April 27-30, 2022; Barcelona, Spain (+Virtual); Day #4 Highlights – Draft

Day #1 Highlights – Omnipod 5 type 2 feasibility and human factors data; plans for a consensus report on CGM-based therapy modification in type 2 diabetes; MiniMed 780G extension phase with Guardian 4; Tempo smart button expected FDA/CE-Mark approval by EOY; Eversense E3 CE-Marking “quite soon”; evidence for CGM in non-insulin using type 2s; beyond A1c posters

Day #2 Highlights – Dexcom G7 has launched in the UK – early feedback from first seven global users; FreeStyle Libre 3 US accuracy data; Dexcom G7 pediatric pivotal data; Dexcom ONE real-world data; head-to-head of 780G vs. 670G in type 1s ages 7-14; RCT data on direct transition from MDI to 780G

Day #3 Highlights – InRange study on Time in Range with Toujeo vs. Tresiba; ATTD Yearbook; more details on Omnipod 5 type 2 feasibility study and limited launch; Dexcom G6 real-world data shows the importance of engagement to outcomes

Top 17 Highlights

1. Beta Bionics’ insulin-only iLet RCT pivotal (n=440): -0.5% A1c improvement and +2.6 hour/day TIR gain on insulin-only vs. standard care of any other insulin delivery method (including MDI, pump, AID) + Dexcom G6 for adults and children with type 1 after thirteen weeks

Speaking to a packed room of rapt listeners, Dr. Steven Russell (Harvard Medical School) read out the highly anticipated results of the investigator-initiated insulin-only iLet pivotal trial, which compared the novel system vs. Dexcom G6 and any other insulin delivery method (including MDI, pump, or AID) in children (n=165, ages 6-18) and adults (n=275). At the end of the 13-week trial, those on iLet saw a +2.6 hour/day Time in Range improvement relative to the standard care group when adjusted for baseline (p<0.001), “almost all” of which occurred within the first day or two of wear. Specifically, those on iLet saw their Time in Range improve from 51% at baseline to 65% at thirteen weeks whereas the standard care arm saw only a slight improvement to 54%. A1c improved a significant 0.5% on iLet vs. standard care (p<0.001 for baseline-adjusted mean difference; n=326). Those on iLet saw their A1c fall 0.6% from 7.9% at baseline to 7.3% at 13 weeks, while the standard care arm saw no change from baseline to 13 weeks, maintaining an A1c of 7.7%.

Topline Results, thirteen weeks

Study Design and System Overview

At baseline, 89% of participants were using CGM and about a third were on MDI (34%), a third were on pump (32%, most on CGM), and a third were on an AID system, most of whom were on Control-IQ (35% total, 23% Control-IQ, 8% MiniMed 670G, 4% PLGS). Children were randomized 2:1 to iLet (n=112) or standard care (n=53), which was defined as Dexcom G6 and their usual insulin delivery method, while adults were randomized 2:1:2 to iLet (n=107), standard care (n=54), and iLet with Fiasp (n=114).

We are eager to (hopefully) see sub-analyses comparing the results for subgroups based on the insulin delivery method used at baseline and in the standard care arm throughout the trial. Furthermore, we hope to see additional outcomes read out, including time above range, time below range, and time >250 mg/dL, which were not reported in today’s presentation, as Dr. Russell focused on the first four outcomes of the hierarchy analysis. Regardless, we’re excited to see these results, given that this system has been closely watched for a long time – for a little bit of additional context, see our coverage of Dr. Ed Damiano’s annual updates from Friends for Life in 2022, 2021, 2020, 2019, and 2018. We’ll be interested to see how the FDA views the data and the system once it begins the review process, the timeline of which has not been disclosed. Furthermore we’ll be watching to see what implications this holds for the bihormonal iLet system, the pivotal of which is now underway, with enrollment beginning in December 2021.

Before we dive into the results, here is a reminder on the features of the insulin-only iLet system. The iLet system uses an in-house tubed pump the “size of a credit card,” which houses the algorithm and a Dexcom G6. The system requires very little user input: (i) only patient weight is needed for initiation (no dosing parameters); (ii) the only setting to adjust is the glucose target (usual, lower, higher), which can be set for different times of day; and (iii) mealtime announcements do not require carb counting (only breakfast/lunch/dinner and categorical meal size [i.e., “more,” “usual for me,” “less,” or “much less”).

Photo from FFL 2019, from left to right: Gen 4 iLet (used in the pivotal study), Gen 3 iLet, Gen 2 iLet, and iPhone X

Finally, we highlight a few differences in the methodology of the pivotal compared to other AID pivotal trials. For one, this pivotal was an RCT like Tandem’s Control-IQ pivotal, whereas Medtronic’s MiniMed 780G and Insulet’s Omnipod 5 pivotal trials were single-arm studies (i.e., before AID vs. after AID). Second, the study compared participants in iLet to participants in the standard care arm using whatever insulin delivery method they used at baseline and included a baseline cohort almost equally split between MDI, pump, and AID. In comparison, the MiniMed 780G pivotal compared MiniMed 780G to sensor-augmented pump (SAP) or MiniMed 670G, the Control-IQ pivotal compared Control-IQ to SAP, and the Omnipod 5 pivotal compared Omnipod 5 to baseline (SAP or MDI). Third, the study included a sample that is far more representative than the usual AID study. Over a fourth of participants were non-Hispanic Black (10%), Hispanic or Latinx (10%), or another non-White race (6%). While still a minority of the sample, the diversity is far greater than in other AID pivotal trials, the populations of which have been almost entirely non-Hispanic White. This racial and ethnic diversity was intentional on the part of the researchers, who aimed for ≥15% participants of minority race/ethnicity. Likewise, the study worked to be more representative of a wider range of A1c values, setting no upper limit for baseline A1c inclusion and aiming for ≤20% A1c <7% (not quite achieved, as 25% of participants had A1cs <7%) and ≥33% with an A1c >8%.

Detailed Results

• Those on iLet saw a +2.6 hour/day Time in Range improvement relative to the standard care group when adjusted for baseline (p<0.001), “almost all” of which occurred within the first day or two of wear. Specifically, those on iLet saw their Time in Range improve from 51% at baseline to 65% at thirteen weeks, whereas the standard care arm saw only a slight improvement to 54%. This Time in Range improvement in the iLet arm was already seen in the first four weeks of the study and was consistently maintained throughout the study. Dr. Russell argued that although the Time in Range at the endpoint is often discussed in evaluating AID systems, he believes that the change in Time in Range is a more apt measure by which to compare studies given the differences in study populations and methodology. He also noted that this analysis included the results for children, who are known to reach a lower Time in Range with other available AID systems as well, thereby lowering the average. While many would certainly agree with this interpretation, some expressed disappointment to see that, on average, participants still did not achieve the 70% Time in Range target, which has been achieved with other AID systems among adults in both pivotal studies and real-world studies, including those who have challenges in managing their diabetes (see more on this below). From our view, these results are impressive and we’ll continue to watch algorithms and how various groups do – again, we state, coming from this baseline is very impressive and we want to see more trials both start at this baseline and start at higher baselines – all of it is great!

• • Dr. Russell also broke out the Time in Range outcomes for children, for adults iLet without Fiasp, and for adults on iLet with Fiasp, all of whom saw Time in Range benefits with iLet. Children saw a 2.4 hour/day Time in Range improvement on iLet compared to those on standard care (p<0.001), from ~47% to 60% within the first four weeks, which was maintained out to week 13. The standard care arm’s Time in Range improved only from 48% to 50%. Among adults, those on iLet (without Fiasp) saw a 2.6 hour/day Time in Range improvement relative to those on standard care, consistent with the overall sample, and saw their Time in Range improve from ~56% at baseline to 69% at 13 weeks, an improvement observed within the first day. The standard care arm saw a slight improvement as well from ~53% to 58% at 13 weeks. Those using iLet with Fiasp saw a further improvement in Time in Range, achieving +3.1 hour/day Time in Range relative the control group, and improving from ~54% at baseline to 71% at 13 weeks, an improvement already achieved in the first four weeks of iLet use. Although Dr. Russell cautioned against looking at the raw Time in Range values achieved, we’d note that among adults, these Time in Range outcomes are lower than achieved with other AID systems (more on this below) and for those using iLet without Fiasp, are lower than the consensus target for Time in Range.
  • Although the specific figures were not read out, Dr. Russell did share that the iLet arm was superior to the standard care arm in time >180 mg/dL and time >250 mg/dL. While iLet was not statistically superior in terms of time <70 mg/dL, during Q&A, Dr. Russell noted that there was a nominal 0.1% decline in time <70 mg/dL among iLet users compared to standard care. This result meets expectations based on the pre-pivotal results, which did not find time <70 mg/dL to be superior with iLet.
• There was no significant between-group difference in time <54 mg/dl, which was already quite low at baseline in both groups (median 0.2% of time <54 mg/dl) and was maintained at 13 weeks, although the iLet group saw a nominal nonsignificant increase to 0.3% (p<0.001 for noninferiority). When separated out, adults and children both saw no significant difference in time <54 mg/dl in with iLet vs. standard care: adults saw a nonsignificant 0.02% difference (p=0.33) and children saw a nonsignificant -0.04% difference (p=0.24). For all time <54 mg/dl data read out, Dr. Russell reported the median value rather than the mean – we’d wonder if there would be a significantly different finding were the other reported.
• At 13 weeks, A1c favored on iLet vs. standard care by 0.5% (p<0.001 for baseline-adjusted mean difference; n=326, doesn’t include those on iLet with Fiasp). Specifically, those on iLet saw their A1c fall 0.6% from 7.9% at baseline to 7.3% at 13 weeks, while the standard care arm saw no change from baseline to 13 weeks, maintaining an A1c of 7.7%. The iLet group also saw a narrowing of its A1c distribution curve (shown below), indicating the elimination of high A1c values and a larger improvement for those with higher A1c values at baseline. Those with A1c values >7% on iLet (n=164) reported a 0.7% A1c improvement relative to their counterparts in the standard care arm (n=76) (p<0.001). This subgroup saw their A1c fall from an average 8.3% to 7.5% on iLet, as compared to falling only 0.1% from 8.2% to 8.1% on standard care.

