Executive Highlights
-
Hello from ADA 2019 in San Francisco!
-
REWIND results headlined the day in diabetes therapy: Lilly’s GLP-1 Trulicity gave a 12% risk reduction on three-point MACE over a median 5.4 years in a population (n=9,901) with a low rate of cardiovascular disease (31%) at baseline. The CV benefit was exactly the same in the study’s primary and secondary prevention cohorts (HR=0.87, 95% CI: 0.74-1.02 in each). This is the first trial to show that a GLP-1 agonist – or any diabetes drug – can reduce MACE risk in people with type 2 diabetes and CV risk factors. Read on below for details, including a 15% risk reduction on the renal composite of new macroalbuminuria, 30% fall in eGFR, or progression to renal replacement.
-
Detailed renal outcomes from DECLARE show consistent improvements across individual and composite cardiorenal outcomes with Farxiga. Dr. Ofri Mosenzon presented new UACR data as well, showing that the SGLT-2 can actually improve albuminuria status and highlighting the importance of targeting both eGFR and UACR to reduce renal risk. We’re keen on the nod she gave to earlier intervention in CKD, too.
-
A morning session featured results from the PRONTO studies for Lilly’s URLi (ultra-rapid insulin lispro) vs. Humalog. In both type 1 and type 2, URLi improved postprandial glucose excursions compared to Humalog, but this didn’t translate to superiority on A1c – though the studies weren’t designed to show that. There was some indication of increased postprandial severe hypoglycemia in type 2, though type 1s saw improved severe hypoglycemia >4 hours post-meal.
-
Also, don’t miss data from Lilly’s 12-week titration study for GIP/GLP-1 tirzepatide, showing that slower dose escalation can minimize (to <5%!) treatment discontinuation rates – a reassuring result following tolerability concerns in the phase 2b data.
-
Finally, see highly positive results for teplizumab in type 1 prevention, showing a median 2-year delay in type 1 onset.
-
-
Diabetes technology:
-
We saw extremely positive results from the six-month, NIH-funded iDCL pivotal trial (n=168) comparing Tandem’s Control-IQ hybrid closed loop/G6 vs. sensor-augmented pump (t:slim X2/G6 without automation). Time-in-range (70-180 mg/dl) was 2.6 hours per day better with Control-IQ – 70% vs. 59% (p<0.0001) – driving a 0.33% A1c advantage at six months (baseline: 7.4%). Hypoglycemia was <2% in both arms, though still in favor of Control-IQ by 13 mins/day. All 168 participants completed the randomized study (unprecedented in Dr. Roy Beck’s experience), and Control-IQ users spent a remarkable 92% (!) of the full six months with closed loop active. Users in both arms on G6 took just one fingerstick every 3-5 days. All in all, it was a home run study for Tandem, Dexcom, NIH, and the entire closed loop field. FDA submission is expected in the “coming weeks” (14+ years), with US launch still expected “this year.”
-
Highly anticipated results from the WISDM trial (n=203) evaluating CGM in elderly (≥60 years) type 1s found CGM was superior to BGM in every outcome measured. The between group difference for time <70 mg/dl was -1.9% in favor of CGM (-27 minutes/day), with a two-hour per day advantage for CGM on time-in-range (70-180 mg/dl). Importantly, CGM was effective regardless of insulin delivery method.
-
Insulet shared strong Omnipod Horizon hybrid closed loop data in a three-day hotel study in 14 children ages 2-6. Children spent four additional hours per day in 70-180 mg/dl (from 55% at baseline to 73%), driven by reduced hyperglycemia. We also learned more about Horizon’s features.
-
In, industry updates, we enjoyed seeing FreeStyle Libre 2 in its first product theater appearance on US soil. A Medtronic analyst briefing offered a detailed look at its AID and CGM pipeline over the next 24 months.
- In a major win for outcomes beyond A1c, the international consensus on time-in-range provided recommendations for CGM targets in a poster (2-LB) and paper published in Diabetes Care. The online paper also includes the new-look AGP, first introduced by Dr. Rich Bergenstal at a symposium last night.
-
Hello from an absolutely tremendous day #3 of ADA 2019! Our highlights below include readout of the REWIND CVOT for GLP-1 Trulicity, very positive phase 2 results for teplizumab in type 1 prevention, the PRONTO studies for Lilly’s ultra-rapid URLi, Tandem’s Control-IQ hybrid closed loop pivotal trial, the WISDM study testing CGM in 60+ year-olds, and so much more. We’ve officially crossed the halfway mark – but there’s much more data coming your way on day #4.
- Diabetes Therapy Highlights
-
- 1. REWIND: Trulicity Gives 12% RRR on 3-pt MACE, Showing that GLP-1 Mediated Cardioprotection Applies to Type 2 Without Established CVD; Effect Driven by Stroke; Trend Toward Mortality Benefit; Encouraging Renal Data
- 2. Teplizumab is the First Drug Shown to Delay Type 1 Diagnosis! Landmark Phase 2 Results Demonstrate ~Two-Year Median Diagnosis Delay in High-Risk Stage 2 Cohort
- 3. DECLARE Renal Outcomes: Farxiga Improves Both eGFR and UACR to Confer Strong Cardiorenal Risk Reduction in Lower-Risk Population
- 4. URLi (Ultra-Rapid Lispro) Significantly Reduces Postprandial Glucose Excursions (~30 mg/dl) and Severe Hypo >4 Hours Post-Meal vs. Humalog; Nonsignificant A1c Reduction; Dosing Post-Meal Only “If Needed”
- 5. PRONTO-T2D: URLi Non-inferior to Humalog on A1c Over 26 Weeks; Improvements in PPG Balanced Against Increase in Hypoglycemia (<54 mg/dL) between 1-4 Hours Post-Meal, but Low Event Rates
- 6. Tirzepatide (GIP/GLP-1) Dose-Escalation Study Shows Minimal (<5%) Discontinuation Due to Adverse Events with Slower Titration Schemes Compared to Phase 2b
- 7. Highly Positive DUAL VIII Trial Finds Longer Sustainment of A1c Values <7.0% vs. Lantus with ~Half the Weight Gain, 56% Reduction in Severe or BG-Confirmed Hypo, and 30% Less Insulin
- 8. BRIGHT Subgroup Analysis Suggests Potentially Improved A1c Lowering vs. Tresiba at Lower Renal Function (eGFR ≤60) with No Increase in Hypoglycemia, but with Small Number of Participants
- 9. Dr. Elizabeth Mayer-Davis Highlights the Synergy between Observational Epidemiology and Clinical Trials in Kelly West Award Lecture
-
- Diabetes Technology Highlights
-
- 1. Terrific Tandem Control-IQ Pivotal Trial: +2.6 Hour/Day Gain on TIR and 0.3% A1c Advantage vs. SAP (t:slim X2/G6); Staggering 92% of Six Months Spent in HCL! FDA Filing in “Coming Weeks”, Launch This Year
- 2. 6-Month WISDM RCT in Type 1s ≥60 Years Shows CGM is Superior to BGM in Every Outcome Measured: -27 Minute Between-Group Difference in Time <70 mg/dl, +2.1 Hour Advantage on Time in Range
- 3. Omnipod Horizon AID Very Strong in 2-6 year-old, ~3 Day Hotel Study: +4 Hr/Day TIR, No Sig Change in Hypo; 3 Investigational Horizon Features – extended bolus, variable setpoint (100-150 mg/dl), Hypo Protect; U500 Omnipod with FDA
- 4. FreeStyle Libre 2 Product Theater Offers First Look on US Soil, First Discussion of Alarms; No Update on FDA iCGM Review
- 5. Medtronic Pipeline Overview in Investor Briefing: MiniMed 780G, Non-Adjunctive CGM Launch in Next 12 Months
- 6. Freelife Kid AP Study of Tandem’s Control-IQ: Interim results show 24-hour wear beats overnight-only (+2 hours/day time-in-range) in pediatrics
- 7. Late-Breaking Poster Details International Consensus on Time-in-Range, Recommendations for CGM-Based Clinical Targets; Updated AGP Included in Diabetes Care Paper
- 8. iDCL Protocol 4 to Test Dexcom G6 CGM/Tandem’s t:slim X2 Pump/Adaptive Zone MPC Algorithm in Six-Month RCT; Algorithm Parameters Adjusted Weekly
-
Diabetes Therapy Highlights
1. REWIND: Trulicity Gives 12% RRR on 3-pt MACE, Showing that GLP-1 Mediated Cardioprotection Applies to Type 2 Without Established CVD; Effect Driven by Stroke; Trend Toward Mortality Benefit; Encouraging Renal Data
In one of ADA 2019’s most highly-anticipated presentations, the full readout of Lilly’s REWIND CVOT (n=9,901) for GLP-1 agonist Trulicity (dulaglutide 1.5 mg) revealed a 12% relative risk reduction (RRR) on 3-point MACE (HR=0.88, 95% CI: 0.79-0.99) vs. placebo. Results were simultaneously published in The Lancet in two separate papers (CV outcomes, renal outcomes). Of critical importance, there was strong consistency on the primary outcome between the trial’s primary (69% of participants) and secondary (31%) prevention cohorts. The hazard ratio on MACE was 0.87 for both groups, with identical confidence intervals (95% CI: 0.74-1.02) – meaning that REWIND offers the first evidence that a GLP-1 agonist can reduce MACE risk in people with type 2 diabetes but without established CVD. Moreover, the “exploratory” renal analysis identified a (nominally) significant impact on the progression of renal disease and a number of other microvascular outcomes – find far more on that below. In addition to the large primary prevention cohort, REWIND is particularly notable for its duration of follow-up (median 5.4 years), the longest of any GLP-1 CVOT, as well as a low baseline A1c (median 7.2%) and large enrollment of women (46%) – the presentation emphasized that REWIND is more applicable to the general type 2 diabetes population than previous CVOTs. Below, dig into our rundown of the eight-part, two-hour symposium, and take a look at our updated CVOT landscape. As a reminder, these data have already been submitted to FDA and EMA.
-
Introduction and research question. Dr. Gilles Dagenais gave an introduction to REWIND and GLP-1 agonist CVOTs, referencing the demonstrated three-point MACE risk reductions from LEADER for Novo Nordisk’s Victoza (liraglutide), SUSTAIN 6 for Novo Nordisk’s Ozempic (semaglutide), and HARMONY for GSK’s discontinued Tanzeum (albiglutide). However, he noted, these trials were in people with mean baseline A1c ≥8.0% and an annual risk of CV events ≥4% – a comparatively higher risk population than that enrolled in REWIND.
-
Background, design, and outcomes. Dr. Rafael Diaz dug deeper into the evidence to-date and baseline characteristics from GLP-1 CVOTs. On a meta-analytic level, all five previously completed GLP-1 CVOTs combine to give a 12% RRR (95% CI: 0.84-0.94) on three-point MACE, including a 13% RRR (95% CI: 0.82-0.92) among those with established CVD but no demonstrated benefit in those with multiple risk factors (HR=1.03, 95% CI: 0.87-1.23). For reference, REWIND enrolled those ≥50 years old with vascular disease, ≥55 with subclinical vascular disease, or ≥60 with two CV risk factors, on any therapy save prandial insulin. All participants had an A1c ≤9.5% (which he characterized as a relatively high limit) and BMI ≥23 kg/m2. Secondary outcomes included a microvascular composite (retinopathy + nephropathy), hospitalization for unstable angina, hospitalization for heart failure, individual MACE components, and total mortality.
-
Participants and follow-up. Dr. Matthew Riddle outlined baseline characteristics of the trial’s 9,901 participants, including mean age of 66 years, 46% female and 76% white enrollment, 31% baseline CVD, 21% prior MI/stroke, 93% hypertension, and 9% heart failure prevalence. Prior CV disease was defined as history of MI, ischemic stroke, unstable angina with ECG change, myocardial ischemia on imaging or stress test, or revascularization (coronary, carotid, or peripheral). Mean baseline A1c was 7.3%, BMI 32 kg/m2, eGFR 75 ml/min/1.73 m2, and diabetes duration 11 years; retinopathy prevalence was 9%, Stage 3 or greater CKD 22%, and albuminuria 35%. Additionally, 81% were on metformin, 46% SUs, 24% insulin, 6% DPP-4s, 2% TZDs, and <1% other; 82% were taking an ACE/ARB, 46% a beta blocker, 57% other blood pressure medication, 66% statin, 9% fibrate, and 54% antiplatelets. In an interview with our team, Dr. Brad Woodward (Global Brand Development Lead and Senior Medical Director, Lilly) noted that 5%-7% of participants were taking an SGLT-2 at the end of the trial. Median follow-up was 5.4 years, and total person years of follow-up was 51,820.
