same more aggressive type of diabetes we see in kids, but one where we’re more likely to see preservation of C-peptide. This is total speculation, but Dr. Gimeno, what are your thoughts?

Dr. Gimeno: In fact, we have some evidence of when the GLP-1 RAs could be used in this population, and they are safe and efficacious. This is very exciting to think about the potential benefit in maintaining and even increasing the beta cell mass, but, in fact, we need trials. We are talking about something similar to people with type 1 diabetes, and we know the GLP-1 RAs have some benefits in that population, with very minor decreases in hemoglobin but important weight benefits and some kind of stabilization of blood glucose metrics. I think this should be highlighted because, due to intensive therapy, we have ~1/3 of the type 1 population with obesity. Probably, this benefit on body weight would be interesting.

Dr. Pratley: Your latter point is very important. In the US, 60% of adults with type 1 diabetes are either overweight or have obesity. They struggle with it as much as my patients with type 2 diabetes. We did a literature review that showed there is no trial or evidence of weight loss medications, or even diets, to help people with type 1 diabetes manage their weight, so I think it’s really an interesting idea about the use of that medication in patients with auto-immune diabetes, but of course, we need more trials.

Dr. Pratley: To summarize, in the type 2 population, which patient segment do you think is ideal for oral semaglutide? And are there any segments in which you wouldn’t use oral semaglutide?

Dr. Cheng: I see a role for it throughout the journey of someone living with type 2 – from the beginning right through to the end, perhaps for different reasons, but I do see the value. In terms of who not to use it in, if I have patients who have frailty, who are struggling to maintain weight or eat properly, or their appetite is already suppressed for other reasons, that is a group where I might shy away from a GLP-1 RA.

vTv CSO Dr. Carman Valcarce discusses type 1 candidate TTP399 for T1D adjunctive therapy as positions “beta cell rest” as an additional potential advantage

Also during the JDRF symposium, vTv’s CSO Dr. Carmen Valcarce reviewed the current data supporting type 1 diabetes adjunct candidate TTP399, and intriguingly, positioned the glucokinase activator as being “complementary to most of the alternatives that are currently being pursued to improve the lives of people with type 1 diabetes” due to its ability to “potentially provide” “beta cell rest.” This was the first time we heard the longstanding theory of “beta cell rest” directly being applied to TTP399. While it seems to align well with the drug’s proposed mechanism of action, we also imagine that as a narrative, it also may be preferred to discussions mainly about adjunctive therapy, which are not as popular after the US regulatory experience of SGLT-2s for type 1. (From our view, that shouldn’t be seen as a negative as much as potential for the future for adequate safety – we are eager to hear more about continuous ketone monitoring in the future. Going back to today’s learnings, the highly regarded Dr. Valcarce explained that by selective activation of hepatic glucokinase, TTP399 increases the uptake of glucose into the liver, thus reducing insulin demand from the pancreatic beta cells, allowing them to rest. While only speculative at this point, we surmise that pairing TTP399 with something like stem cell-derived islet cells in the future may improve the cells’ long-term effectiveness, while keeping TTP399’s known benefits of reducing hypoglycemia and minimizing DKA.     

  • Dr. Valcarce today also presented an analysis of safety data to show that TTP399’s benefits on hypoglycemia (~40% reduction) and ketosis (trend toward benefit, no DKA) are unlikely to be related to reductions in insulin. Because the phase 2 Simplici-T1 study encouraged investigators and patients to adjust insulin to meet treatment goals of FPG 80-130 mg/dL and post-meal glucose <180 mg/dL, some have speculated that hypoglycemia/ketosis benefits may have only occurred in individuals who reduced their insulin dosages. On the contrary, a grouped analysis of participants who had their insulin reduced, kept stable, or increased showed that this was not likely the case, as event decreases were seen in all three groups (hypoglycemia in the increased insulin group could not be accurately measured, however).

  • There was notable interest in TTP399 during Q&A, with multiple audience members asking questions ranging from route of administration (one-daily oral) to whether the drug causes an increase in liver fat (no increases in liver fat identified). While not mentioned during today’s talk, we note that TTP399 is currently progressing through a mechanistic study, and the company plans to launch its pivotal trial in 4Q21. Given the absence of widely used adjunctive therapies for type 1 diabetes, we hope to see TTP399 continue to make progress on this important front.  

New Pooled Analysis Reveals 99.5% of People with Type 1 and Type 2 Achieve Hypoglycemia Rescue Success with Lilly’s Baqsimi Nasal Glucagon

Data continues to build to support the strong safety and efficacy profile of Lilly’s nasal glucagon Baqsimi in both people with type 1 and type 2 diabetes facing insulin induced hypoglycemia. University of Leicester’s Prof. Kamlesh Khunti presented results from a new pooled analysis of three randomized, cross-over studies comparing 3 mg nasal glucagon and 1 mg reconstituted injectable glucagon (n=225 people with type 1 and type 2 diabetes). In the post-hoc analysis, nasal and injectable glucagon were shown to have comparable efficacy, with 99.5% of participants on nasal glucagon and 100% of participants on injectable glucagon receiving treatment success within 30 minutes (the one participant who did not achieve success on nasal glucagon later did so at 40 minutes). Furthermore, these results were consistent in both people with type 1 and type 2 diabetes: 99.4% of people using nasal glucagon with type 1 diabetes achieved success vs. 100% of people with type 2 diabetes. In terms of safety, the frequency of gastrointestinal treatment-emergent adverse events was similar between those on nasal and injectable glucagon, though nasal glucagon had higher rates of nasal administration-related (but transient) side effects like runny nose, nasal congestion, watery eyes, etc. Adverse events were not broken down by type 1 vs. type 2 diabetes, and we are curious to see if any side effects were more common in either condition.

  • Lilly has already begun to reap the benefits of its newer glucagon option. In 1Q21, for the second quarter ever, Baqsimi outpaced traditional glucagon sales at $24 million (+37% YOY and +2% Q/Q) vs. $22 million (-22% YOY and -1% Q/Q). Given the data presented today, we believe more investment could be put into bringing next-generation glucagon options to people with type 2 diabetes, as well as type 1 diabetes, particularly those with longer duration of insulin use at higher risk of hypoglycemia. Back at ATTD 2017, a real-world study of nasal glucagon showed that 98% of caregivers were able to administer nasal glucagon within two minutes to adults with type 1 diabetes experiencing moderate-to-severe hypoglycemia. In the future, we hope to see more data on real-world applications of nasal glucagon in type 2 diabetes to see if these success rates can be replicated.

The Past 100 Years of Insulin Inform the Future of Diabetes Management: What to Expect with New Insulins and Technologies

In a Novo Nordisk-sponsored symposium, Drs. Chantal Mathieu (KU Leuven), Christophe de Block (University of Antwerp), and Andreas Liebl (Fachklinik Bad Heilbrunn) discussed the future of insulin (basal, oral, glucose sensitive) and diabetes devices. Dr. de Block began the session with an overview of the past 100 years of insulin from discovery to now, highlighting those advancements turned diabetes into a manageable condition continually moving toward personalized therapies. However, specific clinical challenges associated with basal insulin include hypoglycemia, complexity, and poor treatment adherence. In her presentation on basal insulin, Dr. Mathieu mentioned several once-weekly basal insulin formulations currently in development: Fc fusion (candidates from AZ, Lilly, and Hanmi), PEGlyated (candidate from Antria/Rezolute), and acylated (insulin icodec from Novo Nordisk). The rest of the conversation focused on insulin icodec, which wrapped up its phase 2 trials in 2020; insulin icodec was comparable to glargine U100 in lowering A1c and required lower doses in insulin-naïve participants with type 2 diabetes. Pending positive phase 3 results for insulin icodec, the once-weekly basal insulin could improve treatment adherence due to its increased ease of use.

  • Dr. Liebl provided a very interesting commentary on the future of insulin in the next 100 years. The four main directions for insulin innovation are (i) improving convenience and adherence; (ii) reducing hypoglycemia; (iii) optimizing insulin therapy through technology; and (iv) moving toward a cure. Once-weekly/oral insulin, glucose-sensitive insulin, devices and digital solutions, and stem cell therapies will be important to each of these directions, respectively. Dr. Liebl posited that once-weekly insulin can be co-formulated with once-weekly incretins (e.g., IcoSema), which would lead to better treatment adherence and added benefits of cardio and renal protection. Development of oral insulin could lead to earlier treatment initiation and increased patient compliance, but these benefits are currently ways away in the future given challenges in permeability, enzymatic stability, food effects, and bioavailability. Dr. Liebl highlighted the potential of oral insulin 338, a long-acting basal insulin tablet with sodium caprate absorption-enhancer, which showed similar glucose-lowering effects to insulin glargine after eight weeks of use. However, the dose requirement for insulin 338 was almost 60 times higher than that of glargine – demonstrating the level of additional R&D required to make the oral insulin a feasible treatment option. Oral injection devices, which release insulin into the stomach lining, and “smart” insulins with glucose sensitivity, which change conformation from active and inactive states based on surrounding glucose concentrations, are other interesting yet far away options for insulin delivery. Diabetes technology holds promise as a more immediate solution to improving insulin dosing, with goals of multi-hormonal closed loop artificial pancreas systems to deliver insulin and glucagon automatically.

  • Panelists reflected on advancements and discussed which options were most feasible during Q&A. In a poll, the majority of the audience stated that they are most excited for glucose-sensitive insulin in the future, followed by artificial pancreas. Dr. Liebl clarified that while this option is exciting, its bioavailability is currently very low, making the artificial pancreas a more thrilling option in the near future. Regarding the use of insulin in type 2 diabetes, Dr. de Block follows recommendations from the most recent EASD/ADA consensus paper, which reserves insulin for those who do not respond well on GLP-1s and those hospitalized in critical conditions. However, as pointed out by Dr. Liebl, we are now seeing more patients with longstanding type 2 that now require insulin therapy- which has potential to be combined with GLP-1s for additional cardio/renal benefits. He was also fascinated by the idea of combining GLP-1s with once-weekly basal insulin as an option to lower the barrier for initiating insulin. It may also be possible to use once-weekly basal insulin in patients with type 1 diabetes, but only if their lifestyles enable them to dose basal the same way daily. Other patients with more flexible lifestyles and dosing schemes, like athletes, would be better candidates for pumps.

Drs. Christophe de Block and Chris Byrne on Best Practices for NAFLD/NASH and Diabetes

In this comprehensive session on NAFLD/NASH, Dr. Christophe de Block (Antwerp University), Mr. Jonathan Mertens (Antwerp University), and Dr. Chris Byrne (University of Southampton) discussed the prevalence NAFLD/NASH is in people with diabetes and best treatment and screening methods. Dr. de Block and Mr. Mertens focused on NAFLD in type 1 diabetes in their presentation, utilizing data on its prevalence from meta-analyses and original data from the Antwerp University Hospital. Findings showed that prevalence of NAFLD can range from 5-8% if screening with MRI to 22-27% if screening with ultrasound, with added effects from referral bias. Thus, Dr. de Block emphasized the need for an accurate diagnostic algorithm to catch NAFLD early and treat it appropriately to avoid negative outcomes. After NAFLD is identified, it is strongly associated with visceral fat accumulation and insulin resistance. With this, people with type 1 diabetes and NAFLD show higher prevalence and incidence of cardiovascular disease. While NAFLD is independently associated with increased rates of CVD, more studies need to be done to determine how NAFLD contributes to CVD in a non-glycemic way. Dr. de Block also stated that NAFLD is independently associated with (i) increased prevalence of CKD and retinopathy; (ii) increased incidence of CKD; and (iii) increased prevalence of distal symmetric polyneuropathy in type 1 diabetes.

  • Shifting gears to type 2 diabetes, Dr. Byrne focused on NAFLD/NASH in metabolic syndrome and early type 2 diabetes. She described the relationship between NAFLD and type 2 diabetes as a “vicious spiral of worsening diseases,” given the overlapping risks of metabolic syndrome, CVD, CKD, and liver fibrosis. A 2018 study published in Diabetes Care showed that NAFLD increases risk of incident diabetes (HR: 2.22). Those with type 2 diabetes and NAFLD were also at higher risk for incidence/recurrent CVD event (HR: 1.70). Dr. Byrne believes that treatment of NAFLD should be in accordance with NAFLD severity and diabetes status; she suggests pioglitazone, Vitamin E, GLP-1 agonists, and SGLT-2 inhibitors for people with NAFLD and type 2. For early type 2 and NASH specifically, the best treatment option may be pioglitazone and/or GLP-1 agonist.

Diabetic Nephropathy-Focused Symposium Highlights the Future Potential of Mineralocorticoid Receptor Antagonists (MRA)

A Bayer and Sciarc GmbH-sponsored symposium focused on the future of treatments for diabetic nephropathy, highlighting the potential of MRA finerenone as a new treatment option. Dr. Jay Skyler (University of Miami) began the session with an overview of current unmet needs in diabetic kidney disease. Notably, diabetic nephropathy is the leading cause of end-stage renal disease (ESRD) in the US. As the rates of diabetes complications like acute MI and stroke have decreased since 1990, rates of ESRD have remained static. Trials like UKPDS, ACCORD, and ADVANCE have shown that tight glycemic control can prevent microvascular complications in type 2 diabetes, but studies of national insurance outcomes (commercial HMO and Medicaid) have shown that control of A1c levels is not improving. Albuminuria is present in at least 20% of youth with type 2 diabetes, indicating that nephropathy begins early in type 2 diabetes progression. Unfortunately, only 85% of people with type 2 are tested for eGFR, while 47% are tested for UACR- both of which are independent predictors of CV mortality. With these unmet needs in mind, Drs. Peter Rossing (Steno Diabetes Center) and Janet McGill (Washington University of St. Louis) presented findings from FIDELIO-DKD that showed MRA finerenone can be a promising treatment for patients with type 2 diabetes and diabetic nephropathy on baseline RAS therapy. Finerenone gave a significant 18% relative risk reduction for the primary composite endpoint of adverse kidney outcomes (i.e., kidney failure, sustained ≥40% decrease in eGFR, or renal death; HR=0.82, 95% CI: 0.73-0.93, p=0.0014). On the secondary composite CV endpoint (i.e., CV death, non-fatal MI, non-fatal stroke, or heart failure hospitalization), finerenone also reduced risk by 14% versus placebo (HR=0.86, p=0.0339). Finerenone has a distinct mechanism of action that targets inflammation and fibrosis to halt CKD progression, making it possible to utilize in tandem with other medications (though extensive combination trials have yet to be performed). In the Q&A, all three speakers discussed whether finerenone could one day be used in combination with both GLP-1 receptor agonists and SGLT-2 inhibitors. While this could be possible in theory, Dr. Skyler stated that trials are necessary to see if all three can work well together. Dr. Rossing added that FIDELIO-DKD data on small subgroups using SGLT-2s and GLP-1s with finerenone will be presented at ADA 2021, but more trials are needed to determine if all three therapies have a synergistic effect on glucose management and cardio-renal outcomes. As FIDELIO-DKD excluded people with type 1 diabetes, further study is also needed to determine if finerenone can benefit type 1 populations who are not able to utilize SGLT-2 inhibitors given the increased risk of DKA.

Drs. Denice Feig and Yariv Yogev Debate Metformin’s Use in Pregnancy

In this session on whether the benefit of metformin use in pregnancy outweighs potential side effects in infants, Dr. Denice Feig (Unniversity of Toronto) debated for metformin and Dr. Yariv Yogev (Tel Aviv University) debated against.

  • Dr. Feig argued that metformin improves both maternal and neonatal outcomes in women with gestational and type 2 diabetes. Alone, metformin’s advantages include effective glucose lowering, low risk of hypoglycemia and weight gain, low cost, and easy to use in combination with other therapies. Results from six meta-analyses of RCTs of metformin versus insulin in gestational diabetes showed that metformin led to less weight gain, preeclampsia, large weight for gestational age (LGA), and neonatal hypoglycemia. However, metformin’s association to lower gestational age remains questionable. Compared to glyburide, metformin use in gestational diabetes led to significantly less maternal weight gain, macrosomia, and LGA. Switching gears to type 2 diabetes, the MiTy study found that metformin use was associated with reduced maternal weight gain, reduced insulin doses, improved glycemic control, reduced c-section rates, and lower adiposity. However, insulin use was also associated with higher rates of infants small for gestational age (SGA), though this was only 3% higher than the average national rate. Longitudinal studies of metformin use in women with gestational diabetes in Australia and New Zealand showed no difference in weight at seven and four years after birth, respectively.  In conclusion, Dr. Feig believes that metformin should be offered to all women with diabetes given its long-term safety.

Dr. Yogev argued against metformin use in pregnancy due to lack of benefit compared to insulin. A 2008 trial comparing the two found no difference in primary composite outcome (neonatal hypoglycemia, respiratory distress, need for phototherapy, birth trauma, 5-minute Apgar score less than 7, or prematurity) between the metformin and insulin-treated groups. A 2019 review also showed that while metformin led to decreased weight gain and preeclampsia in women with gestational diabetes, ~85% of participants needed added insulin to achieve these effects. Dr. Yogev also cited other longitudinal trials, which showed adverse effects of growth inhibition, smaller birthweight, and higher fat mass at four and nine years old after metformin use. In conclusion, Dr. Yogev focused on the more positive effects of insulin over metformin, which are also reflected in the 2018 ADA and ACOG guidelines.