• • The analysis that Dr. Russell presented also broke down the A1c data for children, adults, and the subgroup of adults on iLet with Fiasp, all of whom saw 0.5% relative improvements in A1c on iLet compared to standard care. Specifically, children saw a 0.5% baseline-adjusted relative improvement with iLet compared to standard care (p<0.001). In the pediatric iLet group, A1c fell from 8.1% to 7.5% while the standard care arm saw no change from 7.9% at baseline to 13 weeks. Likewise, adults (excluding those on Fiasp) saw a 0.5% A1c improvement when on iLet vs. standard care (p<0.001), with the iLet group seeing their A1c fall from 7.6% at baseline to 7.1% at 13 weeks while the control group declined only slight from 7.6% to 7.5%. Those on iLet with Fiasp saw a similar 0.5% A1c improvement relative those on standard care and saw their A1c fall from 7.8% to 7.1%.
• Based on the 24-hour mean glucose profile, a majority of the benefit of iLet came at night, as has been seen with other AID systems. Between 2 am and 8 am, the study saw a widening difference in mean glucose between the two groups, growing to a ~30 mg/dL difference at 6 am when those in the standard care arm were still at a mean of  ~170 mg/dL while those on iLet were far lower at ~140 mg/dL. For the vast majority of the day, those on iLet had a lower mean glucose than those on standard care (exception of around 2 pm-4 pm); however, the between-group difference was far lower. Overall, the iLet group also saw both a lower mean glucose and a tightening of the variability of mean glucose over the 24-hour period relative to the standard care arm, suggesting that those on iLet saw more consistent and lower glucose levels. These results were similar in both children and adults.

• The rate of severe hypoglycemia episodes was higher in the iLet group than in the control group, but not statistically significant (excluding Fiasp users; 18 vs. 11 events per 100 person-years; p=0.39). On an absolute basis, 10 severe hypoglycemic events occurred in the iLet group, while 3 events occurred in the standard care group (as a reminder, the iLet sample was about twice as large). There were no DKA events. Despite the same 2:1 randomization, the group on iLet with Fiasp did not see significantly more severe hypoglycemia events (3 vs. 2) and had a non-significantly lower event rate when normalized by person-years (10 vs. 14 events per 100 person-years; p=0.83). There were two DKA events in the iLet with Fiasp group (7 events per 100 person-years), both of which were attributed to infusion set failures.
  
  • Dr. Russell offered several notes on these data: (i) the study was not powered to detect a difference in severe hypoglycemia events (Dr. Russell noted that that would have required data from 5,000-6,000 participants over six months); (ii) he believes the greater event rates of severe hypoglycemia in both arms of this study reflect its representativeness; (iii) the total number of events in all groups was still lower than the average in the T1D Exchange registry, which has a more limiting definition of what counts as severe hypoglycemia; and (iv) none of the severe hypoglycemia events were due to device malfunction. Some said that these results led them to wonder whether the bihormonal system might improve upon these hypoglycemia results given that glucagon will be administered, hopefully preventing these severe hypoglycemia events – we’d certainly say we think they will, although even more, we think they’ll

Comparison with other AID pivotal trials

• As noted above, this iLet trial had a different study population, a different methodology (RCT), and different comparator group (any insulin delivery method) compared to other pivotal trials. Although Dr. Russell cautioned against comparing pivotal trials with different methodologies and in a later presentation, Dr. Boris Kovatchev (UVA) implored the audience never to compare CGM outcomes that use different sensors, we’ve collected a brief comparison of some of the key outcomes across AID system pivotal studies – see more on the other AID pivotal trials here. Based on the results summarized below, the Time in Range and A1c improvements seen with iLet appear to be in line with (or above) those seen with AID systems in other trials in adults.  While it may be that the results with children seem slightly more mixed than in other AID pivotals with an A1c improvement similar to that of other trials but about an hour less/day of a Time in Range improvement, we would seek the views of pediatric experts before making such a pronouncement definitively. Overall, we’re thrilled to see that another AID system option may well become available for people with diabetes soon in the US, particularly near-term, given its lower burden setup and bolus use, and longer-term, the opportunity to move hyperglycemia down more easily with the use of glucagon. We do also note we don’t yet have a sense of what degree if much at all the extra “hassle” that may be required with two hormones vs. one – we look forward to learning more from the field on this.

2. “It is time for a systematic approach to using CGM to adjust type 2 diabetes management in insulin using patients”: Dr. Rich Bergenstal presents novel C²GM Therapy Adjustment Guide

Dr. Rich Bergenstal (International Diabetes Center) took to the stage to deliver one of the most engaging presentations of ATTD 2022 calling for greater action based on CGM data and presenting his novel C²GM (CGM Clinician Guided Management) treatment algorithm. Dr. Bergenstal began his presentation recognizing the last ten years in the adoption of Time in Range – we commend Dr. Bergenstal on the impact he and IDC have had throughout this time. In this talk, he also asserted that some of the field is currently stuck at an “analyze” stage with many clinicians, especially those in primary care, struggling to act on CGM data to optimize glucose management.  Population level glycemic management as well as data from trials like MOBILE demonstrating a lack of CGM-driven clinical action certainly support this sentiment, although we’d say MOBILE was designed to show changes from behavior specifically and not really the impact from therapeutic changes – i.e., in some trials, the design effectively understated the changes that could be made. One could also argue, of course, that RCTs provide more motivation to many PWD through the more thorough care from clinicians helping in the trial.

Additionally, Dr. Bergenstal lamented current best practices in the treatment of type 2 diabetes such as “treat to target” and “fix fasting first,” both of which he expressed actually involve substantial time on the part of providers to walk through current treatment algorithms and calculations including checking for over-basalization. Moving on from current treatment, Dr. Bergenstal presented a novel paradigm for CGM-based management of type 2 diabetes focusing on Time in Range and Time Below Range. In this paradigm, providers walk through a three step process to: (i) determine if a patient has comorbidities where a GLP-1 or SGLT-2 should be considered; (ii) find the % Time in Range and % Time Below Range on the patient’s AGP and ask “is the Time in Range >70%” and “is the Time below Range <2%”?; and (iii) find the Time in Range/Time Below Range category in the treatment algorithm table that corresponds to the patient’s CGM data and adjust treatment as necessary. We love this – so action-oriented – and were absolutely thrilled to see it, since we also think it’s so easy that it can also be explained to patients and recommended that they tell their providers about it if they do not think their providers are working in this way. As providers are only using two metrics from a patient AGP, this creates four simple categories that patients will fall into and for which Dr. Bergenstal’s group has developed simple and specific treatment adjustments allowing providers (as well as patients who have health literacy) to take action quickly and easily to help improve glycemic management. The four categories and immediate goals for therapy adjustment are as follows:

• Time in Range >70% and Time Below Range <2%: Continue regiment by continuing to optimize current therapy and reinforce lifestyle changes and taking insulin as prescribed. Recommended follow-up in 3-4 months.
• Time in Range >70% and Time Below Range >2%: Address hypoglycemia and stop sulphonylurea use if present and reduce background insulin by 10% if Time Below Range is 8-12% or by 15% if Time Below Range is >12%. If the patient is not on a sulphonylurea, decrease the total background insulin dose by 10% if Time Below Range is 2-7%, by 15% if Time Below Range is 8-12%, and by 20% if Time Below Range is >12%. Recommended follow-up of 2 weeks.
• Time in Range ≤70% and Time Below Range <2%: Address hyperglycemia and consider adding or adjusting GLP-1 therapy, otherwise increase background insulin dose by 10% if Time in Range is 51-70%, by 15% if Time in Range is 30-50%, and by 20% if Time in Range is <30%. However, if the patient experiences nocturnal hypoglycemia, consider a smaller increase in insulin dose. Recommended follow-up of 2 weeks.
• Time in Range ≤70% and Time Below Range >2%: Address hypoglycemia immediately and consider referral to a diabetes educator or specialist and stop sulphonylurea use if present and reduce background insulin by 10% if Time Below Range is 8-12% or by 15% if Time Below Range is >12%. If the patient is not on a sulphonylurea, decrease the total background insulin dose by 10% if Time Below Range is 2-7%, by 15% if Time Below Range is 8-12%, and by 20% if Time Below Range is >12%. Additionally, the patient should be referred to a diabetes educator for options to treat hyperglycemia including the potential addition of a GLP-1 or mealtime insulin. Recommended follow-up of 2 weeks.

We see the simplicity of this system as a major, major win for providers with busy schedules (i.e., so many!) or who may be currently less familiar with diabetes treatment, allowing them greater time to learn from patients and engage in collaborative decision-making. Additionally, we imagine the breakdown of recommendations by patient Time in Range will be incredibly beneficial for providers by doing much of the necessary data interpretation for them.

• Dr. Bergenstal’s group came to the decision to use ≤2% Time Below Range as the threshold for clinical adjustment category based on analysis of participant AGPs from the MOBILE study. While this is certainly lower than the Time in Range consensus target for <4% Time Below Range, Dr. Bergenstal explained that he and his collaborators evaluated every available AGP from MOBILE study participants and arranged them based on Time Below Range which varied from 0% to 9%. Of these patients, Dr. Bergenstal and his group came to consensus on whether or not they thought it would be safe to intensify treatment for each percentage increase in Time Below Range ultimately deciding that they were only comfortable intensifying treatment for those with Time Below Range <2%.
• Dr. Thomas Martens (International Diabetes Center) presented case study applications for the C²GM treatment algorithm using participants from the MOBILE study. Dr. Martens created four composite AGP profiles of MOBILE participants such that one composite AGP profile fell into each treatment adjustment category. The first composite profile (n=20) demonstrated a Time in Range of 80% and Time Below Range of 0% leading Dr. Martens to explain that adjustments to therapies were not needed and that providers could comfortably and safely follow-up with these patients 3-4 months later. The second composite profile (n=5) had a Time in Range of 76% and a Time Below Range of 5% for which Dr. Martens recommended immediately addressing hypoglycemia and discontinuing the use of sulphonylureas if present. Importantly, Dr. Martens encouraged providers to follow-up with these patients two weeks later in case these changes impact Time in Range, which could then be addressed in a subsequent appointment. The third composite profile (n=139) demonstrated a Time in Range of 34% and a Time Below Range of 0% and represented the majority of the patients included in this analysis. For this profile, Dr. Martens identified addressing hyperglycemia as the most important treatment goal and advised adding or adjusting GLP-1 therapy or increasing background insulin. Finally, the fourth composite profile (n=12) demonstrated a Time in Range of 51% and a Time Below Range of 5%. Since these patients struggled with both hyper and hypoglycemia, Dr. Martens advocated for a coordinated care-team approach with referral from primary care to a diabetes educator or other specialist. However, in the immediate, Dr. Martens also stressed the importance of addressing hypoglycemia and taking these patients off any sulphonylureas and reducing background insulin to prevent any potentially life-threatening hypoglycemic episodes.