-
At the end of REWIND, final status was known in 99.7% of participants in each arm; 97% attended the final visit, had a primary outcome event, or had a non-CV death. Additionally, for Trulicity and placebo respectively, 82%/83% of total follow-up time was on study drug, 73/71% were on study drug at their last visit, 58%/56% never stopped study drug, and 11%/8% stopped due to an adverse event – this last he characterized as low for a five-year study. Dr. Riddle seemed quite impressed by the overall retention and adherence in the study, as well as the 0.61% treatment difference on A1c at the end of the study (p<0.001) – which he called “rather significant and probably clinically significant,” particularly given the trial was not designed for A1c lowering. See the summary of clinical measures below:
-
-
Cardiovascular outcomes. Prof. Hertzel Gerstein presented cardiovascular results in the REWIND trial, headlined by a 12% RRR on three-point MACE (HR=0.88, 95% CI: 0.79-0.99). Kaplan-Meier curves for the composite cardiovascular outcome and each separate component (CV death, non-fatal MI, and non-fatal stroke) can be seen below. The K-M curve for the composite outcome separates after the first year and demonstrates a steady separation over time, consistent with hypothesized anti-atherosclerotic effects driving GLP-1 cardioprotection. Interestingly, it appears as if benefits in non-fatal stroke (HR=0.76, 95% CI: 0.61-0.95, p=0.017) drove much of the primary outcome benefit, which is similar to a 39% RRR on stroke with Ozempic in SUSTAIN 6.
-
Dulaglutide’s CV benefit was importantly consistent in those with and without established CVD. Remarkably, the point estimate HR for both subgroups was identical (HR=0.87, 95% CI: 0.74-1.02). No significant interactions were seen in most subgroup analyses, with consistent effects across age, sex, diabetes duration, baseline A1c, and BMI above or below 32 kg/m2. A disparity was seen in terms of geography, but this was attributed to the small sample size of participants from the Asia Pacific region; further analysis by ethnicity showed no interaction.
-
There was a trend toward mortality benefit, both CV and all-cause, with Trulicity, but neither reached significance. The hazard ratio for all-cause death was 0.90 (95% CI: 0.80-1.01), and the HR for CV death was 0.91 (95% CI: 0.78-1.06). To our understanding, the lack of significance could be an effect of the lower-risk population, though it’s hard to say for certain.
-
-
Renal outcomes. University of Edinburgh’s Dr. Helen Colhoun took the stage for the renal outcomes; on the primary composite of new macroalbuminuria, 30% fall in eGFR, or progression to renal replacement, Trulicity conferred a significant 15% risk reduction (HR=0.85, 95% CI: 0.77-0.93, p=0.0004). This was driven primarily by reduction in new macroalbuminuria (HR=0.77, 95% CI:0.68-0.87), followed by sustained eGFR decline (HR=0.89, 95% CI:0.78-1.01, p=0.066), and slightly by reduction in renal replacement therapy (HR=0.75, 95% CI:0.39-1.44, p=0.39). Seeing as the eGFR component just barely missed out on superiority, Dr. Colhoun presented sensitivity analyses with more stringent guidelines (which are often used in dedicated renal trials or other CVOTs, e.g., DECLARE). If the eGFR decline threshold had been set at ≥40%, then the component would’ve reached nominal significance in favor of Trulicity (HR=0.70, 95% CI:0.57-0.85, p=0.0004), which further improved when tested at ≥50% (HR=0.56, 95% CI:0.41-0.76, p=0.0002). Dr. Colhoun attributed this trend to less random misclassification at more stringent thresholds. Notably, there were no subgroup differences on the composite outcome based on baseline eGFR (>60 or ≤60 ml/min/1.73 m2, p for interaction=0.65), albuminuria (p=0.66), and ACE or ARB use (p=0.23). The last is particularly intriguing to us, as it suggests the renal benefit from Trulicity comes on top of antihypertensives, some of which are renoprotective themselves.
-
Given the recent, very positive readout of CREDENCE (SGLT-2 inhibitor Invokana in DKD), we find it important to place REWIND’s renal results in perspective. While CREDENCE may have raised the threshold for excitement over renal data, we still view REWIND’s data quite positively, especially in light of the lower-risk population vs. other CVOTs and CREDENCE. It’s important to note that no therapies have been approved for CKD management in nearly 20 years, so there’s a desperate unmet need which ideally could be filled by multiple classes to better suit the diversity of patients with type 2 diabetes. Further, since the SGLT-2 inhibitor and GLP-1 agonist classes operate by such different mechanisms, we wonder whether there could be any additive or synergistic CV/renal benefit, on top of additive glycemic benefit.
-
-
Safety. Dr. Jeff Probstfield’s safety presentation was brief and reassuring: There was no significant difference between the Trulicity and placebo arms on 1st study drug stopping, acute pancreatitis, any cancer, MTC or C-Cell hyperplasia, thyroid cancer, pancreatic cancer, serious renal/urinary event, serious GI event, SVT/CV conduction disorder, or severe hypoglycemia. There was a slight trend toward more hepatic events with placebo (p=0.057) and more immune reactions with Trulicity (p=0.022), but with very low numbers of events. Comparing all GI adverse events, 47% on Trulicity and 34% on placebo were affected (p<0.0001). Dr. Probstfield closed by underscoring a final measure of study population risk: The annual placebo MACE rate in REWIND was 2.7%, compared to 3.9% in LEADER, 4.0% in EXCEL, 4.4% in SUSTAIN 6, 5.9% in HARMONY, and 6.3% in ELIXA.
-
Summary and implications. Dr. Lars Ryden provided context for the REWIND trial, highlighting its influence on our interpretation of existing data for the GLP-1 class. Building on a recent meta-analysis published in Circulation by Zelniker and colleagues, Dr. Ryden showed that incorporating REWIND results leads to an overall weighted MACE hazard ratio of 0.88 (95% CI: 0.84-0.93), solidifying a class effect for GLP-1s. Importantly, REWIND contributes nearly 1/5 of the “weight” for the analysis, due to its large number of participants and long follow-up. In the population with established CVD, incorporating REWIND gives a 13% RRR (HR=0.87, 95% CI: 0.82-0.92, weight=12%); in those without CVD, the HR becomes 0.94 (95% CI: 0.84-1.06, weight=54%). Interestingly, the inclusion of REWIND data shifts the hazard ratio in the “no prior CVD” population from >1.0 to <1.0, as other trials for GLP-1s that included primary prevention cohorts (LEADER for liraglutide, SUSTAIN-6 for semaglutide, and EXSCEL for exenatide) all showed neutral effects on MACE in this subgroup. Seeing as thought leaders have mostly agreed upon cardioprotection as a GLP-1 class effect, we’re curious to see whether this data is compelling enough to convince the field of a cardioprotective effect for the class in primary prevention populations as well – especially since other GLP-1 trials have not shown such benefit.
-
Independent commentary. Monash University’s Dr. Sophia Zoungas provided the independent commentary for REWIND, underscoring the study’s long follow-up and generalizable population as particularly noteworthy, especially when compared with other GLP-1 CVOTs. We appreciated the thoroughness of her analysis, which dissected the trial methods, results, and clinical implications one-by-one:
-
REWIND’s low annual placebo MACE rate (2.7 events/100 patient-years) reflects its low-risk population and generalizability vs. all other GLP-1 CVOTs. By comparison, ELIXA had a placebo MACE rate of 6.3%/100 py, LEADER 3.9%/100 py, SUSTAIN-6 4.4%/100 py, EXSCEL 4.0%/100 py, and HARMONY 5.9%/100 py. To this end, Dr. Zoungas posited that the positive CV results across both the primary and secondary prevention cohorts of REWIND could lay the first brick in a case to update the ADA/EASD Consensus Statement, promoting GLP-1 agonists not only for those with established ASCVD but also for those with CV risk factors. Our sense, however, is that it will take further study, and perhaps more robust data specific to the primary prevention population, to truly move the needle.
-
Although the excellent retention and adherence create a nearly flawless RCT analysis, they do not lend themselves to real-world generalizations. As suggested by Drs. Steven Edelman and Bill Polonsky in Diabetes Care in 2017, real-world data suggest that discontinuation rates for GLP-1 agonists may approach nearly 60% by 1-2 years – far shorter than the 5.4 year follow-up assessed in REWIND. To this end, while the results from REWIND are exciting for a multitude of reasons, providers must focus on promoting adherence to Trulicity to realize the CV and renal benefits, which may present cost-issues for some patients in the long-term. In general, Dr. Zoungas called for more patient-reported-outcomes (yes!) in CVOTs, such as quality-of-life data, though she acknowledged that their primary purpose is to assess the safety of novel therapies.
-
Despite an apparent difference in Trulicity’s effect on stroke (significant reduction) and MI (nearly neutral), we should be cautious about drawing any conclusions on the molecular or class levels. Firstly, the confidence intervals for both endpoints were wide and overlapping in REWIND. Within the GLP-1 class, only SUSTAIN-6 has followed a similar pattern to REWIND (significant stroke reduction, non-significant trend on MI). Other CVOTs for human-based GLP-1s have had very different trends: In LEADER, MI and stroke trended very similarly toward liraglutide, without significance, while MI was significant in HARMONY for albiglutide but stroke was not.
-
Consensus points toward kidney benefit and numerical increases in eye events in GLP-1 CVOTs. The renal benefit seen with Trulicity was positive, “any way you look at it,” while the eye outcome trended toward placebo (HR=1.24, 95% CI:0.92-1.68). As a reminder, the retinopathy signal in SUSTAIN-6 for Novo Nordisk’s Ozempic (semaglutide) prompted an FDA Advisory Committee prior to its approval.
-
The number-needed-to-treat (NNT) to prevent one MACE event in REWIND was comparable to other CVOTs – 60 for the overall population and 18 in those with established CV disease, each over 5.4 years (the median follow-up). Intriguingly, Dr. Zoungas compared this value to SGLT-2 inhibitors instead of GLP-1 agonists: 39 over 3.0 years for Lilly/BI’s Jardiance in EMPA-REG and 40 over 2.5 years for J&J’s Invokana in CREDENCE. For GLP-1’s, this metric is only available for Novo Nordisk’s products: 66 over 3.0 years for liraglutide in LEADER and 45 over 2.0 years for semaglutide in SUSTAIN-6. Notably, NNT is heavily dependent on the study population’s risk level, making comparisons only useful at a very high level.
-
With its diverse population and long follow-up, REWIND contributes significantly to consensus on GLP-1 CV class effects. When added, to Zelniker et al.’s meta-analysis of GLP-1 CVOT outcomes, REWIND contributes 25% of the weight toward an estimated 12% risk reduction on MACE (95% CI:0.84-0.93) for the entire class. Considering only the primary prevention data from these trials, REWIND shifts the overall point estimate from slightly trending toward placebo, to trending toward the GLP-1 class (see below).
-
-
As parting words, Dr. Zoungas quoted the accompanying editorial for REWIND penned by Drs. Subodh Verma, CD Mazer, and Vlado Perkovic (we highly recommend), advocating for earlier intervention with novel agents in type 2 diabetes treatment: “If we are to reduce the burgeoning pump, pipes and filter complications of diabetes, we will need to overcome clinical inertia, and embrace these disease-modifying therapies early, and preferably in combination. The REWIND trial makes a strong case in this regard.”
-
For future study, Dr. Zoungas identified three knowledge gaps related to the clinical impact of GLP-1 agonists: (i) Should GLP-1s with CV or renal benefits be used as first line therapy with lifestyle interventions? (we think displacing metformin would be a tough sell from both cost and efficacy standpoints); (ii) is there any additive CV or renal benefit with GLP-1s and SGLT-2s? If so, in what populations? (as a reminder, additive improvement in CV risk factors has been demonstrated); and (iii) are GLP-1s truly disease modifying? Are there any long-term pleiotropic effects or impacts on metabolic memory?
2. Teplizumab is the First Drug Shown to Delay Type 1 Diagnosis! Landmark Phase 2 Results Demonstrate ~Two-Year Median Diagnosis Delay in High-Risk Stage 2 Cohort
Yale’s Dr. Kevan Herold presented positive results from the highly-anticipated phase 2 teplizumab type 1 diabetes prevention trial, showing marked reductions on cumulative diabetes onset (72% vs. 43%, HR=0.41, p=0.006) and time to diagnosis (48 months vs. 24 months) with the anti-CD3 treatment compared to placebo. This is the first trial to meet a clinical endpoint in delaying the onset of type 1 diabetes and is absolutely phenomenal news for the entire diabetes community. Results were simultaneously published in NEJM and accompanied by an editorial penned by Drs. Clifford Rosen and Julie Ingelfinger – we highly recommend reading both.