Decision Support and Digital Health Highlights

Three-Month Real-World Study (n=66) of DreaMed Advisor Pro Platform Sees Similar Glycemic Benefits and Provider Satisfaction as in Advice4U RCT

A three-month real-world study (n=66) of the DreaMed Advisor Pro platform showed similar glycemic benefits and provider satisfaction as in the 24-week Advice4U RCT (n=122) read out at ATTD 2020. DreaMed’s Advisor Pro is a decision support system that offers insulin dosing recommendations based on pump, CGM, and BGM data from the Glooko diabetes management platform. On Day #2, Dr. Revital Nimri (Schneider Children’s Medical Center, Israel) read out the results of the real-world study, which included type 1s on insulin pumps and CGM with baseline mean glucose values <182 mg/dl, and with at least one Advisor recommendation between baseline and three months. Sixty-six participants from nine pediatric clinics and two adult clinics were included in the study, 63 of whom were children and adolescents. As a reminder, the Advice4U study participants were all pediatric patients and were randomized to either receive expert physician recommendations or Advisor Pro recommendations. At baseline, participants in the real-world study had an average GMI of 8.4%, which was the same as both the control and intervention arms of the Advice4U study. At three months, participants in the real-world study saw a significant ~0.3% GMI reduction to 8.1%, which is nearly identical to the ~0.4% decline to 8.0% in the Advisor Pro group of the Advice4U trial. Participants spent +1.2 hours/day in range (42%) compared to baseline and saw significant reductions in time above range and time in severe hyperglycemia (<250 mg/dl), falling -1.4 hours/day to 56% >180 mg/dl and -58 minutes/day to 26% >250 mg/dl. While these are significant improvements, the three-month averages are still significantly higher than the consensus targets for <25% above range and <5% in severe hyperglycemia and significantly below the consensus target for >70% Time in Range. Participants also saw slight but statistically significant +6 minutes/day increase in time below range to 1.4%; however, that figure still achieves the consensus target of <4% time <70 mg/dl. On average, participants’ total daily dose increased +3 U/day to 41 U/day at three months. The majority of this increase in insulin came from a significant +2 U/day increase in participants’ daily basal dose. Participants’ average daily bolus dose (23 U) and number of boluses a day (6 boluses) did not significantly change from baseline.

 

A1c Change

Time >250 mg/dl

Time >180 mg/dl

Time in Range

Time <70 mg/dl

Time <54 mg/dl

Real-world study

Advisor Pro

8.1% (-0.3%)

26% (-58 min/day)

56% (-1.4 hours/day)

42% (+1.2 hours/day)

1.4% (+6 minutes/day)

0.3% (+1 min/day)

Advice4U RCT

Advisor Pro arm

8.0% (-0.4%)

18%

25%

54%

2.4%

1.0%

Physician arm

8.1% (-0.3%)

18%

24%

55%

2.6%

0.9%

  • Dr. Nimri also presented data on healthcare provider (n=18) satisfaction with the Advisor Pro platform, which suggested that providers are as highly satisfied with the system in the real-world as in the Advice4U RCT. After using the Advisor Pro system for three months, 16 physicians and two nurse practitioners from US pediatric endocrinology practices completed the satisfaction questionnaire from the Advce4U study (n=8 pediatric centers). In both the Advice4U study and the real-world study, participants scored the Advisor Pro platform ~4.5 out of 5 for its utility, simplicity and safety and agreed/strongly agreed (~4.4/5) that their overall experience with the Advisor Pro system was positive. The most notable differences were that the real-world participants scored the Advisor Pro system slightly higher than did Advice4U participants in reliability (4.4 vs. 4.1) and similarity to therapy adjustments they would make (4.4 vs. 3.3) and scored the system slightly lower in saving time than in the Advice4U study (3.9 vs. 4.3). In the real-world study, 13 of 18 providers reported that one of the key benefits of Advisor Pro is that it facilitated useful discussion with patients and their families.

  • DreaMed is also developing its “next-gen” Advisor Pro platform, which will offer support for type 1s and type 2s on pump therapy or MDI. We saw data on the efficacy of the Advisor Pro platform in type 1s and type 2s on MDI at other sessions at ATTD 2021. DreaMed also intends to have the mobile app’s bolus calculator “auto update” based on the Advisor-generated treatment plans after HCP approval. During her presentation, Dr. Nimri also noted that DreaMed intends to integrate Advisor into US clinics’ EMRs, which to us is hugely important in terms of integrating into providers’ workflows. We’re also looking forward to seeing the results of a six-month study on the real-world efficacy of DreaMed Advisor Pro in terms of workflow, system satisfaction, and financial adoption metrics. The study was announced in March as part of a partnership with Yale New Haven Health Systems.

DreaMed’s Advisor Pro Decision Support Software Shown to be Non-Inferior to Expert Provider Recommendations for T1s on MDI; Agreement Rate Between Advisor Pro and Expert Providers Higher Than Agreement Between Providers

Dr. Amir Tirosh (Sackler School of Medicine) presented data from a survey study of 20 international physicians indicating that the DreaMed Advisor Pro decision support system is non-inferior at giving insulin dosing advice to expert providers for patients with type 1 on MDI. As a reminder, Advisor Pro offers insulin dosing recommendations based on CGM or BGM data and requires sign-off from providers before patients are allowed to change dosing regimens to ensure providers remain involved in the insulin dosing process. A number of studies have investigated the non-inferiority of Advisor Pro’s recommendations for patients on pump therapy (most notably, the Advice4U study from ATTD 2020). Today, Dr. Tirosh presented new data on insulin dosing recommendations for case study patients on multiple daily injections. Specifically, DreaMed surveyed 20 providers with an average of over 15 years of experience treating people with type 1. DreaMed provided the physicians with anonymized data from 17 patients with type 1 on MDI who had an average age of 35, average Time in Range of 58%, and A1c of 7.4%. The vast majority of the case study patients were using CGM. Participants were asked to provide the insulin dose adjustment advice they would give to these patients. To analyze survey results, Dr. Tirosh’s group evaluated the “level of agreement/disagreement on the direction of insulin dose adjustments” and “the magnitude of change in insulin dose adjustments” both between providers and between providers and the Advisor Pro recommendations.

  • Starting with direction of change, Advisor Pro agreed with providers on adjustments 67% of the time. For comparison, providers agreed with each other 62% of the time, demonstrating non-inferiority of the recommendations from Advisor Pro (p<0.001). Looking at disagreements, Advisor Pro disagreed with the direction of adjustment suggested by providers 19% of the time, equal to the 19% disagreement rate amongst providers themselves. Again, this result suggests non-inferiority of the Advisor Pro recommendations to those from expert providers (p=0.004).

  • Looking at the magnitude of the suggested insulin dose adjustment, Advisor Pro was again non-inferior to expert advice, but recommended more conservative adjustments than did physicians. As Dr. Tirosh expressed, this is due to the built-in safety features of Advisor Pro that limit the size of recommended changes to roughly 20%-30% of established insulin doses. As decision support tools, including apps, closed loop systems, and smart insulin delivery devices, have become more common, some have said they would be interested to see data on how Advisor Pro may be able to mitigate clinical inertia compared to some of these other interventions. We agree, though of course coming to agreement on how one would standardize definition for inertia isn’t particularly straightforward – and that’s the first step, defining it! As well, we would like to see if any calculations have been done to identify which populations for whom provider-based decision support tools like Advisor Pro have the largest potential to improve glycemic control and outcomes. Sounds like an excellent project for our tech team, the authors of the Digital Diabetes Coaching competitive landscape to suss out before the final ATTD report comes back to you – expect a call, if you’ve got a lot of good experience here, or write to the team to let them know your thoughts! We’ll send a copy of Nudge, above, to the first and fifth of you to respond!

Non-Inferiority of DreaMed’s Advisor Pro Insulin Dose Decision Support Software Compared to Expert Provider Advice in the Management of T2D on MDI; 60% Agreement Rate Between Advisor Pro and Expert Advice Compared to 51% Agreement Between Providers

In the second of back-to-back DreaMed Advisor Pro presentations, Dr. Ofri Mosenzon (Hassadah Medical Center) showed data indicating non-inferiority of Advisor Pro insulin dosing recommendations to advice from expert providers for patients with type 2 diabetes on MDI. In this session, Dr. Mosenzon presented data from a survey of 16 physicians with an average of 18 years of experience treating patients with diabetes. While the majority of physicians came from endocrinology (n=13), DreaMed’s survey population also included two general medicine practitioners and one internal medicine physician. All physicians surveyed in the study were presented 15 case studies from patients with type 2 who had an average age of 63 and an average A1c of 7.5%. Additionally, patients had an average Time in Range of 65% and 83% of patients used CGM technology. Primary outcomes in this study were agreement or disagreement between expert provider and Advisor Pro generated insulin regimen adjustments assessed in terms of direction of dosing change for both basal and bolus insulin doses.

  • Advisor Pro agreed with individual physician recommendations for basal dose adjustments 60% of the time. Similar to the data for type 1s, this is higher than the rate at which expert physicians agreed with each other (51%), demonstrating non-inferiority of Advisor Pro’s dosing recommendations (p<0.001). Rates of Advisor Pro vs. physician disagreement and physician vs. physician disagreement were similar at 26% and 27%, respectively. Turning to bolus insulin recommendations, Advisor Pro again saw a higher agreement rate with individual physician recommendations at 46% of the time, compared to physicians who agreed amongst themselves 43% of the time. Finally, while there was a slight increase in basal rate disagreement between Advisor Pro and expert recommendations, it was still non-inferior to provider-to-provider disagreement (, (34% vs. 33%; p<0.021). Given that this study was conducted with a population of providers well-versed in diabetes management, we are curious what the results would have looked like if DreaMed were to conduct a similar investigation among primary care providers where many individuals with type 2 receive the majority, if not all, of their diabetes-related care.

Google’s Machine Learning-Based Model for External Eye Photos Detects A1c >9%, Diabetic Retinopathy, and Macular Edema Better Than Simple Logistic Regression Using Baseline Characteristics

During a morning symposium on uses for AI in diabetes care, machine learning engineer Dr. Boris Babenko (Google) presented promising results from a new machine learning model that uses external eye images to predict whether a patient has A1c >9%, diabetic retinopathy, or macular edema. The use of external eye images (i.e., pictures of the front of the eye), rather than the retinal photos (i.e., pictures of the back of the eye) used in most previous AI-based studies. In principle, these external photos do not require any special equipment beyond a smartphone camera, while retinal photos require both a specialized fundus camera and a technician operator. The dataset used in this study came from EyePACS and VA Atlanta, with 301 EyePACS sites used for model training and 186 EyePACS sites and VA Atlanta data used for model validation. The images used for training and validation were taken using a fundus camera, but were still front-of-eye images. The validation results discussed below also represent results for images taken with the pupil undilated, as Dr. Babenko noted the goal of the model was to broaden access to diagnostics.

  • Overall, AUC for the AI-based model was around 0.7-0.8 for A1c >9%, diabetic retinopathy, and diabetic macular edema. To assess the model’s value, it was compared against a simple logistic regression model that used self-reported baseline characteristics (e.g., age, sex, ethnicity, years with diabetes). On all three fronts, the photo-based model outperformed the simple model with AUCs of 0.78 vs. 0.71, 0.75 vs. 0.71, and 0.70 vs. 0.65 for macular edema, diabetic retinopathy, and A1c >9%, respectively.

 

  • In general, AUCs of 0.7-0.8 are still too low to have clinical value as a diagnostic tool. However, Dr. Babenko suggested the model would be useful for screening and identifying high-risk patients. For reference, Dr. Babenko noted that the CDC’s Prediabetes Risk Test has an AUC in the similar 0.79-0.83 range. The value of Dr. Babenko’s model for risk stratification was illustrated in a hypothetical scenario in which resources only exist to follow-up for 5% of all patients with diabetes. If the 5% of patients chosen for follow-up were chosen at random, about 30% of those chosen would have an A1c >9%; in other words, about 30% of all people with diabetes have an A1c >9%. Using the logistic regression described above, over half of those identified by the model would have an A1c >9%; in other words, the logistic regression model performs significantly better than random selection at identifying high A1c patients. Finally, using the image-based model, 67% of selected patients would have an A1c >9%; in other words, patients chosen by the model are more than twice as likely as a random patient to have a high A1c.

  • The AI-based model showed surprisingly strong robustness even on lower resolution images. As mentioned above, the images used to train and validate this model were taken by a fundus camera; these images are likely to be higher resolution than those taken by a smartphone camera. In order to explore how the model performed on lower resolution images, the validation dataset was down sampled to various resolutions and model performance was evaluated. As expected, AUC degraded with lower resolution images; however, the model generally showed strong performance with images as small as 150 x 150 pixels. For reference, the original images used for training were 587 x 587 pixels. Of course, as Dr. Babenko noted, image resolution is not the only difference between images taken by a fundus camera and a smartphone camera – other factors to consider would be focus and lighting.

Novo Nordisk Symposium Highlights NovoPen 6’s Current Utility in Sweden and in the Future

In an extensive Novo Nordisk-sponsored symposium on the NovoPen 6 smart insulin pen system, Drs. Richard Bergenstal (International Diabetes Center), Peter Adolfsson (University of Gothenburg), Lena Landstedt-Hallin (Karolinska Institutet), and Johan Jendle (Örebro University) discussed the utility of smart pens in adult and pediatric populations. Dr. Adolfsson began the session with a summary of the results of NovoPen 6’s real-world study, SWEDEN. In adults with type 1, pen use was associated with increased Time in Range (and less hyperglycemia and level 2 hypoglycemia), fewer missed mealtime injections, and more well-dosed mealtime injections. Data also showed varied basal/bolus dosing patterns, with more differences in adults compared to youth. In youth with type 1, pen use was associated with a reduction in daily hypoglycemic events and lower time in hypoglycemia at 12 months. Taken together, these results demonstrate smart pens’ potential to enhance glycemic control. This is paired with positive health economics data, as presented by Dr. Jendle. Specifically, smart pen use can lead to improved clinical outcomes at costs lower than standard care therapy and in a way that utilizes Swedish public healthcare resources efficiently. Though NovoPen 6 is currently only approved in Sweden and Denmark, Dr. Bergenstal mused on the pen’s potential in a virtual diabetes clinic for people with type 2 diabetes. Smart pens could be valuable in closing the loop, but next steps require standardizing the way smart pen data is output into an Ambulatory Glucose Profile Report (AGP) from CGM. Dr. Bergenstal suggested a global template that each company that releases a smart pen can personalize to their pens and features that shows patients’ daily profiles and allows for individualized insulin titration. Overall, smart pen and CGM use can transform insulin management in those with type 2 diabetes in a step-wise fashion.

  • Speakers further discussed the clinical applicability of smart pens in a full panel discussion. Drs. Landstedt-Hallin and Adolfsson discussed how smart pens’ automatic logging of dose amount, date, and time is especially helpful for when people with diabetes forget if they’ve bolused before a meal. When coaching patients, it is also important to normalize making mistakes. Aptly put by Dr. Landstedt-Hallin, “if you have chronic disease, it is normal to sometime miss.” Dr. Adolfsson also made similar remarks, emphasizing that physicians should be coaches and not judges and that missing doses is not indicative of bad values. Smart pens and CGM hold great value as collaborative, educational tools between patient and healthcare provider – with frequent contact being especially important to determine best insulin dosing methods.

UVA Researchers Find That 99% of CGM Profiles (n=42,595) Can Be Classified Under a Finite Set of 483 “Motif” Representative Daily Profiles

During ATTD’s first set of oral presentations, Benjamin Lobo (University of Virginia) discussed UVA finding that 99.0% of CGM profiles (n=42,595) can be classified under a finite set of 483 “motifs” (i.e., representative daily profiles). The researchers identified 483 motifs from a training data set of 9,471 daily CGM profiles.  The robustness of the set of motifs was established by using it to classify 99.0% (n=42,595) daily CGM profiles in a testing data set.  The training and testing data sets were generated using the daily CGM profiles from six different studies (including the iDCL Protocol 3 and DIaMonD studies) which involved both type 1 and type 2 participants using a variety of treatment modes including MDI, pump therapy, and AID systems. Only 430 profiles could not be classified (matched to) one of the 483 motifs; the primary cause of the failure to match was the high percentage of out-of-range sensor readings in the unclassified daily CGM profiles. The 483 motifs were also grouped based on clinical characteristics including Time in Range, time above range, time below range, coefficient of variability, and standard deviation, which allows daily CGM profiles to be classified into an even smaller subset of prespecified groups with similar clinical characteristics.