3. Abundance of MiniMed 780G real-world evidence: Large cohort (n=25,396) achieves 74% Time in Range and 6.8% GMI; crossover study (n=6,299) boasts +2.7 hours/day Time in Range to 74%; longitudinal cohort (n=9,119) sustains glycemic improvements over six months; Dr. Robert Vigersky highlights potential for clinical inertia around changing AID system settings

Immediately following the Beta Bionics iLet pivotal readout, Dr. Robert Vigersky (Medtronic Diabetes) presented three illuminating real-world studies from MiniMed 780G users. Dr. Vigersky spoke in place of Professor Ohad Cohen, who was originally scheduled to present and who also read out three similarly-designed studies at EASD 2021. We got our first look at real-world data (n=4,120) from MiniMed 780G users one year ago at ATTD 2021 after the system’s OUS launch in October 2020 and CE-Marking in June 2020. In the US, MiniMed 780G is still “under active review” with the FDA after being submitted in February 2021, although this timing is highly subject to the company’s warning letter from the FDA. The studies presented by Dr. Vigersky were compiled from Medtronic CareLink data that was uploaded between August 2020 and January 2022 by users “who provided their consent for data to be aggregated” in countries where local data privacy regulation permits data analysis.

• The entire population of MiniMed 780G users in the study (n=25,396) achieved a 74% Time in Range and 6.8% GMI. These results are extraordinary and add to the already strong data we’ve seen from the system at ATTD 2022, including from two RCTs in the ePoster Hall, a readout from the GIF study, and a readout from the extension phase of the 780G pivotal trial. Notably, 70% of users achieved a GMI <7%, 70% achieved a Time in Range >70%, and 84% saw a Time Below Range (<70 mg/dl) of <4%. Putting these together, 66% of participants achieved both a TIR >70% and a GMI <7%, and 54% managed to secure all three: GMI <7%, Time in Range >70%, and Time Below Range (<70 mg/dl) <4%. Dr. Vigersky also broke down the participants by country of residence, and we’ve included this data in a picture below.

• Dr. Vigersky then presented a real-world crossover study (n=6,299), demonstrating that users who initiate MiniMed 780G witness a +2.7 hour/day improvement in Time in Range (+11%). Additionally, participants’ GMI fell by 0.3% over the course of the study from 7.2% to 6.9%, and Time Below Range improved by -6 minutes/day from 2.7% to 2.3%. At baseline, 37% of participants achieved a GMI <7% and 33% achieved a Time in Range >70%, and these figures improved to 70% achieving a GMI <7% and 71% achieving a Time in Range >70% after MiniMed 780G initiation.

• Last, Dr. Vigersky shared data from a longitudinal cohort (n=9,119) showing that the glycemic improvements are observed in the first month after initiating MiniMed 780G and sustained over six months. We are certainly encouraged by these results, as they reinforce the data from the prior two studies and fortify the strong real-world MiniMed 780G outcomes presented at EASD 2021. By the end of the six months, 73% of users attained a GMI <7%, and 73% of users had a Time in Range >70%.

• Importantly, Dr. Vigersky drew attention to clinical inertia around changing AID system settings. In two separate studies of MiniMed 780G users who were not meeting clinical targets for GMI, Time Below Range, and Time Above Range, showing that these users were often using system settings that have been associated with a suboptimal glycemic management. This data further reinforces the sub-analysis of a MiniMed 780G real-world study (n=12,780) that was presented at EASD 2021, showing that even though a lower active insulin time and glucose target correlate with higher Time in Range, only 12% of participants used most aggressive system settings. Of course, we understand that there are many valid reasons for wanting a higher target and a less aggressive active insulin time, especially when it comes to preventing hypoglycemia. Nonetheless, we appreciated Dr. Vigersky’s point that these data highlight a huge opportunity for providers to consider subsets of patients for whom AID system setting adjustments could drive stronger glycemic outcomes. We wonder if Medtronic is thinking about creating a software solution that could identify these patients and provide AID setting adjustment recommendations to HCPs, given the documented association of more aggressive settings with better outcomes.

• In our first look at real-world head-to-head data comparing Guardian 4 and Guardian Sensor 3 in MiniMed 780G, Dr. Vigersky shared that users’ Time in Range, GMI, and Time Below Range were sustained over four weeks of transitioning from GS3 to GS4. This data strongly corroborates Medtronic’s ATTD 2022 ePoster data with extension phase data from the MiniMed 780G pivotal trial, showing that the Guardian 4 CGM demonstrates sustained Time in Range and A1c improvements with an average Time in Range of 73% and an average A1c of 7.1% in the three months after switching from Guardian Sensor 3.

4. Real-world Control-IQ data shows massive ~0.8% A1c reduction for MDI users nine months after switching to Control-IQ from 7.9% to 7.1% (n=426); improvements in both pediatrics and adults

During a potpourri of Saturday morning presentation, Dr. Jordan Pinsker (Tandem) presented a first look at nine-month results from the CLIO trial. To start, Dr. Pinsker described CLIO as a “truly real-world use trial,” in contrast to conventional post-market surveillance studies, which usually run through specialized clinical study sites, Tandem recruited participants for its CLIO study by simply sending emails to all customers who purchased a Tandem t:slim X2 pump with Control-IQ. As a reminder, Tandem first began launch for Control-IQ in January 2020. In the email, customers were able to opt-in to the CLIO study and are simply asked to fill out surveys at baseline, 3-, 6-, 9-, and 12-months, while receiving regular care throughout the 12-month study. As a reminder, we’ve seen three-month data from the CLIO study in adults at ATTD 2021, a sub-analysis by racial and ethnic groups at ADA 2021, three-month quality of life data at Keystone 2021, and three-month pediatric data at ISPAD 2021. While this presentation’s results were limited to A1c (and GMI) outcomes, we look forward to Tandem’s future presentations on CGM metrics and other outcomes.

• Results from the cohort of Control-IQ users who switched directly from MDI were particularly impressive with a median A1c reduction from 7.9% to 7.1% after nine months (n=426). This real-world data continues to support the excellent usability of Tandem’s and Dexcom’s devices in the real-world, with no prior pump experience required to see benefits from using Control-IQ. We believe many would be interested in seeing a sub-analysis of users who were CGM-naïve prior to using Control-IQ. For the pump users in the CLIO study, A1c fell from 7.3% at baseline to 7.1% at nine months (n=1,487). Note that baseline figures are lab-measured A1c, while the nine-month outcome is the CGM-derived GMI, meant to provide an estimate for A1c based on CGM-measured mean glucose.

• Excitingly, Control-IQ was shown to be effective at improving A1c in every age group and baseline insulin delivery method. As shown in the table below, in adults, after nine months on Control-IQ, median A1c was universally in the 6.9% to 7.1% range regardless of age and prior insulin delivery method. In particular, we were excited to see this result for the (admittedly small) sample of elderly patients previously on MDI. This is a population where some have the greatest concerns about diabetes technology uptake, but the results from Tandem suggest it can be done successfully. Additionally, Control-IQ delivered significant improvements to A1c in the pediatric group, who generally started and ended with higher A1cs at baseline and nine months.

5. Dr. Irl Hirsch argues A1c, while “critical” for population insights, can be “problematic” and “dangerous” at individual level; lays out vision for future coexistence of A1c and Time in Range; champions CGM for prediabetes diagnosis and management; EDIC post-analysis of adult type 1s (n=765) shows considerable discordance between TIR and A1c

In one of ATTD 2022’s final sessions, Dr. Irl Hirsch (University of Washington) said that while A1c remains “critical” for population-level insights, it can be “problematic” and “dangerous” at an individual level. Despite being scheduled at the same time as several other impactful sessions, Dr. Hirsch’s presentation drew a huge (!) audience. From our view, Dr. Hirsch’s commentary on the “Beyond A1c” movement marks an evolution from his presentation nine months ago at Keystone 2021. At that time, while he still highlighted several limitations of A1c and noted the “crucial” aspects of GMI and Time in Range for clinical decision making, we perceived a new conviction behind Dr. Hirsch’s confidence around CGM-derived metrics as being much more informative measures of glucose control at the individual level.

Today, Dr. Hirsch noted that during the pandemic, HCPs “had no problems” using GMI instead of A1c due to limitations around point-of-care testing, and that most people have no problems using CGM-based metrics for diabetes management. As Dr. Hirsch said: “In our world, we feel the additional granularity makes [A1c] obsolete.” Still, Dr. Hirsch cautioned that it is “unrealistic” to think that every person with diabetes, let alone person with prediabetes, will have access to CGM in every area of the world. The reality is, Dr. Hirsch argued, that “fingerstick glucose monitoring and [A1c] will be around for a long time,” and that “until there is consensus with regulatory agencies that GMI, [Time in Range], and [Time Below Range] should be primary endpoints, it will be difficult to use only CGM for new drug applications.” Of course, we have already seen movement on this front, with an International Consensus group gathering at ATTD 2022 to discuss the standardization of CGM use in clinical trials.

• “So why do we need A1c?” Dr. Hirsch argued that A1c continues to be the only glycemic metric providing a validated assessment of complications and glucose control, and that it is also the most accessible and relatively cheap measure available “for most populations.” As for why GMI should replace A1c, Dr. Hirsch said that current and future CGMs are a more accurate assessment of mean glucose, and that they can provide more detailed information through Time in Range and %CV. 
• Dr. Hirsch laid out a vision for the coexistence of A1c and CGM-derived metrics, as seen in the picture below. The “most important” point of this diagram, according to Dr. Hirsch, is that he firmly believes in prediabetes management and diagnosis through CGM metrics, given the strong body of evidence pointing to GMI-A1c discordance at the individual level. We were fascinated by his opinions. While some would argue that A1c is a sufficient means to diagnose prediabetes, of course we think CGM would be far better, particularly in that over time, using CGM would make it far easier to see who is hurtling toward a diagnosis but don’t have it yet vs. those who are primarily “steady” over time. Ultimately, we believe the field is become more and more clear  that CGM can offer anyone a wealth of knowledge to understand how their lifestyle affects metabolic health - we were absolutely thrilled to hear Dr. Hirsch’s take on this subject, given the enormous importance of pre-diabetes and Dr. Hirsch’s persuasive talk on what actions will bring fare more focus to this arena. We see steps reducing risk of pre-diabetes as similar to steps being taken currently by many to reduce risks of COVID-19. We hear all the time from people with diabetes that they fear getting long COVID-19, and so work very hard not to get COVID-19 by wearing masks everywhere and taking every precaution to reduce the risk of COVID-19, even though multiple geographies have relaxed recommendations. Similarly, many could begin even more focus on reducing risk of pre-diabetes in order, similarly, to reduce the risk of T2D. While this is not perhaps a perfect metaphor, we note that COVID-19 itself is actually less likely in those without diabetes, so more may be able to achieve two aims at once by working to prevent pre-diabetes (and effectively T2D) as well as COVID-19 (and effectively long COVID).