The study enrolled 76 participants (55 children and 21 adults) who were the relatives of patients with type 1 diabetes who did not have diabetes but were at high risk for developing type 1. All of the study participants were identified by TrialNet’s Pathway to Prevention screening program and in “Stage 2” of type 1 disease progression, meaning that they had two or more type 1 autoantibodies and abnormal blood sugar levels (nearly 100% of those in Stage 2 progress to a clinical diagnosis of type 1 in their lifetime). Participants were randomized to either a single 14-day course of intravenous teplizumab (anti-CD3 monoclonal antibody) or placebo, with follow-up consisting of OGTTs at 6-month intervals.
Type 1 diabetes was diagnosed in only 43% of those on teplizumab compared to 72% of those on placebo. Annualized rates of diabetes were 15% for the teplizumab group and 36% for the placebo group. Very importantly, the median time to diabetes diagnosis was 24 months for placebo and 48 months for teplizumab, indicating that teplizumab treatment could delay type 1 diagnosis by 2 years on average – an extremely meaningful clinical result. See the table below for a snapshot of the trial results. Very notably, Dr. Herold commented that even in those treated with teplizumab who progressed to type 1 diabetes, the progression was slower, indicating that there still was beneficial action occurring. On this front, further work will involve trying to identify responders through biomarker evaluations, along with investigating combo therapies with teplizumab to increase the ability to delay or prevent type 1 progression.
-
Commercially, Provention Bio acquired development rights to teplizumab (now called PRV-031) from MacroGenics in May 2018. Provention’s press release regarding the trial notes that it is now “evaluating a regulatory path forward for PRV-031 in at-risk individuals.” The company is also assessing the candidate in a phase 3 pivotal trial (PROTECT, n=~300) assessing teplizumab’s ability to slow the loss of beta cells and preserve beta cell function in children and adolescent 8-17 years old who have been diagnosed with type 1 in the previous 6 weeks. Provention is holding a conference call Monday morning at 8:30 AM to discuss the results, with a webcast replay available beginning at 11 AM.
-
The history of teplizumab is long and winding, and we salute the perseverance of all those involved in this trial for driving it forward. In 2010, MacroGenics and Lilly announced that their phase 3 trial of teplizumab had failed, as it did not meet its primary efficacy endpoint of decreasing daily insulin usage and A1c after one year in recent-onset type 1 diabetes. Further clinical development was suspended as a result. At the time, we heard from several researchers that the phase 3 failure may have been more attributable to the study design rather than teplizumab’s efficacy, and these phase 2 results from today lend credence to that idea.
-
Phase 2 results in 2013 from Dr. Herold’s group were positive, demonstrating that teplizumab treatment significantly reduced C-peptide decline at two years in people with new-onset type 1 diabetes.
-
The compound was first developed by UCSF’s renowned Dr. Jeffrey Bluestone in the 1980s for use in targeting aggressive CD3+ T cells which at the time had been implicated in organ transplant rejection and multiple autoimmune diseases. See our 2016 interview with Dr. Bluestone here – some of his comments look quite prescient in light of today’s results. Further research by Drs. Bluestone, Herold, and others implicated the candidate in interfering with beta-cell autoimmune destruction, thus prompting clinical investigation of teplizumab in this context.
-
-
GSK is also investigating another anti-CD3 monoclonal antibody, otelixizumab, although the status of this clinical program remains unclear. The candidate was being studied in a phase 2 trial in new-onset type 1 diabetes and was apparently completed in September 2018, but no updates have been given by the company since (and in the wake of the company’s divestment from diabetes). GSK and partner Tolerx had previously reported negative results from a phase 3 study of otelixizumab, where the candidate did not meet the primary efficacy endpoint of a change in C-peptide at 12 months in patients with new-onset autoimmune type 1 diabetes. We wonder whether these positive results from teplizumab may spur additional investment in this space and push further study of other anti-CD3 (and more broadly, other immunotherapies) candidates in type 1 prevention and disease modification. See our full type 1 cure/prevention/treatment landscape here.
Trial Snapshot: Results and Study Population
|
Placebo (n=32) |
Teplizumab (n=44) |
Total T1D Diagnosis Rate |
72% |
43% (p=0.006) |
Annualized Rates of Diagnosis |
36% |
15% |
Median Time to Diabetes |
24 months |
48 months |
Number at end of trial without diabetes |
9 |
25 |
Total Adverse Events |
22 |
43 |
Grade 3 Adverse Events |
2 |
20 |
Grade 2 Adverse Events |
20 |
23 |
Baseline Characteristics |
5.3% A1c Average age: 13 53% Male 94% White |
5.2% A1c Average age: 14 57% Male 100% White |
Primary Outcome K-M Curve: Time to Diabetes Onset and Cumulative Incidence
3. DECLARE Renal Outcomes: Farxiga Improves Both eGFR and UACR to Confer Strong Cardiorenal Risk Reduction in Lower-Risk Population
Dr. Ofri Mosenzon presented very positive in-depth renal outcomes from the DECLARE CVOT (n=17,160) for AZ’s SGLT-2 inhibitor Farxiga (dapagliflozin 10 mg), detailing their potential in CKD prevention and treatment – see the full publication in The Lancet Diabetes & Endocrinology. Results with Farxiga for the full cohort are as follows:
Because DECLARE met only one of its co-primary endpoints, these analyses are considered exploratory. Dr. Mosenzon also paid particular attention to the impact of Farxiga on UACR, which was not included in DECLARE’s two main renal outcomes – unlike in CANVAS and EMPA-REG. Farxiga conferred (i) greater improvements in UACR and (ii) lower rates of UACR deterioration vs. placebo. On the improvement component, Farxiga conferred a 46% increased risk (95% CI: 1.31-1.62) of moving from micro- to normoalbuminuria, an 82% increased risk (95% CI: 1.51-2.2) of moving from macro- to normo/microalbuminuria, a 54% increased risk (95% CI: 1.4-1.69) of moving from macro- to normo/micro- or micro- to normoalbuminuria, and a 41% increased risk (95% CI: 1.27-1.56) of moving from micro/macro- to normoalbuminuria. On the other side of the coin (preventing deterioration), those on the SGLT-2 had a 46% reduced risk (HR=0.54, 95% CI: 0.45-0.65) for moving from normo/microalbuminuria to macroalbuminuria, a 21% reduced risk (HR=0.79, 95% CI: 0.72-0.87) for normo- to micro/macroalbuminuria, and a 27% reduced risk (HR=0.73, 95% CI: 0.67-0.79) for normo- to micro/macro- or micro to macroalbuminuria. That is, in all categories of UACR, dapagliflozin conferred an improvement. To reinforce the importance of impacting both components of renal pathophysiology (eGFR and albuminuria), Dr. Mosenzon showed the cardiorenal and renal-specific event rates across the array of UACR and eGFR categories – see how both contribute to risk below.
-
We also very much appreciated Dr. Mosenzon’s emphasis on the importance of early identification and intervention in renal disease. Given the still-staggering incidence and prevalence of CKD and the contribution of diabetes to this epidemiology, SGLT-2s doubtless represent a huge opportunity area through earlier intervention. Of course, this argument aligns well with the data from DECLARE: Given the study’s large primary prevention population (59% of participants) and labeling restrictions at the time of design, the study had a higher mean eGFR of 85 ml/min/1.73 m2 than other SGLT-2 trials, including CANVAS (77 ml/min/1.73 m2), EMPA-REG OUTCOME (74 ml/min/1.73 m2), and CREDENCE (56 ml/min/1.73 m2). In DECLARE, 48% had an eGFR ≥90, 45% an eGFR 60 to <90, and only 7% <60 (a function of the equation used – this eGFR was technically excluded). Additionally, 69% had a UACR <30 mg/g, 24% ≥30-≤300 mg/g, and only 7% >300 mg/g.
-
AZ’s renal outcomes trial for Farxiga, Dapa-CKD (n=4,000 with CKD, with or without diabetes), is already well underway and expected to complete in November 2020. Given consistently strong evidence around SGLT-2s and renal outcomes, including the recently-reported and groundbreaking CREDENCE trial, Dapa-CKD is very much expected to be positive.
-
In a conversation with our team, AZ’s Head of late phase Cardiovascular and Metabolic development Dr. Elisabeth Björk and VP of Global Marketing Ms. Kiersten Combs underscored the implications of the DECLARE CVOT on, (i) treating CKD earlier; and (ii) treating diabetes beyond A1c. In particular, this new renal analysis underscores that renoprotection may extend to earlier in CKD pathogenesis, as prevention of macroalbuminuria may preempt eGFR decline in some patient groups, according to Dr. Björk. Further, the study’s broad, largely generalizable population, continues to make the case for safety and efficacy, particularly on renal and heart failure outcomes in nearly all people with type 2 diabetes and an eGFR above 60 mL/min/1.73m2 (enrollment below this threshold was not allowed by the label). In the wake of REWIND results, Dr. Björk also added that she sees great potential for combination treatment with GLP-1 agonists and SGLT-2 inhibitors on both cardio and renoprotection, given their very different mechanisms. The two concluded by reminding that this is “only the middle of the journey” for Farxiga: Other running programs include DAPA-HF (heart failure with reduced ejection fraction, with or without type 2, results expected 2H19), as well as DELIVER (heart failure with preserved ejection fraction, with or without type 2) and Dapa-CKD (dapagliflozin in chronic kidney disease, with or without type 2) for which results are anticipated in 2020+.
4. URLi (Ultra-Rapid Lispro) Significantly Reduces Postprandial Glucose Excursions (~30 mg/dl) and Severe Hypo >4 Hours Post-Meal vs. Humalog; Nonsignificant A1c Reduction; Dosing Post-Meal Only “If Needed”
To a standing-room-only audience, Rainier Clinical Research Center’s Dr. Leslie Klaff presented results from the PRONTO-T1D study (n=1,222), demonstrating a significant improvement in postprandial glucose excursions with Lilly’s phase 3 URLi (ultra-rapid insulin lispro) vs. its progenitor Humalog (insulin lispro). Mealtime (dosed 0-2 minutes before meal start) URLi reduced postprandial glucose by ~28 mg/dL (95% CI: 35.3-20.6) at the one hour and ~31 mg/dL (95% CI: 41.1-21.2) at the two hour mark post-meal vs. mealtime Humalog (both p<0.001). Results at the three and four hour marks were also both significantly lower, but no values were given. The post-meal URLi group, a third arm that dosed 20 minutes after meal start, had a significantly higher average postprandial glucose vs. mealtime Humalog after 30 and 60 minutes, but the excursion was reduced to below mealtime Humalog levels by hour 2. At the end of 26 weeks, mealtime URLi demonstrated a non-inferior A1c reduction vs. Humalog (estimated treatment difference: -0.08%, 95% CI: -0.16-0.00), while the post-meal Humalog group had a significantly higher average A1c (estimated treatment difference: +0.13%, 95% CI: 0.04-0.22, p=0.003 vs. Humalog). In an interview with our team, Lilly Senior Medical Director Dr. Tom Hardy reminded us that PRONTO-T1D was a treat-to-target study, meaning that it was not designed to demonstrate A1c benefit over Humalog. He added that even the final A1c of the post-meal URLi group (7.42%) is pretty good for a multi-national study in type 1 diabetes.
On safety, there was no significant difference in postprandial level two hypoglycemia (<54 mg/dL) between the three groups up to four hours post-meal, though mealtime URLi did significantly reduce these events past the four hour mark (2.72 events per patient year vs. 4.35 with mealtime Humalog (p=0.001) vs. 3.88 with post-meal URLi (p=0.006)), which Dr. Klaff attributed to URLi’s faster onset and offset. Overall, there was no change in the rates of overall severe hypoglycemia or nocturnal severe hypoglycemia at 26 weeks. Taken together, he concluded that these results suggest that URLi is a superior agent to Humalog for postprandial control and hypoglycemia >4 hours after a meal and noted that URLi should only be dosed after a meal if needed. King’s College’s Dr. Stephanie Amiel appreciated this qualification during Q&A, stating that the “flexible dosing options” for Novo Nordisk’s Fiasp – which state that it can be dosed up to 20 minutes after a meal – have been a “complete disaster” for patients. Dr. Hardy noted to our team that Lilly will be submitting this data for a post-meal dosing indication, seeing as the hypoglycemia data and area-under-the-curve of the post-prandial glucose curves suggest URLi dosed at 20-min post-meal to be comparable to Humalog dosed at mealtime; he agreed that it should not be generally recommended.