While this may not seem notable at face value, this has hugely important clinical ramifications: Dr. Lobo suggested that clinical guidance, decision support, and predictive modeling can be built around the motif profiles to make actionable insights easy to glean from a patient’s CGM profile. In particular, we see this as hugely important research toward expanding the use of CGM in primary care spaces, as this research is foundational to creating decision support tools for primary care providers that supports them in gathering clinically actionable insights from patients’ CGM profiles. Storing CGM daily profiles as a single data point representing the motif rather than 288 glucose readings per day would also be valuable in data compression to reduce data storage and transmission needs, which will become increasingly important as more and more patient data is collected and stored.

iSpy Carbohydrate Counting App Associated with Fewer Carb Counting Errors than Standard Care; App Use Associated with 0.6% Lower A1c after 3 Months of Use Compared to Control

Drs. Jeffrey Alfonsi (Western University) and Mark Palmert (University of Toronto) presented new data on their photo-ID based carb counting software iSpy designed to help young adults with diabetes more easily count their carbohydrate intake. iSpy uses AI technology to identify foods based on images from users and can help identify “hidden” carbs that users may not have realized were in their meals. iSpy also provides portion advice and allows users to enter carbohydrate estimates so users can see how their own assessment compares to that of the iSpy system. Based on a usability study of 16 patients with type 1 between the ages of 8 and 17 years old, Dr. Alfonsi reported that 93% of users found iSpy easy to use and were satisfied with the platform. Notably, 100% of participants found iSpy was helpful in counting carbohydrates and 80% said they would use the system in their daily carb counting and diabetes management. Turing to the accuracy of iSpy, Dr. Palmert presented data from a randomized controlled trial of 22 iSpy users and 22 comparable patients under standard care as controls. All participants had completed carbohydrate counting classes and used carb counting in their standard care. The primary outcome of the study was a change in carbohydrate counting accuracy over 3 months based on user assessments of 10 foods including vegetables, grains, milks and alternatives, meats and alternatives, and two dessert items. At baseline, both intervention and control groups had similar carb counting errors, but these errors were lower in the intervention arm at 3-month follow-up with an absolute error in % of total grams of 27% compared to 38% in the control group (p=0.008). Additionally, at follow-up participants in the intervention arm had significantly fewer carb counting errors of >10g at 21% compared to standard control with an 32% errors >10g. Finally, iSpy use was associated with a lower A1c among participants in the intervention arm compared to those in standard control with A1c values at 3 months of 8% and 8.8% (p=0.03). While Dr. Palmert did note a decline in user engagement over the course of the 3-month trial, 43% of participants maintained high to medium engagement suggesting iSpy is an acceptable long-term carbohydrate counting tool.

Time in Range and Beyond A1c Highlights

Time in Range (70 mg/dl – 140 mg/dl), Time >140 mg/dl, and A1c at 24 Weeks Gestation are Consistently Predictive of Suboptimal Pregnancy Outcomes in Women with Type 1

Dr. Claire Meek (University of Cambridge) presented secondary analysis from the CONCEPTT trial that assessed the ability of various CGM metrics and biomarkers to predict neonatal outcomes in pregnant women with type 1 diabetes. As a reminder, CONCEPTT showed that using CGM in pregnant women with type 1 diabetes reduced A1c and improved both maternal and neonatal outcomes. In the study presented today, researchers used data from 157 pregnant women with type 1 from the CONCEPTT study to assess the ability of: (i) A1c, (ii) CGM metrics (i.e., mean glucose value, Time in Range (70 mg/dl-140 mg/d), time above range (>140 mg/dl), time below range (<70 mg/dl), coefficient of variation, and standard deviation), and (iii) alternative laboratory markers that measure glycemia over two weeks (i.e., glycated CD59, 1,5 anhydroglucitol, fructosamine, glycated albumin) to predict neonatal outcomes. Four pregnancy-related outcomes were assessed: preterm birth, large-for-gestational-age, neonatal hypoglycemia, and neonatal intensive care unit (NICU) admission. Overall, the study found that A1c, Time in Range (70 mg/dl – 140 mg/dl during pregnancy), and time above range (>140 mg/dl) were the “most consistently predictive” metrics and showed “good predictive ability” for many outcomes. Generally, alternative lab markers and other CGM metrics had a lower predictive value than did A1c, Time in Range, and time >140 mg/dl. However, glycated CD59 showed promising predictive power.

  • Looking at the big picture (shown below), Time in Range was a better early predictor (12 and 24 weeks), while A1c was a better predictor at 34 weeks. At both 12 and 24 weeks, Time in Range, time above range, and mean glucose were the best metrics to predict preterm birth. A1c was only able to predict preterm birth with statistical significance at 24 weeks. Time in Range, time above range, standard deviation, and A1c were able to predict large-for-gestational-age as early as 12 weeks and saw even stronger associations when assessed at 24 and 34 weeks. NICU admission and neonatal hypoglycemia were best predicted at 24 weeks by mean glucose, Time in Range, time above range, glycated CD59, and A1c. At 34 weeks, A1c was a better predictor of neonatal hypoglycemia, preterm birth, and NICU admission than was Time in Range.

  • It was terrific to see the always-fascinating pregnancy CGM data – without a doubt, we will see even more data on the use of CGM in pregnant women in the future. Certainly, to date, CONCEPTT pregnancy data is hard to beat. As background for recent impact, Closer Look readers watching pregnancy will remember that Health Canada temporarily authorized the use of Dexcom G6 in pregnant women with type 1, type 2, or gestational diabetes during the COVID-19 pandemic in July 2020. The European market now has several CGMs indicated for use in pregnancy (or at least, are not contraindicated against use in pregnancy): Dexcom G6, Abbott FreeStyle Libre, and Cascade Waveform, Agamatrix’s CGM that is making waves. As a reminder, this was approved for use in late 2019. For now, there are no CGMs indicated for pregnancy use in the US, but off-label usage is quite common. This is a tough one and a great example where there is data from the US but no FDA approval because the EU approval was easier to get – but, then, the people with diabetes in the US can’t get reimbursement – sigh.

“See the Woods and the Trees”: Drs. Fergus Cameron and Stuart Weinzimer Debate the Value of A1c and Time in Range Zeroing In On Accuracy and Risk Assessment

Dr. Fergus Cameron (Murdoch Children’s Research Institute) and Dr. Stuart Weinzimer squared off in a lively afternoon debate on the value of A1c with the consensus coming that both A1c and Time in Range provide clinical utility in our current diabetes management ecosystem. Dr. Cameron held the pro side of the debate arguing in favor of A1c while Dr. Weinzimer argued against, positing CGM-derived metrics including Time in Range, as preferrable replacements for A1c. While both Dr. Cameron and Dr. Weinzimer presented cohesive and convincing arguments for their respective sides, Dr. Cameron also expressed that he felt the debate inspired a “false dichotomy” between A1c and CGM-derived metrics when the two can provide patients and clinicians with insight into different aspects of diabetes management. Specifically, Dr. Cameron provided an apt metaphor describing the relationships between A1c and Time in Range like being able to “see the woods AND the trees” where A1c represents the “woods” and gives a larger picture of a patient’s glycemic control whereas Time in Range represents the “trees” and can provide a closer look at the various components that influence glycemic control and trends. We certainly agree with this sentiment and see the value of both A1c and Time in Range, and are interested to see in longer term trials the impact of Time in Range on outcomes – many would like to see this become clinical endpoint similar to how A1c is used today to better align results at a population level to clinical care. We hope to see these studies funded and executed, particularly given the impact we believe this trial will have.

  • Dr. Cameron’s argument centered around the value of A1c as an accurate surrogate metric for long-term diabetes-related complications arguing that the main goal of diabetes management is to limit these complications in people living with diabetes. To support this argument, Dr. Cameron pulled on data from follow-up to the DCCT/EDIC study which identified A1c as “the strongest risk factor for CVD death” among the study population also drawing on a 2008 analysis of DCCT data which indicated with A1c could explain “96%” of the difference in risk factors between the intensively and conventionally managed populations in the trial. Dr. Cameron also addressed the often-raised concern that A1c is an average and thus does not represent glycemic excursions that may be detrimental to patient health. On this point, Dr. Cameron highlighted data from the last two decades all of which implied that despite research on glycemic variation, there does not appear to be a correlation between glycemic variability and long-term diabetes-related complications thereby suggesting that glycemic excursions, more easily measurable via CGM, are not necessarily relevant to consider when working to prevent long-term complications. In perhaps the most salient portion of Dr. Cameron’s argument, he directly compared A1c and CGM-derived metrics, highlighting data published by Beck et al. in 2019 where 70% Time in Range was shown to correlate with an A1c of 7%. However, as Dr. Cameron deftly noted, the interquartile range for this correlation of 70% Time in Range varied from an A1c value of 5.6% to one of 8.3%, raising questions about how tightly Time in Range and other CGM-derived metrics, such as GMI, can be correlated with A1c. Finally, while Dr. Cameron did recognize the value of CGM, he highlighted that many patients still do not have access to CGM devices making A1c a necessary metric in patients for whom Time in Range is not available.

    • During Q&A, we caught a comment from debate moderator Dr. Irl Hirsch (University of Washington) in response to Dr. Cameron’s assertion that GMI and A1c discordance implies that GMI may not be an accurate assessment. Specifically, Dr. Cameron cited results from a recently published study from Perlman et al. for which Dr. Hirsch was a co-author, that found “substantial discordance” between laboratory A1c and CGM-estimated A1c (i.e., GMI). While Dr. Cameron interpreted this data as evidence that CGM-derived metrics may not accurately represent a patient’s true risk for diabetes-related complications, Dr. Hirsch shared an opposite interpretation expressing he had understood the data to mean that A1c may not be as accurate an assessment of risk as it has long been thought to be and that GMI may provide a more accurate surrogate for mean glucose, which both Dr. Hirsch and Dr. Cameron identified as the true metric indicative of long-term outcomes. We had made the same assumption as Dr. Hirsch, particularly given other limitations of A1c shown over time, despite its ongoing high value in multiple respects and particularly given what we’ve learned about limitations of A1c from him and multiple other investigators. While Dr. Cameron was unfortunately unable to attend Q&A due to time zone differences between Europe and Australia, we are seeking more information regarding discrepancies between GMI and A1c.

  • Dr. Weinzimer argued ardently in favor of Time in Range and other CGM-derived metrics citing his experience that while providers may consider A1c when creating treatment regiments, care is more easily optimized when patterns around hypo- and hyperglycemia are incorporated into decision-making processes. Specifically, Dr. Weinzimer highlighted four patients all which A1c’s of 8% but with dramatically different Time Above Range and Time Below Range values ranging from 33%-57% and 0.5%-10% respectively. In light of these differences, Dr. Weinzimer argued that most providers would probably prescribe different treatment regimens for the four patients despite their identical A1c values highlighting the clinical utility of Time in Range and CGM data. Dr. Weinzimer also used this data as an example of the glycemic variability CGM has identified within A1c values as these four patients also represented a range of mean glucose values from 156 mg/dl to 195 mg/dl. Dr. Weinzimer also dedicated a significant portion of his argument to the growing body of evidence validating the relationship between Time in Range and micro and macro vascular complications. Specifically, Dr. Weinzimer highlighted four studies from the last 3 years in which increased Time in Range was associated with decreased risk of retinopathy, microalbuminuria, carotid intimal media thickness, and cardiac autonomic neuropathy all of which provide strong support for the use of Time in Range as a clinical target in patients with diabetes. We also expect that we will continue to see additional studies adding to this body of evidence over the next few years.

  • Dr. Weinzimer convincingly concluded his argument by expressing that providers have already begun to use Time in Range and CGM metrics to successfully manage their patients over the last year during the COVID-19 pandemic when many patients have been unable to go to labs and get A1c tests. Despite concerns that the pandemic would disrupt diabetes care, which it certainly did for many, recently published data from 65,067 Dexcom G6 users across eight communities in the US indicated that across all regions and users, patients saw improvements in both Time in Range and mean glucose indicating the efficacy of remote monitoring enabled by CGM technology. While Dr. Weinzimer did acknowledge that not all patients have access to CGM technology currently, he argued that expanding access and provider education are crucial next steps in advancing Time in Range as a clinical glycemic metric.

CITY, SENCE, and WISDM Post-Hoc Analysis Shows Correlations Between A1c and GMI and A1c and Time in Range Are Stronger When %CV Is Higher

Dr. Blake Cooper (Retina Associates) read out results from an interesting post-hoc analysis of three major CGM studies, examining the relationships between A1c and GMI and Time in Range. The analysis utilized data from the SENCE (CGM in young kids, ages 2-6), CITY (CGM in adolescents/young adults, ages 14-24), and WISDM (CGM in older people, ages 60-86) studies. As a reminder, those three studies enrolled a total of 487 people with type 1 diabetes with ~half randomized to use real-time CGM.

 

SENCE

CITY

WISDM

Hours of CGM data

316

311

352

Mean glucose

203 mg/dl

211 mg/dl

167 mg/dl

Time in Range

40%

37%

56%

A1c

8.3%

9.1%

7.5%

GMI

8.2%

8.4%

7.3%

%CV

44%

42%

42%

  • The correlation between A1c and mean glucose (and therefore, GMI) was stronger for those with higher glycemic variability. In the SENCE study, for those with %CV ≤36%, Pearson’s correlation coefficient was 0.22, compared to 0.41 for those with %CV >36%. In the WISDM study, for those with %CV ≤36%, Pearson’s correlation coefficient was 0.58, compared to 0.7 for those with %CV >36%. Interestingly, in the CITY trial, the correlation strength between A1c and mean glucose/GMI was similar regardless of %CV at 0.56 and 0.53 for those below and above 36%, respectively.

  • The correlation between A1c and Time in Range was also stronger for those with higher glycemic variability. In the SENCE study, for those with %CV ≤36%, Pearson’s correlation coefficient was -0.29, compared to -0.39 for those with %CV >36%. As a reminder, the correlation coefficients are negative since a higher Time in Range corresponds to a lower A1c. In the WISDM study, for those with %CV ≤36%, Pearson’s correlation coefficient was -0.43, compared to -0.48 for those with %CV >36%. Interestingly, in the CITY trial, the correlation strength between A1c and mean glucose/GMI was similar regardless of %CV at -0.57 and -0.67 for those below and above 36%, respectively.

Dr. Elena Toschi Calls for Use of CGM-Derived Glycemic Metrics in Clinical Care for Older Patients with Diabetes; “~46%” Older Patients with T1D Experience Differences Between A1c and GMI of ≥0.5%

Dr. Elena Toschi (Joslin Diabetes Center) gave an informative presentation on the use of CGM-derived metrics in older patients with diabetes on MDI therapy highlighting the benefits of CGM and additional data among older populations. Specifically, Dr. Toschi discussed the sometimes rapid changes and deteriorations in cognitive and physical health that can occur in older patients that may not necessarily be reflected on long-term metrics such as A1c. Additionally, Dr. Toschi cited data from 2015 that indicated a larger discordance between mean glucose as measured by CGM and A1c among older adults suggesting A1c may not be the most accurate means of assessing glycemic control in an older population. Turning to more recent data, Dr. Toschi discussed the post-hoc analysis of the DIAMOND trial which indicated older patients on MDI who were CGM naïve saw significant improvements in glycemic control over the six months of the study indicating that prior technology use may not be a prerequisite for successful CGM use in this population. Interestingly, in a recent study conducted by Dr. Toschi’s group, data indicated that while glycemic variability was not related to A1c, higher glycemic variability was associated with increased hypoglycemia among patients with either type 1 or type 2 which is extremely concerning, especially if going undetected in patients not consistently using CGM. Additionally, Dr. Toschi shared data indicating that approximately 46% of older adults with type 1 experience discordance between their A1c and CGM-derived GMI of ≥0.5% which can have clinical implications and raises questions about the value and accuracy of both metrics in treating older patients with diabetes. Specifically, according to Dr. Toschi, in patients where GMI < A1c, clinicians relying only on A1c may choose to increase a patient’s insulin dose and thus inadvertently “potentially increase time spent in hypoglycemia.” Conversely, for patients with GMI > A1c, providers relying only on A1c may “not adjust therapy when it may be necessary” leading Dr. Toschi to call on providers to incorporate CGM-derived glycemic metrics into their clinical practice especially among older patient populations where Dr. Toschi argued the data “can guide clinicians to set individualized glycemic goals and to develop a personalized diabetes management plan.” We certainly agree that more data is better, especially when it comes to diabetes management.