• • Dr. Hirsch also laid out five requirements to create a “different conclusion” than the one he laid out above over the next five years. According to Dr. Hirsch, CGM needs: (i) better accuracy, especially in the hypoglycemic ranges; (ii) to be more affordable; (iii) better uptake with regulatory agencies; (iv) better uptake with non-diabetologists; and (v) ideally, to be noninvasive. Dr. Hirsch has praised G-WAVE historically – see diaTribe’s “Wave of the Future: New Glucose Technology Could Revolutionize Care.” From our view, we think given arrows, accuracy is pretty strong already; we also think regulatory agencies have been phenomenally supportive of CGM over time, though this has faded in the age of COVID-19, when some of the regulatory resources have certainly gone elsewhere. We’d say that beyond non-diabetologists, many diabetologists themselves could be stronger on uptake – see dQ&A data in people with T2D on insulin for more on this front. To boot, while penetration in T1D has risen at a fast pace in the US and some EU countries, that is not the case in most countries globally, endocrinologist or not. We’d also assert that multiple professional organizations representing PCPs haven’t seemed to have been as enthusiastic as most, even claiming that people with diabetes “can’t understand the complexity” – additionally, we’ve noticed some geriatric experts claiming words to the effect of “all that technology is too complicated, especially for families.” This is obviously absurd, although the cost of course is a major area of unmet need. While volume could take care of some costs moving down, we’d like to see this as a bigger area of focus, and/or we’d like to see manufactures focus more on professional CGM, where CGM could at least be used to “correct” therapy, even if not used more than, say, quarterly, or even once or twice a year to make sure therapy continues to be optimized as much as possible. Regarding non-invasiveness, we’d love to better understand the goals here – if it is that the devices could be less expensive, that’s a solid goal, though if accuracy isn’t as high, we wonder about regulatory enthusiasm, since they may be products that FDA or others could worry about being used by those on insulin where lower accuracy could be problematic. Regarding “ease of use,” we wonder the degree to which current products are effectively seen as nearly non-invasive, since virtually none of the products in this day and age appear to have painful insertions like some did back in the day. Stay tuned – we are excited to have heard this provocative talk though where we land on it is that it’s very critical given how much better many PWD feel, especially those on insulin, who have any sort of continuous system. We remember terming early CGM systems such as Dexcom’s STS, circa 2006, that weren’t that accurate (or reliable, or particularly easy to use relative to today) as quite amazing – they were continuous, which was such a major change! It’s true that was over 15 years ago – we’re staying very tuned to all the improved benefits and features of course and have our minds focused on what can drop prices similar to other new technology like smart phones that have dropped in price. Of course, we also note – six billion smart phones are estimated to be in use today, compared to about six million CGM – we’ve got a long way to go.
• Dr. Hirsch dove into a recently published EDIC post-analysis of adult type 1s with a diabetes duration >35 years (n=765), showing a considerable discordance between Time in Range and A1c. Participants had a mean age of 60, mean diabetes duration of 37 years, mean A1c of 7.8%, and used the FreeStyle Libre Pro for 11.9 days. Dr. Hirsch presented a figure showing a plot of A1c vs. Time in Range matched pairs, and noted that while the trend line reflects the conventional wisdom that a 7% A1c correlates with a 70% Time in Range, there is substantial variation within the sample. Plus, this variation is not unidirectional, meaning that there were plenty of people at a given Time in Range with lower-than-expected and higher-than-expected A1cs than the 7% A1c-70% Time in Range correlation would imply. This paper strongly reflects Dr. Hirsch’s argument that while A1c can be a useful population-based metric, it is incapable of capturing the glycemic profile at an individual level.

• On the validation front, Dr. Hirsch highlighted a table from a recent publication in the Journal of Clinical Endocrinology and Metabolism pointing to the association of Time in Range as a marker of long-term complications. In a sample of n=515 type 1s over two years, individuals with microvascular complications spent one month less Time in Range, on average, compared to those without microvascular complications (p=0.022). Although Time in Range was not associated with a risk of macrovascular complications, it was the only independent risk factor for hospitalizations for hypoglycemia or DKA.

• To illustrate A1c’s inability to serve as a glycemic measure at the individual level, Dr. Hirsch presented a graph of severe hypoglycemia (SH) rates in people over 65 years old. Concerningly, there was a huge rise in SH rates following the ADA and National Committee for Quality Assurance’s (NCQA) guideline updates recommending a treatment goal of A1c <7%. Rates of hypoglycemia only decreased after these guidelines were updated to recommend minimizing hypoglycemia. While attaining an A1c <7% has been rigorously proven to minimize one’s risk of developing diabetes-related complications, Dr. Hirsch noted that this example further reinforces that A1c is most useful when thinking about populations, not individuals. From our view, this figure also reinforces the impact of NCQA guidelines on clinician accountability, and has us thinking of the recently published and Helmsley Charitable Trust-sponsored NCQA white paper, championing CGM-derived metrics and psychosocial outcomes as quality measures of diabetes care.

6. Patient-reported outcomes (PROs) from MOBILE study demonstrate increase in “openness” measure of feeling freer from constraints of diabetes management among participants in CGM arm compared to BGM at eight months

Dr. David Price (VP Medical Affairs, Dexcom) presented patient reported outcomes from the landmark MOBILE study that was read out last year at ATTD 2021 and simultaneously published in JAMA. As a reminder, MOBILE enrolled 165 people with type 2 diabetes who were randomized to initiate CGM (n=165) while the control group (n=57) used BGM. At baseline these two groups had average participant ages of 56 and 58 years old, respectively, and the majority of participants identified as racial or ethnic minorities, did not have a college degree, and were on public insurance, which Dr. Price explained is relatively representative of the US population of people with type 2 diabetes. Turning to patient reported outcomes, participants in the MOBILE study took the Glucose Monitoring Satisfaction Survey (GMSS), developed by behavioral psychology expert Dr. William Polonsky, at both baseline and eight months and saw improvements in “openness,” emotional burden, and behavioral burden. Interestingly, there was no significant difference in emotional or behavioral burden between CGM and BGM users at eight months, though both groups did see a in improvement in both metrics across the study period. Discussing the lack of between group difference, Dr. Price hypothesized that this may have been due to the fact that BGM users were still enrolled in a clinical trial during which their data was reviewed by providers and used to adjust diabetes management, likely allowing patients to feel more supported in their diabetes management, which could certainly result in a reduction of both emotional and behavioral burden. Conversely to emotional and behavioral burden, there was a significant difference in the “openness” experienced by participants on CGM versus BGM, with those in the CGM arm feeling significantly more freedom from constant diabetes management after eight months (p=0.003). 

• The GMSS “openness” component assessed patient perspectives on the following four statements: (i) this tool helps me feel more satisfied with how things are doing with my diabetes; (ii) this tool helps me feel less restricted by diabetes; (iii) this tool helps me be more spontaneous in my life; (iv) this tool helps me be more open to new experiences in life. According to Dr. Price, the continued evaluation of patient reported outcomes is a key aspect of research on diabetes technology and Dr. Price also referenced the GMSS developer, Dr. Polonsky, and his view that transitioning toward assessment of “openness” should be incorporated into PRO analysis that has, up until this point, focused largely on diabetes burden. While we see openness and burden as related – with openness reflecting a lack of burden and burden recognizing challenges people with diabetes face – we see this focus on openness as a positive interpretation of data that is more often presented in the negative. Specifically, the focus on openness versus burden is reminiscent of Dr. Diana Isaacs’ DATAA model for diabetes device data interpretation that emphasizes highlighting days and times patients are doing well over times patients may have struggled with glycemic management suing positive reinforcement to help drive behavioral change.

7. Future of nutrition management is bright: Innovative technology like Rocket AP, creative AI like Medtronic’s Klue, and multihormone therapy are potential solutions to help reduce burden

The esteemed Dr. Bruce Buckingham gave a compelling talk highlighting the future of nutrition management with closed loop insulin delivery. Dr. Buckingham began by framing the issue: 65% of patients miss more than one meal bolus/week, which is understandable given the realities of living with a chronic disease, but unfortunately leads to worse glycemic control. Although a truly closed loop system remains the goal and existing hybrid closed loop has reduced this burden to some extent, all available systems still require some sort of meal announcement. Late or missed meal boluses on AID tend to increase prandial hyperglycemia and can increase the risk for late post-prandial hypoglycemia, which is most common when given >1 hour after a meal or when blood sugar is >200 mg/dL. However, there are a number of algorithms in development that are intended to improve meal-time detection, including RocketAP, the algorithm under development with the UVA researchers who developed the Control-IQ algorithm. This algorithm builds on Control-IQ with a new bolus priming system module that is designed to detect unannounced meals quickly and deliver bolus insulin before a lengthy hyperglycemia episode begins. Recall that a study from UVA presented at ATTD 2021 and published in November 2021 showed highly promising results on the algorithm: RocketAP delivered +30% Time in Range over Control-IQ during the six-hour period following unannounced dinner (n=18). Dr. Buckingham also highlighted alternative innovations, like Medtronic’s Klue, a “gesture-sensing” and AI software driven by data from Apple Watch to detect when and how fast someone is eating or drinking, which could help reduce the burden of mealtime bolusing.

• Dr. Buckingham also discussed the potential of insulin co-formulations and multihormone therapy to reduce post-prandial excursions.  In particular, he highlighted a 2020 paper that found insulin+pramlintide increased time in range from 74% to 84% with a major flattening of post prandial glucose. Two additional studies with pramlintide (Majdpour Meng 2021 and Andersen 2021) confirmed these results. Interestingly, a 2020 study randomizing patients on full closed loop to dapagliflozin or placebo found that dapagliflozin stabilized nighttime glucose excursions and significantly increased Time in Range following two unannounced meals.
• We can’t help but wonder if there is greater opportunity with ultra-rapid acting insulins like Lyumjev to help reduce the burden of late boluses, particularly given that Lyumjev, approved for use by the FDA in July 2020, carries an additional indication for post-meal dosing that allows for bolusing in the first 20 minutes of the meal. The greater flexibility to dose during mealtime leads to improvements in post-prandial glucose excursions compared to other rapid acting insulins.