We’ll be particularly intrigued to hear how thought leaders process these results and how Lilly positions URLi relative to Humalog (so far, it has only been filed in the EU and Japan, though US submission is slated for 2019). Thought leaders have historically been split over Fiasp’s added value vs. current mealtime options: Some feel the advantages it brings over NovoLog are merely marginal, while others have identified a unique potential for it in closed loop systems due to the molecule’s faster PK/PD profile. To this end, URLi was also tested in the PRONTO-PUMP study (completed September 2018, results expected at EASD 2019), and Lilly is now conducting the larger safety and efficacy PRONTO-PUMP-2 to support pump use and labeling in the US.
-
We wonder how Lilly will price Ultra Rapid Acting Humalog when it comes to market. Fiasp is currently priced at parity to progenitor NovoLog in most major markets (including the US). Will Lilly pursue a similarly aggressive strategy with URLi?
-
Within PRONTO-T1D there was a small CGM cohort (n=269), which found that mealtime URLi conferred ~44 minutes of extra time-in-range (71-180 mg/dL) vs. mealtime Humalog (p=0.02) between 6 am and midnight, with no difference in overnight time-in-range. To be expected, this was attributed to blunted post-meal highs, which were seen across all three meals (see below). In the grander scheme, we love this addition of CGM to the trial – we believe it more saliently demonstrates the benefits of URLi over Humalog, as well as those of premeal dosing over post-meal dosing.
5. PRONTO-T2D: URLi Non-inferior to Humalog on A1c Over 26 Weeks; Improvements in PPG Balanced Against Increase in Hypoglycemia (<54 mg/dL) between 1-4 Hours Post-Meal, but Low Event Rates
Dr. Thomas Blevins presented the full results of PRONTO-T2D (n=673), the type 2 equivalent of PRONTO-T1D, which again demonstrated that ultra-rapid-acting URLi was superior to Humalog in controlling 1- and 2-hour post-prandial glucose excursions during a mixed-meal-test through 26 weeks. URLi was also, again, non-inferior to Humalog on A1c at the end of the study, with the former group reaching an average A1c of 6.92% vs. 6.86% with the latter – both very well-controlled. While the overall rates of documented and nocturnal hypoglycemia (<54 mg/dL) were nearly identical, URLi conferred significantly higher rates of postprandial hypoglycemia between 1-2 hours after a meal (0.7 events per patient year vs 0.3, p<0.001) and between 2-4 hours post-meal (1.0 events per patient year vs. 0.7, p=0.04). Dr. Blevins did not offer a possible reason for these results but did note that the event rates were extremely low. We also add that there were fewer events of severe hypoglycemia with URLi than with Humalog, although the differences were not statistically significant. That said, the definition of hypoglycemia here corresponds with severe hypoglycemia (see the 2019 ATTD International Consensus Meeting on Time-in-Range), meaning that low event rates should not be dismissed as insignificant, in our minds. Also worth noting, there was a statistically significant increase in injection site reactions with URLi (2.7% of patients) vs. Humalog (0% of patients), though these event rates are also low. Altogether, we wonder whether these results will be enough to differentiate URLi from Humalog in the eyes of prescribers and patients – perhaps this is an area in which real-world-data or real-world-evidence could be used. Indeed, Lilly Senior Medical Director Dr. Tom Hardy noted to us that perhaps these strict, treat-to-target studies do not have the best structure to elucidate differences in rapid-acting insulins, at least between URLi and Humalog. He added that less structure in titration and frequency of HCP interaction, as well as more widespread use of CGM (yes!) for immediate feedback could be important in proving the differentiated profile of URLi to payers, prescribers, and patients, and Lilly is “definitely looking at the next slate of studies that might help us do that.”
-
In Q&A, one audience member asked whether a higher dose of Humalog may confer a similar postprandial glucose profile to URLi; Dr. Blevins believes it could, though at a higher risk of hypoglycemia. We certainly think the question is worth investigating, especially considering the elevated risk of hypoglycemia with URLi at all postprandial timepoints. It would also be very useful information to prescribers and HCPs should URLi be priced at a premium to Humalog.
-
The same audience member noted that the liquid diet used in a mixed-meal tolerance test favors ultra-rapid-acting insulin profiles, calling into question the significance of URLi’s benefits. Dr. Hardy acknowledged that the clinical trial situation is certainly artificial, noting that Lilly has other data in free-living individuals that suggests the postprandial benefits of URLi are still present with solid meals.
-
6. Tirzepatide (GIP/GLP-1) Dose-Escalation Study Shows Minimal (<5%) Discontinuation Due to Adverse Events with Slower Titration Schemes Compared to Phase 2b
A key poster (993-P) presented results from Lilly’s 12-week phase 2 titration study (n=111) for once-weekly GIP/GLP-1 dual agonist tirzepatide, indicating that slower titration meaningfully reduces the severity of GI side effects and treatment discontinuation with tirzepatide. Lilly had previously disclosed that this study achieved discontinuation rates <5% in every study arm, including 15 mg – which was the real sticking point from the phase 2b results presented at EASD 2018. Three different dose escalation algorithms were tested against placebo in people with type 2 diabetes, one to achieve a maintenance dose of 12 mg and two to achieve 15 mg (see schemes below). For comparison, in the main phase 2b study, participants started directly on the 5 mg dose for two weeks, then escalated to 10 mg for four weeks before climbing to 15 mg by week six (depending on group, of course). As shown in the table below, slowing down titration to 15 mg reduced (i) the proportion of patients experiencing GI side effects from 66% to 46%/57% and (ii) the frequency of treatment discontinuations due to adverse events from 25% to 0%/4%, depending on the algorithm. In the entire titration study, only 10 participants on tirzepatide discontinued treatment and only two due to adverse events. While GI events are still very common, it seems that the severity of events improves with slower titration. Per the dosing study’s abstract, GI adverse events were “mild to moderate” in intensity; in an interview with our team, Lilly’s Dr. Brad Woodward (Global Brand Development Lead and Senior Medical Director) explained that a higher proportion of events – about two-thirds – were mild in the titration study, compared to <half of those in the larger phase 2b.
Tolerability and Discontinuation Results: Phase 2b vs. Titration Study
Main Phase 2b Study |
12-Week Titration Study |
||||
Achieved Dose |
GI Event Prevalence (Nausea) |
Treatment Discontinuation (due to adverse events) |
Titration Algorithm |
GI Event Prevalence (Nausea) |
Treatment Discontinuation (due to adverse events) |
5 mg (n=55) |
33% (20%) |
15% (9%) |
#1 (12 mg, n=29) |
48% (24%) |
7% (3%) |
10 mg (n=51) |
51% (22%) |
14% (6%) |
#2 (15 mg, n=28) |
57% (39%) |
21% (4%) |
15 mg (n=53) |
66% (40%) |
34% (25%) |
#3 (15 mg, n=28) |
46% (36%) |
7% (0%) |
Placebo (n=51) |
10% (6%) |
18% (4%) |
Placebo (n=26) |
12% (8%) |
23% (4%) |
Trulicity 1.5 mg (n=54) |
43% (30%) |
15% (11%) |
|
-
Lilly is currently investigating the 5, 10, and 15 mg doses of tirzepatide for type 2 diabetes in the eight-trial phase 3 SURPASS program, which is using titration schemes informed by this dosing study and includes a superiority-powered CVOT. See an infographic outlining the chosen dose escalation schemes below. Additionally, phase 3 for obesity and phase 2 for NASH are set to begin in 2019. Importantly, Lilly plans to use the Trulicity autoinjector across these trials, which should help maximize adherence, and management has tentatively slated regulatory submission for 2022.
-
One key question can’t yet be answered: Will multi-step dose titration present a barrier to tirzepatide initiation and therapeutic dose achievement? Lilly has expressed its belief that packaging tirzepatide in the same autoinjector as Trulicity will make initiation and titration easier than with a standard pen, and that 5 mg itself would be an efficacious maintenance dose for most patients. As such, it would seem tirzepatide would be prescribed similarly to existing GLP-1s, though the escalation doses may be vulnerable to therapeutic inertia.
-
7. Highly Positive DUAL VIII Trial Finds Longer Sustainment of A1c Values <7.0% vs. Lantus with ~Half the Weight Gain, 56% Reduction in Severe or BG-Confirmed Hypo, and 30% Less Insulin
Brigham and Women’s Dr. Vanita Aroda revealed full 104-week results from the DUAL VIII trial (n=1,012 insulin-naïve people with type 2 diabetes), which found significantly greater durability of treatment effect alongside vast improvements in insulin dose, weight, and hypoglycemia with Novo Nordisk’s GLP-1/basal insulin combo therapy Xultophy (liraglutide/insulin degludec) vs. Sanofi’s basal insulin Lantus (insulin glargine) – full results and an accompanying editorial were simultaneously published in Lancet Diabetes and Endocrinology. At the end of the 104-week study period, just 37.4% of those taking Xultophy met the requirements for treatment intensification (defined as back-to-back A1c measurements ≥7.0%) compared to 66.2% of those on Lantus (p<0.0001), though both values were steadily rising. Among those reaching the 104-week mark with an A1c <7.0%, 51.8% on Xultophy did it with no hypoglycemia vs. 25.5% with Lantus (p<0.0001), 20.9% on Xultophy did it with no weight gain vs. 6.3% on Lantus (p<0.0001), and 20.0 % on Xultophy did it with no hypoglycemia or weight gain vs. 6.1% on Lantus. Daily insulin dose was cut down by almost 30% in those taking Xultophy (37 units/day vs. 52 units per day), and Xultophy also conferred half as much weight gain vs. Lantus (1.7 kg vs. 3.4 kg, p<0.0001). Very impressively, there was a 56% lower rate of severe or blood glucose-confirmed symptomatic hypoglycemia with Xultophy compared to Lantus (ERR=0.44, 95% CI: 0.33-0.60, p<0.0001). Altogether, these results paint a stunning picture of the sheer power of GLP-1/basal combinations, particularly Xultophy, as first-line injectables (as a reminder, Dr. John Buse has previously called these agents, “without a doubt the most powerful glucose-lowering agents on the planet”). While it may not have been a fair fight from the get-go – pitting two highly efficacious agents in combination vs. a single, first-gen basal insulin – we do believe these results make a strong case for earlier and more aggressive treatment with GLP-1/basal combinations. As a reminder, FDA earlier this year removed the label restriction requiring that patients already be taking basal insulin or liraglutide before starting Xultophy, meaning that first-line treatment with this agent is just now possible in the US (this restriction never applied OUS). It’s great from our view that this label is gone though memories are long.
-
We continue to be impressed by the clinical performance of Xultophy and other GLP-1/basal insulin combinations but note that commercial uptake has lagged significantly behind. Building on optimal clinical outcomes, a DUAL VII post-hoc analysis at EASD 2018 found that Xultophy treatment was less burdensome to patients than basal-bolus therapy, in terms of dose adjustments, number of injections per day, and total daily insulin dose. Yet, despite these multifactorial benefits, neither Xultophy or Soliqua has truly taken off in the market, particularly in the US. In a conversation with our team, Novo Nordisk CMO in North America Dr. Todd Hobbs confessed his belief that it “might be too late in the US to rebound … once you have a tough product launch, it can be difficult to bounce back.” That said, Xultophy is doing quite well in several countries in the EU, such as France, where reimbursement is strong and clinical algorithms are different.
-
In Q&A, after complimenting the researchers on a job well done, the venerable Dr. Julio Rosenstock noted that this study does not answer a key remaining question: Does simultaneous administration of GLP-1/basal insulin outperform sequential administration? This issue was also brought up in the paper’s accompanying editorial (which we highly recommend), though it is not a new concept. In fact, Dr. Rosenstock has previously contributed to formal debates on the subject, with consensus being that a dedicated RCT will be needed to draw a final conclusion. Perhaps such a study would also improve the commercial outlook for the class (provided simultaneous outperforms sequential), as it is our understanding that many providers are more comfortable prescribing and titrating each agent separately, rather than in a fixed-ratio.
-
Given that its primary purpose was to measure durability, DUAL VIII was designed to mirror clinical practice. Clinic visits occurred just once every 12-14 weeks, vs. the standard 4-6 week visits present in most RCTs, perhaps giving the study greater generalizability to a real-world setting.