Posters

Closing the Loop and Insulin Delivery

Title

Details + Takeaways

Real-world performance of the MiniMed 780G system: Impact of initiating automated basal and correction boluses

  • Real-world analysis of data uploaded to CareLink by 812 MiniMed 780G users in Belgium, Finland, Italy, the Netherlands, Qatar, South Africa, Sweden, Switzerland, and the UK
  • Time in Range improved from 63% to 76% after starting on MiniMed 780G; GMI declined from 7.2% to 6.8%
  • The percentage of users achieving a >70% Time in Range goal increased from 35% to 75% and the percentage of users achieving a <7% GMI goal increased from 38% to 75%

Patient-Reported Outcomes Reveal Potential Impact of Advanced Hybrid Closed Loop Systems on Hypoglycemia Attitues and Behaviors

  • Survey performed on US research panel of adults with type 1 diabetes between March 2019 and June 2020 (n=1,149)
  • Survey respondents skewed female (69%) and higher income (40% >$100,000)
  • Mean total Hypoglycemic Attitudes and Behaviors Scale (HABS) was significantly lower for Control-IQ users (2.14) than other AID system users (2.4) and non-AID users (2.36)

Cost-effectiveness Analysis of the MiniMed 780G System Versus Multiple Daily Injections with Intermittently Scanned Continuous Glucose Monitoring in Individuals with Type 1 Diabetes in Sweden

  • IQVIA Core Diabetes Model was used to assess cost-effectiveness of MiniMed 780G vs. MDI+FreeStyle Libre
  • Simulated patients were assumed to have a baseline A1c of 7.8% and MiniMed 780G was assumed to deliver a 0.5% A1c advantage over MDI+FreeStyle Libre
  • MiniMed 780G was associated with a SEK 373,700 (~$45,000 USD) per quality-adjusted life year gained, driven by reductions in rates of complications

Initial impact of the transition to Tandem Control-IQ on sleep in youth with type 1 diabetes (T1D) and their parents

  • 26 children and 26 parents wore actigraphy watches and kept sleep diaries for fourteen nights to assess sleep quality; data was collected for one week before starting Control-IQ and one week after starting Control-IQ
  • Only two of the sleep measures saw a statistically significant change after initiating Control-IQ
  • Children saw a mean reduction of 0.35 hours in their sleep following Control-IQ initiation (baseline: 7.9 hours)
  • Parents saw a mean reduction of 2 night awakenings in the week following Control-IQ initiation (baseline not provided)

Comparison between a real-time continuous glucose monitoring system and an intermittently-scanned continuous glucose monitoring system as blood glucose source in a do-it-yourself artificial pancreas system

  • A before-after study of 12 participants with type 1 diabetes using AndroidAPS
  • Participants used AndroidAPS with real-time CGM (presumably, Dexcom) for three months, then switched to FreeStyle Libre with a third-party reader/transmitter (e.g., MiaoMiao)
  • Both glycemic control and patient-reported outcomes were similar when using real-time CGM or FreeStyle Libre to drive the AndroidAPS system

Human factors for a closed-loop insulin delivery system: a review of remote methods for assessing the usability of the Tidepool Loop mobile application

  • Tidepool successfully performed formative and validation studies in a remote setting to assess the usability of Tidepool Loop in a virtual setting
  • These studies utilized digital prototyping, teleconferencing, activity tracking (e.g., record tap gestures, etc.), and other software

Prospective analysis of satisfaction and usability of closed-loop Diabeloop DBL-hu treatment in patients with highly unstable type 1 diabetes

  • Seven adults with highly unstable (“brittle”) diabetes and severe hypoglycemia went through two-weeks of run-in with a predictive low glucose suspend system AID system and 16 weeks on Diabeloop’s DBL-hu AID system
  • Use of DBL-hu was associated with a significant increase in treatment satisfaction (48 vs. 39) and perceived frequency of hypoglycemia (1.7 vs. 3.7)
  • All participants wished to continue on the DBL-hu system at the end of the study

SARS-COV2 impact on diabetes type 1 patients using “hybrid closed-loop” artificial pancreas

  • The study analyzed glycemic metrics for one adult type 1 using a custom hybrid closed loop algorithm (“trained with machine learning technology”) before and after contracting SARS-CoV-2
  • Time in Range during the month prior to infection was 82%, compared to 92% during COVID-19 symptoms
  • Time in hypoglycemia (<70 mg/dl)  during the month prior to infection was 8.8%, compared to 7% during COVID-19 symptoms

Reduction of hypoglycaemia using hybrid closed-loop with faster-insulin aspart compared with standard-insulin aspart in adults with type 1 diabetes: a double-blind, multinational, randomized, crossover study

  • 25 adults with type 1 diabetes participated in a multi-center, randomized cross-over study comparing the CamAPS FX AID system with Fiasp and with standard insulin aspart
  • Participants were randomized to start on Fiasp or standard insulin aspart and crossed over at eight weeks
  • Glycemic metrics were nearly identical for both Fiasp and standard insulin aspart, except for hypoglycemia metrics
  • Hypoglycemia slightly favored use of Fiasp compared to standard insulin aspart: for example, time <70 mg/dl was 2.4% during the Fiasp period, compared to 2.9% during the standard insulin aspart period

A novel algorithm to detect priming doses from smart insulin pens

  • Smart insulin pens may have difficulty detecting priming doses (i.e., doses given by patients into the air before actually delivering insulin)
  • The mySugr team performed interviews with experts and people with diabetes to understand the priming strategies and edge cases in order to develop an algorithm for priming dose detection
  • The new algorithm performed significantly better than simple threshold or max dose-based algorithms (98% vs. 76% vs. 60%, respectively) on a simulated dataset

Real-world performance of the MiniMed 670G system across Europe and in South Africa

  • Real-world data uploaded to CareLink from 15,151 MiniMed 670G users from Austria, Belgium, Switzerland, Germany, Denmark, Spain, Finland, the UK, Italy, Luxembourg, the Netherlands, Sweden, Slovenia, and South Africa
  • Mean Time in Range was 72%, ranging from 69% in the lowest country (South Africa) to 75% in the highest country (Spain)
  • Mean Time in Range prior to initiating Auto Mode was 62%
  • Mean glucose was 153 mg/dl, corresponding to a GMI of 7%
  • Users spent an average of 81% of the time in Auto Mode

Telehealth during COVID-19: Patient satisfaction of virtual training on the MiniMed 670G system in people with type 1 diabetes

  • Between February and April 2020, 31 children and adults with type 1 diabetes underwent 100% virtual training for MiniMed 670G
  • At baseline, 71% of participants were using SAP or predictive low glucose suspend and Time in Range was 66%
  • NPS for the virtual training was +87 with 87% of participants “very satisfied” and the remainder “satisfied”
  • Time in Range increased to 76% at follow-up (90 days later)

Algorithm-driven personalized insulin delivery: insights from real-world use of the MiniMed 670G system in Europe

  • Real-world analysis of data uploaded to CareLink by 3,307 people in Europe using MiniMed 670G
  • All users had at least 10 days of CGM data from before and after Auto Mode initiation
  • Users were segmented into GMI <7% before initiating Auto Mode (n=2,673) and GMI >8% before Auto Mode (n=634)
  • Total daily dose increased by 6.2 units for the well-controlled group, compared to +9.9 units for the poorer-controlled group following Auto Mode initiation
  • Time in Range increased from 75% to 78% for the well-controlled group, compared to 36% to 59% for the poorer-controlled group

Case series of four women with type 1 diabetes mellitus using do-it-yourself artificial pancreas systems during pregnancy

  • Case studies of four pregnant women with type 1 diabetes using DIY AID systems (two using G6/Omnipod/Loop, one using G6/MiniMed Veo 722/Loop, and one using G6/Accu-Chek Combo/AndroidAPS)
  • For all three women, strong glycemic control was maintained through all three trimesters of pregnancy
  • Case 1: pre-pregnancy A1c of 5.7%, first trimester A1c of 5.3%, second trimester A1c of 5.2%, third trimester of 5.4%
  • Case 2: pre-pregnancy A1c of 5.9%, first trimester A1c of 4.8%, second trimester A1c of 4.8%, third trimester of 5.5%
  • Case 3: pre-pregnancy A1c of 6.2%, first trimester A1c of 5.8%, second trimester A1c of 5.3%, third trimester of 5.1%
  • Case 4: pre-pregnancy A1c of 6.5%, first trimester A1c of 5.8%, second trimester A1c of 5.3%, third trimester of 5%
  • All four women reported Time in Range >70% (70-140 mg/dl) across all trimesters, except case study 3’s first trimester

Zen Mode of Diabeloop’s DBLG-1 system: stabilization of impact after 90 minutes

  • Data from Diabeloop’s clinical trial  for its DBLG1 AID system in type 1s
  • The clinical trial recorded 367 instances of participants using “Zen Mode” (a setting that reduces risk of hypoglycemia temporarily, e.g., while driving or during an interview) for longer than one hour
  • On average, mean glucose was affected by use of “Zen Mode” after 60 minutes, increasing from 133 mg/dl to 135 mg/dl
  • The effect of “Zen Mode” stabilized around 120 and 180 minutes, plateauing at a higher mean glucose

Hybrid closed loop and difficult foods in children with type 1 diabetes: a pilot study

  • Cross-over study of children with type 1 diabetes using MiniMed 670G and eating “difficult meals” (i.e., meals high in fat and protein)
  • Participants ate margherita pizza and beef lasagna on two nights a week for four weeks (four meals using Auto Mode and four meals on manual mode)
  • No statistically significant differences were found in post-prandial glucose on Auto Mode or manual mode
  • Qualitatively, patients and their families felt more confident with the MiniMed 670G in Auto Mode

A feasibility study assessing an “I-am-eating” meal bolus option for a closed-loop system which eliminates carbohydrate counting

  • 15 adults with type 1 diabetes used a modified version of MiniMed 780G with a set point of 100 mg/dl and active insulin time of two hours
  • For three months, participants performed traditional carb counting; then, for three weeks, participants used a modified MiniMed 780G system with an “I-am-Eating” bolus function (i.e., meal announcements, but no carb counting)
  • Mean Time in Range four hours post-meal was 69% with carb counting, compared to 66% with simple meal announcements; mean glucose during the same period was 156 mg/dl compared to 163 mg/dl, favoring carb counting
  • Mean Time in Range during the entire day was 78% with carb counting compared to 76% with simple meal announcements
  • For meals with ≤60 grams of carbs, posts-prandial Time in Range were comparable for carb counting and meal announcements (68% vs. 69% and 199 mg/dl vs. 201 mg/dl, respectively)

Ultra-rapid lispro (URLi) demonstrates similar time in target range to lispro with the Medtronic MiniMed 670G hybrid closed loop system

  • Randomized crossover trial of 42 adults with type 1 diabetes using MiniMed 670G
  • Participants were randomized to use MiniMed 670G with Ultra Rapid Lispro (URLi) or standard insulin lispro for four weeks before crossing over
  • Mean Time in Range was 77% using URLi compared to 78% with standard lispro
  • The MiniMed 670G may not realize benefits from using faster-acting insulins, since it only modulates basal insulin delivery

Fiasp insulin stability in the extended wear infusion set and on-market infusion sets: an in-vitro samples comparison with the MiniMed 670G system

  • 24 samples of Fiasp insulin were shaken and pumped through a standard 3-day infusion set and Medtronic’s 7-day Extended Wear Infusion Set
  • Samples were shaken for 6 days for the control set and 14 days for the extended-wear set
  • The pumped Fiasp from both the control and 7-day infusion sets met proper specifications
  • There was no incidence of pump malfunction or infusion set inclusion during the study

The Accu-Chek Solo tubeless micropump improves glycemic control and quality of life in adult and pediatric patients with type 1 diabetes: a pilot study

  • Pilot study of nine type 1s in the Spain using Roche’s Accu-Chek Solo patch pump, the first patch pump available in Spain
  • All patients showed a high degree of satisfaction after 1-12 months of follow-up
  • Aspects that could be improved were smartphone control, smaller size, a larger insulin reservoir, and integration with CGM

Characteristics of people with type 2 diabetes mellitus initiating an insulin pump: data from a large commercial claims database

  • A retrospective study of 2,026 people in IBM’s MarketScan Commercial Claims Database with type 2 diabetes using insulin pumps (data from 2016-2018)
  • One-third of patients had a claim for CGM within six months of starting on a pump
  • About one-quarter of people filled their first insulin prescription within 18 months of starting on a pump
  • About half of patients visited an endocrinologist within three months of pump initiation
  • 11% of patients reported at least one hypoglycemic event and mean A1c was 9.5%, suggesting suboptimal glycemic control

Evaluation of a novel CGM-informed bolus calculator with automatic glucose trend adjustment used with sensor-augmented pump therapy

  • 25 participants with type 1 diabetes using Omnipod 5 in manual mode
  • Participants used the standard bolus calculator for one week (user manually enters glucose value), followed by the CGM-informed bolus calculator for one week (calculator that accounts for both CGM value and trend)
  • Percentage of time <70 mg/dl during the four-hour post bolus period was significantly lower with the CGM-informed bolus calculator (2.8% vs. 2.1%)
  • Other glycemic metrics during the four-hour post bolus period were not significantly different

The Medtronic Extended Wear Infusion Set: determining mechanisms of action

  • Summary of three studies, two in humans and one in a diabetic porcine model
  • In one human study and the porcine study, the three-day infusion set was worn for seven-days with survival rates of 33%
  • In the other human study, Medtronic’s Extended Wear Infusion Set was worn for 7-days with a survival rate of 81%

Clinical characteristics of hybrid closed loop system in children and adolescents with type 1 diabetes previously treated with multiple daily injections: one-year experience

  • Real-world data from 30 children with type 1 diabetes after one-year on Minimed 670G
  • A1c declined from 8.2% to 7.1% after one-year on MiniMed 670G; Time in Range increased from 47% to 73%
  • At baseline, about half of the participants had no experience using CGM

Hybrid closed loop system and glucose control in patients with type 1 diabetes: results from a single center study

  • 27 adult type 1s in Italy used MiniMed 780G recorded data using MiniMed 670G for at least 25 days and SAP (MiniMed 640G or Veo) for at least 20 days prior
  • Mean sensor glucose fell from 155 mg/dl in manual mode to 146 mg/dl in closed loop; GMI fell from 7% to 6.8%
  • Time in Range increased from 67% in manual mode to 78% in closed loop

Rapid improvement in Time in Range after Advanced Hybrid Closed Loop initiation

  • 52 adults and adolescents with type 1 diabetes using a predictive low glucose suspend system were initiated on MiniMed 780G
  • Subjects used the system with a set point of 100 mg/dl and active insulin time of two hours
  • Mean glucose fell from 155 mg/dl with PLGS to 133 mg/dl with MiniMed 780G; GMI fell from 7% to 6.5%
  • Time in Range increased from 67% to 82%; the percentage of subjects meeting the TIR>70% target grew from 46% to 89%
  • Bolus insulin increased from 49% of the total daily dose to 56%; 30% of bolus insulin delivered was from auto-correction boluses

Physical activity impact on a type 2 diabetes population, implications for a fully closed-loop system

  • Data from an observational study of 35 adults with type 2 diabetes using Diabeloop’s DBLG-1 AID system
  • 27 participants recorded at least one physical activity period with metabolic equivalent of task >1.5
  • Time in hypoglycemia and %CV were lower for those who participated in physical activity; other glycemic metrics were not statistically significantly different

Lessons learnt with MiniMed 780G Advanced Hybrid Closed-Loop

  • Ten children with type 1 diabetes were initiated on MiniMed 780G; one switched from MiniMed 670G, two from MiniMed 640G, and seven from MDI
  • All participants used MiniMed 780G with a set point of 120 mg/dl and active insulin time of 3 hours
  • Time in Range increased from 68% to 78% with MiniMed 780G

New ultra-rapid acting insulin in sensor augmented pump with predictive low glucose suspension algorithm

  • Retrospective data analysis of 13 people with type 1 diabetes using MiniMed 640G (predictive low glucose suspend) system
  • 7 participants used Fiasp, one used insulin lispro, and five were using glulisine
  • 81% of low glucose suspends prevented glucose from dropping <70 mg/dl using Fiasp, compared to 70% of suspensions with the traditional insulin analogs

A real-world evaluation of automated insulin dosing systems demonstrates superior efficacy and comparable safety with open-source systems as compared to MiniMed 670G

  • Retrospective analysis of AID users (38 on MiniMed 670G, 30 on DIY) examining differences between MiniMed 670G and DIY AID system users
  • On average, DIY users were younger (47 vs. 41), had shorter duration of diabetes (31 vs. 25 years), and lower baseline A1c (7.8% vs. 7.1%)
  • Time in Range was higher in the DIY users at 78%, compared to 68% for MiniMed 670G users

Effect of meal composition on glucose rate of change in patients with type 1 diabetes using a closed-loop system

  • 110 participants with type 1 diabetes used an AID system for five days and recorded data from a total of 1,166 meals (309 meals were analyzed)
  • Meals were classified as high or low fat (defined as more or less than 30% of calories) and high (>45% of calories) or low carb (<26% of calories)
  • Mean glucose rate of change was significantly higher with high fat meals compared to low fat meals; rate of change difference was smaller between high and low carb meals

Prospective analysis of satisfaction and usability of closed-loop Diabeloop DBL-hu treatment in patients with highly unstable type 1 diabetes

  • Prospective randomized study of patient satisfaction with the DBL-hu AID system vs. a predictive low glucose suspend system (PLGS) in 7 adults with highly unstable diabetes (HUD) with severe hypoglycemia
  • Using DBL-hu was associated with a significant increase in treatment satisfaction vs. PLGS (p=0.03)
  • The perceived frequency of hypoglycemia was significantly lower among DBL-hu users (p=0.02)
  • DBL-hu users found the system highly easy to use and autonomous
  • At the end of the study, which is currently in an extension phase, all patients wished to continue on the DBL-hu system

Online, competency-based training to improve competency, confidence and troubleshooting ability when using the CamAPS FX hybrid closed-loop insulin delivery system

  • Assessment of the online Cambridge Diabetes Education Program (CDEP) for educating patients, HCPs, and caregivers on CamAPS FX AID system
  • 595 participants (57% HCPs, 36% patients, 7% teachers) since February 2020
  • All participants reported that their competency, confidence, and troubleshooting either didn’t change or improved
  •  Average score out of 5 (3=no change, 5=significantly improved): 4.4 for competency and confidence, 4.3 for troubleshooting