8. Non-adjunctive inpatient ICU CGM use during COVID-19 enabled by device validation; CGM enabled 72% reduction in POC tests with 0% Time Below Range by Day 2 of sensor wear

Dr. Eileen Faulds (The Ohio State University) presented data from inpatient use of Dexcom G6 during the COVID-19 pandemic demonstrating that CGM was used safely and effectively to help providers manage critically ill patients with diabetes. Specifically, among 50 patients in the medical ICU (92% on ventilatory support, 46% on a vasopressor, 34% on dialysis, and 74% on steroids), sensor MARD ranged from 8.0%-15.3% depending on a number of factors including oxygen saturation, pH, blood pressure, and partial oxygen pressure. Remarkably, Dr. Faulds shared that despite these factors, patients had an average Time in Range while in the ICU of 72% with no recorded hypoglycemia after day #2 of sensor wear. Additionally, Dr. Faulds shared that because nursing staff were able to use CGM non-adjunctively following sensor validation (more on this below), the frequency of point-of-care testing decreased 72% from 24 tests/day prior to CGM use to an average of 10 tests/day on Day 1 of ICU hospitalization and 7 tests/day on Day 2 of ICU hospitalization reducing the amount of PPE needed by nursing staff and reducing the time nurses had to spend collecting blood glucose data. Based on these data, Dr. Faulds explained that CGM allowed nurses to maintain inpatient glycemic management without the need for frequent point-of-care tests, and also shared her view that this is only the beginning of CGM use for inpatient care. As Dr. Faulds expressed, during the COVID-19 pandemic, CGM technology has enabled nurses to get the same necessary glucose data from their patients, but now via a remote technology. However, according to Dr. Faults, this “masks” the potential of CGM to be a revolutionary technology for inpatient management by utilizing information on trends and alerts to better inform care.

• At The Ohio State University, Dr. Faulds worked with nursing leadership, the diabetes consult team, internal medicine providers, and hospital administrators and develop a safe and effective protocol for implementing CGM in the hospital including the use of device validation. More specifically, for hospitalized patients in the ICU on insulin, nurses would apply a G6 sensor and perform one point-of-care test after the sensor’s warm-up period. If the glucose values from both the G6 and the point-of-care test fell within the +/- 20/20 bounds, then the sensor was considered “validated” and the nursing staff could continue to use the sensor non-adjunctively to inform insulin delivery. Nursing staff were asked for validate individual sensors every six hours with a point-of-care test, and if a sensor failed to be validated, nursing staff could wait an hour or two before trying again to validate the sensor and continue with non-adjunctive care. Of note, 67% of sensors were validated based on the first glucose reading they produced, indicating strong early accuracy. To enact this protocol, Dr. Faulds worked with nursing leadership to create a scaffolded education structure wherein a member of the diabetes consult team would teach nursing leadership how to insert and use a CGM as well as interpret the data. Nursing leadership would then place CGMs and train medical ICU staff who would subsequently train each other at any shift changes. While this was the initial education process, Dr. Faulds shared that nurses wanted to take ownership over CGM management for their patients and CGM training is now incorporated into medical ICU annual competencies for nursing staff. We see this as a big win for inpatient CGM use as nursing staff are absolutely critical for managing glycemic levels among hospitalized patients and this level of buy-in is encouraging and absolutely necessary for the sustained use of CGM. Indeed, we await glucose as the sixth vital sign and hope the field doesn’t have to wait too long.
• Discussing the benefits and limitations of inpatient CGM use, Dr. Faulds highlighted the value of a continuous data stream, but recognized there have been some concerns around accuracy. Starting with the potential limitations of inpatient CGM use, Dr. Faulds noted: (i) outstanding questions on the accuracy of CGM among hospitalized patients as well as potential interferents; (ii) the lag time between interstitial and blood glucose levels and any potential treatment implications; (iii) staff unfamiliarity with the technology; (iv) lack of EHR integration; and (v) cost. However, Dr. Faulds also noted numerous benefits of inpatient CGM use including: (i) continuous real-time data stream; (ii) data-enabled predictive capabilities; (iii) threshold and predictive alerts and alarms; (iv) takes less time for hospital staff; and (v) only includes one invasive procedure compared to hourly fingersticks.

9. Update on the International Consensus for AID: 40+ KOLs to recommend that AID is “considered” for and made available to for all type 1s; implores all payers to cover or reimburse AID for type 1 diabetes

Rounding out this week of consensus, a Saturday morning session featured an update on yet another consensus report, this time on AID technology. The update was offered by Dr. Moshe Phillip (Schneider Children's Medical Center, Israel), Dr. Revital Nimri (DreaMed), and Dr. Thomas Danne (Auf der Bult Hospital, Germany), who reviewed the consensus meeting that was held at last year’s virtual ATTD sessions and unveiled some of the takeaways of the consensus report that came out of the meeting. The meeting included about 40 KOLs from a dozen countries, a notably greater proportion of whom were women than in the consensus meeting on CGM metrics in clinical trials. The consensus report has seven keys aims: (i) to agree on the indications for use based on clinical studies; (ii) to give recommendations for initiating AID use; (iii) to determine best practices in education, training, and follow-up; (iv) to give clinical guidelines for treatment; (v) to recommend metrics for and presentation of reporting AID data; (vi) to give recommendations related to psychological burden and behavioral challenges; and (vii) to determine the remaining needs and the future of AID systems. To meet these ambitious goals, the panel met virtually in May 2021 and split into nine working groups, each of which focused on one topic within which to discuss existing literature and provide evidence-based recommendations using the ADA’s evidence-based grading system. The nine subgroups covered the introduction; the summary of clinical evidence; the target populations; the initiation of AID use; education, training, and support; clinical recommendations for AID use; data reporting; psychosocial issues and the perspectives of people with diabetes; and the future of AID. The subgroups them presented to the full group for discussion, after which final report recommendations were reached, although these are yet to be published. Below we summarize our key takeaways from this morning’s session.

• The consensus statement will recommend that AID is “considered” for all people with type 1 diabetes. In particular, the group felt that it is hugely important for those facing challenges with general glycemic management, hypoglycemia, and/or significant glycemic variability. They also noted that AID is particularly useful for those with significant hypoglycemia unawareness and frequent or severe episodes. As part of the recommendation that AID should be considered for all people with diabetes, the authors were sure to point out the challenge to overcome racial/ethnic and social inequities driven by unfounded provider biases around who is best suited to AID technology.
• The group felt that there was strong enough evidence to support the use of AID in type 1 adults, adolescents, and school-age children (ages 7-14), with the most efficacy seen in adolescents, those with high A1cs and those on MDI.  However, they felt that there was insufficient evidence on preschool-age children and older adults (ages >65) to currently recommend AID systems’ use in those populations. That said, Dr. Nimri noted that the evidence continues to accumulate, meaning that there very well may come a day where the evidence supports the use of AID in very young and older people with diabetes. Dr. Nimri also highlighted the need for further research into AID use in pregnancy with type 1, in those with type 2 diabetes on basal-bolus therapy, in those with comorbidities (e.g., renal failure), in people with significant hypoglycemia and hyperglycemia, and in people who are not non-Hispanic White. She suggested that further research into AID in these populations could drive an update to the consensus report to include recommendations for use in these populations. Likewise, the consensus group concluded that there are not enough data to conclude that early initiation of AID will preserve beta cell function but do believe that there are “likely benefits” on long-term glycemic control.
• The group also tackled the accessibility front, strongly recommending that AID systems are made available to all people with type 1 diabetes. As a part of this, they recommended that “all the payers (government and private) should reimburse/cover the AID systems along with initial and ongoing AID education and support for the management of T1D.” This is based on evidence that has repeatedly shown that AID provides greater Time in Range improvements than any other current technology. We were pleased to see much discussion around and a direct recommendation of the need for structured training and onboarding, as well as ongoing support, something that Dr. Laurel Messer has reminded us of repeatedly, including at a particularly strong presentation at ATTD 2021.

10. Real-world patient-reported outcomes from the CLIO study find significant improvements in quality of life, diabetes burden, and device satisfaction after six months on Control-IQ

Dr. Harsimran Singh (Tandem) presented new six-month patient-reported outcomes (PROs) from the Control-IQ Observational (CLIO) study. As a reminder, the CLIO study is a real-world post-marketing study on the efficacy, safety, and quality of life impact of Control-IQ. From the CLIO study, we’ve previously seen three-month glycemic control and quality of life data in adults at ATTD 2021, more in-depth three-month quality of life data at Keystone 2021, a sub-analysis by racial and ethnic groups at ADA 2021, three-month pediatric data at ISPAD 2021, and three-month adverse event rates at DTM 2021.

Dr. Singh shared six-month results on two aspects of psychosocial health: (i) diabetes burden through the Diabetes Impact and Device Satisfaction (DIDS) scale and (ii) diabetes quality of life through the Impact of Diabetes Profile (DIDP) scale. Of the 2,062 participants, 63% were previously on an insulin pump, 36% were previously on MDI, 10% were CGM naïve, and 23% had a baseline A1c ≥8.5%. Notably, the significant benefits of Control-IQ on diabetes impacts, device satisfaction, and quality of life were consistent across previously therapy (pump vs. MDI), baseline A1c (A1c <8.5% vs. A1c ≥8.5%), and age (pediatric vs. adult participants).

• Based on the DIDS scale, participants reported a significant 33% reduction in the impact of diabetes on their overall wellbeing from a score of 4.66 to 3.12 (p<0.001). Dr. Singh highlighted that participants overwhelmingly reported improvement in sleep quality, with a 35% reduction in poor sleep quality due to diabetes. She explained this improvement by pointing out that participants were also significantly less likely to wake up at night to treat low blood glucose. Dr. Singh also expressed enthusiasm about participants’ significantly reduced worry about hypoglycemia and reduced likelihood of missing work or school due to diabetes.
  
  • In terms of device satisfaction, participants reported a 23% increase in device satisfaction from a score of 7.2 to 8.8 (p<0.001). Participants felt significantly more in control of their diabetes and satisfied with their glycemic control. They also reported feeling their device was easier to use than their prior therapy.

• Based on the DIDP scale, participants reported a significant reduction in the burden associated with diabetes management. Dr. Singh said she was most proud of the improvement patients saw in their “freedom to eat as they desired.” She explained that the freedom to eat as desired is the most impacted aspect of qualify of life, irrespective of whether people have type 1 or type 2 diabetes, and this freedom influences many other aspects of quality of life. Participants also reported significantly improved physical health, financial situation, relationships with family/friends/peers, leisure activities, work and studies, and emotional well-being.