8. BRIGHT Subgroup Analysis Suggests Potentially Improved A1c Lowering vs. Tresiba at Lower Renal Function (eGFR ≤60) with No Increase in Hypoglycemia, but with Small Number of Participants
Institute for Clinical and Experimental Medicine’s Dr. Martin Haluzik expounded on Sanofi’s BRIGHT study comparing next-gen insulins Toujeo and Tresiba, uncovering improved A1c values with the former as renal function declined. Originally presented at ADA 2018, BRIGHT found overall similar A1c reductions with the two agents, but a predefined subgroup analysis found significantly different mean A1c changes favoring Toujeo at lower baseline eGFR (p-value for interaction=0.02, see below), prompting this renal deep dive post-hoc. Once broken down by eGFR, there was an increasingly divergent pattern in mean A1c favoring Toujeo at reduced kidney function over the 24-week study. At normal kidney function (eGFR ≥90), the curves nearly overlapped (least-squares mean difference = +0.09% for Toujeo, 95% CI: -0.05 to 0.24), but at stage 2 CKD (eGFR 60 to <90) Toujeo began to prevail (LS mean difference = -0.13% for Toujeo, 95% CI: -0.30 to 0.02), and at stage 3 CKD, the 95% CI no longer crosses unity, favoring Toujeo (LS mean difference = -0.43, 95% CI: -0.74 to -0.12). Importantly, this benefit was not the result of increased hypoglycemia, as there was no discernible difference in incidence or rate of hypoglycemia ≤70 mg/dL or ≤54 mg/dL over the full study period, regardless of renal function (see below). Dr. Haluzik did not offer any hypotheses behind Toujeo’s potential benefit, besides the surface-level, “could be explained by some differences in insulin characteristics,” noting that a dedicated trial in this population would be needed first to confirm the difference.
-
Overall, we think too much time is spent splitting hairs between these next-gen basal insulins. Instead, we should, as a community, focus on bringing them to more patients. As Dr. Alice Cheng said at EASD 2018, “I don’t think the big takeaway from BRIGHT is that it demonstrated non-inferiority in terms of A1c reduction between the two next-gen insulins. The big takeaway is that both insulins were able to lower A1c from an average above 8.5% to 7.0% in just 24 weeks.” Notably, it also seems that the conclusion about which insulin is better depends significantly on the trial sponsor: While DELIVER D+ and BRIGHT found no significant difference between the two in hypoglycemia past the first 12 weeks, the CONFIRM real-world study (n=4,056) suggested that hypoglycemia and treatment discontinuation were both more common with Toujeo than Tresiba, and a Novo Nordisk-sponsored, phase 3b head-to-head trial (n=1,609) of the two agents, which was announced on the company’s 1Q19 earnings call, also found overall lower hypoglycemia risk with Tresiba.
-
As with any stratified analysis, cohort numbers were stretched thin. 467 (50%) of all participants had a baseline eGFR ≥90 ml/min/1.73 m2 (246 Toujeo, 221 Tresiba), 365 (39%) had a baseline eGFR between 60 and 90 ml/min/1.73 m2 (172 Toujeo, 193 Tresiba), and just 96 (11%) had a baseline eGFR <60 ml/min/1.73 m2 (47 Toujeo, 49 Tresiba). With numbers this small, it becomes difficult to draw definitive conclusions, rendering these results purely hypothesis-generating.
9. Dr. Elizabeth Mayer-Davis Highlights the Synergy between Observational Epidemiology and Clinical Trials in Kelly West Award Lecture
University of North Carolina’s Dr. Elizabeth Mayer-Davis (UNC Chapel Hill) received this year’s Kelly West Award, which honors outstanding achievement in diabetes epidemiology. In her award lecture, Dr. Mayer-Davis explained that her career has been defined by synergy between observational epidemiology and clinical trials. She specifically highlighted the landmark SEARCH for Diabetes in Youth study, whose epidemiological insights directly inspired three of the clinical trials she has led: The Tribal Turning Point study to prevent obesity and type 2 diabetes in Native American youth, the FL3X study of behavioral interventions for adolescents with type 1 diabetes, and a forthcoming study on weight regulation in type 1 diabetes by ACT1ON (Advancing Care for Type 1 Diabetes and Obesity Network). We applaud Dr. Mayer-Davis for this achievement, and her enormous contributions to research across the spectrum of diabetes care.
-
The Tribal Turning Point study was borne from the SEARCH finding pointing to a disproportionately large increase in the incidence of adolescent type 2 diabetes within the Native American community. In order to assess whether diabetes education can help break this troubling trend, Tribal Turning Point enrolled Cherokee and Navajo youth (n=62) at risk for developing diabetes, exposing them to a DPP-like intervention emphasizing education on healthy eating and physical activity in a culturally-sensitive manner. The intervention produced a significant beneficial effect after just an 8 month pilot period, yielding a significant decrease in BMI and waist circumference.
-
The FL3X study aims to address poor glycemic control in the pediatric population. It was inspired by an expansion of the SEARCH study, SEARCH 3, which unveiled the presence of diabetes complications among youth with diabetes (both type 1 and type 2) – including DKD, retinopathy, peripheral neuropathy, cardiac autonomic neuropathy, arterial stiffness, and hypertension. In order to assess whether behavioral intervention can improve outcomes for adolescents with poorly-controlled diabetes at risk of developing such complications, FL3X exposed adolescents to a diabetes self-management education program with an emphasis on problem solving skills training and motivational interviewing. Despite near-perfect participant retention (97%) and very high-quality data collection and follow-up, there was no improvement in A1c with the behavioral intervention after 18 months. However, a follow-up study investigating participants’ goal-setting behaviors revealed that participants who set CGM-related goals experienced a 1.3% drop in A1c by the end of the trial, vs. a 0.3% rise in A1c for those who set blood glucose-related goals. This hints that better integration of CGM into self-management education could help people in this age group get more out of this training.
-
ACT1ON’s forthcoming study is designed to establish a scientific foundation for behavioral interventions that improve glycemic control and weight management for young adults with type 1 diabetes – who, as SEARCH data reveals, have the same high rates of overweight and obesity as the general, non-diabetes population. Overweight and obesity confer further CV risk in addition to the existing elevated risk from type 1 diabetes, so interventions that addresses both weight management and diabetes management are sorely needed. ACT1ON’s 18-month pilot study will be two-fold: One aim is to use metabolic chamber techniques to evaluate energy balance and flexibility in type 1 diabetes to inform caloric intake goals; the other is to inform potential counseling strategies, particularly around dietary responses to hypoglycemia.
Diabetes Technology Highlights
1. Terrific Tandem Control-IQ Pivotal Trial: +2.6 Hour/Day Gain on TIR and 0.3% A1c Advantage vs. SAP (t:slim X2/G6); Staggering 92% of Six Months Spent in HCL! FDA Filing in “Coming Weeks”, Launch This Year
To a completely packed room, UVA’s Dr. Sue Brown shared extremely positive results from the six-month, NIH-funded iDCL pivotal trial comparing Tandem’s Control-IQ hybrid closed loop/Dexcom G6 CGM (n=112) vs. sensor-augmented pump (t:slim X2 pump/G6 with no automation; n=56). In participants 14-71 years old, Control-IQ drove tremendous advantages over the tough SAP comparator on every efficacy endpoint at six months: Time-in-range (70-180 mg/dl) was 2.6 hours per day better with Control-IQ – 70% vs. 59% (p<0.0001) – with most of the benefit coming from less time >180 mg/dl (-2.4 hours/day). Control-IQ had a 0.33% advantage in A1c at six months (baseline: 7.4%; p=0.0014), and mean CGM was 13 mg/dl lower with Control-IQ by the end of the study (156 vs. 170 mg/dl; p<0.001). Time <70 mg/dl was low at <2% in both groups, though still a smidgeon better with Control-IQ (-13 minutes/day). Remarkably, all 168 participants completed the randomized study, and Control-IQ users spent a remarkable 92% (!) of the full six months with closed loop active – a stunning result highlighted by both Drs. Boris Kovatchev and Roy Beck in Q&A.
The time-in-range benefits of Control-IQ were realized in the first month, sustained over six months, and extended across the spectrum of baseline A1c levels (5.4%-10.6%) – the 24/7 automated basal insulin delivery (especially overnight) and automated correction boluses drove improvements for a broad population. Notably, 20% of the Control-IQ study group was on MDI and 30% were not on CGM starting the study.
Control-IQ achieved near-perfect scores on a technology acceptance questionnaire: ease of use was 4.7/5, usefulness was 4.6/5, trust was 4.5/5, and desire to continue using was an impressive 4.8/5 – an encouraging sign of the system’s simplicity (no modes to juggle), the no fingersticks G6, and limited alarms. Indeed, use of the Dexcom G6 was impressively high in this study – 96% over six months in the SAP group and 97% in the Control-IQ group – and fingersticks were just 0.37/day with SAP and 0.21/day with Control-IQ (i.e., one fingerstick every 3-5 days).
There were no severe hypoglycemia events and one DKA event in the Control-IQ group due to infusion site failure. The study even deployed a software update in March (Tandem Device Updater) to fix a minor issue, with no impact on outcomes.
All in all, this was a home run study for Tandem, Dexcom, NIH, and the entire closed loop field, especially with the robust RCT design, the real-world inclusion criteria (no entry restrictions on A1c, severe hypo or DKA, or device experience), and six-month duration. There was a real sense in hallway conversations of this study being a landmark moment for the diabetes technology field, with some investigators comparing it to the JDRF CGM trial. Key outcomes are summarized immediately below, followed by notable quotes, a deep dive on all presented study outcomes with baseline levels included, and a comparison to the 670G pivotals. Pictures were not allowed, so we’ve summarized the data in tabular form.
-
On a call this afternoon, Tandem confirmed Control-IQ will be submitted to the FDA in the “coming weeks” (14+ years), with US launch expected “this year, subject to FDA approval.” This is on par with the 1Q19 call, which more specifically called for a launch in “4Q19.” The ~10-minute call summarized the data in prepared remarks – see the press release – with no Q&A taken (“to maximize opportunity for future publications”). The next update will come at the end of July in Tandem 2Q19. Given the positive outcomes, we don’t expect major surprises on the regulatory review, especially with two of the system’s three components already FDA cleared – G6 iCGM and t:slim X2 ACE Pump. Of note, Tandem had no influence on the design of this NIH-funded study or the presentation today.
Efficacy Outcome |
SAP at |
Control-IQ at |
Adjusted Difference |
Time-in-Range |
59% |
71% |
+11% |
A1c |
7.4% |
7.1% |
-0.33% |
Time >180 mg/dl |
38% |
27% |
-10% |
Mean CGM |
170 mg/dl |
156 mg/dl |
-13 mg/dl |
Time <70 mg/dl |
1.9% |
1.4% |
-0.88% |
Time <54 mg/dl |
0.24% |
0.21% |
-0.1% |
-
Dr. Roy Beck in Q&A: “I’ve been involved in running clinical trials for almost 35 years. We’ve never had a six-month trial with 100% treatment retention – both in the control and intervention groups. Time-in-closed loop exceeded what we thought – 92% over six months. Think about all the times when someone is stopping closed loop because they are ill, etc. This is amazing data. We coordinated the JDRF CGM trial, and that data set is still used a lot. I look at this as a landmark study like that. The investment by NIH, our taxpayer dollars here in the US, has been incredibly well spent.”
-
Dr. Boris Kovatchev in Q&A: “I propose a new Data Awesomeness Index. It is study retention rate plus closed loop use plus the difference between time-in-range between control and experimental. In this study, that is 100% + 92% + 11% = 203%. That is above 200%, which is really awesome – a new benchmark for everyone who does studies from now on.”
-
Dr. Aaron Kowalski in Q&A: “Congrats to this amazing team. I was co-organizer of that [NIH/JDRF/FDA artificial pancreas] meeting in 2005, and I’m proud of the progress in this field – the incredible efforts and Boris’ leadership. The proof is in the pudding: people will benefit – you see the overnight data, the reduction in hypoglycemia and hyperglycemia. The benefit to the community will be great. I want to congratulate Tandem and Dexcom, and the competition in this space will provide more options that drive better outcomes.”
-
Gary Scheiner in Q&A: “I can’t wait to start using this personally. Your last slide, on ease of use parameters, really struck me…”
Results
-
Time in range with Control-IQ over six months was 71% vs. 59% with SAP alone (p<0.0001) – an 11% difference or 2.6 hours per day. The difference in time-in-range outcomes was seen immediately at one month and was sustained through the entire six-month study. The nighttime advantage on time-in-range was especially pronounced: 76% with Control-IQ vs. 59% with SAP (p<0.0001); the algorithm is designed to intensify control overnight, pushing users towards 110-120 mg/dl by morning. Indeed, Control-IQ achieved over 90% time-in-range between 4-8am, with a very narrow confidence interval. During the day, Control-IQ also consistently delivered higher time-in-range.