Glucose Sensing Posters

Title

Details + Takeaways

Evaluation of user performance and accuracy of a new blood glucose monitoring system

  • Two studies (laboratory and clinical) evaluating accuracy of new Contour Plus Blue BGM system with Contour Plus test strips with N= 600 and N=332 measurements in the laboratory and clinical studies respectively
  • 99.8% of laboratory test samples fell within +/- 10 mg/dl when compared to YSI reference results; 100% of measurements of blood glucose levels <100 mg/dl fell within +/- 10 mg/dl and 99.8% of all measurements of blood glucose >100 mg/dl fell within +/- 10 mg/dl
  • All laboratory measurements fell within Zone A by Parkes-Consensus Error Grid
  • 99% of clinical study measurements (326 of 329) met ISO accuracy criteria of +/- 15 mg/dl

Use of personal continuous glucose monitoring device is associate with reduced risk of hypoglycemia in a 16-week clinical trial of people with type 1 diabetes

  • Comparison of risk of hypoglycemia among 117 patients with T1D using CGM and 355 patients with T1D not on CGM
  • Patients not on CGM experienced a significantly higher event rate of both symptomatic and asymptomatic hypoglycemia
  • CGM use also associated with reduced time spent in hypoglycemia and decreases in glycemic variability without statistically significant increases in A1c

Glycemic control in young Italian patients with type 1 diabetes during COVID-19 lockdown: the role of age, type of insulin therapy, telemedicine and physical activity

  • Retrospective analysis of 202 patients with T1D using Dexcom G6 during COVID-19 lockdown; average age of 18
  • A1c improved during COVID-19 lockdown to 7.6% from 7.8% at baseline prior to lockdown (p<0.0001)
  • Time in Range improved significantly from 55% at baseline to 59% during the lockdown (P<0.0001) with the most significant improvement among “university students and young adults” who saw Time in Range increases of 6% to 61% from 55% at baseline

Impact of continuous glucose monitoring on glyceamic control in children and adolescents with T1D: real world data from a population-based clinic

  • Observational longitudinal study of 348 children <18 years old with a diabetes duration of at least 2 years on either Dexcom or Medtronic CGM; baseline A1c of 8.5%
  • Participants experienced immediate and sustained A1c reductions following CGM adoption with children ≥14 years old showing the greatest reduction in immediate A1c change
  • Following CGM initiation, children were more likely to achieve the ISPAD target of A1c <7%

Personalized forecasting of severe hypoglycemia using real-time continuous glucose monitoring (CGM) data in patients with Type 2 Diabetes (T2DM)

  • Data from 17 patients with type 2 collected via FreeStyle Libre CGMs was used to train a forecasting model to predict near-term severe hypoglycemia
  • Predictive modeling provided “good” forecasting up to two hours in advance with an accuracy rate of 89% at 15 minutes, 83% at 1 hour, and 74% at 2 hours for all patients
  • Predictive modeling improved in patients with >10 severe hypoglycemia events compared to patients with fewer than 10 severe hypoglycemia events

Interstitial glucose monitoring and skin reactions: real-life study in adults and children

  • Survey of 36 patients with T1D to assess skin irritation caused by CGM; baseline A1c of 7.3%; 22% used FreeStyle Libre, 25% used Dexcom G6, 47% used Guardian Sensor 3, 6% used Eversense
  • 72% of respondents (n=26) experienced “some reaction” with 36% (n=13) reporting “frequent reactions”
  • 17% of skin reactions required medical treatment while the rest were “mild” and were limited to the adhesive dressing area and lasted a few days

Clinical impact in real life of flash glucose monitoring in adolescents with type 1 diabetes

  • Longitudinal observational study of 129 adolescents with T1D evaluating clinical efficacy of flash glucose monitoring after one year of use
  • Time Below Range was significantly lower after 12 months of CGM use at 6.21% compared to 10.4% after 2 weeks of CGM use
  • Patients also experienced reduced glycemic variability after one year of CGM use, but saw no significant changes in Time in Range, Time Above Range, or A1c
  • The authors note that flash glucose monitoring users in this age group would likely benefit from targeted educational programs to improve adherence

Accuracy assessment of the new Glucomen Day CGM system in individuals with Type 1 diabetes

  • N=8 adults with T1D; participants wore 2 GlucoMen Day CGM systems on their abdomen for 14 days at home and conducted at least 5 fingersticks per day plus two 5-hour lab reference tests on days 4 and 10
  • Overall MARD was 9.7 with 98% of all CGM and venous glucose pairs (N=450) in Zones A and B

Association between monitoring frequency with flash technology and glycemic measures in patients with Type 1 diabetes mellitus

  • N=104 adults with T1D using Abbott FreeStyle Libre CGM; assessed relationship between frequency of scans and glycemic measures
  • Results indicated a “significant correlation” between number of scans and A1c, GMI, Time in Range, number of hypoglycemic episodes (p<0.001)
  • A1c was lower in frequent scanners at 7.2% compared to “non frequent” scanners with an average A1c of 8.3% (P<0.001)
  • Frequent scanners had more hypoglycemic episodes than did non-frequent scanners, but Time Below Range and time in hypoglycemia were similar between the two groups

Effect of flash glucose technology on glycemic control in patients with type 1 diabetes

  • N=47 adults with T1D using Abbott FreeStyle Libre CGM; mean age of 39 with mean diabetes duration of 18 years; baseline A1c of 7.4%
  • Data were collected from patients’ first visits after starting CGM and compared to data from a visit at least 6 months later while CGM was in use
  • There was no significant difference in A1c after 6 months of CGM use
  • Patients saw improved Time in Range from 44% at baseline to 56% after 6 months (p=0.01)
  • Patients also saw reductions in Time Above Range from 43% to 34% (p=0.03) and a non-statistically significant reduction in Time below Range from 12% to 10%.

In-hospital real time continuous glucose monitoring during COVID-19 outbreak: experience from a COVID hub

  • Retrospective observational study of Dexcom G6 use among hospitalized patients with COVID-19 and T1D (N=19); mean age of 66; average A1c of 9.1%; patients had a mean hospital stay of 12 days leading to 216 days of CGM data collected
  • Patients had an average Time in Range of 53% with a Time Above Range of 45%
  • The authors concluded that in-hospital CGM use was safe and demonstrated “potential efficacy” in hospitalized complicated patients with diabetes

Evolution of the quality of life of diabetic patients under intensive insulin therapy benefiting from FreeStyle Libre

  • Prospective observational study of 55 patients initiating flash glucose monitoring; 69% participants had T2D and 31% had T1D; average age of 61 years old
  • Participants saw a significant improvement in quality-of-life scores (p=0.34)
  • Participants checked their glucose values significantly more after initiating flash glucose monitoring at 8 times per day up from 4.6 times per day at baseline
  • Patients also saw a significant reduction in A1c after initiating flash glucose monitoring to 7.7% down from 8.3% (p=0.003)
  • The authors found that achieving better glycemic control was significantly associated with improved quality of life (0.008)

Flash glucose monitoring improves quality of life and reduces fear of hypoglycemia in type 1 diabetes patients: a real-world prospective study

  • Prospective real-world 6-month case-control study of 149 patients with T1D; case patients initiated flash glucose monitoring at the beginning of the study period and were compared to participants who used flash glucose monitoring previously
  • 63% of participants were on MDI and 37% used pump therapy; average A1c of 7.6%
  • Participants in both existing and new flash glucose monitoring groups saw significant reductions in A1c at the end of six months
  • Fear of hypoglycemia decreased among new flash glucose users (as measured by EsDQUOL and HFS surveys) compared to existing users (p<0.05)

Insights to the Time in Range (TIR) in patients on flash glucose monitoring (FMG) in patients with type 2 diabetes

  • N=320 patients with T2D managed with standard care who wore FreeStyle Libre pro CGMs
  • 50% participants achieved >70% Time in Range; mean Time in Range was 67%
  • Time in Range was significantly higher among patients ≥60 years old and in patients with a diabetes duration of >10 years
  • Mean Time Above Range was 23% and mean Time Below Range was 10%

The frequency of unlimited flash glycemia scans by patients with type 1 diabetes on pump insulin therapy and its correlation with diabetes mellitus clinical parameters

  • N=58 adults with T1D using insulin pump therapy and flash glucose monitoring for 14 days
  • Frequency of scans ranged from 4/day to 140/day with an average of 29 scans/day for the first 7 days and 30 scans/day for the last 7 days of the study period; there was no significant difference in the number of scans at the beginning, middle, and end of the 14-day study period
  • The authors found a significant correlation between the average number of scans and average glycemia (p<0.05)

First report of gender disparity in technology update [continuous glucose monitoring (CGM)] in patients of diabetes across India

  • N=507 patients in India with diabetes (403 with T2D and 167 with T1D); 55% male, 45% female
  • A higher proportion of male patients were on CGM therapy compared to women (p=0.0032) with an odds ratio of 1.74 implying that there was a 74% higher chance of male patients using CGM compared to female patients
  • Similar gender disparities are seen for insulin pump use in India suggesting larger socio-cultural-economic factors that may need to be addressed to improve outcomes and increase access to technology among women

Proof-of-concept of a novel non-invasive glucose monitor-validity of the algorithm over time without intermittent calibration

  • Study investigated CGM readings from non-invasive wrist-worn monitor; monitor uses microwaves and resonance to measure glucose
  • Glucose values were compared to those from an automated clamp device to measure accuracy and then used to develop an algorithm to predict future glycemic levels
  • Predictions were accurate 10 months after initial glucose readings without additional recalibration

Economic analysis of potential cost reductions through utilization of blood glucose meters with color range indicator for the healthcare system of the Russian Federation

  • Using A1c data from the ACCENTS randomized control the authors estimated the ten year risk of MI events and potential cost savings associated with BGM use
  • Use of the Flex BGM could results in a 10-year MI risk reduction of 2.1%; Use of the Verio BGM could result in a 10-year MI event risk reduction of 2%
  • MI risk reductions are expected to contribute to cost savings of between $780,000 - $819,000 per year in the Russian Health System

Unexpected improvement in glycemic control during forced inactivity due to COVID-19 pandemic in children with type 1 diabetes using CGM

  • N=188; longitudinal observational study evaluated changes in glycemic metrics before and during the COVID-19 pandemic among children with T1D in Italy; 25% participants were on insulin pump therapy
  • Pre-COVID period was defined as January 1-15th, 2020; post-COVID lockdown period defined as June – August, 2020)
  • Participants saw an increase of 3% in Time in Range during the COVID-19 pandemic to 64% from 61% at baseline
  • Participants also saw a reduction in time >250 mg/dl by 1% to 9%

Glucose control after initiation of flash glucose monitoring in type 2 diabetes managed with basal insulin; a retrospective real-world chart review study from Canada

  • Retrospective non-interventional chart review of 103 patients with T2D on basal insulin using FreeStyle Libre; average A1c of 8.9%; average age of 64; 69% males
  • FreeStyle Libre use of at least 3 months was associated with A1c reduction of 0.8% (p<0.0001)
  • Patients saw significant A1c reductions between 3-6 months of FreeStyle Libre use regardless of baseline A1c
  • Patients with A1c ≥9% saw more significant reductions of 1.6% (p<0.0001) compared to 0.5% (p<0.001) for patients with baseline A1c <9%

Improved accuracy of outcomes forecasts from short-term CGM data

  • One Drop poster outlining efficacy of 4-6 month predictive blood glucose, blood pressure, and weight algorithms based on CGM data from 1,694 people using CGM with 87% of patients with T1D
  • Algorithm training set included 6,162 blood glucose values, and testing data set included 8,989 blood glucose values
  • Predictions based on one month of CGM data had a root mean square error (RMSE) of 24 mg/dl; predictions based on two months of CGM data had an RMSE of 19 mg/dl; predictions based on three months of data had an RMSE of 17 mg/dl; all predictions were for 4-6 months after the initial data collection
  • After three months, using additional months of CGM data did not increase prediction accuracy

Impact of continuous glucose monitoring systems on parental quality of life

  • N=95 parents of children with T1D; 78% children used insulin pump therapy; 73% children used CGM
  • Parental wellbeing was assessed via Family Impact Module (FIM) of Pediatric Quality of Life (PedsQoL) survey
  • Parental worry was significantly lower among parents of children on CGM independent of whether or not the child used remote data monitoring with CGM data
  • Insulin pump use was not significantly associated with any decrease in FIM subscales related to family/parent wellbeing
  • Mean total FIM score indicated significant impact of diabetes on parental wellbeing

The relation between cardiometabolic risk factors and glucose profile evaluated by continuous glucose monitoring in type 2 diabetes persons

  • N=30 patients with type 2 diabetes; mean age of 57; baseline A1c of 8.3%
  • Mean amplitude of glucose excursions and glycemic variability were inversely related to weight (r=0.5 and r=0.4 respectively; p=0.004 and p=0.021 respectively)
  • Systolic blood pressure increased with diabetes duration and was directly correlated with total glucose status and inversely related to hypoglycemia
  • Participants with a family history of cardiovascular disease had a lower average A1c at 7.6% versus participants without a family history of CVD with an average A1c of 9.15%

The association between time in range % measured by continuous glucose monitoring (CGM) and physical & functional indices amongst older people with type 2 diabetes

  • Cross-sectional study among 81 patients with type 2 over the age of 60 to assess relationship between Time in Range and physical abilities
  • Participants wore blinded CGM (I Pro2 Carelink) for 1 week and underwent physical-functional assessments
  • A 1% Time in Range increase was associated with a 0.25 higher score on the 6-minute walk test (p=0.019) and a 0.21 lower score on the TUG test of fall risk and balance (p=0.045)
  • Higher Time in Range was associated with better scores on aerobic assessments and reduced risk of falls

Clinical acceptability of flash glucose monitoring (FGM) in pregnant women with diabetes mellitus type 1

  • N=18 pregnant women with type 1, mean age of 31 years
  • Compared flash glucose monitoring (FGM) values with SMBG
  • Average blood glucose values from FGM were lower than those from SMBG at 102 mg/dl and 128 mg/dl respectively
  • Median MARD across glycemic ranges was 15.5% with lower MARD in hypo- and hyperglycemia ranges than in the range of 70-180 mg/dl

Patient satisfaction and clinical efficacy of intermittently scanned continuous glucose monitoring: a 12-month real life study

  • Prospective, single center, single-arm study; N=36; patients evaluated at baseline, 3, 6, and 12 months; baseline A1c of 7.6%
  • All patients were provided with FreeStyle Libre CGMs and completed the validated “Quality of Life associated with Treatment of T1D Questionnaire”
  • Patients with baseline A1c <7.5% saw A1c reductions of 0.3% at 12 months to 6.6% from 6.9% at baseline; Time in Range improved significantly to 72% from 58% at baseline with a 5% reduction in Time Below Range to 3%
  • Patients with baseline A1c ≥7.5 saw a mean A1c reduction of 0.8% to 7.6% from 8.4% at baseline; Time in Range improved to 59% from 44% at baseline with a 3% reduction in Time Below Range to 1% and an 11% reduction in Time Above Range to 40%
  • All participants saw significant improvements across the quality-of-life metrics of: (i) general health (p<0.001); (ii) mental health (p=0.001); (iii) cognitive functioning (p=0.006); (iv) social impact (p<0.001); (v) current and future perspective (p<0.001); and (vi) impact of diabetes (p<0.001)

“Give me a break”: Adolescent and parent perspectives on breaks in CGM use

  • 60 adolescent-parent dyads completed a questionnaire on their CGM use and diabetes self-management as part of an ongoing, multi-site behavioral intervention trial
  • Of those using CGM (72%), breaks were common according to adolescent (40%) and parent (42%) reports; generally lasted 1-2 days with adolescents checking their blood glucose 2-5 times a day during the break
  • Adolescents whose parents reported CGM breaks reported lower diabetes self-care (p=0.033) and non-significantly higher A1c values (p=0.29)

Diabetes technology and alarms: Subjective distress is not associated with glucose parameters

  • DIA-LINK study used data from 203 type 1s who used an unblinded CGM for 17 days
  • Participants experienced elevated distress due to alarms on 22% of days
  • Daily distress due to alarms was associated with higher overall distress, lower mood, and higher stress, but not glucose parameters; glucose parameters were associated with overall diabetes distress
  • Time in Range was significantly associated with daily diabetes distress (association = -0.195, p<0.01)
  • Insulin pump users did not experience higher stress than non-pump users (p=0.124)

 

Digital Health & Telehealth

Title

Details + Takeaways

Survey of US Certified Diabetes Care and Education Specialists on the most important features of diabetes digital coaching programs

  • 403 CDCESs completed an online survey in September 2020, of whom 224 (56%) reported encouraging some (50%) or all (6%) patients to use coaching apps
  • Of the 244, 63% reported that coaching apps are somewhat or extremely helpful in improving patient empowerment, but less than half reported that apps are helpful in achieving goals for A1c (39%), Time in Range (47%), weight loss (41%), diabetes distress (44%), and quality of life (46%)
  • Among the 224, CGM data integration (53%), weight loss programming (44%), and report sharing (40%) were most often ranked as a top 3 coaching app feature; CGM integration was the top choice for the most important feature (26%), followed by insurance coverage (16%) and BGM integration (14%)
  • 72% of all participants (n=403) somewhat or strongly agree that CGM data integration is necessary for strong coaching outcomes