11. Dr. Simon Heller advocates use of diabetes technology in combination with structured education to reduce hypoglycemia incidence and improve hypoglycemia awareness

The esteemed Dr. Simon Heller (University of Sheffield) advocated for a combination approach to improving impaired awareness of hypoglycemia through the use of diabetes technology alongside structured education. Dr. Heller argued that hypoglycemia remains a key barrier to improved glycemic control, citing a study that found an average of 2.4 severe hypoglycemia events per person-year among people with type 1 and type 2 diabetes in North America. Over one-third of people with diabetes (37%) experienced at least one severe hypoglycemia event per year, with this number increasing to 51% in those with type 1 diabetes. Older people with diabetes and those with longer diabetes duration are at a higher risk of impaired hypoglycemia awareness (IAH). IAH affects 25% of those with type 1 diabetes and 10% of those with insulin-treated type 2 diabetes, resulting in six- and 17-fold higher rates of severe hypoglycemia in type 1 and type 2 diabetes, respectively. Given the high prevalence of severe hypoglycemia and especially IAH, Dr. Heller cautioned against “being too satisfied with technology and education” and advocated a combination approach to restore awareness and reduce severe hypoglycemia. On interventions, Dr. Heller briefly touched on novel insulin analogues (including inhaled insulin) and CGM, both of which have been shown to improve glycemic control and reduce the risk of hypoglycemia but do not lead to changes in IAH. For instance, in the SMILE study of hybrid closed loop, MiniMed 640G (predictive low glucose suspend) produced a remarkable 84% reduction in severe hypoglycemia in a high-risk population of people with long duration of type 1 diabetes. However, there were no significant changes in Clarke or Gold scores for hypoglycemia awareness. Meanwhile, structured education programs offer some hope for improving hypoglycemia awareness among people with IAH. In particular, Dr. Heller highlighted the DAFNE (Dose Adjustment for Normal Eating) program in the UK,  which teaches the basics of intensive insulin therapy, and led to improved awareness in those with IAH. Indeed, a 2015 meta-analysis by Yeoh et al. found that both educational and technology interventions improved glycemic control and reduced severe hypoglycemia, but only some educational programs improved awareness and most technology interventions did not improve awareness. In closing, Dr. Heller highlighted the HARPDoc study as a promising educational program to address harmful cognitions surrounding hypoglycemia, with additional mental health benefits as seen in the reduction in scores for diabetes distress, anxiety, and depression.

• Addressing the inconsistent results for interventions to improve hypoglycemia awareness, Dr. Heller emphasized that IAH is a heterogeneous condition that can arise through different underlying pathologies. Broadly, he described two main forms of IAH: (i) IAH that arises due to the underlying pathogenesis of long diabetes duration and (ii) IAH that arises due to a maladaptive stress response, in which the counterregulatory hormonal response to hypoglycemia is suppressed, due to repeated acute hypoglycemia. Given the heterogenous nature of IAH, Dr. Heller stated that it’s naïve to expect a single intervention to improve awareness and ultimately called for larger multi-clinical trials that seek to define the phenotype of IAH as a precursor before evaluating potential interventions.

12. Retrospective, observational chart review in broad type 2 population (n=2,331) at International Diabetes Center: CGM use correlated with 0.9% A1c reduction from 8.9% to 8.0%; fivefold (!) increase in participants taking zero medications, from 5% to 25% of cohort

During an oral presentation session, Dr. Anders Carlson shared data from the International Diabetes Center highlighting CGM use in a broad population of people with type 2 diabetes. To start, Dr. Carlson noted that most studies evaluating CGM use in type 2s examine people on insulin therapy (e.g., the MOBILE RCT), and that large prospective randomized studies of CGM use in non-insulin-treated type 2s are “sparse.” Dr. Carlson did point out that in one such study, albeit a small one (Wada et al. 2020), CGM use was associated with a 0.5% A1c decrease in non-insulin-treated type 2s (n=49) compared to those on BGM (n=51). Turning to the IDC study, Dr. Carlson explained that he and his colleagues performed a retrospective, observational EHR and claims review studying all people with type 2 diabetes who received diabetes care and had insurance through IDC’s parent integrated health system, HealthPartners (n=23,843) between January 2018 and December 2021. Within this population of type 2s, roughly 10% (n=2,331) received a new CGM order, and a majority of participants filled their prescriptions within 30 days of receipt (84%) and ultimately got a FreeStyle Libre (89%) or Dexcom (11%) CGM. Overall, 93% of CGM orders were filled.

• Across the entire type 2 cohort, using CGM use was correlated with a 0.9% decrease in A1c, from 8.9% at baseline, on average, to 8.0% post-CGM (p<0.0001). Dr. Carlson explained that the “baseline CGM value” was the closest A1c between zero and six months prior to starting CGM, and that the “post-CGM A1c value” was the closest A1c between eight weeks and 12 months after using CGM. Notably, the percentage of participants with an A1c under 8% increased by 17%, from 36% at baseline to 53% after CGM use. Dr. Carlson noted that this metric is worth looking at since in many places, including his home state of Minnesota, diabetes care quality metrics  include if HCPs meet a treatment target of A1c <8% and it is currently the HEDIS criteria for diabetes management. He also noted, of course, how much better this was for PWD! Dr. Carlson then turned to a graph (see below) showing the A1c change among participants with a baseline A1c between 8% and 10%, with bubble sizes proportional to the number of patients. As Dr. Carlson noted, it’s interesting to see that while CGM did not bring down A1c for every single participant, it did for most.

• After initiating CGM, Dr. Carlson stressed that the number of individuals in the study taking no medications increased fivefold, from 5% to 25%. According to Dr. Carlson, this result indicates that the mean 0.9% A1c improvement after using CGM likely stemmed from improved nutrition and lifestyle changes, as opposed to pharmacotherapy intensification. The percentage of participants who were taking ≤two medications increased as well, from 62% at baseline to 71% after CGM use. While from our view, less medicine isn’t better for everyone, presumably IDC has ways to determine who should be on a therapy like GLP-1 or SGLT-2s to reduce CV or kidney disease risk, rather than for glycemic management. And indeed, separately from Dr. Carlson, we learned that the observed decrease in pharmacotherapy stemmed almost all from a decrease in analog insulin, sulfonylurea, and DPP-4i usage. For example, SFU use went down from 27% to 16%.
  
  • Insulin usage among participants decreased from 61% pre-CGM to 50% post-CGM, which we find most interesting as it seems to suggest that CGM can drive meaningful lifestyle changes that reduce insulin dosing as well as dependence. Ultimately, we’d love to see Time in Range associated with this population as well as more granular CGM tracings to understand the needs better.
  • Interestingly, GLP-1 usage in the cohort increased regardless of baseline A1c. For those with an A1c over 10%, analog insulin use increased, whereas for those with a baseline A1c ≤10%, sulfonylurea use decreased. At the conclusion of his presentation, Dr. Carlson argued that these data suggest wider CGM coverage in a type 2 population can be beneficial, and he called for more research into CGM use in non-insulin-treated type 2s. We certainly understand the need and call for this.
• Dr. Carlson stressed that at the beginning of the study in 2018, CGM was available to all type 2s insured by the IDC, which may account for the larger number of people not on insulin and on fewer medications in the study. We wonder whether CGM is still offered to all type 2s who are insured through IDC, and if not, we would be eager to learn why this is not the case.
• There was no correlation between ethnicity/race and A1c. This is especially worth noting because the cohort was relatively diverse (1% Native American/Alaska Native, 8% Asian, 18% Black, <1% Native Hawaiian or Other Pacific Islander, 59% White, and 14% Unknown/Other). There were also no correlations between A1c and BMI or the number of concurrent medications. However, male sex (p=0.03), younger age (p=0.001), and filling CGM within 30 days (p=0.0003) were all associated with a lower A1c.

13. FDA-stipulated post-approval MiniMed 670G RCT (n=302) validates 670G single-arm pivotal results: Those on 670G see -0.6% A1c improvement and +2.9 hour/day Time in Range improvement relative to those on sensor-augmented pump therapy at six months

Kicking off a most valuable FDA-mandated oral presentation session focused on AID technology, Dr. Robert Vigersky (Medtronic) read out the results of the RCT evaluating MiniMed 670G vs. sensor-augmented pump (SAP) therapy, which confirmed the clinical benefit of MiniMed 670G demonstrated in the single-arm pivotal. At the start of the presentation, Dr. Vigersky noted that this study was the post-approval RCT that the FDA required when it granted MiniMed 670G its approval in September 2016, an approval that was based on the MiniMed 670G pivotal trial, a single-arm prospective study that compared MiniMed 670G to baseline pump use. Per Dr. Vigersky, the study he read out during this ATTD session is one of three that are being run with MiniMed 670G as part of the post-approval process; however, the other two are currently ongoing.

The six-month MiniMed 670G RCT included 302 AID-naïve participants (ages 2-80, baseline A1c 8.1%) who were randomized to MiniMed 670G (n=151) or SAP therapy (n=151) after two weeks of SAP therapy at baseline. At six months, those on MiniMed 670G saw a 0.6% A1c improvement relative to those on SAP, with those on 670G seeing their A1c improve from 8.3% to 7.3% while those on SAP seeing a slight A1c improvement from 8.1% to 7.7%. In terms of Time in Range, those on MiniMed 670G saw a 2.9 hour/day improvement in Time in Range compared to those on SAP when adjusted for baseline values (p<0.0001). Specifically, those on MiniMed 670G saw their Time in Range improve from 53% at baseline to 67% at six months while those on SAP saw their Time in Range improve only slight from 52% to 55%. Much of this benefit came at night, when those on MiniMed 670G saw their Time in Range improve from 54% at baseline to a whopping 74% at six months (52% to 55% among the SAP group), which translated to a +4.9 hour/day baseline-adjusted Time in Range improvement among 670G users vs. SAP users. These results were challenging to compare exactly to the 670G pivotal studies since there were three of those stratified by age group (ages 2-6, 7-13, 14+) compared to this RCT, which included participants ages 2-80. However, these results are generally in line with those of the pivotal trials, providing further confirmation of the value of AID systems. While some might add “even the first-gen ones,” we’re careful in our references – from our view, the algorithms matter and can make one AID system preferred by one person to another but the comparison at a population level of AID for all compared to AID for a few is stirring, indeed.

• Predictably, those with high A1cs saw greater A1c reductions while those with high time <70 mg/dl saw large improvements in time in hypoglycemia – that, to us, just extends from the understanding that it’s easier to show progress for those doing worse.
• As part of the analysis, the researchers stratified participants into two groups, those with baseline A1c values >8% (n=155) and those with baseline A1c values ≤8% (n=147).
  