Time-in-Range – 70-180 mg/dl
|
Baseline |
26 Weeks |
Adjusted Difference |
SAP |
59% |
59% |
+11% |
Control-IQ |
61% |
71% |
-
Impressively, the benefits on time-in-range extended to every level of baseline A1c– a positive sign that this system can help those already at a low A1c, those around 7%, and those at >8%. Of note was the +21% time-in-range advantage in the middle (7.1%-7.5% A1c) group, a staggering five-hour/day improvement with Control-IQ vs. SAP!
Time-in-Range by Baseline A1c
|
A1c |
A1c |
A1c |
A1c |
A1c |
SAP |
78% |
69% |
49% |
56% |
47% |
Control-IQ |
85% |
76% |
71% |
69% |
60% |
Difference |
+7% |
+7% |
+21% |
+13% |
+13% |
-
Nearly all of the benefit on time-in-range came from less time spent >180 mg/dl – a 2.4 hour/day advantage for Control-IQ vs. SAP. As expected with reduced hyperglycemia, mean CGM glucose was 13 mg/dl lower with Control-IQ. Time >250 mg/dl was not reported.
Hyperglycemia
|
Baseline |
26 Weeks |
Adjusted Difference |
SAP |
38% |
38% |
-10% |
Control-IQ |
36% |
27% |
Average Glucose
|
Baseline |
26 Weeks |
Adjusted Difference |
SAP |
169 mg/dl |
170 mg/dl |
-13 mg/dl |
Control-IQ |
166 mg/dl |
156 mg/dl |
A1c
-
From a starting baseline A1c of 7.4%, Control-IQ drove a 0.3% A1c reduction, compared to no difference in the SAP group.
|
Baseline |
26 Weeks |
Adjusted Difference |
SAP – A1c |
7.4% |
7.4% |
-0.33% |
Control-IQ – A1c |
7.4% |
7.1% |
-
Hypoglycemia was remarkably low in both groups in this six-month study – far below the consensus goal for <4% - highlighting how well participants were doing with G6 and t:slim X2 – with or without automation. Even still, Control-IQ did have a slight (statistically significant) advantage on both time <70 (-13 minutes per day) and time <54 (-1 minute per day).
Hypoglycemia
|
Baseline |
26 Weeks |
Adjusted Difference |
SAP |
2.1% |
1.9% |
Time <70: -0.88% Time <54: -0.1% |
Control-IQ |
2.7% |
1.4% |
Algorithm
-
TypeZero’s Control-IQ hybrid closed loop algorithm uses basal rate modulation combined with automated correction boluses to increase time-in-range (70-180 mg/dl). It uses a combination of targets and target ranges, none of which can be changed by the user. The algorithm specifics by category are listed below. At a high level, the target is to keep people between 70-180 mg/dl as much as possible – it adjusts the basal rate when the prediction is above 160 mg/dl, boluses above 180 mg/dl, decreases insulin below 112.5 mg/dl, and turns off basal below 70 mg/dl. These adjustment ranges are lowered and narrowed when sleeping (target 112.5-120 mg/dl by morning), and raised and narrowed when exercising. Users are still expected to meal bolus. The algorithm requires all the normal pump settings: basal profile, insulin:carb ratio, insulin sensitivity factor; as we understand it, all those parameters influence the system’s aggressiveness. Dr. Sue Brown mentioned that after the initial publication of results, the team does intend to write about the algorithm parameters.
-
Automated Basal Rate Modulation during the day (including after meals): the system targets a range of 112.5-160 mg/dl.
-
Automated Basal Rate Modulation at night: the algorithm becomes more aggressive and tries to arrive at 112.5-120 mg/dl by morning.
-
Automated Bolus Corrections: these automated corrections can occur up to once an hour during the day and are calculated to give approximately 60% of a correction bolus with a target of 110 mg/dl. These boluses occurs on top of the basal rate modulations described above.
-
User Corrections: when a person requests a correction bolus through the usual bolus screen, that gets calculated in the typical manner using the user-defined settings for correction factor with a target of 110 mg/dl.
-
Exercise Mode: there is a button that the user can push if they wish that changes the target range to be closer to 140-160 mg/dl.
-
- In Q&A, Dr. Brown mentioned that in training MDI users on the system, the biggest piece was teaching them about the t:slim X2 pump – not the algorithm. “Once you understand the pump, the actual training for Control-IQ is to hit a button. How the individual interacts with the system is done in a very similar way to how a normal pump would be.” We look forward to hearing more about the user experience of wearing the pump.
Baseline Characteristics and Study Details
-
The $12.7 million NIH-funded DCLP3 study – Grant UC4 DK 108483 with P.I. Boris Kovatchev – is the first large-scale, six-month closed-loop study that included a dedicated control group. There were no exclusion criteria based on A1c, history of complications, or previous experience using an insulin pump or CGM. Following a 2-8 week run-in period, the length of which was determined based on previous pump and CGM experience, 168 patients with type 1 diabetes ages 14 and up were randomized 2:1 to Control-IQ technology (n=112; t:slim X2 with embedded TypeZero Control-IQ algorithm talking to Dexcom G6 transmitter) or sensor-augmented pump therapy (n=56; t:slim X2/G6 without automation) and followed for 26 weeks. The entry A1c for participants in the study ranged from 5.4%-10.6% with a mean of 7.4%. All participants completed the study.
|
SAP N=56 |
Control-IQ N=112 |
Mean age |
33 years |
33 years |
Gender |
54% Female |
48% Female |
Mean A1c |
7.4% |
7.4% |
MDI User |
23% |
20% |
Current CGM User |
71% |
70% |
Median T1D Duration |
15 years |
17 years |
Comparison to 670G Pivotal Trials
-
The slides below taken from ATTD 2019 nicely summarize key CGM outcomes from the MiniMed 670G pivotal trials. It’s important to note that these studies were very different – Medtronic tested the MiniMed 670G by comparing two weeks of open loop (Manual Mode) to three months of Auto Mode use within the same person; Control-IQ was parallel group randomized trial comparing open loop in one group to closed loop in another group for six months each. Control-IQ showed a larger benefit on time-in-range at +11% vs. +5% for the 670G in 14+ years, though Control-IQ participants had a lower baseline time-in-range. Both systems ended up around the same overall time-in-range in 14+ years (71% vs. 72%). Control-IQ had higher time in closed loop (92%) in a study that was twice the length. The 670G drove a slightly larger 0.5% A1c reduction from a similar 7.4% baseline.
2. 6-Month WISDM RCT in Type 1s ≥60 Years Shows CGM is Superior to BGM in Every Outcome Measured: -27 Minute Between-Group Difference in Time <70 mg/dl, +2.1 Hour Advantage on Time in Range
AdventHealth Diabetes Institute’s Dr. Richard Pratley presented the highly anticipated results from the Helmsley- and JDRF-funded WISDM study (n=203) evaluating the effect of Dexcom G5 CGM in older adults (≥60 years) with type 1 diabetes. The six-month, multicenter, randomized controlled trial found CGM conferred a significant advantage over BGM in every outcome measured. To be included in the study, participants had to have a baseline A1c ≤10%, be pump- or MDI-treated, and could not have worn CGM in the last three months. Patients were excluded if they spent at least 10% of the time with glucose <54 mg/dl during the screening phase and experienced a severe hypoglycemic event in the past six months. The BGM group wore blinded CGM (Dexcom G4 Pro) at baseline, 8-, 16-, and 26-weeks. For the primary outcome, time <70 mg/dl, the adjusted treatment group difference for the CGM group was -1.9% (p<0.001), amounting to a difference of 27 minutes/day. Dr. Pratley highlighted that the decrease in time <70 mg/dl occurred for the CGM group by week eight and was maintained throughout the study, while the BGM group showed no significant changes over the study duration. Splitting the results out by pump vs. MDI users, who were divided equally between the two groups, those in the CGM group showed significant decreases in time <70 mg/dl regardless of insulin delivery method – another big win for CGM, building on the evidence base that CGM is effective in both pump and MDI users (e.g., COMISAIR, DIAMOND, GOLD). There was also a significant between-group difference in time <54 mg/dl, with the CGM group showing an adjusted difference of -1.0% (-14 minutes/day) (p<0.001). Once again, this significant improvement was achieved regardless of insulin delivery method. Time-in-range (70-180 mg/dl) improved significantly for the CGM group, with an adjusted difference of +9% (+2.1 hours/day) (p<0.001). Time >180 mg/dl and >250 mg/dl also decreased for the CGM group, with adjusted differences of -6% (-1.4 hours/day; p<0.001) and -4% (-58 minutes/day; p<0.001), respectively. Accordingly, A1c significantly improved in the CGM group, dropping by 0.4 percentage points (baseline A1c: 7.6%) and amounting to an adjusted difference of -0.3% (p<0.001). See the full data tabulated below. The BGM group reported a significantly greater number of participants (10) than the CGM group (1) with a severe hypoglycemia event, defined as requiring assistance of another person – wow! We’re thrilled by the strong results and hope that they serve to quell some misconceptions surrounding use of technology/CGM and the elderly. We also wonder if these data could help Medicare reduce the burdensome restrictions around fingerstick documentation prior to obtaining a therapeutic CGM.
|
|
CGM |
BGM |
Adjusted Difference (95% CI) |
Time <70 mg/dl |
Baseline |
5.1% (1.2 hours) |
4.7% (1.1 hours) |
-1.9% |
Follow-up |
2.7% (39 mins) |
4.9% (1.2 hours) |
||
Time <54 mg/dl |
Baseline |
1.9% (27 mins) |
1.5% (22 mins) |
-1.0% |
Follow-up |
0.5% (7 mins) |
1.6% (23 mins) |
||
Time in Range (70-180 mg/dl) |
Baseline |
56% (13.4 hours) |
56% (13.4 hours) |
9% |
Follow-up |
63% (15.1 hours) |
54% (13.0 hours) |
||
Time >180 mg/dl |
Baseline |
37% (8.9 hours) |
38% (9.1 hours) |
-6% |
Follow-up |
34% (8.2 hours) |
39% (9.4 hours) |
||
Time >250 mg/dl |
Baseline |
14% (3.4 hours) |
15% (3.6 hours) |
-4% |
Follow-up |
10% (2.4 hours) |
16% (3.8 hours) |
||
A1c |
Baseline |
7.6 |
7.5 |
-0.3 |
Follow-up |
7.2 |
7.4 |
||
Mean Glucose |
Baseline |
167 |
168 |
-8 |
Follow-up |
162 |
171 |
||
Coefficient of Variation (CV) |
Baseline |
41% |
42% |
-5% |
Follow-up |
37% |
42% |
-
The CGM group achieved a significant decrease in weekly hypoglycemic events, defined as 15 consecutive minutes with a sensor glucose value <54 mg/dl. The end of the event was defined as 15 consecutive minutes with a sensor glucose ≥70 mg/dl. The CGM group declined from 2.6 events/week to 0.8 events/week, while the BGM group decreased slightly from 2.1 events/week to 1.8 events/week. The adjusted difference was -0.9 (p<0.001).There were no treatment group differences for adverse events, and no device-related adverse events reported.
-
Dr. Pratley found it “really assuring” to see that most patients in the CGM group (81%) used CGM seven days/week. Additionally, by week eight, 83% of participants used CGM to make insulin dosing decisions, which Dr. Pratley interpreted as a positive indication of patients’ trust in and utilization of the system. Surprisingly, there were no treatment group differences for any of the patient-reported outcomes assessed, including fear of hypoglycemia, diabetes distress, hypoglycemia unawareness, and general measures of quality of life. To this end, Dr. Pratley suggested that perhaps given the patients’ lengthy diabetes duration, “they’ve gotten used to it.” Even so, we still would’ve expected the PRO improvements typically seen in other CGM studies to apply to this population.
-
Roughly 30% of participants in the CGM group used the Dexcom G5 mobile app. While Dr. Pratley was not surprised by the low utilization, as many of the participants owned older flip phones, he was a little surprised that just ~10% of participants used the Share function. As he put it, “this is a pretty fragile population” that would likely benefit from monitoring. However, many of the participants explained that they preferred to be independent, as they had been for the majority of their diabetes management. We wonder if (and to what extent) CGM outcomes could’ve been further improved if more of the participants had turned “Share” on. That said, there was only one severe hypo in the CGM group, so it’s unclear how tangible the benefit would’ve been. We’d be interested in seeing how the outcomes of the 10% who allowed remote monitoring compared to the 90% who didn’t turn Share on.