Medication optimization among people with type 2 diabetes participating in a CGM-driven virtual clinic

  • Onduo Virtual Diabetes Clinic (VDC) program for type 2s includes mobile app, coaching, connected devices, telemedicine consultations
  • Used rtCGM ensure safety and efficacy of medication optimization via telemedicine
  • Participants (n=55) had A1c values of 8%-12%; 89% were on ≥2 diabetes medications at baseline; participated in VDC for four months
  • A1c fell -1.6% at four months with no increase in hypoglycemia (p<0.0001)
  • 89% of participants had a medication change during the four months, but only 31% saw a net increase in # of medications
  • Med changes were primarily substituting “ineffective” meds for “effective” meds: GLP-1 prescriptions increased 27% to 53% (p<0.001), and sulfonylurea prescriptions fell 20% to 35%

Improved glycemic control and diabetes distress after using an m-health application: A preparation analysis for the Digital Health Care Act in Germany

  • Retrospective analysis of 5,920 German mySugr app users
  • Those with baseline A1cs <7.5% had a stable A1c of ~6.4% at three and six months
  • Those with baseline A1c values >9% saw a -1.5% and -1.7% A1c reduction at three and six months, respectively
  • Diabetes distress score was significantly lower in mySugr users compared to non-app users (p<0.001)

Efficacy of the My Dose Coach application in basal insulin titration: Effects on glucose control, adherence to treatment, and costs

  • 16-week 2-arm RCT comparing the efficacy of face-to-face conventional titration modifications (n=40) to the My Dose Coach application (n=40), a platform that allows providers to remotely titrate a patient’s basal insulin dose, in insulin-naïve type 2s
  • Significantly fewer insulin doses were missed in the My Dose group (4% vs. 9%), the direct costs were lower ($187 vs. $252, p<0.001), and there was less time spent on titrations (150 vs. 240 minutes)
  • My Dose Coach was noninferior to standard care; the My Dose Coach group saw a -1.8% A1c drop to 6.9% and the conventional treatment group saw a -1.3% drop to 7.2% (not significant difference)

Impact of the frequency of My Dose Coach use on clinical outcomes in type 2 diabetes

  • Retrospective cohort study of My Dose Coach (MDC) usage’s impact on clinical outcomes
  • 2,517 type 2s on basal-only from India, Mexico, and Columbia with high (>3 days/week), moderate (>1 but ≤3 days/week), and low (≤1 days/week) MDC usage
  • More frequent MDC use was significantly associated with greater fasting blood glucose target achievement (p<0.01)
  • No significant difference in hypoglycemia incidence between groups

Effects of the automatic calculation of prandial insulin bolus through a mobile phone application on A1c and quality of life in type 2 diabetes people

  • Prospective study of app-based bolus calculator in 10 type 1s using Accu-Check Instant BGM
  • A1c fell a nonsignificant -0.5% to 8.0% at three months
  • Significant improvements in satisfaction, comfort, and flexibility at three months

Impact of a digital intervention engine on diabetes self-management

  • 246 Dario active members who had not measured their blood glucose in the past week were randomized to the test group that experienced a digital intervention flow (n=127) or the control group (n=119)
  • Significantly more users in the test group measured their blood glucose (14% in first 30 days, 22% in days 30-60)

Real-world evidence of diabetes app use in daily life: Profiling adults with type 2 diabetes who use a blood glucose meter app

  • Four-month REALL study to assess the characteristics and initial motivations of adults with type 2 using CONTOUR DIABETS app paired with CONTOUR NEXT ONE BGM (n=585)
  • Most reported elevated diabetes emotional distress (58%) or diabetes regimen distress (66%); majority had overweight (22%), obesity (44%), or extreme obesity (27%)
  • 46% initially started using the app because it came with the meter, 29% actively sought out an app to improve their diabetes management, 11% and 4% initiated use because of HCP or other recommendation, 10% cited “other” reasons
  • Those with elevated emotional and/or regimen distress were more likely to have sought out the app themselves than those with no/little distress (32% vs. 22%)

Key burden in today’s diabetes management: User needs and requirements for diabetes apps and wearables

  • Survey of 144 German participants (ages 16-70; 96% CGM users, 65% use wearables, 76% type 1)
  • Participants ranked nutrition data logging as #1 (~37%), followed by analyzing: diabetes burden (#2, ~21%), physical activity (#3, ~12%), stress (#4, ~11%), and sleep (#5, ~10%)
  • Most (92%) think that apps would be useful in improving their diabetes management, yet only 64% currently use apps
  • 96% of participants view wearables as very useful in diabetes management

Evolution of AGP parameters before, during, and after lockdown in 80 French persons with type 1 diabetes sharing their FGM data on LibreView

  • Data from 80 French type 1s (age 41, 81% on pumps) during three 90-day periods before and during COVID-19: January 15-March 15 (T1), March 16-June 15 (T2)  June 16-September 15 (T3)
  • Time in Range increased from 54% (T1) to 58% (T2) and remained at 57% (T3) due to a decline in time above range from 40% at T1 to 36% and 37% at T2 and T3, respectively
  • Time below range did not change (6%), but time <54 mg/dl slightly decreased (2% at T1 to 1% at T2 and T3)
  • The number of daily scans (11), GMI (~7.3%), and variability (~40%) did not change

Telemedicine during COVID-19: The disconnect between patient and provider opinions of virtual visit quality

  • Mixed-methods study in June-July 2020; qualitative (n=27 PWD, 24 HCPs), quantitative (n=1,057 patients, 319 HCPs)
  • Overall, agree that telemedicine is useful, but imperfect
  • 30% of HCPs and 37% of patients report shorter wait time with telemedicine
  • 72% of providers and 53% of patients report “subpar” physical exams
  • HCPs more likely to find that telemedicine compromised the overall quality of care (40% vs. 21%) and to believe that telemedicine limited personal connection (53% vs. 27%)

Telemedicine in pediatric diabetes: Patient and clinicians’ views and responses to remote consultations during COVID-19

  • Mixed-methods study of clinicians’ and patients’ views of remote pediatric diabetes consultations
  • 56% of patients/families (n=100) and 75% of clinicians (n=89) would continue remote consultations
  • 25% of patients were concerned with the lack of physical exams
  • Clinicians reported better experience with telemedicine when patients engaged with their devices and had prepared their data; 15% of clinicians and 9% of patients reported lack of understanding and inadequate preparation as sources of negative remote consultation experiences
  • 69% of patients/families had previously experience uploading data; 75% received training

A United Kingdom (UK) wide survey of healthcare provider (HCP) experiences regarding virtual consultations in type 1 diabetes

  • Diabetes Technology Network UK online survey of its provider members (n=143); 50% of consultations via phone, 10% video, 20% face-to-face
  • 64% ranked remote consults as somewhat to very effective
  • Most common barriers are lack of access to device data prior to visit (67%) and patients’ lack of familiarity with tech (72%)
  • Access to patient results (52%) and device data (69%) prior to consultation are key facilitators to effective remote consultations

Transition to remote diabetes care in COVID-19 times: Experiences from a specialized type 1 diabetes clinic

  • Dutch type 1s ages 16+ (n=87) completed a survey after having a remote consultation (86% telephone) in May-July 2020
  • Patients were highly split on whether they would continue telemedicine (33% no, 33% maybe, 33% yes)
  • Higher quality of life, but not demographic or self-reported clinical factors, was associated with a more positive attitude toward future remote consultations
  • 24% of patients reported telemedicine saving them time; the vast majority of patients were satisfied with the audio/video connection (91%), attention to emotions (85%), and relationship with HCP (~78%)

The impact of COVID-19 pandemic on Australian diabetes health professionals and the challenges of rapid pivoting to telehealth services

  • 243 Australian survey respondents: 41% endocrinologists, 33% diabetes educators
  • Telehealth setup (63%) and reimbursement (38%) were the biggest challenges, along with fear of being infected/infecting family (49%)
  • 50% reported increased workload, majority of which was diabetes-related (81%)

Improved glycemic control and prevention of acute hospital admission by using a digital workflow and decision support system in home health care

  • Retrospective analysis of patients treated with GlucoTabMobileCare, insulin titration algorithm integrated into workflow system for geriatric home health care (n=9)
  • Fasting blood glucose improved to ~159 mmol/mol
  • No hypo- or hyperglycemia-related acute hospitalizations occurred while using GlucoTabMobileCare, down from pre- intervention and post-stopping the intervention

Changes in US clinic use of a web-based portal for review of patient continuous glucose monitoring (CGM) data before and during the COVID-19 pandemic

  • Among existing clinic users, Dexcom CLARITY use in the US increased drastically in January-June 2020 compared to previous year
  • Monthly logins increased 34% from January to June 2020, increasing between 49%-99% YOY per month; between April and July, YOY increase was ≥70% each month
  • YOY monthly count of new registered clinics did not change YOY from 2019

Trustsphere (a trusted, secure and privacy-respecting healthcare environment realized for everyone): User engagement to build the trust layer in digital health for type 1 diabetes

  • 18% of HCPs (n=232) and 30% of caregivers of children with type 1 (n=760) reported being “very” or “extremely” concerned about digital privacy and security issues
  • 68% and 42% of caregivers trust their HCP and government regulations to keep their child’s health information secure
  • HCPs reported major challenges in accessing insulin pump (85%), CGM (80%) and BGM (83%) data and helping patients navigate diabetes tech (91%)

Implementing digital diabetes care in a developing country: A qualitative study on physicians’ perspective

  • Cross-sectional, qualitative study on Indonesian physicians’ perspectives on implementing digital diabetes care in Indonesia
  • Opportunities: remote patient monitoring in areas with limited access to care, high smartphone utilization already
  • Barriers: lack of policy support, reimbursement limitations

Big Picture

Title

Details + Takeaways

The relationship of the hypoglycemia fear survey (HFS) and hyperglycemia avoidance scale (HAS) to daily CGM readings

  • Study used baseline data from an RCT trial (n=120 adults) to examine the relationship between hypoglycemia fear and hyperglycemia avoidance and Dexcom G6 readings
  • Hypoglycemia fear did not predict time <70 mg/dl or <54 mg/dl but did predict the likelihood of >10 scans/day (p=0.001)
  • Hyperglycemia avoidance predicted more readings <70 mg/dl and <54 mg/dl

Hypoglycemia Unawareness: Association with Continuous Glucose Monitoring and Autonomic Neuropathy

 

  • Assessment of predictive factors for hypoglycemia unawareness in 111 type 1s in Portugal
  • 14% and 10% had reduced and undetermined hypoglycemia unawareness, respectively; 23% of participants had autonomic neuropathy
  • Time below range and mean duration of hypoglycemia were significantly associated with hypoglycemia awareness
  • Mean duration of hypoglycemia and higher autonomic neuropathy scores were independent predictors of hypoglycemia awareness
  • Mean duration of hypoglycemia ≥91 minutes showed 83% sensitivity and 53% specificity for unawareness

Real-world estimates of severe hypoglycemia and associated healthcare utilization in the US: Baseline results of the INPHORM Study

  • 1,250 American adults with type 1 (17%) or type 2 (83%) completed a survey on their experience of severe hypoglycemia (SH)
  • Retrospective annual incidence proportion of SH was 37% (51% among type 1s, 33% among type 2s); incidence rate of 2.4 events/person/year (type 1 = 3.5, type 2 = 2.2)
  • Nearly half (47%) were treated completely outside of the healthcare system; hospital admission was only necessary in 4% of cases

Diabetes technology usage and self-efficacy in patients with type 1 diabetes

  • 2,054 adults with type 1 participated in a US survey; 72% used pump and CGM, 5% pump only, 16% CGM only, and 7% neither
  • The use of CGM, with or without a pump, may be associated with higher levels of diabetes-related psychological self-efficacy
  • Those using a pump did not see significantly higher self-efficacy scores than did those using neither CGM nor pump

Spotlight consultations: Illuminating patient priorities in type 2 diabetes

  • 18-person feasibility assessment of Spotlight, a novel ‘smart’ adaptive and dynamic patient questionnaire, found that it was effective in rapidly identifying type 2s’ priorities and top concerns
  • 18 adults identified their top two concerns: 100% identified the psychological burden of diabetes as their #1 concern, followed by gaining more skills about aspects of diabetes (39%), improving support around me (39%), and diabetes-related treatment issues (22%)
  • Feeling sad and helpless about living with diabetes (83%), feeling scared (61%), and lacking confidence (56%) were commonly reported
  • Those with diabetes duration >10 years were more likely to report lack of social support as a priority concern while those with diabetes <10 years were more likely to report gaining skills as a priority concern

Psychological care for children and adolescents with diabetes: Preliminary results from the International Pediatric Registry SWEET

  • Data from 92 pediatric endocrinology clinics in the international SWEET registry
  • 94% offer psychological care at type 1 diagnosis; 78% offer later psychological care via either referrals or self-appointment
  • 21% of centers have a systematic, annual psychological consultation
  • ~33% use a structured tool for psychological assessment

The Gatekeeper Study: Provider bias impacts diabetes technology recommendations for youth with type 1 diabetes (T1D)

  • Multi-disciplinary pediatric diabetes providers (n=39) completed a diabetes technology bias assessment tool
  • The majority (n=33, 85%) demonstrated bias, which was defined as recommending more technology for private insurance than public insurance or ranking insurance as one of the top two reasons to offer technology
  • The group with bias had practiced longer than the non-bias group (13 vs. 5 years; even when adjusted for age, sex, race/ethnicity, provider role, and %public insurance served, odds ratio =1.47) and were more likely to be non-Hispanic white (p=0.007)

Time-in-Range without automated insulin delivery systems in children with type 1 diabetes before and during a diabetes summer camp: A glass ceiling?

  • Retrospective analysis of CGM data before and during a 14-day Austrian summer camp for children with type 1 (n=31); no child on AID system, 65% on pumps
  • Time in Range was significantly higher during camp than before camp (66% vs. 56%) and time >180 mg/dl dropped significantly (-2.9 hours/day to 28%), but time <70 mg/dl increased 32 minutes/day to 5.9% during camp
  • There was no difference between those on insulin pumps vs. those on MDI
  • Authors note that Time in Range goals were not achieved during camp despite “high professional effort” and suggest that it could “only be achieved through” AID systems

CGMs and CSII benefits in kidney transplant recipients during and after the intervention

  • Study of 28 type 1s hospitalized in a tertiary transplant clinic, who were evaluated before surgery and two weeks, three months, and six months after kidney transplantation
  • Patients using technology before or after the transplant had statistically significantly better glycemic and renal function outcomes and fewer complications compared to those that didn’t use technology
  • Time in Range increased +1.2 hours/day to 89% and 81% among those using diabetes technology before and after transplant, respectively; those not using technology saw a similar Time in Range improvement but from a much lower baseline (46% to 52%)
  • In the short-term, the group that used technology prior to the transplant saw better outcomes, but that difference was not significant at six months

New Insulin Analogues

Title

Details + Takeaways

Treatment Satisfaction in People with Type 2 Diabetes Receiving Basal Insulin: Results from Real-World Studies and Randomized Controlled Trials with Insulin Glargine 300 U/ml

  • Trial compared patient satisfaction and perceived hypoglycemia frequency between insulin glargine U300, insulin glargine U100, insulin detemir, NPH, or insulin degludec in people with type 2
  • Initiating or switching to U300 was associated with increased treatment satisfaction in people with uncontrolled type 2, which may lead to increased adherence and meeting glycemic targets
  • Benefits of initiating or using U300 was independent of baseline diabetes treatment

Difference in patient-reported outcomes by age and region in adults with type 1 diabetes in the SAGE study

  • SAGE was a multinational, cross-sectional, observational study in people 26 years or older with type 1 diabetes for at least one year
  • This post-hoc analysis collected results from patient-reported outcomes (PROs) questionnaires, which were analyzed by region and age (n=3858)
  • PRO scores indicated low diabetes-related emotional stress and impact on quality of life, low hypoglycemia fear, and high treatment satisfaction across the study’s regions and age groups; unexpected due to poor glycemic control in SAGE population
  • Differences in PRO scores may reflect inequities in type 1 diabetes management, specifically healthcare and cultural factors

Glycaemic Control And Hypoglycaemia In High-Risk Subgroups Of People With Type 1 Diabetes (T1D) In The Sage Study

  • This study analyzed patient characteristics, glycemic outcomes, and metabolic complications in high-risk subgroups with history of hypoglycemia and renal impairment enrolled in the SAGE study
  • Glycemic control in people with type 1 was suboptimal across age groups in SAGE, indicating that high-risk subgroups of people with type 1 are less likely to reach individual glycemic targets
  • History of severe hypoglycemia and pre-existing renal impairment were associated with increased risk of hypoglycemia and hyperglycemia leading to DKA

Association of patient-reported outcomes scores with glycaemic target achievement in type 1 diabetes in the SAGE study

  • SAGE was a multinational, cross-sectional, observational study in people 26 years or older with type 1 diabetes for at least one year
  • This post-hoc analysis collected results from patient-reported outcomes (PROs) questionnaires, which studied the percentage of participants meeting A1c targets and whether there was an association between PRO scores and achievement of A1c targets (n=3858)
  • The 26-44 age group was found to have the highest general A1c target (<7%) achievement (27.6%), while the > 65 age group had the highest physician-directed A1c target achievement (26.2%)
  • Patients tended to have lower diabetes-related emotional distress and higher treatment satisfaction if they achieved general and physician-directed A1c goals