  • Those in the A1c >8% cohort averaged age 35, had had diabetes for 17 years, and had an average A1c of 9.1%, whereas those with A1c values ≤8% averaged age 41, had had diabetes for 23 years, and had an average A1c of 7.2%.
  • As might be expected, those with baseline A1c values >8% saw a far greater A1c improvement than those with baseline A1c values ≤8%. Specifically, those using 670G with baseline A1c values >8% saw their Time in Range improve from 9.2% to 7.7%, good for a 0.8% baseline-adjusted relative A1c improvement compared to the SAP group, which saw its average A1c fall from 9% to 8.2%. Those with baseline A1c values ≤8% still saw a significant improvement relative to the SAP arm (baseline-adjusted relative improvement of 0.3%; p<0.0001).
  • On the flip side, those with baseline A1c values ≤8% saw a greater reduction in time below range than those with baseline A1c values >8%. Specifically, the group with baseline A1c values <8% saw their time <70 mg/dL fall from 8% to 2% whereas their SAP counterparts saw time <70 mg/dL only fall from 9% to 7%. Those with A1c values ≥8% saw a smaller but still significant improvement in time <70 mg/dL from 4% to 2%.
• As was the case in the pivotal trial, MiniMed 670G was found to be very safe with no adverse events occurring among 670G users. There was one DKA event and two severe hypoglycemia events during the run-in period and four severe hypoglycemia events in the SAP arm. 

14. Retrospective payer claims analysis (n=700+) shows professional CGM use associated with a -0.5% A1c improvement in type 2s on multiple non-insulin diabetes medications; professional CGM use associated with increased insulin, GLP-1, and SGLT-2 initiation, but rates still low

In an afternoon oral presentation session, Ms. Poorva Nemlekar (Lead Health Economics and Outcomes Research Specialist in Global Access, Dexcom) presented a large and impressive retrospective RWE analysis. We were thrilled to see more focus on professional CGM initiation, as we believe that it is an often-under-utilized resource that can drive cost-savings for so many individuals. In the analysis, professional CGM was used in a very specific population: people with type 2 diabetes on more than two non-insulin diabetes medications who are not achieving A1c targets. Professional CGM was associated with a -0.5% A1c improvement in this population, compared to patients who did not use professional CGM, which we see as quite meaningful – we have learned extensively from dQ&A about various people with type 2 diabetes who are not at their target A1cs, even people on medications like GLP-1s that are often associated with very positive outcomes in RCTs.

The analysis included n=707 adults ages ≥30 (average 66 years) with baseline A1c values between 7.8% and 10.5% (average 8.7%) with at least one claim of professional CGM use between January 2018 and October 2020 but no prior professional or personal CGM use. Nearly three-fourths of the group was on Medicare (72%), and 50% were non-Hispanic White, making this analysis a more diverse one than is often seen in diabetes technology studies where that figure is often >90%. These 707 participants were compared to a cohort of 14,774 type 2s on ≥2 non-insulin medications who had never used CGM and did not initiate professional or personal CGM between January 2018 and October 2020. To assess the impact of professional CGM, the researchers compared A1c and medication data six months prior and six months after professional CGM initiation (in the professional CGM group) or oral diabetes medication initiation (in the non-CGM group).

• Those who initiated professional CGM saw a -0.5% A1c improvement compared to their non-CGM using counterparts when adjusted for baseline A1c (p<0.0001). Specifically, those who initiated professional CGM saw a -0.8% A1c improvement from 8.7% six months prior to CGM initiation to 7.9% six months after initiation, whereas those who did not use professional CGM saw a 0.3% A1c reduction from 8.5% to 8.2%.
• Professional CGM use was also associated with a higher proportion of insulin initiation (p<0.0001). However, the proportion of professional CGM users who initiated insulin after using professional CGM was still quite low at 20%. This figure is higher than in the non-CGM-using group (10%). However, the fact that only one in five professional CGM users initiated insulin suggests that despite having the added glycemic information provided by CGM, and despite the fact that none of these patients were achieving an A1c <7% at baseline, providers were still generally not initiating insulin.
• When broken down into those who did and did not initiate insulin, the professional CGM users still saw a greater A1c improvement than their non-professional CGM-using counterparts. Specifically, professional CGM users who initiated insulin saw a -0.6% A1c improvement (8.9% to 8.3%) whereas non-CGM users who initiated insulin saw a slight +0.1% A1c increase (8.8% to 8.9%) (p<0.0001 for between-group difference). Likewise, professional CGM users who didn’t initiate insulin saw a -0.9% A1c improvement (8.7% to 7.8%) compared to a -0.4% A1c improvement in the non-professional CGM group (8.6% to 8.2%) (p<0.0001 for between-group difference). Thus, while the low proportion of insulin initiation after professional CGM use is discouraging, these results are still promising, as they suggest that professional CGM can still result in glycemic benefits even if its use is not associated with a therapeutic change. This again hits on the value of CGM to drive behavioral change, reiterating the findings of the MOBILE and FLASH-UK RCTs, both of which showed that those who initiated CGM saw glycemic benefits despite no change in insulin dosing, regardless of the population studied (MOBILE was type 2s on basal-only; FLASH-UK was type 1s).
• The researchers also explored associations between professional CGM use and non-insulin diabetes medication use. Overall, they found that initiating professional CGM was associated with significant reductions in sulfonylurea and biguanide use and significant increases in GLP-1 and SGLT-2 use. There was a particularly stark decline in sulfonylurea use among those who used professional CGM and no change among those who did not use professional CGM. This discontinuation of use among CGM users could be due to previously unknown hypoglycemia appearing in the CGM tracings, although this is purely conjecture on our part. We were also glad to see that GLP-1 and SGLT-2 use increased among those using professional CGM, although we would have liked to have seen further initiation of these therapies, given this population’s higher A1c values and the medication’s complication-risk-reduction benefits. Use of these classes isn’t an end in itself, of course; dQ&A research results from more than 1,700 people with type 2 diabetes using these drug classes in the US show that more than half have A1cs above 7%.

15. Country-level Time in Range data from MiniMed 780G and MiniMed 770G users demonstrates average Time in Range >70% across geographies; increased time in “SmartGuard” associated with increased Time in Range

In the exhibit hall, Medtronic shared remarkable real-world country-level Time in Range data from MiniMed 770G and MiniMed 780G users around the world. Set up as an interactive display, attendees could choose countries where Medtronic’s AID systems are available and then see average Time in Range across users in the chosen geography. Wow! Population-level data from both MiniMed 770G and 780G was very encouraging, showing that, on average, users are achieving >70% Time in Range on these systems. It was a smart way to set it up, putting the two together…

• Medtronic presented real-world MiniMed 780G population-level data from Italy, the Netherlands, the UK, South Africa, Greece, Poland, and Chile.
  
  • Starting in Italy, Medtronic reported data from 3,584 users who had an average Time in Range of 76%, an average Time Above Range of 12%, and an average Time Below Range of 2%. Wow – all impressive! Patients spent 92% of their time with the SmartGuard feature turned on.
  • In the Netherlands, data from 1,830 users spending 94% of time in SmartGuard were evaluated and demonstrated a Time in Range of 75%, Time Above Range of 22%, and Time Below range of 2.4%. Fancy this, having nearly double TAR (22%!) than Italy (12%!) and roughly the same TBR (2% vs 2.4%).
  • In the UK, data from 996 users who spent 91% of time in SmartGuard demonstrated an average Time in Range of 73%, Time Above Range of 24%, and Time Below Range of 2.3%. Here’s another country with higher TAR – we’d love to see the “over 250 mg/dL”.
  • In South Africa, data from 417 users who spent 89% of time in SmartGuard demonstrated an average Time in Range of 74%, Time Above Range of 24%, and Time Below Range of 2.3%.
  • In Greece, data from 132 users who spent an 95% of time in SmartGuard demonstrated an average Time in Range of 81%, Time Above Range of 15%, and Time Below range of 3.3%.
  • In Poland, 418 users who spent 93% of time in SmartGuard demonstrated an average Time in Range of 81%, Time Above Range of 15%, and Time Below range of 3.3%.
  • Finally, in Chile, data from 165 users who spent an average of 92% of time in SmartGuard demonstrated an average Time in Range of 77%, Time Above Range of 20%, and Time Below Range of 3%.

Across these geographies, there also appears to be an association between increased time in SmartGuard (i.e., when the system is delivering automatic correction boluses) and increased Time in Range highlighting the value of these automatic boluses for helping people stay in range – we aren’t sure this is statistically significant.

• Medtronic presented real-world MiniMed 770G population-level data from the US and Canada. Starting in the US, data from 33,713 users who spent an average of 94% of time in SmartGuard demonstrated an average Time in Range of 71%, Time above Range of 27%, and Time Below Range of 2.9%. In Canada, data from 3,068 users who spent an average of 76% of time in SmartGuard demonstrated an average Time in Range of 70%, Time Above Range of 28%, and Time Below Range of 1.7%. Of note, the SmartGuard in the MiniMed 770G system is the same as that in the MiniMed 670G and does not include automatic correction boluses which is likely related to the slightly lower Time in Range values seen among these patients compared to those on MiniMed 780G.

16. GWave non-invasive radio frequency-based glucose monitor starting clinical trials; aggregate data (n=53) demonstrates 96% readings in Zone A compared to venous glucose

Dr. Irl Hirsch (University of Washington) presented a new dataset on the non-invasive GWave glucose monitor developed by Israel-based Hagar where Dr. Hirsch serves as a medical advisor. We first wrote about GWave back in July 2021 – we are still learning more about noninvasive glycemic monitoring and appreciate the opportunity to learn about the area. Since that time, Hagar closed a Series B funding round for $11.7 million in August 2021. As noted a year ago, the GWave system uses radio frequency to measure glucose via a resistor-capacitor model of the skin and underlying blood vessels. This technology is somewhat different from many other non-invasive sensors that use infrared and light-based spectroscopy measuring reflection of light rays off of glucose molecules.

Hagar’s current GWave prototype is a wrist-worn sensor roughly one-third the size of a smartphone (see picture).  Dr. Hirsch explained that the company hopes to miniaturize this technology into a “watch-like” device. There are currently a number of non-invasive glucose monitors under development including other wrist-worn systems including GraphWear, Movano, LifePlus, as well as continued chatter about a potential glucose monitor in a next-generation Apple watch via the company’s partnership with Rockley Photonics - this has yet to be confirmed. We look forward to watching GWave move through the different milestones – reliability, ease of use, pricing, accuracy in larger groups, especially should CGM be commercialized for wellness applications outside of diabetes, sensor insertion, etc. and we remain curious if any non-invasive sensors will be able to achieve non-adjunctive labeling and how they may compare to current next-generation CGMs with Dexcom G7, Abbott’s FreeStyle Libre 3, Medtronic 780G, and Senseonics’ Eversense E3 as those all continue to improve.