-
Prior to the WISDM data readout, Northwestern’s Dr. Grazia Aleppo argued that technology has too many challenges for older adults. Her view was not that technology isn’t effective in the elderly; rather, she pointed out design flaws that make certain devices challenging, and in some cases, impossible for older patients to use. Given that older patients experience cognitive declines, visual and hearing impairment, and reduced dexterity, Dr. Aleppo called for simpler tools requiring fewer steps (i.e., automatic pump priming), large fonts, adjustable alarm volume, and training protocols for specific impairments. She also asserted that long-term facility staff have very limited knowledge of diabetes technology and should receive training.
3. Omnipod Horizon AID Very Strong in 2-6 year-old, ~3 Day Hotel Study: +4 Hr/Day TIR, No Sig Change in Hypo; 3 Investigational Horizon Features – extended bolus, variable setpoint (100-150 mg/dl), Hypo Protect; U500 Omnipod with FDA
Stanford’s Dr. Bruce Buckingham presented very strong data from a study of Insulet’s Omnipod Horizon closed loop in very young children ages 2-6 years old (n=14; mean age 4 years; baseline A1c: 7.4%). The free-living, hotel study – including ample physical activity and meal challenges – compared 48-72 hours of hybrid closed loop with a tablet computer, Dexcom sensor (G4 with software 505) to seven days of patients’ standard therapy. Horizon use resulted in an incredible 4.2 hour/day increase in time in the range of 70-180 mg/dl (73% vs. 55%; p=0.0002). This effect was driven by a dramatic reduction in hyperglycemia, with 3.6 fewer hours/day spent >180 mg/dl (25% vs. 40%; p=0.005) and a remarkable 2.7 fewer hours/day spent ≥250 mg/dl (6% vs. 17%; p=0.002). This amounts to a ~65% reduction in time ≥250 mg/dl, albeit from a high baseline of four hours! Time in hypoglycemia was also reduced, albeit not significantly from a statistical perspective due to the small number of participants. Numerically, time <70 mg/dl fell by a half hour per day with Horizon (5% vs. 3%; p=0.24) and time ≤54 mg/dl was cut by ~20 minutes/day (1.8% vs. 0.4%; p=0.15), a 78% reduction. Mean glucose with Horizon was significantly lower than that on standard care (172 mg/dl vs. 148 mg/dl; p=0.017). Overall, coefficient of variation (CV) dropped from 40% to 36% (p=0.02) – hitting the often-used CV goal of ≤36% (Monnier). Overnight, CV dropped from 37% to a remarkable 25% (p=0.0009) – wow! The modal day chart below clearly shows flatter, narrower, and more in-range glucose values with Horizon, and Dr. Buckingham commented repeatedly that the system performed very well. He shared that a parent stayed with each participant overnight, and when they woke up in the morning, a couple of the mothers were in tears because it was the first night since their child was diagnosed with type 1 that they slept through the night; and when they woke up, the child had a flat, in-range blood sugar. Impressively, though the core MPC algorithm didn’t undergo any major tweaks to better fit the 2-6-year-old population, it performed essentially the same in this age group as it has in the previous studies. Horizon – with a tubeless patch pump, G6 sensor, and algorithm running on the pod – should be very appealing in this age cohort when it comes to market. The persistent hyperglycemia and variability is tough in this population, especially with the fear of hypoglycemia (just glance at the baseline data). We imagine Horizon will be a differentiated system on ease of use that can safely keep glucoses lower without infringing significantly on a kid’s ability to be a kid. We’re not sure how many of the participants were on MDI/pumps at baseline or how much time they spent in closed loop. Future studies, according to Dr. Buckingham, will evaluate Horizon in the setting of truly free-living (algorithm on the Omnipod, not a tablet) and extended use – this presumably refers to the pivotal study set to begin in 4Q19 ahead of a 2H20 launch.
-
Insulet strove to challenge the system and emulate real-world use with ample physical activity (47 sessions total) and meals (115 total). Exercises included parks, ropes courses, laser tag, a farm visit, trampoline, and tandem bike riding (“Kids were in a bouncy house, bouncing off each other, off each other’s Omnipods”) with an average duration of 56 minutes. Meals were unrestricted, with an average of 31 grams of carbs (max 81 grams), and 29% of meals had ≥15 grams of fat.
-
Medtronic presented results from the MiniMed 670G pivotal study in 2-6 year-olds over three months of use (n=46) at ATTD. The studies are not directly comparable – e.g., due to the difference in trial length and setting – but in the 670G study, A1c declined by 0.5% (baseline: 8.0%), time-in-range increased two hours/day (from 55% to 64%), and there was no change in time <70 mg/dl (from 3.6% to 3.5%). Mean glucose declined from 173 mg/dl (run-in) to 161 mg/dl (670G), with a small increase in CV ( from 38% to 39%).
-
Dr. Buckingham discussed a number of Horizon closed loop features that we had not previously recalled hearing – Dr. Ly later underscored that this was an investigational product and not all features will necessarily be in the commercial version:
- Setpoint (target glucose) can be set at 100, 110, 120, 130, 140, or 150 mg/dl. Parents decided to raise the setpoint to 150 mg/dl for 79% of the exercise sessions in the study. During a subsequent Insulet-sponsored dinner event, Dr. Buckingham said that the system was “tested mostly at 120 mg/dl. We also tested down to 110 mg/dl. But you can set it at 150 mg/dl, and we tested that just to see what would happen if you left it there.” The commercial version is likely to have a lower setpoint limit of 110 mg/dl.
-
It allows for extended boluses. Four participants in the study used ≥1 extended bolus for meals – mostly around lunchtime – as decided by their parents.
-
It has a “hypoglycemia protect” mode (we’re not sure if this is commercial naming) that sets the target glucose at 150 mg/dl and cuts basal delivery either to 75%, 50%, or 25%. This feature sounds similar to Diabeloop’s “Zen Mode,” which “virtually ensures” hypo avoidance over a short period of time.
- Setpoint (target glucose) can be set at 100, 110, 120, 130, 140, or 150 mg/dl. Parents decided to raise the setpoint to 150 mg/dl for 79% of the exercise sessions in the study. During a subsequent Insulet-sponsored dinner event, Dr. Buckingham said that the system was “tested mostly at 120 mg/dl. We also tested down to 110 mg/dl. But you can set it at 150 mg/dl, and we tested that just to see what would happen if you left it there.” The commercial version is likely to have a lower setpoint limit of 110 mg/dl.
-
Dr. Ly confirmed that the U500 Omnipod is now under FDA review. This was foreshadowed during the 1Q19 call. January’s JPM talk expected a launch by the “end of 2019/early 2020.”
Previous Omnipod Horizon Safety and Feasibility Studies
|
|||
Duration (hr) |
36 |
54 |
96 |
Setting |
Inpatient |
Hotel – meal and exercise challenges |
Hotel – supervised free-living conditions |
Time in range (70-180 mg/dl) |
70%-73% |
76%-85% |
69%-79% |
Time <70 mg/dl |
≤2% |
~1% |
≤2.5% |
4. FreeStyle Libre 2 Product Theater Offers First Look on US Soil, First Discussion of Alarms; No Update on FDA iCGM Review
Abbott’s Dr. Alexander Seibold (Regional Medical Director, EMEA) provided the first public look at FreeStyle Libre 2 on US soil, covering the system’s optional alarms (via Bluetooth), accuracy, and experience in Germany following the CE Mark last October. FDA 510(k) clearance as an iCGM did not come through before ADA, so the talk had the usual disclaimers about being a non-cleared product in the US. Even still, we learned several new tidbits about the product: (i) FreeStyle Libre 2 is currently available in Germany and Norway (the latter was new to us); (ii) on par with expectations, FreeStyle Libre 2 now works with either the reader or the FreeStyle LibreLink app, meaning alarms appear on the phone lock screen; (iii) the CE Mark for 4+ years includes pregnant women; (iv) the accuracy of Libre 2 in Europe is reported from the same study that supported the 14-day FreeStyle Libre approval in the US (MARD of ~9.5%), but this is the first accuracy improvement in Europe since the original FreeStyle Libre launched. Accuracy was discussed at a fairly high level (see below) and was not bucketed into all the iCGM special controls thresholds; per 1Q19, Abbott is confident the system meets the bar. (We’re surprised the two studies – n=95 adult and n=74 pediatric – were of sufficient size to get tight enough confidence intervals for the special controls; however, it’s possible Abbott has done more studies for the FDA submission.) See below for key slides, discussion of alarms and accuracy, a direct shot at Dexcom on the hypoglycemia front, and Q&A.
-
The focus on optional high/low alarms was stressed (they are defaulted off), and FreeStyle Libre 2 still requires the user to do a manual scan to obtain the real-time glucose reading and trend arrow. “You have to scan – we believe in empowerment, and doing something proactively to get the value.” Abbott is using the slogan: alarm, scan, act. As we’ve reported previously, FreeStyle Libre 2 has three types of optional alarms: low glucose alarm, high glucose alarm, and signal loss.
-
The sensor reads glucose every minute and communicates at a one-minute frequency with the reader/app; a high, low, or lost signal alarm will be generated accordingly. (We’re not sure if the alarm will keep repeating until it is cleared by the user – and if so, at what frequency.) Dr. Seibold positioned the one-minute sampling frequency as an advantage over other CGMs, noting that someone dropping 2.5 mg/dl per minute would have a 10 mg/dl heads up with FreeStyle Libre vs. the longer five-minute sampling with other CGMs. That said, FreeStyle Libre does not have predictive alarms like other systems, so the net-effect may still be equal/less advance notice with FreeStyle Libre.
-
“We believe some patients, under some situations, might benefit from alarms. But I want to stress – up to date, there is no evidence that alarms are beneficial. The study often used, the iHart CGM study, had major methodological flaws, and I’m happy to discuss that afterwards. But we do have evidence that patients are discontinuing use of CGM because of alarm fatigue. That’s why we are stressing the optional alarms.” – Dr. Seibold
-
-
Abbott’s FreeStyle Libre 2 in Europe is reporting similar accuracy to the US 14-day version of FreeStyle Libre: MARD of 9.5% in adults (n=95) and 9.4% in pediatrics (n=74), with ~87-93% of values within 20%/20 mg/dl of YSI reference and ~90% of values in Zone A of the Consensus Error Grid. See the key slides below. Accuracy from these two studies was not broken down by hypoglycemia, which is the key area Abbott needs to improve for iCGM labeling. As we noted in August, based on the US label for FreeStyle Libre 14-day, it appeared to fall below the iCGM special controls hypoglycemia benchmark of >85% of points within ±15 mg/dl for values <70 mg/dl; FreeStyle Libre 14-day was at ~38%-53% of points within ±15 mg/dl for <80 mg/dl, and these were not the lower-bound of the 95% confidence interval. Of course, it’s possible Abbott has done more studies, made further improvements for the US submission, or discussed the special controls with the FDA.
-
A slide suggested G6 underreports the most dangerous hypoglycemia – 40-60 mg/dl – relative to FreeStyle Libre. Pulling from G6’s US label, Abbott compared the % of time the CGM read 61-80 mg/dl when blood glucose was actually in the lower 40-60 mg/dl range (i.e., false highs). This was 21% for the Dexcom G6 vs. 4.9% for FreeStyle Libre 14-Day. “Dexcom G6 system underreports significant hypoglycemia, which may potentially delay treatment.” (See the relevant table on page 300 – Table 3-A in Appendix F – of Dexcom’s G6 user guide.) Dexcom has taken many swipes at FreeStyle Libre – particularly in criticizing the hypoglycemia performance – so it wasn’t too surprising to see this mentioned today. Overall, we hope to see all CGM companies remember that this category is not a zero-sum game – for one sensor to succeed, another does not have to lose.
-
Abbott’s poster 910 shows the “future performance of FreeStyle Libre, which is not yet approved anywhere – just R&D data.” From the looks of the data, it does appear to meet all of the special controls for iCGM, assuming the lower bound of the 95% confidence intervals is close to the point estimates shown below.
-
Dr. Seibold showed an interesting case study of a 44-year-old type 1 (A1c: 7.2%) that started on FreeStyle Libre 2 and benefitted from the alarms. Taking it out of the box (alarms defaulted OFF), he immediately noticed lots of overnight hypoglycemia (first trace). He enabled low alarms at 70 mg/dl (second trace), which reduced the lows but traded them for highs. Adding a 180 mg/dl alarm (third trace) helped him stay in range – all in a 28-day period. We like this sequential approach to adding alarms on an as-needed basis, rather than having everything turned on at the start.