Effectiveness and Safety of Gla-300 Vs Ideg-100 Evaluated with CGM in Adults with T1D in Routine Clinical Practice in Spain – The OneCARE Study

 

  • This trial sought to compare the efficacy and safety of insulin glargine U300 to insulin degludec U100 using CGM in adults with type 1 diabetes
  • There was no significant difference in Time in Range between those using glargine U300 and degludec U100
  • In people who switched from first-generation basal insulin analogues, glargine U300 use led to similar efficacy and safety as degludec, but with higher nighttime Time in Range

Second-Generation Basal Insulin Analogues: First-Choice In Reducing Real-World Severe Hypoglycemia (Baseline Results Of The Inphorm Study, USA)

 

 

  • The crude, real world frequency of severe hypoglycemia in adults taking 1st or 2nd generation basal insulin (n=68,000) was determined, with quantification of potential causal effects
  • Participants had both type 1 (n=10,000) and type 2 diabetes (n=58,000)
  • Real world data confirmed that 2nd generation basal insulins reduce rates of severe hypoglycemia compared to 1st generation basal insulins, irrespective of bolus insulin use
  • Results suggest that clinicians should prioritize 2nd generation basal insulin over 1st generation

Effect Of Insulin Degludec Versus Insulin Glargine U100 On Occurrence Of CGM-Recorded Nocturnal Hypoglycemia In People With Type 1 Diabetes And Previous Nocturnal Severe Hypoglycemia

  • Compare insulin degludec with insulin glargine U100 in 67 participants with T1D and ≥ 1 episode of nocturnal severe hypoglycemia in the preceding 2 years
  • CGM traces over a 6-day period were reviewed for hypoglycemic events lasting > 15 minutes
  • Classified two levels of hypoglycemic events: IHSG Level 1 ≤ 3.9 mmol/L; IHSG Level 2 < 3.0 mmol/L
  • A significant reduction in nocturnal non-severe hypoglycemia was observed in insulin degludec as compared to insulin glargine U100 at both levels of IHSG

The Effect of the Ultra-Rapid Insulin Analog Fiasp® in Pediatric Type 1 Diabetes Patients Under Continuous Subcutaneous Insulin Infusion (CSII)

  • Given that postprandial hyperglycemia is a large challenge for youth with type 1, this study sought to determine the efficacy of Fiasp in metabolic control of youth with type 1 using insulin pumps
  • Using Fiasp led to significantly increased metabolic control (increased Time in Range, decreased time in hyperglycemia)
  • Ultra-rapid insulin holds promise in youth given their preference for carbs and lack of compliance with “wait times” between bolus insulin and eating. No studies have been done on Fiasp in pumps prior to this

Older people with type 2 diabetes benefit more after switching from NPH insulin to glargine 300 U/mL: a post-hoc analysis of a multicentre, prospective, observational study

  • This post-hoc analysis was conducted on an observational study in Polish patients with type 2 diabetes ≥18 years old who switched from NPH insulin to glargine and had A1c≥8% within four weeks before the switch (n=469)
  • Patients were observed through a six-month follow-up period after the switch, with results analyzed by age (≤65 years, n=224, and >65 years, n=245)
  • A1c control was similarly and significantly improved in both older (from 9.15% to 8.20% among those >65 years) and younger (from 9.23% to 8.13% among those ≤65 years) type 2 diabetes patients
  • Greater reductions in hypoglycemia risk were observed in older patients (>65 years) and those with longer disease duration (>13 years)

Clinical Decision Support Systems/Analysis

Title

Details + Takeaways

People with Diabetes and Caregivers Prefer Nasal Glucagon over Conventional Injectable Glucagon: A Discrete Choice Experiment

 

  • Mixed-method patient preference study examined patient and caregiver preferences for glucagon in severe type 1 and type 2 diabetes
  • 78% of patients and caregivers preferred nasal glucagon over injectable glucagon.
  • When glucagon efficacy is comparable, ‘storage temperature’ and ‘delivery method’ had the highest relative attribute importance, driving preferences

The prevalence of NAFLD in a cohort of type 1 diabetes subjects based on non-invasive assessment and its association with macrovascular events

  • This cross-sectional study sought to assess the prevalence of nonalcoholic fatty liver disease (NAFLD) in type 1 diabetes patients (n=400) according to three non-invasive indices – ultrasound (US), fatty liver index (FLI), and controlled attenuation parameter (CAP) – and evaluate if there was an association between NAFLD and prevalent macrovascular disease
  • The NAFLD diagnosis rate was similar using US (22.8%) and FLI (22.3%) but was much higher using CAP (35%), demonstrating the need for consensus on diagnostic tools
  • NAFLD diagnosed by US and FLI – but not CAP – was independently correlated with prevalent cardiovascular events

 

New Medications for Treatment of Diabetes

Title

Details + Takeaways

SGLT-2 Inhibitor-Induced Diabetic Ketoacidosis (DKA) in Acute Medical Setting: Would you recognise it?

  • This poster highlighted four case studies of adults with type 2 diabetes who were in situations that required discontinuation of SGLT-2 inhibitors
  • These case studies showed that DKA risk must be acknowledged when prescribing SGLT-2 inhibitors
  • The risk of DKA increased when patients were experiencing concomitant illness, so people should monitor ketone levels when feeling unwell even in glucose is in range

Dasiglucagon, A Ready-To-Use Glucagon Analog, For Fast and Effective Treatment of Severe Hypoglycemia: A Phase 3, Randomized Controlled Trial in Children with Type 1 Diabetes

  • This phase 3 trial (n=42) was performed to determine the safety and efficacy of dasiglucagon in youth (age 6-17 years) with type 1 diabetes
  • Dasiglucagon showed superiority on the primary outcome of blood glucose recovery, with median time to recovery of 10 minutes compared to 30 minutes with placebo. All participants who used dasiglucagon reached blood glucose recovery within 20 minutes
  • Dasiglucagon is effective in restoring blood glucose levels in youth with hypoglycemia, with a safety profile similar to traditional glucagon

Impact Study: IMCY-0098 Proof Of Action In Type 1 Diabetes, On The Way To A Specific Disease-Modifying Treatment

  • IMCY-0098’s (Imotope in development for treatment of type 1 diabetes) proof of concept study investigated differences in Dried Blood Spots (DBS) fasted c-peptide in people with type 1 versus placebo (n=84)
  • Imotopes are linear synthetic peptides that combine sequences of HLA class II T-cell epitopes. These epitopes are linked to immune-mediated diseases, so treatment with imotopes ideally should generate cytolytic CD4 T-cells that induce apoptosis of antigen presenting cells that lead to tissue damage and disease, all while avoiding autoimmune attacks
  • Secondary endpoints will investigate changes in c-peptide response during mixed meal tolerance test, changes in DBS, and effects on A1c, hypoglycemia, DKA, daily total insulin dose, and CGM measures
  • Exploratory endpoints will investigate the impact of IMCY-0098 on autoantibodies and antigens presented by islet beta cells overtime 

The Dual GIP and GLP-1 Receptor Agonist Tirzepatide Improves Biomarkers Associated with Cardiovascular Risk in Patients with Type 2 Diabetes (T2D)

  • The effect of tirzepatide on lipoprotein and inflammatory and endothelial dysfunction biomarkers was measured in patients with type 2 diabetes and moderate obesity
  • This study aimed to determine tirzepatide’s effects on cardiovascular risk, as lipoprotein and biomarker levels are associated with higher CV risk
  • Tirzepatide decreased levels of lipoprotein (apoB, small LDLP) and inflammation/endothelial dysfunction (hsCRP, YKL-40, MCP-1, ICAM-1, GDF-15) biomarkers
  • Tirzepatide’s effects on CV events in patients with type 2 and eCVD versus dulaglutide are being studied in SURPASS-CVOT

Tirzepatide, a dual GIP and GLP-1 receptor agonist, increases insulin sensitivity and improves pancreatic beta cell function in type 2 diabetes

  • Double-blind phase 2b study of people with type 2 diabetes over 26-weeks, comparing placebo (n=51), tirzepatide 1 mg (n=52), tirzepatide 5 mg (n=55), tirzepatide 10 mg (n=51), tirzepatide 15 mg (n=53), or dulaglutide 1.5 mg (n=54)
  • Tirzepatide showed significantly greater glucose control and weight loss at 26 weeks vs. placebo or dulaglutide (excluding the 1 mg tirzepatide dose)
  • Tirzepatide improved measures of insulin resistance (HOMA2-IR) and improved markers of insulin sensitivity like adiponectin, IGFBP1, and IGFBP2 vs. placebo and dulaglutide (excluding the 1 mg tirzepatide dose)
  • Importantly, insulin-sensitizing effects of tirzepatide were only partially attributable to weight loss, suggesting additional unique mechanisms
  • Tirzepatide also improved markers of beta cell function, including HOMA-2B and intact proinsulin:C-peptide ratio vs. placebo and dulaglutide (excluding the 1 mg tirzepatide dose)

Real world evidence for the glycometabolic durability of sodium glucose cotransporter 2 inhibitors – four year Indian study

  • Longitudinal study using electronic health record data from Lina Diabetes Care Center SGLT-2 registry database
  • Analyzed change in A1c, body weight, and blood pressure evaluating durability of metabolic efficacy and safety over last four years
  • 554 patients with type 2 diabetes on dapagliflozin (n=204), empagliflozin (n=211), and canagliflozin (n=139)
  • Durable glycometabolic effects of SGLT-2 inhibitors were demonstrated: mean a1c decreased from 9.0% to 7.6% (p<0.0001), mean Systolic BP reduced by 4.7 mmHg and Diastolic by 2.2 mmHg, with no significant reduction in body weight

Effect of Dapagliflozin on Body Composition in Patients with Type 1 Diabetes

  • This trial examined the effects of dapagliflozin on body composition (i.e., changes in fat mass and hydration status) in adults with type 1 (n=12)
  • Percentage of change was highest in visceral fat and fat mass after six months of treatment with dapagliflozin
  • Treatment with dapagliflozin led to significant body weight reduction due to decreased fat mass 

COVID-19 and Diabetes

Title

Details + Takeaways

Glycemic Control in Young Italian Patients with Type 1 Diabetes During Covid-19 Lockdown: The Role of Age, Type of Insulin Therapy, Telemedicine and Physical Activity

  • This trial examined glycemic control using DexcomG6 before and during COVID-19 lockdown in Italian youth with type 1 diabetes (n=202, median age 18.2)
  • Data showed improvement of Time in Range from 54.58% to 59.09%
  • Glycemic control improved in all age groups, although physical activity decreased universally during lockdown
  • The COVID-19 lockdown led to an unexpected improvement in glycemic control for young Italian patients with type 1 diabetes

Social media, COVID-19 pandemic and type 1 diabetes – Instagram live promoting information and social support

  • Study assessed Instagram Live as a tool for promoting information and social support about COVID-19 and type 1 diabetes
  • Instagram platform indicators reviewed from March 20-30, 2020: 14,000 of 29,000 followers were reached every week
  • Live Instagram videos received six times more interactions than regular videos, indicating its value as an engagement strategy for spreading knowledge about diabetes and COVID-19

Pediatric diabetes and COVID-19 infection

  • Review of cases of pediatric diabetes ketoacidosis (DKA) associated with COVID-19 infection
  • Five pediatric DKA cases were associated with COVID-19 infection, four of which were new onset type 1
  • The authors suggest that poorly controlled diabetes or unrecognized onset of the disease can cause immune dysregulation and increase the risk of COVID-19 infection
  • There may be a connection between COVID-19 infection and diabetes as an immune trigger

Increasing the number of diabetic patients and COVID-19 epidemic – A call for cooperation between specialists and general practitioners in patient education

  • This study compared the effectiveness of group education for people with type 2 diabetes guided by diabetologists to treatment guided by general practitioners.
  • Group education guided by diabetologists had a significant long-term effect on improving diabetes control and weight loss in diabetic patients.

The COVID-19 pandemic and new onset pediatric type 1 diabetes at Children's National Hospital

  • Study investigated changes in type 1 diabetes initial clinical presentation and patient demographics during COVID-19
  • Retrospective cross-sectional review of youth (≤ 21) diagnosed with type 1 diabetes during the COVID-19 pandemic (n=119) and the time-matched period in 2019 (n=118)
  • Incidence of DKA did not significantly increase, but more youth presented with severe DKA (p=0.01) and fewer with mild DKA (p=0.03)
  • Proportion of youth diagnosed with type 1 diabetes who were publicly insured increased from 40.3% in 2019 to 55.1% during the pandemic (p=0.05)

The Effect of Lockdown and Easing Of Restrictions On Glycaemia In Children With Type 1 Diabetes During The Covid-19 Pandemic

  • Trial investigated the impact of lockdown and the end of restrictions on glycemic control in 13 children and adolescents with type 1 diabetes in the UK
  • Results showed significant differences in time below range (TBR; <50 md/dL); increased during lockdown and immediately after
  • Otherwise, there was no significant change in glycemic variables, like Time in Range, observed during and after lockdown

Short-term impact of COVID-19 pandemic lockdown on metabolic control of patients with type 2 diabetes mellitus

  • This observational study analyzed glycemic control trends in patients with type 2 diabetes (n=120) in Spain throughout the government-initiated lockdown
  • Data from before and after the lockdown was obtained from 75 of the 120 subjects enrolled (The time range when the data was collected was not specified)
  • Significant increases were observed in mean fasting glucose (139.2 to 154.6, p=0.008), blood pressure (131 mmHg to 132.1 mmHg, p<0.001), A1c (7.29% to 7.61%, p=0.025), and BMI (27.9 to 30.3, p<0.001) from before lockdown to after lockdown
  • 40% of subjects (n=30) experienced an increase in A1c greater than 0.3%
  • These findings suggest public health lockdowns have impeded people’s ability to adequately control their diabetes

Improved glycaemia during the COVID-19 lockdown is sustained in adults with type 1 diabetes

  • Retrospective, observational study assessing the impact of lockdown on glycemia and whether these effects were sustained following easing of restrictions in the UK
  • Glucose data from 97 adults aged 41.0 years with diabetes duration of 22 years, collected over 28 days from pre-lockdown, during lockdown, and post-lockdown
  • % TIR (70-180 mg/dL) increased during the lockdown period compared to pre-lockdown (p=0.007)
  • No significant differences in % TIR between during and post-lockdown 
  • Demonstrated an improvement in glycemia during lockdown with sustained effect for two months after easing of restrictions

Effect Of Screen Time On Glycaemic Control With Type 2 Diabetes Patients During Covid19 Outbreak: A Survey Based Study

  • This study examined the effect of screen-time spent on social media per day on glycaemic parameters of type 2 patients
  • The authors analyzed 229 patient surveys. Outcomes examined were poor control of glycaemic parameters like HbA1c (>7%), FBS (>150 mg/dL), and PPBS (>200 mg/dL)
  • The presenters suggest that there is an increased risk of uncontrolled glycaemia if the patient spends more time on social media with reduced physical activity

Type 1 Diabetes During the SARS-CoV-2 Pandemic. Does Lockdown Affect the Metabolic Control of Pediatric Patients?