• Data from Hagar’s initial clinical trial (NCT04658082) of its GWave system (n=5 with 45 data points each) found that following a 75g oral glucose tolerance test, 98% of GWave readings fell in Zone A compared to BGM while 96% of venous glucose comparators fell in Zone A on the Clarke error grid. While this was a small study, this level of accuracy is encouraging and has now been supported by one trial for a total of seven patients. Specifically, Hagar is currently conducting a trial and aims to enroll 250 people with either type 1 or type 2 diabetes to use the GWave sensor. We are glad to see a larger trial coming – many approaches are with smaller numbers of people and we’re wondering how this trial will be powered, what the enrollment criteria will be, etc. We’re hoping it is a trial with multiple different investigators rather than just one hospital so that more people from different geographies can participate.
• Data from the first nine participants enrolled in Hagar’s larger GWave trial followed a similar patter to the company’s initial trial with 89% of capillary glucose comparators falling in Zone A and 100% of venous glucose comparators falling within Zone A on the Clarke error grid. To date, GWave has been assessed in 53 people with 97% of values in Zone A compared to BGM and 96% of values on Zone A compared to venous blood glucose. Again, this is a small sample size, and while GWave’s current form factor may not appeal to all patients, Dexcom’s first one certainly didn’t either! Should the company be able to successfully miniaturize its radio frequency measurement system, we imagine there could well be a place in the market for GWave’s wrist-worn non-invasive sensors. Traditionally, larger companies don’t do extensive tech work like this but we await more news at who will move the company forward, especially given the recent fundraising.
• Notably, as GWave does not use infrared or light-based technology like some other wrist-worn sensors, experts don’t appear to see the potential for sunlight or skin-tone based interferences. This is good news as the potential to reduce system accuracy by interacting with the light receptors in these devices should be lower – stay tuned and we look forward to seeing and hearing more in the months ahead – it’s a big deal to have Dr. Hirsch’s endorsement.

17. rt-CGM use among type 2s on insulin therapy (n=36,080, average A1c of 8.3%) found to meet NHS willingness to pay threshold with an incremental cost-effectiveness ratio of £3,684 per QALY

Health economist Mr. Stephane Roze (Vyoo Agency) presented data from a health economics analysis demonstrating that rt-CGM use among people with type 2 diabetes on insulin therapy meets the NHS willingness to pay threshold. Specifically, rt-CGM use among type 2s has an incremental cost-effectiveness ratio (ICER) of £3,684 per quality adjusted life year (QALY) gained compared to SMBG, falling well-below the cutoff of £20,000 per QALY gained. Mr. Roze based his analysis on patients with type 2 diabetes on insulin therapy (n=36,080) who had an average A1c of 8.3%. Using data from both the MOBILE study and from the Kaiser health system between 2014-2019 of rt-CGM use among type 2s – both of these studies were published in JAMA, wow! – Mr. Roze estimated that rt-CGM use was associated with an average A1c reduction of 0.56% compared to SMBG. Mr. Roze also estimated that rt-CGM use was associated with reduced rates of severe hypoglycemia and hyperglycemia/DKA events of 0 per 100 person-years versus 4/100 person years and 2.5/100 person years, respectively for patients using SMBG. Additionally, Mr. Roze calculated the yearly cost of rt-CGM use at £1,250 (36 sensors per year assuming a 10-day sensor life at the current price of Dexcom G6 in the UK – based on these figures the average cost to the NHS per G6 sensor appears to be ~£35). For comparison,  the yearly cost of SMBG totaled £402 based on results from the DIAMOND trial, which demonstrated an average of 3.8 SMBG tests/day among patients with type 2 diabetes on insulin therapy. Together, these estimates were used to develop a model for rt-CGM cost effectiveness, which was assessed via the IQVIA CORE Diabetes Model. Based on this analysis, Mr. Roze calculated total direct costs related to diabetes of £79,866 for rt-CGM users and £77,172 for SMBG users resulting in a difference in cost of £2,694. In this analysis, the  additional upfront cost of rt-CGM was largely offset by an expected reduction in complications (DKA, severe hypoglycemia, retinopathy, amputations, nephropathy, and cardiovascular disease) for people using rt-CGM compared to SMBG resulting in the relatively close total direct costs of diabetes for both cohorts. Specifically, in 71% of simulations run based on Mr. Roze’s model, rt-CGM use among people with type 2 diabetes on insulin met the willingness-to-pay threshold of costing <£20,000 per QALY gained. Furthermore, in 39% of simulations run based on Mr. Roze’s model, rt-CGM went beyond cost-effectiveness and had the potential to drive cost-savings with a willingness to pay threshold of <£20,000 per QALY gained. Of note, this analysis only included direct costs associated with a diagnosis of type 2 diabetes and did not factor in potential costs from lost productivity, which Mr. Roze argued means real-world cost-efficacy of rt-CGM is likely even higher than demonstrated with this data. We see this analysis as encouraging in terms of building support for CGM use among patients with type 2 diabetes on insulin therapy, which was a key component of updates in the ADA 2022 Standards of Care.

• Updated NICE guidelines published in March recommend offering is-CGM to adults with type 2 diabetes on MDI therapy citing the higher cost of rt-CGM as a reason for not recommending use of the technology in this population. However, the NICE guidelines do note that providers should “consider [rt-CGM] as an alternative to is-CGM for adults with insulin-treated type 2 diabetes if it is available for the same or lower cost.” With the UK launch of FreeStyle Libre 3, which brings rt-CGM functionalities to the FreeStyle Libre franchise CGMs as the same price point as the is-CGM FreeStyle Libre 2 model plus this novel analysis demonstrating cost-efficacy for rt-CGM using Dexcom G6 as the base case among type 2s on insulin, we are curious if NICE may provide further guideline updates or if UK providers will begin to offer rt-CGM to more of their patients with type 2 diabetes.

18. First analysis out of Dexcom’s Type 2 Help study: Three-month observational study suggests Dexcom G6 improves quality of life but doesn’t lead to Time in Range benefits in broad type 2 (n=180) and prediabetes (n=29) population

In the e-poster hall, a 12-week observational study showed that Dexcom G6 improved quality of life in 209 adults with type 2 diabetes and prediabetes in the US. The Dexcom-sponsored study, first-authored by Dr. Margaret Crawford (Dexcom), included 209 adults with type 2 diabetes or prediabetes in the US (average age 56), including 29 people with prediabetes/at high risk for diabetes, 104 type 2s not on insulin, 40 type 2s on basal-only therapy, and 36 type 2s on fast-acting insulin. The study population was more reflective of the general type 2 diabetes population than many other studies in diabetes technology with 15% of participants identifying as Black, 9% as Asian, and 32% as Hispanic, an effort that was intentional in the study design. The analysis evaluated both Time in Range and quality-of-life/treatment satisfaction scores at baseline and after 12 weeks of use. Unfortunately, none of the subgroups saw an improvement in average Time in Range after twelve weeks, although the Dexcom researchers attributed this to the short duration and the observational nature of the study (didn’t include a directed intervention). Regardless of the lack of significant improvement in Time in Range, all subgroups saw improvements in their Patient Health Questionnaire-2 (PHQ-2) score, a measure of depressive symptoms on a 0-6 scale with a score ≥3 indicative of major depressive disorder, which in the study was used as a proxy for quality of life. Across all groups, PHQ-2 scores fell significantly to ≤1. Notably, those in the high-risk/prediabetes group saw additional significant improvements in self-reported self-efficacy, illness perception, and sleep quality. Likewise, the type 2s not on insulin saw significant improvements in self-efficacy and illness perception. CGM satisfaction scores were high across all groups at ~4.0 on a 0-5 scale. Overall, the researchers – who included the esteemed Dr. Katharine Barnard-Kelly – argued that the results suggest that there are significant quality of life benefits to be gained by a wide range of type 2s using rt-CGM.

• This analysis is a part of the broader Type 2 Help study (NCT04503239) and is the first analysis to come out of that study. Based on ClinicalTrials.Gov, the study aims to include 306 people with type 2 diabetes and prediabetes and is collecting data on CGM metrics, medication use, food intake, physical activity, sleep, heart rate, five-hour OGTT, and quality of life outcomes. Based on ClinicalTrials.Gov (last updated October 2021), the study was set to complete in February 2022, suggesting that it may be fully complete, although this wasn’t addressed in the poster or associated recording. Additional analysis from the study will be read out at ADA 2022 on the relationship between behavioral changes and glycemic outcomes in this broad type 2 population.

19. Quick Takes

1. Dr. Roman Hovorka shares that an RCT of a fully closed loop CamAPS AID system in type 2 diabetes has completed and been submitted to EASD 2022 for readout in Stockholm

• During the closed loop update session, Dr. Roman Hovorka (University of Cambridge) discussed AID technology in type 2 diabetes, which he began by acknowledging that while AID is hugely beneficial in type 1 diabetes, we could better attend to the needs of type 2s and how AID could support them as well. While he didn’t present new data in the talk, he did share a meaningful update: the researchers at the University of Cambridge have completed the analysis of an RCT evaluating a fully closed loop CamAPS system in a broad population of people with type 2 diabetes on basal-bolus therapy. The protocol of this study is included on ClinicalTrials.Gov (NCT04701424) and was previously discussed at ATTD 2021. Excitingly, the study’s results were submitted to EASD for readout at EASD 2022, meaning that we could see the results of the first-ever RCT evaluating AID in type 2s in a mere five months in Stockholm.

2. Personalization, structure, engagement, and social support identified as key factors in behavioral weight loss interventions

• Ms. Anne Wolf (University of Virginia) discussed key components of behavioral therapy for weight loss and clinical pearls to maximize effectiveness. Ms. Wolf specifically explored the role of technology in improving patient outcomes, as the challenges that often prevent weight loss program participants from reaching their goals are unique. Ms. Wolf advocated for technology use to reduce cost, connect coaches with participants, and improve patient engagement in order to increase retention for patients while reducing the burden on providers. Ms. Wolf reviewed evidenced from the National Diabetes Prevention Program, which as many know, included 22 group sessions following an evidence-based curriculum over a year long period, ultimately producing average weight loss of 4% (just under the generally accepted 5% threshold of clinical efficacy). While subsequent interventions have had varying degrees of success, Ms. Wolf identified personalization, structure, frequent engagement, and social support as key features that tended to have more positive weight loss results. Additionally, Ms. Wolf said that patients tend to have greater participation and higher retention if they see immediate results, arguing for support for participants early in the process to keep them excited about the program. Generally, though, the weight loss conferred by behavioral interventions does not exceed 5%. Thus, we continue to wonder which patients are an appropriate candidate for these intensive programs and which patients might benefit more from a new weight loss medication that can induce much greater weight loss (ex. 15% for semaglutide, 22.5% for tirzepatide!) This is a question we will continue to explore as we learn more about as the obesity treatment paradigm adjusts to reflect these new pharmacological options.

--by Emily Ye, Jackie Tait, Julia Stevenson, April Hopcroft, Ashwin Chetty, Claire Holleman, Armaan Nallicheri, Albert Cai, Katie Mahoney, Hanna Gutow, and Kelly Close