Questions and Answers
Q: Is FreeStyle Libre 2 approved for other wear locations besides the upper arm?
Dr. Seibold: No.
Q: Can you alternate using the reader and the phone?
Dr. Seibold: Yes, you can use both reading devices. The device to activate sensor is the device where you will receive the alarms. If you activate with the reader, you can read with your smartphone, but you won’t receive alarms on the smartphone.
Q: If you are using the FreeStyle LibreLink app, can you share data with anyone else?
Dr. Seibold: Yes, you can share through the LibreLinkUp app; it’s currently only sharing glucose values, not alarms.
Q: Is there data on use in hemodialysis or peritoneal dialysis?
Dr. Seibold: It is not approved for such patients.
Q: Outside of the EU and US, is there an intent to bring FreeStyle Libre 2 to other geographies?
Dr. Mahmood Kazemi (Abbott): Abbott is trying to make this accessible to all people worldwide. We have to work in the regulatory framework within each geography.
Q: Does FreeStyle Libre 2 integrate currently with an insulin pump or automated insulin delivery system?
Dr. Seibold: No, it is currently standalone. It is not cleared as an iCGM, and it is not coupled with a pump.
Q: If you enable alarms, can you program alarm schedules – e.g., night vs. day?
Dr. Seibold: No, you cannot set a schedule. We have made it very simple to activate/deactivate alarms. Our philosophy is to keep it as simple as possible. The product is not designed as a solution for sophisticated, high-end patients.
Dr. Kazemi: People can enable/disable the alarms. In future iterations of the software, there might be options for that [schedules].
Q: Was the accuracy shown relative to capillary blood glucose (fingersticks) or YSI?
Dr. Seibold: The data shown were relative to venous YSI.
Q: FreeStyle Libre vs. G6: where is the source?
Dr. Seibold: Have a look at the G6 instructions for use and the concurrence grids. (Editor’s Note: see the relevant table on page 300 (Table 3-A in Appendix F) of Dexcom’s G6 user guide.)
Q: Would FreeStyle Libre 2 meet iCGM criteria?
Dr. Kazemi: We leave the designation of iCGM to FDA to determine, the performance data are available, the special controls are publicly known as well. As far as they meet the criteria, that’s for FDA to decide.
Q: Does the alarm take into account rate of change?
Dr. Seibold: It’s not a predictive alarm; just the current reading based on one-minute sampling. It is not taking rate of change into consideration.
5. Medtronic Pipeline Overview in Investor Briefing: MiniMed 780G, Non-Adjunctive CGM Launch in Next 12 Months
Medtronic held a one-hour Diabetes investor briefing today, covering its AID and CGM pipeline for the next 24 months – download the slides here and listen to the webcast here. Following the Tidepool Loop news on Friday and the start of the MiniMed 780G advanced hybrid closed loop and next-gen Zeus iCGM pivotal trial yesterday (day 1 only calibration), there were no major surprises here. Still, the slides did bring more details on timing and product features on what seems to be a clear Medtronic pipeline. The MiniMed 780G and non-adjunctive labeling were positioned for launch in the “next 12 months,” roughly in line with the previous “FY20” window to launch in April 2019-April 2020. Both should hit that goal, as 780G has now started its three-month pivotal and non-adjunctive labeling will be filed this month. The comparator slide to Tandem’s Control-IQ was very interesting, showing that the 780G will be quite a competitive product head-to-head. The CGM discussion was not very convincing, despite management’s best efforts – Medtronic’s offering in 24 months still won’t match the current feature sets of Abbott and Dexcom in terms of factory calibration and 10-14 day wear. Launches in the next 24 months will include the Zeus seven-day iCGM sensor with day 1 calibration only (pivotal trial started yesterday; FDA submission in April-October 2020), the Synergy fully disposable seven-day iCGM sensor (FDA submission in November 2020-April 2021), Personalized Closed Loop (FDA submission in November 2020-April 2021). It was odd not to hear Tidepool Loop mentioned a single time in prepared remarks; it only came up in Q&A, though there was clear commitment to interoperability, including building an ACE pump and iCGM. Note that in the slides below, “FY21” runs from April 2020-April 2021. See product details and pictures below, and get the full slide deck here.
6. Freelife Kid AP Study of Tandem’s Control-IQ: Interim results show 24-hour wear beats overnight-only (+2 hours/day time-in-range) in pediatrics
Following the Control-IQ pivotal data (see #1 above), Dr. Eric Renard (Montpelier University Hospital) presented encouraging 12-week interim results (n=30) from France’s Freelife Kid AP Study. The multi-center, randomized study compares nocturnal-only (dinner and overnight) with 24-hour use of Tandem’s Control-IQ hybrid closed loop and Dexcom’s G6 for children in the real-world. Interim analysis was scheduled after 12 weeks from the first 30 patients included in the study. The interim study group had a mean age of 9 years old, a mean diabetes duration of 6 years, and an average A1c of 7.5%; all were previously on pumps. Impressively, the 24-hour group was in closed loop for ~98% of the 12 weeks, and the nocturnal group maintained time in closed loop of ~54% throughout the 12 weeks (i.e., they did not use during the day) – both confirming the high usability stats on this system. Time-in-range (70-180 mg/dl) for the 24-hour group improved from 60% at baseline to 72% during treatment (+2.9 hours/day) vs. a smaller improvement from 63% to 68% in the nocturnal-only group (+1 hour per day; p=0.009). Most of the increased time-in-range came from reduced time >180 mg/dl: an impressive drop from 36% to 25% in the 24-hour group (-2.6 hours per day) vs. 32% to 30% in the nocturnal-only group (-29 minutes per day; p=0.01). Overnight, the groups were (unsurprisingly) identical, with time-in-range of 82%-83%; during daytime, the open-loop group’s time-in-range fell to 60%, while the 24-hour group’s remained fairly high at 66% (p=0.023). Average daytime CGM reading was also 10 mg/dl lower in the 24-hour closed loop group (166 vs. 156 mg/dl). Interim analysis also demonstrated the system was safe –no DKA, severe hypoglycemic episodes, or hospital admissions occurred for either group. Based on the interim results, the full study (n=120 type 1 children) continued until week 36, with the nocturnal-only arm switching to 24-hour closed loop at week 18. These interim data show Control-IQ/G6 as configured should perform very well in the pediatric population, a plus as Tandem’s Control-IQ study in 6+ years is currently underway in the US. We look forward to seeing these FreeLifeKid AP results, perhaps at EASD 2019 or ATTD 2020.
|
Nocturnal-only |
24-hour |
p-value |
|||
|
Baseline |
Treatment |
Baseline |
Treatment |
To Baseline |
Between groups |
% time between 70-180 mg/dL |
63% |
68% |
60% |
72% |
<0.001 |
0.009 |
% time between 70-140 mg/dL |
42% |
46% |
38% |
49% |
<0.001 |
0.009 |
% time below 70 mg/dL |
4.1% |
2.8% |
3.9% |
2.8% |
0.006 |
>0.05 |
% time below 54 mg/dL |
0.7% |
0.6% |
0.9% |
0.5% |
0.081 |
>0.05 |
% time above 180 mg/dL |
32% |
30% |
36% |
25% |
<0.001 |
0.01 |
% time above 250 mg/dL |
9% |
9% |
12% |
6% |
0.001 |
0.01 |
% time above 300 mg/dL |
2.8% |
2.7% |
3.9% |
1.7% |
0.03 |
0.046 |
Average CGM [mg/dl] |
158 |
157 |
165 |
150 |
0.001 |
0.006 |
7. Late-Breaking Poster Details International Consensus on Time-in-Range, Recommendations for CGM-Based Clinical Targets; Updated AGP Included in Diabetes Care Paper
In a major win for the beyond A1c movement, the international consensus on time-in-range group presented its recommendations for CGM-based clinical targets in a late-breaking poster (2-LB) and paper published in Diabetes Care. Most current CGM users will fall in the first column, where the time-in-range goal is >70% for type 1 and type 2 diabetes, with <4% below 70 mg/dl. See below for a graphical representation of the committee’s recommendations. The initial meeting of CGM and time-in-range experts took place in February prior to ATTD, and the recommendations have since been further refined. Most notably, the consensus recommends <25% time spent above 180 mg/dl and <5% time above 250 mg/dl for those with type 1 and type 2 diabetes. While targets were estimated for type 1 and type 2 pregnancy at the consensus meeting, the official recommendation excludes targets for type 2 pregnancy due to “limited evidence in this area.” The authors conclude that the recommendations may serve to increase routine CGM use and improve diabetes management – we certainly hope so! It will also be interesting to track how the existence of these targets changes clinical adoption/implementation of CGM and whether it impacts payers or the research landscape.
-
The panel also underscored the need to present CGM data in a standardized report. An updated AGP (Ambulatory Glucose Profile) is provided in the poster and paper, clearly depicting the targets for each glucose range, defined as “Very High” (>250 mg/dl), “High” (181-250 mg/dl), “Target Range” (70-180 mg/dl), “Low” (54-69 mg/dl), and “Very Low” (<54 mg/dl). The poster emphasizes the panel’s belief that goal-setting should be collaborative and highly personalized, taking a step-wise approach. A footnote highlights the importance of being conservative with “a strong focus” on reducing time <70 mg/dl and preventing “excessive hyperglycemia.”
8. iDCL Protocol 4 to Test Dexcom G6 CGM/Tandem’s t:slim X2 Pump/Adaptive Zone MPC Algorithm in Six-Month RCT; Algorithm Parameters Adjusted Weekly
Harvard’s Dr. Frank Doyle provided details on Protocol 4 of the NIH-funded International Diabetes Closed Loop (iDCL) trial, including study design and the dual-layer control scheme for the Adaptive Zone Model Predictive Control (MPC) algorithm. In terms of design, after a two-week pilot study (n=10), subjects (n=40) will be randomized 1:1 to closed loop or PLGS in the outpatient setting. The closed loop will consist of a Dexcom G6 CGM, Tandem t:slim X2 pump, and Harvard’s Adaptive Zone MPC algorithm (described at ATTD). The six-month, multi-center, randomized crossover trial will be conducted at four to five US clinical sites. Primary efficacy outcomes include time-in-range (70-180 mg/dl) and time <70 mg/dl between baseline and three-months post-randomization. The algorithm will reside on the study app (iAPS) which runs on users’ smartphones. Dr. Doyle emphasized that these will be common phones, meaning they are not locked-down versions. iAPS has already been tested in 43 subjects across two separate trials in 68 children and adults. The studies are ongoing at Yale (n=48 pediatrics) and Sansum Diabetes Research Institute (n=20 adults) and primarily focus on understanding patient preferences for AID systems, including dietary management and on-body placement. It’s been over a year since we last received an update on Protocol 4 at ATTD 2018 – we’re not sure when the pilot will commence, but we’re eager to see what this system can do in vivo.
-
The Adaptive Zone MPC algorithm is comprised of a lower and upper layer. The lower-layer consists of real-time control algorithms, including: (i) a calculator for basal insulin delivery rate; (ii) a feedback controller dealing with real-time CGM measurements; and (iii) a feedforward controller calculating meal and correction boluses. The upper layer is a long-term parameter adaptation layer, which automatically identifies the correct parameters in the lower layer that need to be adjusted and optimizes them safely and efficiently. Long-term parameters that can be adjusted include the basal insulin rate, insulin:carb ratio, and correction factor. Essentially, it allows for adaptation of the lower layer algorithm at distinct intervals – in the case of Protocol 4, adaptation will occur weekly. The algorithm was tested in a recently published 24-week simulation study involving specific scenarios. For underestimated basal rates and overestimated insulin:carb ratio, the algorithm increased time-in-range (70-180 mg/dl) from 41% to 88% (+11.3 hours/day). For overestimated basal rates and overestimated insulin:carb ratios, the algorithm improved time-in-range (70-180 mg/dl) from 76% to 89% (+3.1 hours) and decreased time <70 mg/dl from 16% to 0% (-3.8 hours). While the algorithm has yet to be tested in humans, Dr. Doyle emphasized that these scenarios were used in previous simulation studies that have since been validated in clinical trials. The study tested the algorithm on the 111-adult cohort of the FDA-accepted UVA/Padova type 1 diabetes simulator.
-- by Adam Brown, Ann Carracher, Abigail Dove, Martin Kurian, Brian Levine, Peter Rentzepis, Maeve Serino, and Kelly Close