  • 128 children with type 1 diabetes (aged 2-18 years) were monitored before and during the SARS-Cov-2 epidemic lockdown in Poland
  • A1c levels improved; average blood glucose, Time in Range and CV did not change
  • Total Daily Dose of insulin (TDD) increased, possibly due to reduced physical activity. This was supported by increased BMI in participants during lockdown 

Improving CGM outcomes in adults with type 1 diabetes mellitus during COVID-19 lockdown in Piedmont, Italy

  • Diasend and Dexcom CLARITY data was collected from 148 adults with type 1 in Italy who used CGM or CGM + pump during the two months prior to lockdown (January-February) and during early lockdown (March-April)
  • During lockdown, participants saw an +30 minute/day increase in Time in Range and a -0.2% decline in GMI to 7.1%; no change in time below range (3.7% vs. 4%)

Description of the glycemic control metrics evaluated by glucose monitoring in a cohort of adult patients hospitalized with hyperglycemia and diagnosis of COVID-19

  • Prospective cohort study of patients with diabetes and/or hyperglycemia and SARS-CoV2 (n=61) to assess the association between CGM metrics and a composite outcome of death, acute respiratory distress syndrome or acute renal failure
  • ~30% had ≥1 severity marker for COVID-19 infection, half of the patients received dexamethasone, and all were treated with MDI therapy
  • There was no association between CGM data and adverse outcomes; although the data trended toward an association between time above range and the composite outcome, it was not significant

COVID-19: Quality of life and risk attitudes among insulin pump users in a time of crisis

  • 833 insulin pump users in Denmark responded to a quality-of-life survey in June 2020
  • Participants reported a somewhat negative impact of COVID-19 on their quality of life; did not differ by gender or age
  • Women and older respondents assessed the risk of becoming seriously ill if infected with COVID-19 as higher than did men and young respondents

Impact of COVID-19 pandemic on provision of continuous subcutaneous insulin infusion (CSII) and continuous glucose monitoring (CGM) starts and renewals in the United Kingdom (UK)

  • Online survey of Diabetes Technology Network (DTN) HCP members who treat type 1s (n=143) conducted between November-December 2020
  • Reduction in new pump starts and renewals reported by 73% and 61% of respondents, respectively
  • 33% reported that >60% of their new pump starts were conducted virtually; respondents reported that virtual CGM and pump starts were as effective as in-person
  • Median 167 pump users, 25 rt-CGM users, 300 Libre users per clinic

Devices Focused on Diabetic Preventions

Title

Details + Takeaways

CHIC-D - cardiovascular health in children with type 1 diabetes – early detection, cardiovascular prevention and treatment monitoring

  • CHIC-D is an ongoing project examining the time course of vascular changes in children with type 1 diabetes and the impact of metabolic control and blood pressure on different layers of the arterial wall
  • So far, data from 36 children (ages 6-15) with type 1 diabetes randomly selected from pediatric diabetes registry SWEDIABKIDS and 23 health controls
  • Preliminary results show a tendency towards increased intima-thickness in radial artery among children with type 1 diabetes (p=0.09), which may be an important marker for early vascular damage

A Randomized Pilot Trial: Using An Activity Tracker To Increase Motivation For Physical Activity In Patients With Type 2 Diabetes In Primary Care

  • 30 patients with type 2 diabetes were given activity trackers in primary care settings to increase motivation for physical activity
  • This intervention lasted three months, with three face-to-face appointments with a kinesiologist
  • Results showed that use of an activity tracker improved cardiometabolic risk and is a promising motivation tool for clinicians
  • This was the first study of its kind to examine activity tracker within the context of a clinician-led physical activity intervention

Trials in Progress

Title

Details + Takeaways

Transition from insulin pump to multiple daily injections using insulin degludec: interim results from a randomized controlled trial

  • This 3-week, randomized control study compared the efficacy and safety of the ‘standard of care’ to an ‘overlap’ strategy of transitioning from an insulin pump to multiple daily injections (MDI) using insulin degludec
  • The study enrolled adults >18 years with type 1 diabetes for >1 year with A1c between 6.5% and 8.5% and excluded those on hybrid closed-loop systems who did not use manual mode
  • Those on the ‘standard of care’ (n=9) method stopped insulin pump use and started insulin degludec in a 1:1 dose on the day of randomization
  • Those on the ‘overlap’ method (n=7) started insulin degludec 1:1 at pump basal dose on the day of randomization and continued pump therapy for the first 48 hours with a gradual basal reduction (50% for 0-24 hours, 75% for 24-48 hours, discontinued after 48 hours)
  • Time-in-range was significantly higher for those on ‘overlap’ (p=0.03), and there were no significant differences in time-below-range between overlap and standard of care

Association of Diabetic Nephropathy with instruments for prediction of cardiovascular risk in patients with Type 1 diabetes

  • This poster examined the association of diabetic nephropathy with type 1 diabetes specific scales predicting cardiovascular disease.
  • 176 people with type 1 were screened for cardiovascular risk using Steno T1 scale and Swedish T1D scale.
  • The study showed significant and direct correlation of albuminuria stage with 5-year and 10-year risk of CVD with Steno and Swedish scales.

Inducing beta cell rest with insulin – a part of the azithromycin insulin diet intervention in type 1 diabetes (AIDIT)

  • Ongoing AIDIT study aims to preserve beta cell function in recently diagnosed children with type 1 diabetes, using (i) Azithromycin to reduce bacteria-induced subclinical pancreatitis; (ii) Insulin-pulses to achieve maximum beta-cell rest; (iii) Dietician support to limit meal volume and encourage time spent per meal to prevent digestive reflux
  • Insulin lispro infusion given for 72 hours (1U/ml) within one week of diagnosis and via intensified subcutaneous infusion with Tandem t:slim insulin pump (100 U/ml) for six to eight hours during one day in study Weeks 5, 9, 13, 17, 25, 34, and 43
  • At present, seven out of 14 participants aged six to 15.9 years old (n=2 female) have been randomized to treatment protocol (all with stimulated C-peptide >200 pmol/l six weeks post-inclusion)
  • 5/7 IV insulin treatments and 27/39 subcutaneous insulin treatments resulted in induction of beta cell rest, defined as C-peptide <100 pmol/l
  • Study provides early evidence that intensive insulin treatment could be used to induce beta cell rest in newly diagnosed T1D

ATTD Yearbook

The COVID-19 Pandemic and Virtual Clinics for Diabetes Care

  • Dr. Satish Garg (Barbara Davis Center) kicked things off on Friday morning with his overview of telemedicine and virtual diabetes care during the COVID-19 pandemic. As Dr. Garg noted, this topic was not planned, and the world changed drastically shortly after the end of last year’s ATTD 2020 meeting in Madrid. Indeed, the pandemic was so impactful on aspects of diabetes care that DT&T ran an entire special issue dedicated to virtual/telehealth. Dr. Garg highlighted a paper from UNC’s Diabetes Care Team, led by Dr. John Buse, that demonstrated success in transitioning inpatient diabetes care to virtual. As shown below, Time in Range did not change at all after transitioning to virtual care. Looking ahead, Dr. Garg expressed his hope that reimbursement parity for telemedicine visits would continue into the future.

Continuous and Intermittent Glucose Monitoring

  • Slovenia’s Dr. Klemen Dovč highlighted three studies during his presentation on this year’s advances in CGM, flash glucose monitoring, and self-monitoring of blood glucose. The chapter contains a total of 14 publications for 2020, representing a selective few out of the 500+ manuscripts published in 2020 and early 2021. To start, Dr. Dovč featured the CITY study, a parallel RCT that assessed the effect of CGM on glycemic control in adolescents and young adults with type 1 (n=153). The study found that CGM incurred a -0.37% adjusted improvement in A1c relative to BGM, as well as a +1.4 hours/day improvement in Time in Range. Dr. Dovč then pivoted to the SILVER study, a continuation of the GOLD study that showed sustained effects of Dexcom G4 vs. SMBG out to 2.5 years in 107 type 1s on MDI. Lastly, we were treated to one of the most important Time in Range validations published thus far (n=6,225 adults with type 2), which was published in Diabetes Care in October 2020 and shows that each 10% reduction in Time in Range is associated with a 5% higher CVD mortality and an 8% increase in all-cause mortality.

Insulin Pumps

  • Dr. Rayhan Lal of Stanford presented the insulin delivery hardware section of this year’s ATTD Yearbook – previously the insulin pumps chapter, but updated this year to also include the growing body of literature around smart insulin pen technology. Dr. Lal began with a study from Adolfsson et al. highlighting a “mild statistically significant increase” in Time in Range and reduced severe hypoglycemia for patients using the NovoPen 6 smart insulin pen. Next, Dr. Lal discussed infusion sets, which, last year, he identified as “the weakest link in the chain” of insulin delivery. Presenting data from Waldenmaier et al. that compared the survival curves of the YpsoPump steel Orbit micro infusion set and Teflon Orbit soft infusion set ultimately concluding that the replacement interval of patients’ infusion sets should be individualized. Staying with infusion sets, Dr. Lal outlined fascinating research from Tschaikner et al. on the accuracy of a combined cannula and glucose sensor (Dexcom G4) which found that CGM accuracy was maintained when the glucose sensor was placed 6mm from the cannula opening. This data provides potential path forward for developing a single on-body dual glucose sensor and infusion set, effectively reducing the number of on-body wearables for people with type 1 diabetes by half, and may be a growing area of investigation moving forward. Finally, turning to insulin pumps specifically, Dr. Lal showed results from Girardot et al. which indicated that the error of basal insulin delivery varies across insulin pumps, with pumps tending to be less accurate at lower basal rates.

New Insulins, Biosimilars and Insulin Therapy

  • Dr. Lutz Heinemann (Profil) gave this year’s update on new insulins, biosimilars, and insulin therapy, first noting that 2021 is a landmark year as it marks the 100th anniversary of the discovery of insulin. The first “hot topic” Dr. Heinemann highlighted in his presentation was once-weekly basal insulins. Indeed, Novo Nordisk’s once-weekly basal insulin icodec read out positive phase 2 results at ADA 2020 and phase 3 trials are ongoing. Lilly’s phase 2 results for its once-weekly basal insulin-FC read out at ENDO 2021. Moving on, Dr. Heinemann discussed a number of open questions around the idea of “smart” (glucose responsive) insulins, including whether or not the current paths of research are realistic. A number of open questions also exist around biosimilars, the first of which started to make their way onto the US market in 2020. A couple of particular questions Dr. Heinemann noted were whether these products could receive regulatory approval as “interchangeable” with traditional insulins and how much they would affect the price of insulin, particularly in the US.

Decision Support Systems and Closed-Loop

  • UVA’s Dr. Boris Kovatchev took on the hugely important topic of decision support and AID systems. Rather than focusing in on the findings of specific studies, Dr. Kovatchev discussed the themes of the chapter’s 15 highlighted studies. The five highlighted decision support studies include two on using advanced data science methods to support diabetes decision support (deep learning and big data analytics), one on the importance of patient input in designing tools, and two on the use of decision support tools in patients on MDI and those with type 2. The ten highlighted closed-loop studies were separated into four themes: (i) pivotal trials and real-life use (the Control-IQ adult pivotal; real-world analysis on pediatric discontinuation of MiniMed 670G; a long-term real-world study of MiniMed670G vs. sensor-augmented pump therapy); (ii) studies of “mobile systems” that are controlled via a smartphone (three-month RCT on inControl AP; an iAPS meal test; an investigation of basal rate and carb ratio learning algorithms); (iii) first trials of new systems including Omnipod 5, MiniMed 780G, and Beta Bionics’ insulin-only Bionic Pancreas; and (iv) one study on the use of faster insulin in closed loop systems.

Using Digital Health Technology to Prevent and Treat Diabetes

  • Canary Health founder Dr. Neal Kaufman discussed the growing field of digital health highlighting two publications on the effectiveness of a social media-based diabetes self-management intervention and methods to enhance patient activation. The first study, published in the Journal of Nursing Scholarship evaluated the effects of a social-media based intervention that also took into account health literacy and was delivered by nurse educators either via social media or telephone. Results were compared to a control group receiving usual care and indicated unsurprisingly, that patients with higher health literacy demonstrated higher levels of patient activation. However, patients who received the telephone-based intervention did see higher levels of diabetes self-care behaviors than those in the usual care group suggesting the intervention may be effective at “lessening the disadvantages faced by people with low health literacy” and that social-media based approaches can be effective when they are created with the target population’s health literacy level in mind. Moving on, the second study Dr. Kaufman discussed evaluated the feasibility of using the US Department of Defense Health Care Environment to enhance patient activation among people with type 2. While patient activation has been shown to be associated with increased self-management behaviors, medication adherence, patient satisfaction, and improved quality of life, this study which provided patients with “behavioral messaging” did not show a significant increase in patient activation suggesting that messaging alone may not be enough to change long-term behaviors and that such changes require more intensive intervention. In his discussion of both studies, Dr. Kaufman also emphasized the importance of recruiting “the right person for the right intervention at the right time” expressing that not all digital interventions will work for all patients and that interventions need to be specific to the needs of their target populations.

Technology and Pregnancy

  • Dr. Jennifer Yamamoto (University of Calgary) gave the very important Yearbook update on diabetes technology in pregnancy. Most notably, she highlighted a paper in DT&T by Castorino et al. evaluating the accuracy of Dexcom G6 in pregnancy women. The trial enrolled 32 participants who wore a total of 63 sensors. On average, across three wear sites, overall MARD of the sensor was 10.3% compared to YSI reference – quite impressive. The posterior arm wear location, in particular, recorded a MARD of just 8.7%. Dr. Yamamoto did note that the study had limitations around the amount of sampling done at very low and high glucose ranges, but the data are nonetheless promising. As a reminder, Dexcom G6 announced CE-Marking in pregnancy at ATTD last year, but there are no currently available systems in the US for pregnant women.

Diabetes Technology and Therapy in The Pediatric Age Group

  • Dr. David Maahs (Stanford) discussed how these technologies differ in pediatric populations compared to adult populations, highlighting five of the 15 selected studies in the chapter. He began with the pediatric Control-IQ pivotal trial (n=101), which showed that Control-IQ users saw a +3.4 hours/day increase in Time in Range (67%) from baseline, a significantly greater improvement than the sensor-augmented pump/Basal-IQ group. He then discussed data from the German DPV Registry that showed that regular CGM use is associated with improved metabolic control and reduced DKA in a real-world pediatric population (n=3,553 type 1s). Next, he pointed to six-month real-world data on pediatric use of MiniMed 670G (n=92), showing a 0.3% A1c drop at month six, but also a 30% discontinuation rate. Shifting gears, he turned to discuss an RCT on the use of Fiasp vs. Aspart in children with type 1, which showed noninferiority of Fiasp. To close out, Dr. Maahs shared the findings of an observational study that suggests that young children have higher variability in insulin requirements on AID systems. Together, Dr. Maahs argued staunchly that yes, diabetes technologies and therapies differ in pediatric populations.

Advances in Exercise and Nutrition as Therapy in Diabetes

  • York University’s Dr. Michael Riddell and Stanford’s Dr. Dessi Zaharieva presented this year’s chapter on exercise and nutrition discussing three studies in particular that evaluated the efficacy of an at-home high-intensity interval training (HIIT) program, the ability of Diabeloop’s closed-loop system to manage large meals and exercise, and the effects of a Mediterranean diet on glycemic control in patients with type 2. Dr. Riddell shared data from the first study demonstrating that six weeks of HIIT exercise among patients with type 1 was associated with a 13% decrease in insulin doses as well as a 7% increase in VO2max. Additionally, the study had a 95% adherence rate demonstrating acceptability of the activities and no cases of exercise-induced hypoglycemia were observed. Turning to closed-loop control, Dr. Zaharieva discussed a study from Hanaire et al. demonstrating a 24% increase in Time in Range for patients on Diabeloop’s DBLG1 closed-loop system compared to sensor augmented pump therapy. The study specifically evaluated the ability to the system to maintain glycemic control when patients consumed large meals and participated in physical activity with the system outperforming sensor-augmented pump therapy in both scenarios (see graphs below with glycemic data from closed-loop users in red). Finally, a recent publication in Diabetes Care suggested that a Mediterranean diet enriched with olive oil was associated with taking fewer diabetes medications at 3.2 year follow-up compared to patients on a typical low-fat diet. This study was especially insightful because patients were enrolled shortly after diagnosis with type 2 suggesting the enriched Mediterranean diet may also have contributed to slowed disease progression.

Primary Care and Diabetes Technologies and Treatments

  • Drs. Thomas Martens and Greg Simonson (International Diabetes Center) provided this year’s update on diabetes technologies and treatments in primary care. Dr. Martens started things on the technology side, highlighting the GP-OSMOTIC study published in Lancet Diabetes Endocrinology in 2020. The Australian RCT randomized 149 adults with type 2 diabetes to use professional CGM (FreeStyle Libre Pro) every three months right before their clinic visit. The other 150 participants followed usual care with professional CGM used at baseline and twelve months for data analysis purposes. During the study, the professional CGM group saw their A1c improve from 8.9% to 8.2%; the control group saw an A1c improvement from 8.9% to 8.5%. The between-group difference of 0.3% was no statistically significant (p=0.06). Dr. Simonson took on the therapy side of diabetes in primary care, highlighting a study published in Diabetes Care evaluating the impact of metformin + various medications on MACE, severe hypo, and all-cause mortality (n=66,807 people with type 2 diabetes in the Danish National Patient Registry). Not too surprisingly, patients on metformin and a sulfonylurea had the highest rates of both MACE and severe hypoglycemia. Interestingly, the addition of a DPP-4 inhibitor to this therapy regimen decreased both MACE and severe hypoglycemia significantly. Metformin + newer diabetes drugs (i.e., GLP-1s and SGLT-2s) were associated with significantly lowered MACE and severe hypoglycemia events.

Practical Implementation of Diabetes Technology

  • Barbara Davis’ Dr. Laurel Messer – who seems to be everywhere we turn at this year’s ATTD 2021 – presented “Practical Implementation of Diabetes Technology.” This year’s chapter includes ten highlighted studies across three themes: (i) benchmarks in the real-world; (ii) person-level barriers/device discontinuation; and (iii) system-level barriers and implications. Dr. Messer highlighted two studies in her brief talk: the FUTURE study and the three-part HUNT study. In the FUTURE study, data following the 2016 nationwide reimbursement for FreeStyle Libre in type 1s in Belgium showed that improving access to isCGM appreciably increased treatment satisfaction and improved real-world health outcomes with fewer hospitalizations, less severe hypoglycemia, and less work absenteeism at six and twelve months. In the Nord-Trondelag Health (HUNT) study, Norwegian researchers found that medium and high socioeconomic status was associated with a higher odds of using new technologies, that education may have a larger impact than income, and that the diffusion of technology may be faster in glucose monitoring than in insulin delivery. Dr. Messer closed out by arguing that improving payer coverage is essential to “leveling the playing field” and that it is not sufficient to only address implicit provider bias and access to specialty care, which are pieces of – but not the entirety of – the issues resulting in unequal access and use.